Liquid Encapsulated Czochralski Process Research Articles

Liquid-encapsulated Czochralski growth of Ga 1−xIn xAs single crystals with uniform compositions

Most alloy semiconductors tend to develop severe macrosegregation along single crystals grown from melts. The feasibility of modifying the conventional liquid-encapsulated Czochralski (LEC) process to reduce macrosegregation was assessed. First, the crystal was pulled from the melt in a stationary inner crucible that was replenished by the melt overflowing from a rising outer crucible. Ga 1− x In x As single crystals of about 1, 2 and 3 mol% InAs were grown. When growing crystals at higher InAs contents, the replenishing melt had a tendency to freeze and disrupt crystal growth. The crystal was then pulled from a melt which was replenished from below by a solid feed held in the same rising crucible as the melt. The solid feed was prepared from a B 2O 3-encapsulated melt that undercooled and then rapidly solidified into an ingot without macrosegregation. Ga 1− x In x As single crystals of about 5 and 7.5 mol% InAs were grown. All crystals were uniform in composition in both the axial and the radial directions. It was also demonstrated that the first modified LEC process can be used to grow inhomogeneous Ga 1− x In x As single crystals from pure GaAs seeds and initially pure GaAs melts, which is convenient when no alloy crystal seeds are available.

Melt motion during liquid-encapsulated Czochralski crystal growth in steady and rotating magnetic fields

During the liquid-encapsulated Czochralski (LEC) process, a single compound semiconductor crystal such as gallium-antimonide is grown by the solidification of an initially molten semiconductor (melt) contained in a crucible. The motion of the electrically-conducting molten semiconductor can be controlled with externally-applied magnetic fields. A steady magnetic field provides an electromagnetic stabilization of the melt motion during the LEC process. With a steady axial magnetic field alone, the melt motion produces a radially-inward flow below the crystal–melt interface. Recently, an extremely promising flow phenomenon has been revealed in which a rotating magnetic field induces a radially-inward flow below the crystal–melt interface that may significantly improve the compositional homogeneity in the crystal. This paper presents a model for the melt motion during the LEC process with steady and rotating magnetic fields.

Magnetic field effects during liquid-encapsulated Czochralski growth of doped photonic semiconductor crystals

During the liquid-encapsulated Czochralski (LEC) process, a single compound semiconductor crystal such as indium phosphide or gallium antimonide is grown by the solidification of an initially molten semiconductor contained in a crucible. The motion of the electrically conducting molten semiconductor can be controlled with an externally applied magnetic field. This paper presents a model for the unsteady transport of a dopant during the LEC process with a steady axial magnetic field. The convective species transport during growth produces significant segregation in both the melt and the crystal. Dopant distributions in the crystal and in the melt at several different stages during growth are presented.

Magnetic stabilization of the buoyant convection in the liquid-encapsulated Czochralski process

This paper presents a linear stability analysis for the buoyant convection during the liquid-encapsulated Czochralski growth of compound semiconductor crystals with a steady, uniform, vertical magnetic field. Results are presented for two values of the Prandtl number, corresponding to indium-phosphide (InP) and gallium-arsenide (GaAs). Most of the results are for a melt depth equal to the crucible diameter, but some results are also presented for a smaller depth. For the cases considered here, the instability involves a transition from a steady axisymmetric flow to a steady nonaxisymmetric flow corresponding to the first Fourier mode in the azimuthal direction. For the weaker magnetic fields, the critical Rayleigh number is close to that for the Rayleigh–Bénard instability in a vertical cylinder. For the stronger magnetic fields, the critical Rayleigh number for GaAs is higher than that for InP because convective heat transfer in the GaAs base flow reduces the vertical temperature gradient.

Transient dynamics and control of indium phosphide LEC furnaces

Abstract A transient 1D model was developed for simulating the dynamics of liquid encapsulated Czochralski (LEC) growth processes. The model considers a melt–solid flat interface and a uniform temperature within the four phases (i.e., air, melt, solid and encapsulant) and predicts the crystal shape evolution during growth given the time histories of the heater temperature and of the pull rate. The model predictions were satisfactorily compared with experimental InP LEC dynamic growth data. Then, a feedback–feedforward control algorithm effective to achieve a cylindrical growth of single crystals in LEC systems was developed. In particular, the model to be used in the feedforward step was derived linearizing the above dynamic model. The controller keeps the crystal radius constant by acting on the heater, taking as input variables heater and air temperature, crystal weight, seed and crucible position. Controller performances have been tested on a transient model of LEC process as controlled system.

