Microstructure of directionally solidified CrAsGaAs eutectic

Microstructure of directionally solidified CrAsGaAs eutectic

Materials Research Bulletin, Vol. 30, No. 4, pp. 453-461,1995 Copyright Q 1995 Elsevier Science Ltd Printedin the USA. All rights reserved 0025-540819...

549KB Sizes 0 Downloads 28 Views

Materials Research Bulletin, Vol. 30, No. 4, pp. 453-461,1995 Copyright Q 1995 Elsevier Science Ltd Printedin the USA. All rights reserved 0025-5408195 $9.50 + .OO

MICROSTRUCTURE OF DIRECTIONALLY CrAs/GaAs EUTECTIC

SOLIDIFIED

Douglas E. Holmes and Linda Y. Koo Electronic

Materials Engineering 829 Flynn Road Camarillo, CA 93012

(Received December 12, 1994; Communicatedby J.F. Ackennann)

ABSTRACT

The microstructure of the CrAs/GaAs eutectic directionally solidified by both the Czochralski (Cz) and vertical Bridgman (Vl3) methods consists of arrays of CrAs rods oriented along the axis of Microdefects in Cz material solidification in a GaAs matrix. including striations, terminations and nucleations, coalesence and branching, and oscillatory instabilities are prevalent and are the result of dynamic morphological adjustment under fluctuating conditions of microscopic solidification. In contrast, selected regions of VB material exhibit near-ideal hexagonal packing of circular rods in the matrix. Cause-effect relationships between microstructure and conditions of solidification were determined and are presented and discussed.

MATERIALS

INDEX: chromium, gallium, arsenides Induction

Directional solidification of eutectic and off-eutectic metallic materials systems have been widely investigated [l-4] as a means of improving the mechanical properties of materials. Multiphase bulk materials with oriented and periodic microstructure, such as the rod-matrix and lamellar types, are so produced. Only a limited number of studies have addressed eutectics containing semiconductor t-5-71materials. The purpose of this investigation was to determine cause-effect relationships between microstructure and solidification conditions of the CrAslGaAs eutectic system. 453

454

D.E HOLMES&

Vol. 30, No. 4

Sample F+eparation Ingots were prepared by both the Czochralski (Cz) and vertical Bridgman (VB) methods. Cz growth was carried out in a high pressure liquid encapsulated system by using boric oxide encapsulant. Melts were prepared by using in-situ synthesis with 6-9’s elemental starting materials. Ingots about 25 mm in diameter and 50 mm long were pulled at a rate of about 6 mm/hr with crystal rotation of 5 rpm. Solidification was initiated with either or <111> GaAs seed crystals. Material synthesized in the Cz system was used for VB experiments. Starting material was first precast in sealed quartz ampoules into ingots about 8 mm in diameter and 20 mm long. Each precast ingot was then re-sealed in a similar quartz ampoule and loaded into an RF-heated graphite susceptor. Solidification was initiated by melting the charge and proceeded from the bottom under a stable vertical temperature gradient by controlled power reduction. Vertical temperature gradients, as determined by thermal profiling in the absence of a charge, were about 5O”Clcm. The average solidification rate was estimated to be between 2.5 and 5 cmlhr. Seeding was not used in the VB experiments. The VB-solidified ingots were sliced both parallel (longitudinal section) and perpendicular (cross section) to the axis of solidification. Individual samples were lapped and polished with colloidal silica. Polished samples were characterized by optical microscopy, secondary electron microscopy (SEM), and energy dispersive X-ray analysis (EDAX).

