The influence of growth conditions on the quality of

INSTITUTE OF PHYSICS PUBLISHING

SEMICONDUCTOR SCIENCE AND TECHNOLOGY

Semicond. Sci. Technol. 16 (2001) 514–520

www.iop.org/Journals/ss

PII: S0268-1242(01)21090-6

The influence of growth conditions on the quality of CdZnTe single crystals J Franc1 , R Grill1 , P Hl´ıdek1 , E Belas1 , L Turjanska1 , P Höschl1 , I Turkevych1 , A L Toth2 , P Moravec1 and H Sitter3 1 Institute of Physics, Charles University, Ke Karlovu 5, Prague 2, CZ-121 16, Czech Republic 2 MTA MFA, Structure Research, H-1535 Budapest-114 POB 49, Hungary 3 Institute of Semiconductor Physics, Johannes Kepler University, Altenbergerstrasse 69, Linz, Austria

Received 18 January 2001, accepted for publication 20 April 2001 Abstract Experimental conditions were investigated for growth of inclusion-free near-stoichiometric CdZnTe single crystals with a minimized concentration of native point defects. The positions of the stoichiometric line P S = 8 × 105 exp(−1.76 × 104 /T ) (atm) and the room-temperature and high-temperature p–n lines were evaluated from high-temperature in situ galvanomagnetic measurements. The Cd pressure at the congruent melting point was estimated at ∼1.15–1.20 atm from analysis of the total inclusion volume of five single crystals fabricated at Cd pressures in the range of 1–1.3 atm. An inclusion-free single crystal was prepared at PCd ∼ 1.2 atm. Calculations based on a model of two major defects, the Cd vacancy and the Cd interstitial, show that a very small deviation of PCd from P S results in a large generation of the native defects. Thus a reproducible production of a high-resistivity material by a slow cooling along the P S seems to be very difficult.

1. Introduction CdZnTe (CZT) has been the principal lattice-matched substrate for epitaxial growth of (HgCd)Te since the mid-1980s. Although alternative substrates (sapphire, Si) are under study, CZT is expected to keep its leading position in the next millennium, especially in the case of fabrication of detectors used in the long-wavelength infrared spectral region, and in advanced applications such as two- and multicolour focal plane arrays. The number of pixel counts now available in photovoltaic HgCdTe detector elements has increased to a value of 10k × 10k. Due to the fact that the pixel size is restricted by the infrared radiation wavelength and focusing possibilities, the increase of pixel counts requires larger near-perfect CZT substrates. One of the critical parameters limiting the substrate quality is the presence of inclusions and precipitates of one of the components in the as-grown crystals, caused by deviations from stoichiometry. A review of results on the formation of inclusions and precipitates and ways of eliminating them in CdTe and CZT was given in [1]. While inclusions are formed by melt-droplet capture near the growing interface at the growth temperature, precipitates are formed due to the retrograde solubility of native point defects during the cooling process of the solidified crystal. Study of precipitates 0268-1242/01/060514+07$30.00

© 2001 IOP Publishing Ltd

on a transmission electron microscope showed their maximum dimensions to be 30 nm [2]. The standard way of reducing the density of secondphase particles is an annealing of CZT wafers in a Cd or Te atmosphere. This technological step, however, increases the cost, and often results in a lower crystallographic quality of the annealed substrate. Therefore the research effort has been focused on the elimination of deviations from stoichiometry during the crystal growth, looking for the optimal Cd pressure and cooling conditions. The purpose of this paper is to find optimal stoichiometric growth conditions based on the evaluation of high-temperature (HT) galvanomagnetic properties of CdTe [3].

