FACTORS AFFECTING THE NEVEMENT AND DISTRIBUTION Of FLUORIDE IN ACOIFERS by

FACTORS AFFECTING THE NEVEMENT AND DISTRIBUTION Of FLUORIDE IN ACOIFERS by

FACTORS AFFECTING THE NEVEMENT AND

DISTRIBUTION Of FLUORIDE IN ACOIFERS by

Eduardo Jorge Utunoff

A Dissertation Submitted to the Faculty of the

DEPARTMENT OF HYDROLOGY AND WATER RESOURCES

In Partial Fulfillment of the Requirements

For the Degree of

DOCTOR OF PHILOSOPHY

WITH A MAJOR IN HYDROLOGY

In the Graduate College

THE UNIVERSITY OF ARIZONA

1988

1

THE UNIVERSITY OF ARIZONA

GRADUATE COLLEGE

As members of the Final Examination Committee, we certify that we have read

the dissertation prepared by Eduardo Jorge Usunoff

entitled Factors affecting the movement and distribution of fluoride in aquifers

and recommend that it be accepted as fulfilling the dissertation requirement for

the Degree of Doctor of Philosophy

Da

Date

A

21:\\ v

.

2,2,

Date

Date

Date g„3

Final approval and acceptance of this dissertation is contingent upon the candidate's submission of the final copy of the dissertation to the Graduate

College.

I hereby certify that I have read this dissertation prepared under my direction and recommend that it be accepted as fulfilling the dissertation requirement.

Date

\

r

'l 1

"

Iteà>

2

STATEMENT BY AUTHOR

This dissertation has been submitted in partial fulfillment of requirements for an advanced degree at the Uhivrsity of Arizona and is deposited in the University Library to be made available to borrowers under rules of the Library.

Brief quotations from this dissertation are allowable without special permission, provided that accurate acknowledgment of source is made. Requests for permission for extended quotation from or reproduction of this manuscript in whole or in part may be granted by the head of the major department or the Dean of the Graduate College when in his or her judgment the proposed use of the material is in the interests of scholarship. In all other instances, however, permission must be obtained from the author.

SIGNED:

3

ACKNOWLEDGMENTS

The carrying out of the various experiments described and the preparation of this dissertation would not have been possible without the help and guidance of many individuals. I am especially grateful to my dissertation director and mentor, Dr. Stanley N. Davis, for his continual support throughout this project. Had not it been for Dr.

Davis' guidance and encouragement, I doubt I would have been able to overcome those many difficulties encountered during the course of this study. Dr. Eugene Simpson, Dr. Soroosh Sorooshian, Dr. Austin Long, and Dr. Joaquin Ruiz served on my ccalmittee, and I thank them for their assistance and understanding Dr. Daniel Evans let me use a storage room to set up my "lab", for which I will never be able to thank him enough. Sincere thanks go to Dr. Roger Bales, who gave me access to his laboratory and many of his instruments.

I would also like to thank all my fellow graduate students, especially Susan Maida for her never-ending patience and dedication in doing a good portion of the analytical work described here, Juan

Estalrich for his fruitless attempts to bring me back to reality, and

David Kebler and James Szecsody for their help and advice on the interpretation of some of my experimental results. Mts. Carla Thies kindly agreed on organizing and typing the manuscript, perhaps not knowing what she was getting into.

Last, but most certainly not least, I want to thank the

Universidad Nacional del Sur for having sponsored my graduate studies at the University of Arizona.

Dedication

To my tribe

(Liliana, GUillermina,

Pablo, and Diego). As Bobby Womack and Alltrinna Grayson would say, "No matter how high I get, I'll still be lookin' up to you all."

To my parents and my parents in-law (now, that's unusual!!)

4

TABLE OF CONTENTS

LIST OF ILLUSTRATIONS

LIST OF TABrPs

ABSTRACT

1.

INTRODUCTION

1.1 Introductory Remarks

1.2 Statement of the Problem

1.3 Dissertation Structure

2.

REVIEW OF THE LITERATURE

2.1 Sources of Fluorine

2.2 Fluorine Geochemistry: Chemical Equilibrium

COncepts

2.3 Fluorine Geochemistry: Some Aspects of

Chemical Kinetics

2.3.1 Generalities on Reaction Rates

2.3.2 Dissolution Rate of Minerals

2.3.3 Fluorine Minerals Dissolution

Kinetics

2.4 Fluorine Geochemistry. Adsorption and Other

Modifying Phenomena

2.4.1 COncepts and Terminology

2.4.1.1 Definitions

2.4.1.2 Adsorption Mbdels

2.4.1.3 Physical Adsorption

2.4.1.4 Chemisorption

2.4.1.5 Retardation

2.4.1.6 Ion Exchange

2.4.1.7 Precipitation

2.4.2 Results of Fluorine Adsorption-Related

Studies

2.5 Distribution of Fluorine Species in Aquifers and Soils

2.6 Significance of Fluorine in Human Health

2.7 Summary of Previous Research

54

54

54

54

57

57

57

59

60

62

Page

8

12

17

19

19

20

21

23

23

28

39

39

43

47

93

95

101

5

TABLE OF CONTENTS- -Continued

3. ADSORPTION OF FLUORIDE BY SOME CallONLY-FOUND

MINERALS

3.1 Introduction

3.2 Results and Discussion of Column Experiments

3.3

Results and Discussion of Batch Experiments

3.3.1 Adsorbate: Quartz Sand

3.3.2 Adsorbate: Vermiculite

3.3.3 Adsorbate: Kaolinite

4. KINETICS OF FLUORITE DISSOLUTION

4.1 Introduction

4.2 Formulation and General Concepts

4.3

Results and Discussion of Column Experiments

4.4 Results and Discussion of Batch Experiments

5. TRANSPORT OF FLUORIDE: AN EXPERIMENTAL APPROACH

5.1 Introduction

5.2 Handling of Chemical Data

5.3 Results and Interpretation of Column Tests

1, 2, and 3

5.4 Results and Interpretation of Column Test 4

6. STATISTICAL STUDY

6.1 Introduction

6.2 Factor Analysis: Definitions, Concepts, and

Terminology

6.3 Correspondence Analysis: Definitions,

Concepts, and Terminology

6.4 Example 1: The Upper San Pedro River Basin

(Arizona)

6.4.1

Factor Analysis: R-Made

6.4.2

Factor Analysis: Q-Mbde

6.4.3 Correspondence

Analysis

6.5 Example 2: The San Bernardino Valley

(California)

6.5.1 Chemical Data Base

6.5.2 Factor Analysis: R-Mbde

6.5.3 Factor Analysis: Q-Mbde

Page

196

196

198

204

206

209

219

222

234

238

241

253

103

103

105

120

120

131

135

140

140

142

145

153

167

167

168

170

192

6

TABLE OF CONTENTS--Continued

7. CONCLUSIONS AND RECOMMENDATIONS

APPENDIX A: PREPARATION OF THE VARIOUS MINERALS

USED IN THE EXPERIMENTS

APPENDIX B: ANALYTICAL DETERMINATION OF F

-

CONCENTRATIONS AND pH

APPENDIX C: CONSTRUCTION OF COLUMNS AND

CALIBRATION TESTS

APPENDIX D: RESULTS OF COLUMN EXPERIMENTS ON F

-

ADSORPTION BY SAND

APPENDIX E: RESULTS OF COLUMN AND BATCH TESTS ON

THE DISSOLUTION RATE OF FLUORITE

APPENDIX F: RESULTS OF OOLUMN COMPOSITE EXPERIMENTS

APPENDIX G: CHEMICAL DATA FOR THE STATISTICAL STUDY

LIST OF REFERENCES

7

Page

255

263

270

274

293

308

335

342

356

LIST OF ILLUSTRATIONS

Figure

1 Concentration C versus distance from a crystal surface r for three rate-controlling processes

2

Effect of pH on the amount of 0H released from a soil clay

3

Diagrammatic representation of possible reactions of fluoride with incompletely coordinated metal ions near the surface of an oxide

4 Isotherms for adsorption of F

-

on goethite

5

Isotherms for adsorption of F

-

on gibbsite

6

Relation between dental caries experience and [F

-

] in drinking water

7 Relation between index of fluorosis and

[F

-

] in drinking water

8 Relation between [F

-

] in municipal waters and fluorosis index for communities with mean annual temperatures of approximately 50 F (Midwest) and 70PF

(Arizona)

9 Breakthrough curve for column experiment 1

10 Breakthrough curve for column experiment 2

11 Breakthrough curve for column experiment 3

12

Breakthrough curve for column experiment 4

13 Breakthrough curve for column experiment 5

14 Breakthrough curve for column experiment 6

15

Breakthrough curve for column experiment 7

16 F

-

adsorption isotherm from batch experiment 3

17 F

-

adsorption isotherm from batch experiment 17

18 F

-

adsorption isotherm from batch experiment 18

19 F

-

adsorption isotherm from batch experiment 19

8

97

98

80

86

86

Page

45

75

110

111

112

113

122

100

107

108

109

123

124

125

9

LIST OF ILLUSTRATIONS--COntinued

Figure

20 Relationship between F

-

adsorbed from and released to solution

21 F

-

adsorption from batch experiments 20 and 21

22 F

-

adsorption fram batch experiments 22 and 23

23 Extent of fluorite dissolution as a function of distance for various mean pore velocities:

Column runs 1, 2, and 3

24 Relationship between the dissolution rate of fluorite and the mean pore velocity

25

Functional relationship between the dissolution rate of fluorite and the mean pore velocity:

COmparison with published data

26

Extent of fluorite dissolution as a function of distance for various pHs: Column runs 1, 4, and 5

27 Relationship between the dissolution rate of fluorite and pH

28 Extent of fluorite dissolution as a function of time for various stirring rates

29 Extent of fluorite dissolution as a function of time for various [NaHCO

3

] in solution

30 Extent of fluorite dissolution as a function of time for various [CaSCO in solution

31 Extent of fluorite dissolution as a function of time for various ion strength of solutions

32 Extent of fluorite dissolution as a function of time for various temperature of solutions

33 Arrhenius plot

34 Breakthrough of non-reactive species: Run 1

35

Breakthrough of non-reactive species: Run 2

36 Breakthrough of non-reactive species: Run 3

Page

130

134

137

146

148

150

152

154

156

157

159

160

163

164

173

174

175

LIST OF ILLUSTRATIONS—Continued

Figure

37

Breakthrough of reactive species: Run 1

38 Breakthrough of reactive species: Run 2

39

Breakthrough of reactive species: Run 3

40 F

-

mass balance: Run 1

41 F

-

mass balance:

Run 2

42 F

-

mass balance:

Run 3

43 Ca" mass balance: Run 1

44

Ca" mass balance: Run 2

45

Ca" mass balance: Run 3

46 Na+

mass balance: Run 1

47

Na mass balance: Run 2

48 Na mass balance: Run 3

49 Breakthrough of F

-

and Na and equilibrium with respect to fluorite

50

Location of the Upper San Pedro River Basin

51

Triangular diagram of factors, confined aquifer samples

52 Triangular diagram of factors, unconfined aquifer samples

53

Sampling sites distribution on a factor I vs.

factor II space

54 Triangular diagram showing the distribution of sample sites through 10-mcde factor analysis

55 Grouping of chemical variables on a factor I vs.

factor II space, confined aquifer samples

56

Grouping of chemical variables on a factor I vs.

factor II space, unconfined aquifer samples

10

223

230

235

Page

177

182

183

184

185

178

179

181

186

187

188

189

194

207

214

217

221

LIST OF ILLUSTRATIONS-- continued

Figure

57 Location of Highland-East Highland area

58 Scree plot

59 Variables on a factor I vs. factor II space

60 Equilibrium with respect to calcite

61

Variables on a factor I vs. factor III space

62 Variables on a factor II vs. factor III space

63

Equilibrium with respect to fluorite

A-1 Ottawa Sand: Granulometric distribution

C-1 Typical elements of a column used in laboratory experiments

C-2 Elements of the multiple input device

C-3

Typical breakthrough curve from continuoussource dispersion tests

C-4 Breakthrough curve from dispersion test:

Quartz-sand column

C-5

Breakthrough curve from dispersion test:

Mixed-materials column 1

C-6

Breakthrough curve from dispersion test:

Mixed-materials column 2

G-1 Trilinear diagram of samples from the confined aquifer (Upper San Pedro River Basin)

G-2 Trilinear diagram of samples from the unconfined aquifer, the San Pedro River, and the McGrew

Spring (Upper San Pedro River Basin)

1

1

Page

237

242

247

248

250

251

252

265

276

277

282

285

289

292

345

346

12

LIST OF TABLES

Table

1

Commonly-found F-bearing minerals

2 Fluoride released from rocks as a function cl pH

3 Solubility of fluorite (89% CaF

2

) and fluorapatite

(containing 1% F) in distilled water, 0.01 M calcium, magnesium, and sodium bicarbonates at 28°C

4 Distribution of F species in % of total F for two natural acidic water samples at 25°C

5 Dissolution rate-controlling mechanisms for various substances arranged in order of solubilities in pure water

6

Correlation coefficients between

ofir

release and amounts of silica, alumina, and ferric oxide obtained by selective dissolution

7

Column experiments: Measured and calculated data

8

Batch experiments: Summary of measured data for quartz-sand

9

Batch experiments: Summary of measured data for vermiculite

10 Batch experiments: Summary of measured data for kaolinite

11 Percentage of variation explained by factors, unconfined aquifer samples

12

Percentage of variation explained by factors, confined aquifer samples

13 Factor loadings on three factors, unconfined aquifer samples

14 Factor loadings on three factors, confined aquifer samples

15

Percentage variation of chemical variables explained by three factors, confined aquifer samples

Page

24

27

33

40

48

77

114

127

133

136

210

210

211

211

213

LIST OF TABLES--Continued

Table

16 Percentage variation of chemical variables explained by three factors, unconfined aquifer samples

17 Factor loadings for three factors, Q-mode factor analysis

18

Coordinates of chemical variables on three factors, confined aquifer samples

19 Per cent absolute (AC) and relative (RC) contributions for three factors, variables: chemical parameters, confined aquifer samples

20

Coordinates of sampling sites on three factors, confined aquifer samples

21 Per cent absolute (AC) and relative (RC) contributions for three factors, variables: sample sites, confined aquifer samples

22 Coordinates of chemical variables an three factors, unconfined aquifer samples

23

Per cent absolute (AC) and relative (RC) contributions for three factors, variables: chemical parameters, unconfined aquifer samples

24 Coordinates of sampling sites on three factors, unconfined aquifer samples

25

Per cent absolute (AC) and relative (RC) contributions for three factors, variables: sample sites, unconfined aquifer samples

26 Percentage of total variation explained by three factors

27 Communalities of variables

28 Variable loadings on three factors after rotation

C-1 Dispersion experiment data. Quartz-sand column

13

Page

232

232

233

244

245

245

284

216

220

224

224

225

226

231

LIST OF TABLES--Continued

Table

C-2 Dispersion experiment data. Mixed-materials column 1

C-3 Dispersion experiment data. Mixed-materials column 2

D-1 Data from column experiment 1

D-2 Data from column experiment 2

D-3 Data from column experiment 3

D-4 Data from column experiment 4

D-5 Data from column experiment 5

D-6 Data from column experiment 6

D-7 Data from column experiment 7

D-8 CFITIM output based on data from column experiment 1

E-1 Data from column experiment 1

E-2 Data from column experiment 2

E-3

Data from column experiment 3

E-4

Data from column experiment 4

E-5

Data from column experiment 5

E-6

Batch tests: Summary of chemical and physical characteristics

E-7

Data from batch test 1

E-8 Data from batch test 2

E-9 Data from batch test 3

E-10

Data from batch test 4

E-11 Data from batch test 5

14

Page

288

305

312

313

314

315

316

291

295

296

297

298

299

300

301

318

319

320

321

322

323

LIST OF TABLES--Continued

Table

E-12

Data from batch test 6

E-13

Data from batch test 7

E-14

Data from batch test 8

E-15

Data from batch test 9

E-16 Data from batch test 10

E-17 Data from batch test 11

E-18 Data from batch test 12

E-19 Data from batch test 13

E-20 Data from batch test 14

E-21

Data from batch test 15

E-22 Data from batch test 16

F-1 Chemical analyses of samples. Column runs

1, 2, and 3

F-2 Chemical analyses of samples. Column run 4

G-1 Upper San Pedro River Basin: Chemical analyses of samples

G-2 Upper San Pedro River Basin: Statistics of chemical variables

G-3 Upper San Pedro River Basin: Correlation coefficients matrix, unconfined aquifer samples

G-4 Upper San Pedro River Basin: Correlation coefficients matrix, confined aquifer samples

G-5 San Bernardino Valley: Chemical analyses of samples

G-6 San Bernardino Valley: Statistics of chemical variables

15

343

347

348

349

350

Page

324

325

326

327

328

329

330

331

332

333

334

339

341

354

LIST OF TABLES -

Table

G

,

7 San BernarclinoWaley: Correlation coefficients matrix

16

Page

355

17

ABSTRACT

This dissertation presents the results of laboratory experiments in which essential aspects of the movement of F

-

in saturated media have been addressed.

The interactions between F-solutions and quartz, vermiculite, and kaolinite were studied through batch and column experiments.

Quartz was found to react slightly with F

-

, giving data described by a quasi-linear isotherm. Vermiculite adsorbed only minute amounts of F.

A large uptake of F

-

by kaolinite was measured. Ion exchange F

-

by 0H

-may not have been the exclusive mechanism operating under the experimental conditions.

The kinetics of the dissolution of fluorite (CaF

2

) were investigated by means of batch and column tests. There appears to be a relationship between the dissolution rate and the mean flow velocity.

Solution pHs greater than 5-6 may accelerate the dissolution process.

For temperatures between 15 and 30°C, the dissolution is characterized by an activation energy of about 7 Kcal/mole, which would indicate that both surface reaction and transport are the rate-limiting step.

The percolation of columns containing quartz, vermiculite, and kaolinite with multi-component solutions (including F

-

) resulted in a late breakthrough of F

-

when compared with that of the other species.

Circulation of distilled water led to an almost complete recovery of the F

-

injected, which exited the column with relative concentrations greater than 1. When the packing included fluorite and distilled water was flushed through the column for 2 days, a concentration of F

-

of

18 about 1.3 pprn was rapidly reached and remained constant throughout the run.

Multivariate analysis techniques (factor and correspondence analyses) were applied to data from two aquifers known to carry high Fwaters. Although helpful in discriminating major and minor associations of species, none of those techniques could help unravel the behavior of F

-

in the study aquifers.

It is suggested that laboratory and field studies be continued and that, for the successful modeling of the movement and distribution of F

-

in aquifers, non-linear source/sink terms should be included in the pertinent differential equation governing the transport of solutes.

19

CHAPTER 1

INTRODUCTION

Introductory Remarks

The study of the behavior of aqueous fluorine (F) species has long concerned hydrologists due to the potential health hazards associated with concentrations in the range of a few milligrams per liter of those species. Even though fluorine is only a minor constituent in most natural waters, concentrations of its natural species commonly impose severe constraints in the development and availability of safe water supplies. Mast of the investigations of areas affected by a high [F

-

] in their ground-water resources have revealed two puzzling characteristics, namely, the sometimes unexplainable degree of association between the variations in [F

-

] and those of the major ions and the lack of a clear distributional pattern of the F species.

The two characteristics mentioned above should be understood in order to determine their relation to the mobility of F species in aquifers, at the regional or even the local scale. Computer models that can couple flow and transport phenomena will not have a direct application to the assessment of the geochemistry of F in a given area without a previous knowledge of the manner in which the corresponding codes are to be modified to account for the particular F-affecting processes. In short, general-purpose computer programs have to be customized if the digital modeling of aqueous F is to have a place among the effective techniques available to hydrologists.

20

The author had a first glimpse of the problems associated with the behavior of F in aquifers in the course of working on his M.S.

thesis (Usunoff, 1984), which consisted of the hydrochemical characterization of an area with a high F

-

content. The mostly tentative conclusions which were reached in that study provided the initial driving force behind the research which is described in this dissertation.

Statement of the Problem

The author's limited experience on the geochemistry of F and the erratic and sometimes contradictory information found in papers dealing with the subject posed several interesting questions:

(1) Can a ccmprehensive review of the literature turn up any underlying facts and subsidiary points worth exploring?

(2) Is it possible to model the movement of F species in aquifers based on thermodynamic chemical equilibrium concepts?

(3) What is the role played by the kinetics of dissolution of F-carrying minerals, and what are the governing parameters?

(4) What is the nature and extent of processes acting when:

(a) waters devoid of F flow through sediments which contain F minerals?

21

(b) F-bearing waters interact with minerals of common occurrence in aquifers?

(5) Given some typical scenarios, can the variations in

[F

-

] be conceivably linked to the variations in the concentrations of selected major ions? Also, can techniques based on multivariate analysis offer any help in the interpretation of the hydrochemical data?

These questions, which constitute the motivation for my research, have been addressed in the various sections of this dissertation.

Dissertation Structure

Each chapter is a separate entity or a module, and the different goals pursued in each of these modules do not always give rise to a smooth integration of the various chapters. Consistency has been achieved, I hope, within each chapter, even though the work as a whole may lack extended connecting paragraphs among certain parts.

Chap er 2 presents the results of an exhaustive literary review of those topics that might contain information of relevance to the interpretation of the results of the experiments which I have completed.

Chapter 3 includes a presentation and discussion of the data from adsoLpLion experiments in which the F-related adsorptive properties of minerals such as quartz, vermiculite, and kaolinite were

22 assessed through batch and column tests. Parameters such as pH, [F] in solution, and mean flow velocity were varied in order to measure the effect on the adsorption of F.

The kinetics of the dissolution of fluorite, perhaps the most common fluorine-bearing mineral, are addressed in Chapter 4. Results from column and batch experiments in which the pH, the mean flow velocity, the chemical composition and the ion strength of solution, the stirring rate, and the temperature were varied are described and interpreted.

Chapter 5 contains a discussion of the data provided by a set of column experiments, which were designed on the basis of the preliminary conclusions coming from the adsorption and the kinetic studies.

Chapter 6 presents two examples of the application of two multivariate analysis techniques, factor analysis and correspondence analysis, to the study of the hydrochemistry of natural systems. The first part of Chapter 6 includes an overview of the main principles behind the techniques employed.

Chapter 7 sunuarizes the relevant conclusions of my research as well as some recommendations and suggestions for future efforts on the subject.

Supporting data and details of experiments are confined to appendices which are placed at the end of the dissertation. Special care was taken to describe those aspects of the experiments that, in my opinion, deserve particular attention because of their involved nature.

23

CHAPTER 2

REVIEW OF THE LITERATURE

Sources of Fluorine

Fluorine, the lightest element of Group VII in the periodic table of the elements, has an atomic number of 9 and an atomic weight of 19. A member of the halogen group, it is found in all naturallyoccurring compounds with a valence of minus one.

The fluorine ion, or fluoride, has an ionic radius of 1.36 c iL, which allows its isomorphous replacement with other species of similar ionic size, in particular the hydroxyl group, (OH

-

) with a radius of

1.40 A (Rankama and Sahama, 1950; Goldschmidt, 1954; Ziauddin, 1977).

Fluoride may also replace iscmorphously chloride, Cl

-

(radius = 1.81

(Fleischer and Robinson, and the oxygen ion, 0 ' (radius = 1.40

1963).

Fluorine does not occur in its free state, although about a hundred minerals are known to contain the element. Some of the most common fluorine-bearing minerals are listed in Table 1.

In igneous rocks, fluorine is the twelfth most abundant element and, even more significant, ranks third among the most abundant anions, after cr and OH

-

(Romo and Roy, 1957). Fleischer and Robinson (1963), after carrying out an extensive literary review, reached the following conclusions:

(1) For intrusive and extrusive rocks, with the exception of phonolites and alkalic intrusives, there is a definite tendency toward higher

Table 1. Commonly found F-bearing minerals (after Kanmakaran,

1977, p. 7).

Narre Formula

%

Fluorine Content

Theoretical Range in Analytical Values

A. FZUCRIDES•

1. Fluorite

2. Sellaite

3. Fluocerite

(Tysonite)

4. Cryolite

B. PHOSPHATES:

CaF

2

MgF

2

(Ce La py)F

Na

3

A1F

6

3

48.67 48.18-48.61

60.98

29.00 19.49-29.44

54.29 53.55-54.88

5. Fluorapatite

(Carbonate-apatite)

6. Wagnerite

Ca5 (PO4)

Mg2PO4F

3

F

7. Triplite

8. Amblygcnite

(Mil,Fe,Mq,Ca)2FPO4

(Li,Na)A1PO4(F,OH)

C. SILICATES:

9. Topaz

10. Humite group

11. Sphene

12. Vesuvianite

13. Tourmaline

MICA GROUP

14. Muscovite

15. Phlogopite

16. Biotite

17. Lepidolite

18. Amphibole group

3.80

11.68

12.85

2.57- 5.60

5.06-11.48

6.02- 9.09

0.57-11.26

A1(F.OH)

2

SiO

4

Mg(OH,F)

2

.nMg

2

SiO

4

CaTi(Si0)

4

(0,0H,F)

Calcium-Magnesium silicate with (OH,F)

4

Complex boro-silicate with (OH,F)

4

Sheet silicates with

(OH,F)

4

Potash mica

Magnesian mica

Iron-Magnesian mica

Lithium mica

Chain silicates with

(OH,F)

2

20.70 13.23-20.37

Tr. -13.55

0.61- 1.40

Tr. - 3.22

0.07- 1.16

0.00- 2.06

0.56- 9.20

0.32- 5.02

4.93- 8.08

Tr. - 2.95

24

25 fluorine contents with an increasing content of

Si01

2

.

(2) For the most common extrusive rocks, the arithmetic average of the fluorine content reported in a large number of analyses is (in ppm of F): basalts 360, andesites 210, rhyolites 480, phonolites 930. The average content for the corresponding intrusive rocks is: gabbros and diabases 420, granites and granodi-

°rites 810, alkalic rocks 1000.

(3) Similar averages for sedimentary rocks are: limestones 220, dolomites 260, sandstones and graywackes 180, shales 800, oceanic sediments

730, soils 285.

Inasmuch as fluorapatite was thought to represent the chief carrier of fluorine in igneous rocks, many investigators previously estimated the amount of fluorine based on the phosphorus content

(Barth, 1947). Consequently, the average fluorine was reported to be about one-tenth that of the phosphoric acid. As more reliable analytical methods to determine fluorine were found, evidence started accumulating to the effect that fluorine does not increase in proportion to phosphorus. Commercial types of domestic phosphate rocks, for example, were found to contain from 12 to 52% more fluorine than theoretically required to form fluorapatite. Mansfield (1940) interpreted those

26 findings as the result of the formation of a solid solution between fluorapatite and fluorite.

Given the presence of ample fluorine in minerals of common occurrence, the next question to pose is to what extent fluorine is released from rocks. Natarajan and Mbhan Rao (1977), in reporting the results of laboratory studies, summarized their information in the way presented in Table 2.

In shifting the attention to hydrology, it is evident that the fluorine content in natural waters should be studied in relation to the geological environment. The bedrock supplying sediments to many basins may contain large amounts of fluorine (Corbett and Manner, 1984;

Christiansen, 1986). Consequently, sediments derived from the erosion of those rocks are bound to contain fluorine-bearing minerals, a fact certainly true for alluvial sediments of southern Arizona's basins

(Kister and Hardt, 1966).

Although the dissolution of many different minerals (Table 1) would constitute a potential source of fluoride in ground waters, it is in particular the dissolution of fluorite, F-bearing micas, phosphates, and certain clay minerals which probably represents the main source of fluorine in virtually all types of ground waters (Dobrovol'skiy and

Lyal'ko, 1983). Although there seems to be a general consensus about the above statement, generalized gecchemical predictions of expected fluoride concentrations cannot be made because of the very localized features that sometimes control the concentration of fluorine in a given geological setting. As an example, Travi (1984) postulated a

Table 2.

Fluoride released from rocks as a function of pH (after

Natarajan and Mbhan Rao,

1977, p.

39).

Mineral/Rock

F

-

released (mg/1) in 24 hrs. pH values (between parentheses)

1. Fluorite

2. Fluorapatite

8.30

7.00

8.00

8.00

9.50

9.00

(3.1) (5.6) (7.2) (8.1) (9.0)

(11.0)

2.20

0.40

0.45

0.50

0.65

1.50

(3.5) (6.2) (7.6) (8.8) (9.5) (11.0)

0.15

0.15

0.15

0.15

0.15

0.15

(3.2) (9.5) (9.5) (9.5) (10.2) (11.0)

3. Biotite (1‹,mig,

Fe, Al silicate with OH and F)

4. Phlogopite mmkg,

Fe, Al silicate

0.15

nil nil nil nil nil

(4.7) (6.4) (8.3)

(8.3) (10.2) (11.1) with OH and F)

5. Tourmaline (Na,Mg,

Fe,Mn,Li,A1 silinil nil nil nil nil nil

(3.0) (5.0) (6.5)

(7.3) (8.8)

(11.0) cate with B, OH and F)

6. Actinolite Schist 0.15

0.15

0.15

nil nil nil

(3.5) (4.5)

(6.0) (8.5) (10.6) (11.0)

7. Anthophyllite

Schist nil nil nil nil nil nil

(3.8) (4.4) (7.1)

(8.0) (10.1) (11.0)

8. Hornblende Schist 0.15

nil nil nil nil nil

(3.0) (4.3) (7.0) (8.6)

(9.9) (11.0)

27

28 common origin of fluorine and magnesium in ground waters from a local area of calcium phosphate associated with attapulgite.

With respect to the fluorine content of soils, Robinson and

Edgington (1946) have indicated that the alteration of muscovite, biotite, and hornblende, all exceedingly common soil minerals, would represent the main source of the element in soils. Of equal importance, if not more important, in determining the fluorine content of soils are the sources of artificial fluorine enrichment. As a result of the effort to achieve a world-wide green revolution, the application of superphosphate and rock phosphate as fertilizers has been a common practice since the early 1930's (Hart et al., 1934) until a few years ago. Some local enrichment of fluoride in soils has been related to the widespread use of fluorine in various manufacturing processes

(aluminum, phosphatic fertilizers, bricks, enameled ware, high octane gasoline, etc.) as well as to the use of fluorides as insecticides

(Robinson and Edgington, 1946).

Fluorine Geochemistry: Chemical Equilibrium Concepts

For the sake of simplicity, the following discussion assumes that the fluoride ion (F

-

) is the only F species in solution. Specietion and ion pairing is discussed later in this section. Viewed from purely theoretical grounds, the amount of F

-

to be found in aqueous environments should be controlled by that F-bearing mineral phase having the lowest solubility. For such a rule to hold, even approximately, the following conditions must be met:

29

(1) Fluoride (F

-

) is the only fluorine species in solution, that is to say, no complex ions or ion pairs are formed at the expense of the F

-

in solution.

(2) Interactions among the F

-

ions, the bulk solution, and other mineral phases present are neglected.

(3) The low-solubility, F-carrying mineral is present, and its supply is unlimited. If such a mineral is not originally present, its solubility product is soon reached, and precipitation occurs at once.

(4) Regardless of the processes taking place in the aqueous phase, the pH is approximately constant and corresponds to the equilibrium pH for the low-solubility fluorine mineral.

(5) Effects of chemical kinetics are not considered (i.e., precipitation an/or dissolution occur without delay as soon as the solubility product is reached or concentrations fall below it, the rate cf mass transfer is disregarded, etc.).

It can be concluded right away that, although attractive for its simplicity, the model above cannot be realistically assumed to hold in natural ground-water systems. Indeed, the degree of complexity of natural aqueous environments is such that serious objections can be made to each of the constraints listed above. I will now present a summary of the conclusions and facts observed by numerous investigators, which illustrate errors resulting from the possible use of

30 simple models to describe the behavior of fluorine species in saturated or partially-saturated porous media. Unless otherwise defined, [F

-

] stands for the concentration of fluoride ions in solution.

One common characteristic of many studies related to fluoride is the finding of a direct correlation between [F

-

] and alkalinity, coupled with a negative correlation between [F

-

] and hardness

(Carlston, 1942; Cederstrcm, 1945; Bhakuni and Sastry, 1977; Handa,

1975 and 1977; Rao et al., 1977; Rao and Rajyalakshmi, 1977;

Viswanadham and MUrty, 1977; Zack, 1980; Sarma and Swamy, 1982; Bahr and Corbett, 1985). As an example, the experimental data reported in

Viswanadham and MUrty (1977) indicate that distilled water In contact with fluorite (CaF

2

) can dissolve as much as 8 ppm of F

-

. If the water contains H(0

3

, the amount of dissolved F

-

is increased to 24 ppm. If the water contains 1./4" and/or Ca" (but not HCO

3 -

), [F

-

] comes down to about 3 ppm. Finally, if the water contains Ca++, v4", and HCO

3 -

, about 15 to 18 ppm of F

-

can be found in solution.

Handa (1975) has given his interpretation of such characteristics by assuming that the ground water is in contact with fluorite and calcite solid phases and that an overall chemical equilibrium is established. If so, the equilibrium relationships can be represented as follows (ion-strength disregarded, so that concentrations equal activities): cacop

3

(s) + Br <====> calcite

+ HCO

3 -

; K c

= [Ca++][HCO3

-

]/[H+]

(2.1)

CaF

2

(s) <====> Ca++ + 2F

-

; K t

= [CS++][F

-

]

2

(2.2) fluorite

31

Dividing (2.1) by (2.2):

K c

/K

I

= [HCO

3

-

]/[11

4-

][F

-

]

2

(2.3)

Handa (1975) pointed out that if the pH does not vary much, as is found in natural environments, then any increase in the HCO

3

concentration should be accompanied by an increase in the F

-

concentration in order to keep the ratio [HCO

3

-

]/[F

-

]

2

in (2.3) unchanged.

Incidentally, if dolomite (CaMg(CO

3

)

2

) or magnesite

(migcoo

are considered instead of calcite, the conclusions are just about the same, provided that one recognizes that, in those cases, any increase in

[H03

3

-

]should lead to a decrease in

[

tr

.

.]

or an increase in [F], or both, with the ultimate result that, whatever the case, the ratio

[N4r+)/[F

-

] always decreases.

On the other hand, Equation (2.2) indicates that [Ca++] and

[F

-

] will increase until the solubility product is reached. Beyond that point, any addition of Ca" or F

-

ions from other sources will cause the removal of, respectively, F

-

or Ca"

-

ions from solution so that the solubility product is not exceeded. It is on the basis of those rather simple considerations that, according to Handa (1975 and

1977), the negative correlation between [F

-

] and hardness and the positive correlation between [F

-

] and alkalinity can find a suitable explanation. Travi (1984) argued that, in same cases, it is only the calcium hardness that is opposed to [F

-

] and presented evidence of a positive correlation between [F

-

] and the ratio [Mkg"]/[Ca++] for waters from a Paleocene aquifer in Senegal.

32

Many of the papers reviewed support the idea that fluorite

(CaP

2

), in spite of not being the least soluble among the F-containing minerals, controls the F

-

content in ground waters. Pathak (1977) suggested that NaF (villiaumite) should control the [fil in ground waters with high [Na'], but such an assertion does not appear very realistic. If one considers the extremely high solubility of NaF,

42.2 x 10

3

pin (Weast, 1984), and assumes a rather high F

-

content of

10 ppm, the corresponding amount of Na needed for NaF to control the

F

-

solubility is more than 44 x

106

wit which is, of course, physically impossible.

The actual amount of F

-

ions released by fluorite (up to 15 ppm according to Graham et al., 1975) is highly dependent on a whole variety of factors such as the presence and accessibility of the mineral, the chemical composition of the solution, and the time of contact between the source mineral and the water, among the most important variables (Sharma, 1977; Kraynov et al., 1980). Comparative studies on F

-

release by fluorite and fluorapatite under various conditions have concluded that fluorite released 2 to 50 times more

F

that fluorapatite (Natarajan and Mbhan Rao, 1977; Sharma, 1977), and that phosphate nodules do not release F

-

easily (Zack, 1980). Sharma's

(1977) findings are summarized in Table 3.

The actual attainment of chemical equilibrium between ground waters and F minerals is still under discussion. Bassett and Wood

(1978) found that artificially-recharged ground water in

the

Ogallala aquifer (Texas), where the source of F might be in the numerous volcanic ash beds, approached equilibrium with respect to fluorite

Table 3. Solubility of fluorite (89% CeF

2

) and fluorapatite

(containing 1% F) in distilled water, 0.01 M calcium, magnesium, and sodium bicarbonates at 28 °C (after

Sharma, 1977, p. 305).

Water or 0.01

M. Solutions

Distilled water

Ca(HC00

2

!(Hœ3 )2

NaHCO

3

Distilled water

Ca(HCO

3

)

2

Mg(HCO

3

)

2

NaHCO

3

Mineral

Fluorapatite

Fluorapatite

Fluorapatite

Fluorapatite

Fluorite

Fluorite

Fluorite

Fluorite

Time of

Contact

Overnight

Do.

Do.

Do.

Overnight

Do.

Do.

Do.

FReleased

(PrIn)

10

10

25

20

4

Nil

1

Nil

33

34 after about 1 to 2.5 months, which encouraged the authors to use chemical equilibrium computer mcdels for describing the movement of F

an a regional scale. The opposite view has been given by other authors who concluded that, inasmuch as fluorite is such a low-solubility mineral (Kt = 10-l o

.

4o

, Elrashidi and Lindsay, 1986), ground water with

[F

-

] close to saturation levels with respect to fluorite is not common.

Dobrovol'skiy and Lyalyko (1983) suggested that high [F

-

] can be measured in waters coming from deep horizons in artesian basins (where rocks are rich in fluorite and water exchange is difficult), in waters from scattered areas of fluorite deposits, or even in waters from hot springs (usually high in [F

-

]). Those same authors, although convinced that the maximum [F

-

] in ground waters is determined by the solubility product of fluorite, have pointed out that a set of very special conditions must be met to have precipitate fluorite. Indeed, the available evidence indicates that fluorite is hardly deposited in natural aqueous environments, and that supersaturation factors of 6-7 or even larger can be reached without any noticeable CeLF

2

precipitation

(Dobrovol'skiy and Lyal'ko, 1983).

In light of the material reviewed up to this point, it may be concluded that the assumption of chemical equilibrium can be questioned when applied to the geochemistry of fluorine in aquifers. The problem is that many investigations fail to take into account the fact that natural heterogeneous systems are usually in a non-equilibrium state.

Hence, the application of the methods based on equilibrium therncdynamics to the geochemistry of ground waters, unless done for exploratory purposes, is bound to produce some unexpected results. As

35

Kraynov et al. (1980, p. 17) have stated it, "while there is a relatively good agreement between the calculations and the observed element distributions in ground water, there is abundant evidence for considerable discrepancies between the concentrations, which means that forecasting is completely unreliable." In those cases, an alternative method of analysis, and certainly a most useful one, is that of local or partial equilibria. As defined in Kraynov et al. (1980, p. 18) "the principle of partial equilibrium implies that a system that is not in equilibrium as a whole may be split up into a set of spatially distinct parts for which the conditions of chemical equilibrium are met at a particular instant. These individual parts with their local equilibria are not in equilibrium with one another, and this reflects the general

=equilibrium in any major hydrochemical system." Although references to the application of the aforementioned principle as applied to the study of fluorine in aqueous systems have only been found in the scarce

Russian literature on the subject, it is hoped that this principle will soon be used by many gecchemists for the elucidation and forecasting of hydrochemical phenomena in natural systems. It is my sincere opinion that the method deserves a critical look and testing by those involved in hydrochernical studies. In order to demonstrate the potential range and/or limitations of the partial equilibrium principle, Kraynov et al.

(1980) applied it to the study of fluorine and calcium in ground waters. Two assumptions underlay their model: the rock (or sediment) is a continuous source of F

-

ions, and the final equilibrium concentrations are determined by the solubility product of CaF

2

. Using Fbearing granite columns through which solutions of NaHCO

3

, NfaSO

4

, and

36

NaC1 at various temperatures and flow rates were circulated, the authors found that: (1) as the rate of flow decreased, the loss of F from the granite decreased, and the [F•] and [Carl] increased up to saturation with respect to fluorite, and (2) the equilibrium was fairly rapidly achieved, after which the concentrations became independent cl the rock type, although dependent on the overall solution composition.

After calibrating their model, Kraynov et el. (1980) were able to reproduce measured (i.e., real-aquifer) concentrations of Ca" and F

-

, whose activity products became equal to the solubility product for fluorite as the measured flow velocities decreased. One of the most interesting conclusions was that the dissolution of fluorite occurs in the external-diffusion range, which means that the mass transfer is limited by convective diffusion at the surface of the mineral being dissolved. As the rate of flow is lowered, the rate of outflow of reaction products is also lowered, and the chemical potential gradient becomes smaller. Under those conditions, the system, locally speaking, has a better chance of approaching equilibrium, although large portions in the regional system may exhibit substantially large differences in the composition of the solutions, which is accounted for by the very definition of the partial equilibria principle. Finally, the authors gave a word of advice in indicating that the whole idea is not to identify a model that can reflect all the processes but to select one such that the results can be compared with the actual distribution of components in discrete segments of the thermodynamic system. Kraynov et al. (1980) also suggested that the lack of kinetic information may

37 hinder the hydrochemical forecasting, particularly in the study of highly dynamic systems.

With respect to the equilibrium chemistry of fluorine in the unsaturated zone, the analysis becomes even more cumbersome than that for aquifers due to the larger number of solid phases and other substances (soil acids, organic matter) that can participate in establishing an equilibrium concentration of F species. Because of its fine chemical approach, perhaps the most enlightening work on the subject is that by Elrashidi and Lindsay (1985 and 1986). In their work of 1985, the authors performed a series of batch experiments in which surface soil samples with textures varying between sand and clay were put in contact with F-carrying solutions at pHs ranging from 4.11 to 8.14.

Following an equilibrium period of about 24 hours, several chemical variables were measured, which led to the following conclusions:

(1) In neutral and calcareous soils, the concentrations of F

-

in water are controlled by the solubility of fluorphlogopite (10/4

3

A1S1

3

0

10

F

2

).

(2) In slightly acid soils, the F

-

concentrations are controlled by fluorite.

(3) In strongly acid soils, the mineral cryolite

(A1F

3

) controls the concentration of F.

In a later study, Elrashidi and Lindsay (1986) focused on the three minerals mentioned above and further examined the hypothesis that the F

-

solubility in soils may be governed by one or more of them. The method used was the same as in their 1985 work, although in this case,

38 the equilibrium concentrations were approached from both undersaturation and supersaturation. It was concluded that (2) and (3) above may not hold true because the equilibrium concentrations turned out to be undersaturated with respect to fluorite and cryolite. No mention is made of alternative solid phases influencing F

-

concentrations. It was found, nevertheless, that conclusion (1) above still held under the new experimental conditions.

A final annotation should be made about the F species that can be found in solution. The point is particularly relevant because the

F

-

ion is one of the important ligands responsible for complexing trace metals and keeping them in solution in natural waters. Indeed, F

-

can complex metals such as Be, Al, Sc, Ti, Zr, Hf, Nb, Ta, Fe, Si, and Sn

(Pittwell, 1974). Hem (1968) treated extensively the complexing of

Al

3

+ by F

-

, and found that for pHs below neutrality (ion strength = 0.5

NI, and [F

-

] varying between 0.5 and 12 ppm), the most important complexes with aluminum are A1F

2

+, A1F

2

+, A1F

3

°, and A1F4

-

. Slavek et al.

(1984), in studying the interaction between clays and dilute F

-

solutions, reached similar conclusions and added A1F

5

= to the list of potential complexes in those cases where enough F

-

is present in solution. Above neutral pH, the complexing of Al

3

+ by OH

-

can predominate to form Al(OH)

4 -

, thus freeing F

-

(Hem, 1968). At large sulfate concentrations, the radical SO

4

= can also compete with F

-

to form complexes with Al

3

+, giving rise to the formation of species like

Al(SO

4

)* and Al(SCO

2 -

(Hem, 1968).

According to Roberson and Barnes (1978), the only significant complex that F

-

can form with Si is SiF e r, provided that the pH is

39 below 5. The authors, however, stated that in most natural waters, the complexes that F

-

form with A1

84

and Fe

3

+ are more important than those with Si. As an example, Table 4 presents the F-speciation measured in two acidic natural water samples.

At low pHs, and even in the presence of small amounts of metal cations, the species HF° is formed, which greatly enhances the dissolution of quartz (Iller, 1979). The magnitude of such a phencmenon can be realized by considering the figures given by Iller (1979) for experiments conducted at 25°C: when the solvent is water, r (rate of dissolution of quartz, in mg mr

2

hr

-1

) is 4 x10

-8

; if the solvent is a solution 0.1 M HF, r is increased to 2.8-3.0; and if the solvent is a solution 1 M HF, r is as large as 110. This effect, however, disappears at a pH above 7.