Double crucible LEC growth of In-doped GaAs using inner crucibles with a bottom tube

In order to reduce InAs segregation along In-doped GaAs single crystals, the liquid encapsulated Czochralski (LEC) process commonly used for GaAs was modified, by using an improved double crucible technique and the concept of zone leveling. This avoided the long-existing dilemma of the double crucible technique: the hole in the inner crucible either is too small to let the replenishing melt go through or too large to suppress mixing between the growth melt and the replenishing melt through the hole. A tube was provided at the bottom of the inner crucible; its 6 mm inside diameter was large enough for the replenishing melt to pass through without being blocked by B 2O 3, and its 70 mm length was more than long enough to suppress mixing. The initial growth melt was leveled (enriched) to C o/ k, where C o is the targeted crystal composition of 0.04 mol% InAs and k is the effective segregation coefficient 0.116. Effective segregation reduction was obtained and a crystals was grown with a uniform InAs concentration of 0.0422±0.0024 mol% InAs. A similar crystal was grown by using the conventional LEC process; the InAs concentration rose monotonically from 0.0463 mol% at the starting end to 0.245 mol% at the finishing end.

Electron irradiation effects on photoconductive semiconductor switches (PCSSs) used in sub-nanosecond transient generators

Radiation-induced damage occurs in GaAs photoconductive semiconductor switches used in sub-nanosecond transient generators when subjected to 600 keV and 6 MeV electron irradiation. These switches are made from semi-insulating (SI) compensated material through a EL2/carbon compensation mechanism, and the liquid encapsulated Czochralski process. New defect levels are formed as a result of the non-ionizing energy loss (NIEL) process. The formation of new defect levels in the device alters the compensating balance between the existing deep level EL2 trap/donors and carbon acceptors, and changes the material properties. As a result, two important parameters of the device are adversely affected-the hold-off voltage of the switch at the pulse-charging (off) state, and the rise time during the conduction (on) state. The hold-off voltage shifts to a lower value since there are more trap-filled regions available that can fill up and alter the homogenous nature of the device material. Unstable filamentary conduction then occurs at a lower voltage and leads to premature breakdown. As with EL2 trap levels, new defect states induced by electron irradiation will further contribute to the delay in the rise time of the switch. The rise time determines the maximum energy transferred to the load. The electron damage mechanism and its effects on the switch characteristics depend on the material properties. Intrinsic material, or material made through compensation other than through the deep donor and shallow acceptor balancing process are not expected to behave similarly. Simulation results at higher bias show a marked degradation of material properties. The switch current-voltage (I-V) characteristic when the bias increases to the kilovolt range is similar to trap-dominated semiconductors. An initial sublinear current regime at low bias is followed by a super-linear regime of current flow at higher bias, and is in agreement with earlier observations.

In situ optical assessment of semi-insulating iron doped InP grown by liquid encapsulated Czochralski process

In semi-insulating, Fe doped InP, we demonstrate that infrared absorption technique provides quantitative measurements of Fe concentrations. We show that the quantum efficiency is reduced more than 70% in Fe doped samples, in comparison to undoped samples. Relative near band edge photoluminescence (PL) intensity is antiproportional to Fe concentration. This can provide a calibration method to transform scanning PL images to quantitative Fe distributions. We demonstrate that infrared absorption mapping images Fe distribution in InP, in a similar way to EL2 imaging in GaAs. The images show that the slip-like defects decorated by Fe centers predominantly lie along [011] direction.

Thermal Field Simulation of LEC Process by Finite Volume Method

Thermal field is obtained for the Liquid Encapsulated Czochralski process by Quasi-Steady State Model. Our model includes the effects of convective heat transfer in the melt and radiational heat transfer of the equipment surfaces. To calculate the radiational heat flux, the view factor calculation is performed. Encapsulant layer is assumed to be totally opaque or transparent to the radiation. The thermal field formation and the temperature of the heater depend on the thermal transparency of the encapsulant layer. The effects of melt flow and thermal transparency of the encapsulant layer are analysed. The proposed model can be extended to the semi-transparent model and used as thermal field prediction model for optimization of LEC process.