2. Experiment CZT single crystals (with a Zn content ∼4%) were fabricated by the vertical gradient freeze method [4]. The crystals were prepared in a quartz ampoule with a diameter of 105 mm, in which a growth crucible with a diameter 100 mm was located at Cd overpressure 1–1.3 atm. 6N purity starting elements were used in all growth runs. The Cd pressure at the maximum temperature during the growth was reached by adding the

Printed in the UK

514

The influence of growth conditions on the quality of CdZnTe single crystals

Table 1. Basic parameters of crystals used in the study. Crystal

PCd (atm)

Type

Concentration (cm−3 )

ρ ( cm)

µh (cm2 V−1 s−1 )

Le , 300 K (µm)

Le , 80 K (µm)

A B C D E

1.3 1.2 1.2 1.05 1

P N P P P

6.8 × 1014 1.8 × 108 4.6 × 1015 1.8 × 1016 1.9 × 1016

110 1.1 × 108 17 6.2 3.4

85 210 80 35 97

3.5 — 1.3 1.2 1.7

71 — 4.1 1.3 7.6

corresponding amount of Cd to the ampoule. The cooling speed during the crystallization process at temperatures 1120– 1080 ◦ C was 0.5 ◦ C h−1 . All crystals were very high quality single crystals filling about 90% of the crucible volume with no macroscopical defects (grain boundaries, twins). A complex study of the physical properties of as-grown single crystals was performed. The concentration of the inclusions was investigated using an infrared microscope enabling the observation of inclusions larger than 1 µm. A wafer of thickness 1.5 mm was cut along the (111) growth axis of the crystal. The average concentration of inclusions was calculated at several positions along the growth axis. Optical properties were investigated by a FTIR spectrometer (BRUKER IFS 66/S) with Si and TGS detectors in a continuous flow He cryostat. Spectra were corrected with respect to the sensitivity of the detectors. Two types of contacts were used for Hall effect and conductivity galvanomagnetic measurements—chemically prepared Cu contacts for lowresistivity samples, and molten Pt contacts in the case of the high-resistivity crystal B (see table 1). The diffusion length of minority carriers was measured by an electron-beam-induced current (EBIC) method using a JEOL-50 scanning electron microscope with a cooled holder. An Au-evaporated Shockley barrier was used to separate electron–hole pairs in the EBIC method. In order to find the intrinsic stoichiometric line, we performed two types of experiments. In the first, the influence of annealing conditions (T , PCd ) on the electrical properties of the material quenched to room temperature (RT) was studied. Samples were annealed in the two-zone or one-zone furnace at temperatures of 700, 750, 800, 850 and 900 ◦ C and defined values of PCd within the whole stability region. RT Hall effect and conductivity measurements were performed in a classical arrangement and the position of the stoichiometry line was established from the sign change of the Hall coefficient at RT. The second experiment was based on the in situ measurements of the Hall effect and conductivity at temperatures 600–1050 ◦ C [3]. The theoretical analysis of such measurements allowed us to establish the HT intrinsic conditions in the P –T diagram and to correlate them with the stoichiometry of the samples.

3. Theory The effort to obtain high-quality CdTe and CZT samples for substrates or radiation detectors is connected with the demand to prepare the stoichiometric single-crystalline material without inclusions and precipitates and with a low density of point defects. The thermodynamic conditions which characterize such a treatment are defined by the position of the

stoichiometry line (P S ) in the P –T diagram at which the lattice defects are mutually compensated, and the anion and cation atomic fraction relation xCd +xZn = xTe = 0.5 is fulfilled. Due to the small amount of Zn in our single crystals and in order to simplify the model calculations, the Zn-related defects will be omitted in this paper. Here we are interested in undoped materials, and therefore include only native defects in our analysis. The slow cooling of the as-grown crystal along the P S results in sample formation without precipitates, which are formed due to the native defect supersaturation. The annealing of the stoichiometric CZT at lower temperature results in the recombination of the defects producing pure materials, without claiming to add some species by the diffusion from an external source. The process rate is limited only by the necessary thermal balancing during the cooling, where the temporary nonequilibrium between the solid CZT and the surrounding gas phase can cause the escape of one component (mostly Cd) from the sample and the ensuing stoichiometry deviation. Ignoring one- and two-dimensional defects, the stoichiometry deviations are determined by native defects present in the ideal lattice. The native defects (all divalent) which should be taken into account are [5] the Cd or Te interstital donors CdI and TeI , the Te vacancy VTe , and the Te antisite TeCd . The acceptor is the Cd vacancy VCd . The nature of the Cd antisite CdTe has not yet been identified. Due to the low content of Zn in our samples we do not include Zn-native defects in the list. The densities of neutral native defects are given by reaction constants [6] √ X n0 T PCd CdI = K 1 vib vib , (1) × exp − + S(CdI ) E(CdI ) + U (CdI ) kB T X n0 K TeI = √ TP Cd 1 (2) × exp − E(TeI ) + U (TeI )vib + S(TeI )vib , kB T X n0 K VCd = √ TP Cd 1 E(VCd ) + U (VCd )vib + S(VCd )vib , (3) × exp − kB T √ X n0 T PCd VTe = K 1 × exp − (4) E(VTe ) + U (VTe )vib + S(VTe )vib , kB T 2 X n0 T PCd CdTe = 2 K 515

