Modeling of a DR Shaft Operated with Pure Hydrogen Using

Proceedings of the 4th Ulcos seminar, 1-2 October 2008
SP12 - New Direct Reduction / n°4-8
Modeling of a DR Shaft Operated with Pure Hydrogen
Using a Physical-chemical and CFD Approach
A. Ranzani da Costa, D. Wagner, F. Patisson and D. Ablitzer
LSG2M, INPL, CNRS, Nancy-Université, Nancy, France
The hydrogen-based route could be a valuable way to produce steel considering its low carbon dioxide emissions. In ULCOS, it is regarded as a long-term option, largely dependent on the emergence of a hydrogen economy. To anticipate its possible development, it was decided to check the feasibility of using 100% H2 in a Direct
Reduction shaft furnace and to determine the best operating conditions, through appropriate experimental and
modelling work.
We developed from scratch a new model, called REDUCTOR, for simulating this process and predicting its performance. This sophisticated numerical model is based on the mathematical description of the detailed physical,
chemical and thermal phenomena occurring. In particular, kinetics were derived from experiments. The current
version is suited to the reduction with pure hydrogen, but an extension of the model to CO is planned so that it
will also be adapted to the simulation and optimisation of the current DR processes.
First results have confirmed that the reduction with hydrogen is much faster than that with CO, making it possible to design a hydrogen-operated shaft reactor quite smaller than current MIDREX and HYL.
Nancy, France, to check the feasibility of the reduction of iron ore by pure hydrogen and to define the
best processing conditions for the direct reduction in
a shaft furnace, through appropriate experimental
and modelling work.
Breakthrough ways for making steel are investigated
within the ULCOS (Ultra low CO2 steelmaking) European program [1], with the target of at least 50 %
reduction in CO2 emissions as compared to these of
the current integrated steel plants, i.e. 1844 kg of
CO2/t of HRC (hot rolled coil). The use of hydrogen
as a reducing agent, instead of CO, with an associated production of H2O instead of CO2, was evaluated in ULCOS SP4 (Hydrogen-based steelmaking)
and is further studied in SP12 (Advanced direct reduction), as well as by other groups in the world [2].
The ULCOS SP4’s preferred route to steel is shown in
Figure 1. Its performance regarding CO2 emission is
promising: down to less than 300 kg CO2/t HRC, if
hydrogen is produced by water electrolysis using
hydraulic or nuclear electricity, without any CO2
capture necessary [3].
Although the reduction of iron ore by hydrogen was
extensively studied in the 1970s and 80s (e.g. [4-7]),
few publications report kinetic expressions suited to
the reduction of industrial pellets and directly usable
for modelling. To clarify the mechanisms of the reaction and get accurate kinetic data for reactor modelling, we undertook a series of reduction experiments
by thermogravimetry, completed by sample characterization by XRD (X-ray diffraction), SEM (Scanning
electron microscopy), and Mössbauer spectrometry.
The thermobalance used was a Setaram Tag24
model, with two symmetrical furnaces, which presents the advantage of eliminating the disruptive
effects of buoyancy and drag forces, ensuring µg
accuracy. A specific steam generator was attached to
possibly add controlled water content to the reaction
gas. Samples were small hematite cubes (5-mm side,
550-mg weight) shaped in CVRD industrial pellets.
The reason for making such cubes was the limiting
total weight loss imposed by the balance. Reduction
experiments were run under isothermal conditions: a
single cube was wrapped in a Pt wire and directly
hung to the balance beam; when the desired temperature was reached and stabilized, H2 was introduced into the system and maintained until the end
of the reaction. Gas compositions were 100 % H2
and 60 vol. % H2 in He. In some cases, up to 4
vol. % H2O were added to study the effect of water
in the reducing gas. Smaller cubes were also alterna-
Figure 1. Hydrogen-based route to steel
However, the future development of such a process
heavily depends on the emergence of the so-called
hydrogen economy, which could result from the
demand of other industrial sectors, like energy and
transportation industries. It is here assumed that H2
would become available in this frame, in large quantities, at competitive cost, and with low CO2 emission
for its production. We believe that steelmakers
should anticipate and be ready to take profit of such
a situation by using hydrogen to a large extent. The
present research was performed, at the University of
tively used. Temperature varied from 600 °C to
990 °C. Its effect (Figure 2) deserves attention and
is discussed below.
