Nearly diffusion controlled segregation of tellurium in GaSb

Transcription

1 Journal of Crystal Growth 191 (1998) Letter to the Editors Nearly diffusion controlled segregation of tellurium in GaSb Partha S. Dutta, Aleksandar G. Ostrogorsky* Department of Mechanical Engineering, Aeronautical Engineering and Mechanics, Rensselaer Polytechnic Institute, Troy, New York 12180, USA Received 22 August 1997; accepted 15 April 1998 Abstract 111 B and 1 00 oriented tellurium-doped GaSb single crystals have been grown by the vertical Bridgman technique, with and without a baffle in the melt. The axial and radial segregation has been investigated. Nearly diffusion controlled steady-state segregation was reached with a baffle submerged in the melt above the melt solid interface. The experimental data has been compared with the theoretically calculated curves using a segregation coefficient of 0.37 and distribution coefficient of 3 10 cm /s for Te in GaSb Elsevier Science B.V. All rights reserved. PACS: Eq; Fq; Fb; #g Keywords: GaSb; Segregation; Submerged baffle; Single crystals 1. Introduction During normal freezing, the compositional homogeneity of semiconductor crystals is strongly affected by melt convection. Since horizontal temperature and concentration gradients always yield some level of convection, virtually all semiconductor crystals exhibit significant axial and radial inhomogeneity. In the presence of gravity, the dopant concentration of directionally solidified crystals follows the theoretical profiles computed assuming * Corresponding author. Fax: # ; ostroa@rpi.edu. complete mixing in the melt [1], C (g) "k (1!g), (1) where C (g) is the dopant concentration in the solid at the location where a fraction g of the melt is solidified, C is the initial concentration of the solute in the melt and k is the equilibrium segregation coefficient, The ideal diffusion controlled steady-state segregation was reported only in the space-grown Tedoped InSb, grown by Witt et al. in Bridgman-type configuration [2]. In the space-grown GaSb crystal, steep initial transient in Te concentration was reported [3]. The transient does not seem to follow /98/$ Elsevier Science B.V. All rights reserved. PII S (98)

2 P.S. Dutta, A.G. Ostrogorsky / Journal of Crystal Growth 191 (1998) the Tiller s equation for diffusion controlled segregation [4], (x) "k#(1!k) 1!exp!k R D x, (2) where (x) is the dopant concentration in the solid at a distance x measured from the beginning of the specimen, D is the diffusion coefficient of the solute in the melt and R is the growth rate. The diffusion controlled steady state was not achieved [3]. To reduce convection driven by destabilizing radial (horizontal) temperature gradients, Müller used a vertical Bridgman-type configuration with thermally insulated side walls [5,6]. The heat was supplied to the melt and extracted from melt through two graphite cylinders positioned at the top and the bottom of the melt, respectively. A significant reduction of melt convection was achieved. The axial distribution of Te in the GaSb crystals was between the theoretical profiles computed assuming complete mixing (Eq. (1)) and diffusion controlled segregation (Eq. (2)). For Te-doped GaSb melt, Mu l- ler used k"0.37 and D"3 10 cm /s [5]. Numerical simulations and experiments have demonstrated that the use of a submerged heater in vertical Bridgman configuration drastically reduces the level of melt convection [7]. In most experiments, unpowered submerged heater which should be more appropriately designated as baffles were used. Using the submerged baffle, diffusion controlled steady-state impurity segregation has been achieved in Sn-1% Bi and in Te-doped InSb [8,9]. Diffusion controlled steady-state impurity segregation was not achieved in Ga- and Sb-doped Ge [9], because of the low equilibrium segregation coefficients of the dopants. Significant improvement of radial segregation was obtained by setting the baffle into oscillatory rotation [10]. The present work aims at obtaining telluriumdoped n-type GaSb with uniform electrical properties; a material system which is gaining technological importance [11]. 2. Experimental procedure The crystals were directionally solidified in a multizone Mellen furnace using the vertical Bridgman configuration with and without baffle in the melt. The details of the growth equipment and configurations are described elsewhere [10]. The 32 mm diameter Te-doped GaSb crystals were grown in silica crucibles with oriented seeds along 111 B and 100. The charges were doped with tellurium to obtain n-type conductivity. The growth runs were done under argon atmosphere in open crucibles with LiCl : KCl eutectic (58% : 42%) salt for liquid encapsulation to minimize volatilization of antimony from the melt and prevent direct contact and sticking of the crystal to the silica ampoule [12]. Freshly etched polycrystalline GaSb charge-doped with Te (synthesized from 6 N pure starting elements) was used for the experiments. The charges were etched using CP4, followed by thorough wash in de-ionized (DI) water and methanol with ultrasonic vibrations and nitrogen dried. The silica ampoules and silica baffles were cleaned with soap solutions and then in organic solvents followed by acid treatments and de-ionized water wash. The ampoules and baffle were thoroughly dried by blowing nitrogen and hot air. Prior to growth, the melt was homogenized by vertically moving the baffle (up and down) in the ampoule for 30 to 40 min to ensure uniform initial Te concentration in the melt. In the experiment conducted for comparison without baffle, the baffle was withdrawn from the melt after premixing. After seeding, growth was initiated by lowering the ampoule in the furnace through the axial temperature gradient (see Fig. 1). The temperature gradient near the melt solid interface was 15 K/cm. In the experiment conducted with the baffle, the baffle was placed around 1 cm above the melt solid interface. During the entire growth run, the baffle remained at a fix axial position (elevation), while the ampoule was lowered at a constant rate of 3.3 mm/h. At the end of growth, the furnace was slowly cooled down to room temperature. 3. Results Each of the grown single crystals contained one twin. They were characterized for the impurity distribution along the axial and radial directions by