J Franc et al

1 exp − E(CdTe ) + U (CdTe )vib + S(CdTe )vib , (5) kB T X n0 K 2 TeCd = 2 T PCd 1 vib vib + S(TeCd ) , × exp − E(TeCd ) + U (TeCd ) kB T (6) where E are relevant defect formation energies, and the vibrational energy U vib and entropy S vib determine contributions to the vibrational free energy. Here, K = kB T

3

mCd kB 2π¯h2

3/2 ,

n0 = 1.48 × 1022 cm−3 is the number of unit cells per volume, kB is the Boltzmann constant, and mCd is the mass of a Cd atom. The label X is used for neutral defects overall. The densities of multiply ionized defects are calculated for acceptors and donors: z − X gX z − zµF − Ea1 − · · · − Eaz X exp = X , (7) gX X kB T

Xz

+

1 g z+ Ed + · · · + Edz − zµF = XX X exp , gX X kB T

(8)

where Ea and Ed are the acceptor and donor one-electron ionization energies and µF is the Fermi energy, which is obtained by solving the electric neutrality condition. gXz− (gXz+ ) is the degeneracy of the defect. The index z signifies the degree of ionization. The list of experimentally determined ionization energies of native defects is given in [3]. The experimentally determined energy gap Eg according to [7] Eg = 1.622 − 3.5 × 10−4 T − 1.1 × 10−7 T 2 (eV)

(9)

is used at all temperatures. It has been shown by ab initio calculations [5] and confirmed by in situ galvanomagnetic measurements [3] that the dominant donor (acceptor) native defects are CdI (VCd ), both being doubly ionized near P S . In contrast, the shallow donor TeCd proposed in [5] was not approved in [3]. The simplified electric neutrality condition in which and L electrons [8], holes and singly and doubly ionized native defects are included reads 2− − + 2 VCd = p + Cd+i + 2 Cd2+ (10) n + nL + VCd i . Though CdI and VCd are sufficient to describe electric properties of CZT, there is still a discrepancy at the optical density measurements of CZT homogeneity range [9, 10], which extend both at Cd and at Te-rich conditions about 3–5 times further from 50% atm than is obtained from electrical measurements. Consequently, a considerable amount of neutral defects which are not detected by electrical measurements should be present in CZT. At Te saturation TeCd was suggested [5,11] to be the most acceptable defect. We are not familiar with any attempts to describe the Cd-saturated side. We suppose that the deep-donor VTe could be a proper defect for this case. 516

