(435a) Numerical Analysis of the Flow, Transport, and Interfacial Phenomena Associated with Growth of Crystalline CZT Under Crucible Rotation | AIChE

(435a) Numerical Analysis of the Flow, Transport, and Interfacial Phenomena Associated with Growth of Crystalline CZT Under Crucible Rotation

Authors 

Divecha, M. S. - Presenter, University of Minnesota
Derby, J. J., University of Minnesota

Cadmium zinc telluride
(CZT) crystals are widely used for advanced gamma ray detectors due to
their room temperature operability and high band gap. In order to achieve
the high carrier mobility and lifetime properties necessary for its
functionality, precise crystal growing techniques are required. CZT is
traditionally grown via the gradient freeze method, where molten CZT is
directionally solidified in a cylindrical ampoule by drawing it through a
temperature gradient, thus encouraging growth of single crystalline material.
Directional solidification methods such as this result in complex transport
phenomena, as the translating solid-liquid interface induces thermal-buoyancy
driven flows, and a non-unity segregation coefficient leads to inhomogeneous
concentration gradients. One problem that commonly arises during gradient
freeze growth of CZT is the accumulation of thermodynamically supercooled liquid near the solidification interface. This
occurs due to local compositional and thermal inhomogeneity created by the
growing conditions, which results in constitutional supercooling.
Such supercooling promotes an unfavorable
morphological instability, changing the solid-liquid interface from planar to
cellular. The subsequent cellular morphology can result in the entrainment and
entrapment of melt, resulting in micron-sized tellurium-rich inclusions that
reduce detector performance and require lengthy annealing process to remove.

The accelerated crucible
rotation technique (ACRT), which repeatedly rotates the crucible back and
forth, has been shown to result in a lower density of these tellurium-rich
particles and dramatically improve detector performance. Under suitable
operating conditions, ACRT dramatically changes the patterns of flow in the
melt, inciting new rotational phenomena such as Ekman flows and Taylor-Gšrtler instabilities. Enhanced convection also impacts the
thermal and solutal fields. Conventional wisdom
asserts that ACRT works by homogenizing the composition in the melt. However,
little is actually understood from a thermodynamic and mechanistic perspective
on how the ACRT works to reduce constitutional supercooling,
and, in turn, tellurium-rich inclusions in the crystal. Additionally, little is
known on how to select the optimal rotation schedule to achieve high-quality
material.

We aim to address these
questions using a comprehensive computational model of the flow,
heat, and mass transfer within the CZT growth system. We employ an in-house, Galerkin finite-element code with elliptic mesh generation
that allows for the computation of flow patterns,
thermal field, composition, and interface shape. Using the results
from our simulations, we are able to track the quantity and locations of
locally supercooled melt. We will show that, contrary
to the conventional wisdom, ACRT does not necessarily homogenize the melt, but
rather disrupts the local compositional field, which, in turn, affects the
thermodynamic stability of the melt. Moreover, we have developed a thermodynamically-consistent metric that allows for the
systematic evaluation and optimization of ACRT schedules. Optimal
ACRT schedules and generalized ACRT recommendations for this geometry will be
presented.

Figures

Fig. 1: a) One example of an ACRT scheme, with four
distinct regimes: acceleration, constant, deceleration, and stopped, repeated
in the opposite direction, and b) a typical result of the streamlines (left)
and thermodynamically undercooled regions (right) in the system during the
deceleration regime.

This work is supported by the Department of Energy,
National Nuclear Security Administration, under Award DE-NA0002565; no official
endorsement should be inferred.