evolution (Figure 3). The hematite pellet is made up
of large, dense grains, separated by thin cracks. The
grainy structure only changes slightly from hematite
to magnetite and wustite. The main point is the
appearance and growth of pores on the sides of the
grains. At the stage of wustite, however, crosssections of grains (bottom right photo) show that
these become very porous and sub-divided into
smaller grains that we termed “crystallites”. From
wustite to iron, the solid structure changes dramatically. Iron forms larger, dense grains. Yet the final
inter-grain porosity remains high.
Figure 3. Morphological evolution of the solid during the
reduction of hematite cubes at 800 °C, 60 % H2 in He
Single pellet model
A mathematical model simulating the reduction of a
single pellet was designed as a link between the
experiments and the reactor model. Used independently, it helps to understand, analyze and simulate
the reduction experiments. Used as a subroutine of
the reactor model, it predicts the reaction rate as a
function of the local reduction conditions (temperature and gas composition). This pellet model was
built according to the experimental findings. The
pellet (typically 12-mm diameter) is viewed as a
cluster of grains (25 µm diameter), with an intergranular porosity of about 10 %. Grains are themselves porous (50% intra-grain porosity) and, in the
case of wustite, assumed to be composed of dense
crystallites (2 µm) separated by thin pores.
Figure 2. Influence of temperature on the reduction curves
of hematite cubes, under 60 % H2 in He
For experiment temperatures up to 800 °C, the
higher the temperature, the faster the reaction,
except at 700 °C where the reaction is slower. Above
800 °C, an increase in temperature first accelerates
the reaction rate, but, when reaching a high conversion, the reaction slows down and goes on at a lower
rate, the total reaction time being longer than at
800 °C. Thus, from these experiments, 800 °C appears as an optimum temperature for achieving a full
conversion in minimum time.
To keep the mathematical description simple, which
is quite desirable for subsequent integration in REDUCTOR, we decided to use the concept known as
the law of additive reaction times [8] for calculating
the overall reaction rate. This law considers that
mass transfer processes take place in series so that
mass transfer resistances, represented by characteristic times, can be added. A single, complicated but
analytical, relationship can be derived for the reaction rate. The law is approximate but its validity was
demonstrated for several gas-solid systems [9]. Its
top interest is its ability to represent intermediate
(mixed) kinetic regimes in a closed-form equation,
very fast to compute as a subroutine of larger reactor-scale codes.
A series of interrupted experiments, where the hydrogen flow was stopped before complete reduction,
were carried out to allow for the characterization of
samples at different degrees of conversion. Partly
reduced samples were then observed by SEM and
analyzed by Mössbauer spectrometry and X-ray diffraction. These experiments helped reveal the course
of the reduction, through the formation of intermediate oxides and the corresponding morphological
A. Ranzani da Costa, D. Wagner, F. Patisson and D. Ablitzer
The detailed equations used for calculating the different characteristic times and the corresponding
reaction rates are given in [10]. The following phenomena are taken into account: mass transfer of H2
SP12 – ULCOS-4, October 2008
and H2O through the boundary layer surrounding the
pellet, gas diffusion in the inter-grain pores, in the
intra-grain pores, and in the pores of the iron layer
around the crystallites, the last two ones involving
Knudsen-type diffusion, solid phase diffusion of oxygen through the iron layer when it is dense, the
three heterogeneous reactions of reduction,
equations considered are the local mass, energy, and
momentum balances written below. Notation: r =
radius, z = height, ct = total gas concentration, xi =
molar fraction of i in gas, ug = superficial gas velocity, Da and Dr = axial and radial dispersion coefficients, r1, r2, r3 = rates (mol s-1 m-3) of the three
reduction reactions: hematite → magnetite, magnetite → wustite and wustite → iron, wj = mass fraction
of j in solid, Mi = molar weight of i, ρg = density of
the gas, ρs = apparent density of the solid bed, cpg,
cps = specific heat of the gas and solid, λg = thermal
conductivity of the gas, Tg, Ts = temperatures of the
gas and solid, ag = specific surface area of the bed,
h = heat transfer coefficient, λeff,a , λeff,r = axial and
radial effective conductivities of the solid bed, ΔrHi =
heat of reaction i, p = gas pressure, ε= inter-pellet
porosity of the bed, dp = diameter of the pellets, μg =
viscosity of the gas.