Results and Discus&on The predominant microstructural feature observed in cross sectional and longitudinal samples of Cz-solidified material consists of rods oriented along the axis of solidification, as shown in Figure 1. Rod diameters range from 2 to 10 pm corresponding to a density of about 5 x 106cmP2. EDAX compositional mapping showed that the rods consist of CrAs in the GaAs matrix. Accuracy of the EDAX method as determined by measuring standard GaAs samples was about 1% at. %. Within this range of accuracy, solubility of Cr in the GaAs was no more than 1 at. %, and that of Ga in the CrAs was no more than, but perhaps as high as 3 at. %. Although the microstructure was distinguishably rod-matrix throughout the entire ingot, a considerable variation of features as well as a variety of defects (with respect to an ideal, hexagonally closed-packed structure of continuous rods) were observed. For example, cross-sectional shapes of the rods were typically distorted circles or severely elongated. Elongated shapes apparently resulted from coalesence or branching of the rods, as observed in longitudinal views (Figure 2). Longitudinal sections (Figure 3) also revealed spontaneous termination and nucleation of rods. In addition, an oscillatory instability also shown in Figure 3 was observed, which was caused by alternating and successive terminations and nucleations. Finally, ideal hexagonal packing of rods in the matrix (as viewed in cross section) was rarely observed to extend from an isolated

Vol. 30, No. 4

ARSENIDE CRYSTALS

455

rod to more than one set of 6 nearest neighbors due to the variability of the microstructure. Although the length of the longest continuous rods was about 2.5 mm, the typical length was about 50 pm. Furthermore, the longest continuous rods were invariably observed toward the center of the ingot, and the length gradually decreased toward the edge. In contrast to Cz-solidified material, the microstructure in selected regions of VB ingots displayed strikingly regular arrays of circular CrAs rods arranged in nearly ideal hexagonal packing in the GaAs matrix, as shown in Figure 4. However, such regions extended to only a few hundred micrometers in width. Otherwise, dendritic growth dominated the microstructure. On the macroscopic scale, a feature common to all Cz and VB ingots was the presence of grains, which were easily seen by eye in polished samples. Grains were typically a few millimeters wide, and extended along the axis of solidification up to about 15 millimeters in length. Microscopic examination revealed that the relative orientation of CrAs rods in adjacent grains changed at the boundary. Otherwise, no distinct lineage was observed. Apparently, the misorient&ion between the assemblage of rods across a boundary was sufficient to alter the optical reflectivity of polished samples. There was no apparent effect of seed orientation on the incidence of grain formation. Another dominant feature of the macroscopic distribution of rods observed only in Cz material viewed in cross section was a pattern of concentric striations emanating from a point located away from the geometric center of an ingot, as Striations were visible by eye in polished samples. seen in Figure. 5. Examination at higher magnification showed that the structure of the striations consisted of alternating bands of small diameter rods packed at relatively high density and large diameter rods packed at relatively low density, as shown in Figure 6. The patterns of concentric striations was continuous throughout the entire cross section of an ingot, and was superimposed over all grain structure. In these samples, grain boundaries were not distinguishable at any magnification due to the presence of the striations. A final observation revealed in SEM micrographs was intensity banding within the matrix material. Typical bands were on the order of 5 micrometers Since secondary wide. Banding was not observed in optical micrographs. electron emission is dependent on crystallographic orientation, it was tentatively concluded that the GaAs matrix is twinned. Based on published work on directionally solidified [8,91 materials, a “single crystal’ of the CrAs/GaAs eutectic would consist of a single crystalline GaAs matrix in which CrAs rods would have a common, preferred crystallographic orientation, and all CrAs rods would be crystallographically aligned to each other. Although compliance with this model might be anticipated for intra-granular microstructure, the likelihood of multiple micro-twinning in the matrix precludes drawing any conclusions about the self-orientation of the Further characterization needs to be done to understand the CrAs. crystallography of the material. The general variability of the microstructure of the Cz CrAs/GaAs and the persistence of micro-defects can be attributed to the spatial dependence of the conditions of microscopic solidification in the Czochralski configuration. For example, the striation pattern (Figure 5) of varying rod diameters can be directly