In view of the complicated defect structure of CZT and the number of unknown parameters, the endeavour to establish the position of the stoichiometry line looks rather hopeless. There is, however, a good amount of knowledge which can be used to simplify this task, and to reach favourable result without demanding the characterization of unknown defects. We can take advantage of a well known feature of the defect statistics, which is expressed by the pressure dependence of the neutral and ionized defect densities. Based on the mass action relations [12], the PCd dependence of charged defect density is significantly more moderate than that in the case of neutral species. For the above-mentioned dominant 1/3 −1/3 2− charged defects we obtain [Cd2+ I ] ∝ PCd and [VCd ] ∝ PCd . With regard to the position of P S , which is below 1000 ◦ C located by more than one order of magnitude in PCd from the three-phase line, the concentration of all neutral (≡ deep) defects is significantly reduced (10×–100×) compared to the moderate decrease (∼2×) at the doubly ionized (≡ shallow) defects. Consequently, although the neutral defects dominate at saturated conditions, their effect is minimized inside the homogeneity range far from the three-phase line and ionized defects are sufficient to be included in the theory looking for the position of P S . The same reason for the omission is also valid at the native defect complexes, which are created by the native defects reactions [5] and can be connected with the stoichiometry deviations (e.g. VCd TeCd , CdI VTe , TeI TeCd , etc). The density of all such complexes is strongly dependent on PCd and can be thus eliminated from the discussion. The thermodynamic diversity between shallow and deep defects mentioned above are convenient for the adjustment of stoichiometric samples. If the slightly nonstoichiometric sample is cooled to the temperature at which only shortrange diffusion is active and the solid–gas material exchange is insignificant, the internal effective PCd moves to the saturation limit and the shallow/deep defect density rate exceeds the equilibrium distribution at such T . Consequently, a large proportion of shallow defects convert into deep ones, decreasing the effect on the electrical properties of CZT. The effect must be assisted by other types of lattice defects, e.g. dislocations, to allow the defect conversion. For example the conversion of two shallow VCd into possibly deep donor TeCd requires a pair Cd+Te from an edge dislocation. The estimated concentration of dislocations >104 cm−2 also in very good single crystals is sufficient to guarantee such process above 450 ◦ C. P S (defined by xCd = xTe = 0.5 for CdTe) is usually coupled with the p–n line (P I ), which defines the intrinsic material given by the electron and hole concentration equilibrium n = p. Though both lines are located close together in the P –T diagram and their association is roughly tolerable, we should emphasize their different nature here in order to appreciate the demands which are put on the CZT samples. Generally, we can specify two p–n lines in the P –T diagram (see figure 1), which differ by a method to be I established. The first one (PRT ) is defined by the change of the type of conductivity p ↔ n at RT of samples quenched I after annealing at defined T and PCd . PRT is practically S identical to P if shallow defects dominate the deep defects. The precision of the assignation is limited by the quenching


101

1100

1000

900

800

700

T ( oC)

600

500

10

19

10

18

PIHT

10-1 10-2

10

17

T=900 C

-3

Defect concentration (cm )

10

PCd (atm)

20

o

0

[13]

[1] [7]

PIRT

10-3 10-4 P

10-5

S

h 2Cdi

1/IeRHI

+

2+

nL

10

16

10

15

10

14

nΓ

V h

Cdi

+

V

10-6 10-7 0.7

10

10

-3

10

-2

+

2V

-

2Cd

Cd

-

n rt

rt

x

Cd

Cd

10

-1

10

0

x i

10

1

PCd(atm) 0.8

0.9

1.0

1.1

1.2

1.3

1000/T (K)

Figure 1. Partial pressure of Cd along the three-phase loop of CdTe(S) and the set of stoichiometric and p–n lines. The full curve represents the stoichiometric line P S obtained within our model, in which only electrically active native defects are included. I Dash-dotted and dashed lines plot the HT intrinsic line PHT and RT I line PRT , which was obtained by quenching. Long–short dashed and dotted lines show the results of [7] and [13].

velocity. Also, in case of very fast cooling, it takes about 1 min to decrease the sample temperature significantly (below ca 550 ◦ C) to stop the diffusion in the sample and to prevent the more volatile component (Cd) escaping. Consequently, the p–n line is systematically shifted to the higher PCd in this case. I ) is based on the HT The second procedure (PHT in situ galvanomagnetic measurements taken under the thermodynamic conditions at which the position of the p– n line is to be established. The method is based on the fact that in the case where both dominant defects are doubly ionized, the sample behaves as intrinsic near the stoichiometry line. The p–n line is located in the P –T diagram, where the galvanomagnetic properties are identical to the properties of the intrinsic semiconductor at the same temperature. We profit from the fact that the electron mobility is dominated by the phonon scattering at HT, being practically independent of the defect densities. Such a line is obtained by the thermodynamic equilibrium measurements and is not influenced by the undefined transformation during the cooling.