3 906 P.S. Dutta, A.G. Ostrogorsky / Journal of Crystal Growth 191 (1998) Fig. 1. Experimental set up for synthesis and crystal growth along with the axial temperature profile in the furnace. measuring their resistivity. The spatial resistivity of crystals was measured by the four-point probe technique at room temperature. The measurements were repeated using various surface preparations of the specimens, different probe currents and pressure conditions. To correlate the measured resistivity with the Te content of the specimens, a simple methodology has been adopted. From Hall measurements, the resistivity versus net donor concentration in Te-doped n-gasb has been evaluated. The values of net donor concentration has been calculated from the Hall coefficient measured by Sagar [13], with the corrections for the two-band conduction in GaSb. The correlation between net donor concentration and actual Te-content in the crystal has been comprehensively obtained by Sunder et al. [14]. The Te-concentration has been measured either through secondary ion mass spectrometry (SIMS) or atomic absorption (AA) techniques. Combining the above two data, one can obtain resistivity as a function of Te-concentration. This has been used to calculate the Te-content in our crystals from the experimentally measured resistivity. The accuracy of the Hall measurements carried out by Sagar [13] was within $5%. The AA measurements by Sunder et al. [14] had an error within $2%. The error in our four-point probe resistivity measurements was $2%, except in the last-to-freeze portion of the crystal, where Te concentration exceeded 5 10 atoms/cm. In the last-to-freeze portion, the error approached $5% because of the low absolute value of resistivity. Taking into account the cumulative error of the Hall, AA and the four-point probe measurements, the total error in the estimation of the axial and radial Te concentration is: $9% of the measured values ($12% in the last-to-freeze portion).