Buoyant Convection During the Growth of Compound Semiconductors by the Liquid-Encapsulated Czochralski Process With an Axial Magnetic Field and With a Non-Axisymmetric Temperature

This paper treats the buoyant convection of a molten semiconductor in a cylindrical crucible with a vertical axis, with a uniform vertical magnetic field, and with a non-axisymmetric temperature. Most previous treatments of melt motions with vertical magnetic fields have assumed that the temperature and buoyant convection were axisymmetric. In reality, the temperature and resultant buoyant convection often deviate significantly from axisymmetry. For a given non-axisymmetric temperature, the electromagnetic suppression of the axisymmetric part of the buoyant convection is stronger than that of the non-axisymmetric part, so that the deviation from an axisymmetric melt motion increases as the magnetic field strength is increased. The non-axisymmetric part of the buoyant convection includes relatively strong azimuthal velocities adjacent to the electrically insulating vertical crucible wall, because this wall blocks the radial electric currents needed to suppress azimuthal velocities.

Modelling of thermal fluid flow in the liquid encapsulated Czochralski process and comparison with experiments

The thermal fluid flow in liquid encapsulated Czochralski (LEC) process has been studied by a combined experimental and numerical modelling approach. Experiments using gallium melt and a stationary dummy crystal have been performed in an industrial puller with the crucible rotating at 30 rpm. This system is numerically simulated by a global conduction-radiation and a three-dimensional (3D) melt convection model. Significant difference in thermal fluid flow structures is found depending on the crucible position. This illustrates the strong influence of the thermal boundary conditions on the fluid flow. From practical viewpoint, it is therefore important to optimize the crucible, i.e. the growth position in the LEC process.

Improved uniformity of GaAs by carbon control

Gas analysis was used to monitor the atmosphere of a high-pressure crystal during the growth of semi-insulating gallium arsenide by the liquid-encapsulated Czochralski process. Apart from the background argon, the only gases detected were carbon monoxide and hydrogen. These were formed by reaction between the graphite components of the puller and residual water vapour. The quantities of CO and H2 increase with atmospheric humidity and this results in an increase in the electrical resistivity of the crystal. Diffusion through the boric oxide encapsulant appears to act as a rate-limiting step in the carbon transport. The CO concentration was increased by injection and decreased by gettering or by dilution, and by these means it was found possible to grow crystals with very uniform electrical properties and tightly controlled resistivities at selected values within the range 107-2*108 Omega cm.

Numerical simulation of flow structures and instabilities occurring in a liquid-encapsulated Czochralski process

This paper concerns the prediction and the analysis of the steady and time-dependent flow regimes that develop in liquid-encapsulated Czochralski (LEC) systems for material processing. The combined effects of temperature gradient and rotation on the flow and the temperature field, and especially the influence of the encapsulant layer (of Prandtl number 10≤Pre≤3270 ) on the melt (Prm=0.015), are investigated using axisymmetric modeling. The accuracy of the solutions based on a finite element method is discussed. The effects of the Grashof number (up to Gr=1.9×107 ) and of the rotation Reynolds numbers of the crystal (up to ‖Res‖=104 ) and of the crucible (up to Rec=1.5×104) are analyzed.

Autonomous liquid encapsulated Czochralski (LEC) growth of single crystal GaAs by “intelligent” digital control

An “intelligent” digital growth control approach developed at Arizona State University (ASU) which combines deterministic and heuristic techniques allows virtually total automation of the high-pressure LEC process for GaAs growth. A number of 4 kg, 3 inch semi-insulating GaAs crystals have been grown reproducibly on an adapted Cambridge Instruments CI 358 puller without any significant operator intervention, with process yields of more than 75% single crystals. Crystals produced with the digital system were superior in diameter control, compared to conventionally grown crystals, and exhibited a better uniformity of their electrical and crystalline properties.

The Production of High Quality, III-V Compound Semiconductor Crystals

A new technique, called the Gault process, has been developed at AT&T Engineering Research Center for growing large, single crystals of gallium arsenide, gallium phosphide, and indium phosphide. It can produce seeded, 50-mm diameter crystals that have lower dislocation densities and more uniform properties at all levels of doping than crystals grown by existing methods. In this paper, we describe the Gault process, which is based on the vertical gradient freeze method. Its features include seeded crystal growth in a boron nitride crucible, with much lower axial and radial thermal gradients than the conventional, liquid-encapsulated Czochralski process. We show that establishing an appropriate phosphorus or arsenic vapor pressure over the melt is an effective method of stoichiometry control. Also, a nearly planar liquid-solid interface has positive implications for radial uniformity and the occurrence of twinning and polycrystalline nucleation.