As a final reference, Travi (1984) has suggested the possibility of formation of the complex mkg*, although it may be argued that for it to happen, the

mgr.'

content in solution should be very large and other species should be either absent or in negligible concentrations.

Fluorine Geochemistry: Some Aspects of Chemical Kinetics

Generalities on Reaction Rates

Before presenting those cases in which the time-dependent reactivity of fluorine minerals has been studied, it seemed appropriate to proceed with a brief review of the essential principles on the subject in order to be able to understand better the implications and scope of the reviemxiworks.

40

Table

4.

Distribution of

F species in

% of total F for two natural acidic water samples at

25°C

(after Roberson and Barnes,

1978, p.

254).

Fluorine

Species

F

-

HF°

AlF**

A1F2

4

A1F3 °

FeF**

FeF2

4

SiF6 =

BF

4 -

Sample

(pH

0.01

Nil

Nil

1

= 0.86)

0.004

0.6

92.0

2.8

0.004

4.7

Sample

2

(pH

= 2.25)

0.007

0.05

81.0

2.3

0.002

16.0

0.03

Nil

41

Briefly stated, chemical kinetics is the study of rates of chemical reactions. Because the word "rate" is part of the label, it follows that time is the master variable in this type of study.

Indeed, the primary goal of considering kinetics is to determine how far and at what pace a given reaction proceeds as time goes by.

Eventually, such information will allow the investigator to assess the prevalent reaction mechanism(s), that is, the collection of individual molecular steps of elementary reactions accounting for the net reaction.

The rate of a reaction in aqueous solutions can be influenced by a number of variables (Espenson, 1981) such as: (1) concentrations of reactants, products, and possible catalysts, (2) temperature and pressure, as the primary physical conditions, and (3) the properties of the solvent (viscosity, dielectric constant, ion strength). Temperature is certainly one of the most important variables. For example, some rate constants can change by several orders of magnitude as the temperature is varied by 100°C (Lasaga, 1981). The reason for such a drastic variation is the exponential dependence of the rate constant on temperature, which is commonly assumed to comply with the equation proposed by Arrhenius in 1889: k = A exp(-E a

/RT) (2.4) where k is a rate constant, A is the pre-exponential factor, E a

is the activation energy (kcal/mol), R is the universal gas constant (1.987 x

107

3

kcal/mol°K), and T is the absolute temperature (°K).

42

According to Equation 2.4, a plot of log k versus 1/T is linear, with slope equal to -E a

/2.303R and intercept equal to log A.

An alternative form of tenperature dependence, which has been derived from the transition state theory, is presented by Lyman et al.

(1982) as: k = (1/h)k

0

T exp(-AFP/RT) exp(AS°/R) which can also be expressed as: log k/T = log k b

/h - LIP/2.303RT + AS

(5

/R where kb is the Boltzmann's constant; h is the Planck's constant; Ali° is the enthalpy cl activation; S

0

is the entropy of activation, and k,

R, T are as above defined. According to the last expression, AH° can be calculated from the slope of a plot of log k/T versus log 1/T, and

ASP from the intercept of such plot.

Other investigators choose to fit the k and T data using relationships of the form (Lyman et al., 1982): log k = -A/T + B log T + C where A, B, and C are empirical constants.

In practice, any one of the equations presented usually gives a good fit to experimental data because of the small number of data points and the uncertainties in the individual values measured. For

43

example, as applied to the Arrhenius equation (2.4), it can be seen that the pre-exponential factor

A

and the activation energy E a

are adjustable parameters. Thus, the error in k associated with a particular uncertainty in T is given by (Lyman et al., 1982) as:

k/k = ( E a

/RT ) 6T / T

For example, for a reaction taking place at about room temperature with a typical activation energy of 8 kcal/mol, the error in k associated with an uncertainty of ±0.5

0

C is:

Sk/k (%) = (8/1.987 x 10

-

3

300) (0.5/300) 100 = 2.2 %

Lasaga (1981) has pointed out that, whereas the Arrhenius equation is not derived fully from molecular theory (i.e., it is a phenomenological relationship), detailed molecular theories often result in a temperature-dependent pre-exponential factor. Furthermore, the activation energies for overall reactions may also depend on temperature.

This characteristic would be seen in curvatures in the plot of log k versus log 1/T (Lasaga, 1981).

Dissolution Rate of Minerals

In the material presented above, major emphasis has been placed on the relationship between temperature and mineral reactivity.

Mbntion should also be made of the processes which limit the rate at which minerals dissolve in undersaturated solutions. The so-called

44 rate-limiting mechanisms can be classified into three main categories

(Berner, 1978): (1) transport of solutes away from the surface of the dissolving mineral, (2) detachment of ions or molecules from the surface of the mineral, and (3) canbination of transport and detachment. In the first case, we are dealing with pure "transport-controlled" kinetics (termed also "solution-transport" mechanism in Lasaga

(1981), and even "external diffusion" mechanism in Kraynov et al.

(1981), and in Dobrovol'skiy and Lyal'ko (1983)), in the second with

"surface reaction-controlled" kinetics, and in the third case with a mixture of the two previous types or "partial surface reaction-controlled" kinetics. The characteristics of those three processes can be visualized in Figure 1 and from the precise explanation offered by

Berner (1978, p. 1237):

"In pure transport-controlled dissolution, ions are detached so rapidly from the surface of a crystal that they build up to form a saturated solution adjacent to the surface. Dissolution is then regulated by transport of these ions via advection and diffusion into the surrounding undersaturated solution. The rate of dissolution depends upon flow velocity and the degree of stirring with increased stirring and flow resulting in accelerated transport and, therefore, faster dissolution. (In the special case where crystals are so all that they are carried along with eddies, dissolution is not enhanced by further stirring...).

In pure surface reaction-controlled dissolution, ion detachment is sufficiently slow that ion buildup at the crystal surface cannot keep pace with advection and diffusion, and the resulting concentration level adjacent to the surface is essentially the same as that in the surrounding solution (see Figure 1). In this case, increased flow and stirring rate have no effect on dissolution rate.

Intermediate cases between pure transport-control and pure surface reaction-control exist where surface detachment is sufficiently fast that the surface

45

A.

B.

C.

C eq

cea

Ceq

C o

.

0 r-p

0

,•••nn•

0 r

Figure 1. Concentration C versus distance from a crystal surface r for three rate controlling processes (C eq

= saturation concentration; C= concentration out in solution).

(A)

Transport-control; (B) Surface-reaction control;

(C) Mixed transport and surface-reation control.

(After Berner, 1978, p. 1238).

46 concentration builds up to levels greater than the surrounding solution but lower than that expected for saturation.

This is also shown in Figure 1."

Dissolution controlled by surface reaction proceeds at a slower pace than by transport control because the surface detachment is more rapid in the latter, which makes the rate of removal be the limiting step. If one considers that the slowest fashion in which a mineral can dissolve through transport-controlled kinetics is by pure diffusion (no advection, water stagnant over the time and distance scales of molecular diffusion), it follows that the rate of diffusion-controlled dissolution should define the boundary between the more rapid transportcontrolled rates and the slower surface reaction-controlled rates

(Berner, 1978). This is a very important point because it highlights the usefulness of the laboratory-determined dissolution rates in distinguishing the rate-limiting mechanism operating in the system. As a matter of fact, and according to Berner (1978), that is the most important piece of information that laboratory-measured rates can provide.

Berner (1978) indicated that it is extremely difficult to predict accurate absolute values for dissolution rates from laboratory data due to the differences between the field and the laboratory in factors such as the surface composition and configuration of the mineral particles, the effect of trace species subject to adsorption onto the mineral surface (this "surface poisoning" effect has also been mentioned by Mbrse, 1983), and the catalytic or inhibiting role of organisms in natural waters.

47

With respect to activation energy E a

values characteristic of the mechanisms mentioned above, Berner (1978) and Lasaga (1981) agreed that

Ea

's in the vicinity of 4-5 kcal/mole may typify diffusioncontrolled dissolutions, whereas surface reaction-controlled kinetics should result in an overall empirical E a

distinctly greater than 4-5 kcal/mole.

Berner (1978, 1981) has also postulated the existence of a good correlation between the solubility of minerals and their ratecontrolling mechanisms of dissolution. Based an numerous data (a few examples are shown in Table 5), the author has concluded that for solubilities below 10r

4

M (moles/liter), surface reaction dominates, for solubilities in the range 10

-4

- 10

-3

M either surface reaction or transport is the rate-controlling mechanism, whereas for solubilities above 2 x 10-

3

transport control dominates.

It is important to point out that, based on his omprehensive work, Berner (1978) concluded that many of the dissolution processes occurring under normal earth surface conditions are controlled by surface reaction and not by transport or diffusion.

Fluorine Minerals Dissolution Kinetics

In spite of its relevance in determining the tine-dependent features of fluoride ion concentrations in solution, little work has been done to assess the kinetics that govern the dissolution and precipitation of F-bearing minerals. Although many investigations provide clues or data which can be interpreted by reference to laws of kinetics, only one published report was found in which the kinetics

Table 5. Dissolution rate-controlling mechanism for various substances arranged in order of solubilities in pure water. (After Berner, 1981, p. 126)

Substance

Solubility

(mole per liter)

Dissolution

Rate Control

Ca

5

(P00

3

01-1 2 x 10

-8

Surface-reaction

KA1S1

3

0

8

3 x 10

-7

Surface-reaction

NaA1Si

3

0

8

6 x 10

-7

Surface-reaction

BaSO

4

1 x 10

-5

Surface-reaction

AgC1

1 x 1 0r

5

Transport

Sr(K)

3

3 x 10r

5

Surface-reaction

CaCC1

3

6 x 10-

5

Surface-reaction

Ag

2

0

4

1 x 10

-4

Surface-reaction

PbSO

4

1 x 10

-4

Mixed

Ba(I0

3

)

2

8 x 10r

4

Transport

SrSO

4

9 x 10r

4

Surface-reaction

Opaline Si0

2

2 x 10

-3

Surface-reaction

CaSO

4

.2H

2

0 5 x 10

-3

Transport

Na

2

SO

4

.10H

2

0

MgSO

4

.7H

2

0

Na

2

KC1

CO

NaC1

M3C1

2

3

.10H

.6H

2

0

2

0

2 x 10

-1

Transport

3 x 10P

Transport

3 x 10P

5 x 10P

Transport

4 x

1

0° Transport

Transport

5 x 10° Transport

48

49 associated with the dissolution of fluorite are dealt with throughout the whole paper. This report is by Dobrovol'skty and Lyarko (1983) and will be summarized below.

In referring to the differential equation describing solute transport, Dobrovol'skiy and Lyal'ko (1983) pointed out that the always present source/sink term cannot be treated as a constant and that it should inclnde the dissolution rate. Such a term would be independent of the infiltration rate if the rate-limiting step were the reaction at the surface; conversely, it would be flow velocity-dependent if the transport of products away from the dissolving surface were the dissolution rate-controlling mechanism. In order to study the problem, the authors performed batch and column experiments in which crushed fluorite was put in contact with solutions of various compositions and the chemical evolution of the system closely monitored. Briefly presented, the conclusions that Dcbrovol'skiy and Lyal'ko (1983) drew from their batch experiments are as follows:

(1) The dissolution rate increased as the stirring rate of the batch solutions was increased, which led the authors to indicate that the transport stages play a considPrable role (i.e., external diffusion is the rate-controlling mechanism). The same suggestion had been made by Kraynov et al. (1980) in discussing the kinetics of fluorite dissolution, although it may be important to recall that Berner (1978) had shown that, under average crustal conditions of temperature, pressure, and solutions composition, most minerals

should have their dissolution rates controlled by surface reactions rather than by transport mechanisms.

Such a difference of opinion might be linked to the fact that in their model, Dobrovol'skiy and Lyal'ko

(1983) considered that the specific surface area of the mineral particles being dissolved did not change throughout the time span of their experiments. This assumption, certainly attractive for its simplicity and not critical in many cases, may not hold true in some circumstances. For example, Baer and Lewin (1970), in dealing with the kinetic replacement of calcite by fluorite, found that the rate of conversion increased as the surface area of the exposed mineral increased.

(2) In those experiments where the dissolution rate was measured as a function of temperature and the Arrhenius model assumed to describe the data, the authors found a distinct change of slope (i.e., activation energy) in the plot, which was interpreted as a change in macrokinetics: between 0.5 and 13.5°C, the dissolution rate is determined by surface reaction, between 23 and 100

°C external diffusion governs the dissolution process, whereas in the range 13-22°C, the state was defined as a transitional one in which the rates determined by surface reaction and diffusion are comparable (i.e., mixed control).

50

(3) In examining the relationship between the dissolution rate and the ion strength (p) of solutions, the authors concluded that the effect of increased solution concentrations was negligible as long as p was below

0.05 NI; that is to say, changes In the dissolution rate due to changes in TDS (total dissolved solids) should not be expected in low-salinity waters.

(4) The addition of phosphate to the batch solutions substantially retarded the dissolution of fluorite.

Indeed, the presence of phosphate will result in minute amounts of F

-

in solution because of the solubility control on F

-

exerted by fluorapatite (solubility product = 1.4 x 10

-119

; Handa, 1977). If this is the case, then the dissolution of fluorite will proceed at a much slower pace and will be controlled by surface reaction (Berner, 1978).

51

With reference to the data from column experiments, which were designed to obtain information about the effect of the infiltration rate on the dissolution of fluorite, Dobrovol'skiy and Lyaliko (1983) offered the following conclusions:

(1) The dissolution of fluorite became stationary after a few hours, and the magnitude of the dissolution rates measured was directly correlated to the water infiltration rate through the column. The empirical relationship of the authors shows that k (dissolution rate) is

a function of v

(infiltration rate) raised to a certain power. At this point, it may be pertinent to recall that Berner, in his

1978 paper, cautioned that extreme care must be taken in assuming that the dissolution rate of minerals is directly proportional to the flushing rate. Berner

(1978) indicated that the only clear case in which flushing controls the rate of dissolution is in the processes occurring in the unsaturated zone of a soil simply because of the need of water; in other cases, the main aspect to consider is the rate of flushing relative to the rate of mineral reaction.

(2)Due to the common-ion effect, the dissolution rate of fluorite decreased as the Ca" concentration in solution increased, although such a decrease was noticeable only at very high

[Ca

4-4

] (2000-4000 ppm).

The authors indicated that the activity product for fluorite depends on the concentrations of Ca" and

F

-

, as well as on the activity coefficients, and that in natural waters, the rates of reduction in the activity coefficients at low and medium concentrations are greater than the rate of increase in the Ca" concentrations.

(3)Although theory predicts that an increase in the concentrations of those ions not contributing to the common-ion effect should tend to increase the dissolu-

52

53 tion rate (because of the corresponding decrease in the activity coefficients), the authors found the opposite effect: when the concentration of NaC1 in solution was inammuNI up to 1

NE,

a regular decrease in the dissolution rate was measured.

(4) Based on the whole set of experimental data,

Dobrovol'skiy and Lyaltko (1983) concluded that the variations in the dissolution rate did not follow the activity product for fluorite, which was interpreted as an indication of the dependence of the dissolution rate on the bulk solution ccmposition and concentration.

Finally, Dobrovol'skiy and Lyal'ko (1983) presented the results of a computer simulation of the leaching of fluorite-bearing rocks in aquifers. They found, again, that flow velocity was a critical parameter. An explanation was offered in the following terms: the larger the flow velocity, the better the chances that dissolved products have to be carried away, with a corresponding enlargement of the region in the aquifer containing water unsaturated with respect to fluorite.

Furthermore, the model employed highlighted the central role of porosity in reducing (or increasing) the active specific surface area of the mineral in contact with flowing solutions, thereby increasing

(or reducing) the size of the unsaturation region.

54

Fluorine Geochemistry: Adsorption and Other Mbdifying Phenomena

Concepts and Terminology

Definitions. Simply defined, adsorption refers to the fixation of molecules and/or ions from a solution on the surface of a solid phase. Adsorption may be specific or non-specific (Mbrrill et al.,

1982). Specific adsorption occurs when particular charged sites of a surface exert strong forces on molecules or ions of a particular makeup. Ions absorbed by this process are termed "specifically adsorbed" or "ligand exchanged" (Hingstan, 1981). Non-specific adsorption, on the other hand, is more general and includes all adsorption that cannot be accounted for solely by electrostatic forces (Hingston,

1981).

Adsorption is essentially a two-dimensional phenomenon in the sense that it deals with the interactions between solutes and solid phases at the surface of the solids. It should not be confused with absorption, a three-dimensional process by means of which there is an effective diffusion of molecules or ions within the mass of an absorbing phase. To include both adsorption and absorption, the term sorption is used.

Adsorption models. The amount of a given compound adsorbed onto the surface of a solid depends on the number of potential adsorption sites, and is usually a function of the specific surface of the solid particles and their charge density. Mbst information on adsorption is obtained from batch experiments by using a "slurry" technique, i.e., shaking a known mass of solids in a solution having known concentrations of dissolved substances until equilibrium is reached. The

55 difference between the initial and final contents of the species in solution, normalized by the weight of solids, is designated as adsorbed by the solid phase. The process is repeated at constant temperature for a number of different initial concentrations, from which "adsorption isotherms" are obtained. Although such isotherms may have at least four different shapes (Mbrrill et al., 1982), the results are normally desm'itxkl by either the Langmuir or the Freundlich equations.

The Langmuir equation, originally developed for adsorption of gases on solids (Mbrrill et al., 1982), can be written in the following form for adsorption of species from solution: x/m = K

1

K

2

C/(1 + K

2

C) or,

C/(x/m) = 1/K

I

K

2

+ C/K

2

(2.5) where x/m is the amount adsorbed (x) per unit mass (m) of the solid phase, C is the equilibrium solute concentration in solution after adsorption, K

1

is a constant related to bonding energy, and K

2

is a constant describing the maximum adsorption. According to Equation

(2.5), the Langmuir constants are obtained by plotting C/(x/m) against

1/C. The intercept of such a plot gives 1/K

I

K

2

1

and the slope is 1/K

2

On many occasions, the good fit between the data and the

Langmuir equation has led to the conclusion that the assumptions underlying the method are met (Mbrrill et al., 1982). Some authors,

56 however, have raised serious objections indicating that such a good fit might be the result of fortuitously compensating factors (Braman and

Chesters, 1975). Veith and Sposito (1977) and Sposito (1982), for example, have pointed out that the interpretation cl the parameters of the Langmuir is possible only if independent evidence exists that adsorption is the sole phenomenon causing the solute to disappear from solution because the method cannot actually distinguish between pure adsorption and precipitation.

In those cases where the Langmuir equation fails to adequately describe the experimental data, the Freundlich equation is used, which can be written as follows:

S

= K d

O/b where S, equalled to x/m, is the mass of adsorbate (x) per unit mass of adsorbent (m), E' d

is the equilibrium constant, also known as partition coefficient or distribution coefficient (Freeze and Cherry, 1979), C is the equilibrium concentration of solution after adsorption, and b is a constant, usually less than one and indicative of the degree of linearity of the adsorption process. The Freundlich equation can be expressed in logarithmic form as log S = log K a

+ lib log C, and the experimental data can be plotted as log S against log C, which yields log K d

as the intercept and 1/b as the slope. According to Mbrrill et al. (1982), K t1

is roughly an indicator of the adsorption capacity and l/b of the adsorption intensity.

57

Physical adsorption. The phenomenon of physical adsorption results from electrostatic interactions between atoms, ions, and molecules due to electron fluctuations that produce instantaneous dipoles (Mbrrill et al., 1982). It is, in other words, the result of van der Waals forces. Such forces are weak and decrease very rapidly with the distance between the interacting species. The energy of adsorption due to van der Waals interactions, which are exothermic in nature, has values typically in the range of 1 to 2 cal/mol, although some authors have indicated values as high as 2-6 Kcal/mol (Mbrrill et al., 1982).

Chemisorption. In dNamisorption, electrons are redistributed in new orbitals and, when the electrons spin align according to Pauli 's principle, a relatively permanent chemical bond is formed between the surface atoms and the adsorbed species (Mbrrill et al., 1982). An exothermic phenomenon, chemisorption is characterized by heats of adsorption usually greater than 20 Kcal/mol (Lasaga, 1981). According to Mbrrill et al. (1982), in short-term adsorption studies, chemisorption is not a very active process and generally slower than physical adsorption.

Retardation. If the Freundlich nodal can be applied (i.e., the adsorption isotherm is linear) and the reactions are fast and reversible, the distribution coefficient K a can be used in modeling the transport of the solute. A given species in solution which interacts with the surrounding solid phase(s), as in the case of adsorption, will not travel at the same velocity of that of the solvent. Said in an equivalent manner, the solvent moves ahead of the reactive solute, and

58 the adsorbed species is then said to be retarded. In those cases, the retardation of the solute front relative to the bulk mass of solvent is described by the following relationship (Freeze and Cherry,

1979): vc/v = 1/[1 (Pb/n)K,al where I/ is the average linear velocity of the solvent, v c

is the velocity obtained from dividing the length traveled by the reactive species by the time required to measure a concentration of approximately

50% of that of the solution at the source,

Pb is the bulk mass density, n is the porosity, and

K

d

is the distribution coefficient.

The ratio v c

/v is known as relative velocity, and the term

[1 +

(P b

/n)K a

] is referred to as the retardation factor, R.

In order to quantitatively appreciate the effects of chemical retardation, Freeze and Cherry

(1979) assumed that the average mass density of minerals in unconsolidated deposits is about

2.65, with porosities in the range of

0.2 to

0.4.

Thus, the range of bulk mass densities that corresponds to such a porosity range is

1.6 to

2.1

gm/an.

The term

Pb

In, therefore, will have values ranging from

4 to

10 gm/ce.

Hence, the retardation factor will take on values within the following range:

R = v/ve c

= (1 + 4K t1

) to

(1 +

'OK' d

)

Keeping in mind that

K

4

can be expressed as the ratio between the mass of solute on the solid phase per unit mass of solid phase and the

59 concentration of solute in solution (with reduced dimensions of L

3

/M, usually ml/gm), the expression above indicates that, if K u 1 mil/gm, the mid-concentration point of the solute would be retarded by a factor between 5 and 11 with respect to the bulk solvent flow. Inasmuch as measured values for K ol

cover a wide range, from near zero to 10

3

ml/gm or greater (Freeze and Cherry, 1979), for values of

K

d

that are orders of magnitude larger than 1 ml/gm, the solute is immobile for most practical considerations.

Ion exchange. Simply defined, ion exchange is a phenomenon in which a given ion is withdrawn from solution in exchange for another ionic species contained or adsorbed on a solid. Ion exchange is influenced by the solvation of the adsorbent in the region of the adsorption site and by the solvation of the adsorbate ion (Morrill et al., 1982).

Because most of the solid phases ccumnly encountered in aqueous natural environments are negatively charged, the type of exchange chiefly taking place is cation exchange. It is for this reason that not much information is usually available on anion-exchange reactions. Nonetheless, the phenomenon does take place, thereby anions are attached either electrostatically or with a certain degree of chemical bonding to natural solids (Morrill et al., 1982). The main control variables in anion-exchange reactions are the concentration of exchangeable anions, the concentration and composition of the bulk solution, the presence of potential inhibiting species, and, above all, the pH. The general consensus is that anions are not strongly exchanged within the normal pH range of natural systems, although the

60 process may sometimes be qualitatively significant and worth studying in those cases where there is no clear distinction between anion exchange and isomorphic substitution, such as in minerals like montmorillonite (Mbrrill et al., 1982).

Precipitation. Although similar in many respects, specific adsorption and precipitation have their own characteristics. Adsorption, as said before, is a two-dimensional process, whereas precipitation involves the consideration of a three-dimensional domain. Such a statement, although theoretically correct, constitutes an oversimplification, because it is sometimes very difficult to be positive as to which process is actually taking place. Corey (1981, p. 161) has summarized in the following manner the possible consequences of adding an adsorbate ion to a system containing mineral adsorbents:

"1. Crystal growth. This occurs if the adsorbate is a component of the mineral adsorbent.

2.

Crystal growth and/or diffusion into the solid phase. If the adsorbate is not a component of the adsorbent but can substitute iscmorphously for a component of the adsorbent and form a stable, three-dimensional solid solution, this will take place.

3.

Formation of a stable surface compound (two-dimensional solid solution). This takes place if the adsorbate is not capable of forming a three-dimensional solid solution with the adsorbent.

4.

Stabilization of metastable, polynuclear ions. This occurs by adsorption onto oppositely-charged surfaces of the adsorbent.

5. Heterogeneous nucleation of a new solid

phase.

This involves a new phase composed of the adsorbate and a component fran the solution (hydroxides, carbonates, etc.).

6. Heterogeneous nucleation of a new solid

phase.

This refers to a new phase composed of the adsorbate and a component of the adsorbent resulting in dissolution of the adsorbent."

61

Corey (1981) pointed out that purely adsorption reactions should be defined in terms of the formation of a stable surface compound in equilibrium with the solution (that is, number 3 above), and that errors in the interpretation have sometimes been traced to the failure to recognize the possibility of occurrence cf other reactions.

Another important point, which complicates the picture even further, is that supersaturation is needed to initiate precipitation.

As explained by Corey (1981), a critical degree of supersaturation is required because the solubility of the precipitate increases with decreasing particle size, and the size of the nucleus is initially very small. For the case of homogeneous nucleation (i.e., when nucleation that leads to precipitation occurs through chance collisions of reactants in solution), Stumm and Mbrgan (1981) have calculated that degrees of supersaturation as much as 10 times larger than the solubility product would not result in precipitation even within geological time spans.

As pointed out by Corey (1981), the fact that an ion is specifically adsorbed onto the surface of a solid suggests that it has a tendency to form an insoluble compound or stable complex with ions of opposite charge in the substrate. Therefore, as more and more ions adsorb on the solid surface, chances are that the nucleation of a new solid phase will start at some point, and the solubility of the adsorbate ion would then be controlled by the solubility of the new solid phase rather than by the on-going adsorption reaction.

Two of the most relevant conclusions of Corey's work (1981) are that (1) the exact boundary between adsorption and precipitation seems

62 difficult to determine, even for simple systems, because many laboratory experiments have been carried out in supersaturation conditions with respect to a potential new solid phase involving the adsorbate ion, and that (2) the concentrations of ions that are minor constituents in natural systems are usually controlled by adsorption reactions or solid-solution equilibria involving poorly-defined systems.

Results of Fluorine Adsorption-Related Studies

Due to the potential toxic effect that fluorine can have on humans, animals, and plants, many studies have been carried out to assess the mobility of the various compounds of fluorine in natural systems. Of particular interest to scientists has been the fate of fluorine species which have been artificially added to soils (fertilizers, sewage sludge, airborne emissions from aluminum smelters, phosphate processing plants, and enamel factories, etc.). It is precisely for this reason that most of the information available on fluorine adsorption processes comes from the soil research literature.

Although some obvious differences between the chemical phenomena in unsaturated and saturated media can be found, it is assumed that the overall conclusions can be applied to the study of either environment.

The study of the behavior of fluorine species in soils or aquifers is not a simple task, and the simultaneous consideration of several hydrochesnically-related mechanisms is complicated. As a direct consequence of such a fact, the reader will find a certain heterogeneity (both in methodology and in treatment of information) in the following presentation of data and conclusions from studies dealing

63 with fluorine. Realistically speaking, it is close to impossible to carry out experiments in which all or most of the modifying phenomena can be accounted for. Hence, investigators usually choose to focus on particular processes and riPqign their studies so that the effect of other mechanisms can be neglected. Given those circumstances, the interplaying of quality and quantity should be borne in mind by those interested in doing a worthy literary review. It is my opinion that a clear and convincing picture on the subject will emerge only after as many sources as possible have been reviewed and their results have been evaluated and compared. This, which may well be deemed to be an obvious truth (verdad de Percgrullo) applicable to any type of study, is particularly relevant in fluorine geochemistry research.

In line with the introductory remarks above, some of the investigations done on fluorine adsorption and related topics will now be reviewed in chronological order.

In trying to determine the availability of nutrients in soils,

Dickman and Bray (1941) carried out experiments in which kaolinite was treated with neutral sodium fluoride solutions. The amount of F

-

that had vanished from solution after equilibration was found to correlate with a noticeable increase in pH (40 meq 0H

-

/100 gm clay were released as 41 meq r/100 gm clay were adsorbed), which led the authors to conclude that the stoichiometric replacement of hydroxyl ions by fluoride on the clay surface (i.e., ion exchange) was the mechanism operating in the system. The increase in pH was calculated by keeping count of the amount of diluted HC1 needed to maintain a pH of about 7 during the course of the experiment.

64

The active participation of F- in ion-exchange reactions has also been mentioned in such classical works as those by Rankama and

Sahama (1950) and Goldschmidt (1954).

In a study of fluoride adsorption by clay minerals and hydrated alumina, Samson (1952) described an experiment in which kaolinite and huatmorillonite were shaken with neutral NI-1

4

F solutions for three hours, which resulted in: (1) the solution becoming alkaline, (2) F

being adsorbed, and (3) aluminum being detected in solution. Samson

(1952, p. 269) offered the following explanation: "The adsorption process appears to consist of an immediate displacement of surface hydroxyl groups, due to co-ordinative binding between fluoride ions and aluminum ions in the octahedral layer. Adsorption sites are thus available at the edges of clay mineral platelets. In addition, a kaolinite crystal has one basal surface composed of Al-OH linkages.

The more accessible aluminium ions are evidently dissolved quite readily, and co-ordinate six F

-

ions in solution. The removal of these aluminums at plate edges exposes OH ions for replacement by F

-

, especially in the kaolinite structure, where Al: 01-I = 1:2."

Samson pointed out that, when comparing the OH

-

released and the F

-

adsorbed, the buffer capacity of the clays is to be taken into account because the amount of

a"

-

released would otherwise be underestimated. However, when allowance was made for such a buffering effect, the F

-

adsorbed was still larger than the OH

-

release, which the author took as an indication of the alai nature of fluoride adsorption: replacement of hydroxyl groups and dissolution of surface alumina.

65

Romo and Roy (1957), motivated by the description of the two effects made by Samson (1952), investigated further the reactions between neutral solutions of fluoride and clay minerals, in particular with reference to the apparent formation of various solid complex fluoride phases. When kaolinite was brought into contact with F

solutions (NaF), the authors found that the fluoride uptake increased with time, temperature, and [F

-

] in the soaking solution. The positive correlation between the amount of F

-

adsorbed and the [F

-

] in solution was interpreted as a clear evidence that the purely-exchange F

-

by

OHhypothesis might not offer a unique answer as to the active mechanism because the presence in solution of excess F

-

ions would have had no effect when equilibrium was attained. It should be mentioned that the experiments, which lasted three months, were carried out at neutral pH and that the buffering capacity of the clay minerals was disregarded based on earlier studies by Romo (1952). The authors, however, did indicate that there was a reasonable agreement between the amount of

ccr

released to solution and that of F

-

found in the solid phases after the experiment concluded, which supported the exchange theory. On the other hand, mention was also made of the formation of a new phase, cryolite (NatA1F

6

), which was in favor of what the authors called the

"decomposition" hypothesis, in essential agreement with Samson's findings (1952). Rom and Roy (1957) suggested the following mechanisms to account for their observations:

66 exchange: Al

2

S1

2

0

5

(OH)

4

+ xNaF <===> Al

2

S1

2

(OH)

4

_ x

F x

+ xNaOH decomposition: Al

2

S1

2

01

5

(OH)

4

+ 6NaF <===> Na

3

A1F

6

+ A1Si

2

0

5

(OH) +

3NaOH

With regard to the exchange hypothesis, the authors indicated that two simultaneous and independent processes may be taking place, one involving direct exchange with the solution and the other a slow diffusion of fluoride ions into the lattice. For kaolinite, the exchange of exposed hydroxyls by F

-

ions was characterized by a rate 10 times faster than that of the diffusion process. Although recognizing the fairly extensive and rapid transfer of F

-

ions from solution to the solid phase when brought into contact with layer silicates, Romo and

Roy (1957) could not present conclusive evidence of the mechanism of such a fluorine fixation.

In a report describing the effect of artificially-added HF to sandy soils (characterized by a high content of quartz), Spetch and

MbIntire (1961) concluded that most of the fluorine had been retained by conversion to CaF

2

through reactions with the calcium carbonate of the soil, which also produced a moderate elevation of the soil pH.

According to the authors, such a finding would explain why certain

Florida soil systems retained 98% of the fluorine contained in 4932 pounds per acre of calcium fluoride even after ten years of leaching by rain water.

Huang and Jackson (1965) attempted to elucidate the uncertainty as to the mechanism(s) of interaction by studying the effects of mixing

67 silicate clays (carmercial vermiculite, bentonite, kaolinite, halloysite, chlorite, allophane), oxides (hematite, goethite, gibbsite, quartz), and soils of various canpositions with neutral KF solutions.

It was found that the pH of the batch solutions increased as F

-

was absorbed, and that the amount of CEr released increased as the soilwater ratio r decreased. The authors, however, reported that Xray tests for mineral-bonded F

-

conducted right after the experiments were negative, which was attributed to leaching of the bonded F

-

by a water-ethanol mixture used to wash the samples five times before the analytical measurements were made. The authors interpreted such a loss as an indication of the weak nature of the F

-

bonding, and that of the minor significance of the replacement of lattice

am

-

by F

-

. After equilibrium was reached, at a pH higher than that of the neutral initial solution, aluminum and iron were detected in solution, which

Huang and Jackson (1965) claimed to be a strong evidence of the occurrence of complexing reactions with dissolution of sesqui-oxides and formation of flucroaluminate and fluoroferrate. Defining M as a moncvalent cation, Huang and Jackson (1965) postulated the following reactions:

Al(OH)3 + GMF <===> N6A1F

6

+ 3MIDE

Fe

2

0

3

+ 104F + 3H

2

0 <===> 2diFeF

5

+ aNIDE

Another proof of the effect of fluoride solutions attack on minerals is presented by the authors in the form of microplx*.ographs shaaing

68 fractured surfaces of otherwise smooth muscovite cleavage planes. The authors pointed out that quartz turned out to be rather inert to KF solutions, which was taken as an indication of the negligible interactions between F

-

and xSi-OH groups at a neutral pH.

New evidence on the subject was presented by Semmens and Meggy

(1966a), who studied the reactions between kaolinite and neutral and acid NaF solutions. For the effect of neutral F

-

solutions on kaolinite, a period of study of three months was taken at a temperature of about 20°C. After such a reaction period, the amount of F

-

lost awl solution was correlated with the appearance of a new solid phase, fluoroaluminate (cryolite).

The effect of acid NaF solutions (with the desired pH obtained by means of the addition of acetic acid) was investigated by treating kaolinite samples with three solutions at pHs of 3.4, 4.7, and 5.0 (at a constant temperature of 20 °C). In the three cases, the pH was observed to increase as the reactions proceeded. After an overall equilibrium was reached, X-ray diffractometer traces and powder camera photographs indicated that the kaolinite residue contained at least two new additional solid phases which were identified as sodium fluorosilicate and sodium fluoroaluminate. Examination of the kaolinitic residues after treatment with both neutral and acid NaF solutions revealed the presence of small, spherical particles dispersed among the hexagonal kaolinite crystals, which were thought to be colloidal silica. Taking into account that the ratio between the ar released to solution and the F

-

lost from it was around 0.18, and that positive evidence existed an the appearance of new solid phases, the authors

69 concluded that the earlier reported stoichiometric exchange of hydroxyls in the clay lattice by F

-

ions may be fortuitous because the basic reaction is one of destruction of the crystal lattice, with the F

-

ions being lost from solution to the solid phase in the form of an insoluble complex. To summarize their results, Semmens and Meggy (1966a) proposed the following reactions:

For pH < 7 and in the presence of Na ions:

Al2Si205(OH)4 + 10Nae + 24F

-

+

141

3

0r

<===> 2Na

3

A1F

6

+ 2Na

2

SiF6 + 2311

2

0

For pH > 7 and in the presence of Na ions:

Al2 Si

2

0

5

( OH )

4

+ 6Ne + 12F

-

+ 6H

3

Cr

<===> 2Na

3

A1F

6

+ 2S10

2

+ 11H

2

0 or, if the presence of orthosilicic acid is accepted instead of gel formation:

Al

2

Si

2

0

5

(OH )

4

+ 6Na* + 12F

-

+ 6H

3

<===> 2Na

3

A1F

6

+ 2Si(OH)

4

+ 7H

2

0

Although of limited relevance for the study of natural systems, it may be mentioned that, in a related study, Semmens and Méggy (1966b) treated kaolinite samples with solutions of hydrofluoric acid (HF). It was calculated that there is a preferential loss of aluminum from the kaolinite crystal lattice to the acid solutions (no formation of new phases was detected in the kaolinite residues), especially at higher temperatures, with formation of colloidal silica. The reaction mechanisms postulated were as follows (Semmens and Meggy, 1966b):

Al

2

S1

2

0

5

(OH)

4

+ 24F

-

+

1411

3 cr <===>

2A1F

6 3-

+ 2SiF

6

+ 23H

2

0

70 and, for higher temperatures:

Al

2

S1

2

0

5

(OH)

4

+ 12F

-

+ 611

3

0' <===> 2A1F

6

3

+ 2Si0

2

:H

2

0 + 9H

2

0

Pointing out that most of the work done up that point used solutions of rather high [F

-

], Bower and Hatcher (1967) decided to study the reactions between fluorine species and soils and minerals at low (i.e., more natural) F

-

contents. Five alkaline soils, an acid soil, and hydroxyl-containing minerals commonly found in soils were submersed in NaF solutions containing from 2 to 16 ppm of F

-

. The conclusions from their study can be summarized as follows:

(1) Gibbsite, kaolinite, halloysite, freshly precipitated

Al(OH)

3

, and all tested soils adsorbed significant amounts of F

-

with release of OH

-

.

(2) The LangmArmodel fitted the adsorption data.

(3) Goethite, montmorillonite, and vermiculite adsorbed only traces of F

-

.

The authors indicated that, while the pH of the suspensions increased with increased F

-

adsorption, the amount of

cm

-

released was very small compared to the amount of F

-

adsorbed. Inasmuch as the systems exhibited buffer effects, the authors pointed out that the hypothesis of purely ion exchange could not be rejected. Moreover, Bawer and Hatcher

(1967, p. 152) added that "As a mass-action type of ion-exchange equation, such as would apply to F

-

-ar exchange, reduces to the

Langmuir equation if the activity of one ion in solution remains

7 1 constant, the very small change in

CEr activity over the range of adsorption studied probably explains the ability of the Langmuir equation to describe the adsorption data." However, and based on (a), the little effect on F

-

adsorption caused by enormously increasing the amount of exposed OH- upon converting dehydrated alloysite to the hydrated form, (b) the high adsorption capacity of freshly-precipitated

Al(OH)

3

, and (c) the increase in adsorption capacity upon acid washing, the authors concluded that the F

-

adsorption occurs primarily by exchange with OH groups from Al(OH)

3

and from basic Al polymers adsorbed on mineral surfaces, rather than by exchange with lattice OH groups. In line with this hypothesis, Bower and Hatcher (1967) suggested that, except for alkaline soils with low F

-

adsorption capacity, excess F

-

in water can be removed by adsorption on soils, and that this characteristic should be very practical in artificial ground-water recharge schemes. They also mentioned that, whenever the adsorption process can be described by the Langmuir equation, the dynamics of F

adsorption can be successfully studied by soil column experiments.

Another study involving low [F

-

] in solution was done by Larsen and Widdowson (1971). Coming out at a time when the specific-ion fluoride electrode had just become available and the [F

-

] measurements were, therefore, less subject to uncertainty, the primary goals of

Larsen and Widdowson's work were to investigate the effect of the soil to solution ratio (r ww

) and the time of shaking of the batch solutions on the F

-

uptake by soils, as well as the effect of type and amount of electrolytes on soluble soil fluoride. After reporting that a r sw

of

0.32 was found to be optimal for studies of soil F

-

, the authors indi-

72 cated that the soluble F

-

decreased with increasing concentrations of

KC1 and CaC1

2

in the batch solutions. It was interpreted as the control on F

-

solubility exerted by a sparingly soluble salt, not fluorite because of the particular [F

-

] and [Cole+] used, but most likely fluorapatite, Ca

10

(PO

4

)

6

F

2

. A marked increase in soluble F

-

was measured after increasing the AlC1

3

concentration in solution, which was linked to the formation of aluminum-fluoride complexes. The variations of adsorbed F

-

with depth for two soils were also investigated, and Larsen and Widdowson (1971) found that in a sandy loam, the

F

-

adsorbed decreased dramatically with depth, whereas in a clay soil, it decreased less and showed some tendency to increase below a depth of

60 cm. In the first case, a pH decrease of nearly one unit from top to bottom of the profile was suggested to be the cause for the corresponding decrease in F

-

adsorbed, whereas for the second soil, in spite of having its CaCO

3

content and pH increased with depth, the adsorbed F

increased in the lower part of the profile. Larsen and Widdowson

(1971), in spite of their careful analysis, could not decide on the main parameters governing the F

-

solubility in soils. It was concluded that the F

-

geochemistry appears quite complex and that many factors, such as pH of the system, solubility of F-bearing minerals in soils, exchange reactions between F

-

and OH", etc., have to be considered simultaneously.

In treating the anion adsorption characteristics of goethite and gibbsite, Hingston et al. (1974) came up with a good set of sound conclusions on the desorption behavior of specifically adsorbed ions, such as the F

-

ion. Of particular relevance to fluorine chemistry, the

73 authors found that F

-

adsorbed by goethite from a solution of a pH of

4.5 (0.1 M NaC1 was used as supporting electrolyte), was later desorbed when the goethite was washed with a solution at the same pH but containing no fluoride. That is to say, the adsorption of F

-

on goethite is reversible as opposed to other multicharged ions (e.g., phosphate) which can form multidentate bonds with the mineral surface, making their desorption only partially reversible or non-reversible at all. The authors indicated that dentate ligands from mcnocharged species such as F

-

and 0W are less likely to act as "bridging" ligands on the surface complex, which could provide a reason for their more reversible behavior.

The possibility of hysteresis in the adsorptiondesorption isotherms is also mentioned, an affect probably due to nonattainment of equilibrium during the experiment.

A,

most interesting remark was made by Hingston et al. (1974) in saying that any meaningful investigation of ion transport should pay close attention to the desorption branch of adsorption isotherms, in particular when the adsorption process is partially reversible or nonreversible.

Perrott et al. (1976), in reviewing previous research on the interactions between fluoride with soils and soil minerals, called attention upon the fact that while some authors indicated the disruption of the minerals structure and the formation of new compounds, others suggested that only an adsorption reaction occurs, but all mentioned an increase in pH. Inspired by such a fact, Perrott et al.

(1976) set out to study the CEr release by soil minerals and soils as a

74 function of pH, soil-water ratio, and [F

-

] in solution. The results of their investigation can be summarized as follows:

(1) When the pH was maintained constant at 6.8, and with

0.85 M NaF in solution, the amount of cEr released in

25 minutes was independent of the soil-solution ratio up to r sw

x 10r

3

. For r sw

in the range of 5 x 10

-3 to 1.6 x 10

-2

, the CEr released in 25 minutes was also constant, but smaller than in the previous case.

(2) With a [NaF] of 0.85 M in the batch solutions, the amount of OH

-

released decreased as the pH was increased, in the fashion shown in Figure 2. The authors suggested that the high values around pH 6 can be accounted for by the greater stability of 6bi1F0

3at low pHs and the enhanced stability of (A10H

4

)

-

at high pH values, which basically agrees with the conclusions presented by

Hem (1968) as to the species likely to be formed. The rather low reactivity at high pH values, as explained by Perrott et al. (1976), may be related to the non-reactive nature of siliceous components above pH 7.6.

(3)

As expected, the amount of OH

-

released at a pH of 6.8

increased with the concentration of F

-

in solution.

With regard to this, it is important to mention that the range of [F

-

] used (950 to 19000 ppm) does not reflect the range expected in nature.

Figure 2. Effect of pH on the amount of OFr released frcrn a soil clay (25 mg clay treated with 5 ml of

0.85 M NaF at 25

°C). (After Perrott et al., 1976, p.

60).

75

76

(4) The contribution of crystalline minerals to the amount of OH

-

released is very small compared with that of non-crystalline or poorly-ordered materials. Such a finding suggested to the authors to use the amount of

CEr released as an indication of the content of noncrystalline inorganic materials of soils, provided that their carbonate content is previously eliminated with acids.

