A research team led by Eugene Haller has provided important
new insights into self-diffusion in solids. In a paper published
in Nature, they reported their discovery that in the semiconductor
gallium antimonide (GaSb) the diffusion coefficient of antimony
is at least 1000 times smaller than that of gallium. A completely
new diffusion mechanism may be responsible for this unexpected
phenomenon.
The most fundamental mass transport process in solids is self-diffusion,
the natural movement of individual atoms within the material.
It was first suggested in the 1920's that this process does not
occur through the simple exchange of positions of two neighboring
atoms, but rather, must be mediated by defects. In metals and
in some semiconductors, it has been shown that vacancies (missing
atoms) are the most important such defect. In this so-called "vacancy
mediated diffusion," an atom diffuses by moving into a neighboring
vacancy, leaving behind a vacancy in its former position. [Another
important diffusion mechanism in semiconductors involves motion
of atoms located between lattice sites (self-interstitials).]
Historically, it has been difficult to measure self-diffusion
rates accurately in solids, in part because of the need to label
some but not all of the atoms in a pure material. Radioactive
isotopes are often used for this purpose but are limited by their
short half lives. Another difficulty is that diffusion rates are
highly sensitive to the quality of the crystal under study; for
accurate measurements material of high crystalline quality is
required. For these reasons, even for silicon, accurate measurements
have been available only in the last 10 years. Self-diffusion
in compound semiconductors has an addition level of complexity;
the two elements might have different diffusion coefficients.
The Haller group performed these self-diffusion measurements using
highly enriched stable isotopes as labels. Compared to the radioactive
probe approach, diffusion studies with stable isotopes are neither
limited by the half-life of the isotope nor affected by the degradation
associated with the implantation or deposition of the isotopes.
Thus, experiments can be extended to much longer times and conducted
over a wider temperature range.
GaSb offers a unique opportunity for this type of experiment.
Two stable isotopes exist for both Ga (69Ga and 71Ga) and Sb (121Sb
and 123Sb), allowing the simultaneous study of diffusion of Ga
and Sb on their respective sublattices. A collaborating team from
Spain grew a special test structure of high crystalline quality:
a natural GaSb substrate under a 100 nm layer of 71Ga123Sb under
a 100-layer of nm 69Ga121Sb under a 200-nm natural GaSb cap (see
figure). To measure the self-diffusion, the team annealed the
structure at various temperatures and measured the movement of
the isotopically labeled atoms into the neighboring layers via
ex-situ secondary ion mass spectrometry (see figure). The results
were surprising: Ga atoms diffused much faster than Sb atoms:
Ga diffusion could be measured at temperatures as low as 575 C
while little Sb diffusion was observed even after 18 days at 700
C only 12 C below the melting point. The observed thousandfold
asymmetry of the self-diffusion on the two sublattices is extraordinary.
This discovery is unprecedented in the field. Similar disparities
in self-diffusion rates of compounds have been seen in some transition
metal silicides, but in these systems the diffusion coefficients
of the two elements approach each other near the melting point.
The team has hypothesized that the Ga and Sb atoms diffuse independently,
each on its own sublattice. The temperature dependence of the
Ga self-diffusion coefficient is consistent with that predicted
for Ga-vacancy-mediated diffusion. The situation with Sb is different,
however. If an Sb vacancy tries to form, the electronic properties
of GaSb make it favorable for a neighboring Ga to fill it by forming
a so-called Ga "antisite" defect. Surrounded by its
4 Ga neighbors, the Sb vacancy is therefore unstable and cannot
mediate Sb diffusion. Other diffusion pathways apparently are
also slow
The results may be applicable to other III-V and II-VI compound
semiconductors, and are expected to inspire new thinking into
the self-diffusion mechanisms in these materials. The discovery
may have important ramifications in that the manufacturing and
operation of all semiconductor-based microelectronic devices,
including microprocessors and optoelectronic devices, relies critically
on the control of diffusion processes. For example, controlled
diffusion is required for the introduction of impurity atoms to
certain areas of semiconductors to create the three-dimensional
structures required to enable electronic function. On the other
hand, diffusion processes must be understood and controlled such
that these 3-D structures are maintained during thermal processing
and device operation.
H. Bracht, S. P. Nicols, W. Walukiewicz, J. P. Silvera, F.
Briones, and E. E. Haller, "Large disparity between gallium
and antinomy self-diffusion in gallium antimonide," Nature
408, 69-72 (2000).
Ernest Orlando Lawrence Berkeley
National Laboratory
One Cyclotron Road, Mail Stop 66, Berkeley, California 94720 USA