- Email: [email protected]
Luminescence of CuInS2
Journal of Luminescence 27 (1982) 35—53 North-Holland Publishing Company
35
LUMINESCENCE OF CuInS2 I. THE BROAD BAND EMISSION AND ITS DEPENDENCE ON THE DEFECT CHEMISTRY
J.J.M. BINSMA, L.J. GILING and J. BLOEM RIM Laboratory of Solid State Chemistry, Catholic University, Toernooiveld, 6525 ED Nijmegen, The Netherlands
Received 21 December 1981
The emission of CuInS2 is studied as a function of the exact composition in terms of deviations from molecularity and stoichiometry. By means of temperature-dependent (4.2—220 K) and excitation-intensity-dependent measurements, the two broad emission bands usually present are correlated with donor—acceptor transitions. In In-rich material the acceptor is located at 0.10 eV above the valence band. This acceptor is ascribed to V~. In Cu-rich CuInS2 the acceptor level is located at 0.15 eV above the valence band, which is attributed to either Vh, or Cu1,,. Two donor levels are identified, their ionization energies being about 35 and 72 meV. These donors are present in both In and Cu-rich samples and may originate from either intrinsic defects or impurities (Fe). Also, the influence of quenching of the crystals as well as the effect of different crystal growth methods (melt-growth and vapour-growth) on the emission is studied.
1. Introduction CuInS2 is one of the I—Ill—V!2 chalcopyrite compounds, the ternary analogues of the Il—VI zincblende compounds. The optical and electrical properties of CuInS2 crystals and thin films have been investigated for fundamental reasons as well as in view of practical applications, e.g. in solar cells [1—6].A prerequisite for its practical application(s) is a thorough understanding of the defect chemistry of the material. For CuInS2, like for the other chalcopyrites, the information about the defect chemistry is scarce. In order to get more insight into the defect chemistry of the compound, we have studied the photoluminescence as a function of deviations from molecularity and stoichiometry, defined in terms of the ratios [Cu]/[In] and [S]/[Cu] according to ref. [7]. CuInS2 is particularly attractive for such a defect-chemical study, because it can be made both n- and p-type conductive [2], and, having a band gap of about 1.5 eV [1], it might be very well suited for application in solar cells [8]. 0022-231 3/82/0000—0000/$02.75
©
1982 North-Holland
36
J.J.M. Binsma et a!.
/
Luminescence of CuInS
2. I
The photoluminescence of CuInS2 crystals has been investigated by Tell and Shay [1]. They found a series of five sharp emission lines in the region 1.50—1.54 eV (“exciton emission”) and two broad emission bands at 1.44 and 1.40 eV. From their experiments they concluded that CuInS2 possesses a direct band gap (1.5 eV at 2K). Verheijen [3], was the first to correlate the observed broad band emission characteristics of the material to its conductivity type (n or p) and to its deviation from molecularity. In In2S3-rich CuInS2 (obtained only as p type) broad band emission was observed at 1.44 eV. Cu2S-rich CuInS2, (always obtained as n type) showed two broad band emissions at 1.40 and 1.37 eV. Recently two papers appeared about the cathodoluminescence of CuInS2 crystals. Lahlou and Masse [9] studied the 1.44 eV emission by time-resolved spectroscopy and concluded that the emission originates from donor to acceptor transitions. Masse et al. [10] reported subsequently about the influence of annealing (in In, S, In + S or vacuum) on the emission. These annealing procedures, however, do not correspond with equilibrium conditions. In addition, their defect-chemical interpretation does not take into account cation disorder and the role of impurities. From these consideratons it is clear that the defect chemistry of CuInS2 needs a more detailed study. In this paper we will first discuss the broad band emission of CuInS2 as a function of preparation conditions. Then the defect-chemical implications will be discussed, the validity of which may extend, in our opinion, not only to CuInS2 but to other chalcopyrites as well. The near-band-edge emission (exciton emission) will be dealt with in Part II [11].
2. Experimental Nominally pure polycrystalline CuInS2 was prepared from a melt of the pure (~SN) elements in sealed evacuated (2 X l0~ Torr) silica tubes. In the same way samples with small deviations from molecularity were prepared. X-ray diffraction revealed that the as-grown material was of single phase with chalcopyrite structure. CuInS2 single crystals were grown by chemical vapour transport (CVT) method using iodine as the transporting agent according to the time-varying temperature profile method described in3refs. The and [12,13]. a maximum prepared crystals had typical dimensions of 1 X 2 X 10 mm size of 2 X 4 X 20 mm3. Fixed deviations from molecularity (ax) and from stoichiometry (z~y)could be obtained by equilibrating the as-grown CuInS 2 samples for 60 h (no change in the properties was observed for longer annealing times) at 750°Cin the presence of CuInS2 powder and two other components of the Cu—In—S system. For the points a—d indicated in fig. 1 these components are (a) Cu2S and 5, (b) CuIn5S8 and S, (c) CuIn5S8 and In and (d) Cu2S and In saturated with Cu. The sulfur-rich samples (a, b) were p-type
J.J.M Binsma et a!.
/
Luminescence of CuInS
2. I
37
CuIn5S8
______________
/Cu_ln
In
Fig. I. Schematic isothermal portion of the Cu—In—S phase diagram around the homogeneity region of CuInS2. The exact composition is fixed by the deviation from molecularity (~x) and the deviation from stoichiometry (Ay). Points a—d correspond to three-phase equilbria. The Cu—In—S diagram is given as an insert.
conductive, whereas the metal-rich samples (c, d) were of n type, as determined by the hot probe analysis. CuInS2 was found to be not in equilibrium with Cu2S and Cu at the annealing temperature [14]. The polycrystalline (meltgrown) samples were cleaved or cut from larger boules. The luminescence efficiency of samples with cleaved surfaces was a few orders of magnitude higher than that of cut samples. After etching in 1: 1 HC1/HNO3 an increase of the efficiency of the latter samples, to almost the value for the cleaved samples, could be obtained. The single crystals prepared by chemical vapour transport were first washed in an aqueous K! solution, CS2 and acetone in order to remove iodine, iodides and sulfur present on the crystal surfaces (see ref. [13]). The photoluminescence measurements were performed as a function of temperature (4.2 to 220 K), and excitation power. An argon laser (Spectra Physics 164-06) was used as an excitation source at a wavelength of 514.5 nm. In order to achieve excitation powers of 20 mW and less, neutral density filters were used. The maximal excitation power used was 200 mW; at higher values local heating of the sample was observed. The exciting beam was passed through a laser monochromator (Anaspec 300S) to eliminate argon-gas discharge lines. The sample was mounted in a continuous flow cryostat (Oxford Instruments), which was cooled by means of liquid helium. The emission spectra were analysed by means of a 0.5 m monochromator (Monospek 600) used at 1 meV resolution for most experiments and a photomultiplier tube (EM! 9684) provided with an SI cathode and cooled to 70°C.The spectra were corrected for the photomultiplier sensitivity by means of a computer program. —
38
J.J.M. Binsma et a!.