4. Results and discussion At first we concentrated on finding a position for the stoichiometric and p–n lines in the material. This problem was addressed by several authors experimentally [7], and theoretically [7, 13] (see figure 1). These results disagree significantly. In the case of [7] the steep slope of the line indicates that data were affected by undetermined donors at a concentration of ∼5 × 1016 cm−3 in the samples. The theoretical calculation in [13] is based on HT in situ Hall measurements [12] which roughly agree with our experimental data. The model does not, however, take the occupation of Lminima at HT into account and the p–n line is consequently shifted about 4× to higher PCd . The results of our calculations are based on annealing/quenching experiments and the in situ galvanomagnetic measurements in CZT [3]. In figure 1 we show the positions

Figure 2. The neutral (X) and ionized (+; −) native defect − concentrations together with and L electron (n− , nL ) and hole (h+ ) concentrations, all at 900 ◦ C, are represented by full curves. The experimental (circles) and theoretical (dotted curve) Hall concentrations are involved. The dashed curves show RT n− RT and h+RT . The arrows show the saturation limits.

of stoichiometry and p–n lines defined in section 3. We see that the RT p–n line 1.58 × 104 I = 2.4 × 105 exp − (atm), (11) PRT T which was obtained by the quenching of annealed samples, is shifted to a higher pressure relatively to the HT intrinsic p–n line: 1.73 × 104 I 5 (atm). (12) PHT = 4.5 × 10 exp − T I PHT crosses the three-phase line at PCd = 1.4 atm. Based on the fit of HT experiments we could establish the position of the stoichiometry line where only electrically active (shallow) defects were included: 1.76 × 104 (atm). (13) P S = 8 × 105 exp − T

With respect to the reasoning in section 3 we prefer this line as the best approach, which conforms to the present knowledge about CZT. P S crosses the three-phase line at PCds = 1.9 atm, which is higher than the congruent melting point pressure (PCdm = 1.266 [10]). The effect of the four L minima in the conduction band to the native defect statistics at 900 ◦ C is outlined in figure 2. Due to the occupation of the L bands the relevant electron concentration is not expressed satisfactorily by 1/|eRH |. If higher minima in the band structure are occupied, only the simultaneous analysis of both RH and σ describes the carrier concentrations and mobilities correctly. Consequently the concentration of native defects can be two or more times higher than electrons being assumed to be only in the -point. The dashed curves show the concentration of electrons and holes at RT. Fabrication of inclusion- and precipitate-free single crystals depends on several critical technological steps: solidification of the melt at the congruent melting point (CMP) PCdm to avoid segregation of inclusions near the growth 517

J Franc et al 1000

900

800

700

10

500

19

-2x10 17

100

-10

17

-5x10

-1

10

16

-2x10

D

16

-10

10-2

16

10

-5x10

18

-3

E A

17

10

10

15

-2x10 15

5x

PCd (atm)

T ( oC)

600

-3

1100

∆(cm )

101

10 2x

B 10

15

10 10 2x 5x

17

10-4

17

10-5

15

16

10

16

0.7

0.8

0.9

1.0

1.1

1.2

1.4

1.3

1000/T (K)

Figure 3. P –T diagram with isoconcentration lines for deviation from stoichiometry.