As a way of backing up some of their conclusions, Perrott et al. (1976) presented the results of a statistical analysis in which a multiple regressionmcdel was applied to data from freely-drained soils

(with measurable quantities of poorly-ordered inorganic materials) and from poorly-drained soils (with scarce content of such materials). The dependent variable was the amount of al

-

release, and the independent parameters were the amounts of alumina, silica, and ferric oxide. For both soils, the amount of ar released was highly correlated with the amount of alumina (Table 6). The correlation improved when terms related to silica and ferric oxides were included. The term silica was applied to amorphous silica because the authors also reported that a sample of quartz from Brazil released only 1 mmol of CH

-

per 100 gm of sample in 25 minutes (Perrott et al., 1976). The mean values for the amounts of OH- released were 43.1 and 79.6 mmo1/100 gm (standard error of ±5.1) for poorly-drained and freely-drained soils, respective

This is highly significant in terms of supporting the suggestion of

Table

6. Correlation coefficients between OH- release and amounts of silica, alumina, and ferric oxide obtained by selective dissolution. (After Perrott et al.,

1976, p.

64)

77

Variables

Correlation coefficients

Freely-drained Poorly-drained

Soils Soils

OH xAl

2

0

3 xSi0

2

OW

x.Fe

2

0

3

OH

-

xSi0

2

and

Al

2

0

3 xSi0

2

and

Fe

2

O

3

OH

-

XA1.

2

0

3

and

Fe

2

0

3

26i0

2

,

Al

2

0

3

, and

Fe

2

0

3

0.89***

0.58***

0.44***

0.91***

0.72***

0.89***

0.92***

0.90***

0.45***

0.17*

0.90***

0.47***

0.91***

0.91***

*: significant at

5% level;

***: significant at

0.1% level

78

Perrott et al. (1976) to use the Cir release as a chemical index to characterize the inorganic, non-crystalline content of soils.

The results of a ccmprehensive study involving the interactions between fluorides and soil were published in 1977 by Barrow and Shaw.

Samples of five surface soils were treated with solutions containing F

in concentrations varying between 0.05 and 3 ppm. It was found that, for a given [F], an increase in both the temperature and the incubation period of the batch solutions led to an increase in adsorbed F

end to a decrease in the amount of F

-

that could later be desorbed.

For the adsorption process, Freundlich isotherms were fitted. Unlike the F

-

adsorption on goethite (Hingston et al., 1974), Barrow and Shaw

(1977) indicated that adsorption and desorption of F

-

on the tested soils proceeded at a different pace (i.e., adsorption and desorption data for a given experiment did not plot on the same isotherm). The authors concluded that, after an initial adsorption stage, there were slow changes in the nature of the link between the F

-

ions and the surface of the soil particles. Furthermore, in making a parallel between fluoride and phosphate, Barrow and Shaw (1977) postulated that the increase in negative charge as a result of adsorption reduces the capacity to adsorb further anions. This statement, however, seems to be describing a Langmuir-type of behavior rather than the Freundlich model the authors used. It is my opinion that, had the authors employed higher F

-

concentrations in the batch solutions, the adsorption data wculd probably have been better described by the Langmuir equation. Barrow and Shaw (1977) presented a possible mechanism for

79 adsorption, which is shown in Figure 3, and offered the following explanation (Barrow and Shaw, 1977, p. 274):

"The slow reactions between soil and anion could then be regarded as reactions that make this increase in negative charge [as a result of anion adsorption] more difficult to reverse.

A possible mechanism for this is effective migration of the negative charge from the site of adsorption, as is illustrated in Figure 3. The upper reactions in this figure are suggested to be relatively rapid; the balance between the uncharged and the negatively-charged sites is determined by ionic strength of the solution and by pH, as was observed for selenite by Hingston et al. (1972). It is suggested that the lower product is formed more slowly and that further migration of protons from adjacent sites could move the negative charge even farther from the site of adsorption. This could be interpreted as a range of reaction rates. Desorption of fluoride from the uncharged site so produced could be by reaction with an anion, such as hydroxide, as in the reverse of the reaction in the top of the figure, by the site acquiring a positive charge, or by reversion to the negatively-charged state by reverse migration of protons.

The longer the period of contact, the farther the negative charges may have to move to return to the adsorption site and, hence, the desorption. Because desorption is slower--that is, the rate of movement of fluoride ions to the left of the diagram decreases-further adsorption occurs when fluoride is incubated with the soil. This scheme thus seems to be consistent with the observations on fluoride and to be applicable also to phosphate and molybdate."

Omueti and Jones (1977) performed batch experiments in which 14 surface and 6 subsurface horizons from representative Illinois soils were incubated with F

-

solutions (2-20 ppm F

-

). For all cases, the authors fitted Langmuir isotherms to the adsorption data, although it was acknowledged that such fitting was done out of practical considPrations (the Langmuir equation allows the estimation of an adsorption capacity term) because the data also followed the Freundlich isotherm.

OW r

H20

\m /

F a

N ow

\yowl

1 1

N20

Figure 3.

Diagrammatic representatiorx of possible reactions of fluoride with incanpletely coordinated meta1 ions near the surface of an oxide. (After Barrow and Shaw,

1977, p. 227).

80

81

Interesting enough, Omueti and Jones (1977) reported that liming a soil sample to above pH 7 not only caused a reduction In the amount of adsorbed F

-

but also lowered the bonding energy.

In all cases, there was a significant pH increase as more F

was adsorbed, although the correlation between those two parameters was not a simple one because of buffering by the various soil constituents.

The pH, therefore, was assumed to be the chief variable in the adsorption process, with adsorption maxima occurring between pH 5.5 and 6.5

for the tested soils.

When kaolinite was equilibrated against a 2-ppm F

-

solution,

Omueti and Jones (1977) found that the adsorption isotherm was consistent with the dissolution of kaolinite by fluoride as a function of pH with formation of fluoroaluminate and fluorosilicates, a remark also made by Semmens and Meggy (1966a). In view of such facts, the authors decided to carry out a multiple regression analysis on the soil chemistry data, which resulted in aluminum content and soil pH being the variables with the highest degree of association with the measured F

adsorption. As for the Illinois soils, the authors indicated that, at normal pHs, aluminum polymers were responsible for the F

-

adsorption.

Although the retention mechanism was not discussed, Pal iwal and

Sowani (1977) presented a case in which F

-

-containing waters were applied to experimental plots. It was later determined that about 80% of the F

-

applied had been adsorbed in the first 7.5 cm of the surface soil, a trend that was kept irrespective of the [F

-

] and IDS in the irrigation water used and the type of crop grown.

82

In a study aimed at assessing the effect of soil sodicity on the behavior of fluorine, Chhabra et al. (1980) treated 25 surface soil samples of varying ESP (exchangeable sodium percentage) with solutions containing the F

-

ion in concentrations ranging from 0.2 to more than

11 ppm (inferred from Figure 2, p. 35 of Chhabra et al., 1980). The effects of soil-solution ratio, pH, and incubation period were also examined. The relevant conclusions were as follows:

(1) The adsorption process ceased within one hour, which indicated that equilibrium was rapidly attained.

(2) For the pH range of 8.4 to 10.0, the F

-

adsorption decreased as the pH went up, which was interpreted as the replacement of F

-

by

our

groups on the soil particles.

(3) As ESP increased (associated with high pH), the F

adsorption decreased, probably due to the greater anionic repulsion related to the increased negative double-layer repulsion operative under such conditions.

(4) For [FA up to 6 x 10

-4

M (11.4 ppm), the F

-

adsorption was described by the Langmuir model. For higher fluoride contents, the adsorption increased linearly with the [F

-

] in solution (although not following neither Langmuir nor Freundlich equations) up to [F

-

]

133 ppm, beyond which data points curved sharply upwards (enhanced F

-

attachment), which suggested as the result of F

-

irrrnobilization owing to precipitation of fluorapatite.

83

(5) There was an inverse correlation between the adsorption maximum (from Langmuir plots) and the bonding energy.

Ntkrshina (1980) carried out F

-

adsorption studies on Soviet soils in which [F

-

] between 10 and 12000 ppm were used as the adsorbate. The Freundlich equation fitted the adsorption data. Again, there was a direct correlation between the amount of F

-

adsorbed and pH, which was taken as the evidence of F

-

exchanging for air on the solid material. When the pH was kept constant, the F

-

adsorption increased as more acid solutions were employed. Although the range of

[F

-

] used may not be representative of natural conditions, Mbrshina

(1980) suggested the existence of chemisorption as one of the mechanisms for the uptake of F

-

by soils. Indeed, the formation of poorlysoluble calcium and magnesium fluorides made the contents of Ca" and

mg-

in the equilibrium solutions drop to zero; in addition, a parallel decrease of total exchangeable Ca and Mg was noticed.

Mbrshina (1980) was not able to draw sound conclusions on a highly hypothetical relationship between the organic matter content of the tested soils and the amount of F

-

fixation. Morshina (1980) also carried out a column experiment at a flow rate of about 1 ml/min in order to characterize the dynamic adsorption of F

-

by a soil sample, and the result was the obtention of a breakthrough curve typical of ion-exchangers.

Inasmuch as ion exchange has just been mentioned, it may be interesting to add that, according to Zack (1980), most of the F

concentrations found in ground waters pumped from a sandy aquifer in

84

South Carolina could be accounted for by an exchange between air and F

ions

.

In referring to some basic principles and their application in adsorption studies, Schindler (1981, p. 4) pointed out that "A hydroxylated oxide particle can, to a certain degree, be understood as a polymeric oxo acid. Therefore, one could, partially, predict its reactions from the chemical properties of corresponding monomeric acids. For most of the interesting oxides, these monomers (Mn(OH)

4

°,

Al(OH)

3

°, Fe(OH)

3

°, etc.) are unfortunately not well known; however, the monomer of silica has been well investigated." Following, the author summarized some important reactions involving the orthosilicic acid, H

4

S10

4

, one of which is of particular relevance to this review

(Schindler, 1981, p. 5):

Ligand exchange: Si(OH)

4

+ 6F

-

<===> S1F

6

= + 40Er log K= 24.9 ±0.1 (25 °C; I= 0.5 M NaC10

4

)

Farther on, Schindler (1981) indicated that in compounds such as Si0

2

,

Ti0

2

, A100H, and Fe0OH, the surface hydroxyl groups may be replaced by fluoride or by oxyanions. Mention is also made to the effect that numerous studies have confirmed that F

-

and other anions are specifically adsorbed (i.e., ligand-exchanged). Together with many other authors, Schindler (1981) concluded that pH is the master variable that governs the extent of most adsorption processes.

Hingston (1981) treated extensively the adsorption of F

-

on goethite and gibbsite based on the experimental data shown in Figures 4

85 and 5. The author pointed out that, although the general behavior can be described by the Langmuir model, the existence of a maximum is difficult to prove, even for goethite (Figure 4). Hingston (1981) suggested that the all-too common use of the Langmuir equation might not be appropriate, and that the most useful mechanistic interpretations for the characteristics of specific adsorption would be given by a model that (1) does not require consideration of maxima at constant pH, (2) incorporates adsorption parameters pertinent to particular anionic species, and (3) estimates the effect of surface charge. As an example of the need of such "customized" adsorption models, Hingston

(1981) presented a case in which, in order to fit the data of F

adsorption on goethite, the model had to be adjusted to allow the small

F

-

ion to enter the same plane of coordinated CH

-

and H

2

0. In making comments about the same case, the author indicated that only the adsorption of the anion was allowed in the model, and that the field of investigation is open to consider the possibility that both the anion and its conjugate (i.e., weak) acid are adsorbed.

Although the increase in the ion strength of solutions should theoretically increase adsorption at pH values below the point of zero charge (pzc) and decrease adsorption for pHs above the pzc, such an effect has not been experimentally confirmed without ambiguities

(Hingston, 1981).

Hingston (1981) reported that fluoride was reversibly adsorbed by goethite at pH 4.5 and in 0.01 M NaC1 solution, a phenomenon also described in Hingston (1974).

200

Figure

4.

Isotherms for adsorption of F

-

on goethite (23 °C, supporting electrolyte 0.1 M NaC1).

(After Hingston,

1981, p.

55).

86

Figure 5.

Isotherms for adsorption of

F

-

an gibbsite (23 °C, svporting electrolyte 0.1 M NaC1).

(After

Hingston,

1981, p. 56).

87

Noting that F

-

is specifically adsorbed, Hingston (1981) indicated that the reactions to form surface complexes cannot be explained as a simple ion exchange but as a rather complex process in which surface coverage, pH, and the presence of other specifically adsorbable ions are to be considered.

Fluhler et al. (1982), after saying that the term adsorption as used in their paper included adsorption "per se", ion exchange, precipitation, formation of mixed solids, and even complexation, reported the results of two types of tests, batch experiments in which various soils were brought into contact with [F

-

] solutions, and transport experiments performed by percolating soil columns at constant flow rate.

For the batch experiments, Fluhler et al. (1982) stated that the best fit was offered by the Langmuir equation, although they also said that such a model might not be a reliable (i.e., unbiased) estimator, in particular at low [F

-

] (Gupta et al., 1982, reported the same lack of a good fit of the Langmuir equation for [F

-

] < 10

-4

M).

Desorption was found to exhibit hysteresis. Inasmuch as the individual branches of the hysteresis loops turned out to be very steep when [F

-

] in solution was between 10 and 20 piml, Fluhler et al. (1982) suggested that the explanation might be linked to the kinetic effect of a hypothetical CaF

2

formation and dissolution. The results of their column experiments allowed Fluhler et al. (1982) to draw the following conclusions:

(1)

The F

-

ion did not seem to interact much with quartz surfaces because the breakthrough curves for F

-

and Cl were very similar in the quartz-packed columns.

(2)As the [F

-

] in the feed solutions was increased, the breakthrough of F

-

at the top of the column took place earlier, which was interpreted as the non-linear nature of the adsorption isotherm (otherwise a single breakthrough curve would have been obtained for all initial

[F

-

(3)For a fixed [F

-

] in the feed solution, the adsorption process was not affected by the change of the supporting electrolyte in solution.

(4)The F

-

retention in columns packed with soils containing CaCO

3

was less than that in columns containing acid soils.

(5)The F

-

transport models should include a rate-dependent, non-linear sink term.

88

Gupta et al. (1982) studied the fluorine retention in alkaline soils and analyzed the roles of pH and sodicity. After fitting

Langmuir isotherms to the experimental data, the authors found that, when the pH was increased from 5.5 to 10.5, the F

-

adsorption decreased. In comparing the F

-

adsorption capacity of calcareous and non-calcareous soils, it was found that there was no difference at a pH of 10.4, possibly due to the

cEr

coating of calcite particles which inhibited the F

-

adsoip Lion by

caco

3

.

At pH 8.95, however, the calcareous soil adsorbed more F

-

than its counterpart. At pH 8.35, calcite is reported to adsorb little F

-

, and the authors argued that it may be due to the closeness to its isoelectric pH of 8.4. When the

89

Car* content in solution was increased and the SR (sodium adsorption ratio) lowered, the F

-

uptake was slightly larger. It was then concluded that (1) the pH of the soil system markedly influenced the adsorption of F

-

, and that (2) at constant pH, changes in SR or ESP

(exchangeable sodium percentage) did not promote a substantial change on F

-

retention by soils.

In a 1983 paper, Murray described the results of an evaluation of the F

-

retention in highly-leached sand podzols of Australia. The

1-year column study, in which input solutions of varying concentrations of sodium fluoride and cryolite were employed, showed that, once breakthrough of F

-

was achieved, the rate of F

-

leaching reached a plateau and then declined, possibly due to an increase of the firmness of the link between F

-

ions and soil particles. Murray (1983) claimed that such behavior could be compared to the results obtained by Barrow and Shaw (1977).

The overall F

-

retention was high, in spite of the low-clay and low-cation contents of the tested soils. The F

-

binding in the soil column was associated with the presence of calcium, iron, aluminum, clay, and/or humus commonly found in the illuvial B soil horizon and, to a lesser extent, in the mineral A soil horizon, a fact already suggested by Spetch and McIntire (1961) and Omueti and Jones (1977).

In investigating the reactions between dilute F

-

solutions and clays (montmorillcnite, kaolinite, and illite), Slavek et al. (1984) indicated that the amount of F

-

removed from the batch solutions and the fraction complexed increased with decreasing pH and increasing [F

-

] in solution. The observed trend led the authors to conclude that the

90 main process involved is chemical interaction rather than specific adsorption. Inspired by the findings of Semmens and Meggy (1966b), the authors reasoned that, if HF° species promoted clay dissolution, there would be some relationship between the HF° levels in solution and the

F

-

uptake. Such a relationship could not be clearly established, which

Slavek et al. (1984) presented as an evidence of the contributing nature of the acid attack as distinct from the dominating process. The profound effect of pH changes on F

-

uptake was deemed to be consistent with a mechanism based on the precipitation of cryolite (Na

3

A1F

6

), in agreement with the conclusions drawn by Semmens and Meggy (1966a). The authors also concluded that in acid soils, a remarkable F

-

uptake is to be expected; whereas in alkaline soils such adsorption is reduced unless conditions are such that CeLF

2

is formed.

A final remark made by Slavek et al. (1984, p. 219) highlights the vital role of kinetics in these type of studies:

"Expressed in terms of mmol/kg, the amount of fluoride lost from solution [when F

-

solutions were brought into contact with kaolinite, illite, and mcntmorillonite] is of similar order to the observed capacity of the clays to adsorb metal ions. Such ccmpariscns have limited value, however, because two different mechanisms are involved. Allowance has to be made for the slaw rate at which heterogeneous reactions can proceed. Accordingly, the data recorded in this paper primarily provide an indication of short-term effects; if contact between the fluoride solutions and clay suspensions had been maintained for weeks and months, enhanced fluoride losses may have been observed."

Farrah et al. (1985) studied the F

-

adsorption by soil ccmponents such as calcium carbonate, humic acids, manganese dioxide, and silica. The main conclusions can be best summarized as follows:

(1) }Kunio acids adsorbed F

-

according to the Freundlich model, and the effect was linked to the formation of

AlF2

x species and even to cryolite precipitation (the aluminum was present as an inorganic impurity in the organic material).

(2)No detectable amounts of F

-

were adsorbed by manganese oxides (cryptomelane, hydrous manganese oxide, 6-Mln)

2

, pyrollusite), or silica (precipitated Si0

2

, silica gel).

(3) When solutions of [F

-

] > 19 ppm were added to CaCO

3 suspensions, the F

-

uptake was proportional to [F

-

]

2

, which is consistent with equilibrium with fluorite according to the reaction Ca" + 2F

-

<===> CaF

2

. Mbreover, the F

-

retention was independent of the weight of solid present, which the authors interpreted as a clear chemisorpticn process involving surface exchange of F

by CO

3

= with the calcite surface providing Ca" nucleation centers from which the new solid phase could extend in the form of a coating. When acid was added to eliminate the calcite particles, the F

-

uptake was considerably reduced, even though the solubility product for fluorite was exceeded. The authors, however, did not rule out the possibility of fluorite formation. They thought rather that small precipitate particles were formed and later dissolved in the buffer solution used to measure F

-

. Cautiously, Farrah et al.

91

92

(1985) indicated that equilibrium with fluorite may be approached, but at a rate slower than that in the presence of CaCC1

3

particles.

Mare data emerged from experiment done by Peek and Volk (1985), in which ten surface soil samples were mixed with solutions containing the F

-

ion in varying concentrations. It was concluded that the adsorption at low [F

-

] followed the Langmuir model, whereas higher }equilibrium concentrations resulted in deviations from such a model, as also noticed by Cmueti and Jones (1977). The authors suggested that such a feature may be related to the existence of more than one set of adsorption sites or mechanisms. Freundlich isotherms, however, were found to conform the whole data set. Multiple regression analyses, in which the parameters of the Freundlich and Langmuir equations were the dependent variables, concluded that the single most important predictor of F

-

adsorption is the aluminum content in the solid phase.

The F

-

desorption was found reversible for the most part, with cases in which hysteresis was present due to a hypothetical precipitation of CaF

2

. Finally, Peek and Volk (1985) pointed out that the mechanism of F

-

adsorption on soils is more complex than a simple mcnodentate ligand exchange reaction given the irreversibility of the process in soils high in amorphous Al materials, a concept earlier postulated by Hingston (1981).

The last paper reviewed is that by Schoeman and McLeod (1987), in which the variables that affect the fluoride removal by activated alumina were investigated. It was found that the single most important

93 factor affecting the fluoride removal efficiency was the alkalinity of solutions (because HC)

3

-

and 0H groups enter into competition with F

for adsorption sites on the activated alumina), and that the fluoride removal takes place through an ion-exchange process.

Distribution of Fluorine Species in Aquifers and Soils

The numerous study case presented above should give a clear indication to the effect that the investigation of the geochemistry of fluorine in natural environments is, at best, a complex and cumberscme task.

A multitude of sometimes not-well defined processes may overlap, which IPA& to confusion and uncertainties as to how general conclusions on the subject are to be drawn. Therefore, it comes as no surprise to find out that, according to many researchers, regular patterns of variation of F

-

concentrations in both saturated and unsaturated media are not easily established. A few examples will now be presented which illustrate the efforts to shed some light on the general problem.

Aimed at establishing the link between the high incidence of dental diseases and the quality of the water supplies, Smith and

Cammack Smith (1932) carried out a comprehensive survey throughout the

State of Arizona. In spite of the large number of water samples taken, it was not possible to relate the F

-

content to the depths of the wells sampled.

In describing the fluorine disL.Libution with depth in soil profiles, Robinson and Edgington (1946) indicated that an overall increase of [F

-

] with depth was found, although the largest concentra-

94 tions were sometimes found in the C horizon and not necessarily in the

B horizon (where clays usually accumulate). The authors suggested that fluorine in micas of surface soils is leached down along with potassium.

Kister (1966) could not relate the [F

-

] measured to the depths of the surveyed wells, and concluded that the observed F

-

contents were rather dependent upon the type of rock making up the aquifers and the flow regime.

Although of minor significance for their study, Graham et al.

(1975) described a case in which the fluoride content in ground waters of Ontario, Canada, was found to vary seasonally.

Murthy and Mbrthy (1977) presented a case in which the variation of the F

-

content with depth in soils and local rocks did not show any special trend. Moreover, most of the ground waters containing high

[F] were not related to the F

-

content of the sediments which they were flowing through.

Anomalously-high [F

-

] zones in local ground waters were reported by Natarajan and Mbhan Rao (1977) to be unevenly distributed, with a lack of areal and in-depth distribution pattern.

In presenting the results of a survey in which almost 6,000 water samples from wells were taken and analyzed, Rao et al. (1977) pointed out that no definite relationships could be noticed between fluoride levels and pH, [Cl

-

], [NO

3

-

], and TDS. The depth of the wells was not a good predictor of the F

-

content because, for example, wells located only one or two hundred meters apart had widely-varying amounts of the F

-

ion.

95

Rao (1977) surveyed an area affected by fluorine-related diseases, and concluded that the pattern of distribution of fluoride concentrations in local ground waters was irregular, both areally and seasonally.

Eccles and Klein (1978) were able to define a definite increase of [F

-

] with depth in a southern California aquifer, a pattern very likely associated with an upward ground-water flow through faults.

Gomez Artola et al. (1983) reported the lack of a clear correlation between the F

-

concentrations in ground waters and the various water-bearing units in and around the city of Madrid, Spain.

In the report that laid out the basis for this study (Usunoff,

1984) and in an accanpanying paper (Usunoff and Nelda, 1988), it was not possible to identify areal or vertical distribution patterns of

[F

-

] in the aquifer system in southern Arizona which was studied. It may be relevant to add that, in referring to this and similar alluvial basins in Arizona, Robertson (1985) indicated that pH-dependent sorption-adsorption reactions and/or ion-exchange processes appear to be important controls of the usually high [F

-

] found in local ground waters.

Significance of Fluorine in Human Health

An unquestionable fact, long ago found and conspicuously reported in the pertinent literature, is that both high and low concentrations of fluoride in water supplies and diets do have a profound effect on the development of human bones and teeth.

96

Fluoride has been found to act as an inhibitor of the bacterial enzymes which, in turn, are believed to produce the acids that initiate the enamel caries (Jenkins, 1970). Although most diets contain measurable amounts of the element fluorine, the total intake is mainly determined by the concentration of fluoride in the local water supply and the total amount of water ingested (Srikantia, 1977). Hodge and

Smith (1965) showed that the absence of fluoride in water supplies or its presence in low concentrations (below 0.5 ppm F

-

) increases remarkably the number of dental caries in children. The empirical relationship the authors came up with is presented in Figure 6. It is precisely the low fluoride content (i.e., below 1 ppm F

-

) in public water which supplies the explanation that Gomez Artola et al. (1983) gave to the epidemic of dental caries in Madrid, Spain.

On the other hand, the ingestion of excess fluoride is responsible for the disease known as fluorosis (mottled enamel). Dental fluorosis, a condition of endemic nature, is a dental aberration which leads to a varying degree of brownish-yellow discoloration of teeth.

Although the prolonged consumption of marine food and/or tea--a strong fluorine accumulator, according to Robinson and Edgington (1946)— contributes to the appearance of fluorosis (Rao and Murty, 1977), its degree of severity is chiefly related to the fluoride content in drinking water, in the approximate fashion depicted in Figure 7. Hodge and Smith (1965), certainly a most comprehensive work on the subject, have pointed out that the sharp fluoride-related defects which appear above 1 ppm of F

-

in water (Figure 7) are a typical toxicological

• s

o

1:0 1.5 2.0 2.5 3.0

FLUORIDE CONTENT OF THE

PUBLIC WATER SUPPLY (PPM)

Figure 6.

Relation between dental caries experience and

[F

-

] in drinking water.

(After Hodge and Smith, 1965, p.

464).

97

SEVERE

MODERATE o

MILD

VERY MILD

ZERO

.1

1

FLUORIDE CONCENTRATION IN WATER (PPM)

10

Figure 7. Relation between index of fluorosis and [F] in drinking water. (After Hodge and Smith,

1965, p.

449).

98

99 response inasmudh as responses to many drugs increase with the logarithm of the dose.

The effects on the teeth caused by fluoride ingestion is of accumulative nature. Thus, for a given fluoride concentration above

0.8-1.0 ppm, the effects depend upon the amount of fluoride-carrying water consumed. Aside from very special circumstances, the consumption of water is agreed to be a direct function of the air temperature, which has inspired regulatory agencies to set safe [F

-

] limits based on mean annual temperatures for a given region. The U. S. Department of

Health, Education, and Welfare (1962), for example, has followed such a criterion, an illustration of which is presented in Figure 8.

For a given water consumption, however, disagreement still exists in determining the fluoride-content threshold above which the effects on the teeth are noticeable. Hodge and Smith (1965) have tentatively suggested a concentration of 0.5 ppm F

-

, although many authors have pointed out that fluoride concentrations even lower than that can have toxic effects on people who were born and grew up in warm regions. Minoguchi (1977) advocated such a concept based an a study that revealed that the human perspiration in areas of high mean annual temperature is less saline, which would increase the amount of fluorine that is accumulated and retained in the body.

Waters containing high fluoride concentrations (above 4-5 ppm

F

-

), if consumed for a long time, can produce severe structural deformations of human bones such as in the case of the diseases known as kyphosis (forward bending of spine), syndrome of genu-valgum

1.2

o o

702

OBJECTIONABLE

FLUOROSIS

00'

BORDERLINE

o

• 5

0Q

o

NEGATIVE

0

"V/

0

I

4

I Y

.6 .8

1.0

1.2 14

1.6

FLUORIDE

CONCENTRATION (PPM)

ll

1.8

I

2.0

.021* F .353 ( for 50

2

)

COMMUNITY

FLUOROSIS

INDEX:

-.291 + F 1.132 ( f or

Tz. 70

2

)

F: fluoride concentration

(ppm)

Figure 8. Relation between

[r] in municipal waters and fluorosis index for

carmunities

with mean annual temperatures of

approxixnately

50 °F

(Midwest) and 70

°F (Arizona).

(After Hodge and Smith, 1965, p. 457).

1

00

101

(development of knocked-knee), osteosclerosis, etc. (Krishnamachari,

1977).

Summary of Previous Research

Compiled under the various headings found in this chapter, an attempt has been made to provide the reader with some insight into the research done on the geochemistry of fluorine. It may be concluded that a variety of interpretations exists as to the behavior of fluorine species in natural environments. The main conclusions and/or uncertainties an the topics reviewed can be stated as shown below.

The dissolution of F-carrying minerals appears to be the ultimate natural source of the fluorine species encountered in natural settings. As artificial sources of the element, the use of fertilizers and the careless disposal of products used in various manufacturing processes can be mentioned.

Based on principles of chemical equilibrium, the common association of waters high in [F

-

] with high alkalinity and low hardness levels is explained by dissolution of fluorite in the presence of calcite at a relatively constant pH. The application of the partial equilibrium approach seems to yield better results in modeling such a dissolution process. Several elements can complex fluorine, chiefly Al and Fe. In unsaturated media, the study of the fluorine geochemistry is highly complicated because of the many factors having a direct or indirect incidence on the mobility of F species through the soil profile.

102

The review of the kinetic aspects related to the dissolution of fluorite has turned up two theories to explain the mechanism in control of the dissolution rate: transport stages and surface reactions. Other variables important in establishing a given dissolution rate are the temperature and the bulk solution composition and concentration.

The adsorption of fluorine species on various mineral surfaces has many times been described by the Langmuir model, although a group of researchers has used the Freundlich isotherm with equal success, and others could apply neither model. Adsorption of F

-

does occur, but the mechanism is not as simple as physical adsorption; phenomena such as ion exchange F

-

-(Er, formation of new phases, complexation, etc., have to be taken into account and properly addressed. Acid environments are found to adsorb more F than alkaline ones, with variables such as pH, presence of hydrous minerals and oxides, and solution composition having a decisive effect on the adsorption process. The adsorption of

F

-

has been found to be mostly reversible, with desorption curves exhibiting hysteresis.

No clear areal, vertical, and seasonal patterns in the distribution of F species have been found in soils and aquifers.

As for the human health implications, the consumption of waters containing about 1 ppm F

-

appears to have a beneficial effect on dental hygiene. Below such F

-

concentration, the appearance of dental caries is highly enhanced, whereas waters with higher fluoride contents can produce fluorosis or have even more dramatic bone-deforming consequences.

103

CHAPTER 3

ADSORPTION OF [F

-

] BY SCME OCtvMDNLY-FOUND MINERALS

Introduction

The main objective of this dissertation is to evaluate the mechanisms that affect the movement of fluorine species in aquifers.

Therefore, minerals to be tested for their sorptive characteristics of

F

-

were (1) minerals commonly present in aquifer sediments, and (2) minerals known or suspected to interact with fluorine. The three minerals selected were quartz, kaolinite, and vermiculite. Information on the quartz sand and kaolinite used, as well as the details on their preparation for the various experiments, are given in Appendix A. As for the vermiculite, a commercially-available type (i.e., exfoliated by heating) was used "as is". The selection of quartz and kaolinite, aside from mineralogical implications, was intended to represent the sand and clay fractions regularly present in sediments. Vermiculite is not a conspicuous constituent in sediments, and its inclusion was to test the effect of mica-like minerals.

The characterization of the adsorption phenomenon was done through a series of batch experiments involving the minerals mentioned above and F

-

solutions at various pHs. Surprisingly enough, a group of preliminary tests showed that the quartz sand was able to adsorb small yet measurable quantities of F

-

. This was thought to be worth pursuing, and a set of percolation tests was designed using quartz sandpacked columns. The batch experiments tested the effect of changes in pH and [F

-

] in solution on the F- adsorption by the selected minerals.

104

The column experiments were aimed to characterize the dynamics of the adsorption process, that is, the influence of the adsorption phenmenon in the transport of fluorine species, and the mean pore velocity (v) was added to the list of independent variables.

A summary of the type of experiment, the materials used, and the purpose is presented below.

Type of

Experiment Purpose

Column

Solid Phase Solution Variables

Present

Used Tested

* To observe Quartz sand Solutions adsorption- of varying desorption behavior

* To determine

[FpH

[F-

Batch * To measure Quartz sand Solutions pH

K d

Kaolinite

* To fit of [F- ]

Vermiculite varying adsorption model

[F- ]

Before proceeding with the description of the experiments, it may be convenient to recall briefly the definition of two parameters closely related to adsorption processes and movement of solutes, namely, the distribution coefficient and the retardation factor. The distribution coefficient, K id

[L

3

/M],

normally reported as milliliters per gram, expresses the ratio between the mass of solute on the solid phase per unit weight of the solid phase and the concentration of the

105 solute in solution. The retardation factor, R, a dhnensionless quantity, is functionally defined as R = 1 + PA/n, where

P b

is the bulk

mass density [M/L

3

], E' d

is the distribution coefficient [L

3

1M], and n is the porosity [L

3

/L

3

] (Freeze and Cherry, 1979). If R = 1, the solute travels with the same velocity as that of the water (or any other solvent), whereas if R> 1, the solute interacts with the medium, and its movement is said to be retarded with respect to the bulk flow.

Results and Discussion of Column Experiments

A plexi-glass column, whose construction and testing details are given in Appendix C, was packed with the quartz sand described in

Appendix A. The flow, which took place from bottom to top, was driven by a peristaltic pump which was connected through a two-way valve to two containers carrying the F

-

solution and distilled water. Flow measurements and samples were taken at the top of the column, and [F

-

] and pH were measured as described in Appendix B.

Seven runs were performed, numbered 1 through 7, in which the

[F

-

] and the pH of the input solutions and the flow velocity were varied. Typically, a run would proceed as follows: distilled water at a pH approximately equal to that of the input F

-

solution was circulated until the effluent pH stabilized, time at which the circulation of the traced solution started. Samples of the effluent were then taken until their [F

-

] was about the same as the input [F

-

], then the double-position valve was switched to allow the entrance of the pHadjusted distilled water. Sampling then continued until the effluent

[F

-

] was negligible. The adsorption-desorption data so obtained were

106 the input of a computer program -CFIT1M- (van Genuchten, 1981) used to calculate transport parameters from observed column effluent curves by a non-linear least-squares curve fitting procedure. Based on earlier evidence, a linear equilibrium adsorption option was selected, which resulted in outputs containing, among other parameters, the fitted value for R, the retardation factor. The computer program assumes that analytical solutions of the governing equations are available. If analytical solutions are to be used, the governing equations must be properly linearized. That is the reason why CFIT1M can incorporate only linear adsorption. Details on the operational Characteristics of

CFI= and a sample of a typical output are presented in Appendix D.

The data collected at each cf the seven runs are also presented in

Appendix D, and what follows is a discussion of the results.

The breakthrough curves obtained from the various runs are depicted in Figures 9 through 15, and a summary of the relevant input and output parameters for each run is presented in Table 7. The fact that the right-hand limb of the calculated breakthrough curves did not match properly, the observed data (Figures 9 through 15) should not be taken as an indication of a poor matching performance by CFITIM.

Inasmuch as equilibrium kinetics was assumed, the computation led to sigmoidal curves. The asymmetric or non-sigmoidal shape of the observed data indicates, at the very least, that non-equilibrium uptake/release of F

-

could have occurred. MOasurement errors are disregarded based on the rather systematic appearance of an early breakthrough of F.

I

5

4

TIME (Pore volume)

Figure 9. Breakthrough curve for column experiment

1

(solid line: fitted by CFIT1M; dots: data).

107

1.0 •

0.8

U

0

0.G -

(

-

3

OA

0.2 -

• •

5

TIME (pore volume)

Figure

10.

Breakthrough curve for column experiment

2

(solid line: fitted by

CFIT1M; dots: data).

8

108

J

3

TIME (pore volume)

6

8

Figure 11. Breakthrough curve for column experiment

3

(solid line: fitted by CFIT1M; dots: data).

109

5

TIME (Pore volume)

7

Figure 12.

Breakthrough curve for column experiment

4

(solid line: fitted by

CFITIM; dots: data).

8

110

1.0 •

0.8

0 0.G -

0.4

0.2

1

5

4

TIME

(pore volume)

Figure 13.

Breakthrough curve for column experiment

5

(solid line: fitted by

CFITIM; dots: data).

7

111

1.0 •

0.8 o 0.G -

0.4 •

0.2

3 4 5

TIME (pore volume)

0

Figure

14.

Breakthrough curve for column experiment

6

(solid line: fitted by

CFITIM; dots: data).

7

112

1.0 -

0.8 •

0 0.G

0.4 •

0.2 •

1 2

A

TIME (pore volume)

Figure

15. B re akthrough curve for column experiment 7

(solid line: fitted by CFITIM; dots: data).

7

113

Table 7. Column experiments: measured and calculated data.

pH of pH of [F-] in an

Run Input

#

Effluent Input Flaw Retardation

Solution Solution Solution Velocity Factor

(pH units) (pH units) (ppm F

-

) (cm/min)

1

3

2

4

5

6

7

6.75

4.88

3.11

6.09

6.13

6.08

6.31

7.05

5.00

3.20

6.44

6.22

6.12

6.54

6.00

6.00

5.45

6.00

4.00

2.00

6.00

0.5589

0.5403

0.5133

0.4179

0.4086

0.4020

0.2894

1.43

1.49

1.50

1.30

1.35

1.41

1.22

114

115

Runs 1, 2, and 3 were designed to assess the relationship between F adsorption and pH. Taking the retardation factor R as the diagnostic variable, R is seen to increase as the pH decreases. That is to say, low pH values enhanced the F adsorption on sand particles.

Such a relationship may be related to at least two factors:

(1)The zero point charge of quartz, for which Stumm and

Mbrgan (1981) have given a value of 2. Indeed, as the pH increases above 2, the surface of quartz particles becomes negatively charged, with a corresponding decrease in anion sorption. Hingston (1981) has pointed out that, theoretically, the effect can be magnified by increasing the ion strength of solutions.

(2)The fluorine complexation as the pH is lowered. At pHs below neutrality and in the absence of cations such as

A1

3

* and Fe

3

+, the F species able to be present are

HF, HF

2 -

, and F

-

(Hem, 1968; Roberson and Barnes,

1978). The concentrations of HT° and HF

2 -

(@ I = 0 and

25°C) can be calculated as follows (Broene and De

Vries, 1947):

[HP] = [H+] [F

-

] 10

3

-

17

, and [HF

2 -

] = [H+] [F

-

]

2

10

3

76

One can see immediately that, unless the pH is extremely low and the total concentration of F is very high, the amount of HF

2 in solution can be neglected. Of importance, however, are HT° and F, whcse representative percentages for the various runs were as follows:

116

Run 1: almost all added F is in the F

-

form

Run 2: [HY°] 1

-

93% of total F, and [F

-

] 7% of total

F

Run 3: [HTe] 1.5% of total F, and [F

-

] 98.5% of total F

As for the possible species that fluorine can form with silicon, Roberson and Barnes (1978) concluded that SiF

6 2-

is the most important one. However, the solubility of quartz at 25°C is so low [6

14E1 S10

2

(Mbrey et al., 1962; Marion et al., 1976; Roberson and Barnes,

1978)] that, even for the column run at the lowest pH (run 2), the concentration of SiF

6 2-

barely exceeds 10

-8

M, which indicates that its contribution can be disregarded. Hcwever, the presence of HT° will enhance the dissolution of quartz, as described by Iler (1979), with a resulting increase in the concentration of SiF

6 2-

, although this possibility was not investigated by laboratory experiments.

Based on (1) or (2) above, or on both, the conclusion can be drawn that F adsorption depends on the type of F species present, which, in turn, are linked to the prevailing pH in the system, with low pH values (below 4.5 or 5) promoting the formation and adsorption of

HT°. It may be pertinent to recall that Hingston (1981) suggested the consideration of the role of both anion (F

-

) and conjugated acid (HF) in adsorption studies.

Runs 4, 5, and 6 were to test the degree of F adsorption as the

[F

-

] in the feed solutions was varied. The observed trend indicates that adsorption occurred, with retardation factors becoming smaller as the input F

-

concentrations were increased. The results are certainly puzzling because, in general terms, the adsorption of inorganics onto

117 mineral surfaces has been found to increase with the concentration of the adsorbate in solution. Inasmuch as [Br] should have been minimal at the pH of the three input solutions, the conclusion is that the

F

ion did not interact much with the mineral surface, or that the adsorption process was overcome by the highly dynamic nature of the experiment. It may be interesting to add that Fichier et al.

(1982) reported that an earlier breakthrough of

F

-

at the top of columns as the

[F

-

] in the feed solutions was increased could be interpreted as the non-linear nature of the adsorption isotherm.

Runs

1, 4, and

7 were intended to reveal the influence of the mean pore velocity in the adsorption phenomenon. Care was taken to select velocities so that the flow would be laminar. For example, if a fluid density of

1 gm/ce and a fluid viscosity of

0.01 poises are assumed, the Reynolds number for run

1

(the one at the highest pore velocity) is about

0.02, that is, well within the laminar flow region.

The results indicate that larger retardation factors are associated with larger velocities. Again, this observation is contrary to the expected trend. It can be thought, at least theoretically, that an enlarged time of contact between the adsorbate and the adsorbent (i.e., lower velocities) would result in better chances for the species in solution to attach to the mineral surface. Such an explanation, however, ignores the status of the surface charge of the adsorbent when adsorption is taking place. As stated earlier, the quartz particles are negatively charged above a pH of

2, which would certainly inhibit anion adsorption. If so, and in the light of the results already

118 discussed, such repulsion seems to be more effective at lower pore velocities.

Two common characteristics were found in the column experiments, namely, an increase in the pH of the effluent solutions as compared to the pH of the corresponding input solutions, and the existence of hysteresis in the otherwise reversible adsorption process.

Allowance was not made to account for the buffering effect of quartz, although it is assumed to be negligible at the pHs used in the experiments. Therefore, the observed systematic increase in pH can be ascribed to the release of cc"

-

from the exchange sites on the sand particles, a fact commonly found to happen on minerals other than quartz in contact with F-carrying solutions. In this regard, Perrott et al. (1976) found that a sample of quartz from Brazil released 1 mmol

0Hi/100 gm of sample after 25 minutes of being soaked by a 0.85 M NaF solution. The recession limb of the breakthrough curves indicates that, for all cases, desorption proceeded at a faster pace than adsorption. Such an effect may represent the lack of equilibrium during washing (Hingston et al., 1974) and, at the very least, gives a clear indication of the weak bond between the adsorbed species and the sand particles.

Inasmuch as the distribution coefficient

K d

is embodied in the definition of the retardation factor R, the fitted values for the latter can be used to calculate the magnitude of the former. If the assumption is made that the mass density of the siliceous sand is 2.65, then the

mass

density P b

that corresponds to a porosity

n

of 39% is around 1.62 gm/ce, and the disLiibution coefficient can be calculated

119 from the expression for R presented earlier. Having done so, the following figures were obtained for the various runs:

Run # K

1

(ml/gm sand)

5

6

7

3

4

1

2

1.04 x10-1

1.18 x10-1

1.21 x10-1

7.24 x10-2

8.44 x10-2

9.89 x10-2

5.31 x10-2

The underlying assumption in calculating K u

is that the values for both n and R are correct and, hence, not subject to uncertainties.

However, that is not strictly so. If, instead of a porosity of 39%, a range of porosities between 35 and 45% is considered (i.e., a change of

29%) with other variables remaining constant, a 52% change in the values for

K u

is calculated. On the other hand, when n is kept constant and R is varied between 1.1 and 1.6 (i.e., a 45% change), the corresponding values for

K ci

change by 400%. Such a simple exercise demonstrates quite explicitly the important effect that uncertainties in the measured values for n and the fitted values for R can have on the estimation of the distribution coefficient. Extreme care, therefore, should be taken by those applying values for E' d

reported in the literature, particularly when those values are to be used in environmental-impact studies.

120

Results and Discussion of Batch Experiments

Adsorbate: Quartz Sand

Using the results obtained with the column runs, a set of batch tests at varying pHs was designed in order to obtain values for the disUibution coefficient which would, later on, be compared with those figures obtained from the column experiments. A second purpose of the batch study was to assess the degree of linearity of the adsorption phenanenon, a characteristic not obtainable from the (column percolation probes.

Based on experience from previous experiments, a soil-to-water ratio (r s

.) of 0.6 was used in four batch experiments (numbered 3, 17,

18, and 19). Plastic centrifuge tubes of 50-ml capacity were used as the reaction containers, in which 15 gm of sand and 25 ml of F-ccntaining solutions were poured. The sand had been prepared as described in

Appendix A. The pH of the solutions was measured before pouring the sand, and its value labeled as initial pH. All batch solutions contained F

-

in concentrations comparable to those found in natural waters

(about 0.5 to 10.0 ppm), except for batch 3 in which larger concentrations were used, up to about 45.0 ppm.