456

DE. HOLMES et

Vol. 30, No. 4

related to the fluctuating, microscopic solidification rate associated with a rotating crystal in a thermally asymmetric environment, as discussed below. Experimental data on metallic eutectic systems has invariably shown that the characteristic spacing, y, (lamellar spacing or inter-rod spacing) is related to the solidification [lo,111 rate, R, as follows: y2R = B/A, (1) where B and A are constants of the material. Equation 111is theoretically derived from the assumption 1121 of eutectic solidification at the extremum, where growth occurs at the maximum rate for a given thermal undercooling, or, equivalently, at a minimum undercooling for a given rate. Furthermore, extensive investigations of Czochralski crystal growth of dilute binary semiconductor materials by the application of interface demarcation 113,141 have shown that continuous crystal rotation through hot and cold regions of the melt generally leads to periodic fluctuations of the microscopic solidification rate. Accordingly, the inter-rod spacing of CrAs/GaAs in regions of high solidification rate adjusts to smaller-than-average dimensions, and vice versa in regions of low solidification rate. Because the average volume fraction of CrAs and GaAs must remain constant, rod diameters in regions of high solidification rate are smaller than average, and vice versa in regions of low solidification rate. In this investigation, the concentric nature of the striations (in cross section) in the eutectic are analogous to the morphology of impurity striations characteristic of dilute binary semiconductor materials, which arise due to the dependence of the effective distribution coefficient on solidification rate. The complicated pattern of concentric (tree-ring-like) striations appear in cross sections as a general consequence of the superposition of time- and spatiallySince the solid-liquid interface of Cz-grown crystals is dependent effects. generally non-planar, a polished cross sectional sample presents a surface that has solidified at different times. For example, in a crystal with a convex (with respect to the crystal) interface, the center of the polished sample would have solidified before the edge. Superimposed on this surface are the effects of a spatially dependent microscopic solidification rate. Concentric striations are the result. Microdefects observed in the Cz material, including branching, coalesence, termination, and nucleation of rods, and the generally irregularly shaped cross sections of the rods are indicative of the dynamics of solidification as the solidliquid interface continuously adjusts to maintain stability in response to a adjustment to a higher fluctuating solidification rate. Thus, a microstructural density of smaller diameter rods in response to an increased solidification rate is achieved by branching of rods or by the nucleation of new rods. (Rod nucleation would correspond to branching of the matrix.) Likewise, an adjustment to a lower density of larger diameter rods in response to a reduced solidification rate is achieved by rod termination or coalescence. It is noted that rod branching and coalesence are believed to be analogous to the lamellar [ll] fault, a mechanism that accounts for the adjustment of a lamellar eutectic to changes in growth conditions. Based on our observations of CrAs/GaAs, it is concluded that whether branching (or coalesence) or nucleation (or termination) occurs depends on the relative acceleration or de-acceleration of the instantaneous microscopic solidification rate. Stable branching (and coalesence) involves simultaneous

Vol. 30, No. 4

ARSENIDE CRYSTALS

451

lateral redistribution of the compositional field in the liquid ahead of the solidification front by diflksion. If an acceleration is sticiently low as to accommodate lateral diffusion, motion of the rods across the interface can occur. However, high acceleration would lead to accumulation of Cr ahead of the matrix promoting the formation of an indentation in the GaAs (due to the reduction of the GaAs melting temperature with increased Cr composition) and eventual rod nucleation. Termination would occur under a large de-acceleration when adequate time is available for excess Cr concentration gradients to substantially Re-melt could also cause terminations. Therefore, oscillatory homogenize. instabilities would correspond to conditions of alternating high acceleration and de-acceleration. It is also important to note that the defects had an apparent distribution across the Cz ingots, which may shed further light on the mechanism of defect formation. In general, branching and coalesence were most prevalent toward the edges of the ingot where the length of continues rods was lowest. Terminations and nucleations, on the other hand, although not as prevalent, were generally observed toward the center of the ingots where the length of rods was the greatest. Such a spatial distribution might seem contradictory to the discussion above concerning the relationship between the nature of the defect and the relative acceleration and de-acceleration of the interface. This is because greater accelerations (and de- accelerations) are generally expected at the edge of an ingot due to the larger thermal asymmetry toward the edge of the crucible. In fact, the observed distribution of the rod lengths across the diameter of the ingots would confirm that more steady solidification occurred toward the center. Therefore, terminations and nucleations may have been caused by timedependent fluctuations of interface temperature associated with turbulent convection in the melt rather than to spatially dependent fluctuations of microscopic solidification rate. In regard to V&solidified material, the predominance of dendritic solidification can be explained as follows. Since the starting material for these experiments was prepared by the Cz method, the composition of VB and Cz melts should have been the same. However, the composition was undoubtedly not exactly the eutectic. Therefore, a minimum ratio of temperature gradient (normal to the interface), G, to the solidification rate, R, would have been required for stable, planar solidification 141. In this investigation, VI3 material was solidified at rates four to eight times greater than Cz material probably causing the G/R ratio to fall below the level required for stability.

