nonanoyed to beam GaAs

nonanoyed to beam GaAs
Low-resDstance nonanoyed ohmuc contacts to Si-doped
beam epitaxia~ GaAs
P. D. Kirchner, T. N. Jackson, G. D. Pettit, and J. M. Woodall
IBM T.J. Watson Research Center. Yorktown Heights. New York 10598
(Received 4 March 1985; accepted for publication 8 April 1985)
We have found evidence that the surface depletion charge density in molecular beam epitaxial nGaAs doped heavily with Si approaches the Si concentration. In situ metallization of the as-grown
surface of GaAs uniformly doped with Si at 1 X 10 20 cm - 3 yield.s a specific contact resistivity of
1.3/1fl cm 2 , indicating a space-charge density about equal to the silicon density despite a
measured bulk electron density of 4 X 10 18 cm -3. This contact resistivity is among the lowest for
nonaHoyed ohmic contacts to n-GaAs. We attribute the large discrepancy between surface spacecharge density and bulk electron density to the amphoteric behavior of silicon in GaAs. Surface
Fermi-level pinning and arsenic stabilization create a surface depletion region where donor site
selection predominates, whereas the extrinsic electron density in the bulk causes selfcompensation.
Bulk thermochemistry establishes that for a given crystal growth condition, an amphoteric dopant such as silicon
in GaAs has an equilibrium ratio of donors to acceptors that
is proportional to the ratio of holes to electrons. I Consequently, carrier density saturates at high dopant concentrations by causing the ratio of donors to acceptors to approach
unity. However, in GaAs, a wide variety of"conditions pins
the surface Fermi level near midgap, depleting the near-surface region of free carriers and fixing the hole-to-electron
ratio there near unity. Thus, if Fermi-level pinning occurs at
the surface during growth, the equilibrium donor-to-acceptor ratio in the surface depletion region does not depend
upon dopant concentration, and saturation of the surface
space-charge density should not occur. This is important because most metal contacts to GaAs produce midgap pinning
which, in order to exhibit low-resistance ohmic behavior,
requires the depleted GaAs at the metal interface to have a
high space-charge density.
In this letter, we report that for molecular beam epitaxy
(MBE) of(I00) GaAs doped with Si, measured bulk electron
concentrations and surface space-charge densities that are
inferred through contact resistance indicate that initial selection of donor and acceptor sites by Si atoms is dominated
by the effects of arsenic stabilization and the consequent Fermi-nevel pinning. In situ metaIlization ofGaAs doped with Si
at 1 X 1020 cm- 3 yields a specific contact resistivity of 1.3
/1fl cm 2 at 300 K. The solubility limit of Si in GaAs of
- 2 X HY o cm -3 (Ref. 2) suggests that a contact resistivity as
low as 0.5/1fl cm 2 might be achieved.
Molecular beam epitaxial GaAs doped with Si was
grown with elemental sources at a growth rate of 1/1/h using
semi-insulating (1 (0) GaAs substrates held at 550-61O·C
with AS 4 or AS 2 pressures roughly ten times the Ga pressure.
Silicon was subliimed. from an etched silicon wafer with
many parallel filaments.~ Electrically, the source resembles
the single filament used by Miller and co-workers,4.5 because
thermal runaway causes only one filament to heat. This
source allows high-purity Si doping at high growth rates.
For growth of GaAs at micron-per-hour rates, the conventional Si effusion ceIJ is unsuitable for doping at percent levels, because the high temperatures required decompose its
Appl. Phys. Lett. 47 (1), 1 July 1985
boron nitride crucible and insulators.
The layers were doped uniformly to allow both bulk
and surface characterization. The silicon content was measured prior to growth with a mass spectrometer at the substrate's growth position. Following growth, the silicon density in heavily doped layers was verified by electron
microprobe. Bulk carrier concentrations and mobHities were
measured by Hall effect. Surface space-charge density was
inferred from the specific contact resistivity of nonaUoyed
metallurgy applied in situ folIowing growth. Ag was chosen
for its lack of reaction with GaAs and ease of patterning into
transmission lines. 6 Actual contact spacings, measured by
optical microscopy, were the major source of uncertainty in
the resistivity measurement. Surface space-charge density
was estimated by reference to the work of Chang et al. 7 and
Schroder and Meier.8 For an effective mass m *, Richardson
constant A **, Boltzmann constant k, temperature T,
Planck's constant h, a net space-charge density (ND - NA. ) of
N, electron chargeq, permittivity E, and barrier <P, the specific contact resistivity Pc is
Pc o::::=.(kIqTA U)exp[(21T<Plh )~(Em*IN)].