The batch solutions were shaken for 48 hours in a continuousrotation shaker, after which the pH of the supernatant was measured following centrifugation and separation of the solid phase. After adding TISAB (a pH-buffer and ion-strength-adjuster solution), the F

content of the supernatant was measured, and the reading labeled as

[F

-

] in solution. The amount of F

-

adsorbed was obtained by subtracting the [F

-

] in solution after 48 hours from the initial [F

-

]. It is

121 important to mention that the selected time span of 48 hours was deemed long enough to attain equilibrium, inasmuch as previous experiments had shown that no noticeable changes in the [F

-

] in solution were measured after 14-16 hours. The measurements of pH and [F

-

] proceeded as detailed in Appendix B.

The pH and [F

-

] readings were then normalized by considering the soil-water ratio, and the results were plotted in the conventional fashion cl initial [F

-

] in solution (M) versus [F

-

] adsorbed (mole/gm sand). Such plots turned out to be approximately linear, and a

Freundlich-type equation was used to fit the data. In doing so, estimates for the distribution coefficient and the degree of linearity were obtained.

The results of the four batch experiments carried out are shown in Figures 16-19. The solid lines represent the power function obtained from a regression analysis of the collected data. Such a regression, which considered the well-known model S = K d

CP (Freeze and

Cherry, 1979), yielded the following parameters:

Batch #

3

17

18

19

Kd (mi/gm sand)

5.70 x10-1

2.50 x10-1

1.60 x10-1

7.78 x10-2

1.1365

1.0622

1.0355

0.9868

122

35

30

ci7

25

E

20

es

o

15

.o

o

.13

ex)

L.

1 0

5

4

12

1G

F

-

in solution (M x10

-4

)

24

Figure

16. F

-

adsorption isotherm from batch experiment

3

(solid line: power function from regression analysis of data -two last points on the right not considered-; dots: data).

1

2 3 4

F" in solution (M x10

-4

)

5

G

Figure

17. F

-

adsorption isotherm from batch experiment 17

(solid line: power function from regression analysis of data; dots: data).

123

1 2

,

3

.

,

4

F" in solution (M x10

-4

) i G

Figure 18. F" adsccrpticn isotherm £i..ut

batch experiment

18

(solid line: power functicn fran regressicn analysis of data; dots: data).

124

5

1 3 4

F

-

in solution

CM x10

-4

)

G

Figure 19. F

-

adsorption isotherm from batch experiment

19

(solid line: power function aum regression analysis of data; dots: data).

125

126

To supplement the figures above, information on the initial and final values for those variables measured at each experiment are presented (Table 8).

By considering the two pieces of information presented, the first conclusion that can be drawn is that the F

-

adsorption, as judged by the magnitude of the K u

parameter, increased as the pH decreased.

In addition, the departure from linearity of the adsorption process, of which the fitted parameter b is the diagnostic feature (a perfectly linear behavior would yield b = 1), is more pronounced at decreasing pHs.

The measured data show a very good alignment along the fitted adsorption functions for the range of [F

-

] used, except for batch 3 in which the concentration range was expanded. In that case, the points corresponding to rather high [F

-

] in solution fall below the adsorption isotherm, as if a threshold were reached above which reduced adsorption takes place. If so, the Langmuir model can be used to fit the measured data when a large range of [F

-

] in solutions is to be used. This possibility was not further explored because it would have required the use of large F

-

concentrations of little significance in natural systems.

A common characteristic was observed in all batch experiments, namely, the pH of the solutions after adsorption had taken place was always higher than the corresponding initial pHs. For a given experiment, the change in pH was larger as the [F

-

] in solution increased, whereas for a given [F

-

] in solution, the pH change was larger in those batches at higher initial pHs. Viewed as an OH-releasing mechanism,

Table 8. Batch experiments: summary of measured data for quartz-sand.

A

H

3

Initial [F] in solution

(M x 10

-4

)

Initial pH of soin.

(at 24.0PC)

[F

-

] in soin.

after 48 hrs.

(Mx

10

-4

)

Soin. pH after 48 hrs.

(at 23.5°C)

A

Initial [F

-

] in solution

(M x 10

-4

)

Initial pH of soin.

1

7

H

(at 24.0°C)

1F

-

] in soin.

after 48 hrs.

x 10r

Sian.

pH

4

) after 48 hrs.

(at 22.5°C)

1

8

A

H

Initial [F

-

] in solution

(M x 1Cr

4

)

Initial pH of soin.

(at 23.7°C)

[F

-

] in soin.

after 48 hrs.

(M x 1Cr

4

)

Soin.

pH

after 48 hrs.

(at 23.5°C)

0.521 1.032 2.026 2.979 3.900 9.737 23.842

2.950 2.939 2.951 2.943 2.941 2.937

2.953

0.479 0.942 1.826 2.684 3.500 8.789 22.000

2.950 2.939 2.951 2.944 2.942 2.939

2.955

0.521 1.032 2.026 2.979 3.900 5.632

4.225 4.225 4.223 4.226 4.224 4.225

0.484 0.947 1.858 2.732 3.574 5.163

4.226 4.230 4.237 4.265 4.268 4.273

0.521 1.032 2.026 2.979 3.900 5.632

4.412 4.410 4.411 4.408 4.409 4.413

0.489 0.963 1.889 2.779 3.637 5.253

4.418 4.420 4.439 4.470 4.479 4.492

127

Table 8 - -COntinued

1

9

A

H

Initial [F

-

] in solution

(M x 10

-4

Initial pH

)

of

soin.

(at 22.5°C)

CF

-

] in soin.

after 48 hrs.

(Mx 10

-4

Soin.

pH

) after 48 hrs.

(at 24.0°C)

0.521

1.032

2.026

2.979

3.900

5.632

5.020

5.020

5.022

5.023

5.019

4.953

0.495

0.979

1.926

2.832

3.711

5.353

5.082

5.094

5.190

5.433

5.502

5.786

128

129 the conclusion is that mechanism operates more effectively as both the

[F

-

] and the pH of the soaking solutions are raised. This observation brings about, once again, the hypothesis of anion exchange between OH

and F

-

species. To test the relevance of such a hypothesis, the F

adsorption versus the Cair release was plotted as depicted in Figure 20.

If anion exchange were to be the only mechanism causing fixation of F

-

and release of OH

-

, the data points in Figure 20 would all lie close to or on the

1:1 stoichiometric relationship line. This relationship was clearly not found. All points lie below such a line, which indicates that a simple anion-exchange model cannot account for all the F disappearing from solution. Anion exchange may be part of the process, but certainly not the only mechanism nor the most important one. If more than one mechanism determines the adsorption of F on the quartz sand, the role of anion exchange should be analyzed considering the variables that define the original status of the adsorbing system, in particular its pH. In this respect, it is important to point out that the information in Figure 20 reveals that the anionexchange hypothesis seems to be more justified as the pH of the initial solution is increased. Although not shown in Figure 20, the results from batch 3 would yield a line of meager slope, very flat and close to the x-axis. That is to say, the measured uptake of

F

-

was not related to increased pH.

Inasmuch as diluted nitric acid was used to adjust the pH of the batch solutions, NO

3

- groups may possibly replace OH

-

on the sand particles surfaces, thereby inhibiting the F

-

adsorption and providing another source of ar going into solution. The limited scope of the

160

140 x

120

a a

X

20

• a

Batch

17

a

x

Batch

18

X

A

Batch

1g

X

A

160 260 360 460 560 360 760

Fœadsorbed

(mole/gm sand x10

-10

)

800

Figure

20.

Relationship between

F

-

adsorbed from and

cur

released to solution.

130

131 batch experiments planned did not allow pursuing that hypothesis, although it should be considered in follow-up studies.

The batch experiments were designed to establish a relationship among the values of the distribution coefficient calculated from both the batch and the column tests. With this in mind, it can be seen that the batch experiments yielded distribution coefficients sensibly higher than those obtained from the column runs, which is to be expected due to the very nature of batch techniques (prolonged residence time and no renovation of the soaking solution) as opposed to the dynamic characteristics of solute transport in columns. Such differences (i.e., in

K ia

's measured in batch experiments and those calculated from the estimated retardation factors in the column test), however, were not extreme, and a range for

K d

of 10

-1

to 10r

2

ml/gm sand can be assumed to characterize the observed adsorption. In other words, the magnitude of the measured distribution coefficients seems to confirm the fitted values for the retardation factor, which were in the range of 1.2 to

1.5.

Adsorbate: Vermiculite

The use of vermiculite, a group of micaceous minerals closely related to chlorite and with the general formula (Mg,Fe,A1)

3

(A1,Si)-

4

(OH)

2

.4H

2

0, was intended to acquire some insight on the adsorption of fluoride onto hydroxylated surfaces. Samples of commercial vermiculite

(i.e., exfoliated by heating) were brought into contact with solutions of known pH and varying [F

-

], at a soil-water ratio of 0.04 (1 gm vermiculite per 25 ma of solution). Readings of both [F

-

] in solution

132 and pH were taken after an equilibration period of 48 hours. The process was repeated at two differential initial pHs, which resulted in two sets of data named batch 20 and 21. Those data are presented in

Table 9.

The equilibration period selected led to a low adsorption of F

of similar extent at the two tested pHs. The analysis of the data indicated a highly non-linear nature of the adsorption process, which did not allow an interpretation based on the Freundlich equation. The fitting of the Langmuir model to the data, on the other hand, was somewhat better (Figure 21), particularly when the far-right data point for batch 20 was dropped. With so few data, however, no firm conclusions could be drawn.

The data presented in Table 9 reveal strikingly high equilibrium pHs as compared to the respective initial pHs. Moreover, the final pH values were all in the approximate range of 7.4 to 7.9, independently of the initial pH and the [F

-

] in solution. Such a phenomenon, undoubtedly linked to the interactions between protons in solution and

car

groups on the vermiculite surface, can have a profound effect on the movement of F

-

in aqueous environments. As presented earlier in this chapter, the degree of interaction of dissolved fluorine is partly dependent upon the type of species present which, in turn, is mainly a function of pH. If vermiculite (or any other mineral for that matter) has the ability to induce pH changes, the whole set of

F-related chemical equilibrium relationships are bound to be altered.

The pH increase, in this case, would eliminate most of the F-complex ions (i.e., F

-

should be the prevalent species in solution), allowing

Table 9. Batch experiments: summary of measured data for vermiculite.

Initial [F

-

] in solution

(M xl0r

4

)

A

H

Initial pH of soin.

(at 24.0°C)

[F

-

] In soin.

after 48 hrs.

(M x10

4

) 2

0 Soin. pH after 48 hrs.

,

(at 22.0°C)

0.521 1.032 2.026 2.979 3.900 5.632

4.225 4.224 4.225 4.223 4.225 4.224

0.516 1.026 1.947 2.832 3.595 5.189

7.382 7.385 7.409 7.456 7.620 7.553

A

H

Initial [F

-

] in solution

(M xlCr

4

)

Initial pH of soin.

(at 23.8°C)

[F

-

] in soin.

after 48 hrs.

(M x10

-4

) 2

1 Soin.

pH

after 48 hrs.

,(at

22.0°C)

0.521 1.032 2.026 2.979 3.900 5.632

6.482 6.483 6.482 6.481 6.482 6.483

0.521 1.026 1.974 2.842 3.653 5.211

7.591 7.653 7.954 7.813 7.859 7.946

133

6

5

o

2

1

0

• 5

0 Batch 20

Batch

21

1.60

2.00

0.40

0.80 1.20

1/C (1/M x 10

4

)

Figure 21. F

-

adsorption fran batch experiments

20 and 21

[C= equilibrium [F

-

]

(

4); S= F

-

adsorbed (mole per gram of vermiculite)].

134

135 the F

-

ions to participate in other equilibria (with Ca", for instance). Inasmuch as increased pHs would also affect the carbonate chemistry, other chemical changes probably take place, with the establishment of new conditions once equilibrium is reached in the local geochemical environment. Again, as before, pH stands out as the master variable in the process.

Adsorbate:

Kaolinite

Samples of kaolinite, whose preparation characteristics are given in Appendix A, were poured into plastic vials containing solutions of known [F

-

] and pH. A soil-water ratio of 2 x 10r

2

was used

(0.4 gm of clay per 20 ml of solution), which is about halfway between the r sw

of 4 x 10

-2

used by Posner and Quirk (1966), and that of 8 x

10

-4

in Slavek et al. (1984) in similar experiments. As in the case above, two batch experiments at different pHs were carried out, labeled batch

22 and 23. The initial readings and the measurements taken after

72 heurs are presented in Table 10.

The consideration of the Langmuir model did not produce acceptable results. The Freundlich equation, on the other hand, offered a better fit of the data (Figure 22).

A

regression analysis of the data for the model S = K a

CP designed to obtain numerical values for the model's parameters yielded the following (fitted function not shown in Figure 22):

Table 10. Batch experiments: summary of measured data for kaolinite.

A

Initial [F

-

] in solution

(M x10

-4

)

Initial pH of soin.

H

2

2

(at 24.2°C)

[F

-

] in soin.

after 48 hrs.

(M x10

-4

)

Soin. pH

, after 48 hrs.

(at 24.0°C)

0.521 1.032 2.026 2.979 3.900 5.632

4.224 4.223 4.220 4.225 4.225 4.223

0.042 0.121 0.332 0.632 0.963 1.653

5.464 5.522 5.811 5.956 6.077 6.124

A

H

2

3

Initial [F

-

] in solution

(M x10

-4

)

Initial pH of soin.

(at 23.1°C)

[F

-

] in soin.

after 48 hrs.

(M x10

-4

)

Soin.

pH

after 48 hrs.

(at 22.5°C)

0.521 1.032 2.026 2.979 3.900 5.632

6.002 6.003 6.001 6.002 6.000 6.003

0.084 0.163 0.442 0.737 0.984 1.674

6.208 6.333 6.385 6.415 6.426 6.571

136

20

L•

a

4

0

0

0

, cl.

0

0.6 1.2„

Fin

solution (Mx 10)

0 Batch 22

• Batch 23

1.8

Figure

22. F adsarpticn frail batch experiments 22 and 23.

137

138

Batch # KU (ml/gm kaolinite)

22

23

3.04

10.72

0.5750

0.7192

The values above are presented only for illustration purposes; their strict quantitative meaning is uncertain. Notice that, according to the data in Table 10, adsorption in batch 22 was slightly higher than that in batch 23, yet its KU value from the regression analysis is smaller than that of batch 23. Such an apparent contradiction is due to the special characteristics of the regression scheme, although, in my opinion, it does not invalidate the observed trend, that is, adsorption does occur and the partition coefficients are greater than

1. It may be important to recall that, if K u

= 1 ml/gm, the mid-concentration point of the solute would be retarded by a factor between 5 and 11 relative to the bulk ground-water flow (Freeze and Cherry,

1979). Therefore, in any situation involving the movement of F

through sediments containing kaolinite-type clays, a great deal of retardation is to be expected, with the bulk of F

-

ions lagging well behind the solvent that originally carried them.

In both experiments, the equilibrium pH was higher than the corresponding initial pH, although the magnitude of such a change was not so significant as to postulate that the only mechanism involved was a direct exchange of F

-

by OH

-

, which is in agrearent with most of the research done up to date. Unfortunately, provisions were not taken to monitor the concentrations of alumina and silica in solution, which

139

would have helped determine the hypothetical formation of new insoluble phases, as also postulated by many scientists as one of the mechanisms responsitde for the F

-

uptake from soluticns.

140

CHAPTER 4

KINETICS OF FLUORITE DISSOLUTION

Introduction

In reading the literature review presented In Chapter 2, one soon realizes that many conclusions drawn from studies on the behavior of aqueous fluorine assume that the principles of equilibrium chemistry hold. The implications of embracing such assumptions are undoubtedly related to the scope attached to a given study and the degree of precision sought. Strictly speaking, however, it constitutes an oversimplification. While the magnitude and sign of some thermodynamic parameters can be used to decide whether a certain reaction will occur, those same quantities do not provide any indication of the time frame attachable to the process. The problem is by no means trivial, particularly in the case of potentially toxic solutes such as the fluorine species.

The purpose of the kinetic study presented in this chapter was to determine which of the most common chemical and physical characteristics of aqueous environments have the greatest influence in the dissolution of fluorite, CaF the most conspicuous F-carrying mineral in the crust. Each experiment was designed so to consider a range of the variable under testing comparable to that found in natural systems.

Another objective of this study was posed as to re-assess the ratelimiting mechanism in the dissolution of fluorite.

As in the adsorption study, column and batch experiments were performed. The column runs tested the effect of changing pHs and mean

141 flow velocities an the dissolution rate of fluorite, whereas the batch experiments focused on the changes of such rate when the degree of stirring, the bulk solution composition, and the temperature are varied. A summary of the type of experiments carried out, the variables considered, and the general objectives is offered below.

Type of

Experiment Purpose

Column

Solid Phase Solution Variables

Present

* To observe how Ground the dissolution fluorite rate changes as the pH is varied.

* To establish a functional relationship between the dissolution rate and the mean flow velocity.

Used

Neutral and acidified milli

-Q water

Tested pH

(mean flow veloc.)

Batch * To assess the quantitative changes in the dissolution rate as the solution composition, its temperature, and its degree of agitation are varied.

* To determine the rate-limiting step in the dissolution of fluorite.

Ground fluorite

* Neutral milli-Q water

* Neutral solutions of varied compositions.

stirring rate

[NaHCO

3

]

[CaSO4]

[NaCl]

Temperaature

Details on the experiments carried out as well as on their results are given in Appendix E. Concentrations of F

-

and pH were

142 measured as described in Appendix B, whereas Appendix A gives an explanation an the manner in which the specific surface(s) of fluorite particles was calculated.

Formulation and General Concepts

When water flows through permeable materials, the distribution of the solutes which the flow carries can theoretically be determined at any point and at any time by means of the following three-dimensional model:

EX,y,z v 2C +

Vx,y,z grad

C + w = n aC,

at

(4.1) where D is the dispersion coefficient, v

2

is the Laplace operator, C is the concentration of solute in solution, V is the pore velocity, w is a source or sink term, n is the porosity, and t is the time. In those cases where the dissolution of minerals can release solutes of the type used to trace the ground-water flow, then the term w above becomes a source. The actual functional relationship that expresses such rate of release, however, can be very complicated. In the dissolution of salts, for example, the term w takes on the following general form

(Dobrovol'skiy and Lyal'ko, 1983): w =

143 where k is the dissolution rate constant, s is the specific surface of the dissolving solid, C s

is the saturation concentration, m is an empirical exponent, and w and C are as defined before. Even if C s

is taken as a constant (actually, it is a function of, at least, temperature, pressure, and solution composition), we see that w cannot be considered solely as a function of the subsaturation (C

5

-C)because, according to Equation 4.1, C is by itself dependent on the flow velocity. This is also valid for k, the dissolution rate constant.

The objective of the column experiments will be to establish the dependence of k on the flow velocity. To do so, let us assume that when the dissolution of a given mineral such as fluorite becomes stationary, the distribution of [F

.

] along the main direction of flow

(x) can be described by (Dobrovol'skiy and Lyal'ko, 1983):

C = C s

[1 - exp(-ksx/v)] (symbols as above)

Rearranging that expression: ln (1-C/C

5

) = -ksx/v

Therefore, if one assumes that s, v, and C

5

are known constants, and measurements of C at various distances are available, the dissolution rate constant k can be obtained from the slope of the line on a graph of ln (1-C/C

5

) versus x. On a later step, the relationship between the so-obtained values for k and the flow velocities can be worked out.

144

For

the batch-like experiments, the basis of the analysis are slightly different in order to account for the non-renovation of solutions. If the assumptian is made that the dissolution of fluorite is controlled by the transport of products away from the dissolving mineral surface, the equation that describes the changes in concentration over time for a constant solvent volume is as follows

(Dobrovol'skiy and Lyal'ko, 1983):

= ks(C s

-C) = y(C.-C)

Furthermore, if k and s are assumed to remain essentially unchanged during the course of a given experiment, then Y becomes a constant, and the variables in the differential equation above can be separated: dC

= y dt

C s

-C

The integration of such an expression, with the initial conditions of C

= 0 at t = 0, yields:

In

C =

It

C s

-C

145

Similarly to what has been presented before, if measurements of C at different times are available, Y (and consequently k) can be obtained from the slope of the line on a graph of ln(C.X.-C) versus t.

Results and Discussion of Column Experiments

The percolation experiments used a plexi-glass column (see

Appendix E for its characteristics) packed with crushed, treated fluorite. The flow, which took place from bottom to top, was driven by a speed-controllable peristaltic pump. Five runs were carried out, numbered 1 through 5. Typically, water at a given discharge rate and pH was circulated, and samples were taken at the various sampling ports until the two last [r] readings at a given port did not differ by more than 1%, at which point it was assumed that a steady-state distribution of [F

-

] in the system had been achieved. In all cases, the dissolution became stationary after 4 to 5 hours, which has also been reported in

Dobrovol'skiy and Lyal'ko (1983). The final [F

-

] readings at each sampling port (i.e., C) where then related to the theoretical saturation concentration (C s

)--calculated on the basis of the temperature of the output solution--and the results processed as shown in Appendix E.

Runs 1, 2, and 3 were designed to assess the relationship between the mean flow velocity in the system (v) and the dissolution rate constant (k). Data from each run are presented in Figure 23 from which, as explained before, the individual k's can be estimated by considiaring the slope of the various lines. The numerical results obtained can be summarized as follows:

(

5

0/3

-

1 ) Ul r;

) i

i

146

147

Run #

Slope of Line

(ks/v) from Fig. 23

(air'))

1

2

3

- 0.0161764

- 0.0171119

- 0.0151796

Mean Flow

Velocity

(v)

(cm/min) (cm/sec)

0.9577

0.4447

1.3890

1.60 x10

7.41 x10-

2.32 x10

-2

-2

3

Diss. Rate

Constant (k)

(cm/sec)

7.64 x10

-7

3.75 x10

-7

1.04 x10

-6

When the logarithm of v and k were taken and the results plotted as shown in Figure 24, the relationship obtained turned out to be approximately linear. A regression analysis was then done in order to obtain estimates for the slope and intercept, and the results converted back to arithmetic scale, which yielded the following function to relate the dissolution rate constant to the mean flow velocity: k = 3.11 x 10

-5

v

°

.

(k, v in cm/sec)

In order to establish canparisons with published information, the data given by Dobrovol'skiy and Lyal'ko (1983) were similarly processed, as a result of which the following functional relationship was obtained: k =

1.66 x

10r

4

IP

78

(k, v in cm/sec)

-1

1

.8 -17

- I.G

Figure

24.

Relationship between the dissolution rate of fluorite and the mean pore velocity.

148

149

In an attempt to visualize better the differences between those pieces of data, the two functions were plotted as Shown in Figure 25.

Obvious differences exist, which result in discrepancies of about one order of magnitude between the k's calculated by each function for a given v.

The lack of agreement, however, may not prove to be critical in coming up with conclusions. Of the two parameters characterizing the power functions fitted, the exponent is the relevant one. If so, one may conclude that the two functions are essentially in agreement to each other. It is the constant term what establishes the greatest differences between the two expressions, and a tentative explanation is in order. It is my opinion that such a difference may arise from two facts:

(1) Dobrovol'skiy and Lyal'ko (1983) performed their experiments at 18 ±1°C, while the runs presented here were carried out at 22 ± 1.5°C.

(2)The calculation of k depends heavily on the value assumed for s, the specific surface of the particles undergoing dissolution. Dobrovol'skiy and Lyal'ko

(1983) used a value of 123 cm

-1

, whereas I employed a figure almost three times as large, 338 cm

-1

. I must concede that the russian authors may be able to rely more on their figure because the specific surface was measured by means of the thermal desorption of nitrogen, whereas the value I obtained came from a method bound to yield less precise estimates (see

Appendix A).

A

final annotation should be made on the fact that s is

I

o

-4

Kr

!

Dobrovol'skiy and Lyiarko (1983)

/ N

/

Runs I, 2, and 3

150

Range of used by

Dobrovorskiy and

Lyial l ko (1983)

Range of used in runs

1, 2, and

3

o

-9

10-

4

16-3 _

V

(cm/sec)

RS-2

Figure 25. Functicnal relationship between the dissolution rate cf fluorite and the mean pore velocity:

Comparison with published data.

assumed to remain constant during the course of the experiment, which may not constitute an appropriate assumption.

151

The overall conclusion is that there seems to be a quantitative relationship between the flow velocity and the dissolution rate of fluorite, which suggests that the dynamic conditions under which ground-water flow takes place may be the relevant aspect to look at in that they would control the extent of the dissolution process.

Runs 1, 4, and 5 were intended to reveal the influence of changing pHs on the dissolution rate of fluorite. Results from those tests are plotted in Figure 26, from which the following estimates of k were made:

Run #

1

4

5

Output pH

6.71

5.59

3.16

k (cm/sec)

7.64 x 10

-7

2.28 x 10

-7

1.97 x 10

-7

Although there is a slight increase in the value of k as the pH is increased, the actual functional relationship is not so clear. Four types of regression analysis on the data were tried (linear, logarithmic, exponential, and pcwer), and the best fit (r = 0.8) was given by the following exponential function:

(

'D/

0- i

) ul

152

153 k = 5 •93 xlCr

8

exp(0.33 pH) (k in cm/sec)

The exponential function above and the estimated values for k are plotted in Figure 27. Indeed, the relationship is not well established; more data would have been needed to support a firm analysis.

As given by the data, a pH change from 3 to about 5.6 led to a small increase in the dissolution rate, whereas the increase in pH from 5.6

to 6.7 made the dissolution rate increase considerably. Such behavior would suggest the existence of some pH threshold above which the dissolution process of fluorite might take place at an increased rate.

The point deserves further attention, and a series of well-planned column experiments should be carried out to provide the information needed to formalize the hypothesis above.

Results and Discussion of Batch Experiments

In order to determine the way in which selected variables can affect the dissolution rate of fluorite, sixteen batch tests were implemented, the results of which are summarized below:

Batch

Experiment #

5

6

7

3

4

1

2 k Measured

(cm/sec x10 8)

4.64

3.45

1.18

1.74

not calculated not calculated

2.19

10

8.

0 c G.

Ln

4.

2.

pH

Figure 27. Relationship between the dissolution rate of fluorite and pH.

8

154

155

12

13

14

15

16

10

11

8

9 not calculated

2.01

6.16

6.16

3.77

3.23

4.22

4.86

6.19

Data from experiments 1,2, and 3, graphically displayed in

Figure 28, indicated that the dissolution rate of fluorite and the solubility increased with increasing stirring rates. Similar results had been obtained by Dobrovol'skiy and Lyal'ko (1983), in what can be interpreted as the central role that the transport stages may play when other potentially affecting variables are kept constant.

Batch experiments 4, 5, and 6 were intended to establish the relationship between the dissolution rate constant and the presence in solution of variable amounts of Na and H00

3

-

. In the case of batch 5 and 6, the simultaneous presence of HCO

3

-

and CW+ led to precipitation of CaCO

3

. The calculation of k, therefore, was not done because it would have rendered little information inasmuch as (1) not just one solid phase was present at the moment fluorite was being dissolved, and

(2) the pH was neither constant nor close to neutrality (see Appendix

E). The data collected were nevertheless plotted for illustration purposes (Figure 29). Results from batch 2 were included to establish comparisons with a case in which the batch solution did not contain neither Na nor HCC1

3

-

species.

Experiments 7, 8, and 9 had the purpose of testing the hypothetical changes in k due to changes in the concentration of Ca++ and

o

go

Time (minutes)

460

GOO

Figure 28. Extent of fluorite dissolution as a function of time for various stirring rates.

156

I

1.0

re)

d

,

I in

N

d

( 0-sO/s0) u

I d

0

1•n• d

157

158

SO

4

= in solution. Some unexpected problems rendered data from batch 8 useless (see Appendix E). The data have been plotted as shown in

Figure 30, to which the results from batch 2 (no Ca" or SO

4

= in solution) have been added. The tendency observed indicates that, as

[Ca"] and [SO

4

=] in solution increased, k as well as the F

-

solubility decreased slightly.

Tests 10, 11, and 12 were designed to show how changes in the ion strength of the batch solutions (p) would affect the dissolution rate of fluorite. The data are presented in Figure 31, which also contains the batch 2 data. The trend is not clear. Although k seems to increase as decreases, two facts do not lend support to the postulate, namely, (1) batch 10 and 11 yielded just about the same k value, in spite of the 5-fold difference in p between them, and (2) the k value obtained from batch 2 should have been larger. As pointed out by

Dobrovol'skiy and Lyal'ko (1983), the kinetic equation w = ks(1-C/C s

) should more properly be written as follows: w = ks [1 -

(Ily i

C)/SP],

1 where

Tly i g is the product of ion activities for fluorite, and SP is

1 the solubility product. That expression indicates that any increase in the concentration of ions in solution (non-common ions) should result in an increase of the dissolution rate k to counter the corresponding decrease in the activity coefficients. However, Dobrovol'skiy and

Lyallko (1983) reported a regular decrease in the dissolution rate of fluorite when the [NaCl] in solution was increased to values as high as

In to d

In

N d

(0--s3/s0)ul

In o

159

GOO

Figure 31. Mftent of fluorite dissolution as a function of time for various ion strength of solutions.

160

161

1

M. For the low range of p (i.e., more natural range), between 0 and

10

-2

NI, the authors did not find a noticeable change in the value of k, which indicates that the effect of low-salinity waters can be neglected in those

cases

where not very accurate estimates are to be made. The more general conclusion drawn by the Russian authors states that "the variations in dissolution rate therefore do not follow the activity product for calcium and fluoride, which indicates that the main factor is the dependence of k on the solution canpositicn and concentration.

Therefore, the possibility of theoretical prediction is very problematic unless an experimental study is made of the dissolution kinetics"

(Dobrovol'skiy and Lyal'ko, 1983, p. 76).

I should mention that, up to this point, the various figures presented did not include the plotting position of the last datum taken at each test (see Appendix E). It so happens that, if such points were to be included, they would almost always lie below the lines drawn for each test. That might indicate that, in the long run, k's would be smaller than those calculated. Perhaps the k's calculated correspond to the early dissolution of smaller fluorite particles, in which case the experiments should have lasted longer to obtain representative estimates of k. Dobrovol'skiy and Lyaliko (1983) disregarded the possibility of changes in the dissolution rate beyond the time frame of their experiments (few hours), but I am inclined to believe that the point deserves a closer look.

The last series of batch tests, numbered 13 through 16, were carried out to determine the changes experimented by k as the temperature is varied, and, from these data, to be able to assess the dissolu-

162 tion-controlling mechanism. It should be said that high-precision constant temperature baths were employed, and that the batch solutions were closely monitored to avoid even the smallest temperature change during the course of a given experiment. The analysis of the data collected, which are plotted in Figure 32, concluded (as expected from theoretical considerations) that the increase in temperature enhanced the dissolution of fluorite. The results were then converted as shown below to obtain the coordinates of the log k versus 1/T plot (Figure

33), better known as the Arrhenius diagram.

Batch Diss. Rate, k Temperature

(cm/sec x10

-8

) (°C) (°K)

13

14

15

16

3.23202

4.22214

4.85700

6.19272

log k

15.0

288.18

-7.4905

20.1

293.28

-7.3745

24.8

297.95

-7.3136

30.9

304.05

-7.2081

1/T

[(1/°K)x10-3]

3.47005

3.40971

3.35627

3.28894

As reviewed in Chapter 2, the slope of the line in Figure 33 can be used to estimate the activation energy

(E.)

which characterizes the dissolution process because such slope is equal to (-E a

/2.303R). For the case presented here, E a

turned out to be 7.184 kcal/mole, which suggests that the rate of dissolution of fluorite is controlled by surface reaction (Berner, 1978; Lasaga, 1981). It should be said that the straight line in Figure 33 was fitted by hand in view of the good alignment of three of the four data points (a linear regression of the data, with a correlation coefficient of -0.996, yielded an estimated

163

500

Figure

32.

act

it of fluorite dissolution as a function of time for various temperature of solutions.

-7.50

328 : . 3..3 g

G ' 3.40

I/T (1/°K x10

-3

) g

.

3.44

3.48

Figure

33. Arrhenius

plot.

164

165 regression coefficient such that the Ea would have been 6.972 kcal/mol, a value very close to that obtained from the fitting by hand).

Although the value of Ea obtained may find support in the statement made by Berner (1978) to the effect that most dissolution processes occurring under normal earth surface conditions are controlled by surface reaction, it should not be disregarded that Lasaga (1981) pointed out that such processes are characterized by an overall empirical E a

distinctly greater than 4-5 kcal/mole. Inasmuch as I cannot claim that the value of 7.2 kcal/mole that I measured is conclusively greater than 4-5 kcal/mole, and based also on the results of my column experiments, I am inclined to conclude that both surface reaction and transport conditions seem to control the solubility of fluorite, perhaps with a little more weight given to the constraints imposed by the surface reaction mechanist. Incidently

,

my conclusion would agree with the observation made by Berner (1978, 1981) that, for minerals with solubilities in the range of 10r

4

to 10

-3

M (fluorite is one of them), the rate-controlling mechanism is a mixture of transport and surface reaction.

Dobrovol'skiy and Lyal'ko (1983) reported an E a

of 13.8 kcal/ mole for the temperature range of 0.5 to 13.5°C, and an Ea of 4.8 kcal

/mole for temperatures in the range of 23 to 100°C. Their conclusion was that surface reaction dominated the picture at the lower temperature range, whereas external diffusion (i.e., transport) was the relevant controlling mechanism at the higher range of temperatures. The temperature range 13-23°C was interpreted as a transitional state in which both mechanisms operate, a conclusion that was elaborated from a

166 change in the slope of the Arrhenius plot which the authors presented in their 1983 paper.

A final comment should be made to the effect that, independently of the variables tested in each case, the dissolution rates obtained from the batch experiments carried out resulted in values about one order of magnitude smaller than those measured in the column runs. If the existence of measurement errors is disregarded, an explanation for such a difference can be afforded by reference to the very distinct nature of the two experimental techniques. In batch

tests, the solutions (stirred or not) are not renewed, whereas in column experiments, the products of a given reaction between the solution and the surrounding medium are carried away with the flow and replaced by fresh solution. Whatever the amount of solutes locally incorporated to the bulk flow, therefore, the degree of undersaturation of column solutions is always greater than the equivalent batch solutions, which should lead to larger dissolution rates.

167

CHAPTER 5

TRANSPORT OF FLUORIDE: AN EXPERIMENTAL APPROACH

Introduction

The purposes of the experiments described in this chapter were two-fold:

(1) to assess the F-affecting chemical changes taking place when F-carrying solutions are flushed through material commonly found in aquifers, and

(2) to observe the extent of a F-mineral dissolution when water circulates through systems containing conceivable

(i.e., natural) amounts of such mineral along with other solid phases of common occurrence in aqueous environments.

Those objectives were investigated by means of column tests, which were labeled composite experiments. The experiments relevant to

(1) above consisted of three column runs (numbered 1 through 3) in which column 1 was used. Objective (2) was studied through a single column test, numbered run 4, which employed column 2. Characteristics of the columns used, the packing materials, and the results of dispersion tests are given in Appendix C. The minerals used for packing the columns were prepared as described in Appendix A.

For runs 1, 2, and 3, once the input solution was prepared and a given run got started, the test continued in a fashion similar to that described in Chapter 3; that is, samples were taken at the outlet

168 until the electrical conductivity readings indicated that the bulk solution had arrived at the top of the column, at which time a two-way valve was switched to let distilled water In. Sampling continued until the electrical conductivity at the outlet was comparable to that of the background solution before the experiment started.

Run 4, on the other hand, consisted of a continuous input of distilled water at a constant rate, with samples being taken at the outlet (unlike the runs mentioned above, there was no switching to other type of solution).

Samples were then taken to the laboratory for the analytical determination of several species. Results of such analyses are presented in Appendix F.

Handling of

Chemical

Data

As reported in Appendix F, [Cr], [SO

4

=], and [F

-

] were determined by ion dhnornatograp [Ca"] by atomic adsorption techniques, whereas pH, temperature, and conductivity were measured at sampling time. Concentrations of Na' and HCC1

3

-

, however, could not be analytically determined due to some unexpected problems with the measuring apparatus. Inasmuch as those parameters were deemed to be essential for this study, an attempt was made to estimate their approximate concentrations on the basis of the assumptions offered below.

Having readings of electrical conductivity (C), the corresponding value of total dissolved solids (TDS) can be obtained from the following empirical relationship (CUstodio and Llamas, 1976, p. 209):

169

C (micromhos/cm) = 1.35 TDS (ppm)

The expression above is valid for a temperature of 18°C. The C measurements taken throughout the runs, however, were normalized to 25

°C; therefore, a correction factor should be applied to account for such a 7-°C difference. Electrical ccnolictivity, on the average, increases about 2% per °C of temperature increase (CUstodio and Llamas,

1976, p. 208), which means that, for 25°C, the relationship between C and TDS reads as follows:

C = 1.54 'IDS (units same as above) which was used to estimate the TDS of all samples taken. The composition of the various input solutions were known, so that the initial TDS

(TDS,) and the initial concentration of HCO

3

-

([HCO

3

-

1

0

) are data. If the assumption is made that HCO

3

-

behaves as a non-reactive species, the concentration of bicarbonate of any sample x can be calculated as follows:

[HCO

3

-

], = [HCO

3

-

], TDS

TDS

0

Having so obtained the bicarbonate concentration of all samples, and considering that 'IDS represents the sum of the conceatrations of all species, the Na concentrations were obtained by difference.

Several objections can be raised to the use of the approach described above, in particular to the categorization of HCO

3

-

as a non-

170 reactive ion. Consequently, any interpretation involving [Na] and

[HCO

3

-

] data will be treated cautiously. It should be added that the samples have not been discarded, and that [Na] and [HCCI

3

-

] analytical measurements will be taken as soon as the facilities are available to do so.

Results and Interpretation of Column Tests 1, 2, and 3

Inasmudh as evidence presented in earlier chapters had demonstrated that pH was a master variable in many phenomena related to the geochemistry of fluorine, the input solutions used in these tests were of similar ccmpositicn but having a distinct pH (4.07, 6.25, and 8.49).

However, the buffer capacity of the system, mostly attributable to the presence of vermiculite, was of a magnitude high enough to virtually erase the pH imprint of the feed solutions, which resulted in output solutions with an average pH of around 7 (range: 6.7 to 7.5). Although data presented in Chapter 3 can give an idea on the buffer capacity of the packing materials, a simple test was decided to gain a better insight on the observed pH fluctuations. Samples of the potentiallybuffering packing materials (vermiculite and kaolinite) were brought into contact with solutions of a composition similar to those used in runs 1 and 2, and their pH adjusted by adding proper amounts of diluted

HC1 or NaOH. After a few hours, the pH of the mixtures was measured and recorded. The batch containers were kept closed throughout the test, and the pH measurements were taken in such a way so as to avoid equilibration with the atmosphere. The results can be summarized as follows:

171

For vermiculite: (vermiculite/solution = 5 gm/60 ml)

Initial Values pH T (°C)

3.872

6.032

8.511

21.6

21.6

21.6

Final Values

(after 12 hours) pH T (°C)

6.705

6.884

7.977

21.5

21.7

21.6

For kaolinite: (kaolinite/solution = 5 gm/60 ml)

Initial Values pH

T (°C)

3.955

5.985

8.553

21.4

21.1

20.9

Final Values

(after

8 hours) pH T (°C)

4.005

4.325

5.533

22.0

22.5

22.2

The data above point to an opposite buffering effect, with vermiculite making the pH stay in the range of 6.7-8.0 and kaolinite lowering it. In the column, where both minerals are present, those buffer effects overlap.

The result is a stabilization of pH around neutrality, which may suggest the prevalent buffering effect by vermiculite because of its being present in larger amounts and its increased surface area (this is industrial vermiculite, and the

172 artificial exfoliation should have resulted in an increased surface activity of the mineral).

The chemical data displayed in Appendix F indicate that, in general terms, two groups can be distinguished on the basis of ion concentrations at the output, namely, one made up of non-reactive species (C1

-

and 50

4

= ), and a second group of reactive ions (Na+, Ca", and F

-

). The HCO

3

-

ion was not included In the interpretation because, due to the very nature in which its concentrations were calculated, it would have shown a systematic (and probably incorrect) non-reactive behavior. Of course, terms such as "reactive" and "non-reactive" can be used only after some standard behavior is defined to compare against. In this particular case, it was thought that conductivity, due to its "averaging" nature, would provide the proper raw data for the transport analysis. Therefore, the conductivity data available for each run were used as input information for CFITIM, a computer code specifically designed to help process data from percolation tests (see

Chapter 3 and Appendix D). For runs 1, 2, and 3, CFIT1M estimated retardation factors of 1.67, 1.49, and 1.97, respectively. The breakthrough curve based on the conductivity measurements as well as the

[Cr] and [SO

4

=] data are presented in Figures 34 through 36. As for the non-conservative species, the [Na+], [Cai

4

], and [F

-

] data have been plotted in Figures 37 through 39, to which the breakthrough curve for conductivity has been added for comparison purposes.

Figures 34 and 35 reveal that Cl

-

and SO4= follow the conductivity breakthrough pattern quite closely, although in run 3 (Figure

36), there seems to be an early breakthrough of those species which

o

Figure 34. Brwadthrough of non-reactive species: Run 1

(solid line: CFITIIIM fit based on conductivity data).

173

2

Figure 35. Breakthrough of non-reactive species: Run 2

(solid line: CFITIM fit based on conductivity data).

174

175

I . 0

o 415

0.8-

0.G

-

(

o

0.4

X

0.2-

0 2

4

G

8

Time

(pore volumes)

10

12

Figure 36. Breakthrough of non-reactive species: Run 3

(solid line: CFITIM fit based on ccnductivity data).

14

176 could be attributed to ion-exclusion effects. It may also happen that, in this case, the conductivity readings are showing the effect of relevant modifying phenomena and, as such, the use of conductivity as a conservative parameter would be ruled out. If so, a much proper nonreactive breakthrough curve can be obtained from the Cl

-

or SC1

4

= data.

As for the non-conservative species, Figures 37 through 39 present common characteristics, namely, a great deal of Cac" has been retained by the sediments, at the upper limb of the breakthrough curve the relative concentrations of Na+ and F

-

exceed 1, and the peak of the

F

-

breakthrough is displaced (i.e., retarded) to the right of the conductivity breakthrough function. The transport behavior of Na + and

Ca++ ions would be consistent with a mechanism based on ion exchange on kaolinite.

It should be recalled that, by virtue of its preparation

(Appendix A), most of the exchange sites on the clay surface must have been occupied by Na+. Therefore, the

[Na+] measured at the outlet represent the sum of the input [Na] plus that amount of Na+ coming out from the exchange reactions with CA ++, which would result (as observed) in relative concentrations of Nfac' greater than

1.

An explanation for the odd

F

-

breakthrough may prove harder to find. Although not reaching immobility, the movement of F

-

ions in the system appears to be retarded (reactions with kaolinite?) until distilled water starts circulating through the column. The distilled water seems to have had a sweeping effect on the

F

-

ions. If the assumption is made that the F

-

ions were being subject to adsorption or to some other reversible mechanism, the arrival of distilled water has produced a fast desorption and agglomeration of

F

-

ions in a front that

0 2 4

G

Time

(pore volumes)

8

10

Figure 37. Breakthrough of reactive species: Run 1

(solid line: CFITIM fit based

an oonductivity data).

177

I.G

1.4 -

Cond.

6

Na +

0

Ca++

x

F- •

o

0

0

1.2-

1.0-

(5)

n

0.8

O

-

0.G-

0.4-

o

X

0.2-

o

2

4

G

Time (pore volumes)

10 r

12

Figure 38. Breakthrough of reactive species: Run 2

(solid line: CFITIIM fit based on conductivity data).

178

o

Cond.

Na+

Ca+

+

F

179

1.4

1.2-

1.0-

0.4-

0.2-

12

Figure 39. Breakthrough of reactive species: Run 3

(solid line: CFIT1M fit based on conductivity data).

14

180 exits the column, with relative concentrations of F

-

greater than 1.

This effect is worth investigating by future experiments for the implications that it may have in envbxmmental impact studies.

Although none of the runs lasted long enough to observe the F tailing, an attempt was made to carry out a mass balance of the species. The measured F

-

ccricentrations were plotted as a function of time, and the areas under the breakthrough curves were compared with those representing the continuous source (Figures 40 through 42). A summary of the results obtained is presented below:

Run #

1

2

3

Mass of F

-

Mass of F

-

% Mass

Injected (mg) Recovered (mg) Recovery

7.89

11.46

11.08

6.21

11.25

10.12

79

98

91

As given by the figures above, the proportion of F

-

not accountable for is very small, and probably can be neglected due to the approximate nature of the approach used to obtain the mass balance (tailing could have been more pronounced, peak may not have occurred where indicated, etc.).