4 P.S. Dutta, A.G. Ostrogorsky / Journal of Crystal Growth 191 (1998) Fig. 2. Axial Te-concentration in GaSb measured along the centerline and periphery, grown with a submerged baffle in the melt. R"3.3 mm/h with initial concentration of Te in the melt of 2 10 atoms/cm. The solid and the dashed lines are theoretically calculated profiles for complete mixing and pure diffusion respectively. k"0.37 and D"3 10 cm /s according to Ref. [5]. Fig. 3. Axial Te-concentration in n-gasb grown without a baffle in the melt. R"3.3 mm/h with initial concentration of Te in the melt of atoms/cm. The axial profile was measured in the center of the crystal along the growth axis. The solid and the dashed lines are theoretically calculated profiles for complete mixing and pure diffusion, respectively. k"0.37 and D"3 10 cm /s according to Ref. [5]. The experimental data for the axial Te-distribution in the GaSb grown with a baffle in the melt is shown in Fig. 2 along with the theoretical profiles computed assuming diffusive transport of Te (Eq. (1)) and for complete mixing (Eq. (2)). Following Müller [5], theoretical profiles were computed using k"0.37, D"3 10 cm /s and assuming that the rate of crucible translation is equal to the growth rate, R"3.3 mm/h. Although, the data show significant small scale variation, the distribution of Te is close to the theoretical profiles for diffusion controlled solidification at a constant growth rate. Based on the error bars, the scatter in the data shown in Figs. 2 and 3 is due to the actual variation in the Te content, not due to measurement errors. There is a 2 3 cm long plateau of nearly steady state segregation with (x)". The transient in the last-to-freeze portion occurred when the baffle separated from the melt. The decrease in the Te-concentration in the last centimeter of the crystal may be explained by enhanced mixing caused by Marangoni convection. There is little difference in Te-concentration along the centerline and along the periphery of the crystal (Fig. 2). The low level of convective interference with segregation is particularly surprising considering the low growth rate, only 3.3 mm/h. The low growth rate may have resulted in particularly flat growth interface. The axial Te-distribution in the crystal grown without the baffle is shown in Fig. 3 along with the theoretical plots of the Tiller s and Pfann s curves. The tellurium profile clearly follows the Pfann s curve for normal freezing with complete mixing. The radial variation in Te-concentration is more pronounced than in the crystal grown with the baffle. The small scale variations in Te-concentration are less pronounced. This may be explained by a thinner solute layer, which should be less sensitive to variations in the growth rate, convection, or perhaps other effects. The mobility of the crystals grown with carrier concentration in the range of mid-10 to mid- 10 cm is around 1800 and 6000 cm /V s at 300 and 77 K, respectively. The etch pit densities varied in the range of cm. 4. Summary Nearly diffusion controlled segregation of Te in GaSb was demonstrated using a submerged baffle in the vertical Bridgman configuration, despite the

5 908 P.S. Dutta, A.G. Ostrogorsky / Journal of Crystal Growth 191 (1998) low growth rate (3.3 mm/h). Further studies will include growth at higher growth rates and interface demarcation through current pulsing. Acknowledgements This work is supported by the Microgravity Science and Applications Division of the National Aeronautics and Space Administration. The growth equipment used in the present work was provided by the Lucent Technologies. References [1] W.G. Pfann, J. Metals 194 (1952) 747. [2] A.F. Witt, H.C. Gatos, M. Lichtensteiger, M.C. Lavine, C.J. Herman, J. Electrochem. Soc. 122 (1975) 276. [3] T. Nakamura, T. Nishinaga, P. Ge, C. Huo, Abstracts of the Joint Xth European and VIth Russian Symp. on Physical Sciences in Microgravity, St. Petersburg, Russia, 1997, p [4] W.A. Tiller, K.A. Jackson, J.W. Rutter, B. Chalmers, Acta Met. 1 (1953) 428. [5] G. Mu ller, Convection and Inhomogeneities in Crystal Growth from the Melt Crystals, vol. 12, Springer, Berlin, [6] G. Mu ller, German patent P (1984). [7] A.G. Ostrogorsky, G. Mu ller, J. Crystal Growth 137 (1994) 64. [8] A.G. Ostrogorsky, F. Mosel, M.T. Schmidt, J. Crystal Growth 110 (1991) 950. [9] A.G. Ostrogorsky, H.J. Sell, S. Scharl, G. Mu ller, J. Crystal Growth 128 (1993) 201. [10] S. Meyer, A.G. Ostrogorsky, J. Crystal Growth 171 (1997) 566. [11] P.S. Dutta, H.L. Bhat, V. Kumar, J. Appl. Phys. 81 (1997) [12] J.P. Garandet, T. Duffar, J.J. Favier, J. Crystal Growth 96 (1989) 888. [13] A. Sagar, Phys. Rev. 117 (1960) 93. [14] W.A. Sunder, R.L. Barns, T.Y. Kometani, J.M. Parsey Jr., R.A. Laudise, J. Crystal Growth 78 (1986) 9.