/
Luminescence of CuInS
2. I
3. Results and discussion 3.1. General observations In fig. 2, typical emission spectra are shown for In-rich (points b, c in fig. 1) and Cu-rich (points a, d in fig. 1) CuInS2. For both types of CuInS2, broad emission bands in the region 1.35—1.45 eV together with narrow lines in the region 1.50—1.55 eV near the band gap were observed. The location on the energy scale of the narrow lines, among which are several exciton emission lines, is the same for In-rich and Cu-rich CuInS2. The constancy of the photon energy of these near-edge emission lines indicates that the band gap of the material is not affected by differences in composition. The narrow line emission will be treated in detail in Part II. The broad band emission, 0.10 to 0.20 eV below the band gap, is made possible by defects which introduce localized levels within the forbidden band. When going from In-rich to Cu-rich CuInS2 a decrease of 50 meV in the energy of the maxima of the broad band emission is observed (fig. 2). Because of the invariance of the band gap, the changes in the peak energies of the broad band emission should be due to changes in the nature and concentration of the defect levels themselves. In table 1, the peak energies of the broad band emission of samples prepared in different ways are summarized. In all cases, except for quenched CuInS2, two broad emission bands are present in Cu-rich samples at about 1.39 and 1.36 eV and in In-rich samples at about 1.44 and 1.41 eV. These energy values have to be interpreted as characteristic values; for different samples there is a variation of up to about ± 10 meV, depending on the composition, temperature and laser intensity (see subsects. —
—
—
4.2K Cu-rich
/
1.35
\n_rich
—
Ji 1.30
/
~
1.40
1.45
hv (eV)
1.50
155
Fig. 2. Emission of Cu-rich (dashed line) and In-rich (solid line) CuInS2 at 4.2 K and constant excitation intensity (50 mW).
J.J.M. Binsma et a!.
/
Luminescence of CuInS
2. I
39
3.2 and 3.3). Independent of the type of sample, temperature and laser intensity, the separation between the high- and low-energy bands was found to be of constant value 37 ± 2 meV for both Cu- and In-rich samples. In fig. 3, the emission of In-rich CuInS2 is shown at different temperatures. The peaks shift to higher energy, leaving the separation between high- and low-energy emission unchanged. The intensity ratio of the high- and low-energy emissions remains nearly constant, although this is somewhat obscured by the broadening of the emissions. From table 1 it follows that there is no difference in peak energy between samples prepared by annealing in an atmosphere containing one of the constituent metals and as-grown samples prepared from a melt with excess In or Cu. Another important observation is that the type of emission is not influenced by the electronic conductivity type, which is determined by the deviation from stoichiometry (z~yin fig. 1) [2—4,14].Cu-rich samples annealed under minimal sulfur presence (~y<0, point d in fig. I and n-type conductive) show emission at the same photon energy as Cu-rich samples annealed under maximal sulfur pressure (z~y>0, point a in fig. 1 and p-type conductive). This similarity is also observed for In-rich samples (points b and c in fig. 1). From these observations an important conclusion can be drawn, that the type of luminescence mainly depends on the deviation from molecularity, where the type of conductivity is fixed by the deviation from stoichiometry. As-grown samples, which were not intentionally grown with deviations from the ideal composition, showed either the behaviour of Cu-rich samples or that of In-rich samples. This was also observed for CVT-grown CuInS2 crystals. Typical broad band emission spectra of CVT-grown CuInS2 (both as grown and annealed) are shown in fig. 4. After annealing, the characteristic emissions of Cu-rich or In-rich CuInS2 are obtained, just as for melt-grown samples. The intensities of
Table I Peak energies for broad band emission of as-grown and annealed CuInS2 at 4.2 K under constant excitation intensity Type of sample
E~(eV)b)
As grown, n.p. ~) (melt, CYT) Cu103In099S2 (melt) Annealed (Cu2S/S or Cu2S/Cu—In) Cu097In101S2 (melt) Annealed (CuIn5S8/S or CuIn5S8/In) Quenched from melt
1.44, 1.41 or 1.39, 1.36 1.39, 1.36 1.39, 1.36 1.44, 1.41 1.44, 1.41 1.44, 1.41, 1.39, 1.36
b)
n.p. = nominally pure. The accuracy of all values is ±0.01eV.
J.J.M. Binsma et a!.
40
/
Luminescence of CuJnS
2. I
• E~i~1
1.30
1.35
1.40
1.45
hv (eV)
1.50
1.55
Fig. 3. Broad band emission of In-rich CuInS2 as a function of temperature at constant excitation intensity (50 mW). (1) 4.2 K, (2) 80 K and (3) 120 K.
the low- and high-energy emission bands are almost equal for as-grown
CVT-crystals. Samples, which were quenched from the melting point, 1090°C, [15] to room temperature by pulling the ampoule out of the furnace, exhibited all four broad emission bands, both those of Cu-rich and In-rich samples. After annealing, these samples showed only the emission bands belonging to the type of annealing atmosphere. This experiment strongly indicates that the defects involved in the emission processes are in equilibrium with the gas phase.
Cu anneal
In anneal
~
1.30
I
1.35
1.40
1.45
hv (eV)
1.50
155
Fig. 4. Broad band emission of CVT-grown CuInS2: as grown (I), Cu-annealed (2) and In-annealed (3), at 4.2 K and constant excitation intensity (50 mW).
J.J.M. Binsma et a!.
/
Luminescence of CuInS
2. I
41
Before going into the details of the defect chemistry (see sect. 4), it is necessary to discuss the temperature and excitation intensity dependence of the broad emission bands. 3.2. Peak energy as a function of temperature As already shown in fig. 3, all broad emission peaks shift to higher energies with increasing temperature. In fig. 5 the shift in the peak energy i.~E = E~(T) E~(4.2) of the high-energy emission (1.44 eV) is plotted for three In-rich CuInS2 samples, viz. (I) annealed with Cu1n558 and 5, (2) as grown and (3) annealed with CuIn5S8 and In. Also the temperature dependence of the energy gap E5 as determined from measurements of the exciton emission (see part II) is shown in this figure. The separation between the high- and the low-energy emissions remains constant, viz. 37 ±2 meV, independent of the temperature. The characteristic variation of the peak energies for any sample can thus be represented by that of the high-energy emission (Er) alone. It is clear that the peak energies are increasing more rapidly than Eg with increasing T in the range 4.2—80 K. Above 80 K when Eg decreases, still an increase in is found. For most samples, 8Ev/aT was about 4 X l0~ eV K but also larger values were observed, occasionally up to a maximum of 17 X I0~ eV K The additional shift of E1, with respect to Eg is an indication of donor— —
~,
/
/ ~
(meV)
/
2
zv~ —
GO
r-~!.0
I 50
I 150
100 T(K)
Fig. 5. Shift in the peak energy ~E = E~(T)— E~(4.2)as a function of temperature for three In-rich CuInS2 4.2)samples as reported (1.44 in eVPart emission). II. (1) The annealed dashed with lineCuIn represents the shift in the energy gap Eg(T)_ Eg( 5S8 and S, (2) as grown and (3) annealed with CuIn5S8 and In.
42
J.J.M. Binsma et a!.
/
Luminescence of CuInS
2. I
acceptor (DA) emission [16]. The emission energy for a donor—acceptor pair at distance r is, in the first approximation, 2/r. (1) hv=Eg—EA—ED+e
When the temperature is increased, the migration of mobile charge carriers to more favourable sites (smaller r) is enhanced, thus giving a shift in hv additional to the shift in E 5. Because the additional shift depends on the degree of compensation [17], different samples will show different temperature coefficients as is illustrated by fig. 5 for In-rich samples. Also Cu-rich samples show this behaviour. On the contrary, for free to bound emissions, which involve the valence band or the conduction band, the same temperature coefficient is expected for all samples, as follows from the corresponding peak energy in that case [16]: Ep=Eg~Ei+kT,
(2)
where E1 stands for the (temperature independent) ionization energy of the centre (either donor or acceptor) which captures the free carrier. The temperature coefficient for all samples is given in the latter case by aE aE p— 3 8T~T which is not observed. So it can be concluded that the broad band emission of both In-rich and Cu-rich CuInS2 is of the donor to acceptor type. At high temperatures (> 100 K), however, an increasing contribution of a conduction band to acceptor emission is found at the high-energy edge (see the forthcoming subsections). This shift in the character of the emission may be partly responsible for the increase in E~at high temperatures. 3.3. Peak energy as a function of excitation intensity The peak energy of the various observed broad band emissions appears to be a function of the intensity of the exciting beam, the general trend being an increase with ii~creasingexcitation intensity. This is demonstrated by fig. 6, where the shift in E~for the high-energy emission (1.41 eV) is plotted for the series of In-rich CuInS.2 samples already presented in subsect. 3.2 over three decades of excitation power P1 at 4.2 K. As already mentioned in subsect. 3.1, the separation between the high- and low-energy emissions is unaffected by the excitation intensity and remains constant at 37 meV. E~,as a function of P1 is behaving according to the relation which usually applies to the DA emission in compensated semiconductors [17]: E~=E~+f3(logP~ —logP10),
(4)
with E1, as the peak energy at P1 and /3 the energy shift per decade shift in P1.