boundary, annealing the excess Te present in the solidified ingot due to the fact that the CMP is shifted towards the Te site of the T -x diagram, and cooling the crystal along the stoichiometric p–n line. For the case where inclusions are formed during solidification of the melt, a method for their removal in a temperature gradient has been proposed [13]. The annealing must be done under stoichiometric conditions at high temperatures, where the Cd diffusion coefficient is high enough to reach the equilibrium in the whole crystal at real times. In the ideal case when the crystal is cooled exactly along the p–n line, precipitate-free highresistivity materials could be obtained. We show in figure 3 the isoconcentration lines of deviation from stoichiometry "n = 4n0 (xCd − 0.5) due to the electrically active defects. This figure illustrates the fact that it is very complicated to prepare the high-resistivity undoped CdTe. If we assume the high-resistivity material to be limited by a free carrier concentration (∝"n(p)) less than 109 cm−3 at RT and the minimum temperature (∼700 ◦ C) where the diffusion in large scale (∼1 cm) samples really occurs, then we get an extremely narrow interval "PCd /P S ∼ 10−3 % at which such a sample should be annealed. Such experimental precision is very difficult to match in real experiments and for the highresistivity materials the proper deep defect doping should therefore be used. The maximum resistivity which can realistically be reached in undoped CdTe can be obtained pinning the Fermi energy to the deep VCd level, which is located at ≈470 meV above the valence band. Such a situation corresponds to a p-type conductivity and a typical resistivity of 106 cm. In order to optimize the growth conditions from the point of view of minimization of inclusions in the material we performed the growth of five crystals under different Cd pressures in the range of 1–1.3 atm near the CMP in order to verify and precisely identify its position. The position of the CMP has been measured by the authors of [10] as PCd = 1.266 atm. It should be noted that HT measurements of deviations from stoichiometry are very difficult and that so far only relatively few points are available in the region around the CMP in the P –T diagram. 518

1.2

pCd(atm)

-7

10

18

1.0

10

16

0

10

C 10

2x

1 5x

-6

Figure 4. The deviation from stoichiometry calculated from an average volume of inclusions with dependence on PCd in the ampoule.

A list of single crystals used in the study is given in table 1. Figure 4 depicts the volume of second phase inclusions with dependence on the Cd overpressure estimated from the concentration of inclusions and their average dimensions. Crystals prepared at Cd pressures above 1.2 atm contained inclusions of diameter ∼10 µm, crystals fabricated at Cd pressure ∼1 atm contained a higher concentration of inclusions with a smaller diameter (1–5 µm). One single crystal grown at ∼1.2 atm of Cd pressure contained no inclusions in the whole volume. Based on the results presented we conclude that the position of the CMP can be estimated in the range PCd = 1.15– 1.20 atm, which is only slightly lower than the results in [10]. As is apparent from figure 5, the deviation from stoichiometry caused by a small deviation from the CMP is relatively large (∼1018 cm−3 ), and is caused by the flat shape of the T –x diagram, when a small change of T results in a large "x [1]. The position of the CMP from our experiments is depicted in figure 5. Only a qualitative comparison between RT galvanomagnetic data (table 1) with the calculated deviation stoichiometry curves with dependence on PCd (figure 3) can be given. Figure 3 was calculated based on the assumption of no significant exchange between the Cd vapour and solid phases. This is probably a good approximation in the case of large crystal volumes and real cooling speeds used during the crystal growth. However, at temperatures below the melting point, when the Cd diffusion coefficient is high [14], Cd can diffuse several centimetres into the crystal volume. This can cause deviations from the model calculation presented in figure 3. All except one of the crystals studied are low-resistivity p-types at RT. This is in good agreement with the range of PCd = 1 used, i.e. 1.3 atm, which is well below the crossing of the stoichiometric line with the three-phase boundary at PCd = 1.9 atm (see figure 5). The high resistivity of crystal B can be explained by the fact that during cooling, due to some exchange with the gas phase, the path through the P –T diagram was close to the stoichiometric line. Only a small concentration of unintentional contamination with some deep impurity can result in pinning of the Fermi level in the middle of the gap and a high crystal resistivity.


Figure 7. The optical transmissivity of the crystals studied at 300 K. Figure 5. Detail of the P –T diagram. Open triangles represent experimental points determining the region of stability from [11]. The position of the CMP estimated from our experiments is shown by a full diamond. Dash-dotted and dashed lines plot the HT I I intrinsic line PHT and the RT line PRT , which was obtained by quenching. 100

5. Conclusion

80 A 60

L(µm)

resulting from a high concentration of VCd at the Te-rich side of the phase diagram. Absorption of free carriers also causes a drop of transmissivity in the long-wavelength infrared part of the spectrum of the low-resisitivity inclusion-free crystal D.