Mass balances were also done for Ca" and Ne. The graphical approach described above was used to calculate the percentage of mass recovered based on the plots in Figures 43 through 48. The numerical results Obtained were as follows:

I

11)

L.

>

(1)

0

181

co

Tr

(Wdd) _J

;0

uo!4D.44.ua3uo3

f

0

......./..

I

M

CL

C

....

_

I

0

1

co

t

I

(.0 it

( Ul d d

) _ j

4

0

UO

I

4.

D

J

.1.

UQ 0 U

Op

t c\I

o

.. it

•MIINP

182

183

o

(pore volumes)

Figure 42. F

-

mass balance: Run 3 (curve fitted by hand; tailing inferred).

Figure 43 • Ca" mass balance: Run I

(curve fitted by hand).

184

185

2 5-

4

G

8

Time (pore volumes)

Figure 44.

ce +

mass balance: Run 2

(curve fitted by hand).

10 12

0 4

8

Time

(pore volumes)

Figure

45. CW+ mass balance: Run

3

(curve fitted by band).

12

186

(Ludd)

+

DN

JO

uoilDipapuop

0

187

(Ladd)

+

DN io uoi4.D.14.ua3uo0

0

.MMA

IND "

=MEND

CO

188

..

(D

189

tu

0

CS

CO

CD

0 cr

1

0

(.1

(wdd) +

DN 0 uou.D.quaDuoo

0

For calcium:

Run

#

Mass of CaL++

Injected (mg)

Mass of

Ca,*

Recovered (mg)

%

Mass

Mass of

Ca++

Recovery

Lost (meg)

1

2

3

26.30

39.16

37.57

7.71

9.41

7.19

29

29

19

9.3 x 10

-4

1.5 x 1Cr

3

1.5 x 10

-3

For sodium:

Run

Mass of Na

+

Mass of Na %

Mass

Excess

Mass

#

Injected (mg) Recovered (mg) Recovery of Na + ( meg)

1

2

3

63.67

93.22

93.93

112.57

153.01

128.50

177

164

137

2.1 x 10

2.6 x 10

1.5 x 10

-3

-3

-3

For both ions:

Run # Excess

Na

/Ca' lost (meg/meg)

1

2

3

2.29

1.75

0.99

190

191

It is important to notice that, had ion exchange been the only active mechanism, the ratio shown above should have always been close to 1.

That has only been achieved in run 3, with a tendency for the ratio to decrease as the runs proceeded. In view of these results, and although it may be said that the hypothesis of having ion exchange occurring in the system is plausible, mention should be made of some factors which can altere the interpretation, namely, (1) the possible interactions between Ca" and H00

3

-, which could not be checked because of the lack of [HC10

3

-

] measurements; (2) the way in which Na concentrations were calculated, which leaves a wide margin for errors; and (3) before proceeding with the various runs, a dispersion test was carried out, and a solution of 1000 ppm NaC1 was used. Although the column was later cleaned up by flushing several pore volumes of distilled water, some

Na may have remained in the system, which would act as a source not included in the considPrations made above.

Aside from what has been discussed, a series of suggestions can be made applicable to future experiments of the kind presented here:

(1) The input solutions should be allowed to circulate through the system long enough for all constituents to reach some output equilibrium concentration and the results analyzed within the frame of the continuoussource transport theory. Experiments of this sort can provide valuable information, which could be used later in the interpretation of tests in which the recession limb of the various breakthrough curves has also been measured (such as in those cases presented here).

(2) In order to "standardize" the status of the system and, thus, facilitate the interpretation, the clays used should be regenerated after a given run is completed by circulation of solutions containing excess amounts of the exchangeable species.

(3)Complete chemical analyses are definitively needed if the investigator is to have good insight into the processes taking place in the system.

192

Results and InteLpLetation of Column Test 4

As

given in Appendix C, column 2 included 2% of crushed, pretreated fluorite as part of the packing along with 88% sand, 6% v

e

rmiculite, and 4% kaolinite. Ublike earlier experiments, run 4 used milli-Q water as the input solution throughout the whole time-span of the test. The idea was to test the degree of saturation with respect to fluorite that waters devoid of solutes could achieve when flowing through material carrying the mineral. The proportion of fluorite in the sediments was selected so as to reproduce what might be found in natural settings. The run proceeded at a constant rate of 1.02 ml/min, which led to mean pore velocities in the order of 0.26 cm/min. Samples were taken at the outlet and sent to the laboratory for analysis of

[Cac"], [F], [NW], [SO

4

=], and [C1

-

]. Electrical conductivity, pH, and temperature were measured at sampling time. A description of the samples and the results of the chemical analysis are given in Appendix

F.

193

As revealed by the analyses, the first striking feature is that output

[Car'] was null (or below the detection limit). Concentrations of

F

-

rapidly climbed to 1.3 ppm and remained constant until the end of the test. [Na+] behaved quite steadily, too, with output concentrations around 3 ppm. [Cr] and [SO4'] were systematically below the detection limit.

The breakthrough of F

-

and Na+ can be seen in Figure

49, which also includes the concentrations of

F

-

and Car' corresponding to the equilibrium dissolution of fluorite. These concentrations were calculated after correcting the equilibrium constant for the dissolution of fluorite for temperatures different from

25°C, with the new temperatures being those monitored throughout the run. The obvious conclusion from Figure 49 is that equilibrium with fluorite was not reached. If it did, [F

-

] would have been around 7.5 ppm or even higher because most of the Ca" coming out of the mineral dissolution would have exchanged with Na+ groups on the kaolinite, which would have resulted in increased amounts of F

-

in solution.

Mbreover, the Na breakthrough curve does not show any effect of a relevant ion-exchange process taking place. In a way, these results are not surprising if one considers that fluorite is a sparingly soluble mineral phase and, as such, is not expected to release species to solutions easily. Notice, on the other hand, that a concentration of F

-

above 1 ppm was rapidly acquired by the passing solution, which indicates that waters entering a region with a all mineral content in the sediments will incorporate enough F

-

to be above the safe limit for human consumption, a figure

194

EQUILIBRIUM WITH FLUORITE

MEASURED

x

No+

o

2

.

4

.

G

Time

(pore volumes)

8

.

.

10 12

Figure

49.

Breakthrough of

F

-

and Na and equilibrium with respect to fluorite.

195 around

1 ppm

of F

-

(U.S. Dept. of Health, Education, and Welfare,

1962).

Inasmuch as the dissolution of fluorite appears to be a very slow process, it would have been interesting to continue run

4 for a long time, perhaps several weeks or even months, in order to observe the equilibrium concentration of

F

-

that would have been reached after such a period. It would also be interesting to find out how the results from such experiments compare with the conclusions drawn in

Chapter

4, in which the kinetics of the dissolution of

CaF

2

were treated.

196

CHAPTER 6

STATISTICAL STUDY

Introduction

Due to the complexity in the chemical evolution of ground waters and the frequent availability of a consideTable amount of chemical data, hydrologists are often unable to obtain a clear picture of the system under study. The basic behavioral model must be known because it constitutes the necessary framework upon which more sophisticated interpretations or detailed explanations can be built. It is in those first stages of any investigation that multivariate analysis may be able to fulfill the needs of the hydrologist, especially in these days of high-speed electronic computers.

A typical hydrochemical study, or, for that matter any study in which several variables are considered, involves the collection of a significant amount of data. If z represents the attribute (variable) measured and x the sample, the data base is usually presented in the following form:

(xj. )

(x

ç, ) zp

(x i

)

• zp (x n

)

197

Each row represents an individual observation, while each column contains the information of the value of a given attribute at the various sampling sites. Given such an array, the researcher will always be interested in the following (Myers, 1985):

(1) generating new rows; that is, to be able to estimate

(predict) the value that the attributes may take on at sites for which samples are not available.

(2) completing the data matrix. Some entries may be missing because the measurement of all attributes at all sampling sites may be neither possible nor practical.

(3) identifying the correlation among attributes and among variables.

(4) reducing the number of variables in order to facilitate the interpretation.

Among those techniques currently employed in multivariate analysis, the following can be mentioned as those of special interest in hydrology: principal components analysis, factor analysis, correspondence analysis, discriminatory analysis, cluster analysis, and kriging.

Two cases of the application of factor and correspondence analyses to the study of hydrochemical data will be presented in this chapter. The first study case involves the hydrochemical evaluation of an area in southern Arizona, the upper San Pedro River basin, where I collected data for my M.S. thesis. After carrying out a hydrogeologi-

198 cal and hydrochemical assessment of the area, the application of the techniques mentioned was undertaken to define the observed chemical trends.

The second example is the application of only factor analysis to the study of the hydrochemical mechanisms in a portion of the San

Bernardino Valley (California), and the way in which those phenomena relate to the geochemistry of F

-

in local aquifers.

Factor Analysis: Definitions, Concepts, and Terminology

Factor analysis (FA), a technique with basis in the statistics, was so extensively used in psychology in the first half of this century that is commonly considered incoLiectly as a technique used exclusively by psychologists. The ready accessibility to computers that started in the early 1950's, however, resulted in the application of FA in numerous fields other than psychology. Among those disciplines in which FA is of present common use, Harman (1976) listed the following: economics, medicine, physical sciences, political science and sociology, and regional science. In chemistry, and as a result of the revolutionary marriage of chemistry and computer science, a new subdiscipline has been born called chemometrics, in which FA has proven to be one of the most potent techniques (Malinowski and Flowery, 1980).

Formally defined, and in words by Davies (1973), "the purpose of factor analysis is to interpret the structure within the variancecovariance matrix of a multivariate data collection". Such an analysis of the covariance structure results in a linear model dependent on random and unobservable variables, the so-called common factors. This

199 characteristic establishes the difference between FA and the more common multivariate regression model, in which the independent variables can be observed (Johnson and Wichern, 1982).

The presentation of the theoretical foundations of FA is far beyond the scope of this section, and only a brief review of the main definitions and concepts is presented below. For specific details on the subject, the interested reader is referred to the key work by

Harman (1976), which is the basis of the summary that follows.

Any multivariate analysis starts by obtaining the simplest statistics out of the data base. If N is the number of observations, the sample (as opposed to population) mean of any variable X i

is =

/N. The observed values of the variables are commonly transformed to a more convenient form. One of those transformations is the fixing of the origin which, when taken as the sample mean, produces the quantity x

3

= X ii

- Ti: j

, called a deviate. The sample variance of variable can then be defined as:

E x..

2 iN

J1 '

When the sample standard deviation (i.e., the square root of the sample variance) is taken as the unit of measurement (another of the transformation above mentioned), the standard deviate of variable j for individual i is given by:

200

And, for any two variables j and k, the sample covariance is defined bar:

S i

k

=

Ex

1

x ki

/N

Finally, the correlation coefficient between those two generic variables j and k is given by: r ik

=

S i

k

iS i

S k

Ezi z ki

/N

The computation of the correlation coefficient and the covariance between all possible pairs of variables is the first step in FA, which results in the correlation coefficient and the covariance matrices.

The FA model, designed to reproduce the maximum observed correlations, can be expressed in the following form: z i

= a ji

F i

+ a j2

F

2

+ . . . + a im

F. + u.Y. (j = 1,2,...,n), where each of the n variables is described linearly in terms of m common factors

(F) and a unique factor (Y). Notice that if m < n, the goal of FA, the description of the system under study. becomes simpler.

The coefficients of the factors (a) are referred to as "loadings". The common factors and the unique factor have zero means and unit variances. Furthermore, the common factors are usually assumed to be

201 uncorrelated among themselves, and the unique factor is always uncorrelated with the common factors.

The total contribution of a factor to the variance of all the variables (VO is defined as follows:

V

P

= z a2

.

J j=1

P

(p = 1,2 m) and the total contribution of all the common factors to the total variance of all the variables is given by:

V=

E p=1

The ratio Vin is sometimes taken to be an indicator of the completeness or robustness of the FA performed.

The communality of a variable, which is given by the sum of the squares of the factor loadings on that variable (i.e., 10 j

= a

2 a

2 j2

+ .

+i.m'

• j = 1,2,...,n), is a measure of how well that variable can be explained by linear combinations of the other variables. Thus, a high communality value reveals an effective relationship between that variable and the others, which opens up the possibility of defining it in terms of those other variables. Conversely, a low communality value indicates that the variable is unique or, better defined, that its variation is mostly explained by unique factors or by variables not included in the analysis.

202

Once the correlation matrix (i.e., the array of all possible pairs of correlation coefficients) is found, the system is diagonalized to obtain its principal directional components, or eigenvectors, with their associated magnitudes (eigenvalues). Factor 1 will be related to the largest eigenvalue and explaining the greatest amount possible of variance in the data set. The second factor, orthogonal to the first one (i.e., independent) and associated to the second largest eigenvalue, explains the greatest of the remaining variance, and so forth.

A key advantage of the FA is the possibility of plotting the position of the variables onto the space determined by two given factors. In the ideal case in which two factors account for most of the observed variance, the two-dimensional plot will help considerably in the interpretation stage. Even when three factors are to be considered, the representation can still be done in a two-dimensional domain by carrying out some basic manipulations (Kiayan, 1966). The procedure is as follows: inasmuch as the communality of a given variable (h

2 i

) is defined as the sum of the squares of the loadings of the factors on the variable (a"), a new coordinate A" of variable i in factor j can be defined as follows:

=

For each variable, the sum of the A"'s over the factors (3, in this case) is 1, which allows the plotting on a triangular diagram.

With particular reference to the use of FA in hydrochemical studies, Dalton and Upchurch (1978) indicated that the technique

203 emphasizes groups or relationships shown on trilinear diagrams (Piper,

1944; Durov, 1948) and adds variables that such diagrams cannot include. According to Dalton and Upchurch (1978), the key advantages of the FA approach over more traditional graphical methods can be summarized as follows:

(1) Neutral chemical species (e.g., Si0

,

2

) and non-chemical data (temperature, well depth, etc.) can be included in the interpretation.

(2) Variations of ions in small concentrations (F', Br

-

) are not masked by the variations in chemically similar ions in greater concentrations (C1').

(3) Secondary mixing trends are emphasized.

A final reference should be made to the types or nodes of FA currently in use. When the data matrix is presented so that the rows represent individual samples and the columns are the values that the variables take at each observation, the FA carried out is termed Rmode. Such a data matrix can be transposed (i.e., turned sideways so that each row represents the value of the variables at the various sampling points, and each column an individual sample) and the technique applied, in which case the FA is said to be of the 10-node. In short, the R-mode describes similarities among variables, whereas the

10-node gives the correlation pattern(s) among samples.

204

Correspondence Analysis: Definitions, Concepts, and Terminology

Cbrrespondence analysis (CA) was one of the data analysis techniques developed by a group of distinguished French researchers led by Jean-Paul

Benzébri. The philosophical framework of those data analysts, summarized quite clearly in Benzeeri's second principle of data analysis, "The model must fit the data, not vice versa" (Greenacre,

1984), was built upon inductive reasoning, proceeding from the particular to the general. The efforts of the research group resulted in the development of mainly geometric descriptive techniques, among which

CA stands out as the best product.

The raw data for CA are the same as for FA, that is, a table containing the values taken up by certain variables at several sampling locations, known also as contingency table or simply data matrix. The scaling transformations made in FA's R and Q-modes, however, are nonsymmetrical and, hence, need separate interpretations. In CA, the scaling and weighing procedures used make both analyses equivalent so that only one diagram is produced to show the plotting positions of both variables and samples (David et al., 1977). According to

Greenacre (1984), although the inspection of raw data can provide a clear picture, such a display of points is much more informative and very convenient in order to make canpariscns and to identify patterns in the data.

The so-called canal display constitutes a key advantage of CA in that it allows the calculation of what Greenacre (1984) called the column profiles (i.e., the relationships among variables) and the row profiles (i.e., the relationships among samples), and a certain degree

an excellent way of detecting errors in the data.

Inasmuch as all the variables (or samples) are "responsible" for the appearance of a given factor, the sum of the AC's for that factor would add up to 1 (or 100% if percentually expressed).

(2)

The relative contributions (RC) of the factors, which are indicative of how much each factor has contributed to the departure of a given variable (or sample) from an average condition defined as the center of gravity of both clouds (one of the clouds is determined by the plotting positions of the samples in the variables space, and the other by the positions of the variables in the samples space). The sum of the RC's of the factors on a given variable (or sample) is, as before,

1 or 100%.

205

Before proceeding with the presentation of examples of application of this technique, it is important to mention that Greenacre

(1984) suggested that CA be used as a exploratory tool rather than a confirmatory one. Such a statement, which can also be made extensive to FA, is indeed very appropriate in that it highlights the actual scope of a much-needed multi-disciplinary approach in the study of cases involving many variables and observations. In the context of hydrological studies, and if a careful balance is achieved between the information provided by these techniques and the field hydrological, geological, and chemical data, hydrologists will most likely come to

206 appreciate the relevant contribution of multivariate analysis to the better understanding of hydrological systems (Usunoff and Guzman-

Guzman, 1988).

Example 1: The Upper San Pedro River Basin (Arizona)

The study area, a typical section of the Basin and Range

Province, lies in the northeast part of the upper San Pedro River basin, Cbchise County, Arizona, with a N-S length of 24 miles and a width of 18 miles (Figure 50).

Hydrogeologically, two water-bearing units can be distinguished: a lower, confined aquifer (locally artesian) in Miocene-Pliocene sedimentary rocks [valley-fill deposits], and an upper, unconfined aquifer in Pleistocene sediments [flood-plain alluvium], separated by a low-permeability unit of more than 1000 feet thick in the center of the valley. Recharge from precipitations takes place at the mountain fronts, whereby the general pattern of the ground-water flow is from the mountains toward the valley axis and then northward. Additional sources of recharge to the unconfined aquifer are infiltration from runoff, seepage from irrigated lands, and water moving upward from the confined aquifer through breaks or discontinuities in the confining beds or through wells tapping both aquifers (Montgomery, 1963; Roeske and Wrrell, 1973). DeWald (1984) has suggested that recharge along washes and stream channels may represent an important source of water for the confined aquifer.

A composite interpretation of earlier studies (Smith and

Cammack Smith, 1932; Wallace and Cooper, 1970; Roeske and Werrell,

15

• ,

• .

• r-4..

.•

,

" •

, •

LOWER

SAN

,

1

PEDRO

BASIN s,

'• •.„,„"

• "

UPPER

SAN

PEDRO BASIN

,•

3200'

I?

1,7

......•••••• n

Boundary of the POsin

---4000---Contour elevation line (feet 0.m.al.)

Alluvial plain along

rivers and wasnes.

Piedmort sidoe

aric

aiii;viai fans.

1.77

mountains.

6

• unseal feed aquifer sample

• confined aquifer temple

'er spring sample r isisr sampl•

was

Figure 50. Location of the Upper San Pedro River Basin.

207

208

1973; Konieczki, 1980; Usunoff, 1984) has revealed that, hydrochemically, the following processes are active in both aquifers:

(1) dissolution of gypsum-carrying sediments along the flow direction, with a considerable increase in 50

4

= and

Ca++ concentrations;

(2) increase of HCO

3

-

concentrations, correlated with a possible dissolution of carbonatic sediments;

(3) ion exchange, favored by the existence of clay lenses, with softening of the ground water (CW+ retained, Na+ released); and

(4) F

-

concentrations in the area are generally well above the recommended limits, and their disLiibution is rather erratic. Such a puzzling characteristic might be explained by assuming the presence of F-carrying minerals (chiefly fluorite, CaF

2

) randomly distributed in the sediments and/or the very localized changes on some determinant chemical parameter (pH, for example).

In order to verify the above findings and, if possible, to better explain them, the application of factor analysis and correspondence analysis was chosen. The original data matrix consisted of 49 samples from both aquifers and the river, each sample characterized by

12 measured chemical parameters (concentrations of Ca++,

mg-,

Na+, K.+,

Si0

2

, Cl

-

, SO

4

=, HCO

3

-

, F', and Et

-

, pH, and Electrical Conductivity).

The data base used in this study is presented in Appendix G (Table G-

1), as well as a table with the statistics of the variables measured

209

(Table G-2) and two trilinear diagrams (Figures G-1 and G-2). Factor analysis was carried out by the use of the standard SPSS (Statistical

Package for the Social Sciences) available through the University of

Arizona Computer Center. The correspondence analysis used a computer code generously made available by Dr. Donald !tiers (University of

Arizona, Tucson).

Factor Analysis: R-Mode

Samples from the unconfined aquifer OU group) and the confined aquifer (C group) were treated separately in an attempt to isolate those hydrochemical characteristics pertaining to each aquifer. The first step was the calculation of the correlation matrices, which are shown in Tables G-3 and G-4 (Appendix G). Once the first factor was identified, the computation proceeded with Varimax rotation and Kayser normalization (Kayser, 1958). The percentage of variance that each factor is able to explain is presented in Tables 11 and 12. Only the three first factors have eigenvalues greater than 1, so the system was thought to be adequately represented by considPration of factors I, II, and III (in the case of the confined aquifer, four factors could have been considPred, but that would have made the graphical interpretation very difficult). The factor matrix (loadings) for three factors is shown in Tables 13 and 14.

The second step consisted of the evaluation of the completeness of the analysis (Harman, 1976) based on the three-factor model.

Table 11. Percentage of variation explained by factors, unconfined aquifer samples.

3

4

1

2

5

6

9

10

11

12

7

8

Factor

Eigenvalue

7.06

1.85

1.06

0.86

0.61

0.25

0.17

0.08

0.03

0.02

0.01

0.00

Percentage of

Variation

58.8

15.4

8.8

7.2

5.1

2.1

1.4

0.7

0.3

0.2

0.1

0.0

Cumulative

Percentage

58.8

74.2

83.1

90.2

95.3

97.4

98.8

99.5

99.8

99.9

100.0

100.0

210

Table 12. Percentage of variation explained by factors, confined aquifer samples.

Factor

10

11

12

8

9

5

6

7

3

4

1

2

Eigenvalue

4.25

2.33

1.61

1.14

0.93

0.70

0.44

0.23

0.17

0.13

0.06

0.02

Percentage of

Variation

35.4

19.5

13.4

9.5

7.7

5.8

3.6

1.9

1.4

1.0

0.5

0.2

Cumulative

Percentage

35.4

54.9

68.3

77.8

85.5

91.3

94.9

96.8

98.2

99.3

99.8

100.0

Table 13. Factor loadings for three factors, unconfined aquifer samples.

Variable Factor I

Ca"

Mg

++

Na r

Si0

2

Cl

-

SO

4

=

HCC

3

-

F

-

Br

pH

Conductivity

0.61352

0.75283

0.94439

0.51442

0.02239

0.82382

0.89149

0.65985

-0.10039

0.92883

-0.25647

0.87474

Factor II

0.32950

0.45256

0.32025

-0.16559

0.82439

0.08491

0.19635

0.63498

0.11032

0.14136

-0.90033

0.44643

Factor III

-0.10480

-0.18404

0.07472

0.09879

0.07316

-0.16689

-0.01315

-0.09212

0.92062

-0.24870

-0.09458

-0.04360

Table 14. Factor loadings for three factors, confined aquifer samples.

Variable Factor I

CW+

Mg*+

Na

F:*

SiOs

Cl

-

SO

4

=

HCOs

-

F

-

Br

pH

Conductivity

0.61874

0.81210

0.18993

-0.04818

0.46339

-0.04114

0.27718

0.83774

-0.40710

0.27917

-0.54962

0.91065

Factor II

-0.07967

0.01671

0.75437

-0.50262

-0.31579

0.12399

0.57929

0.37650

0.62384

-0.17754

-0.02667

0.23084

Factor III

0.05723

0.39785

0.15004

-0.11856

0.05758

0.93120

0.13785

0.21245

-0.22096

0.78221

0.09088

0.15121

211

212

For the U group:

Factor I: V = 5.71275

Factor II: = 2.61399

Factor III: Ni n

= 1.26152

= 9.58826

V/n = 9.58826/12 = 0.799 ===> 80% complete

For the C group:

Factor I: V = 3.45083

Factor II: = 1.89545

Factor III: sf = 1.82451

= 7.17079

V/n = 7.17079/12 = 0.589 ===> 60% complete

The interpretation of the factors' significance must be made in terms of the squares of the factor loadings (Dawdy and Feth, 1967) shown in Tables 13 and 14.

For the C group, factor I accounts for 34.5% of the variation in the data set, factor II for 19.5%, and factor III for 13.4% (Table

12).

Table 15 shows the percentage variation of the chemical variables that each of the selected factors is able to explain. The transformation of the coordinates of the variables on the factors proposed by

Klovan (1966) was adopted, and the results were plotted in a triangular diagram (Figure 51), which constitutes a good way of visualizing the grouping of variables.

Regarding the ions, inspection of factor I shows that most of the variance in the system may be accounted for by variations in the concentrations of bicarbonate, magnesium, and calcium, which would be related to dissolution of carbonates contained in the sediments. On

Table

15. Percentage variation of the chemical variables explained by three factors, confined aquifer samples.

Chemical Percentage variation explained by

Total

% variation

Variable Factor I Factor II Factor III factors account for

Ca"

M9

+

+

Na

K+

Si0

Cl

-

SO

4

2

=

HCO

3

-

F

-

BrpH

38

66

4

0

21

0

8

70

17

8

30

Cbnductivity 83

10

2

34

14

57

25

1

0

39

3

0

5

0

87

2

61

1

5

5

2

0

16

2

1

31

89

44

89

39

82

63

26

61

72

31

90

213

214

215 the other hand, calcium is negatively correlated with fluoride, which might indicate that both species are controlled by equilibrium with fluorite in the aquifer. This explanation follows the suggestion made by Dawdy and Feth (1967) as to how the presence of mutually-exclusive components can be interpreted. Variations of the concentrations of the main chemical species make the conductivity show a great deal of variation, most of which factor I explains.

Factor II is dominated by variations in the concentrations of sodium, potassium, sulfate, and fluoride. This strengthens the suggestion that dissolution of gypsum followed by ion exchange produces the variation pattern observed. Surprisingly enough, calcium is of little importance in this factor. Notice that the square of the factor loading for sulfate (0.57959

2

= 0.33557) approaches the same value for fluoride (0.62384

2

= 0.38917). This almost identical loading of the two species may be related to the ability of sulfate to keep fluoride in solution suggested by Bhakuni and Sastry (1977).

Factor III is associated with bromide and chloride, which could be indicative of their common source (input by precipitations?).

For the U group, factor I explains 58.8% of the variation, factor II 15.4%. and factor III 8.8% (Table 11). Table 16 presents the percentage variation of the chemical variables explained by each factor. A triangular diagram (Figure 52) with modified coordinates

(Klovan, 1966) displays the clustering of variables.

Factor I explains most of the variance observed in the system, although simultaneous variations of several chemical parameters make the analysis difficult. The squares of the factor loadings for calcium

Table

16.

Percentage variation of the chemical variables explained by three factors, unconfined aquifer samples.

Chemical Percentage variation explained by

Total

% variation

Variable Factor I Factor II Factor III factors account for

Ca"

Mg

+

+

Na

,

r

SiO

2

Cl

-

SO

4

-

68

79

44

HCI0

3

-

r

BrpH

1

86

7

Conductivity 77

37

57

89

26

0

11

20

10

3

68

1

4

40

81

20

1

2

86

6

1

0

1

1

1

3

1

3

0

1

49

80

100

30

69

72

83

85

88

94

89

97

216

i!

217

218

(0.61352

2

= 0.37641) and for bicarbonate (0.65985

2

= 0.43540) are very similar, which once again is suggestive of dissolution of carbonates contained In the sediments. The squared loadings for sodium (0.94439

2

= 0.89187) and for chloride (0.823822 = 0.67868) also yield reasonably close numbers, which might point to a common source. In the absence of evaporite formations, such a common source would be the leaching of the saline residues left in the upper layers of the soil by successive cycles of irrigation-evaporation. The variation of the sulfate concentrations, mostly explained by factor I, does not appear to be correlated with those of other species. Fertilization practices could be causing that effect because, if ammonium sulfate is being used as reported, the sulfate content would show a great deal of variation that does not necessarily correlate with variations of the other compounds considered in this study. As for the C group, factor I explains most of the variation of the conductivity measurements.

Factor II is mainly associated with variations of pH and silica. The narrow range of variation in the concentration of those species, however, does not lend itself to a clear interpretation.

Factor III is almost exclusively related to fluoride, which constitutes a clear sign that the variation of [r] is independent of other constituents in the system (or, as said at the beginning of this chapter, related to variables not included in the analysis).

The absence of numerically-comparable, mutually-exclusive components at a given factor (Table 13) suggests that none of the species is strictly controlled by equilibrium with minerals in the unconfined aquifer.

219

Factor Analysis: QI-Mbde

As presented before, although conceptually similar to the Rmode, the analysis takes into account a correlation matrix based on comparisons between sample sites instead of comparisons between chemical variables. Each sample site is described by the twelve chemical measurements taken. Besides the samples belonging to the confined and unconfined aquifers, samples from the San Pedro River and a sample from the MbGrew Spring were included.

Factor I accounts for 96.6% of the variance in the system. If factor II is included in the interpretation, such a percentage improves to 98.8%.

The factor loadings for three factors are given in Table 17, whereas Figure 53 presents the distribution of the sampling sites on a plane formed by factors I and II. Figure 53 reveals that water samples from the confined aquifer taken mostly in the St. David area have high loadings on factor I and medium loadings on factor II. On the other hand, samples from the unconfined aquifer at the Benson-Pcmerene area and from the San Pedro River tend to plot at higher factor II and medium factor I coordinates. This particular distribution does not coincide with the distribution of the two main types of water identified, namely, the bicarbonate-calcium and the bicarbonate-sodium types of water. Wells which withdraw water from more than one waterbearing unit and a certain degree of water mixing due to upward flow from the confined aquifer are believed to be the cause of such absence of correlation. The separation achieved in Figure 53, however, is useful in that it helps verify the difference in chemical behavior as

Table 17. Factor loadings for three factors, Ql-mode factor analysis.

Sample

Site

Factor

I

Factor

II

Factor

III

44

45

46

47

3

6

13

14

17

18

8

10

11

15

16

4

5

1

2

7

19

24

27

29

42

0.996

0.982

0.963

0.979

0.987

-0.065

0.058

0.259

0.198

-0.129

-0.007

0.147

0.443

0.013

0.001

0.987

0.151

-0.025

0.981

0.135

-0.125

0.995

-0.061

0.005

0.962

0.067

0.056

0.989

0.029

0.095

0.930

0.361

0.023

0.995

-0.048

-0.033

0.990

0.105

0.057

0.993

0.099

-0.031

0.996

-0.034

-0.023

0.961

0.262

0.050

0.987

0.121

-0.049

0.954

-0.285

0.073

0.923

0.371

0.104

0.998

0.198

-0.053

0.998

-0.006

0.048

0.991

0.088

-0.009

0.992

-0.085

0.058

0.995

-0.010

-0.104

0.991

-0.123

0.016

33

34

35

36

37

26

28

30

31

32

20

21

22

23

25

38

39

40

41

43

48

49

Sample

Site

Factor

I

Factor

II

Factor

III

0.986

-0.094

0.058

0.996

-0.062

-0.063

0.993

-0.070

-0.049

0.994

-0.101

-0.002

0.989

-0.134

-0.023

0.994

-0.105

-0.030

0.993

-0.113

-0.029

0.993

-0.110

0.021

0.991

-0.065

-0.025

0.983

0.358

-0.203

0.991

-0.130

-0.011

0.993

-0.101

-0.040

0.996

-0.083

-0.034

0.992

-0.126

-0.031

0.991

-0.126

-0.031

0.997

-0.049

-0.019

0.989

-0.118

-0.012

0.989

-0.102

-0.075

0.916

-0.025

0.335

0.997

-0.064

0.048

0.992

-0.129

0.013

0.996

-0.041

-0.013

220

.9 —

0

x o

.8 -

7 .

0 .

6

0

0 o

o

0

O

X x x

Rox„

sample 41 x

.5 •

.4 sample

46 x

3

.4 .5 .6

FACTOR I

.8 x: CONFINED AQUIFER SAMPLE o: UNCONFINED AQUIFER SAMPLE

4

3:

SAN PEDRO

RIVER SAMPLE

•: SPRING SAMPLE

.g

Figure 53. Sampling sites distribution on a factor I vs. factor II space.

4

221

222 given by the samples from the confined and unconfined aquifers, which would arise from distinct hydrochemical mechanisms acting in the aquifers, or even from the degree of intensity with which such processes are operating.

If three factors are taken into considPration, the percentage of explained variation is improved to 99.5%. A triangular diagram has been prepared (Figure 54) by using the system of coordinates advocated by Klovan (1966) in order to emphasize the separation of sample sites.

Figures 53 and 54 suggest that samples 41 and 46 are unique in their variance. The reason for their individuality was not identified.

Correspondence Analysis

The input data for the computer program that carry out the correspondence analysis were the 47 available samples (samples 9 and 12 were not included), each of them described by the twelve parameters measured. Again, the program was run separately for the C and U groups.

For the C group, three factors accounted for 86.8% of the variance in the data set. It should be mentioned that, due to the normalization done in correspondence analysis, the factors are not necessarily the same as those extracted through factor analysis. Table

18 presents the coordinates of the variables in the three factors, whereas Table 19 lists the absolute and relative contributions. Tables

20 and 21 are equivalent to Tables 18 and 19, respectively, but in this case, the sample sites replace the chemical parameters in the analysis.

223

224

Table 18. Coordinates of chemical variables on three factors, confined aquifer samples.

Variable Factor I pH

Ca"

Mg

++

Na+

K+

Si0

2

Cl

-

SO

4

HCO

13

= r

Br

-

-

Conductivity

0.09512

0.15780

0.09705

-0.38968

0.00249

0.11240

0.01436

-1.09110

0.03118

-0.12566

0.11569

0.04992

Factor II

-0.11285

0.31305

0.26120

-0.28261

-0.14701

-0.04117

-0.05552

0.32672

-0.05335

-0.21859

0.09981

0.02339

Factor I

-0.23834

-0.10043

0.39465

0.03264

-0.24725

-0.21433

-0.18329

-0.06052

0.00585

-0.41324

0.04023

0.03697

Table 19. Per cent absolute (AC) and relative (RC) contributions for three factors, variables: chemical parameters, confined aquifer samples.

Variable Weight AC(I) RC(I) AC(II) RC(II) AC(III)

RC(III) pH

Ca"

Mg"

Na+

IC

S

10

2

0.013

0.049

0.008

0.3

11.5

1.2

2.9

18.7

33.8

0.2

4.0

4.0

0.057

20.9

65.2

32.3

0.003

0.052

0.0

0.0

1.6

21.0

0.5

0.6

Cl

-

S O

4

=

HCO

F

3

-

0.013

0.024

0.286

0.003

Br

-

0.000

Conduct.

0.492

0.0

70.3

0.7

0.1

0.0

3.0

0.6

91.5

25.2

6.7

53.6

56.6

0.3

18.5

5.8

1.2

0.0

1.9

16.2

73.7

29.2

34.3

26.1

2.8

8.4

8.2

73.9

20.4

39.9

12.4

10.5

7.1

18.5

0.9

2.7

34.4

6.3

1.3

0.1

8.5

0.0

9.8

72.3

7.6

66.7

0.5

73.9

76.2

91.1

0.3

0.9

72.9

6.5

31.0

Table 20. Coordinates of sampling sites on three factors, confined aquifer samples.

Sample

37

38

39

40

41

32

33

34

35

36

43

46

48

49

18

20

21

22

23

13

14

3

6

17

25

26

28

30

31

Factor I Factor II Factor III

0.04862

-0.20534

-0.19964

-0.08802

0.19691

0.01407

-0.15827

0.04370

0.15607

0.05867

0.17863

0.13242

0.12886

0.44500

0.11187

-0.03419

0.11272

0.13542

0.11567

0.25895

0.16545

-0.02754

0.08644

0.14783

-0.84268

-0.12099

0.08614

0.07323

-0.03476

0.09149

0.06277

0.33601

-0.14955

0.06434

-0.11019

-0.17580

0.05232

0.14800

0.05197

0.05971

-0.02831

-0.04855

-0.07068

-0.04125

-0.07469

-0.12861

-0.06087

-0.05267

0.09078

-0.00971

-0.04356

0.00354

-0.04621

-0.06086

0.06657

-0.19363

-0.01275

-0.08016

0.10329

0.01050

-0.00433

0.06386

0.04506

0.05855

0.17685

-0.02530

0.02904

0.03635

0.02891

-0.08287

-0.10052

0.69400

0.00490

-0.08417

-0.08428

-0.10211

-0.06870

-0.14825

-0.07014

0.05738

0.06782

-0.09193

-0.12170

-0.00316

-0.19000

0.03075

-0.13870

225

226

Table 21. Per cent absolute (AC) and relative (RC) contributions for three factors, variables: sample sites, confined aquifer samples.

Sample

18

20

21

22

23

13

14

3

6

17

32

33

34

35

36

25

26

28

30

31

43

46

48

49

37

38

39

40

41

Weight AC(I) RC(I) AC(II) RC(II)

AC(III) RC(III)

0.040

0.046

0.057

0.038

0.032

0.041

0.055

0.035

0.042

0.042

0.2

11.0

4.7

91.2

2.4

1.3

5.6

26.1

46.1

0.7

22.7

6.0

3.0

86.3

0.9

0.0

1.3

3.6

3.3

28.7

12.1

0.2

36.1

2.5

51.7

0.4

46.1

0.7

6.6

0.8

1.0

0.2

0.4

1.5

0.3

0.040

0.027

0.026

0.042

0.025

0.025

0.025

0.027

0.027

0.022

3.1

0.8

87.9

1.1

69.6

1.0

57.1

0.2

16.8

87.9

0.1

8.5

0.8

35.0

1.2

56.5

0.9

64.1

3.6

68.9

0.030

0.038

0.037

0.026

2.0

0.1

0.7

1.4

84.5

12.8

61.8

67.4

0.035

60.3

97.5

1.0

2.9

0.7

0.5

1.3

0.0

0.5

0.0

0.4

0.9

0.041

0.021

0.041

0.020

1.5

76.7

0.4

9.2

0.5

82.9

0.1

4.5

1.3

5.6

0.1

0.9

49.9

0.2

0.0

11.9

4.5

21.7

35.9

12.1

1.8

17.7

2.3

27.2

34.8

40.9

0.2

51.2

19.5

32.1

22.6

22.6

15.2

55.3

38.1

26.1

2.0

0.1

44.6

14.6

71.6

2.1

1.8

2.5

3.2

7.5

0.0

11.0

0.6

5.5

0.5

2.7

3.8

2.9

0.0

2.5

2.6

4.0

1.8

7.0

6.2

0.1

0.0

2.2

0.9

2.1

24.9

0.3

0.5

0.8

40.3

45.5

11.4

13.3

8.5

0.3

31.9

0.1

6.6

0.5

23.2

46.3

2.5

23.9

39.1

8.5

73.9

65.4

9.2

77.0

35.4

51.8

46.5

36.2

9.8

3.2

8.1

42.4

12.0

227

The representativity of the factors, that is, the percentage variation that factors are able to explain, is in indication of the magnitude of such variations. Thus, factor I explains regional variations, whereas factors II and III explain more localized variations, or variations attributed to local causes (David et al., 1977). Of course, terms such as regional or local have to be related to the scale of the area investigated.

Table 19 shows that sulfate, sodium, and calcium are "responsible" for the appearance of factor I (the sum of their absolute contributions is 94.17%), whereas Table 18 reveals that sodium and sulfate are associated (same sign) and both opposed to calcium. This might be interpreted as a regional hydrochemical process which involves an increase (decrease) in both sodium and sulfate concentrations and a parallel decrease (increase) of calcium content, which is in accordance to what has been suggested to happen In the confined aquifer; that is, dissolution of gypsum followed by ion exchange reactions that soften the ground water by releasing sodium ions and trapping calcium ions.

Mast of the samples with significant absolute contributions in factor I (Table 21) are from wells in the Pomerene area and close to the mountains, which confirms the suggestion that had been made

(Usunoff, 1984) about a distinct chemical behavior of waters north of

St. David.

Following the same interpretation guidelines, factor II reveals characteristics similar to those of factor I, although in this case, sulfate is associated with calcium, and both ions are opposed to sodium. Opposition of calcium and fluoride would indicate that con-

228 centrations of both ions are governed by equilibrium with minerals in the aquifer (fluorite?). However, and according to data in Table 19, the absolute contributions of calcium and fluoride on factor II are not comparable. That may mean that, while the equilibrium hypothesis may still hold, calcium variations are mostly explained by other phenomena

(notice, for example, that the absolute contributions of calcium and sodium on factor II are very similar and that those species are in opposition, which points directly to the ion-exchange hypothesis).

Factor III explains very localized variations which occur mostly in the St. David area. Again, fluoride is found in association with calcium, both ions having about the same absolute contributions.

This adds more evidence to the hypothesis that the incorporation of fluorides to the ground-water flow is related to the particular characteristics of the confined aquifer in and around St. David (local lithology). A more general conclusion regarding the geochemistry of F in the area would be that the observed F

-

concentrations and their distribution cannot be related directly to the hypothetical presence of fluorite homogeneously distributed in sediments. The lack of a clear statistical correlation may prove that the variation of [F

-

] is related to variables not included in the study, and/or that the mineral

(fluorite) has a random, pocket-like distribution in the sediments.

This latter possibility is attractive in the sense that it would explain the shape of the F

-

variograms for this area (Carrera et al.,

1984; Carrera-Ramirez and Samper Calvete, 1985).

Inasmuch as factors I and II account for 77.15% of the variance in the data set, the clustering of chemical variables is assumed to be

229 properly visualized in the factor space of factor I vs. factor II

(Figure 55).

For the U group, three factors explain 89.60% of the overall variance. Table 22 presents the coordinates of the chemical variables in the three factors considered, whereas Table 23 lists the absolute and relative contributions. Likewise, the coordinates of the sampling sites on the three factors are listed in Table 24, and the absolute and relative contributions for the sample-focused analysis are found in

Table 25.

By adding up the corresponding absolute contributions from

Table 23, sulfate and bicarbonate are seen to contribute 82.75% to the appearance of factor I, both ions being in opposition (Table 22), which would correspond to a case in which the regional hydrochemistry is controlled by equilibrium with gypsum and calcite (common-ion effect).

The data in Table 22 also indicate that fluoride and bicarbonate concentrations vary concomitantly, which would constitute an expression of the chemical equilibrium involving those ions (Banda, 1975) as presented earlier in Chapter 2. As given by Table 25, a group of selected samples (1, 10, 11, 19, 24, and 47) have greatly contributed to factor 1. Those samples are from wells which are along the San

Pedro River near the mountains and close to Pomerene.

Factor II is 88.62% (Table 23) dominated by variations in opposition of the calcium and sodium concentrations (Table 22). Ionexchange reactions may provide an explanation for such behavior.

Samples which load heavier on factor II correspond to wells spread out over the whole study area (Table 25).

230

Table

22. Coordinates of chemical variables on three factors, unconfined aquifer samples.

Variable Factor I

PH

Ca"

Mg

Na

++

K*

Si0

2

Cl

-

50

4

0.56925

-0.03465

0.10528

-0.18132

0.41486

0.48564

0.19499

-0.59270

HC01

3

-

F

-

0.32186

0.53574

Et

-

-0.02662

Conductivity -0.02317

Factor II

-0.10972

-0.39851

0.14715

0.28583

0.12316

0.05841

0.19444

-0.07152

-0.01384

0.03687

0.21781

0.00707

Factor I

0.39021

-0.05129

0.09492

-0.07059

0.39215

0.12305

0.30314

0.04055

-0.04368

0.25027

0.20893

-0.00024

231

232

Table 23. Per cent absolute (AC) and relative (RC) contributions for three factors, variables: chemical parameters, unconfined aquifer samples.

Variable Weight AC(I) RC(I) AC(II) RC(II) AC(III) RC(III) pH

Ca"

0.006

0.041

3.0

0.1

66.4

0.7

0.5

48.1

Mg++

Na

0.013

0.068

0.2

26.6

K*

0.002

0.4

50.5

Si0

Cl

SO

-

4

HCCs

F

-

2

-

0.023

0.014

8.2

92.7

0.8

22.7

0.096

51.7

98.1

0.196

31.1

98.0

0.001

0.6

81.8

Br

-

0.000

Conduct.

0.539

0.0

0.4

0.8

91.5

2.1

3.4

27.5

40.5

0.2

0.6

3.9

3.6

0.3

0.0

0.1

0.2

2.5

97.6

51.9

68.3

4.5

1.3

22.5

1.4

0.2

0.4

51.7

8.5

23.1

2.7

3.0

8.4

6.4

8.7

32.1

4.0

9.4

2.2

0.2

0.0

31.2

1.6

21.6

4.2

45.1

6.0

54.8

0.5

1.8

17.8

47.6

0.0

Table 24. Coordinates of sampling sites on three factors, unconfined aquifer samples.