The barrier <P is reduced from the Itow-field barrier height t/J
by 4 Jq3 Nt/J -IS?€'". This expression for Pc is approximate,
and the temperature dependence of the pre-exponential factor is uncertain.
Bulk electron concentrations appear in Fig. 1 as triangular symbols. Up to _10 19 em -3, the electron density is
nearly equal: to the silicon density, as noted by others,s,9 from
mobilities similar to non amphoteric dopants. 1O Electron
density did not vary detectably with substrate temperature,
arsenic species, or arsenic pressure, although Ga-rich
growth was not examined. Observed variations" may be due
to unintentional impurities. II At silicon densities above
0::::=.10 19 cm --\ the electron concentration peaks and drops to
a level dependent upon substrate temperature.
The theoretical and actual values of Pc are presented in
Fig. 2. The square symbol shows the result of Barnes and
ChO l2 forSn doping. The measured vaJ!ues ofPc for the in situ
Ag contacts are plotted versus silicon density using circles.
Silicon densities of 0.4, 1, and 2 X 1020 cm -3 yield Pc of
@ 1985 American Institute of Physics
Si Doping of MBE GaAs (100)
1urine, Resistance vs. Doping
LSi] or (No-N A) 10 ,9 cm- 3
~ 1
10 20
S Ts~550C
1 Ts ;610C
Si doping (This Work)
, a,a
(50lubihty Limit)
cm- 3
8x10- 11
FIG. 1. Electron density and surface space-charge density of MBE (1001
GaAs IS plotted vs silicon concentration. The Hall electron density is shown
f?r layers grown at 550 (V) and 610 'C 161. The surface space-charge densIty tnferred from the specific resistivity of nonalloyed contacts (0) is shown
from Fig. 2.
5( ± 2), l.3( ± 0.7), and l.5(:c 1) f,J.fl cm~, respectively.
Within the accuracy of both measurement and theory, the
resistivities obtained correspond with the theoretical values
for a 0.7-D.8-eV barrier if the silicon density and the spacecharge density are equivalent. The lowest resistivity indicates a space-charge density of ~ 1 X 1020 cm". which exceeds the highest electron concentration obtained in GaAs
with any dopant 12 and exceeds by tenfold the maximum electron density obtained in bulk GaAs with Si doping.
The effects of native oxide formation due to air exposure confirmed the presence of a high space-charge density
layer confined to the near-surface region. Cr was evaporated
following an HCI rinse oflayers that had been exposed to air
for several days. A layer with a Si density of 5 X 10'9 cm- 3
yielded Pc = 22( ± 2) J.Lfl cm z, several times that expected
for a similar in situ contact. The resistance of any residual
oxide only partly explains the increased Pc, because higher Si
densities produced Pc above 100 pfl cm 2 , typical of bulk Si
doping limits. Ifhigh space-charge density exists only within
the original surface dep.letion zone, then oxidation of the top
few nanometers of GaAs extends the depletion zone into low
space-charge density material, increasing Pc' For the
5X 10'9 cm- 3 Si density, oxidation consumes some of the
high space-charge density layer, yielding Pc lower than bulk
doping yields but higher than a similar in situ contact. At
higher silicon densities, all of the high space-charge density
layer is destroyed, exposing bulk material that exhibits high
contact resistivity.
Silicon can be a donor or acceptor in GaAs. At equilibrium, the ratio of donors to acceptors is given by the expression
= SiG./SiAs
k (T)Xlpln)XP AS "
where N D and NA are the donor and acceptor concentrations, k (T) is a temperature-dependent constant incorporating the equilibrium arsenic pressure over GaAs, pin is the
carrier concentration ratio, and PAs, is the arsenic dimer
pressure. At a given temperature and 'arsenic pressure, if the
GaAs is intrinsic, the ratio of donors to acceptors is constant
becausepln~ 1. MBE ofGaAs commonly uses an excess As
flux, yielding a e(2 X 4) reconstruction, which Katnani et
Appl. Phys. Lett., Vol. 47, No.1, 1 July 1985
10 17 ,":'7-'--'-~~~-L-l.~:;;-'---'-'-'--~-"----'---'10 17
10 '9
o Sn doping (Barnes and Cho)
10 '9
o No-NA from Contact Resistivity
1.2xIO- IO
1.6xlO- lo
2x10- 10
(N -N A)-1/2 cm 3 / 2
FIG. 2. Specific contact resistivity Pc for nonalloyed ohmic contacts to n·
GaAs is plotted vs the reciprocal square root of the space-charge density.