/
J.J.M. Binsma et a!.
30.0
Luminescence of CuInS
2. I
43
-
AE
1,,,
(meV) 2G0
100
0.0
/
-
/
-
0.0
A
/~
~ I
1.0
3.o
2.0 Iog~—
Fig. 6. Shift in the peak energy ~E as a function of the relative excitation intensity at 4.2 K for the 1.44 eV emission of three In-rich CuInS2 samples (numbering of samples as in fig. 5). ~ E = E~(P1) —
E~(P1).
The parameter /3 increases with increasing degree of compensation and decreases with increasing acceptor and donor concentrations [181.Indeed various values for /9 were found at low temperatures (4.2 K) ranging from 0 to 25 meV for In-rich CuInS2 (peaks at 1.44 and 1.41 eV). For Cu-rich CuInS2 (peaks at 1.39 and 1.36 eV) /9 ranges from 0 to 15 meV at T 4.2 K. The increase of E~with increasing P1 can be explained for small values only a few meV of /3 on the basis of a decrease of the mean r value in eq. (1) at higher excitation levels [19]. Another mechanism which can also explain larger values > 10 meV of /3 is band perturbation by high concentrations of defects [17]. Because high concentrations of, probably, intrinsic defects can be present in CuInS2 [2,3,14] this explanation may hold for sample I. The parameter /3 appears to be a function of temperature: it decreases rapidly with increasing temperature and becomes ~ I meV at 80 K. At higher temperatures the migration of charge carriers to more favourable sites is enhanced (see sect. 3.2) thus giving rise to a smaller possible influence of the excitation intensity. This accounts for the small /9 values at higher temperatures. For free to bound emission also an increase in E~with increasing P~should be observed [17]. In contrast to DA emission, however, the shift in this case should be independent of the type of sa.mple, for no substantial broadening of the acceptor levels is to be expected because of the large ionization energy of the trapping center. In addition, the high-energy edge of the emission band —
—
—
—
/
Luminescence of CuInS
1.50
1.55
J.J.M. Binsma et a!.
44
1.35
1.40
1.45
hv (eV)
2. I
Fig. 7. Broad band emission of In-rich CuInS2, at 4.2 K, at different excitation intensities, showing the structure at the high-energy edge for high excitation.
should be abrupt corresponding to the quasi Fermi level of the free carrier band [17]. The high-energy edge of the emission of both Cu-rich CuInS2 is not very abrupt, however, and becomes even less sharp at increasing P1. At high temperature (T> 100 K), free to bound emission may contribute to the high-energy emission, as was also observed in ref. [9]. This makes the interpretation of the spectra difficult for T> 100 K (see sect. 3.4). At the highest excitation intensities and low temperature (4.2 K), some structure is observed in the high-enegy edge (fig. 7). This structure may correspond to donor—acceptor emission lines for relatively small separations. Weak DA lines are commonly observed at the high-energy edge of DA emission bands with strong phonon coupling (bell shaped bands) [16]. The low-energy emission shifts to higher energy with increasing P1 in the same way as the high-energy emission, leaving the difference (37 + 2 meV) between both peak energies unchanged. This indicates that the concentrations of the donors and acceptors involved in both transitions are not very different. 3.4. Emission intensity as a function of temperature In order to obtain additional information about the centers involved in the emission processes in CuInS2, the intensity of the emission was studied as a function of temperature. When trapped electrons or holes are playing a role in the recombination, quenching of the emission will occur at high temperatures by ionization of the trapping center [16]. The thermal activation energy for quenching can be related then to the ionization energy of the trapping center. In the case of two trapping centers, thus for DA emission, the interpretation of
J.J.M. Binsma et a!.
/
Luminescence of CuInS
2. I
45
the activation energy is not quite unambiguous, but may still give useful information. In fig. 8 the intensity (approximated by peak height times peak width) of the (high energy) 1.44 eV emission of an In-rich sample and the intensity of the (high energy) 1.39 eV emission of a Cu-rich sample are plotted as a function of temperature. The observed behaviour of the intensity I as a function of T is in accordance with a thermally activated probability for non-radiative recombination as described by Curie [20]: I(T)= ‘~°~ (5) 1 + Cexp[—LIE/kT] with LIE as activation energy for non-radiative recombination and C being a measure for the capture cross section of the center which is thermally emptied. From the curves plotted in fig. 8, LIE is calculated to be 90 meV for In-rich material (1.44 eV emission) and 145 meV for Cu-rich material (1.39 eV emission). For DA emission LIE should be equal to the ionization energy of the shallower level at low temperature, and to the sum of the ionization energies, EA + ED, at higher temperatures. The values of LIE which are calculated here are too high to correspond to the shallower and of are0.11 too eV lowfor to In-rich be equaland to 2/erlevel a value EA 0.16 + ED.eVEq. yields for EA + ED e of for(1)Cu-rich samples. For e2/er a value of about 0.03 eV can be assumed [9], which gives for EA + ED 0.14 and 0.19 eV, respectively. On the other hand, the LIE values are in good agreement with the acceptor levels found in p-type CuInS 2 by electrical measurements [14]. In p-type In-rich CuInS2 a level of 100 meV was observed whereas p-type Cu-rich material possesses an acceptor with EA = 150 meV. The 150 meV acceptor was also —
~E—
1(K)
100 GO
Cu-rich
~
~E.145eV
—2.0 .
/
50
~—~°
/
S
~
~
/
In—rich /AE~.O90eV
/1 -40
10
20
io~
30
T (K~)
—.
250
Fig. 8. Relative intensity of the broad band emission (1.44 eV) of In-rich CuInS2 (x) and the broad band emission (1.39 eV) of Cu-rich CuInS2 (9) as a function of temperature at constant excitation intensity (50 mW).
46
J.J.M. Binsma et a!.
/
Luminescence of CuInS
2. I
reported by Look and Manthuruthil [2], and Verheijen [3]. They performed Hall measurements on CuInS2 annealed only in the presence of sulfur. Because of the high vapour pressure of the M111-sulfides above M1 M111 X2 chalcopyrites compared to the pressure of the M1-sulfides [21], material annealed only in the presence of sulfur will tend to become M1-rich. So it is clear, that in sulfur-annealed CuInS2 only the 150 meV acceptor belonging to the Cu-rich material has been observed. These considerations lead us to the conclusion that different acceptors with ionization energies of 100 and 150 meV, respectively are involved in the DA emissions of In-rich and Cu-rich materials. An activation energy of 85 meV for quenching was found by Lahlou and Masse [9] for In-rich CuInS2. They also did not observe EA + ED as an activation energy. This was explained by assuming a large contribution of free to bound (here, conduction band to acceptor) emission, which might be possible at high temperature (T> 100 K). The intensities of the emissions at lower energies (1.41 and 1.36 eV) decrease simultaneously with the high-energy emission when the temperature is increased. This leads to an almost temperature-independent ratio of the intensities of the high- and low-energy emissions up to 100 K (see fig. 3). At higher temperatures an accurate determination of the intensity of the low-energy emission becomes difficult because of the broadening of the emission bands. So the activation energy for quenching could not be determined and no indication about the ionization energies of the donors and acceptors involved in the emissions can be obtained. The observed weak temperature dependence of the intensity ratio together with the varying intensity of the low-energy emission for different samples provide the arguments for the interpretation as an additional DA emission rather than in terms of a phonon wing (see also subsect. 3.6). As is already argued above, the shift of the entire broad band emission over 50 meV to lower energy when going from In-rich to Cu-rich material can be ascribed to a change in the nature of the acceptor involved. From the difference in peak energy between the high- and low-energy emission (37 ± 2 meV, see subsect. 3.1), it can now be deduced that the donor involved in the low-energy emission possesses an ionization energy which is larger than that of the donor belonging to the high-energy emission by the same amount 37 meV. —
—
—
3.5. Emission intensity as a function of excitation intensity
Apart from the peak position as a function of excitation intensity also the emission intensity (see fig. 9 for In-rich CuInS2) has been studied. in the temperature range 4.2 to 150 K over three decades of excitation intensity. The emission intensity I varied as I = CP~with C and x as constants and P1 as the excitation intensity. At 4.2 K, a ranges from 1.20 at lowest excitation intensities to 0.63 at highest excitation intensities. This applied for both Cu-rich and
J.J.M. Binsma et aL
4.0
/ Luminescence of CuInS2. I
47
,74.2K
LogI (au.)
/
//
30
/65K /X
/5
/
/
20
1.0
.
0.0 0.0
7
//
/
/X /128K
(1*10)
/1 /5
I
I
1.0
2.0
I
3.0 Log9 (au.)
Fig. 9. Intensity of the 1.44 eV emission of In-rich CuInS2 at three different temperatures (4.2, 85 and 128 K) as a function of the relative excitation intensity.
In-rich CuInS2. With increasing temperature but constant excitation intensity a increases, the range being 1.5—1.0 at T> 70 K. The temperature and excitation intensity dependence of a observed here is in good agreement with the theory deduced by Maeda [22] for DA transitions (compare figs. 8 and 9). According to this theory, temperature regions with different temperature dependences of I have their characteristic ranges of a values. In the low-temperature region (no temperature dependence of I), the excitation intensity dependence of I should range from linear (low P1) to sublinear (high P1). At high temperatures, where I quenches with an activation energy LIE, the dependence should vary from supra-linear or even quadratic to linear at high P1. This behaviour is reflected in fig. 9. Care has to be taken, however, in the interpretation of the curve at 128 K, because the emission may originate in part from free to bound transitions (see the foregoing subsection). The ratio of the intensities of the high- and low-energy emissions, 11/12, appeared to be nearly independent of the excitation intensity. Only when an appreciable contribution of the deep donor—acceptor transition is present (11/12 <<5), the high-energy emission is slightly favoured with increasing excitation intensity with respect to the low-energy emission (11/12 increases slightly). A saturation of the low-energy emission at high excitation intensities
J.J.M. Binsma et at
48
/
Luminescence of CuInS
2. I
may be responsible for this behaviour, as already noted by Verheijen [3], who reported similar observations for the broad band emission of CuInS2. 3.6. Interpretation in terms of defect levels In the foregoing the (high energy) emissions at 1.44 eV for In-rich and at 1.39 eV for Cu-rich materials were correlated to recombinations between the same donor (ionization energy ED1) and two different acceptors, their ionization energies being 100 and 150 meV. The low-energy emissions (1.41 and 1.36 eV) are attributed to donor—acceptor recombinations from a deeper donor (ionization energy ED + 37 meV) to the acceptor levels which participate also in the high-energy emission. Because the separation between high- and low-energy emissions is close to the optical phonon modes of CuInS2, which are lying in the range 35—40 meV [23], the possibility of a phonon-assisted transition should be considered too. Two arguments can be given against this interpretation, however, viz. (i) the strongly varying intensity ratio between the high- and low-energy emissions for different samples, ranging from 0.7 to 5, and (ii) the weak temperature dependence of the intensity ratio (see subsect. 3.4 and fig. 3), which decreases at most 10% from 4.2 to 100 K. The intensity of a phonon-assisted transition involving emission of a phonon to the lattice should quench with respect to the zero-phonon transition according to the factor exp[hw/kT], h~being the phonon energy [24]. The only unknown ionization energy so far is that of the shallower donor, ED. There are strong indications, however, that the correct value of ED is about 35 meV. In the first place, we found by electrical measurements on n-type CuInS2 a donor level of 35 meV [14]. Further, a donor to valence band transition is observed at 1.52 eV which also yields a donor-ionization energy of 35 meV (see Part II). Finally, fitting results of time-resolved measurements by Lahlou and Masse reveal a level of 45 meV [9]. The donor and acceptor levels present in Cu-rich and In-rich CuInS2 are summarized in table 2 together with the peak energies of the donor—acceptor Table 2 Peak energies and donor and acceptor levels for Cu-rich and In-rich CuInS2 (accuracy for E0 is ±5 meV and for EA, ±10meV) Molecularity
E~(eV)a)
ED(meV)
EA(meV)
Cu-rich
1.39 1.36
35 72
150 150
In-rich
1.44 1.41
35 72
100 100
a)
Measured at 4.2 K.
J.J.M. Binsma et a!.
cond. band ~\\~\\\\\
—
/
Luminescence of CuInS
2. I
49
cond. band ~\\
~1\1
3~i.T
T~\j.:oev
f~v
I.5nej
~,,
,,,,,,,,,I~,,
vat, band
val. band
Cu-rich CuInS 2
(a)
In-rich CuInS2 (b)
Fig. 10. Energy level diagrams for (a) Cu-rich CuInS2 and (b) In-rich CuInS2, showing the various radiative transitions and the donor and acceptor levels involved. Also the donor to valence band transition at 1.52 eV, which will be reported in Part II, has been indicated. Energy balance is obtained for the DA emissions when the Coulomb energy ( 0.03 eV) is included. A1 is assigned to either V1,, or Cu1,, and A2 to Vt,,. The nature of D1 and D2 is not known up to now (see Sect. 4).
emissions originating from these levels. In fig. 10, the emission processes involving the donor and acceptor levels are schematically drawn for Cu-rich (a) and In-rich (b) CuInS2. Also the donor to valence band transition found near the band edge [11] has been indicated. When energy balances are2/r constructed for DA for fig. 10, one should take into account the Coulomb energy e emission (see eq. (1)), which corresponds to about 0.03 eV at the emission peak [9]. The nature of the various defects corresponding to the donor and acceptor levels and their dependence on molecularity and stoichiometry will be discussed in the following section.
4. Defect-chemical considerations The experiments on samples which were quenched from the melting point and those on samples which were equilibrated under various conditions strongly indicate that the luminescence properties are governed by intrinsic defects. In a ternary compound, a wide variety of intrinsic defect equilibria is present. These equilibria can be classified into internal equilibria involving only defects in the solid and interphase equilibria between solid and gas phase [7,25]. Compared to binary compounds, additional mechanisms for defect formation originate from deviations from the ideal molecularity and from cation-disorder. In terms of defect-equilibria these mechanisms can be formulated as follows: ~
J.J.M. Binsma et a!.
50
—
—I
a ~Pcu2sL
rv*Cul12F~j*1_L I S J
/ Luminescence of CuInS2. I
‘
Cu~~ + In~~ Cu’r~+ In~~ Kh=[Cu~fl][In~U].
(7)
The equilibrium given by (6) may also be formulated in terms of the indium sulfide pressure and V~.Because the copper and indium sulfide pressures are coupled via the formation reaction, both representations are equivalent. In the equilibria (6) and (7) the point defects have been represented in their neutral state. Charge carriers are not involved in these equilibria, by which comparable amounts of donors and acceptors are formed. This in contrast to the equilibrium which describes deviations from stoichiometry: 2S~v2V~+S2(g) 2p K~ [V~] 5.
(8)
The formation of n-type material as a result of sulfur deficiency is directly given by (8), because V~’acts as a donor. When the equilibrium is formulated in terms of sulfur interstitials (which are acceptors) it will be clear that a sulfur-excess gives rise to p-type material. Here charge carriers are thus indirectly formed. The occurrence of disorder corresponding to (6) and (7) may explain why for the chalcopyrites usually high compensation is found in contrast to the binary compounds, e.g. Il—VI compounds [25]. The attribution of specific defects to the observed levels (see table 2) will require further measurements (e.g. of the electrical properties), but some conclusions can already be drawn. As the nature of the acceptor level changes with annealing or quenching it may be attributed to an intrinsic majority defect. Table 3 gives a survey of the various defects which can act as acceptors in CuInS2 together with their conditions for the deviation from molecularity (LIx>0: Cu-rich; LIx<0: In-rich). These conditions can be deduced on the basis of the relations between exact composition and defect contents given in ref. [7]. For In-rich CuInS2 the acceptor at 0.10 eV may correspond to V1~ when we exclude the energetically unfavourable 5,. The acceptor at 0.15 eV is Table 3 Possible intrinsic majority defects acting as acceptors and the conditions for their occurrence in terms of the deviation from molecularity, ~x = [Cu]/[Inl — Defect Vc~
<0
V1,, Si Cu1,,
>0 <,=,>0 >0
J.J.M. Binsma et a!. / Luminescence of CuInS
2. I
51
correlated to Cu-rich CuInS2 as revealed by this work and that of other authors (see subsect. 3.4). Table 3 gives two possible defects for this level, viz. V1,, and Cu ~ when S, is again excluded. It is difficult to discriminate between the remaining possibilities, but that the acceptors are of intrinsic character and not due to impurities is indicated by the fact that an acceptor with an ionization energy of about 0.15 eV is found for various Cu-containing chalcopyrites [26] independent of their band gap. The energy levels of a cation vacancy (surrounded by four anions for the chalcopyrites) are solely related to the valence band as shown by Lang et al. [27]. Changing the type of cation (e.g. in Al~GaI _~As)does not alter the environment of the cation vacancy, resulting in an acceptor level which is fixed relative to the valence band (condition of unchanged environment). Impurity levels on the other hand are related to both the valence band and the conduction band and are thus shifting with the band gap when varying the nature of the cations. For the same reasons the ionization energy of the CuM acceptor (corresponding to a copper atom surrounded by four anions) will also not be affected by the band gap. It is clear from the defect-chemical considerations summarized in table 3 that the attribution of the 0.15 eV acceptor to either V1,, or Cu In is much more plausible than that to Va,, as given in ref. [26]. In addition, V1~and Cu1~fulfil the condition of unchanged environment in the CuM111 X2 compounds even better than Vc~ does, if second nearest neighbours are included. For the donor levels it is not clear up to now whether they correspond to majority or minority defects and a definite attribution is thus not yet possible for these levels. Various defects are possible such as V5, Cu,, In,, Inc~ or foreign elements incorporated in the lattice. Especially the undesired incorporation of iron may be significant as has been demonstrated by EPR measurements for CuInS2 and other chalcopyrite compounds [28—30].Spectrochemical analysis reveals the presence of 3Fefor as both by farmelt-grown the most important impurityCuInS up to and CVT-grown a concentration of 5 X 1018 cm 2 [14]. This concentration is equal to the measured majority defect concentrations, which means that a defect-chemical model of CuInS2 is not complete when the role of iron is ignored. Measurements which should lead to a better understanding of this role are under way. Finally, it may be noticed that the interpretation of the levels at 45 and 90 meV as given by Masse et al. [10] needs to be reconsidered. They ascribed the 45 meV level to the VCU acceptor on the basis of the analogy with some other chalcopyrites. From our optical and electrical experiments it can be deduced that the 90 meV level corresponds to ~ whereas the 45 meV level originates from a donor. The nature of this donor has still to be revealed (see above).
52
J.J.M. Binsma et a!.
/ Luminescence of CuInS2. I
5. Conclusion By studying the broad band emission as a function of the exact composition, intrinsic defects were found to play an important role in the defect-chemistry of CuInS2. On the one hand, deviations from molecularity (Cu-excess or In-excess) determine the type of broad band emission. This shows that the ternary character of the compound is an important factor in fixing the defect-chemical state of the compound. The influence of deviations from stoichiometry, on the other hand, is reflected in the electrical properties, sulfur-rich material being p type and metal-rich being n type. The various types of CuInS2 can be converted into each other by annealing in specific atmospheres corresponding to the possible three-phase equilibria. Finally, it should be mentioned that the melt-grown and CVT-grown crystals showed no fundamental differences in emission.
Acknowledgements The authors would like to thank Mr. H.W.J.M. Hoekstra, who participated in the emission measurements. The investigations were performed as part of the research program of the Stichting Scheikundig Onderzoek in Nederland (SON) with financial support from the Nederlandse Organisatie voor Zuiver Wetenschappelijk Onderzoek (ZWO).
References [I] B. Tell, J.L. Shay and H.M. Kasper. Phys. Rev. B4 (1971) 2463. [2] D.C. Look and J.C. Manthuruthil, J. Phys. Chem. Solids 37 (1976) 173. [3] A.W. Verheijen, Thesis, University of Nijmegen, Nijmegen (1979). [41B. Tell, J.L. Shay and H.M. Kasper, J. Appi. Phys. 43 (1972) 2469. [5] L.L. Kazmerski, M.S. Ayyagari and G.A. Sanborn, J. AppI. Phys. 46 (1975) 4865. [6] H. Neumann, B. Schumann, D. Peters, A. Tempel and G. Kuhn, Kristall und Technik 14 (1979) 379. [7] J.A. Groenink and PH. Janse, Z. Phys. Chem. N.F. 110 (1978) 17. [8] J.M. Meese, J.C. Manthuruthil and D.R. Locker, Bull. Am. Phys. Soc. 20 (1975) 696. [9] N. Lahlou and G. Masse, J. AppI. Phys. 52 (1981) 978. [10] G. Masse, N. Lahlou and C. Butti, J. Phys. Chem. Solids 42 (1981) 449. [Ill J.J.M. Binsma, L.J. Giling and J. Bloem, J. Lumin. 27(1982)55. [12] C. Paorici, L. Zanotti and 0. Zucalli, J. Crystal Growth 43 (1978) 705. [13] JiM. Binsma, W.J.P. van Enckevort and G.W.M. Staarink, J. Crystal Growth, in press. published. [14] J.J.M. Binsma, to be published. [15] J.J.M. Binsma, L.J. Giling and J. Bloem, J. Crystal Growth 50(1980) 429. [16] P.J. Dean, in: Progress in Solid State Chemistry, Vol. 8. Eds. JO. McCaldin and G. Somorjai (Pergamon, New York, 1973) p. 1.
J.J.M. Binsma et a!.
/
Luminescence of CuInS
2. 1
53
[17] J.I. Pankove, Optical Processes in Semiconductors (Dover, New York, 1975). [18] MI. Nathan and T.N. Morgan, in: Physics of Quantum Electronics, Eds. P.L. Kelley, B. Lax and P.R Tannenwald (McGraw-Hill, New York, 1966) p. 478. [19] E. Zacks and A. Halperin, Phys. Rev. B6 (1972) 72. [20] D. Curie, Luminescence Cristalline (Dunod, Paris, 1961). [21] S.S. Strelchenko, S.A. Bondar, A.D. Molodyk, LI. Berger and A.E. Balanevskaya, Izv. Akad. Nauk. SSSR, Neorgan. Mater. 5 (1969) 593. [22] K. Maeda, J. Phys. Chem. Solids 26 (1965) 595. [23] W.H. Koschel and M. Bettini, Phys. Stat. Sol. (b) 72 (1975) 729. [24] H. Barry Webb and E.W. Williams, in: Semiconductors and Semimetals, Vol. 8, Eds. R.K. Williardson and A.C. Beer (Academic Press, New York, 1972) p. 182. [25] F.A. Kroger, Chemistry of Imperfect Crystals (Norh-Holland, Amsterdam, 1973). [26] H. Neumann, D. Peters, B. Schumann and G. Kuhn, Phys. Stat. Sol. (a) 52 (1979) 559. [27] D.V. Lang, R.A. Logan and L.C. Kimerling, Phys. Rev. B 15 (1977) 4874. [28] G. Brandt, A. Rauber and J. Schneider, Solid State Commun, 12 (1973) 481. [29] H.J. von Bardeleben, A. Goltzené, C. Schwab and R.S. Feigelson, AppI. Phys. Lett. 32 (1978) 741.
[30] Hi. von Bardeleben and R.D. Tomlinson, J. Phys. Cl3 (1980) Ll097.
35
LUMINESCENCE OF CuInS2 I. THE BROAD BAND EMISSION AND ITS DEPENDENCE ON THE DEFECT CHEMISTRY
J.J.M. BINSMA, L.J. GILING and J. BLOEM RIM Laboratory of Solid State Chemistry, Catholic University, Toernooiveld, 6525 ED Nijmegen, The Netherlands
Received 21 December 1981
The emission of CuInS2 is studied as a function of the exact composition in terms of deviations from molecularity and stoichiometry. By means of temperature-dependent (4.2—220 K) and excitation-intensity-dependent measurements, the two broad emission bands usually present are correlated with donor—acceptor transitions. In In-rich material the acceptor is located at 0.10 eV above the valence band. This acceptor is ascribed to V~. In Cu-rich CuInS2 the acceptor level is located at 0.15 eV above the valence band, which is attributed to either Vh, or Cu1,,. Two donor levels are identified, their ionization energies being about 35 and 72 meV. These donors are present in both In and Cu-rich samples and may originate from either intrinsic defects or impurities (Fe). Also, the influence of quenching of the crystals as well as the effect of different crystal growth methods (melt-growth and vapour-growth) on the emission is studied.
1. Introduction CuInS2 is one of the I—Ill—V!2 chalcopyrite compounds, the ternary analogues of the Il—VI zincblende compounds. The optical and electrical properties of CuInS2 crystals and thin films have been investigated for fundamental reasons as well as in view of practical applications, e.g. in solar cells [1—6].A prerequisite for its practical application(s) is a thorough understanding of the defect chemistry of the material. For CuInS2, like for the other chalcopyrites, the information about the defect chemistry is scarce. In order to get more insight into the defect chemistry of the compound, we have studied the photoluminescence as a function of deviations from molecularity and stoichiometry, defined in terms of the ratios [Cu]/[In] and [S]/[Cu] according to ref. [7]. CuInS2 is particularly attractive for such a defect-chemical study, because it can be made both n- and p-type conductive [2], and, having a band gap of about 1.5 eV [1], it might be very well suited for application in solar cells [8]. 0022-231 3/82/0000—0000/$02.75
©
1982 North-Holland
36
J.J.M. Binsma et a!.
/
Luminescence of CuInS
2. I
The photoluminescence of CuInS2 crystals has been investigated by Tell and Shay [1]. They found a series of five sharp emission lines in the region 1.50—1.54 eV (“exciton emission”) and two broad emission bands at 1.44 and 1.40 eV. From their experiments they concluded that CuInS2 possesses a direct band gap (1.5 eV at 2K). Verheijen [3], was the first to correlate the observed broad band emission characteristics of the material to its conductivity type (n or p) and to its deviation from molecularity. In In2S3-rich CuInS2 (obtained only as p type) broad band emission was observed at 1.44 eV. Cu2S-rich CuInS2, (always obtained as n type) showed two broad band emissions at 1.40 and 1.37 eV. Recently two papers appeared about the cathodoluminescence of CuInS2 crystals. Lahlou and Masse [9] studied the 1.44 eV emission by time-resolved spectroscopy and concluded that the emission originates from donor to acceptor transitions. Masse et al. [10] reported subsequently about the influence of annealing (in In, S, In + S or vacuum) on the emission. These annealing procedures, however, do not correspond with equilibrium conditions. In addition, their defect-chemical interpretation does not take into account cation disorder and the role of impurities. From these consideratons it is clear that the defect chemistry of CuInS2 needs a more detailed study. In this paper we will first discuss the broad band emission of CuInS2 as a function of preparation conditions. Then the defect-chemical implications will be discussed, the validity of which may extend, in our opinion, not only to CuInS2 but to other chalcopyrites as well. The near-band-edge emission (exciton emission) will be dealt with in Part II [11].
2. Experimental Nominally pure polycrystalline CuInS2 was prepared from a melt of the pure (~SN) elements in sealed evacuated (2 X l0~ Torr) silica tubes. In the same way samples with small deviations from molecularity were prepared. X-ray diffraction revealed that the as-grown material was of single phase with chalcopyrite structure. CuInS2 single crystals were grown by chemical vapour transport (CVT) method using iodine as the transporting agent according to the time-varying temperature profile method described in3refs. The and [12,13]. a maximum prepared crystals had typical dimensions of 1 X 2 X 10 mm size of 2 X 4 X 20 mm3. Fixed deviations from molecularity (ax) and from stoichiometry (z~y)could be obtained by equilibrating the as-grown CuInS 2 samples for 60 h (no change in the properties was observed for longer annealing times) at 750°Cin the presence of CuInS2 powder and two other components of the Cu—In—S system. For the points a—d indicated in fig. 1 these components are (a) Cu2S and 5, (b) CuIn5S8 and S, (c) CuIn5S8 and In and (d) Cu2S and In saturated with Cu. The sulfur-rich samples (a, b) were p-type
J.J.M Binsma et a!.
/
Luminescence of CuInS
2. I
37
CuIn5S8
______________
/Cu_ln
In
Fig. I. Schematic isothermal portion of the Cu—In—S phase diagram around the homogeneity region of CuInS2. The exact composition is fixed by the deviation from molecularity (~x) and the deviation from stoichiometry (Ay). Points a—d correspond to three-phase equilbria. The Cu—In—S diagram is given as an insert.
conductive, whereas the metal-rich samples (c, d) were of n type, as determined by the hot probe analysis. CuInS2 was found to be not in equilibrium with Cu2S and Cu at the annealing temperature [14]. The polycrystalline (meltgrown) samples were cleaved or cut from larger boules. The luminescence efficiency of samples with cleaved surfaces was a few orders of magnitude higher than that of cut samples. After etching in 1: 1 HC1/HNO3 an increase of the efficiency of the latter samples, to almost the value for the cleaved samples, could be obtained. The single crystals prepared by chemical vapour transport were first washed in an aqueous K! solution, CS2 and acetone in order to remove iodine, iodides and sulfur present on the crystal surfaces (see ref. [13]). The photoluminescence measurements were performed as a function of temperature (4.2 to 220 K), and excitation power. An argon laser (Spectra Physics 164-06) was used as an excitation source at a wavelength of 514.5 nm. In order to achieve excitation powers of 20 mW and less, neutral density filters were used. The maximal excitation power used was 200 mW; at higher values local heating of the sample was observed. The exciting beam was passed through a laser monochromator (Anaspec 300S) to eliminate argon-gas discharge lines. The sample was mounted in a continuous flow cryostat (Oxford Instruments), which was cooled by means of liquid helium. The emission spectra were analysed by means of a 0.5 m monochromator (Monospek 600) used at 1 meV resolution for most experiments and a photomultiplier tube (EM! 9684) provided with an SI cathode and cooled to 70°C.The spectra were corrected for the photomultiplier sensitivity by means of a computer program. —
38
J.J.M. Binsma et a!.
/
Luminescence of CuInS
2. I
3. Results and discussion 3.1. General observations In fig. 2, typical emission spectra are shown for In-rich (points b, c in fig. 1) and Cu-rich (points a, d in fig. 1) CuInS2. For both types of CuInS2, broad emission bands in the region 1.35—1.45 eV together with narrow lines in the region 1.50—1.55 eV near the band gap were observed. The location on the energy scale of the narrow lines, among which are several exciton emission lines, is the same for In-rich and Cu-rich CuInS2. The constancy of the photon energy of these near-edge emission lines indicates that the band gap of the material is not affected by differences in composition. The narrow line emission will be treated in detail in Part II. The broad band emission, 0.10 to 0.20 eV below the band gap, is made possible by defects which introduce localized levels within the forbidden band. When going from In-rich to Cu-rich CuInS2 a decrease of 50 meV in the energy of the maxima of the broad band emission is observed (fig. 2). Because of the invariance of the band gap, the changes in the peak energies of the broad band emission should be due to changes in the nature and concentration of the defect levels themselves. In table 1, the peak energies of the broad band emission of samples prepared in different ways are summarized. In all cases, except for quenched CuInS2, two broad emission bands are present in Cu-rich samples at about 1.39 and 1.36 eV and in In-rich samples at about 1.44 and 1.41 eV. These energy values have to be interpreted as characteristic values; for different samples there is a variation of up to about ± 10 meV, depending on the composition, temperature and laser intensity (see subsects. —
—
—
4.2K Cu-rich
/
1.35
\n_rich
—
Ji 1.30
/
~
1.40
1.45
hv (eV)
1.50
155
Fig. 2. Emission of Cu-rich (dashed line) and In-rich (solid line) CuInS2 at 4.2 K and constant excitation intensity (50 mW).
J.J.M. Binsma et a!.
/
Luminescence of CuInS
2. I
39
3.2 and 3.3). Independent of the type of sample, temperature and laser intensity, the separation between the high- and low-energy bands was found to be of constant value 37 ± 2 meV for both Cu- and In-rich samples. In fig. 3, the emission of In-rich CuInS2 is shown at different temperatures. The peaks shift to higher energy, leaving the separation between high- and low-energy emission unchanged. The intensity ratio of the high- and low-energy emissions remains nearly constant, although this is somewhat obscured by the broadening of the emissions. From table 1 it follows that there is no difference in peak energy between samples prepared by annealing in an atmosphere containing one of the constituent metals and as-grown samples prepared from a melt with excess In or Cu. Another important observation is that the type of emission is not influenced by the electronic conductivity type, which is determined by the deviation from stoichiometry (z~yin fig. 1) [2—4,14].Cu-rich samples annealed under minimal sulfur presence (~y<0, point d in fig. I and n-type conductive) show emission at the same photon energy as Cu-rich samples annealed under maximal sulfur pressure (z~y>0, point a in fig. 1 and p-type conductive). This similarity is also observed for In-rich samples (points b and c in fig. 1). From these observations an important conclusion can be drawn, that the type of luminescence mainly depends on the deviation from molecularity, where the type of conductivity is fixed by the deviation from stoichiometry. As-grown samples, which were not intentionally grown with deviations from the ideal composition, showed either the behaviour of Cu-rich samples or that of In-rich samples. This was also observed for CVT-grown CuInS2 crystals. Typical broad band emission spectra of CVT-grown CuInS2 (both as grown and annealed) are shown in fig. 4. After annealing, the characteristic emissions of Cu-rich or In-rich CuInS2 are obtained, just as for melt-grown samples. The intensities of
Table I Peak energies for broad band emission of as-grown and annealed CuInS2 at 4.2 K under constant excitation intensity Type of sample
E~(eV)b)
As grown, n.p. ~) (melt, CYT) Cu103In099S2 (melt) Annealed (Cu2S/S or Cu2S/Cu—In) Cu097In101S2 (melt) Annealed (CuIn5S8/S or CuIn5S8/In) Quenched from melt
1.44, 1.41 or 1.39, 1.36 1.39, 1.36 1.39, 1.36 1.44, 1.41 1.44, 1.41 1.44, 1.41, 1.39, 1.36
b)
n.p. = nominally pure. The accuracy of all values is ±0.01eV.
J.J.M. Binsma et a!.
40
/
Luminescence of CuJnS
2. I
• E~i~1
1.30
1.35
1.40
1.45
hv (eV)
1.50
1.55
Fig. 3. Broad band emission of In-rich CuInS2 as a function of temperature at constant excitation intensity (50 mW). (1) 4.2 K, (2) 80 K and (3) 120 K.
the low- and high-energy emission bands are almost equal for as-grown
CVT-crystals. Samples, which were quenched from the melting point, 1090°C, [15] to room temperature by pulling the ampoule out of the furnace, exhibited all four broad emission bands, both those of Cu-rich and In-rich samples. After annealing, these samples showed only the emission bands belonging to the type of annealing atmosphere. This experiment strongly indicates that the defects involved in the emission processes are in equilibrium with the gas phase.
Cu anneal
In anneal
~
1.30
I
1.35
1.40
1.45
hv (eV)
1.50
155
Fig. 4. Broad band emission of CVT-grown CuInS2: as grown (I), Cu-annealed (2) and In-annealed (3), at 4.2 K and constant excitation intensity (50 mW).
J.J.M. Binsma et a!.
/
Luminescence of CuInS
2. I
41
Before going into the details of the defect chemistry (see sect. 4), it is necessary to discuss the temperature and excitation intensity dependence of the broad emission bands. 3.2. Peak energy as a function of temperature As already shown in fig. 3, all broad emission peaks shift to higher energies with increasing temperature. In fig. 5 the shift in the peak energy i.~E = E~(T) E~(4.2) of the high-energy emission (1.44 eV) is plotted for three In-rich CuInS2 samples, viz. (I) annealed with Cu1n558 and 5, (2) as grown and (3) annealed with CuIn5S8 and In. Also the temperature dependence of the energy gap E5 as determined from measurements of the exciton emission (see part II) is shown in this figure. The separation between the high- and the low-energy emissions remains constant, viz. 37 ±2 meV, independent of the temperature. The characteristic variation of the peak energies for any sample can thus be represented by that of the high-energy emission (Er) alone. It is clear that the peak energies are increasing more rapidly than Eg with increasing T in the range 4.2—80 K. Above 80 K when Eg decreases, still an increase in is found. For most samples, 8Ev/aT was about 4 X l0~ eV K but also larger values were observed, occasionally up to a maximum of 17 X I0~ eV K The additional shift of E1, with respect to Eg is an indication of donor— —
~,
/
/ ~
(meV)
/
2
zv~ —
GO
r-~!.0
I 50
I 150
100 T(K)
Fig. 5. Shift in the peak energy ~E = E~(T)— E~(4.2)as a function of temperature for three In-rich CuInS2 4.2)samples as reported (1.44 in eVPart emission). II. (1) The annealed dashed with lineCuIn represents the shift in the energy gap Eg(T)_ Eg( 5S8 and S, (2) as grown and (3) annealed with CuIn5S8 and In.
42
J.J.M. Binsma et a!.