Sample

15

16

19

24

10

11

1

8

27

29

42

45

47

Factor I

Factor II

Factor III

0.36682

0.14800

0.42496

0.31752

-0.14068

0.18940

-0.26681

0.41328

0.06684

0.18810

0.23991

0.16266

-0.31501

-0.07103

0.02021

0.02763

-0.16332

0.03376

-0.26219

-0.17237

0.11224

0.11770

-0. 04291

0.02696

0.14060

0.05918

-0.03927

0.18946

0.00090

-0.05854

0.00533

-0.01551

-0.00884

0.17337

-0.09421

0.12862

-0.00386

-0.03368

0.01642

233

Table

25.

Per cent absolute (AC) and relative (RC) contributions for three factors, variables: sample sites, unconfined aquifer samples.

Sample

15

16

19

24

27

29

42

45

47

1

8

10

11

Weight

AC(I) RC(I) AC(II) RC(II) AC(III) RC(III)

0.045

0.054

0.120

9.2

95.3

0.029

1.0

37.6

0.040

11.0

99.6

1.7

0.1

0.2

8.2

77.0

10.5

3.6

94.4

1.0

0.048

2.7

34.2

24.3

0.130

14.1

70.5

28.2

0.028

0.105

0.046

7.2

80.0

2.5

0.7

16.4

10.7

2.5

65.8

0.6

0.044

0.100

3.9

98.7

0.210

31.9

96.3

0.2

4.1

55.9

14.5

5.4

3.6

0.7

0.4

20.4

5.4

65.6

29.4

5.9

50.9

3.4

1.3

41.7

3.4

1.7

25.6

0.0

4.6

0.1

0.3

0.3

20.7

23.4

19.1

0.0

2.8

1.4

1.1

61.7

0.0

2.6

0.1

0.2

0.1

14.1

32.6

30.8

0.0

2.4

0.3

234

The analysis of the information provided by factor III did not lead to clear conclusions, which may mean that very localized features are masked by major hyarchemical processes acting at a larger scale.

Out of 89.60% variation explained by three factors, the two first factors account for 85.29% of variance explained. Therefore, the plot of the chemical variables on a factor I-factor II plane can produce a clear picture of the distribution and relationships among variables. Such a plot is presented in Figure 56.

Example 2: The San Bernardino Valley (California)

In order to demonstrate the application of the described statistical-based techniques to the study of areas with high [F

-

] in their ground waters, the Highland-East Highland area of the San

Bernardino County, California, was selected. The reasons for having selected that particular area are that:

(1) the number of wells sampled (54) is large enough to produce statistically meaningful results, and

(2) the chemical analyses can be considered as complete

(Davis, 1987).

Studies of this area started in 1976 sponsored by the U. S.

Geological Survey and the San Bernardino Valley Municipal Water

District in response to the concern created by the high [F

-

] and

[NO

3

-

] found in local aquifers. The results of the first phase of the investigation, which dealt primarily with the distribution of dissolved

NO

3

-

and F

-

in waters from the saturated zone, were presented by Eccles

4. •

1.

u

L.

L.

CD

2*

H

1

tr

.

)

235

236 and Klein (1978). What follows is a description of the principal hydrogeological and water quality characteristics of the area as given in the referenced research report.

The study area (Figure 57) is in the east end of the San

Bernardino Valley. The boundaries of the triangle-shaped area are the base of the San Bernardino Mbuntains (which corresponds to the trace of the San Andreas fault zone) for the north boundary; an east-west line through the Santa Ana Wash for the south boundary; and a north-south line approximately 1 1/2 miles east of San Bernardino for the west boundary.

The aquifers in the area are mostly unconfined and in alluvial valley-fill deposits composed of gravel, sand, silt, and boulders. The predominant ground-water gradient is westerly in the eastern part of the study area and southerly near the western margin. The water level near the east end of the study area and near the northwest corner is at a depth of 200 feet. Near the southwest corner, it is only about 100 feet deep, and in the central-western part it is 100-200 feet. Waterlevel discontinuities due to faulting are not noticeable, but the existence of two major faults in the area may be significant with respect to water quality.

As for the main purpose of the study, Eccles and Klein (1978, p. 26) summarized its results in the following terms:

"Increasing urbanization in the Highland-East Highland area of San Bernardino Valley has placed an increased demand on the ground-water resources for public supplies. Unfortunately, the water pumped from some wells used for public supplies contains concentrations of nitrate-nitrogen that exceed the limit of 10 mg/1 reoannerNied by the U.S. andiarriental Protection Agency

e m

-

V

..::-

0

.3$

• •

..

.

n

111.... $

.

.:..:.

,

4

44

.

'

, ‘

/: /

I/.

,= 4

Si

//

//

It

.

/

i

A

/ 7

.".

..0 /

''

1

VOS4

a.

AV

/

//I ay

WAt

- /

7

'''4 0111:wyz41/21/2=6.10,// -

/

AY

A,

J

.

/

I..

0

I=

a

cn

0

a

< z ec c c z

..

<

lb

n

Y

W

44.

J

,e•

A

J "t

VS

<

>

0

z

cr

< z m w o

Z

< o ,

I i

o

v

IV

.....

0—

.

.

...)

0 g im -

In tu _ i

237

238 and remedial action is required. One form of action commonly used consists of blending high-nitrate water with low-nitrate water before distribution; however, in the study area, some of the low-nitrate water has a fluoride concentration that exceeds the California

Department of Health's optimum dissolved-fluoride concentration of 0.8 mg/l.

In general, nitrate-nitrogen concentrations exceeding

10 mg/1 are near the center of the study area and decrease toward the perimeter. Concentrations of dissolved nitrate-nitrogen exceeding 20 mg/1 occur in water from four public-supply wells. Wells drilled near areas of recharge, such as the Santa Ana Wash, generally yield water with dissolved nitrate-nitrogen concentrations less than 5 mg/l. Data are insufficient to establish a continuous relation between dissolved nitrate-nitrogen and depth; however, plots ... indicate that the concentrations are substantially less at depths greater than 500 feet.

Concentrations of dissolved fluoride in water from wells show an increase with depth in some places.

Shallow wells drilled near faults may intercept water moving up from depth and also yield water with excessive fluoride concentrations. Wells north of Base Line

Road drilled to depths greater than 340 feet will generally yield water with concentrations of dissolved fluoride exceeding 1 mg/l. The maximum observed fluoride concentration was 6.4 mg/l. Water from wells to the south near the Santa Ana Wash shows the effects of dilution due to recharge with water having dissolved-fluoride concentrations generally less than 0.4

mg/l."

Chemical Data Base

As presented in Eccles and Klein (1978), samples from 54 wells were analyzed by the U.S. Geological Survey and the Babcock Laboratories (Riverside, California). As presented in Table G-5 (Appendix

G), each sample is characterized by the depth of the sampled well, temperature, pH, specific conductance (SP), and the concentrations of

239 silica, iron, calcium, magnesium, sodium, potassium, bicarbonate, sulfate, chloride, fluoride, nitrate, and boron.

Before undertaking the statistical analysis, the quality of the data was checked by means of a series of control tests. A computer code (CHEMAN; Utunoff, 1987) specifically designed for this purpose was used. Such a program does the cation-anion balance for each sample as well as the analytical error evaluation based on a criterion described in CUstodio and Llamas (1976, p. 233). CHEMAN is not capable of incorporating in the calculations the concentrations of species such as

F

-

, Fe

2

+, and BP+ (all of them part of the data), so the decision as to whether to accept or reject a given chemical analysis based on CHEMAN's output was somewhat flexibilized. As a result of the check by CHEMAN, four samples (38, 43, 47, and 53) appeared to have large analytical errors, so their analyses were not included in the statistical study.

The fact that some of the variables had not been measured at all sites created additional problems as to the way that the data matrix is handled by the program which does factor analysis. In order to cope with this problem, several possibilities were tried out, namely,

(1) samples containing missing values were deleted. It resulted in a matrix ill-conditioned for factor analysis, which means that the results obtained would not be as reliable as when more samples were considered.

(2) missing values were substituted by the variable mean.

This was not actually considered a proper alternative

because, in order to do so, one has to have independent evidence that such an action is justifiable by hydrological and hydrochemical data.

(3) cases with missing values were deleted pairwise, that is, all cases with non-missing values for each pair of variables correlated were used to compute that correlation, regardless of whether the cases had missing values on any other variable (SPSS Inc., 1986).

Although this option gave better results than (1), additional tests of sampling adequacy (SPSS Inc., 1986) indicated that the use of factor analysis had to be reconsidered.

(4) variables with missing entries were deleted. Although this meant eliminating four variables from the data base (depth, temperature, and the concentrations of iron and boron), this alternative yielded the best results in terms of the appropriateness of the reduced data matrix to be treated by factor analysis.

240

In summary, the data used for factor analysis consisted of a matrix similar to that presented in Table G-5 (Appendix G), in which samples 38, 43, 47, and 53, and the columns of well depth, temperature, iron, and boron concentrations were disregarded. The main statistical descriptors of this new data base are shown in Table G-6 (Appendix G).

The Statistical Package for the Social Sciences (SPSS) in its PC

241 version was used, made generously available to the author by Dr.

Soroosh Sorooshian (Ublversity of Arizona).

Factor Analysis: R-Mode

The first step consisted of the calculation of the correlation matrix, i.e., the array of correlation coefficients for all possible pairs of variables (Table G-7, Appendix G). Inspection of the F

-

row in Table G-7 indicates that [F

-

] are significantly correlated with

[Ca], [r], [SO

4

=], and [Cl

].

Based on the correlation matrix, factor analysis then proceeded with the calculation of the variance explained by each factor, which is quantified by means of the eigenvalues. It should be recalled at this point that, before choosing an appropriate number of factors, the method takes into consideration as many factors as variables are in the original data matrix. Strictly speaking, 100% of the variance in the system will be explained using only a number of factors equal to the number of variables. The graphical display of the amount of variance

(eigenvalue) that each factor accounts for is what is known as "scree" plot (Figure 58). As explained by SPSS Inc. (1986, p. B-47), "typically, the plot shows a distinct break between the steep slope of the large factors and the gradual tailing off of the rest of the factors.

This gradual tailing is called the "scree" because it resembles the rubble that forms at the foot of a mountain. Experimental evidence indicates that the scree begins at the kt

h

factor, where k is the true number of factors." From Figure 58, it appears that a three-factor model should be sufficient for the study. Another way of explaining

4.0-

a

1•I

1.0-

a

• •

• •

I l

2. 3 4

Factor number

• •

I

G

1

7 (3 9

6

1 0

II 12

Figure 58.

Scree plot.

242

243 the criterion for deciding on the number of factors to be used is by recalling that, before calculation of the eigencvalues, the variables are standardized, that is, they are transformed so that their mean is zero and their standard deviation unity. Therefore, any factor whose eigenvalue is less than one is no better than any given variable.

In the rotation stage (done by principal components analysis), several pieces of information were extracted such as the amount of total variance explained by three factors (Table 26), the oanmunalities of the variables (Table 27), and the loadings of the variables in each of the three factors (Table 28).

The paper that contains the data used in this example (Eccles and Klein, 1978) offers only a slight interpretation of the hydrochemistry of the study area, so the conclusions drawn from the statistical study cannot actually be independently verified by existing information. On the other hand, the lack of time and other hydrological data impeded my carrying out a complete study under a more classical approach. I have, therefore, chosen to present the results of the application of FA and to add a few comments as to what they might signify, although the interpretation is probably neither complete nor accurate.

The rather high communalities of variables such as calcium, magnesium, sodium, bicarbonate, sulfate, chloride, fluoride, and SP

(specific conductance) (Table 27) indicate that their variations can be explained by linear combinations among them. This, plus the fact that three factors explain about 73% of the total variance in the system,

Table 26. Percentage of total variation explained by three factors.

244

Factor Eigenvalue Pct. of Var. Cum. Pct.

1

2

3

4.34884

3.18421

1.19005

36.2

26.5

9.9

36.2

62.8

72.7

Table 27. Communalities of variables.

Variable

Si0

2

Ca"

Na+

HCO

3

-

50

4

Cl

-

F

-

NO

3

-

SP pH

Ccrrrnunality

.58047

.88439

.72964

.91170

.65360

.75133

.86491

.86488

.77885

.50983

.92628

.26720

Ca"

H00

3

-

Mci f

+

1\0

3

-

SP

PH

Na+

SO

4

'

Cl

-

F

-

IV

-

Si0

2

Table 28. Variable loadings on three factors after rotation.

Factor 1

.93221

.86581

.77312

.70466

.67699

-.50913

-.21371

.12283

.41323

-.40546

-.03519

.29286

Factor 2

-.11463

.00640

.01997

.06214

.66967

-.07609

.92741

.92182

.83117

.67385

.21099

-.02816

Factor 3

.04725

.04085

.36268

.09709

.13967

.04688

.07704

.00820

.05733

.40047

.77965

.70279

245

246

(Table 26) gives a positive sign as to the reliability on the threefactor model selected.

Variables with high loadings on factor I are calcium, bicarbonate, magnesium, nitrate, SP, and pH (Table 28). Notice, from data in Table 28 and the factor I vs. factor II plot of Figure 59, that calcium and bicarbonate have similar loadings and the same sign and that both ions are opposed to pH, which would be consistent with a mechanism based on non-equilibrium dissolution of carbonate mineral, favored by lowering of the pH and resulting in an overall increase in

SP.

This hypothesis is supported by the pattern shown in Figure 60, in which the [HCC1

3 -

] and [Ca++] of all samples delineate a parallel increase with plotting position mostly in the field of undersaturatian with respect to calcite. The equilibrium relationship depicted in

Figure 60 considered the following data:

CaCCI

3

+ H' <====>

Ca" + HCC1

3 -

; K = 10

1

-

97

(Hands, 1975) ion strength = 0.007 M (average over all samples) pH = 7.43 (average over all samples)

Factor II is mainly associated with variables such as sodium, sulfate, chloride, fluoride, and SP (Table 28). Unlike the lithologically-controlled factor I, factor II seems to be related to culturallyinfluenced surface runoff or waste sources (Hem, 1959). In line with this hypothesis, it should be said that Eccles and Klein (1978) suggested that the existence of high-SP spots in the area could be related directly to the fertilization of citrus groves and other agricultural lands, animal wastes, and septic tank effluent.

1.0

0.8

F

-

0.2

0 pi-1

-0.2

-0.6

-OA

-0.2

K

+

sio2

0

0.2

0.4

Factor I

(36%)

0.6

• m n

4+

`-'

HCO3--

Ca"

0.8

1.0

Figure 59. Variables an a factor I vs. factor II space

(% variance explained by factor in parenthesis).

247

o

3

o o o o o

10

1

I

I

I I

10

I

I I

I to

2

Concentration of

Ca" (ppm)

Figure 60. Equilibrium with respect to calcite

(ion strength= 0.007 M; pH= 7.43).

io

3

248

249

Factor III owes its appearance mostly to variations of IC+ and

Si0

2

concentrations (Table 28). It is not easy to find an explanation for that, although some kind of secondary lithological control is suggested.

The spatial relationship of the variables in the planes formed by the pairwise combination of the three extracted factors is shown in

Figures 59, 61, and 62.

Notice that, although preponderate on factor II, the variable

[F

-

] has loadings on the three factors (Table 28). Said in an equivalent manner, there is no reason to believe that the observed [F

-

] in the local aquifers are explainable by means of a single mechanism. The possibility of local waters being in equilibrium with fluorite was evaluated by considering the following information:

CaF

2

<====> Ca" + 2F

-

; K = 10

-1°-57

(Banda, 1975) ion strength = 0.007 M (average over all samples)

The calculated equilibrium relationship and the measured calcium and fluoride concentrations have been plotted in Figure 63. It can be appreciated that not only all samples plot in the region of undersaturation with respect to the mineral, but also that F

-

concentrations seem to vary somewhat independently of the Ca" content, at least in the range of [F

-

] between 0.1 and 2 ppm. The finding of the mineral phase and/or the processes which control [F

-

] in the aquifer would require a substantially larger amount of information, particularly in

0.8

0.7 •

0.G •

K

4 e

Si 02

6

.4". 0.5 •

o

•-•

0.4 • o

4

u la. 0.3

m

.?

4 4

0.2 •

0. I •

0

-0.G pH

-OA

Na*

-0:2

C i -

SO

4

-

Og.2

0I4

Factor I (WO

0.

6

• N 0"

3

Ca**

Htoi

1.0

Figure 61. Variables on a factor I vs. factor III space

(% variance explained by factor in parenthesis).

250

0

0.4• o

4-

Li..

0.34

0.8

0.7•

0.G•

• K

+

F

N o

0-1

0.1-

0

-0.2

C0+*pH

• • o

*HCOi

02

0.4

Factor II

(2770)

01.6

S.P

CI

_

No+

SO;

01.8 1.0

Figure 62.

Variables on a factor II vs. factor III space

(%

variance

explained by factor in parenthesis).

251

252

1

0

7

0 0

CO

0000 0

,,..

1

10

I 1111

1

J

10

2

Concentration of Ca" (ppm)

I

Figure

63.

Equilibrium with respect to fluorite (ion strenght= 0.007 M).

253 view of the hypothesis of Eccles and Klein (1978) that high [F] may be associated with deeper water moving upwards along faults in the area.

Factor Analysis: Q-Mbde

The hypothesis of mixing of waters mentioned in the paragraph above was the driving force that decided the application of the Q-mbde of FA. The program used to carry out the Q-mode of FA could not handle properly the amount of data available and the size of the data matrix had to be reduced in order to continue the analysis. Therefore, the data base for the Q-mbde analysis is the transpose of the matrix in

Table G-5 (Appendix G), in which the new variables of analysis are 45 sampling sites characterized by 10 chemical measurements. As it frequently happens in IQ-mode FA, the correlation matrix turned out to be ill-conditioned because of the large number of variables as compared to the number of observations. It was then important to resort to the application of some tests in order to decide whether the analysis may continue. One of those tests is the calculation of the so-called

Kayser-Mayer-Olkin measure of sampling adequacy or EMO (SPSS Inc.,

1986), which is an index between 0 and 1 for comparing the magnitudes of the observed correlation coefficients with the magnitude of the partial correlation coefficients. Without going into much detail, it should be sufficient to say that, as related to the possibility of using FA, measures of such an index can be characterized "in the 0.90's as marvelous, in the 0.80's as meritorious, in the 0.70's as middling, in the 0.60's as mediocre, in the 0.50's as miserable, and below 0.50

254 as unacceptable" (SPSS Inc., 1986). For the case presented here, the

KMO was 0.986, which encouraged the continuation of the analysis.

It turned out that one factor was enough to account for 99% of the variance in the systems, with all variables having carnunalities greater than 0.95. That is a clear indication that, as far as given by the variables included in the analysis, the samples come from a unique source, and hence no evidence of a possible mixing of waters exists.

On the other hand, and if the hypothesis of mixing of waters is plausible, then the results of the Ql-mode analysis indicate that the variables that characterize such a process are not included in the data matrix which was used.

255

CHAPTER 7

CONCLUSIONS AND RECOMENDATIONS

Although the conclusions drawn from the various experiments carried out have already been presented in the individual chapters, this chapter will attempt to integrate the various conclusions in order to evaluate the extent to which the objectives of the research outlined in Chapter 1 have been fulfilled. When deemed proper to do so, suggestions for the improvement of future research efforts have been made.

The adsorption study demonstrated that uptake of fluorine species (chiefly, fluoride) does occur, and that pH appears to be the variable governing the ultimate extent of the process. Slight F

adsorption by quartz sand was found, which provided data that follow the Freundlich isotherm. Desorption is completely reversible and takes place without delay, evidencing a weak bonding between the fluorine species and the surface of the silica sand grains. The rather strong dependence on pH of the adsorption process suggests that the ultimate extent of the F

-

attachment to quartz surfaces would be determined by the type of F species present, chiefly HF. The adsorption mechanism remains uncertain, although it is believed that the F

-

uptake might be the result of more than one such mechanism. Ion exchange F

-

by cur was found, but could not account for all the adsorbed F. The experiments did not incorporate surface area as a variable. Future experiments should undoubtedly incluri surface area measurements in order to determine if the F

-

uptake is indeed attributable to physical adsorp-

256 tion or if it may be explained by other mechanisms not made apparent by the experiments presented here.

Vermiculite adsorbed minute amounts of F

-

, giving data which can probably be described by the Langmuir model. Noticeable changes in pH were brought about by the interactions between the various solutions and the mineral, which might have substantially affected the adsorption process.

Of the minerals tested, kaolinite was the most effective adsorbent of F

-

. The nature of the adsorption phenomenon was not made evident by the experiments, although the data did indicate, once again, that the ion-exchange theory (F

-

by OH

-

) cannot account for the large amounts of F

-

lost from solution. Future batch experiments using kaolinite as adsorbent should include measurements cf aluminum and silica in the equilibrium solutions to assess the hypothetical formation of F-carrying phases.

The study of F

-

adsorption on minerals through properlydesigned batch tests should be continued. Data such as surface area and charge distribution of potentially-adsorbent minerals must be known in order to enhance the interpretation and to quantify (rather than qualify) the adsorption process. Careful planning and monitoring are essential if batch experiments are to draw the seemingly fine line between adsorption per se and other chemical reactions such as precipitation, dissolution, and formation of complex

phases.

The results of the kinetic study, in which the dissolution rate of fluorite as a function cf several parameters was obtained, indicated that the dissolution of CaLF

2

is characterized by a rate constant, k, in

257 the approximate range of

10

-7

to

1

0-

8

cm/sec. A functional relationship between k and the mean pore velocity was found which, coupled with the results of batch experiments that established a direct relation between k and the rate of stirring of solutions, gave a clear sign as to the important influence of the transport stage in the dissolution process. Such transport-related features, however, do not seem to constitute the only rate-limiting step. Data from batch experiments in which k variations as a function of temperature were measured concluded that reactions at the mineral surface also play an important role in constraining or bounding the dissolution process. In the light of those and other considerations, the general conclusion to be drawn is that the dissolution of fluorite appears to be influenced by both the extent of surface reaction and the transport characteristics. Changes in the concentration of the batch solutions to cover a natural range of ion strength resulted in no major variations of k, which led to the conclusion that natural changes in ion strength can probably be disregarded in studies of fluorite dissolution in aquifers. A distinct increase in k in the pH range of 5.6 to

6.7 was found, although its causes were not further investigated. This appears to be a subject which should be pursued through laboratory experiments which could inclurip

,

percolation tests at pHs varying in the expanded range of

5 to

8

(most natural waters fall within those extremes).

The kinetic information obtained from the various experiments carried out indicates that several factors can affect the dissolution rate of fluorite (velocity, pH, etc.). This means that, even assuming a homogeneous distribution of

CaF

2

in aquifers, the

[F

-

] released to

258 circulating ground waters will depend on the very local chemical environment (pH) and the many features determining the flow regime

(hydraulic gradient, porosity, etc.). Inasmuch as those parameters can exhibit a very erratic distribution in aquifers, it is virtually impossible to come up with an acceptable functional relationship to be used in modeling the F

-

distribution. It is thought that better results may be obtained if it were possible to partition the flow domain Into several portions showing a similar behavior or sharing common characteristics.

The idea is to understand the geochemistry of each of those portions and, on a later stage, to integrate the various pieces in order to build a general, full-scale explanation. Given such a conceptual approach, the application of the principles of partial equilibrium should be explored.

The actual modeling step may be lengthy and complicated. This is to be expected because, in general terms, any attempt to predict water and solutes movement will soon reveal the multitude of variables with their complex functions and interactions. The obvious result is that any useful and accurate procedure must also be of some complexity.

At any rate, the conclusions above strongly suggest that an adequate knowledge of the kinetics of the geochemical processes is needed. Detailed and informative mass-transport studies can no longer afford to neglect the time-dependent characteristics because in many cases--and this research has presented evidence in this regard--it cannot be assumed that the concentration of a given solute is determined by equilibrium with minerals containing such species, even in

259 those cases where a continuous supply of solutes is suggested by the abundance of minerals in the sediments.

The transport characteristics of F- solutions and F-dissolving solutions were investigated through column tests, in which an attempt was made to reproduce the type and relative amounts of minerals found in aquifers as well as the composition of natural waters. The behavior of a selected group of species was compared to that of the F

-

ion. F

had a tendency to breakthrough late at the various tests, with peak concentrations that exceeded the input content, possibly due to a hypothetical re-solvation (desorption?) and fast breakthrough at the top of the columns promoted by a drastic change in the composition of the solution percolating the system. The movement of F

-

was considerably retarded as compared to other species, although mass balances demonstrated that most of the injected F

-

had been recovered. A different planning strategy may be suggested for future tests. These would start by carrying out experiments at a constant-rate, continuous source and measuring the concentration of the various species when real

(steady-state) equilibrium is reached. It is thought that such information would help interpret data from later tests in which the source is changed before actual equilibrium conditions prevail.

Distilled water-flushing of columns containing small amounts of fluorite and other minerals resulted in output [F

-

] well below those values corresponding to equilibrium with the mineral. Although the amount of F

-

dissolved would make most ground waters unsafe for human consumption, the actual pace of the dissolution process does not appear to match a model based on equilibrium with the mineral, at least as

260 given by the rather short time frame of the tests carried out. In this regard, future experimental research should consider tests of longer duration in which pore velocities substantially lower than those presented here would be used.

The application of multivariate analysis techniques (factor analysis and correspondence analysis) to help define and understand the hydrochemical patterns in two areas with high [F

-

] in their ground waters had mixed results. The techniques mentioned above were able to segregate and group those variables with decisive influence in determining the overall chemical behavior, which certainly constitutes a great help in interpreting the type of modifying phenomena in aqueous environments. However, as far as the geochemistry of F in those study areas is concerned, factor and correspondence analyses provided little information. It may actually be more proper to say that multivariate analysis did not add evidence to what was already known, that is, that the geochemistry of F in those aquifers is not simple to explain. The information available on the study areas and the data from the multivariate analysis pointed to a lack of clear correlation between the variations of [r] and those of the other parameters considered. Such a lack of definite correlation patterns may well signify that, in many cases, the behavior of the F species in aquifers is related to variables which are either unknown (i.e., not measured) or not included in the final interpretation. It would be very interesting to carry out a study of the geochemistry of F in an area for which abundant background information is available. If it were possible, the interpretation would consider al least three sets of variables:

261

(1)Chemical variables: concentration of major and minor species, which would provide an idea on the relationship between the [F

-

] variations and the distributions of the ounamtration of the primary solutes, and would serve to check on the possible formation of F-carrying complex ions.

(2)Physical and geological variables: such as ground-water temperature (of influence in the value taken up by some thermodynamic quantities), type of sediments making up the aquifer and their distribution, mineralogy, depth to the production zone of wells (screened sections), presence and distribution of clayey materials.

(3) Hydrological variables: ground-water flow regime, location of sample points with respect to areas of recharge, transport, and discharge of ground waters, possible connection between distinct ground-water bearing horizons.

Given such data availability (a Utopian thought, perhaps), the study may start by using some of the statistical-based methods presented in order to obtain an early idea on the type of processes present and the variables/sectors on which a close look should be kept. On a later stage, laboratory setups can conceivably simulate natural ocandi-

262

tions, so that the field observations can be reproduced and studied. I believe that a research project which proceeds along the lines just described will be bound to produce meaningful results in terms of reducing the uncertainty usually associated with the movement and distribution of fluorine species in aquifers.

APPENDIX A

PREPARATION OF THE VARIOUS MINERALS

USED IN THE EXPERIMENTS

263

264

Sand

The quartz sand used, commercially known as bond sand, was available through the Ottawa Silica Company. In order to characterize the particle size distribution, a sieve analysis was done, the results of which are reported below and plotted in Figure A-1.

U.S. Mesh #

Mesh

Opening (mm)

%

Retained % Cumulative

40

50

70

100

140

200

0.425

0.300

0.212

0.150

0.106

0.075

2

27

40

24

6

1

2

29

69

93

99

100

FiLut

Figure A-1, the following size-characterizing parameters can be compu

ted: medium size = d

50

O.25 Mn; uniformity coefficient = d40/d90 z-1.65

According to the classification by size presented in Custodio and Llamas (1976), the sediment wculd fall close to the boundary between fine and coarse sand (International Society of Soil Sciences), or near the boundary between medium and fine sand (European classification). Visual inspection of the sand did not indicate the presence of minerals other than quartz. The grains looked clean and free of coatings. licwever, and to ensure the elimination of sub-microscopic carbonate minerals and other impurities, the sand samples were boiled

o

ts

CD CD

W

cr

03 N

I V132:1 %

0

266 in 10% HC1 for ten minutes and then washed with excess distilled water, a procedure inspired in the work by Dobrovol'skiy and Lyal'ko (1983).

Washings continued until the pH of the leachate approximately equalled that of the distilled water. The samples were then oven-dried (@ 105

°C) and kept in close containers. Later inspection under the microscope confirmed that the grains were entirely of quartz composition.

Fluorite

Fluorite from Sonora (northern Mexico) was used, available through a local merchant. The samples were clean, well crystallized, and exempt of macroscopic minerals other than fluorite. The samples were then crushed and sieved and the fraction retained between sieves

35 (0.495 mm) and 60 (0.246 mm) saved for further treatment.

As with the sand, the sieved fraction was boiled in 10% HC1 for ten minutes and washed with excess distilled water until the pH of the leachate equalled the pH of the water used for washing, which guaranteed the elimination of impurities, in particular carbonatic minerals.

Samples were then dried in an oven at 105°C and stored in isolated containers.

Visual inspection of the fragments so treated revealed a somewhat restricted variety of particle shapes. Inasmuch as the kinetic study required the previous definition of the specific surface of the mineral, s [L

2

/L

3

], three basic shapes were considered: sphere, tetrahedron, and cube. The maximum dimension of the particles was assumed to be represented by the average of the sieves openings mentioned above: (0.495 + 0.246)/2 = 0.3705 mm.

267

For the sphere-Shaped particles,

Volume = 4/3 7r

3

= 4/3 Tr(0.03705/2)

2

= 2.66 x 10r

5

cuP

Area

= 4n1

2

= 4 n

(0.03705/2)

2

= 4.31 x 10r

3

cn?

Specific surface, s s

= Area/Volume = 161.94 cmc l

For the tetrahedron-shaped particles,

Volume = a

3

V5212 = (0.03705Ni712 = 5.99 x 10

-6

ce

Area

= a

2

Nri = (0.03705)

2

E= 2.38 x 10r

3

ce

Specific surface, s t

= Area/Volume = 396.68 cmc l

Likewise, for the cube-shaped particles,

L=

M

2

= a

2

+ (a\ri)

2

Notice that the assumed maximum dimension (0.3705 mm) is M.

==> a = M/V3

-

= 0.03705/\F = 0.02139 cm

Volume

= a

3

= (0.02139)

3

= 9.79 x10

-6

ce

Area

= 6a

2

= 6(0.02139)

2

= 2.75 x10

-3

ce

Specific surface, s c

= Area/Volume = 280.50 curl

Based on visual inspection, the percentage representativity of the three shapes distinguished was determined to be as follows: spheres

20%, tetrahedrons 70%, and cubes 10%. These weighing coefficients, when applied to the figures calculated above, allowed the obtention of

268 a representative value for the specific surface of the fluorite fragments used in the study: s = (161.94) 0.2 + (396.68) 0.7 + (280.50) 0.1 = 338 curl

NOTE: As used in the various experiments, the term "specific surface" refers to the ratio between the total area of a particle and its volume. Such a relationship, although not widely used, actually constitutes one of the three ways in which the term has been defined by soil scientists (Hillel, 1982).

Kaolinite

Kaolinite samples were obtained from the Georgia Kaolin Co. and treated according to the procedure suggested by Slavek et al. (1984).

Such treatment, pioneered by Posner and

Quirk

(1964), has the double purpose of obtaining a homo-ionic clay and removing the exchangeable aluminum without causing undue damage to the clay surface. The procedure, a free adaptation of that of Posner and Quirk's, included the following steps: a sample of the clay was suspended in a 1 M NaC1 solution, and diluted HC1 was added until the pH of the suspension remained at a value of 3 for thirty minutes. At that point, the clay was already dispersed, and a 2%

suspension

in water was then prepared and its pH adjusted to 7 by using diluted NaOH. The suspension was then centrifugated at the proper velocity and time to collect the less than 2

pm

equivalent spherical diameter fraction. After addition of

NaC1 to flocculate the clay, the centrifugation was repeated twice.

269

The residue, as a

10% suspension in 1 M NaC1, was treated with diluted

HC1 until the pH remained constant for six hours at a value of 3.

After discarding the supernatant, the procedure was repeated three times. The clay was then washed five times with

IM NaC1, the fifth wash being for 72 hours. The residue was then dialyzed against eight liters of distilled water until no chloride was detected in the dialyzate, which took a total of five days and ten renovations of the distilled water

(80 liters). The clay was then oven-dried at 50

0

C and the residue lightly grounded by hand and stored In close containers.

APPENDIX B

ANALYTICAL DETE:RMINATION OF FLUORIDE

CCNCENTRATIONS AND pH

270

271

Measurement of Fluoride Concentrations

The F

-

concentrations were measured with a Fisher Accumet pH/mV meter (model 620) and a specific electrode (Orion, model 94-09).

Before measuring a given set of samples, fresh solutions of known [F

-

] were prepared to obtain the corresponding calibration curve, and a rechecking of such a curve was done between samples measurement. When the sampling lasted for more than one day, new [F

-

] standards were prepared for the electrode calibration. The preparation of standard solutions and the electrode calibration followed the guidelines given by Orion Research Inc. (1973) and by the U.S. Environmental Protection

Agency (1979).

Both samples and standard were mixed in proper amounts with

TISAB, a ccomercially-available solution designed both to buffer the solution at a pH between 5 and 6, and to provide a high total ion strength background which swamps out variations in total ion strength between samples (Orion Research Inc., 1973). Such ion-strength normalization is done so that the readings are total F

-

concentrations.

Before measuring the F

-

content of acid samples, and after addition of the TISAB solution, the pH of the resulting mixture was measured to make sure it fell within the range of 5 to 6. Replicate [r] readings were generally reproducible to ± 0.5 mV. The precision and accuracy of the measuring method can be appreciated by reference to the following quotation (U.S. Environmental Protection Agency, 1979, p. 274):

"A synthetic sample prepared by the Analytical Reference Service, PHS, containing 0.85 mg/1 fluoride and no interferences was analyzed by 111 analysts; a mean of

0.84 mg/1 with a standard deviation of ± 0.03 was obtained.

272

On the same study, a synthetic sample containing 0.75

mg/1 fluoride, 2.5 mg/1 polwilosphate, and 300 mg/1 alkalinity was analyzed by the same 111 analysts; a mean of 0.75 mg/1 fluoride with a standard deviation of

± 0.036 was obtained."

Although the specific electrode measures ion activities, it is calibrated against solutions of known [F

-

], which results in readings of total concentration of F (as F

-

). Such [F

-

] are labelled as total because TISAB contains CDTA (cyolchexylene dinitrilo tetraacetic acid), an organic chelate in charge of preferentially complexing those ions that can bind F

-

(chiefly Si

4

+, A1

3

*, and Fe

3

+), thus freeing the potentially-ccmplexed F

-

ions (Orion Research, Inc., 1973). An attempt was made to obtain measurements of both complexed and total F

-

by preparing TISAB without CDTA. If we call T the [F

-

] read using regular

TISAB, and Tw the [F

-

] readings that used TISAB without CDTA, the amount of complexed F

-

is obtained by simply subtracting Tw from T.

However, when two sets of standard solutions at pHs in the range of 3 to 7 were measured (one set containing regular TISAB and the other

TISAB with no CDTA), the electrode yielded readings of little difference for samples with the same [F

-

]. The observation was deemed irreconcilable with theory, and the attempt was abandoned. It may be added that Farrah et al. (1985) attempted to determine the amount of complexed F

-

by using TISAB solutions with and without CDTA, but the results were equally unsuccessful. On the other hand, Elrashidi and

Lindsay (1985) reported the successful use of TISAB without CDTA to measure complexed F-.

273

Measurement of pH

For the pH measurements, the pH meter mentioned above was used with a standard glass-combination electrode. The pH meter was operated on the expanded scale mode (i.e., variations of 10

-3

pH units are detectable) to be able to register minute changes of pH in the various solutions. Before proceeding with the measurement of samples, the electrode was calibrated against cœmercial buffer solutions of known pH.

NOTE: Unless specifically stated, all pH and [F

-

] readings were taken at 23 ± 1°C.

APPENDIX

C

CONSTRUCTION OF CCCUMNS AND

CALIBRATION TESTS

274

275

Generalities

Countless number of published reports can be found in which the results of column experiments are presented and discussed. However, most of those reports do not include even the vaguest reference to the steps followed in constructing, packing, and calibrating the columns used. This observation should not be taken lightly because, in my opinion, the various tasks involved in column construction, if not carefully planned and executed, may lead to the obtention of results difficult to interpret or downright meaningless, let alone the waste of time that an inappropriate construction may entail. It is my firm conviction that the quality of the information obtainable from column experiments is in intimate relationship with the care and attention with which some previous steps have been followed. The many frustrating experiences I have had with columns have made me realize the need of warning others as novice as I am about the potential problems in columns construction and their possible solutions, which I crystallized in a recent report (Usunoff, 1988). Conveniently excerpted portions from that report will now be presented.

With slight customizing features, any column used in laboratory experiments has elements shown in Figure C-1. The multiple input device (MID), details of which can be seen in Figure C-2, is needed to ensure the homogeneous distribution of flow at the input surface (i.e., the bottom of the column). Such homogeneity is required not to introduce components of hydrodynamic dispersion other than those legitimately exhibited by the system, and, ultimately, to justify one of the primary assumptions of dispersion tests in columns such as the obten-

n input of distilled

ter

input of tracer

Figure C-1.

Typical elements of a column used in laboratory experiments.

276

sample multiple input device

Figure C-2. Elements of the multiple input device.

bottom of column tubes of flexi ble plastic

277

278 tion of a "square-shaped" input wave. The MID must be calibrated before putting it to use, a task that requires the carrying out of two types of tests:

(1)Test of homogeneous input: due to the particular geometry of each MID and the dispersion inside it, the tracer will select preferential paths in its upward movement, as a result of which some tubes will conduct the tracer faster than others. In order to have the tracer circulating through all the tubes and reaching the input surface simultaneously, the length of some tubes (or even the diameter) should be varied. The question of which tubes should be affected by this procedure is solved by the injection at a low flow rate of a dye flowing throouçtithe water-saturated MID.

(2)Test of arrival time: assuming that water is circulating through the system and that at a given time the double-position valve is switched to allow the entrance of the tracer, it will take a certain time for it to reach the input surface (t = 0 for the experiment).

Such circulation time can simply be calculated by dividing the volume of the MID by the flow rate. In most real situations, however, the so-implicit assumption of piston flow may not hold, and a more refined approach is to be used to calculate the arrival time

(t a

)• One way of doing it is by using a conductivimeter and the setup shown in Figure C-2. Water is

circulated at a known rate Q and, at a given set time, the entrance of a solution ccntaining a strong electrolyte at high concentration (1,000 ppm NaC1, for example) is allowed, after which well-stirred samples are taken at the outlet (base of the column) and their electrical conductivity C measured. Inasmuch as the conductivity of the input solution (CO and that of the background solution (CO are known, the relative concentration (CO can be calculated from C r

= (C-00/(C in

-

C). If the values of C r

are then plotted against time, a typical sigmoidal curve is obtained. The test is repeated at several Q's, trying to cover the range of discharges to be used later in the columns experiments, which will provide a set of curves of relative concentration versus time for a given flow rate. By reading on those curves the time value (t a

) corresponding to C r

= 0.5

(or 0.9 using a very conservative approach), a new plot can be made of Q versus t a

to be used as the calibration curve for the map.

279

The packing of the column, if properly done, will not only avoid the preferential settling of the sediments (unless heterogeneity is deliberately sought) but will also provide an initial estimate of the porosity in the system. In coming up with a value for porosity, the relationship between the column dimensions and the mean particle diameter can be essential to avoid errors in the estimation. Ward

280

(1966) reported that when the ratio between the mean particle diameter and the container diameter and the ratio between the mean particle diameter and the container length were both about 0.1, an error of about 7.3% in the observed porosity was found. Air entrapment during packing can also affect the hydrodynamic characteristics of the system.

Orlob and Radhakrishna (1958) concluded that a 10% increase in the air content of media voids can produce a 15% reduction in effective porosity, a 35% decrease in permeability, and about a 50% reduction in hydrod y namic dispersion. The assumption of homogeneous flow can also be unfulfilled for small values of the ratio between the diameter of the column and the mean particle diameter (Schwartz and Smith, 1953).

Moreover, Mbrris and Kulp (1961) recommended the use of vibratory packers to achieve a uniform packing. in order to avoid air entrapment, which later on would impede or distort the flow within the column, the packing should be done in saturation conditions. To do so, a known volume of water is poured into the column from its top to reach a height of about one inch inside the column. The sediment is then poured from the top until its level within the column reaches that of the water. The process is repeated until the top of the column is reached.

If the volume of added water has been recorded and the total volume of the column is known ("dead" spaces such as volume of sampling ports should be accounted for), the porosity can be estimated.

In order to know the dispersive characteristics of the system, and after the column has been packed and cleaned up by circulation of excess distilled water, a dispersion test is in order. Its preparation and execution are the same as for the MID's test of arrival time, with

281 the advantage that t a

for the selected Q is now known. If the experiment is carried out until C C

I

,

(as measured at the outlet), the plot of C r

. versus time will resemble that presented in Figure C-3.

Defining:

L = column length or length between input surface and sampling port [L]

Q = flow rate through the column [L

3 1

/T]

A = column cros-sectional area [L

2

]

= mean pore velocity [LIT]

0 = effective porosity [L

3

/L

3

]

= x

, ordinate corresponding to Cr= 0.5 [T] dCr/dt (@ vt = L) = slope of the breakthrough curve at C r

= 0.5 [T

-1

]

= longitudinal dispersion coefficient [L

2

/T] a l

= longitudinal dispersivity [L]

Notice that L, Q, A

,

t

*

, and dCr/dt are known, and that the other parameters can be calculated as follows (Neuman, 1984):

= L/t

*

;

= Q/vA; VP (pore volume) = OAL;

= [vP2(dC r

/dt)]

2

(7L)

-1

; a

1

=/v

It is a well-known fact that the value cf the longitudinal dispersivity depends on the scale taken for its determination. In this regard, and in those cases where solute adsorption does not play a major role, a l should be one or two orders of magnitude larger than the mean diameter

(d

50

) of the particles in the column.

OCr

Dt

t*

Time

Figure C-3. Typical breakthrough curve from continuoussource dispersion tests.

282

283

Although somewhat lengthy, this introduction was thought to be necessary for the better understanding of the information pertaining to each of the columns used, which is preamatedbelow.

Column Used in the Adsorption Experiments

Packing material = Ottawa sand.

Column length = 30.48 cm.

Internal diameter of column = 3.15 cm.

Column material = plexi-glass.

Porosity (estimated at packing time) = 38%.

The dispersion experiment considered the followitag previous data:

Q = 0.7305 ml/min.

Tracer solution = 1,000 ppm NaCl.

q n

= 1,985.00 pmhos/cm.

C b

=

3.00 pmhos/cm.

t a

= 36 minutes.

Table C-1 presents the data of the dispersion experiments, whereas the resulting plot is shown in Figure C-4. Based on the analysis already presented and on the data obtained from Figure C-4, the following parameters were calculated:

O (porosity) = 39% (value adopted for later calculations)

PV (pore volume) = 92 crn

3

= 5.63 x 10

-2

cm

2

/min

Table C-1. Dispersion experiment data. Quartz-sand column.

Elapsed time, t

(min.)

0

16

36

56

76

96

117

138

158

178

Elect. 010

, nd.

measured, C

( milcs/cm)

3.00

3.03

3.24

3.55

5.63

37.20

564.00

1,603.00

1,916.00

1,940.00

Relative ccnc.,

(%)

0.00

0.00

0.01

0.03

0.13

1.73

28.33

80.73

96.52

97.73

284

10

0

8 0-

20-

GO

• r

80

— er"

.

• •

100 120

Time (minutes)

140

IGO 180

Figure C-4. Breakthrough curve from dispersion test:

Quartz-sand column.

285

a

= 0 . 22 am

Column Used in the Dissolution Rate of Fluorite Experiments

Packing material = crushed fluorite.

column length = see sketch.