The lines show theoretical (see Refs. 7, 8)p, for 0.7- and O.8-eV barriers. (0)
The result of Barnes and Cho lsee Ref. 12) showing the accuracy of the
theory. 10) Pc for our contacts with lSi] substituted for space-charge density.
The agreement with theory shows that surface space-charge density in these
samples is equivalent to the silicon density.
01. 13 show yields Ec-Ef~O. 8 eV. Free electrons are depleted
from the surface region, where initial dopant site selection
occurs. In this depletion region, the pin ratio is near 1 and,
therefore, in this region the equilibrium NDINA should be
that of intrinsic material even though the bulk may be extrinsic. Therefore, for conditions that create low-compensation
n-type material at low dopant concentration, the surface
equilibrium N DINA should be large for any dopant concentration.
Kinetic factors are expected to prevent equilibration in
MBE. However, the typical micron-scale surface mobility of
Ga adatoms and the low density of stoichiometric defects in
MBE GaAs show that surface kinetics may allow some reactions to reach equilibrium. At high silicon concentrations,
initially depleted GaAs with high N DINA encounters bulk
conditions as more material is grown atop it. To reach bulk
equilibrium where NDINA is near 1, up to nearly half the
silicon atoms must switch from donor to acceptor sites. The
Ga vacancies and As interstitials produced must diffuse to
the surface for site switching to continue. Their diffusivity is
enhanced by proximity to the surface by exponentially increasing near-surface vacancy concentrations and linearly
increasing concentration gradients. Therefore. the kinetics
of site switching can approach surface rates at high silicon
concentrations because the high space-charge density
shrinks the surface depletion layer so that bulk electron concentrations occur only a few nanometers from the surface.
The kinetic barrier to equilibration can be seen in Fig. 1,
where the electron density peaks near 10 19 cm -3. Increasing
the Si density shrinks the surface depletion layer, increasing
site switching so that the electron density drops to its temperature-dependent maximum equilibrium value.
The low contact resistances obtained (1.3-5 J.Lfl cm l )
are inconsistent with the bulk doping behavior of silicon in
GaAs. However, Miller et 01.4.5 also used silicon doping at
high concentrations for MBE tunnel diodes and obtained
unexpectedly high conductances. Both their diodes and our
contacts have Si-doped GaAs that is depleted of free e1ecKirchner et al.
trons during growth due to the p-n junction and surface potentials. Among alternative explanations, a deep donor at
high density would yield both a high space-charge density
and low electron density, but it is unlikely that the deep level
would be as tightly confined to the surface region as the airexposure experiments indicate. Similarly, a substantial reduction of cP would cause low Pc despite the low bulk electron concentration, but then the increase in Pc with
increasing Si density for the air-exposed samples is unexplained. Thus, high space-charge density in the surface depletion region is the likely mechanism by which the low contact resistivity is obtained.
In conclusion, in situ Ag metallization of MBE GaAs
doped at 1 X 1020 cm- 3 with Si yields a nonalloyed specific
contact resistivity of 1.3 /lfl cm 2 • This result indicates a
space-charge density in the GaAs at the metal interface approximately equal to the silicon density, despite a bulk electron density of 4X 10 18 cm- 3 . This effect is explained by
mid gap Fermi-level pinning of the arsenic-stabilized GaAs
surface during growth. Silicon's solubility in GaAs is
~2 X 10 cm -3, suggesting that nonalloyed contacts applied in situ could yield a specific contact resistivity as low as
0.5 /lfl cm 2 •
NpnN doub~e-heterojunction
The authors gratefully acknowledge N. Braslan for
helpful discussions and A. Boulding, F. Cardone, and R.
Savoy for microprobe analysis.