3Fe2 O3 + H2 = 2Fe 3 O 4 + H2 O
H =
Fe 0.95 O +
19 2 19
19 2
Fe 0.95 O + H2 = 0.95 Fe + H2 O
Fe 3 O4 +
and possible sintering of the iron phase.
The description of this last phenomenon is an original feature of our model, introduced to reflect two
experimental observations: first, the reaction rate
decrease noted (Figure 2, top) at high conversions
and at temperatures greater than 850 °C; and second, the growth and densification of the iron grains
(not shown) above the same temperature. We explain this behaviour by the tendency of the freshly
formed iron layer, at the level of the crystallites, to
sinter, i.e. to decrease its surface area. As a result,
the intra-crystallite pores get thinner, which makes
gas diffusion through these pores more difficult, and
the pores eventually disappear, which makes solid
diffusion, a slower process, the only possible mass
transfer process for the rest of the reaction. The
corresponding equations in the model are again
given in [10].
Figure 4 gives the main results of the pellet model.
The comparison with the experiment (top graph)
shows that the model satisfactorily simulates the
experiment. The slowing down at 80% conversion
can be noted. The succession of the three reactions
(hematite → magnetite → wustite → iron, bottom
graph) is illustrated. Hematite disappears rapidly,
whereas the reduction of wustite is the longest step.
REDUCTOR: model of a shaft furnace operated with 100% H2
To predict the performance of the reactor, for a
process that does not yet exist, mathematical modelling seemed to be the best suited approach. It can
give access to all of the relevant variables of the
process (local solid and gas temperatures, compositions and velocities, reactions rates, conversion,
etc.). Our modelling approach is similar to CFD
(computational fluid dynamics): it is based on the
numerical solution of the local mass, energy and
momentum balances. Therefore, beyond a global
assessment of the process like obtained from flowsheeting softwares, the detailed, spatially-resolved
behaviour of the reactor can be known. Moreover,
the influence of operating conditions and physical
parameters can be analyzed quantitatively and optimal conditions derived.
Figure 4. Results of the single pellet model, 900°C, 60 %
H2 in He. Top: comparison with experiment; bottom: evolution of the mass fractions of the different oxides and iron.
Mass balances for the gases:
REDUCTOR is a 2-dimensional (cylindrical coordinates), steady-state, numerical model. The principal
A. Ranzani da Costa, D. Wagner, F. Patisson and D. Ablitzer
SP12 – ULCOS-4, October 2008
)+ ∂ (c x u )=
1 ∂ rc t x i u gr
system is iteratively solved using the Gauss-Seidel
algorithm. Meshing involves 201 (vertical) × 21 (radial) cells. The code is written in Fortran 90. Computations were run using a PC cluster (16 processors
AMD Opteron) and lasted typically 40 h.
∂x i ⎞
∂x i ⎞
1 ∂ ⎛
∂ ⎛
⎜ rc t Dr
⎜ ct Da
⎟ + Si
r ∂r ⎝
∂r ⎠ ∂z ⎝
∂z ⎠
with i = H2 or H2 O and − S H = S H O = r1 +
A reference case was first simulated. The domain
modelled corresponds only to the cylindrical section
of the shaft furnace, where the reduction takes
place, above the lateral reduction gas inlet. Figure 5
gives the geometry of the furnace modelled and the
inlet flows. The solid flow rate corresponds to a
production of 1 MtFe yr-1, and the total gas flow rate
(3734 mol s-1) is 3.8 times the stoichiometric one.