Internal diameter of column = 0.46 cm.

Column material = plexi-glass.

NOTE: No attempt was made to determine the porosity of the packing or its dispersive parameters because they were deemed not essential for the type of experiments for which this column was intended.

out

SP4

A

8E

(no

SP3 dE

(nu

E

.4

mo

SP2 u spi

I input

Columns Used in the Composite Experiments

Column 1

Packing material: mixture of sand, vermiculite, and kaolinite in the following percentages by weight: sand = 90%, vermiculite = 6%, and kaolinite = 4%.

NOTE: After packing the column, and during the regular cleanup procedure, some fine material was observed in the leachate, most likely kaolinite. Therefore, the figure above (4%) is subject to questioning.

Column length = 95 cm.

Internal diameter of column = 3 cm.

Column material = plexi-glass.

Porosity (estimated at packing time) = 48%.

286

287

The dispersion experiment considered the following previous information:

Q = 3.06 ml/min.

Tracer solution = 1,000 ppm NaCl.

= 1,970.00 pmhos/cm.

C b

=

10.90 pmhos/cm.

t a

= 21 minutes.

Data collected during the dispersion experiment are presented in Table C-2, whereas the breakthrough curve is shown in Figure C-5.

The analysis based on the considerations presented before, and using the data from Figure C-5, yielded the following values:

O (porosity) = 47% (value adopted for later calculations)

PV (pore volume) = 315

ce

D

1

=

5.75 cm?/min a

=

6.24 an

Column 2

Packing material: mixture of sand, vermiculite, kaolinite, and crushed fluorite in the following percentages by weight: sand = 88%, vermiculite = 6%, kaolinite = 4%, and fluorite = 2%.

NOTE:

After packing the column, and during cleanup operations, some kaolinite was deserved in the leachate. It should then be assumed that its original percentage (4%) was reduced.

Column length = 93 cm.

Table C-2. Dispersion experiment data.

Mixed-materials column

1.

93

102

111

120

129

138

147

157

167

176

191

0

19

53

71

83

Elapsed

Elect.

Cond. time, t measured,

C

(min.) ( mbcs/cm)

10.90

10.90

22.70

251.00

560.00

782.00

973.00

1,159.00

1,312.00

1,433.00

1,519.00

1,603.00

1,673.00

1,729.00

1,756.00

1,815.00

Relative conc., C r

.

(%)

0.00

0.00

0.60

12.26

28.03

39.36

49.11

58.60

66.41

72.59

76.98

81.27

84.84

87.70

89.08

92.09

288

289

I00

80-

20-

O

50

,

.•

70

.

9 1

1

0

1

3

0

Time (minutes)

1k)

I

10

Figure C-5. Breakthrough curve fran dispersion test:

Mixed-materials column 1.

190

290

Internal diameter of column = 3 cm.

Column material = plexi-glass.

Porosity (estimated at packing time) = 52%.

The dispersion experiment considered the following previous information:

Q = 2.40 ma/min.

Tracer solution = 1,000 ppm NaCl.

C m

= 1,980.00 pmhos/cm.

C b

= 12.51 pmhos/cm.

t a

= 12 minutes.

Data collected during the dispersion experiment are presented in Table C-3, whereas the breakthrough curve adopted the shape depicted in Figure C-6.

The analysis based on the considerations presented before, and using the data from Figure C-6, yielded the following values:

O (porosity) = 62% (value adopted for later calculations)

PV

(pore volume) = 404 ar1

3

• = 3.56 ce/min a l

=

6.44 cm

Table

C-3.

Dispersion experiment data. Mixed-materials column

2.

Elapsed time, t

(min.)

149

158

167

177

189

199

209

219

228

109

119

129

139

0

52

75

97

238

257

Elect.

COnd.

measured,

C

( mhos/cm)

12.51

12.51

14.42

50.60

141.20

272.00

413.00

559.00

713.00

847.00

981.00

1,106.00

1,215.00

1,310.00

1,391.00

1,460.00

1,512.00

1,575.00

1,666.

0

0

Relative conc.,

C r

(%)

0.00

0.00

0.10

1.94

6.54

13.19

20.36

27.78

35.60

42.41

49.22

55.58

61.12

65.95

70.06

73.57

76.21

79.42

84.04

291

.

6

.

6

( ok) in re)

Jo fUOI1D4U23UO3 6A14.Dia8

.

4

,

0

6

00

292

APPENDIX D

RESULTS OF COLUMN EXPERIMENTS ON

F

-

ADSORPTION BY SAND

293

294

Presentation of Results

Seven column runs were carried out to characterize the

F

adsorption by Ottawa sand. The information pertaining to the column construction and calibration is given in Appendix C. As for the sand preparation and the measurements of pH and

[F

-

], the reader is referred to Appendices A and

B, respectively.

For each of the seven runs, fresh

F

-

solutions were prepared by dissolving proper amounts of

NaF and their pH adjusted to the desired value by using diluted solutions of

BIA0

3

or

Na0H.

In spite of the additions of acid or base, the ion strength of the resulting solutions was never above

10r

3

FL

The relevant information for the various columns experiments in presented in Tables D-1 through D-7.

Table D-1. Data iLutoolumn experiment 1.

EXPERIMENT: Fluoride adsorption in sand columns - RUN # 1

Column content = Ottawa sand

Mass of sand in column = 369.5 gm

Effective porosity (from disp. exp. data) = 39%

Column pore volume (from disp. exp. data) = 92

Ce

Average discharge through column = 1.655 ml/min

Mean pore velocity = 0.5589 cm/min

Date experiment performed: 05-01-87

Input solution parameters:

6.00 ppm F-

7.26 ppm Na+ pH = 6.75

T = 22.5°C

Total

Pulse (# of PV of trace solution injected): 3.24

Output pH = 7.05 (at 23.2°C)

12

13

14

10

11

8

9

3

4

1

2

5

6

7

Sample #

PV

2.78

3.06

3.41

3.85

4.24

4.54

4.85

0.00

0.83

1.20

1.54

1.83

2.19

2.52

[r]

(ppm F

-

) C r

=C/C,

0.000

nil.

0.840

4.570

5.430

5.710

6.000

6.000

6.000

5.730

4.920

2.020

0.650

nil.

0.000

0.0

00

0.140

0.762

0.905

0.952

1.000

1.000

1.000

0.956

0.821

0.334

0.107

0.000

295

Table D-2. Data from column experiment 2.

EXPERIMENT: Fluoride adsorption in sand columns - RUN # 2

Column content = Ottawa sand

Mass cf sand in column = 369.5 gm

Effective porosity (from disp. exp. data) = 39%

Column pore volume (from disp. exp. data) = 92

CIO

Average discharge through column = 1.520 ml/min

Mean pore velocity = 0.5133 cm/min

Este experiment performed: 05-06-87

Input solution parameters:

5.45 ppm F-

6.60 ppm Na+ pH = 3.11

T = 22.7°C

Total Pulse (# of PV of trace solution injected): 4.23

Output pH = 3.20 (at 23.1°C)

Sample # PV

7

8

5

6

3

4

1

2

9

10

11

12

13

14

0.00

1.00

1.37

1.75

2.15

2.56

2.95

3.34

3.71

4.59

4.99

5.42

5.81

6.20

[F

i

prin

F

-

)

=C/C,

0.000

0.330

2.770

3.920

4.780

5.070

5.160

5.260

5.260

5.160

4.400

2.080

0.680

0.350

0.000

0.061

0.509

0.719

0.877

0.930

0.947

0.965

0.965

0.947

0.807

0.382

0.125

0.06

296

11

12

13

14

15

16

9

10

7

8

5

6

3

4

1

2

Table D-3. Data from column experiment 3.

EXPERIMENT: Fluoride adsorption in sand columns - RUN # 3

Column content = Ottawa sand

Mass

cf sand in column = 369.5 gm

Effective porosity (from disp. exp. data) = 39%

Column pore volume (from disp. exp. data) = 92

ce

Average discharge through column = 1.600 ml/min

Mean pore velocity = 0.5403 cm/min

Date experiment performed: 05-08-87

Input solution parameters:

6.00 ppm F-

7.26 ppm Na+ pH = 4.88

T = 22.3°C

TOtal

Pulse (# of PV of trace solution injected): 4.25

Output pH = 5.00 (at 22.9°C)

Sample # PV

[F

-

]

(ppm F

-

)

C=C/C

0

0.00

0.63

0.98

1.35

1.72

2.17

2.60

2.99

3.38

3.78

4.59

4.99

5.42

5.81

6.20

6.60

0.000

0.310

0.590

3.260

0.000

0.052

0.098

0.544

4.250

4.910

0.708

0.819

0.840

5.040

5.720

5.870

5.950

5.710

4.840

2.290

0.750

0.953

0.979

0.992

0.952

0.807

0.382

0.125

0.380

0.064

nil.

0.000

297

9

10

7

8

5

6

3

4

11

12

13

14

1

2

Table

D-4.

Data from column experiment 4.

EXPERIMENT: Fluoride adsorption in sand columns - RUN # 4

Cblumn content = Ottawa sand

Mass of sand in column = 369.5 gm

Effective porosity (from disp. exp. data) = 39%

Column pore volume (from disp. exp. data) = 92

ce

Average discharge through column = 1.270 ml/min

Mean pore velocity = 0.4179 cm/min

Date experiment performed: 05-16-87

Input solution parameters:

6.00 ppm F-

7.26 ppm

Na+ pH

= 6.09

T = 22.3°C

Total Pulse (# cf PV of trace solution injected): 3.61

Output pH = 6.44 (at 23.0°C)

SAMPLE # PV

[r]

(Ran F

-

) C r

=C/C

6

3.90

4.28

4.69

5.05

5.41

5.73

0.00

0.60

0.95

1.44

1.83

2.32

2.79

3.12

0.000

nil.

0.870

4.220

5.530

5.650

5.900

5.960

5.950

5.710

3.140

0.950

0.190

nil.

0.994

0.992

0.952

0.523

0.158

0.031

0.000

0.000

0.000

0.145

0.703

0.921

0.942

0.983

298

Table D-5. Data from column experiment 5.

EXPERIMENT: Fluoride adsorption in sand columns - RUN # 5

Column content = Ottawa sand

Mss of sand in column = 369.5 gm

Effective porosity (from disp. exp. data) = 39%

Column pore volume (from disp. exp. data) = 92

an

Average discharge through column = 1.242 ml/min

Mean pore velocity = 0.4086 cm/min

Date experiment performed: 05-18-87

Input solution parameters:

4.00

pprn

F-

4.84

pprn

Na+ pH = 6.13

T = 22.5°C

Total Pulse (# of PV of trace solution injected): 3.58

Output

pH

= 6.22 (at 23.2°C)

9

10

7

8

5

6

3

4

11

12

13

14

1

2

Sample # PV

0.00

0.45

0.85

1.24

1.61

1.96

2.31

2.66

3.04

3.88

4.24

4.62

4.97

5.25

[F]

(ppm F

-

) C i

,=C/C

0

0.000

nil.

nil.

1.450

3.570

3.700

3.950

3.960

4.000

3.960

3.580

2.150

0.670

0.330

0.000

0.000

0.000

0.363

0.891

0.926

0.987

0.990

1.000

0.989

0.896

0.537

0.167

0.080

299

Table D-6. Data from column experiment 6.

EXPERIMENT: Fluoride adsorption in sand columns - RUN # 6

Cblunn content = Ottawa sand

Mass of sand in column = 369.5 gm

Effective porosity (from disp. exp. data) = 39%

Column pore volume (from disp. exp. data) = 92 ce

Average discharge through column = 1.222 ml/min

Mean pore velocity = 0.4020 cm/min

Date experiment performed: 05-20-87

Input solution parameters:

2.00 ppm F-

2.42 ppm Na+ pH = 6.08

T = 22.5°C

Total Pulse (# of PV of trace solution injected): 2.75

Output pH = 6.12 (at 23.1°C)

Sample #

PV

1

4

5

2

3

10

11

12

13

8

9

6

7

0.00

0.44

0.81

1.17

1.53

1.88

2.21

2.53

3.23

3.62

4.02

4.34

4.71

[r]

(ppm F

-

) C r

=C/C

6

0.000

nil.

nil.

0.590

1.362

1.750

1.904

1.982

1.950

1.668

0.756

0.324

0.174

0.000

0.000

0.

000

0.295

0.681

0.875

0.952

0.991

0.975

0.834

0.378

0.162

0.087

300

Table D-7. Data from column experiment 7.

EXPERIMENT: Fluoride adsorption in sand columns - RUN # 7

COlumn content = Ottawa sand

Mass of sand in column = 369.5 gm

Effective porosity (from disp. exp. data) = 39%

Column pore volume (from disp. exp. data) = 92 cuP

Average discharge through column = 0.907 mi/min

Mean pore velocity = 0.2894 cm/min

Date experiment performed: 05-23-87

Input solution parameters:

6.00 ppm F-

7.26 ppm Na+ pH = 6.31

T = 23.0°C

Total Pulse (# of PV of trace solution injected): 3.49

Output pH = 6.54 (at 23.4°C)

SAMPLE #

PV

9

10

11

12

13

7

8

5

6

3

4

1

2

3.00

4.00

4.36

4.71

5.06

0.00

0.44

0.84

1.20

1.56

1.93

2.29

2.64

[F

-

(pin )

C r

=C/C.

0.000

nil.

0.260

3.510

5.220

5.530

5.660

5.800

5.890

4.920

3.850

1.380

0.400

0.000

0.000

0. 043

0.585

0.871

0.922

0.943

0.966

0.981

0.820

0.643

0.230

0.067

301

302

Estimation of Transport

Parameters

The adsorption desorption data presented in Tables D-1 through

D-7 were used as input of a computer program, CFIT1M, (van Genuchten,

1981), used to obtain estimates of the transport parameters. What follows is a description of how the programs operate as given in van

Genuchten (1981).

When chemical adsorption is considered, the one-dimensional convective-dispersive equation can be written as:

ac+E, as=cs a

2

c—v

ac,

a

t

ri at

ax2 ax

where C is the solute concentration [M/1,

3

], P b

is the bulk mass density

[M/L

3

], n is the porosity [L

3

/L

3

], S is the adsorbed concentration

[M/MI, D is the longitudinal dispersion coefficient [L

2

IT],

y is the pore velocity [L/T], t is the time [T], and x is the space [L]. If the relationship between the solute adsorbed and its concentration in solution can be described by a linear isotherm of the form S = K d

C

(where

Rd

is the distribution coefficient {L

3

/M]), the transport equation above becomes:

R

a

C=D

a

2

C - at a

X

2

ax

(D.1) where R, the retardation factor, is given by:

303

R = 1 + PO% (Freeze and Cherry, 1979)

As done by CFIT1M, Equation (D.1) is solved subject to the following conditions:

Initial condition: C(x,0) =

Boundary conditions: upper boundary: C(0,t) = Co lower boundary:

DC (co ,t) = 0 x van Genuchten (1981) indicated that for an analysis of effluent data, it is convenient to introduce the following dimensionless variables:

T = vt/L

P = vL/D

C r

= C-

Co-C i

1 z = x/L

L= column length [L]

(Notice tht T is the number of pore volumes and P is the Peclet number for the column)

Introducing these variables into Equation (D.1), the result is as follows:

R aC =1

2

C - aC r

.

@T --7Z2- az

The program accepts the

Cr

's measured at various T's as input data and, by using a non-linear, least-squares procedure, fits a curve

304

to the data and outputs the oarresIxrding transport parameters (R and

P). Although CFIT1M can handle only those cases formulated in terms of linear adsorption, it

does

give the user the opportunity of selecting among five different options depending on the existence or lack of equilibrium of the adsorption phemamaxm. For the cases presented here, the linea equilibrium adsorption model was selected based on earlier evidence. For illustration purposes, a typical output of the program is presented in Table D-8.

Table D-8. CFITIIIM output for data from column experiment 1.

************************************************************

• NONLINEAR LEAST SQUARES ANALYSIS

EQUILIBRIUM TRANSPORT (MODEL A)

FIRST-TYPE BOUNDARY CONDITION

* FLUORIDE ADSORPTION BY SAND - COLUMN RUN # 1

************************************************************

INITIAL VALUES OF COEFFICIENTS

NO

1

2

NAME peclet retard

INITIAL VALUE

34.000

1.400

OBSERVED DATA

OBS. NO.

1

2

3

4

7

8

5

6

9

10

11

12

13

PORE VOLUME

.8300

1.2000

1.5400

1.8300

2.1900

2.5200

2.7800

3.0600

3.4100

3.8500

4.2400

4.5400

4.8500

CONCENTRATION

.0000

.1400

.7620

.9050

.9520

1.0000

1.0000

1.0000

.9560

.8210

.3340

.1070

.0000

10

11

8

9

6

7

4

5

2

3

ITERATION

0

SSQ peclet retard

4.8559350

34.00000

1.40000

4.1881474

18.02725

1.46131

2.7659679

14.27629

1.67325

1.6572467

15.05110

1.80987

.6740808

5.56488

1.74272

.3624650

12.63394

1.33153

.0882361

21.08067

1.49603

.0282468

37.23396

1.41925

.0165183

51.91983

1.43250

.0157998

56.59684

1.42654

.0157862

57.57889

1.42584

.0157862

57.57936

1.42584

305

Table

D -8 -

-Continued

1

2

3

CORRELATION MATRIX

2 1

1.0000

.0364

-.0716

1.0000

-.6776

3

1.0000

NON-LINEAR LEAST SQUARES ANALYSIS, FINAL RESULTS

VAR

NAME

VALUE

1 peclet 57.57936

2 retard

1.42584

S.E.COEFF.

T-VALUE

9.2289

95%

CONFIDENCE LIMITS

LOWER UPPER

6.24

37.0161

78.1426

.0221

64.42

1.3765

1.4752

----ORDERED BY

CCMPUTER

INPUT

PORE CONCENTRATION

RESI-

NO VOLUME

OBS.

DUAL

1

.830

.000

.002

-.002

2 1.200

.140

.201

-.061

3

1.540

4

5

6

1.830

2.190

2.520

.762

.905

.952

1.000

.694

.925

.068

-.020

.992

-.040

.999

.001

7 2.780

1.000

1.000

.000

8

9

3.060

3.410

1.000

1.000

.000

.956

1.000

-.044

10 3.850

.821

.863

-.042

11 4.240

.334

.312

.022

12 4.540

.107

.073

.034

13

4.850

.000

.011

-.011

-ORDERED BY RESIDUALS

PORE CONCENTRATION RESI-

N° VOLUME CBS.

Mau) DUAL

3 1.540

.762

.694

.068

12 4.540

11 4.240

.107

.334

.073

.312

6 2.520

1.000

.999

7 2.780

1.000

1.000

.034

.022

.001

.000

8 3.060

1.000

1.000

1 .830

.000

.002

.000

-.002

13 4.850

4 1.830

.000

.905

.011

.925

-.011

-.020

306

Table D

-8 - -Cbntinued

5 2.190

10 3.850

9 3.410

2 1.200

END OF

PROBLEM

.952

.992

-.040

.821

.863

-.042

.956

1.000

-.044

.140

.201

-.061

307

APPENDIX

E

RESULTS OF COLUMN AND BATCH TESTS ON

THE DISSOLUTION RATE OF FLUORITE

308

309

Introduction

As explained in Chapter

4, column and batch experiments were carried out to determine the numerical value of the dissolution rate of fluorite when selected chemical and physical parameters are varied. The following is a detailed description of the many characteristics and results from the individual experiments, the interpretation of which can be found in Chapter 4.

Column Experiments

Dimensions of Column

A plexi-glass column of an internal diameter of 1.9 an and a total length of 121.5 an was used in the experiments. Samples were taken at four ports situated along the column (see sketch in Appendix

C). Once packed, the column contained 592.7 gm of fluorite previously treated as presented in Appendix A, being the porosity of the packing material around

46%.

Solubility of Fluorite

The dissolution of fluorite, CaF

2

, under equilibrium conditions can be simply expressed as CaF

2

<===> Cac" + 2F

-

, a reaction characterized by an equilibrium constant of

10-10.41

(Smith and Martell, 1976), that is,

K

= lo r io.41 = [ca.] [

F

.]2

Inasmuch as 2[Ca

4-4

1 = [F

-

], it follows that:

310

[Ca"] 4[Ca"1

2

101

-

10.41

Therefore,

[car

.

]

(lo r io.41/4

)

,

1/2 =

2.135 x 1 0r

4

M (8.54 ppm Ca")

[F

-

] = 2[Ca] = 4.269 x 10r

4

M (8.11 ppm F

-

)

The calculations above disregard the effect on ion concentrations of temperatures different from 25°C and the changes in the ion strength of solutions. Although the experiments described in this section were mostly performed at low solution ion-strength and at temperatures close to 25°C, correction factors were applied to reflect the change in the value of the equilibrium constant due to small deviations from the standard conditions. The change in K due to changes in temperature was accounted for by the use of the van't Hoff equation. Corrections due to ion-strength effect were made by using the GUntelberg equation (Pytkowicz, 1983), which has the following form: log 1', = -0.5

z

1

2

07,

1 +

vTr

where Y = activity coefficient of species i, z i

= charge of species i, p = ion strength of solution (

p.

C

1

z

1

2

), and Ci = molar concentration of species i.

311

The effect of ion pairing was for the most part disregarded, except in column experiment

5.

In that experiment, the pH was low enough to give raise to the appearance in solution of the neutral species liF o

.

The inclusion of the new species was done by the use of a computer code,

MIIINEQL, whose main operating principles can be found in

Wrestall et al.

(1976).

312

Results of the Column Experiments

Table E-1. Data from column experiment 1.

Input solution: milli-Q water (at a pH of 6.17 at 21.9°C)

Q

(mean flow rate) = 1.249 ml/min v

(mean pore velocity) = 0.9577 cm/min

Temperature = 22.0°C

C s

= 7.46 pLan F

pH of output solution (at 22.0°C) = 6.71

The readings presented below were taken after the [F

reached a steady-state distribution inside the column.

] had

Time (minutes after reaching steady state)

[F-] at given sampling port, C (pprn F

-

)

SP1 SP2 SP3 SP4

50

92

126

165

6.03

6.45

6.90

7.09

Ordinates of scattergram of distance from input vs. ln (1-C/C s

):

Distance from

Input, x (cm)

28.0 (SP1)

58.4 (SP2)

89.0 (SP3)

119.1 (SP4) ln (1-C/C

-1.650

-2.000

-2.590

-2.996

s

)

313

Table E-2. Data from column experiment 2.

Input solution: milli-Q water (at a pH of 6.66 at 21.4°C)

Q (mean flag rate) = 0.58 ml/min

y

(mean pore velocity) = 0.4447 cm/min

Temperature = 22.2°C

C s

= 7.46

ppm

F

pH of output solution (at 22.2°C) = 6.82

The readings presented below were taken after the [F

-

] had reached a steady-state distribution inside the column.

Time (minutes after reaching steady state)

[F] at given sampling port, C (ppm F

-

)

SP1 SP2 SP3 SP4

54

108

173

224

6.64

6.99

7.16

7.28

Ordinates of scattergram of distance from input vs. In (1-C/C s

):

Distance from

Input, x (cm)

28.0 (SP1)

58.4 (SP2)

89.0 (SP3)

119.1 (SP4) ln (1-C/C

-2.208

-2.757

-3.228

-3.737

s

)

314

Table E-3. Data from column experiment 3.

Input solution: milli-Q water (at a pH of 6.38 at 23.5°C)

Q

(mean flow rate) = 1.812 ml/min v

(mean pore velocity) = 1.389 cm/min

Temperature = 22.0°C

C s

= 7.46 pin

The readings presented below were taken after the [F

-

] had reached a steady-state distribution inside the column.

Time (minutes after reaching steady state)

[F-] at given sampling port, C (ppm F

-

)

SP1 SP2 SP3 SP4

26

63

107

143

5.75

6.34

6.73

7.04

Ordinates of scattergram of distance from input vs. In (1-C/C s

):

Distance from

Input, x (cm) in (1-C/C s

)

28.0 (SP1)

58.4 (SP2)

89.0 (SP3)

119.1 (SP4)

-1.473

-1.900

-2.322

-2.869

315

Table E-4. Data from column experiment 4.

Input solution milli-Q water (at a pH of 5.05 at 24.0°C)

Q

(mean flow rate) = 1.209 ml/min v

(mean pore velocity) = 0.9266 cm/min

Temperature = 25.9°C

= 8.29 ppm F

pH of output solution (at 25.9°C) = 5.59

The readings presented below were taken after the [F

-

] had reached a steady-state distribution inside the column.

Time (minutes after reaching steady state)

[F-] at given sampling port, C (ppm F

-

)

SP1 SP2 SP3 SP4

31

73

140

178

6.42

6.71

6.81

7.10

Ordinates of scatteiyiam of distance from input vs. In (1-C/C s

):

Distance from

Input, x (cm) in 1-C/CS)

28.0 (SP1)

58.4 (SP2)

89.0 (SP3)

119.1 (SP4)

-1.489

-1.658

-1.723

-1.941

316

Table E-5. Data from column experiment 5.

Input solution: milli-Q water (at a pH of 2.99 at 25.5°C)

Q (mean flow rate) = 1.194 ml/min v (mean pore velocity) = 0.9157 cm/min

Temperature = 26.0°C

= 13.69 ppm F

-

(ion pairing accounted for) pH of output solution (at 26.0°C) = 3.16

The readings presented below were taken after the [F

-

] had reached a steady-state distribution inside the column.

Time (minutes after reaching steady state)

[F-] at given sampling port, C (ppm F

-

)

SP1 SP2

SP3

SP4

51

95

145

186

9.80

11.20

11.45

11.75

Ordinates of scattergram of distance from input vs. In (1-C/C s

):

Distance from

Input, x (cm)

28.0 (SP1)

58.4 (SP2)

89.0 (SP3)

119.1 (SP4) in (1-C/C

-1.258

-1.704

-1.810

-1.954

s

)

317

Batch Experiments

Sixteen batch tests were carried out, in which several parameters were varied to assess the effect on the dissolution rate of fluorite.

A

summary of the chemical and physical conditions under which the experiments proceeded is given in Table

E-6, whereas the measurements taken at each test are presented in Tables

E-7 through

E-

22.

318

Results of the Batch Tests

Table

E-6.

Batch

tests:

summary of the chemical and physical characteristics.

Batch

#

Batch solution Stirring rate (rpm) pH

T (°C)

1

2

3

4

7

8

5

6

milli -Q

water milli

-Q water milli

-Q water

2x1

0 r3 M NaHCO

3

6x10-3 M NaHCO

3

1x1Cr2 M Nal-E0

3

8x10-4 MaZO

4

1.2x10 M CaSO

4

9 1.6x10r3 M CaSO

4

10

1 x1

0 r3 M NaC1

11 5x1

0

0 M NaCl

12

1 x10-2 M NaC1

13

1 x1

0 r3 M NaC1

14 1x1Cr3 M NaC1

15 1x100 M NaC1

16

1 x1

0

0 M NaC1

300

150

150

150

150

150

150

150

150

150

0

150

150

150

150

150

6.16

6.15

6.14

10.36(*)

10.36(*)

10.37(*)

6.46

6.53

6.57

6.73

6.75

6.66

6.29

6.20

6.37

6.16

23.3

26.1

23.0

24.3

23.7

24.9

25.0

24.5

25.9

26.8

26.7

28.4

15.0

20.1

24.8

30.9

(*)

High pH values most likely due to precipitation of calcium carbonate.

NOTES:

1. For all the batch tests, r s

,

(soil/water ratio) was

0.0125

[25 gm of fluorite per

2000

ml of solution].

2. The reported pH and temperature values represent the average of the readings taken throughout the experiment.

319

Table E-7. Data from batch test 1.

Batch experiment 1 (high stirring rate)

Time (minutes since test was started)

120

200

280

355

510

1460

[F] in solution,

C (ppn )

2.70

3.35

3.63

3.82

4.56

5.10(*) ln ( /C s

-C)

0.4373

0.5790

0.6414

0.6811

0.8748

1.0310

(*) Not a reliable figure; the stirrer stopped working sometime after the first 510 minutes.

NOTE: CS was calculated at sampling time by making the oanrwixxxiing corrections for temperature and ion-strength effect.

320

Table E-8. Data from batch test 2.

Batch experiment 2 (medium stirring rate)

Time (minutes since test was started)

120

200

280

355

510

1460

[r] in solution,

C (ppm r)

1.24

1.66

2.02

2.28

2.98

4.12

In (C s

/C.-C)

0.1673

0.2309

0.2840

0.3244

0.4418

0.6321

NCuh: C s

was calculated at sampling time by making the corresponding corrections for temperature and ion-strength effect.

321

Table

E-9.

Data from batch test

3.

Batch experiment

3

(not stirred)

Time (minutes since test was started)

120

200

280

355

510

1460

[F-] in solution,

C (ppm F

-

)

0.45

0.60

0.68

0.80

1.12

1.91

In

(C s

/CS-C)

0.0610

0.0815

0.0932

0.1101

0.1550

0.2846

NOIE: CS was calculated at sampling time by making the corresponding corrections for temperature and ion-strength effect.

322

Table

E-10.

Data from batch test

4.

Batch experiment

4

(low

[NaHCO

3

] in solution)

Time (minutes since test was started)

130

200

250

315

380

440

1440

[F-] in solution,

C (ppm r)

0.99

1.12

1.36

1.54

1.68

1.89

4.23

in

(CS/C s

0.1249

0.1387

0.1688

0.1922

0.2091

0.2372

0.6235

-C)

NOTE: C s

was calculated at sampling time by making the corresponding corrections for temperature and ion-strength effect.

323

Table E-11. Data from batch test 5.

Batch experiment 5 (medium [NaHCO

3

] in solution)

Time (minutes since test was started)

130

200

250

315

380

440

1440

[F-] in solution,

C (ppm F

-

)

1.19

1.60

1.85

2.13

2.35

2.77

17.49(?) ln ( /C s

-C)

0.1452

0.1953

0.2258

0.2616

0.2893

0.3490

(*)

(*) The calculation yields a negative number, which indicates that the solubility C s

should not be based on the assumption that CaLF

2

is the only solid

phase

present. This fact, plus the high equilibrium pH measured point to a hypothetical precipitation of CaCO

3

.

NOTE: C s

was calculated at sampling time by making the corresponding corrections for temperature and ion-strength effect.

324

Table

E-12.

Data from batch test

6.

Batch experiment

6

(high

[NaHCO

3

] in solution)

Time (minutes since test was started)

130

200

250

315

380

440

1440

[F] in solution,

C (ppm r)

1.19

1.45

1.64

1.97

2.22

2.44

10.70(?) ln

(C s

/C s

-C)

0.1354

0.1620

0.1826

0.2206

0.2496

0.2578

(1)

(*)

Same comment as above (batch

5).

NOTE: C was calculated at sampling time by making the corresponding corrections for temperature and ion-strength effect.

Table E-13. Data from batch test

7.

Batch experiment 7 (low [CaSq l

] in solution)

Time (minutes since test was started)

130

200

275

350

420

490

1440

[F-] in solution,

C (prm F

-

)

2.10

2.45

2.66

2.86

3.07

3.25

3.51

ln (

0.2661

0.3147

0.3422

0.3695

0.3984

0.4252

0.4916

-C)

NOTE: C s

was calculated at sampling time by making the corresponding corrections for temperature and ion-strength effect.

325

326

Table E-14. Data from batch test 8.

Batch experiment 8 (medium [CaS00 in solution)

Time (minutes since test was started)

130

200

275

350

420

490

1440

[F-] in solution,

C (ppm F

-

) [*]

2.55

3.03

3.99( 7)

3.50

3.62

3.79

5.93

ln (C s

/C s

-C)

0.3313

0.4003

0.5580

0.4665

0.4807

0.5089

1.0424

[*] The stirring bar was found partially grounded, which provided loose particles that acted as nucleii onto which some precipitation occurred.

The solution was filtered and the precipitate collected. When acidified, it dissolved as if it were CaCO

3

. Consequently, the results are reported but not used.

NOTE: C s

was calculated at sampling time by making the corresponding corrections for temperature and ion-strength effect.

327

Table

E-15.

Data from batch test

9.

Batch experiment

9

(high

[CaSCO in solution)

Time (minutes since test was started)

130

200

275

350

420

490

1440

[F-] in solution,

C (ppm F

-

)

1.48

1.81

2.13

2.35

2.61

2.82

2.86

In

(C s

/C i

,-C)

0.1708

0.2082

0.2448

0.2707

0.3011

0.3275

0.3454

NOTE: C s

was calculated at sampling time by making the corresponding corrections for temperature and ion-strength effect.

328

Table E-16. Data from batch test 10.

Batch experiment 10 (lcw [NaCl] in solution)

Time (minutes since test was started)

130

210

270

340

405

470

1440

[F-] in solution,

C (ppm F

2.32

3.11

3.50

4.04

4.47

4.87

7.37

-

) ln (C s

/C s

0.3083

0.4238

0.4861

0.5789

0.6609

0.7428

1.7203

-C)

NOTE: C s

was calculated at sampling time by making the corresponding corrections for temperature and ion-strength effect.

329

Table

E-17.

Data from batch test

11.

Batch experiment

11

(medium

[NaCl] in solution)

Time (minutes since test was started)

130

210

270

340

405

470

1440

[F-] in solution,

C (ppm F

-

)

2.79

3.56

3.99

4.59

5.01

5.45

8.44

in

(C i

,/C s

0.3488

0.4565

0.5215

0.6195

0.6950

0.7800

2.0140

-C)

NOTE: C s

was calculated at sampling time by making the corresponding corrections for temperature and ion-strength effect.

330

Table

E-18.

Data from batch test

12.

Batch experiment

12

(high

[NaCl] in solution)

Time (minutes since test was started)

130

210

270

340

405

470

1440

[F-] in solution,

C (ppm F

-

)

2.22

2.55

2.93

3.50

3.80

4.19

6.11

In

(C s

/C i

,-C)

0.2432

0.2749

0.3167

0.3859

0.4231

0.4763

0.8388

NOTE: C is

was calculated at sampling time by making the corresponding corrections for temperature and ion-strength effect.

331

Table E-19. Data from batch test 13.

Batch experiment 13 (low temperature)

Time (minutes since test was started)

150

230

290

340

390

445

500

[F-] in solution,

C (ppm F

-

)

0.90

1.30

1.49

1.55

1.70

1.79

1.94

ln (C

5

/C

5

-C)

0.1859

0.2591

0.3049

0.3249

0.3621

0.3902

0.4342

NOTE:

C s

was calculated at sampling time by making the corresponding corrections for temperature and ion-strength effect.

332

Table E-20. Data from batch test 14.

Batch experiment 14 (low to medium temperature)

Time (minutes since test was started)

150

230

290

340

390

445

500

[r] in solution,

C (ppm F

1.70

2.14

2.44

2.57

2.85

3.09

3.20

-

) ln (C

2

/C i

,-C)

0.2593

0.3362

0.3950

0.4187

0.4732

0.5297

0.5578

NOTE: C, was calculated at sampling time by making the corresponding corrections for temperature and ion-strength effect.

333

Table E-21. Data from batch test

15.

Batch experiment 15 (medium to high temperature)

Time (minutes since test was started)

130

180

230

290

350

410

465

[F-] in solution,

C (ppm )

2.16

2.62

2.90

3.22

3.65

3.82

4.15

in

(

0.2875

0.3615

0.4103

0.4622

0.5421

0.5835

0.6603

-C)

NOTE: C s

was calculated at sampling time by making the corresponding corrections for temperature and ion-strength effect.

334

Table E-22. Data from batch test 16.

Batch experiment 16 (high temperature)

Time (minutes since test was started)

130

180

230

290

350

410

465

[F-]

C in solution,

(ppm F

3.42

4.01

4.33

4.77

5.06

5.50

5.71

-

) ln (CS/C s

0.4287

0.5254

0.5813

0.6639

0.7211

0.8218

0.8703

-C)

NOTE: CS was calculated at sampling time by making the corresponding corrections for temperature and ion-strength effect.

APPENDIX F

RESULTS OF COLUMN OZMPOSITE EXPERIMENTS

335

336

Data

from Column

Composite Run 1

The dimensions and dispersive characteristics of the column used in this experiment (column 1) are given in Appendix C. A total of

4.04 PV (pore volumes) of solution (source strength) were pumped through the column before switching the two-way valve to give access to milli-Q water at a pH equal to that of the input solution.

Description of Samples

Sample #

3

4

5

6

7

8

1

2

Remarks

Input solution

Background output solution (taken right before run 1 started)

Output solution (between 0.78 and 1.27 PV)

H 11

(

" 1.87 and 2.37 PV)

11 11 11

H 11

(

(

II

3.29 and 3.84 PV)

4.83 and 5.37 PV)

H

11

H

11

(

(

H

H

5.82 and 6.39 PV)

7.29 and 7.80 PV)

Data

from

Column Composite Run

2

The dimensions and dispersive characteristics of the column used in this experiment (column 1) are given in Appendix C. A total of

6.06 PV

(pore volumes) of solution (source strength) were pumped through the column before switching the two-way valve to give access to milli-Q water at a pH equal to that of the input solution.

337

Description of Samples

Sample

#

9

10

11

12

13

14

15

16

17

18

Remarks

Input solution

Background output solution (taken right before run

2 started)

Output solution (between

0.80 and

1.34 PV) tt n

( "

1.81 and

2.17 PV) u ti

(

Il

3.48 and

4.03 PV) ti it

(

Il

5.24 and

5.86 PV)

Il

II

(

II

7.16 and

7.67 PV) it u tt tt

(

(

If

8.06 and

8.53 PV) n

8.80 and

9.31 PV) n ti

( tt

10.93 and

11.40 PV)

Data from Column Composite Run 3

The dimensions and dispersive characteristics of the column used in this experiment (column 1) are given in Appendix C. A total of

5.96 PV

(pore volumes) of solution (source strength) were pumped through the column before switching the two-way valve to give access to water at a pH equal to that of the input solution.

Description of Samples

Sample

19

20

21

22

23

24

#

Remarks

Input solution

Background output solution (taken right before run

3 started)

Output solution (between

1.41 and

1.86 PV)

II

II

(

" 2.40 and

2.89 PV) it

I/

it

If

(

(

"

"

3.26 and

3.79 PV)

5.32 and

5.76 PV)

6.75 and

7.25 PV)

8.48 and

8.87 PV)

9.31 and

9.76 PV)

10.33 and

10.90 PV)

12.17 and

12.74 PV)

338

Table F-1. Chemical Analyses of Samples. Column Runs 1, 2, and 3.

339

Sample

Namber [Ca']

[Na]

[SO

4 " ] [Cl-] [H00 3

-

]

[F

-

]

Elect.

Cond.

pH

Temp.

21

22

23

24

17

18

19

20

25

26

27

28

29

13

14

15

16

9

10

11

12

5

6

7

8

3

4

1

2

<0.20

<0.20

20.00

<0.20

1.83

3.46

3.56

4.12

20.50

0.33

0.43

4.82

5.58

5.58

6.60

<0.20

3.02

<0.20

<0.20

<0.20

<0.20

20.65

<0.20

0.33

6.68

6.36

4.16

0.57

<0.20

50.0

45.1

103.0

9.7

1.0

3.6

12.0

6.4

67.8

41.7

19.4

96.1

73.1

45.8

101.0

76.4

34.9

79.3

32.7

16.9

7.2

1.5

48.8

44.7

15.1

5.1

91.8

4.7

7.8

64.8

39.2

65.2

45.7

67.6

45.4

46.7

29.7

4.9

16.4

10.8

2.0

0.9

50.0

44.7

15.0

0.7

16.5

29.7

43.5

44.9

49.2

45.7

64.6

45.2

54.9

33.2

14.4

10.4

8.6

1.9

1.3

1.0

22.1

83.8

88.9

91.7

56.8

6.5

1.5

<0.5

50.7

<0.5

50.7

51.4

51.5

51.5

38.3

1.7

1.0

1.1

79.0

6.2

507

0.0

<0.3

22

8.0

<0.3

71

66.3

<0.3

429

71.0

<0.3

458

61.7

4.1

401

16.6

4.1

77.3

8.4

2.9

6.0

124

47

450

11

10.4

<0.3

70

69.0

<0.3

403

73.7

<0.3

430

76.8

3.4

447

51.9

7.1

306

9.2

---

7.2

2.5

6.7

1.5

52

25

79.3

5.9

398

0.0

<0.3

10

42.3

62.7

65.8

75.7

59.9

0.0

0.0

0.0

5.8

217

317

332

380

6.3

3.5

2.2

7.6

303

69

2.3

1.3

0.7

41

27

21

4.066

20.7

7.203

21.6

6.993

22.2

6.693

22.6

6.758

22.2

6.904

22.8

7.095

22.9

7.097

22.9

6.246

20.3

7.341

21.0

6.982

22.1

6.971

21.9

6.961

21.5

6.880

21.1

6.742

21.3

6.862

22.0

7.012

21.1

7.134

22.0

8.491

21.1

7.263

20.9

6.886

21.8

6.864

20.7

7.027

20.5

7.205

21.5

7.463

20.6

7.416

21.6

7.271

22.2

7.355

21.7

7.285

21.6

NOTE: All values are ppm, except for conductivity (micrcmhos/cm), pH, and temperature (°C). Conductivity, pH, and temperature were measured at sampling time. [C1

-

], [SO

4

=], and [F

-

] were determined by ion chrunatography. [Ca++] measurements used a atomic adsorption device.

[HCO

3

-

] and [Na] were not determined analytically but calculated as described in Chapter 5.

340

Data from Cblumn Ccmposite Run 4

The dimensions and dispersive characteristics of the column used in this experiment (column 2) are given in Appendix C. A total of about 12 PV (pore volumes) of milli-1Q water were pumped through the column at a constant rate, with samples being taken at the outlet.

Description of Samples

Sample #

30

31

32

33

34

35

36

37

Remarks

Background output solution (taken right before run 4 started)

Output solution (between 1.01 and 1.21 PV)

ti

2.94 and 3.19 PV)

3.96 and 4.17 PV)

4.94 and 5.18 PV)

6.77 and 7.00 PV)

7.88 and 8.09 PV)

10.31 and 10.62 PV)

341

Table F-2. Chemical Analyses of Samples. Column Run 4

30

31

32

33

34

35

36

37

Sample

NuMber Ca" Na+ Cl -

<0.2

2.66

<0.5

<0.2

3.05

0.9

<0.2

3.08

<0.5

<0.2

2.94

<0.5

<0.2

2.95

<0.5

<0.2

2.79

<0.5

<0.2

2.82

<0.5

<0.2

2.65

<0.5

Kt:

0.7

0.9

0.9

0.8

0.9

0.8

0.8

0.7

F

-

Elect.

Cond.

pH

Tentp.

0.7

10.80

1.0

16.65

1.3

16.18

1.3

15.59

1.3

15.96

1.3

15.04

1.4

15.25

1.3

14.46

6.981

22.1

7.036

22.2

6.936

21.2

7.000

22.0

6.986

22.1

6.964

21.3

7.043

22.6

6.947

21.0

NOTE: Units and analytical methods used same as above, except for N'a!

which was determined by flame.

APPENDIX

G

CHEMICAL

DATA

FOR STATISTICAL STUDY

342

-

343

1

1

1 1

1 !

;. g g

!

n

;i

sr

!

:

41

!

!

! ! .!

!

1

!I

.

h. b.

1.1 11.

14 1 R

E E 1. 1 f,

F.

il

al al al

51

31 ii

Si ii Bi

Si

Ri fi fi fi

El

fi fi 1; fi fi

.ft

Ds g g g g g

-• - • -.:.

,E ,g

g

_

0i mN 21 21

--m

-A

-m mE

-A -A

EE 2! mi tE 21 RI rIN al

L-E 21 2R mE 21 eE 2E 0, -0 el

=1

-A -.

4

'A

-

A m

4

-

A

sZ

'A

"4 '4 -A

-

A

-

.7

:

'A

22

21

RA

e! 21 xi e! 21 RE .4 mi

_ _ sE :TR 21 mi g .4

5

RI

- -^ ri mE 01 si -E

E=

.

2

3

.* ' * 1 .

2

.E E .! .R .R .E .R m= .E

e el...cm-true simeesucRmstecem s i g g g 1 s i !i

R i

2 i g g '

71

R i s i s i s i s i 1

R i

R i zi

.E * E

.

Ai

-

.E .5 .

1

1

1

Ai

- - 2 = mi ft. ft. eff;

S"

2

17 «, 1.

= 2 2 = 2

W.

2. v'

= -

-

=

-X

"2 eR sE ef -5 ei -E 21 g cf ei 2i

_ r.a. 21R;

I

-R -;

I

.; -z

• m!