40 E C

20

D 0 0.004

0.008

0.012

0.016

0.020

-1

1/T(K )

Figure 6. The temperature dependence of diffusion length of minority electrons in the crystals studied.

RT data of the diffusion length of minority electrons L and the corresponding mobility–lifetime product are also shown in table 1. Results of temperature dependences (60–300 K) of L of p-type samples from table 1 are shown in figure 6. Diffusion length L in all samples studied has a tendency to increase with decreasing temperature. Samples with a higher deviation from stoichiometry and a very low resistivity at RT (C, D, E) show only a moderate increase. In agreement with [15] where a correlation of increased recombination, which is manifested by low L with a VCd -related complex was observed, we conclude that the deviation from stoichiometry towards Terich conditions favouring the formation of VCd results in the deterioration of L. This effect which can be seen at RT data in table 1 increases strongly with decreasing T . Results of measurement of optical transmissivity are shown in figure 7. Crystals grown at PCd = 1.2 atm or higher exhibit very good transmissivity within the whole spectral range studied. Crystals prepared under low PCd ∼ 1 atm have a poor transmissivity due to the prevailing absorption on free carriers in these high-concentration, low-resistivity crystals,

The parameters important for growth of inclusion- and precipitate-free CdTe and CZT single crystals with a minimum concentration of point defects were studied, namely the position of the CMP and of the stoichiometric P S line in the P –T diagram. The Cd pressure at the CMP was estimated at ∼1.15–1.20 atm. An inclusion-free single crystal was fabricated at PCd ∼ 1.2 atm. The positions of the intrinsic p–n lines at high and room temperatures were established. The set of isoconcentration lines of deviation from stoichiometry based on a model of two dominant native defects, the Cd vacancy and the Cd interstitial, shows that a very small deviation of PCd from P S results in a large generation of the native defects. Thus a reproducible production of a high-resistivity material by a slow cooling along the P S seems to be very difficult. However, cooling of crystals close to P S results in improved optical properties, a higher mobility–lifetime product and higher resistivity.

References [1] Rudolph P 1998 Melt growth of II-VI compound single crystals Recent Development of Bulk Crystal Growth ed Milssiki (Research Signpost) [2] Rai R S, Mahajan S, McDewitt S and Johnson D J 1991 J. Vac. Sci. Technol. B 9 1892 [3] Franc J, Höschl P, Grill R, Turjanska L, Belas E and Moravec P J. Electron. Mater. at press [4] Höschl P, Ivanov Yu M, Belas E, Franc J, Grill R, Hl´ıdek P, Moravec P, Zvára M, Sitter H and Toth A L 1998 J. Cryst. Growth 184/185 1039 [5] Berding M A 1999 Phys. Rev. B 60 8943 [6] Berding M A, van Schilfgaarde M and Sher A 1994 Phys. Rev. B 50 1519 [7] Nobel D 1959 Philips Res. Rep. 14 361 Nobel D 1959 Philips Res. Rep. 14 430 [8] Franc J, Grill R, Turjanska L, Höschl P, Belas E and Moravec P 2001 J. Appl. Phys. 89 786 [9] Greenberg J H 1999 J. Cryst. Growth 197 406 [10] Fang Rei and Brebrick R F 1996 J. Phys. Chem. Sol. 57 443

519

J Franc et al

[11] Brebrick R F and Fang Rei J 1996 J. Phys. Chem. Sol. 57 451 [12] Chern S S, Vydyanath H R and Kröger F A 1975 J. Solid State. Chem. 14 33 [13] Vydyanath H R, Ellsworth J, Kennedy J J, Dean B, Johnson C J, Neugebauer G T, Sepich J and Liao Pok-Kai

520

1992 J. Vac. Sci. Technol. B 10 1476 [14] Zanio K 1978 Semiconductors and Semimetals vol 13, ed R K Willardson and A C Beer (New York: Academic) p 125 [15] Franc J, Hl´ıdek P, Sitter H, Belas E, Franc J, Toth A L, Turjanska L and Höschl P 1999 Physica B 273 883