Nitrogen concentration in GaP measured by optical absorption and by proton-induced nuclear reactions

A method is described for the determination of the nitrogen concentration in GaP crystals from optical absorption measurements. The method has been calibrated by measuring the absolute nitrogen concentration by nuclear reaction quantitative analysis in the same samples for which the optical absorption was measured. The relationships between the nitrogen concentration, the integrated absorption of the A line (2.3172 eV), and the optical absorption at Ax (2.3275 eV) at 4.2°K are given by [N] cm−3=8.5×1014∫αA d E=3.4 ×1016αAx, where α is measured in cm−1 and d E in meV. Previous formulas which have commonly been used to determine the nitrogen concentration in terms of optical absorption overestimate the nitrogen concentration by about a factor of 5. Also the application of these formulas has varied between laboratories. Section I describes the determination of nitrogen concentrations by nuclear microanalysis. Measurements have been made in the range from [inverted lazy s]5×1016 to [inverted lazy s]1018 cm−3 in GaP grown by the liquid encapsulated Czochralski process in a nitrogen-pressurized crystal puller and in nitrogen-doped GaP grown by liquid phase epitaxy at [inverted lazy s]1000°C. Two distinct charged-particle techniques have been employed. In crystals containing nitrogen of normal isotopic composition, the 14N concentration was determined by proton activation using the reaction 14N(p,α)11C with proton energies in the range 6–10 MeV. The reaction product, 11C, was chemically separated after irradiation. The 11C is a positron emitter, and the γ rays from the positron annihilation were measured. Boron interference through the reaction 11B(p,n)11C was independently determined by measuring the prompt α-particle emission from the reaction 11B(p,α1)8Be*→2α2 during irradiation with 685-keV protons. In 15N-doped LPE layers, the 15N concentration was determined by measuring the prompt α-particle emission from the reaction 15N(p,α)12C during irradiation with 685-keV protons. Section II describes the optical absorption measurements and the correlation with the nuclear measurements. The temperature dependence of the integrated absorption and the line shape of the A line have also been measured. For determination of the nitrogen concentration, the A-line absorption measurements should be made at a temperature at which the absorption line shape is amenable to accurate integration. The integrated absorption can then be corrected to its value at 4.2°K and the nitrogen concentration calculated from the above relationship.

Liquid encapsulated Czochralski growth of 35 mm diameter single crystals of GaP

Liquid encapsulated growth of GaP is complicated by two features not present in ordinary Czochralski growth: molten B 2O 3 encapsulates the melt, and high pressure gas (55 atm) fills the growth chamber. The growth of large ingots of GaP (200 g, 35 mm diameter) under these conditions is more difficult than the previously described growth of smaller ingots. Diameter control is more difficult, and the large ingots often crack spontaneously. In the liquid encapsulated Czochralski process, the growth conditions are dominated first by radiative heat transfer and later by convective heat transfer. Evidence to support this conclusion includes the shapes of growing crystals and the configurations of the growth interface. By using a knowledge of the heat transfer mechanisms, a reliable technique has been developed for adjusting rf power input to obtain uniform diameter single crystals. Unless precautions are taken, however, such ingots will crack spontaneously. The vertical temperature gradient in the growth chamber can be as high as 500 °C/cm. The resulting thermal stress causes plastic deformation in the growing ingots, leading to frozen-in stresses which may exceed the yield stress of GaP. These stresses have been reduced sufficiently to avoid cracking by using an afterheater which is an integral part of the rf susceptor. The use of an afterheater can lead to the occurrence of twin lamellae, but these have been eliminated by using a suitably shaped crucible (e.g., flat bottomed).

Boron and nitrogen im gallium phosphide grown by the liquid encapsulated czochralski process

Boron and nitrogen concentrations have been determined, using charged particle activation analysis, in a number of GaP ingots grown by the Liquid Encapsulated Czochralski process in order to determine the extent of the chemical interaction between the Gap melt and the liquid B2O3 encapsulant, and the nitrogen ambient used to pressurize the crystal puller. The effects of deliberate oxygen, nitrogen and boron doping have also been investigated. Boron concentrations have been found in the range 3 x 1018 to3 x 1018 cm-3, and boron incorporation is seen to be suppressed by the presence of oxygen and nitrogen. Channeled nuclear reaction and local mode absorption measurements have shown that the boron is predomin-antly located on substitutional gallium sites. The nitrogen gas used to pressurize the crystal puller results in nitrogen incorporation in the range 1–2 × 1017 cm−3. Comparison with optical absorption measurements indicates that the formulas used to determine nitrogen from optical measurements overestimate the nitrogen by a factor of 4.