A! 2! e! c! 2.! =.! r! s: .4 2: 2! 2! 2! 2! r! R! =! m! 2! R! e! r!

22228 2

V z

1?

2: :eery m 2 2

6

4.0..wz.glzztv.zz

t t !t: t t

ft ft „ ft ......

,v na

.

.

...

.

88

ea

Ii t

T.

7

.

7

. 7 7 - 7 7 - 7 7 r r •

- 7

'F-fEif

- ci

I s i 11

2211; sai .istsss isssttsttstit

^2

;Ji!li

-

!

!fill

«

R «

- r 2 « R R 2 R R

R

4 4 ; 4 4 4 4 4

Illf!IlliiJil

R R

444 4 4444444 4 Ass4Asies444

2 Zr

344

.: a a sat tat a

I il

2

`"

:::::14 s 1

:

I

1

7

Ft

ti

i

Il

I

4

a

I

:

I e. t :

!

1 r.

a

-!.-

13

I1. i il

II

: it ii

' 2 2 2 •

Z.

• - ..

I" il t. i a 1 E ER

5 5 5

F.: L'i

2

! ii

A

E E E

A E E

4

.3

A F.

1

3

Ali ii

4

1 gi Ei 'a ii gl gl II

Ri

RI il ' il

41

RI

RI

$0 0

4i il 4i :i il

-i l _. m

ft..;

z „ox

. 2 2 '"

.4 :! :! :!

r

! ..

:

E

. 2

:.::

..4 -E mE -.5 -5

1

-I m5 rE -.

7,

q

Pi

el AZ a

S1

YE

-.; -

-,, .-. =4

4

:

3

sl mi

RE mi g!

-I mil RE

-5 mi mi -5 -5

RI

_ m _

LI

'2

X m.

R ;SS 2. ; S 5 5 le 5 7. m-

S

x :Isaac sx g xr.gx r nag *g x

4i 45

9 41 4i 4i 4 4 4 ei

3= 4i i 4i 4 4i 4i ei 5q '4 a ll ai at ai 45 sE ai

Rg 41 2: 2

1

:

RE ai gR

RE ai aZ mR gi aE xi xi tR

*I ai

_ rtigi -s =i

-

3

g cti e!

4

' 4" g

3

R

5

"

8

^

2

' 5

2! 2!

= =

-E "

pS r4°

-3

2R

-E e

-

R

-6 4

_

E

41

_ _

.x

x

2 .2

7,

2 x

2 -S t

8

5

2

R rn

3 .7.

5 5 5

I

7'

R R

14.

g

s

It

d

I

I

FA

EZELL E L L L a

ES SS

2

r. r F

7 : ... : ...

7

F r F F

:...-v

. i

1

1

11111111/11f11121111

-

111

.:

1

-1

2.

t" tt

II

a

I

Ul X

t

2

2

g

111

.., fifiii! 11

1

[(I

I

;

a :y

E

2 2

11

4

Pm

;-7

CD

ri

a

a a a

. 1 a d a a

: it a a a g

I d a i

2 ."

'

F.

PE

A

-7-

"

2 2. I t St

P 21 R X

St R 2 St

XV

:V ZsCs

a s s

E-I

345

346

Table G-2. San Pedro River Basin: Statistics of chemical variables.

347

Variable pH

CW

-

F

Mci f

+

Na

IC'

S10

2

Cl

-

SO

4

=

HC10

3

-

F

-

Br

-

Conductivity

7.32

27.79

4.70

32.59

1.71

29.55

7.38

13.93

163.76

1.97

0.04

282.00

Confined Aquifer

Mean Standard

Deviation

0.27

15.49

3.81

20.51

0.65

5.85

2.54

19.13

47.13

1.25

0.03

82.83

7.40

50.39

16.00

82.46

2.03

27.85

17.00

117.31

239.00

1.71

0.18

655.23

Unconfined Aquifer

Mean

Standard

Deviation

0.56

38.79

12.45

86.81

1.41

7.69

12.60

165.43

97.76

0.99

0.16

470.99

All values are ppm, except for electrical conductivity (micromhos/cm) and pH.

Number of samples considered: 29 from the confined aquifer (sample 9 disregarded), and 13 from the unconfined aquifer (sample 12 disregarded).

Table G-3. San Pedro River Basin: Correlation coefficients, unconfined aquifer samples.

Ca++

Mg++

Na K* S102 C1

Ca

Mg

++

Na

K+

Si0

2

Cl

-

SO

4

=

HCO

3

-

F

-

Et

-

PH

Cond.

1.00000

.52568

1.00000

.65305

.83726

1.00000

.23389

.30500

.46156

1.00000

.21507

.44737

.30512

-.05345

1.00000

.36705

.78126

.83722

.39996

.04557

1.00000

.86909

.73646

.87591

.42166

.12149

.65785

.59861

.77422

.85496

.20066

.53193

.66824

-.19021

-.18714

.01400

.01155

.15111

-.20475

.54795

.85545

.90617

.42524

.12661

.92339

-.49426

-.57559

-.53023

.04273

-.74683

-.30073

.82377

.85266

.94039

.36007

.39647

.70190

SO

4

HCO3FBrpH

Cond.

SO

4

=

H003

-

F

-

Br

-

PH

Cond.

1.00000

.66007

-.08452

.80081

-.41046

.93695

1.00000

-.07389

.75776

-.74690

.84707

1.00000

-.28904

-.17475

-.08094

1.00000

-.33819

.85630

1.00000

-.58567

1.00000

348

Table G-4. San Pedro River Basin: Correlation coefficients, confined aquifer samples.

Ca++

Mg++

Na'

Si02

Ca

Mg++

Na +

K+

Si0

2

Cl

-

SO

4

=

HCC3

-

F

-

Br

-

PH

Cond.

.57884

1.00000

-.21381

.18150

1.00000

.01861

-.03279

.26273

1.00000

.37716

.38264

-.10337

-.19101

1.00000

.07556

.35293

.17719

.04670

.01664

1.00000

.28573

.23348

.65864

.20515

.13304

.20698

.43097

.78515

.58476

.10840

.25248

.19613

-.14835

-.35016

.20389

.53444

-.57900

-.10039

.23785

.54754

.00689

-.21790

.13557

.71570

-.39646

-.39186

-.04878

-.18630

-.19294

.12196

.59048

.82196

.40425

-.02374

.26803

.10139

SO

4

=

HCO3F-

Br-

PH

Cond.

SO4=

HCO

3

-

F

-

Br

pH

Cond.

1.00000

.37949

.23673

.03785

-.00550

.43342

1.00000

-.19568

.34500

-.53388

.86506

1.00000

-.36516

.15344

-.28692

1.00000

-.16577

.33378

1.00000

-.47041

1.0

349

350

Table G-5. San Bernardino Valley: Chemical analyses.

Sample

Number

Well depth

(ft.)

[Fe"j

(PPb)

[SiCs]

(PPII)

[Ca"]

(Wm) f] [M

(PPm)

[Naf]

(Wm)

[W]

(Wm)

[HCCs

-

]

(PPm)

36

37

38

39

40

41

42

31

32

33

34

35

15

16

17

18

11

12

13

14

9

10

7

8

5

6

3

4

19

20

21

22

23

24

25

26

27

28

29

30

1

2

28.0

32.0

22.0

25.0

28.0

20.0

20.0

30.0

25.0

12.0

32.0

10.0

32.0

10.0

29.0

31.0

30.0

30.0

38.0

30.0

33.0

33.0

30.0

30.0

30.0

10.0

30.0

22.0

25.0

10.0

10.0

10.0

23.0

28.0

12.0

20.0

10.0

20.0

27.0

30.0

30.0

30.0

128.0

340.0

552.0

425.0

396.0

382.0

798.0

381.0

80.0

30.0

40.0

540.0

448.0

498.0

190.0

546.0

569.0

590.0

375.0

648.0

414.0

20.0

200.0

100.0

360.0

60.0

500.0

20.0

40.0

230.0

418.0

110.0

200.0

100.0

464.0

2400.0

311.0

257.0

225.0

10.0

180.0

20.0

10.0

292.0

390.0

600.0

750.0

110.0

120.0

140.0

245.0

245.0

50.0

100.0

40.0

10.0

20.0

800.0

440.0

100.0

80.0

0.0

30.0

290.0

256.0

-- 0

415.0

40.0

130.0

20.0

10.0

500.0

38.0

18.0

48.0

50.0

32.0

55.0

64.0

52.0

48.0

48.0

57.0

9.0

52.0

56.0

32.0

58.0

26.0

40.0

64.0

76.0

56.0

68.0

48.0

30.0

55.0

50.0

51.0

74.0

46.0

68.0

67.0

68.0

84.0

74.0

76.0

94.0

80.0

23.0

84.0

76.0

55.0

57.0

5.0

6.0

45.0

46.0

1.0

100.0

6.0

6.0

34.0

45.0

3.0

6.0

34.0

54.0

2.0

4.0

6.0

4.0

94.0

45.0

32.0

82.0

1.0

120.0

8.4

5.0

3.0

59.0

58.0

37.0

50.0

13.0

13.0

10.0

9.0

50.0

69.0

66.0

17.0

10.0

5.0

6.0

5.0

50.0

31.0

24.0

13.0

3.0

5.0

5.0

5.0

18.0

24.0

24.0

24.0

40.0

4.0

10.0

12.0

12.0

28.0

26.0

30.0

8.0

7.0

6.0

9.0

4.0

26.0

31.0

34.0

24.0

20.0

6.0

8.0

6.0

6.0

6.0

29.0

28.0

29.0

31.0

26.0

27.0

3.0

170.0

2.0

160.0

6.0

98.0

3.0

160.0

3.0

170.0

3.0

160.0

3.0

180.0

2.0

140.0

2.0

130.0

3.0

190.0

3.0

160.0

3.0

110.0

2.2

160.0

3.0

170.0

3.0

130.0

4.0

190.0

6.0

190.0

1.0

160.0

3.0

170.0

3.0

180.0

3.0

170.0

3.0

190.0

2.0

170.0

3.0

110.0

2.0

160.0

3.0

160.0

2.0

160.0

2.4

230.0

3.0

170.0

4.0

190.0

2.8

210.0

2.6

180.0

3.0

200.0

2.0

190.0

4.0

170.0

2.0

250.0

3.0

200.0

3.0

160.0

4.0

160.0

4.0

180.0

3.0

130.0

4.0

130.0

351

Table G-5 --Continued

Sample

Number

Well depth

( ft. )

[FW+]

(PPb)

[Si0

2

]

(PPrn)

[Ca 'j

(MO

[M g ]

(P121n)

[Na+]

(Pin)

[1(1 ]

(PPITI)

[FICC1

3

-

]

(Pi:rn)

43

44

45

46

47

460.0

733.0

385.0

460.0

400.0

48

49

952.0

818.0

50 1052.0

51 280.0

52

53

972.0

54 1100.0

40.0

40.0

13.0

<10.0

100.0

20.0

<10.0

140.0

40.0

200.0

20.0

20.0

20.0

20.0

20.0

20.0

15.0

20.0

20.0

20.0

20.0

36.0

30.0

32.0

46.0

30.0

72.0

30.0

30.0

53.0

37.0

25.0

130.0

20.0

22.0

6.0

3.0

3.0

4.0

3.0

7.0

3.0

2.0

5.0

3.0

10.0

2.0

16.0

14.0

12.0

21.0

16.0

29.0

16.0

15.0

26.0

25.0

57.0

35.0

2.0

120.0

2.0

120.0

1.0

120.0

2.0

150.0

2.0

64.0

3.0

140.0

2.0

110.0

3.0

120.0

2.0

150.0

3.0

120.0

4.0

230.0

3.0

120.0

Table G-5- -Cbntinued

27

28

29

30

23

24

25

26

31

32

33

34

35

36

37

38

39

40

41

42

17

18

19

20

21

22

13

14

15

16

7

8

5

6

9

10

11

12

3

4

1

2

Sample

Number

[S014=]

(10Pm)

[C1

-

]

(PPm)

[F]

[N10

3

]

(PPm) (Prom)

S.P.

( Sim) pH

Temp.

(°C)

[3:

4-

]

(ppb)

71.0

20.0

35.0

38.0

52.0

63.0

64.0

48.0

63.0

50.0

70.0

65.0

74.0

50.0

72.0

88.0

64.0

34.0

55.0

66.0

120.0

110.0

65.0

96.0

90.0

33.0

17.0

42.0

66.0

70.0

100.0

45.0

68.0

35.0

50.0

140.0

57.0

35.0

74.0

160.0

31.0

35.0

20.0

23.0

20.0

14.0

23.0

14.0

21.0

25.0

16.0

18.0

21.0

39.0

23.0

18.0

12.0

20.0

28.0

27.0

23.0

18.0

21.0

2.2

1.2

2.0

310.0

8.5

580.0

6.4

2.0

530.0

1.2

14.0

460.0

1.1

20.0

600.0

1.4

5.0

390.0

0.2

16.0

570.0

2.6

<1.0

590.0

1.4

10.0

460.0

1.4

13.0

510.0

0.9

18.0

600.0

2.6

4.0

650.0

1.5

1.0

0.8

4.9

16.0

6.0

2.2

560.0

560.0

390.0

1.3

9.4

535.0

3.2

16.0

640.0

2.2

14.0

680.0

1.4

3.0

600.0

0.5

1.0

2.0

490.0

520.0

17.0

19.0

20.0

21.0

20.0

25.0

18.0

18.0

23.0

23.0

0.6

0.5

9.0

0.4

14.0

0.2

2.0

590.0

3.0

355.0

2.0

250.0

9.0

440.0

16.0

14.0

0.3

10.0

460.0

0.2

10.0

425.0

23.0

0.4

17.0

680.0

14.0

0.4

5.0

380.0

23.0

28.0

21.0

16.0

20.0

0.3

1.3

0.3

0.2

0.4

0.2

0.2

0.4

0.2

0.2

0.2

0.2

0.3

14.0

10.0

15.0

22.0

19.0

19.0

8.0

15.0

23.0

20.0

18.0

16.0

5.0

580.0

530.0

550.0

650.0

610.0

560.0

620.0

590.0

620.0

610.0

630.0

480.0

460.0

7.1

20.0

<100.0

7.1

20.0

7.5

200.0

--

7.6

24.5

<100.0

7.6

23.0

100.0

7.4

19.0

7.4

7.3

7.3

7.8

7.6

7.6

7.4

40.0

7.6

7.6

7.4

24.0

7.4

24.0

23.0

28.0

36.5

33.0

7.4

22.0

7.1

21.5

22.0

400.0

100.0

500.0

100.0

100.0

200.0

300.0

190.0

--

300.0

90.0

<100.0

500.0

500.0

7.5

22.0

<100.0

7.6

100.0

7.5

21.0

<100.0

7.3

17.0

200.0

7.4

21.0

0.0

7.2

17.0

7.6

19.5

7.0

18.0

7.3

24.5

7.7

7.4

20.0

7.4

21.0

7.2

18.5

7.1

7.5

25.0

7.2

20.0

7.4

--

7.4

--

7.4

24.0

7.1

7.4

28.0

7.4

23.0

7.1

25.0

200.0

--

100.0

100.0

<20.0

200.0

30.0

200.0

30.0

100.0

20.0

30.0

390.0

--

100.0

400.0

352

Table G-5 - -Cbntinued

Sample

Number

49

50

51

52

43

44

45

46

47

48

53

54

[SO4']

(WO

[Cr]

(WO

[F

-

] []s10

3

-

]

(Prm) (PPm)

S.P.

Sian)

PH

Temp.

(°C)

[BP+]

(ppb)

21.0

120.0

0.3

16.0

5.0

0.5

2.5

320.0

1.0

250.0

16.0

19.0

17.0

110.0

11.0

12.0

52.0

32.0

12.0

260.0

130.0

22.0

7.0

9.0

7.0

25.0

9.0

9.0

18.0

9.0

0.4

2.0

250.0

0.3

11.0

380.0

0.4

1.6

225.0

0.4

0.5

0.4

0.4

0.5

0.6

0.5

3.0

4.0

3.0

2.0

2.0

1.0

1.0

560.0

260.0

260.0

365.0

340.0

970.0

280.0

7.5

19.0

7.7

7.4

18.5

7.6

17.0

22.0

200.0

20.0

--

200.0

7.3

7.7

100.0

7.9

18.0

<100.0

7.5

18.5

7.6

20.0

100.0

100.0

7.6

20.0

7.6

22.0

7.7

21.5

100.0

100.0

<10.0

353

Table G-6. San Bernardino Valley: Statistics of Chemical Variables

354

SiOs

Ca"

Mg++

Na+

K+

HCCs

-

SO

4

=

Cl

r

NOS

-

SP pH

Variable ?an

Std. Dey.

Minimum

Maximum

23.24

52.72

6.07

38.42

2.88

160.76

58.02

18.52

.93

9.10

492.40

7.43

7.77

18.91

3.45

22.84

.95

32.08

32.79

6.33

1.07

6.55

127.42

.21

10.0

9.0

1.0

12.0

1.0

98.0

11.0

5.0

.2

1.0

250.0

7.0

38.0

94.0

18.0

120.0

6.0

250.0

160.0

39.0

6.4

22.0

680.0

7.9

Number of Observations = 50

All units are ppm, except for SP (micrcmhos/cm) and pH (pH units)

Table G-7. San Bernardino Valley: Correlation Coefficients si0

2

ce+

mg++

Na r

Hco3-

Si0

2

Ca"

1.00000

.25763

1.00000

.49971

.66620

1.00000

Na

H00

3

-

SO

4

=

Cl

-

F

-

-.00161

-.37093

-.10298

1.00000

.21849

.06335

.11590

.21564

1.00000

.19024

.15980

.11411

.81456

.05481

.32646

.72348

.13404

.35613

-.13873

.78687

.63234

.01239

.11739

.27132

1.00000

.06704

.31446

.05455

-.44543

-.11249

.14677

.58358

.43602

.73359

-.00777

.42988

-.23443

.17453

.50154

NO

3

-

SP pH

.26330

.56866

.54810

-.19413

-.39527

-.29702

.49895

.06362

.24583

.59008

.04920

-.30783

355 so4=

C l-

F

-

NO3-

SP pH

SO

4

=

Cl

-

F

-

NO3

-

SP pH

1.00000

.82411

.49846

-.00586

.69830

-.10731

1.00000

.34791

.29540

.80221

-.25645

1.00000

-.20000

.20211

.05652

1.00000

.60434

-.32278

1.00000

-.27229

1.00000

356

LIST OF REFERENCES

Baer, N. S., and Lewin, S. Z., 1970, The replacement of calcite by fluorite:

A

kinetic study: Am. Min. 55, pp. 466-476.

Bahr, T. J., and Corbett, R. G., 1985, Formation of sodium bicarbonate ground water and the effect on fluoride solubility: Geological

Society of America Annual Meeting (Orlando, 1985), AbbLLacts with

PLuyiams 17(7), p. 517.

Barrow, N. J., and Shaw, T. C., 1977, The slow reactions between soil and anions: 6. Effect of time and temperature of contact on fluoride: Soil Sci. 124(5), pp. 265-278.

Barth, T. F., 1947, On the geodhemical cycle of fluorine: Jour. Geol.

50(5), pp. 420-426.

Bassett, R. L., and Wood, W. W., 1978, Computer simulation as an aid to interpreting the geochemistry of fluoride in the Ogallala aquifer,

Texas: Proceedings of the XII Symposium on Trace Substances in

Environmental Health (D. D. Hemphill, Ed.), pp. 89-98.

Berner, R. A., 1978, Rate control of mineral dissolution under earth surface conditions: Am. J. Sci. 278, pp. 1235-1252.

Berner, R. A., 1981, Kinetics of weathering and diagenesis, in Lasaga,

A. C., and Kirkpatrick, R. J., Editors, 1981, Kinetics of geochemical processes: Reviews in Mineralogy 8, Mineralogical Society of

America, pp. 111-134.

Bhakuni, T. S., and Sastry, S. A., 1977, Fluoride removal by anionexchange resins: Proceedings of the 1974 Symposium on Fluorosis,

Indian Academy of Geosciences, pp. 285-295.

Bower, C. A., and Hatcher, J. T., 1967, Adsorption of fluoride by soils and minerals: Soil Sci. 103(3), pp. 151-154.

Broene, H. H., and De Vries, T., 1947, The thermodynamics of aqueous hydrofluoric acid: J. Am. Chem. Soc. 69, pp. 1644-1646.

Browman, M. G., and Chesters, G., 1975, The solid-water interface: Am.

Chem. Soc. Environ, Chem. 15, pp. 10-13.

Carlston, C. W., 1942, Fluoride in the ground water of the Cretaceous area of Alabama: Geol. Surv. of Alabama Bull. 52, 67 p.

Carrera, J., Usunoff, E., and Szidarovszky, F., 1984, A method of optimal observation network design for groundwater management:

Jour. Hydr. 73, pp. 147-163.

357

LIST OF REFERENCES--Continued

Carrera Ramirez, J., and Samper Calvete, J., 1985, Curso sobre metodos geo-estadisticos aplicados a la Hidrologia Subterranea: Servicio de

Publicaciones de la Escuela Tecnica Superior de Ingenieros de

Ceminos, Cenales y Puer

tas de Valencia, 262 p.

Cederstrom, D. J., 1945, Geology and ground-water resources of the

Coastal Plain in southeastern Virginia: Virginia Geol. Surv. Bull.

63, 384 p.

Chhabra, R., Singh, A., and Abrol, I., 1980, Fluorine in sodic soils:

Soil Sci. Soc. Am. J. 44, pp. 33-36.

Christiansen, E. H., 1986, Fluorine and chlorine in granitoids from the

Basin and Range Province, Western United States: EOS 67(16), p.

401.

Corbett, R. G., and Manner, B. M., 1984, Fluoride in the ground water of northeastern Ohio: Ground Water 22(1): pp. 13-17.

Corey, R. B., 1981, Adsorption vs. precipitation, in Anderson, M. A., and Rubin, A. J., Editors, 1981, Adsorption of inorganics at solidliquid interfaces: Ann Arbor Science, pp. 161-182.

CUstodio, E., and Llamas, M. R., 1976, Hidrologia subterranea, Vol. 1:

Edic. Omega, 1157 p.

Dalton, M. G., and Upchurch, S. B., 1978, Intrepretation of hydrochemical facies by factor analysis: Ground Water 16(4), pp. 228-233.

David, M., Dagbert, M., and Beauchemin, Y., 1977, Statistical analysis in Geology: Correspondence analysis method: Quarterly of the

Colorado School of Mines 72(1), 60 p.

Davies, J. C., 1973, Statistics and data analysis in Geology: John

Wiley and Sons, 550 p.

Dawdy, D. R., and Feth, J. H., 1967, Applications of factor analysis in study of groundwater quality, Mbjave River Valley, California:

Water Resources Research 3(2), pp. 505-510.

DeWald, L. B., 1984, The occurrence of dissolved oxygen in ground waters of the upper San Pedro Basin, Cochise Cbunty, Arizona:

Unpublished M.S. Thesis, Dept. of Hydrology and Water Resources,

University of Arizona, Tucson, Arizona, 68 p.

Dickman, S. R., and Bray, R. H., 1941, Replacement of adsorbed phosphate from kaolinite by fluoride: Soil Sci. 52, pp. 263-273.

358

LIST OF REFERENCES-- Continued

Dobrovol'skiy, Ye. V., and Lyal'ko, V. I., 1983, Dynamics of groundwater fluoride: A model for the effects of kinetic and infiltration factors: Geochem. Internat. 20(4), pp. 68-81.

Durov, C. A., 1948, The geometrical method in hydrochemistry: Proc. Ac.

Sci. U.S.S.R. 59(1), pp. 87-96.

Eccles, L. A., and Klein, J. M., 1978, Distribution of dissolved nitrate and fluoride in ground water, Highland-East Highlands, San

Bernardino County, California: U. S. Geol. Surv. Water Resources

Invest. 78-14, 42 p.

Elrashidi, M. A., and Lindsay, W. L., 1985, Solubility relationships of fluorine minerals in soils: Soil Sci. Soc. Am. J. 49, pp. 1133-

1136.

Elrashidi, M. A., and Lindsay, W. L., 1986, Solubility of aluminum fluoride, fluorite, and fluorophlogopite minerals in soils: Soil

Sci. Soc. Am. J. 50, pp. 594-598.

Espenson, J. H., 1981, Chemical kinetics and reaction mechanisms:

McGraw-Hill Series in Advanced Chemistry, McGraw-Hill, 218 p.

Farrah, H., Slavek, J., and Pickering, W. F., 1985, Fluoride sorption by soil components: Calcium carbonate, humic acid, manganese dioxide and silica: Aust. J. Soil Res. 23, pp. 429-439.

Fleischer, M., and Robinson, W. O., 1963, Some problems of the geochemistry of fluorine: R. Soc. Can. Spec. Pap. 6, pp. 58-75.

Fluhler, H., Polomski, J., and Blaser, P., 1982, Retention and movement of fluoride in soils: J. Environ. Quai. 11(3), pp. 461-468.

Freeze, R. A., and Cherry, J. A., 1979, Groundwater: Prentice-Hall, 604

P-

Goldschmidt, V. M., 1954, Geochemistry: Oxford University Press, 730 p.

Gomez Artola, C., Borregon Martinez, J., Llamas Madurga, M. R., and

Sanchez Saez, J. J., 1983, Analisis preliminar del contenido de fluor de las aguas de Madrid y de su incidencia en las caries dentales de los nihos: III Simposio Nacional de Hidrogeologia, pp.

157-166.

Graham, G. S., Kester, S. E., and Van Loon, J. C., 1975, Fluorine in ground waters as a guide to Pb-Zn-Ba-F mineralization: Ec. Geol.

70(2): pp. 396-398.

359

LIST OF REFERENCES - -Continued

Greenacre, M. J., 1984, Theory and applications of correspondence analysis: Academic Press, 364 p.

Gupta, R. K., Chhabra, R., and Abrol, I., 1982, Fluorine adsorption behavior in alkali soils: Relative roles of pH and sodicity: Soil

Sci. 133(6), pp. 364-368.

Handa, B. K., 1975, Geochemistry and genesis of fluoride-containing ground-waters in India: Ground Water 13(3): pp. 275-281.

Handa, B. K., 1977, Presentation and interpretation of fluoride ion concentrations in natural waters: Proceedings of the 1974 Symposium on Fluorosis, Indian Academy of Geosciences, pp. 317-347.

Harman, H. H., 1976, Modern factor analysis

(3 rd

Ed.): The University of Chicago Press, 487 p.

Hart, E. B., Phillips, P. H., and Bchstedt, G., 1934, Relation of soil fertilization with superphosphates and rock phosphate to fluorine content of plants and drainage waters: Am. J. Pub. Health 24, pp.

936-940.

Hem, J. D., 1968, Graphical methods for studies of aqueous aluminum hydroxide, fluoride, and sulfate complexes: Geol. Surv. Water-

Supply Paper 1827-B, 33 p.

Hillel, D., 1982, Introduction to soil physics: Academic Press, 364 p.

Hingston, F. J., 1981, A review of anion adsorption, in Anderson, M.

A., and Rubin, A. J., Editors, 1981, Adsorption of inorganics at solid-liquid interfaces: Ann Arbor Science, pp. 51-89.

Hingston, F. J., Posner, A. M., and Quirk, J. P., 1972, Anion adsorption by goethite and gibbsite. I. The role of the proton in determining adsorption envelopes: Jour. Soil Sci. 23(3), pp. 177-192.

Hingston, F. J., Posner, X. M., and Quirk, J. P., 1974, Anion adsorption by goethite and gibbsite. II. Desorption of anions from hydrous oxide surfaces: Jour. Soil Sci. 25(1), pp. 16-26.

Hodge, H. C., and Smith, F. A., 1965, Fluorine chemistry, Vol. IV:

Academic Press, 789 p.

Huang, P. M., and Jackson, M. L., 1965, Mechanism of reaction of neutral fluoride solutions with layer silicates and oxides of soils: Soil Sci. Soc. Am. Proc. 29, pp. 661-665.

360

LIST OF =UMW - -Continued lier, R. K., 1979, The chemistry of silica: John Wiley and Sons, 866 p.

Jenkins, G. N., 1970, in Rahamatulla, Md., 1977, Dental fluorosis as a dental health problem: Proceedings of the 1974 Symposium on

Fluorosis, Indian Academy of Geosciences, pp. 487-495.

Johnson, R.

A.,

and Wichern, D. W., 1982, Applied multivariate statistical analysis: Prentice-Hall, 594

p.

Karunakaran, C., 1977, Fluorine bearing minerals in India--Their geology, mineralogy and geochemistry: Proceedings of the 1974

Symposium on Fluorosis, Indian Academy of Geosciences, pp. 3-18.

Kaiser, H. F., 1958, The varimax criterion for analytic rotation in factor analysis: Psychometrika 23, pp. 187-200.

Kister, L. R., 1966, Maps showing fluoride content and salinity of ground water in the Willcox basin, Graham and Cochise Counties,

Arizona: U. S. Geol. Surv. Hydrologic Investigations Atlas HA-214,

2 sheets.

Kister, L. R., and Hardt, W. F., 1966, Salinity of the ground water in

Western Pinal County, Arizona: Geol. Surv. Water-Supply Paper 1819-

E, 21 p.

Klovan, J. E., 1966, The use of factor analysis in determining depositional environments from grain-size distributions: Jour. Sedim.

Petrol. 36, pp. 115-125.

Kcnieczki,

A.

D., 1980, Maps showing ground-water conditions in the upper San Pedro Basin area, Santa Cruz and Cochise Counties,

Arizona, 1978: IL S. Geol. Surv. Wat. Res. Inv., Open-File Rep. 80-

1192, 71 p.

Kraynov, S. R., Ryzhenko, B. N., Shvarov, Y. V., Vasil'kova, I. V., and

Kraynova, L. P., 1980, Hydrodynamic conditions for computer simulation of physicochemical interactions in a water-rock system in hydrochernical forecasting: Application to

CaF

2

with solutions of various compositions: Geochem. Internat. 17(3), pp. 17-29.

Krishnamachari, K. A., 1977, Some new aspects of fluorosis in South

India--A new approach to prevent the problem: Proceedings of the

1974 Symposium on Fluorosis, Indian Academy of Geosciences, pp.

519-529.

Larsen, S., and Widdowson, A. E., 1971, Soil fluorine: Jour. Soil Sc!.

22(2), pp. 210-221.

361

LIST OF REFEREWES--Continued

Lasaga, A. C., 1981, Rate laws of chemical reactions, in Lasaga, A. C., and Kirkpatrick, R. J., Editors, 1981, Kinetics of geochemical processes: Reviews in Mineralogy 8, Mineralogical Society of

America, pp. 1-68.

Lyal'ko, V.I., and Dobrovol'skiy, Ye. V., 1976, page 71 in Dobrovol'skiy, Ye. V., and Lyaleko, V. I., 1983, Dynamics of ground-water fluoride: A model for the effects of kinetic and infiltration factors. Geochem. Internat. 20(4), pp. 68-81.

Lyman, W., Reehl, W., and Rosenblat, D., 1982, Handbook of chemical property estimation methods: McGraw-Hill, 960 p.

Malinowski, E. R., and Flowery, D. G., 1980, Factor analysis in Chemistry: John Wiley and Sons, 251 p.

Mansfield, G. R., 1940, The role of fluorine in phosphate deposition:

Am. Jour. Sci. 238, pp. 863-879.

Marion, G. M., Hendricks, D. M., Dutt, G. R., and Fuller, W., 1976,

Aluminum and silica solubility in soils: Soil Sci. 121(2), pp. 76-

84.

Mincguchi, G., 1977, The correlation of chronic toxic effects in tropical and subtropical areas between fluoride concentration in drinking water and climate, especially mean annual temperature:

Proceedings of the 1974 Symposium on Fluorosis, Indian Academy of

Geosciences, pp. 175-186.

Montgcrnery, E. L., 1963, The geology and ground water investigation of the Tres Alamos Dam Site area of the San Pedro River, Cochise

County, Arizona: Uhpublished M.S. Thesis, University of Arizona,

Tucson, Arizona, 61 p.

Morey, G. W., Fournier, R. O., and Rowe, J. J., 1962, The solubility of quartz in the temperature interval from 25°C to 300°C: Geoch. et

Oosmoch. Acta 26, pp. 1029-1043.

Mbrrill, L. G., Mahilum, B. C., and Mchluddin, S. H., 1982, Organic compounds in soils: Sorption, degradation, and persistence: Ann

Arbor Science, 326 p.

Mbrris, D. A., and Kulp, W. K., 1961, Mechanical uniform packing of porous media: U. S. Geol. Surv. Short Papers in the Geologic and

Hydrologic Sciences, Art. 305, pp. D31-D32.

362

LIST OF REFERENCES - -Continued

Morse, J. W., 1983, The kinetics of calcium carbonate dissolution and precipitation, in Reeder, R. J., Editor, 1983, Carbonates: Mineralogy and chemistry. Reviews in Mineralogy 11, Mineralogical Society of America, pp. 227-264.

Mbrshina, T. N., 1980, Fluorine adsorption by soils: Soviet Soil Sc!.

12, pp. 413-416.

Murray, F., 1983, Fluoride retention by sandy soils: Water, Air, and

Soil Poll. 20, pp. 361-367.

Murthy, D. S., and MUrthy, D. V., 1977, A study of high fluoride bearing waters around Nalgonda District, Andhra Pradesh: Proceedings of the 1974 Symposium on Fluorcsis, Indian Academy of Geosciences, pp. 311-315.

Myers, D., 1985, Notes from the Course MATH 579-2: Topics in Applied

Mathematics--Geostatistics, Fall Semester 1985: Dept. of Mathematics, University of Arizona, Tucson, Arizona.

Nataraj an, V., and Mohan Rao, V. R., 1977, Hydrochemical investigation for fluorine-bearing minerals in the Kangall and Hallia River

Basins, Nalgonda DistLict, Andhra Pradesh: Proceedings of the 1974

Symposium on Fluorosis, Indian Academy of Geosciences, pp. 37-47.

Neuman, S. P., 1984, Notes from the Seminar on Contaminant Transport

(HYDR 696fx), Fall Semester 1984: Dept. of Hydrology and Water

Resources, University of Arizona, Tucson, Arizona.

Omueti, J. A., and Jones, R. L., 1977, Fluoride adsorption by Illinois soils: Jour. Soil Sci. 28, pp. 564-572.

Orlob, G. T., and Radhakrishna, G. N., 1958, The effects of entrapped gases on the hydraulic characteristics of porous media: Trans. Am.

Geo. Union 39(4), pp. 648-659.

Orion Research, Incorporated, 1973, Instruction manual. Fluoride electrodes models 96-09 and 94-09: Orion Res., Inc., 28 p.

Paliwal, K. V., and Somani, L. L., 1977, Interactive effect of fluorine and salts in irrigation waters on the growth of wheat and peas:

Proceedings of the 1974 Symposium on Fluorosis, Indian Academy of

Gosciences, pp. 145-152.

Pathak, B. D., 1977, Occurr e nce of fluoride in groundwaters in Uttar

Pradesh: Proceedings of the 1974 Symposium on Fluorosis, Indian

Academy of Geosciences, pp. 237-247.

363

LIST OF REFERENCES - -COntinued

Peek, D. C., and Volk, V. V., 1985, Fluoride sorption in soils: Soil

Sci. Soc. Am. J. 49, pp. 583-586.

Perrott, K. W., Smith, B. F., and Inkson, R. H., 1976, The reaction of fluoride with soils and soil minerals: Jour. Soil Sci. 27, pp. 58-

67.

Piper, A. M., 1944, A graphic procedure in the geochemical interpretation of water analyses: Trans. Am. Geophys. Union 25, pp. 924-928.

Pittwell, L. R., 1974, Metals coordinated by ligands normally found in natural waters: Jour. Hydr. 21, pp. 301-304.

Posner, A. M., and Quirk, J. P., 1964, Adsorption of water from concentrated electrolyte solutions by mcntiorillonite and illite:

Proc. Roy. Soc. London A278, pp. 35-56.

Pytkowicz, R. M., 1983, Equilibria, nonequilibria, and natural waters:

John Wiley & Sons, 351 p.

Ramamohana Rao, N. V., and Rajyalakshami, K., 1977, Studies of water quality and incidence of fluorosis in Andhra Pradesh: Proceedings of the 1974 Symposium on Fluorosis, Indian Academy of Geosciences, pp. 273-284.

Ramamohana Rao, N. V., Rajyalakshami, K., Krishna Sastry, K. R., Subba

Rao, P., Krishna Rao, V. V., and Narasimha Rao, K., 1977, Incidence of fluorosis in drinking waters resources in Andhra Pradesh:

Proceedings of the 1974 Symposium on Fluorosis, Indian Academy of

Geosciences, pp. 227-235.

Rankama, K. R., and Sahama, T. G., 1950, Geochemistry: University of

Chicago Press, 912 p.

Rao, A. S., 1977, Hydrogeological studies in parts of Nandigama Taluk,

Krishna DisLiict, Andhra Pradesh, with special reference to the problem of fluorosis: Proceedings of the 1974 Symposium on Fluorosis, Indian Academy of Geosciences, pp. 209-216.

Rao, K. V., and Murty, B. V., 1977, Cn the possible role of defluoridating agents in treatment of fluorosis: Proceedings of the 1974

Symposium on Fluorosis, Indian Academy of Geosciences, pp. 471-475.

Roberson, C. E., and Barnes, R. B., 1978, Stability of fluoride complex with silica and its distribution in natural water systems: Chem.

Geol. 21, pp. 239-256.

364

LIST OF REFERENCES--Continued

Robertson, F. N., 1985, Occurrence and solubility controls of trace elements in ground water in alluvial basins of Arizona, in Anderson, T. W., and Johnson,

A.

I. (Editors), 1985, Regional aquifer systems of the United States. Southwest alluvial basins of Arizona:

American Water Resources Association Mbnograph Series 7, pp. 69-80.

Robinson, W. O., and Edgington, G., Fluorine in soils: Soil Sci. 61, pp. 341-353.

Roeske, R. H., and Werrell, W. L., 1973, Hydrologic conditions in the

San Pedro River Valley, Arizona, 1971: Arizona Water Commission

Bulletin 4, 76 p.

Rom, L. A., 1952, Ph.D. Dissertation: Chemistry Dept., University of

Wisconsin, in Romo, L. A., and Roy, R., 1957, Studies of the substitution of OH- by F- in various hydroxylic minerals: Am. Min.

42, pp. 165-177.

Rom°, L. A-, and Roy, R., 1957, Studies of the substitution of OH- by

F- in various hydroxylic minerals: Am. Min. 42, pp. 165-177.

Samson, H. R., 1952, Fluoride adsorption by clay minerals and hydrated alumina: Clay Min. Bull. 1, pp. 266-271.

Sarma, V. V., and Swamy, A. N., 1983, Common compositions characteristics of fluoride bering groundwaters: Water, Air, and Soil Poll.

20, pp. 29-39.

Schindler, P. W., 1981, Surface complexes at oxide-water interfaces, in

Anderson, M. A-, and Rubin, A. J., Editors, 1981, Adsorption of inorganics at solid-liquid interfaces: Ann Arbor Science, pp. 1-49.

Sdhoeman, J. J., and MacLeod, H., 1987, The effect of particle size and interfering ions on fluoride removal by activated alumina: Water SA

13(4), pp. 229-234.

Schwartz, C. E., and Smith, J. M., 1953, Flow distribution in packed beds: Ind. and Eng. Chem. 45(6), pp. 1209-1218.

Semmens, B., and Meggy, A. B., 1966a, The reaction of kaolin with fluorides. I. Effect of neutral and acid sodium fluoride solutions:

J. Appl. Chem. 16, pp. 122-125.

Semmens, B., and Mbggy, A. B., 1966b, The reaction of kaolin with fluorides. IL Effect of aqueous hydrofluoric acid: J. Appl. Chem.

16, pp. 125-128.

365

LIST OF REFERENCES—Ccntinued

Sharma, M. G., 1977, Incidence of fluoride in the natural waters of

Kanigiri Taluk, Prakasam District, Andhra Pradesh: Proceedings of the 1974 Symposium on Fluorosis, Indian Academy of Geosciences, pp.

301-306.

Slavek, J., Farrah, H., and Pickering, W. F., 1984, Interaction of clays with dilute fluoride solutions: Water, Air, and Soil Poll.

23, pp. 209-220.

Smith, H. V., and Cammack Smith, M., 1932, Mbttled enamel in Arizona and its correlation with the concentration of fluorides in water supplies: University cl Arizona Agricultural Experiment Station

Technical Bulletin 43, 287 p.

Smith, R. A., and Martell, A. E., 1976, Critical stability constants.

Vol. 4: Inorganic complexes: Plenum Press, 217 pp.

Spetch, R. C., and McIntire, W., 1961, Fixation, leaching, and plant uptake of fluorine from additions of certain fluorides in representative sandy soils: Soil Sci. 92, pp. 172-176.

Sposito, G., 1982, On the use of the Langmuir equation in the interpretation of "adsorption" phenomena: II. The "two-surface" Langmuir equation: Soil Sci. Soc. Am. J. 46, pp. 1147-1152.

Srikantia, S. G., 1977, Permissible levels of fluoride in drinking water: Proceedings of the 1974 Symposium on Fluorosis, Indian

Academy of Geosciences, pp. 407-409.

Stumm, IAL, and Mbrgan, J., 1981, Aquatic chemistry, 2nd Ed.: John Wiley and Sons, 780 p.

Travi, Y., 1984, Origene des fortes teneurs en fluor

des

eaux souterraines de l'aquifere paleocene de la region de Mbour (Senegal).

Role de l'ion magnesium: C. R. Acad. Sc. Paris, T. 298, Serie II,

No. 7, pp. 313-316.

U. S. Dept. of Health, Education, and Welfare, 1962, Drinking water standards: Public Health Service, 61 p.

U. S. Environmental Protection Agency, 1979, Methods for chemical analysis of water and wastes, 1978: Env. Mbnit. and Support Lab.,

Office Res. and Develop., U.S. E.P.S., 441 p.

Usunoff, E. J., 1984, Hydrochemistry of the San Pedro River basin near

Saint David, Cochise County, Arizona, with special emphasis on the behavior of fluoride: Unpublished M.S. Thesis, Dept. of Hydrology and Water Resources, University of Arizona, 191 p.

366

LIST OF REFERENCES--Continued

Usunoff, E. J., 1988, Aspectos practicos de experimentos con columnas sedimentarias: Bolet-in Hidrologico de la Univ. Autonoma de Cliihuahua, in press.

Usunoff, E. J., and Guzman-Guzman, A., 1988, Multivariate analysis in hydrochemistry: An example of the use of factor and correspondence analyses: Ground Water, in press.

Usunoff, E. J., and Maida, S., 1988, Elevadas concentraciones de fluoruro en aguas subterraneas: Caso estudiado en Arizona: Boletin

Hidrologico de la Univ. Autonoma de Chihuahua, in press.

van Genuchten, M. Th., 1981, Non-equilibrium transport parameters fiLm miscible displacement experiments: U. S. Salinity Lab. Res. Rep.

119, 88 p.

Veith, J. A., and Sposito, G., 1977, On the use of the Langmuir equation in the interpretation of "adsorption" phenomena: Soil Sci.

Soc. Am. J. 41, pp. 697-702.

Viswanadham, C. R., and Marty, V. S., 1977, Fluoride in gieundwaters of

Rajasthan and Gujarat: Proceedings of the 1974 Symposium on

Fluorosis, Indian Academy of Geosciences, pp. 187-200.

Wallace, D. E., and Cooper, L. R., 1970, Dispersion of naturallyoccurring ions in ground water from various rock types in a portion of the San Pedro River Basin, Arizona: Jour. Hydr. 10(4), pp. 391-

405.

Ward, J. C., 1966, Porosity in the vicinity of container walls:

Colorado State University Sanitary Paper No. 3, pp. 1-26.

WP-Ast, R. C., Editor, 1984, CRC Handbook of Chemistry and Physics, 65 th

Ed.: CRC Press.

Westall, J. C., Zachary, J. L., and Mbrel, M. M., 1976, MINEQL, a computer program for the calculation of chemical equilibrium composition of aqueous systems: Parsons Laboratory, Massachusetts

Institute of Technology, Tech. Note 18, 91 p.

Zack, A. L., 1980, Geochemistry of fluoride in the Black Creek aquifer system of Harry and Georgetown Counties, South Carolina--and its physiological implications: U. S. Geol. Surv. Water-Supply Paper

2067, 40 p.

Ziauddin, M., 1977, Endemic fluorosis in parts of Dharward and Raichur

Districts, Karnataka State: Proceedings of the 1974 Symposium on

Fluorosis, Indian Academy of Geosciences, pp. 85-92.

Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement