Preparation and properties of CdSnAs2

J. Phys.

Chem. Solids

Pergamon

Press 1961. Vol. 17, Nos. 314, pp. 278-283.

PREPARATION

AND

A. J. STRAUSS Lincoln Laboratory,?

Massachusetts

PROPERTIES and

Printed in Great Britain.

OF CdSnAs,

A. J. ROSENBERG*

Institute of Technology,

Lexington

73, Massachusetts

1 July 1960)

(Received

Abstract-n-type samples of Cd&Ass (melting point = 590”-600°C) have been prepared by freezing from the melt. The highest room temperature Hall mobility attained was 1.2 x lo4 cm*/volt set for a sample with a free electron concentration of 5.5 x 101’ cm-*. From infrared absorption data the energy gap is estimated to be approximately 0.23 eV at room temperature. The absorption edge shifts to shorter wavelengths with increasing electron concentration. The electrical and optical data indicate that the conduction band of CdSnAss is characterized by a low electron effective mass, of the order of a few hundredths of the free electron mass. INTRODUCTION THE SYNTHESIS of ten intermetallic

semiconductors of the composition A~BrvCvz-_where AI1 is Cd or Zn, BIv is Si, Ge, or Sn, and Cvis P or As-has been reported.(r*a) Except for ZnSnAsa, these compounds crystallize in the tetragonal chalcopyrite structure, in which each atom is tetrahedrally bonded. They are isoelectronic with the III-V compounds, from which they are formally derived by the ordered substitution of one Group I1 and one Group IV atom for each pair of Group III atoms. An electron mobility of 3000 ems/volt set has been reported for CdSnAsz,(r) but no other data on the electrical properties of the II-IV-V compounds have been published. Energy gaps for six of the compounds (not including Cd&Ass) have been estimated from optical absorption measurements. (2) The present investigation was undertaken because of our interest in compound semiconductors with high carrier mobilities. Its most notable result has been the preparation of CdSnAss with a room temperature Hall mobility of 1.2 x 104 cma/ volt set at an electron concentration of 5.5 x 1017 cm-s. This mobility is by far the highest reported for any ternary compound, and is exceeded at room temperature only by values measured for four binary compounds with the zinc blende + Now at TYCO, Inc., Waltham, Massachusetts. t Operated with support from the U.S. Army, Navy, and Air Force. 278

structure-InSb, InAs, HgTe, and HgSe. Infrared transmission measurements indicate that the energy gap of CdSnAsa is probably about 0.23 eV at room temperature. The absorption edge shifts to shorter wavelengths with increasing electron concentration as observed in the case of n-type InSb@) and InAs.(4) SAMPLE PREF’ARATION Samples of the compound were synthesized by melting stoichiometric quantities of the elements in evacuated and sealed quartz tubes. Three different crystallization techniques were used: vertical Bridgman, quenching, and slow cooling. The Bridgman apparatus consisted of two resistance furnaces, the lower at 750°C for melting the material and the upper at 600°C for maintaining the arsenic vapor pressure at approximately one atmosphere. Directional freezing was accomplished by lowering the sample tube out of the apparatus at O-1 in/hr. The ingot prepared in this manner consisted mostly of polycrystalline CdSnAsz as a single phase, but the last-to-freeze portion contained two phases, due to the loss of arsenic by condensation on the upper part of the sample tube. The single phase material was threaded with many fine cracks, but it was possible to isolate a crack-free sample for electrical measurements. The grain size of this sample was only about 20 cc, as estimated from a Laue pattern. Quenched samples were prepared by immersing quartz tubes containing the molten compound in

PREPARATION

AND

PROPERTIES

liquid nitrogen. The ingots obtained were virtually free of cracks but were polycrystalline, with an estimated grain size of only 10 CL. A number of samples were prepared by slow cooling in the course of thermal analysis experiments performed to determine the freezing point of CdSnAss. In these experiments, a chromelalumel thermocouple was inserted into a re-entrant in the bottom of a quartz sample tube. The sample tube was placed in a massive stainless steel cylinder, and the whole assembly was heated in a wellinsulated vertical resistance furnace until the compound had been melted. The furnace power was then reduced sufficiently to establish a cooling rate of approximately 1 deg/min until freezing (non-directional) was completed, after which the furnace was turned off. Although the ingots prepared in this manner were very severely cracked, the grain size was much larger than for the Bridgman-grown and quenched ingots. In several cases, it was possible to obtain crack-free single crystals of the order of 1 cm in their longest dimension. THERMAL ANALYSIS Three cooling curves and two heating curves have been determined in thermal analysis experiments on Cd&Ass. The solid-liquid transformation occurs between 590°C and 600°C; a more exact value for the transition temperature cannot be given, since both cooling and heating curves reveal complexities in the phase diagram which are not well understood. One of the cooling curves is shown in Fig. 1, while a heating curve is shown in Fig. 2. Fig. 1 shows only the voltage readings for the thermocouple inserted into the re-entrant in the sample tube, while Fig. 2 also gives the differences between these readings and those for a thermocouple inserted into the wall of the stainless steel cylinder. Both curves shown exhibit three thermal arrests, rather than the single temperature plateau characteristic of a solid-liquid transformation in a simple phase diagram. The other curves obtained have the same general features but are somewhat different in detail. The presence of two plateaus in the cooling curves suggests the existence of a metastable solid phase whose liquidus temperature, at the composition corresponding to CdSnAss, is lower than the liquidus temperature of the stable

OF CdSnAsa

26

279

I

25

Y 0 24 z 8 z i

Y 23 4 I

-I---

22

21

1

50

TIME

(mid

FIG. 1. Cooling curve for CdSnAsa. compound. If thisis the case, of the stable phase permits metastable phase, and the served. This phenomenon

sufficient supercooling crystallization of the initial plateau is oboccurs in the Cd-Sb

26

-

SAMPLE THERMOCOUPLE

----REFERENCE

TC-SAMPLE

TC

25

12

23

0

10

20 TIME

30

40

6nin)

FIG. 2. Heating curve for CdSnAsp.

A. J.

280

STRAUSS

and A. J.

system, where the metastable phase CdsSba crystallizes from a slowly cooled melt of the composition CdSb, even though its liquidus is approximately 50°C below that of the stable compound CdSb.(b) If the metastable phase then transforms rapidly into the stable phase, and the transformed solid comes into contact with the remaining liquid, crystallization of the stable phase is nucleated. The temperature then rises rapidly to the stable liquidus, and a second plateau is observed. This type of nucleation frequently occurs in alloys of CdSb with ZnSb.ta) The third arrest, beginning at 552°C in Fig. 1, could result from an allotropic transformation of the stable phase, whether this phase had been formed by direct crystallization from the melt or by transformation of metastable material. While the model proposed can explain the cooling curves obtained for CdSnAsa, it does not account for the observed heating curves. On the basis of this model, the heating curves should exhibit a thermal arrest at 5%560°C due to the allotropic transformation of the stable phase, followed by a plateau at the melting point of the hightemperature form. Instead, the measured curves exhibit three poorly defined arrests, all of which begin at significantly higher temperatures than 56O”C, and the last of which extends to a slightly higher temperature than the upper plateau in the cooling curve. It is clear, therefore, that a more extensive investigation of the Cd-Sn-As system

ROSENBERG

will be required to elucidate the phase diagram in the vicinity of CdSnAsa. LATTICE

PARAMETIlRs

X-ray diffractometer measurements on two powdered samples of CdSnAsa, one prepared by slow cooling and the other by quenching, gave the same results within the limits of experimental error. The following average values were obtained for the tetragonal lattice parameters: a = 6.093 + 0.003, and c/a = 1.959 of: 0406. These data are in agreement with the published values of 6.092 and 1.957, respectively. (1) ELECTRICAL

PROPERTIES

The Hall coefficient (&) at 7800 G and resistivity (p) were measured by conventional d.c. potentiometric techniques on crack-free samples of CdSnAsa prepared by the three methods described above. All samples were n-type. Data for resistivity, carrier concentration (l/&e), and Hall mobility (&/p) at room temperature and liquid nitrogen temperature are given in Table 1, together with values for the thermoelectric power at room temperature. Fig. 3 shows the temperature dependence of p and RH for sample A of Table 1 between 85°K and 350°K. All the samples measured were extrinsic over the temperature range investigated. It can therefore be concluded that the intrinsic carrier concentration in CdSnAsa at 300°K is less than 5 x 1017 cm-s

Table 1. Properties of n-type CdSnAsa i Sample

A 2 1A 4A 4B 4c SA

Method of preparation Bridgman Quenched Slowcooled cooled SlowSlowcooled Slowcooled Slowcooled

300°K

_P

(ohm-cm)

1z

--

(cme3)

77°K

_Pm

[&g-l)

(cmsV-lsec-l)

I

72

(cme3)

Pn

(cmsV-lsec-l)

_-

2.7 x 101* 2.5 x 101s 2.9 x lo13

.5*6x 103 2.4 x 10” 6.1 x lo3

-54 -51

, 9.3 x 10-4 5.5 x 1017

1.2~104

-111

4.2 X 10v4 1.1 x 10e3 3.6 x 1O-4

P

(ohm-cm) 3.7 x 10-4

2.8 x 101*

6.1 x 103

3.4 x 10-4

2.8 x lo’*

6.6 x 103

8.2 x 1O-4

5.5 x 1017

1.4x

104

5.4 x 10-4

1.1 x 101s

1-l x 104

-83

1.1 x 10-3

5.8 x 1017

1~0x104

-119

9.5 x 10-4

5.4x

10'7

1.2~104

8.9 x 1O-4

2.8 x 101s

2.5 x 103

-71

8.3 x 10-4

3.0 x ‘IO13

2.5 x 10s

1 -

PREPARATION

AND

PROPERTIES

and that the lattice mobility of electrons at the same temperature is at least 12,000 ems/volt sec. The mobilities of the various samples, with one or perhaps two exceptions, exhibit the usual variation with impurity concentration and crystal perfection. The highest mobilities were observed

Cd SnAsa

I

3-3

---I

OF

281

CdSnAsa

crystal does not have a higher mobility than the Bridgman sample, in view of the great difference in their grain sizes. INFRARED

TRANSMISSION

Infrared transmission data detween 2 and 15 p for samples 4C and 5A of Table 1 are shown in Fig. 4. The measurements were made at room temperature with a Perkin-Elmer Model 221 double beam spectrophotometer. Both curves are

I

I

I

I

I

I

1

i

WAVELENGTH

i/d

FIG. 4. Infrared transmission of n-type CdSnAss at room temperature. Sample thickness: 4C, 200 /J; 5A, 80 B.

--+--103/T”K FIG. 3. Temperature dependence of electrical resistivity and Hall coefficient for n-type Cd&Ass.

in the three samples of lowest carrier concentration, all of which were single crystals prepared by slow cooling. Among the other samples, all of which have carrier concentrations of approximately 3 x 1018 cm-s, the polycrystalline quenched sample and one slow-cooled single crystal have mobilities of about 2500 ems/volt sec. Although the quenched and Bridgman samples have approximately the same grain size, the quenched sample might be expected to contain a higher density of imperfections, and its lower mobility is therefore reasonable. On the other hand, the fact that one of the slow-cooled crystals has the same mobility as the quenched sample is anomalous. In addition, it is rather surprising that the other slow-cooled

seen to exhibit typical semiconductor features: an opaque region at the shortest wavelengths which results from interband absorption, a relatively sharp absorption edge, and a partially transparent region beyond the edge in which free carrier absorption is predominant. The gradual decrease in transmission observed on the long wavelength side results from the usual increase in free carrier absorption with increasing wavelength. Since sample 4C contains only one-fifth the free electron concentration of sample 5A, its transmission beyond the edge is much greater than for sample 5A, although it is more than twice as thick as sample 5A. In fact, it seems probable that between 6 TVand 7 p the transmission for sample 4C is determined mostly by reflection losses, with absorption playing only a minor role. The absorption edge for sample 5A is located at shorter wavelengths than the edge for sample 4C. Similar shifts of the absorption edge to higher energies with increasing electron concentration have been observed for n-type I&b(s) and InAs.(4) These shifts are attributed to the low density of states in the conduction bands of these compounds. (7~4)When the free electron concentration is sufficient to fill an appreciable fraction of the

282

A.

J.

and A. J.

STRAUSS

lower states in the conduction band, photons of higher energy are required to promote electrons from the valence band to higher unoccupied states. Therefore the shift of absorption edge in CdSnAss may be taken as evidence for a low electron effective mass in this compound. From the absorption edge for sample 4C, the energy gap for CdSnAss is estimated to be approximately 0.23 eV; the data for InSb suggest that further decreases in free electron concentration will not shift the edge much further. BAND

STRUCTURE

Although the present experimental data are too limited to permit a detailed discussion of the band structure of Cd&Ass, it is clear from the infrared

ROSENBERG

It was noted previously that Cd&Ass resembles InSb and InAs in exhibiting a shift of its optical absorption edge to higher energy with increasing electron concentration. These three compounds, together with HgSe, also exhibit marked similarities in their electrical and thermoelectric properties, as shown in Table 2. These similarities indicate that the conduction band of Cd&Ass, like those of the other compounds, is characterized by a low electron effective mass of the order of a few hundredths of the free electron mass. They also suggest that the conduction band of Cd&Ass may have the non-parabolic form described by WE for InSb,(ll) which probably is also characteristic of HgSe, as HARMAN has recently suggested.

Table 2. Properties of h&h-mobility compound semiconductors (300°K) CdSnAss

--

(crtF3-3)

5.5 x 101’ 1 *l x 1018 2.7 x 10’8

(a) (b) (c) (d) (e) (f)

(pVi&)

( cmaV%sec-l)

-111 -83 -54

1.2~104 1.1x104 5.6 x 103

InSb

InAs

aa (PVdeg-I)

@ (cmsV-lsec-l)

UC (PVdeg-l)

-106 -82 -51

2.5 x 104 1.9x 104 1.3x104

-140 -110 -7s

HgSe @e

-

( cm2V-lsec-1)

u* (PVdeg-1)

p* (cmsV-lsec-1)

1.3x104 1~0x104 0.9 x 103

-91 -70 -49

1.7x104 1.2~104 7.0 x 103

BARRIE R. and EDMONDJ. T., J. Electronics 1, 161 (1955). H-AN T. C. and STRAU.~~ A. J. (unpublished). Estimated from data of WEISS H., 2. Nututf. lla, 131 (1956). HARMANT. C., GOERING H. L. and BEER A. C., Phys. Rev. 104, 1562 (1956). SCHILLMANE., 2. Naturf. lla, 463 (1956). HARMANT. C. (private communication).

transmission measurements that at room temperature the optical energy gap (N 0.23 eV) is significantly less than the gap for InAs (O-35 eV),(s) the isoelectronic III-V compound. GOODMAN@) found that the II-IV-Vs chalcopyrites which he investigated also have lower energy gaps than the related III-V compounds. As in numerous other cases, the reduction in energy gap is correlated with a reduction in cohesive energy, as shown by the decrease in melting point from 943°C for InAs to 590-6OO”C for Cd&Ass. These effects are presumably associated with the reduction in the average electronegativity difference of the atoms from 0.3 for InAs to 0.25 for CdSnAss, on PAULING’S scale. (10)

Acknowledgements-The authors are indebted to Mr. A. A. MENNA and Mrs. L. B. FARRELL for assistance in the experimental work.

RBFBRBNCBS 1. FOLBERTH 0. G. and PFISTER H., Hulbleiter und Phosphore, p. 474. Vieweg, Braunschweig (1957). 2. GOODMANC. H. L., Nature, Lond. 179,838 (1957). 3. TANENXIAUMM. and BRIGC~ H. B., Phys. Rev. 91, 1561 (1953). 4. TALLEY R. M. and STERN F., J. EZectronics1, 186 (1955). 5. HANSEN M., Constitutionof Binary Alloys (2nd Ed.) p. 438. McGraw-Hill, New York (1958). 6. STRAUSSA. J., Semiconductor Symposium of the Electrochemical Society, October, 1959 (nnpublished).

PREPARATION

AND

PROPERTIES

7. BURSTJZIN E., Phys. Rev. 93,632 (1954). 8. DIXON J. R., Bull. Amer. Phys. Sot. (ser. 2) 4, 133 (1959). 9. VAN DERB~~MGAARLIJ. and SCHOL K., Philips Res. Rep. 12, 127 (1957). 10. PAULING L., The Nature of the Chemical Bond

OF

CdSnAss

283

(3rd Ed.) p. 93. Cornell University Press, Itham, New York (1960). 11. KANE E. O., J. Phys. Chem. Solids 1, 248 (1957). 12. HARMMTT. C., Bull. Amer. Phys. Sot. (ser. 2) 5,152 (1960).