'R. L. Longini and R. F. Greene, Phys. Rev. 102, 992 (1956).
2M. E. Greiner and J. F. Gibbons, App!. Phys. Leu. 44, 750 (1984).
3p. D. Kirchner, J. M. Woodall, and S. L. Wright, Abstract in IEEE Trans.
Electron Devices ED·30, 1590 (1983).
40. L. Miller, S. W. Zehr, and J. S. Harris, Jr., J. App!. Phys. 53, 744 (1982).
sO. L. Miller, H. T. Yang, and S. W. Zehr, "Cascade AIGaAs-GaAs Solar
Cell Research using MBE," Los Angeles SPIE Technical Symposium,
SPIE Proceedings, 1982, p. 323.
6H. H. Berger, 1. Electrochem. Soc. 119, 509 (1972); Solid State Electron.
IS, 145 (1972).
7c. Y. Chang, Y. K. Fang, and S. M. Sze, Solid State Electron. 14, 541
"D. K. Schroder and D. L. Meier, IEEE Trans. Electron Devices ED-31,
637 (1984).
"Y. G. Chai, R. Chow, andC. E. C. Wood, App1. Phys. Lett. 39, 800(1981).
IOJ. De-Sheng, Y. Makita, K. Ploog, and H. J. Queisser, J. App!. Phys. 53,
999 (1982).
"R. A. A. Kubiak and E. H. C. Parker, App!. Phys. A 35,75 (1984).
12p. A. Barnes and A. Y. Cho, App!. Phys. Leu. 33, 651(1978).
13 A. D. Katnani, P. Chiaradia, H. W. Sang, Jr., and R. S. Bauer, J. Vac. Sci.
Techno!. B 2, 471 (1984).
transistor on lnGa.AsP/~nP
l. M. SU,a) N. Grote, R. Kaumanns, and H. Schroeter
Heinrich-Hertz-Institutfiir Nachrichtentechnik Berlin GmbH. Einsteinufer 37, D-JOOO Berlin, 10,
West Germany
(Received 20 November 1984; accepted for publication 10 April 1985)
Double-heterojunction bipolar transistors have been fabricated on InGaAs(P)llnP with current
gains of up to 200. Transistors with a p+ -lnGaAslN-InP base/collector junction exhibited
drastic gain reduction at low collector bias voltages which is ascribed to the electron repelling
effect of the conduction-band spike formed at the colJector heterojunction. To overcome this
complication a thin n-InGaAs transition layer was inserted between the ternary base and the InP
wide-gap collector. The resulting nN double-layer collector structure leads to excellent current!
voltage characteristics.
In the last few years, 111- V compound heterojunction
bipolar transistors (HBT) have gained increasing interest for
microwave amplification and high-speed switching as wen
as for optical detection. Microwave HBT's on GaAIAsl
GaAs have already been reported with excellent high-frequency properties, namely, unity gain cut-off frequencies of
up to 25 GHZ. I - 3 HBT's on InGaAsP/lnP, however, have
been mostly developed as phototransistors (e.g., Refs. 4 and
5) for long wavelength (1-1.6 /lm) optical communication
applications. Only recently InP-based "electrical" bipolar
transistors with base terminals have been published. These
devices mainly comprise a base-collector homojunction of
InGaAs (x ln = 0.53).6--S Moreover, doubl1e-heterojunction
"On leave from Peking Electron Tube Factory, Peking, China.
Appl. Phys. Lett. 47 (1),1 July 1985
transistors employing an InP wide gap col.lector in conjunction with a quaternary InGaAsP (Ag = 1.1 /lm) base layer
have been fabricated to form a driving circuit together with a
monolithically integrated laser diode providing 1.6 Gbit/s
modulation capability.9
In order to optimize the high-frequency performance of
InGaAsP IlnP HBT's a device structure should be chosen
which uses an 1110.53 G~147 As layer for the base and InP as a
wide gap collector. In,) 53 Gao.47 As is known to exhibit the
highest electron mobilities (/l300K ~8-11 X loJ cm 2/Vs at
n~1016 cm-· l ) within the InGaAsP/lnP alloy system,
which result in short transit times in the (neutral) base. On
the other hand, for the collector of a microwave transistor
the high-field drift velocity rather than the (low field) mobility is of importance. This is because the drift velocity deter-
© 1985 American Institute of Physics
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