r2 + r3
Mass balances for the solids:
∂ ρs u sw j
)= S
with j =Fe2 O3 , Fe 3 O 4 , Fe 0.95 O, or Fe and
S Fe O = −3MFe O r1
2 3
2 3
S Fe O = MFe O 2r1 − r2
3 4
S Fe
3 4
= MFe
⎛ 60
⎜ r2 − r3 ⎟
⎝ 19
S Fe = 0.95 MFe r3
Heat balance of the gas:
ρg c pg ⎜ u gr
+ u gz
+ag h T s − T g
∂Tg ⎞ 1 ∂ ⎛
∂Tg ⎞ ∂
⎟ =
⎜ r λg
∂z ⎟⎠ r ∂r ⎜⎝
∂r ⎟⎠ ∂z
− ⎜ r1 +
r + r ⎟ c − c PH O
19 2 3 ⎠ PH2
⎛ ∂T ⎞
⎜ λg
∂z ⎟⎠
− Ts
Heat balance of the solid:
− ρs u s c ps
∂Ts ⎞ ∂ ⎛
∂Ts ⎞
1 ∂ ⎛
⎜ r λeff ,r
⎜ λeff ,a
r ∂r ⎝
∂r ⎠ ∂z ⎝
∂z ⎠
) (
) (
ag h Tg − Ts + r1 −Δ r H1 + r2 −Δ r H2 +
r3 −Δ r H 3
+ ⎜ r1 +
r2 + r3 ⎟ c PH − c PH O Tg − Ts
Momentum equation combined with gas continuity
Figure 5. Conditions of computation in the reference case
Figure 6 shows the calculated mass fractions of the
different oxides and iron.
1 ∂ ⎛ c t ∂p ⎞ ∂ ⎛ c t ∂p ⎞
⎟ =0
r ∂r ⎝ K ∂r ⎠ ∂z ⎝ K ∂z ⎠
(1 − ε ) μ
with K = 150
ε dp
+ 1.75
(1 − ε )ρ u
ε 3d p
The reaction rates of the three reactions are those
determined according to the single pellet model,
which are derived from the kinetic experiments.
These partial derivative equations are complemented
by the set of appropriate boundary conditions
(known inlet temperatures and compositions, zero
fluxes at the symmetry axis and wall, etc.). All details
are given in [10], together with the correlations used
for calculating the various physical, chemical, and
thermal parameters.
The numerical solution is that of the finite volume
method [11]: the partial derivative equations are
rendered discrete and the resulting algebraic linear
A. Ranzani da Costa, D. Wagner, F. Patisson and D. Ablitzer
SP12 – ULCOS-4, October 2008
Introducing a higher water content in the gas decreases the driving force of the (reversible) reactions
and thus the kinetics. The map of iron mass fractions
with 10% water (Figure 7 right), when compared
with that of the reference case where only 2% water
were present in the injected reduction gas, shows
that the reaction is not complete, in particular in the
centre region.
Figure 8. Calculated mass fractions of iron for different
pellet sizes: dp=6 mm (left) and dp =24 mm (right)
When changing the pellet size, with 24-mm diameter
pellets, the conversion is not complete (75 % Fe on
average at outlet), whereas with 6-mm pellets, a full
conversion is obtained only 2 m downstream from
the solid inlet, which can be compared to 4 m with
the 12-mm pellets of the reference case. Therefore,
decreasing the pellet size seems a quite attractive
option for the industrial practice.
Figure 6. Calculated mass fractions of the solid species in
the reference case: gas in 800 °C, 98 %H2, dp =12 mm
Similar to the single pellet model, reductions of
hematite and magnetite are fast. That of wustite
takes longer but, at z=2 m, all oxides are fully converted to iron, irrespective of the radial position (in
this simulation). This is an important result because
it shows that about 4 m in height are required to
obtain 100% Fe, whereas usual DR furnaces operated with syngas (CO+H2) are typically 9 m high
(with the same 6.6-m diameter for a classical
MIDREX furnace) and only permits one to attain
92 % Fe.