The microstructure of directionally solidified CrAs/GaAs eutectic is the rod-matrix type. Although this microstructure is easily produced by Cz solidification, it is highly defective due to dynamic morphological adjustment under fluctuating conditions of microscopic solidification inherent to the Cz process. As a result, highly periodic, three-dimensional structures extending over substantial regions of an ingot are difficult to produce. On the other hand. our results indicate that with greater attention to promoting conditions of stability, including control of temperature gradients and solidification rates,

458

D.E. HOLMES &&

solidification in the VB configuration should microstructure over ingot-scale distances.

Vol. 30, No. 4

enable the production

of periodic

Acknowledgments The authors gratefully acknowledge the support of the Air Force Office of Scientific Research under Contract No. F49620-90-C-0087 and the encouragement of Lt. Col. Gernot S. Pomrenke. List of RefelWlces 1. 2. 3. 4. 5. 7: 8. 9. 10. 11. 12. 13. 14.

F.D. Lemkey and N.J. Salkind, in ‘Crystal Growth’, p. 171, Pergamon, Oxford (1967). F.S. Galass, J. Metals, 12,17 (1967). H. Bibring, G. Seibel, and M. Rabinovitch, Mem. Sci. Rev. Met., 49(5), 341 (1972). F.R. Mollard and M.C. Flemings, Trans- AIME, 239, 1526 (1967). W.K. Liebmann and E.A. Miller, J. Appl. Phys., 34, 2653 (1963). B. Riess and T. Renner, Zeitschrift fiir Naturforschung, 21, 546 (1966). A. Muller and M. Wilhem, J. Phys. Chem. Solids, 26, 2021 (1965). L.M. Logan, R.W. Kraft, and F.D. Lemkey, in “Advances in Materials Research,” H. Herman (ed.), 5, 83 (1971). F.R. Mollard and M.C. Flemings, Trans. AIME, 239, 1534 (1967). J.D. Hunt and K.A. Jackson, Trans. AZME, 236,843 (1968). K.A. Jackson and J.D. Hunt, ibid., 1129. W.A. Tiller, “Liquid Metals and Solidification,” p. 276, ASM, Cleveland, (1958). A.F. Witt and H.C. Gatos, J. Electrochem. Sot., 115, 70 (1968). R. Singh, A.F. Witt, and H.C. Gatos, J. Electrochem. Sot., 115, 112, (1968).

Vol. 30, No. 4

ARSENIDECRYSTALS

459

Fig. 1 Cross section (a) and longitudinal in GaAs matrix of Cz ingot.

section 6) showing

Fig. 2 Longitudinal view showing coalescence and branching Direction of solidification is left to right.

CrAs rods (lighter

phase)

of rods in Cz materials.

460

Examples of rod termination Cz-solidified materials.

Cross section of VB materials in GaAs matrix.

D.E. HOLMES et

and

Fig. 3 nucleation,

Fig. 4 showing near-ideal

Vol. 30, No. 4

and

oscillatory

instability

packing of circular

in

CrAs rods

Vol. 30, No. 4

Low magnification

ARSENIDE CRYSTALS

461

Fig. 5 view of striations in cross section of Cz material.

J



@a)

J ‘09

Fig. 6 Higher magnification view of striations shown in Fig. 5. Striations consist of alternating bands of high density, small diameter rods (a) and low density, large diameter rods (b).