REDUCTOR, a new, steady-state, 2-dimensional,
CFD-type model of a shaft furnace operated with
pure hydrogen has been developed. It describes gas
and solid flows, mass and heat transfers, and the
three-reaction reduction of hematite to iron. Kinetics
are expressed using a specific single pellet model
based on the law of additive reaction times and
experimental data. First results confirm the high
potential of hydrogen direct reduction: CO2 emission
of the reactor itself is zero and, due to the faster
kinetics of iron oxides reduction with hydrogen (compared with CO), a full conversion to iron can be
obtained with a much smaller reactor (typically half
the size) than usual MIDREX or HYL direct reduction
furnaces. Future work on this model will consist in
taking into account CO, and the related chemical
reactions, to be able to simulate current DR processes and thus to validate the model against plant
Other simulations were conducted to study the influence of operating conditions and physical parameters. For sake of brevity of the paper, only the effects of the water content of the inlet gas and of the
pellet size will be reported here (Figure 7 and 8).
Figure 7. Calculated molar fractions of water vapour in the
gas (left) and mass fraction of iron in the solid (right),
when introducing 10% H2O in the gas
A. Ranzani da Costa, D. Wagner, F. Patisson and D. Ablitzer
SP12 – ULCOS-4, October 2008
The authors thank all the ULCOS SP4 and SP12 partners for interesting exchanges on this piece of work,
and Dr. J.P. Birat, ULCOS chairman, for his continuous support.
The present work is part of the ULCOS program,
which operates with direct financing from its 48
partners, especially of its core members (ArcelorMittal, Corus, TKS, Riva, Voestalpine, LKAB, Saarstahl, Dillinger Hütte, SSAB, Ruukki and Statoil), and
has received grants from the European Commission
under the 6th Framework RTD program and the RFCS
[2] H.Y. Sohn, “Suspension Hydrogen Reduction of Iron
Oxide Concentrate”, 2008, US DoE report, DE-FC3697ID13554,
[3] ULCOS SP4 conclusions, ULCOS internal report, 2006.
[4] E. T. Turkdogan and J. V. Vinters, "Gaseous Reduction
of Iron Oxides.1. Reduction of Hematite in Hydrogen",
Metallurgical Transactions, 2 (11) (1971), 3175-3188.
[5] D. H. St John and P. C. Hayes, "Microstructural Features
Produced by the Reduction of Wustite in H2/H2O GasMixtures", Metallurgical Transactions B-Process Metallurgy,
13 (1) (1982), 117-124.
[6] M. Moukassi, P. Steinmetz, B. Dupre and C. Gleitzer,
"Mechanism of reduction with hydrogen of pure wustite
single crystals", Metallurgical Transactions B: Process Metallurgy, 14B (1) (1983), 125-32.
[7] Towhidi, N. and J. Szekely, "The influence of carbon
deposition on the reduction kinetics of commercial grade
hematite pellets with CO, H2, and N2" Metallurgical Transactions B (14) (1983), 3, 359-367.
[8] Sohn, H. Y., "The law of additive reaction times in fluidsolid reactions" Metallurgical Transactions (9B) (1978), 8996.
[9] F. Patisson, B. Dussoubs, and D. Ablitzer “Using Sohn’s
law of additive reaction times for modeling a multiparticle
reactor. The case of the moving bed furnace converting
uranium trioxide into tetrafluoride”, Sohn International
Symposium “Advanced processing of metals and materials”,
27-31 Aug. 06, San Diego. Proceedings edited by F. Kongoli
and R.G. Reddy, TMS, vol. 1 “Thermo and physicochemical
principles: non-ferrous high-temperature processing”, pp.
[10] D. Wagner, “Etude expérimentale et modélisation de la
réduction du minerai de fer par l’hydrogène”, PhD dissertation,
[11] Patankar, S. V. (1980), “Numerical heat transfer and
fluid flow”. New York, Hemisphere Publishing Corp.
Priority 3 of the 6th Framework Programme in the area of “Very low CO2 Steel
Processes”, in co-ordination with the 2003 and 2004 calls of the Research Fund
for Coal and Steel
A. Ranzani da Costa, D. Wagner, F. Patisson and D. Ablitzer
SP12 – ULCOS-4, October 2008