/
Luminescence of CuInS
2. I
acceptor (DA) emission [16]. The emission energy for a donor—acceptor pair at distance r is, in the first approximation, 2/r. (1) hv=Eg—EA—ED+e
When the temperature is increased, the migration of mobile charge carriers to more favourable sites (smaller r) is enhanced, thus giving a shift in hv additional to the shift in E 5. Because the additional shift depends on the degree of compensation [17], different samples will show different temperature coefficients as is illustrated by fig. 5 for In-rich samples. Also Cu-rich samples show this behaviour. On the contrary, for free to bound emissions, which involve the valence band or the conduction band, the same temperature coefficient is expected for all samples, as follows from the corresponding peak energy in that case [16]: Ep=Eg~Ei+kT,
(2)
where E1 stands for the (temperature independent) ionization energy of the centre (either donor or acceptor) which captures the free carrier. The temperature coefficient for all samples is given in the latter case by aE aE p— 3 8T~T which is not observed. So it can be concluded that the broad band emission of both In-rich and Cu-rich CuInS2 is of the donor to acceptor type. At high temperatures (> 100 K), however, an increasing contribution of a conduction band to acceptor emission is found at the high-energy edge (see the forthcoming subsections). This shift in the character of the emission may be partly responsible for the increase in E~at high temperatures. 3.3. Peak energy as a function of excitation intensity The peak energy of the various observed broad band emissions appears to be a function of the intensity of the exciting beam, the general trend being an increase with ii~creasingexcitation intensity. This is demonstrated by fig. 6, where the shift in E~for the high-energy emission (1.41 eV) is plotted for the series of In-rich CuInS.2 samples already presented in subsect. 3.2 over three decades of excitation power P1 at 4.2 K. As already mentioned in subsect. 3.1, the separation between the high- and low-energy emissions is unaffected by the excitation intensity and remains constant at 37 meV. E~,as a function of P1 is behaving according to the relation which usually applies to the DA emission in compensated semiconductors [17]: E~=E~+f3(logP~ —logP10),
(4)
with E1, as the peak energy at P1 and /3 the energy shift per decade shift in P1.
/
J.J.M. Binsma et a!.
30.0
Luminescence of CuInS
2. I
43
-
AE
1,,,
(meV) 2G0
100
0.0
/
-
/
-
0.0
A
/~
~ I
1.0
3.o
2.0 Iog~—
Fig. 6. Shift in the peak energy ~E as a function of the relative excitation intensity at 4.2 K for the 1.44 eV emission of three In-rich CuInS2 samples (numbering of samples as in fig. 5). ~ E = E~(P1) —
E~(P1).
The parameter /3 increases with increasing degree of compensation and decreases with increasing acceptor and donor concentrations [181.Indeed various values for /9 were found at low temperatures (4.2 K) ranging from 0 to 25 meV for In-rich CuInS2 (peaks at 1.44 and 1.41 eV). For Cu-rich CuInS2 (peaks at 1.39 and 1.36 eV) /9 ranges from 0 to 15 meV at T 4.2 K. The increase of E~with increasing P1 can be explained for small values only a few meV of /3 on the basis of a decrease of the mean r value in eq. (1) at higher excitation levels [19]. Another mechanism which can also explain larger values > 10 meV of /3 is band perturbation by high concentrations of defects [17]. Because high concentrations of, probably, intrinsic defects can be present in CuInS2 [2,3,14] this explanation may hold for sample I. The parameter /3 appears to be a function of temperature: it decreases rapidly with increasing temperature and becomes ~ I meV at 80 K. At higher temperatures the migration of charge carriers to more favourable sites is enhanced (see sect. 3.2) thus giving rise to a smaller possible influence of the excitation intensity. This accounts for the small /9 values at higher temperatures. For free to bound emission also an increase in E~with increasing P~should be observed [17]. In contrast to DA emission, however, the shift in this case should be independent of the type of sa.mple, for no substantial broadening of the acceptor levels is to be expected because of the large ionization energy of the trapping center. In addition, the high-energy edge of the emission band —
—
—
—
/
Luminescence of CuInS
1.50
1.55
J.J.M. Binsma et a!.
44
1.35
1.40
1.45
hv (eV)
2. I
Fig. 7. Broad band emission of In-rich CuInS2, at 4.2 K, at different excitation intensities, showing the structure at the high-energy edge for high excitation.
should be abrupt corresponding to the quasi Fermi level of the free carrier band [17]. The high-energy edge of the emission of both Cu-rich CuInS2 is not very abrupt, however, and becomes even less sharp at increasing P1. At high temperature (T> 100 K), free to bound emission may contribute to the high-energy emission, as was also observed in ref. [9]. This makes the interpretation of the spectra difficult for T> 100 K (see sect. 3.4). At the highest excitation intensities and low temperature (4.2 K), some structure is observed in the high-enegy edge (fig. 7). This structure may correspond to donor—acceptor emission lines for relatively small separations. Weak DA lines are commonly observed at the high-energy edge of DA emission bands with strong phonon coupling (bell shaped bands) [16]. The low-energy emission shifts to higher energy with increasing P1 in the same way as the high-energy emission, leaving the difference (37 + 2 meV) between both peak energies unchanged. This indicates that the concentrations of the donors and acceptors involved in both transitions are not very different. 3.4. Emission intensity as a function of temperature In order to obtain additional information about the centers involved in the emission processes in CuInS2, the intensity of the emission was studied as a function of temperature. When trapped electrons or holes are playing a role in the recombination, quenching of the emission will occur at high temperatures by ionization of the trapping center [16]. The thermal activation energy for quenching can be related then to the ionization energy of the trapping center. In the case of two trapping centers, thus for DA emission, the interpretation of
J.J.M. Binsma et a!.
/
Luminescence of CuInS
2. I
45
the activation energy is not quite unambiguous, but may still give useful information. In fig. 8 the intensity (approximated by peak height times peak width) of the (high energy) 1.44 eV emission of an In-rich sample and the intensity of the (high energy) 1.39 eV emission of a Cu-rich sample are plotted as a function of temperature. The observed behaviour of the intensity I as a function of T is in accordance with a thermally activated probability for non-radiative recombination as described by Curie [20]: I(T)= ‘~°~ (5) 1 + Cexp[—LIE/kT] with LIE as activation energy for non-radiative recombination and C being a measure for the capture cross section of the center which is thermally emptied. From the curves plotted in fig. 8, LIE is calculated to be 90 meV for In-rich material (1.44 eV emission) and 145 meV for Cu-rich material (1.39 eV emission). For DA emission LIE should be equal to the ionization energy of the shallower level at low temperature, and to the sum of the ionization energies, EA + ED, at higher temperatures. The values of LIE which are calculated here are too high to correspond to the shallower and of are0.11 too eV lowfor to In-rich be equaland to 2/erlevel a value EA 0.16 + ED.eVEq. yields for EA + ED e of for(1)Cu-rich samples. For e2/er a value of about 0.03 eV can be assumed [9], which gives for EA + ED 0.14 and 0.19 eV, respectively. On the other hand, the LIE values are in good agreement with the acceptor levels found in p-type CuInS 2 by electrical measurements [14]. In p-type In-rich CuInS2 a level of 100 meV was observed whereas p-type Cu-rich material possesses an acceptor with EA = 150 meV. The 150 meV acceptor was also —
~E—
1(K)
100 GO
Cu-rich
~
~E.145eV
—2.0 .
/
50
~—~°
/
S
~
~
/
In—rich /AE~.O90eV
/1 -40
10
20
io~
30
T (K~)
—.
250
Fig. 8. Relative intensity of the broad band emission (1.44 eV) of In-rich CuInS2 (x) and the broad band emission (1.39 eV) of Cu-rich CuInS2 (9) as a function of temperature at constant excitation intensity (50 mW).
46
J.J.M. Binsma et a!.
/
Luminescence of CuInS
2. I
reported by Look and Manthuruthil [2], and Verheijen [3]. They performed Hall measurements on CuInS2 annealed only in the presence of sulfur. Because of the high vapour pressure of the M111-sulfides above M1 M111 X2 chalcopyrites compared to the pressure of the M1-sulfides [21], material annealed only in the presence of sulfur will tend to become M1-rich. So it is clear, that in sulfur-annealed CuInS2 only the 150 meV acceptor belonging to the Cu-rich material has been observed. These considerations lead us to the conclusion that different acceptors with ionization energies of 100 and 150 meV, respectively are involved in the DA emissions of In-rich and Cu-rich materials. An activation energy of 85 meV for quenching was found by Lahlou and Masse [9] for In-rich CuInS2. They also did not observe EA + ED as an activation energy. This was explained by assuming a large contribution of free to bound (here, conduction band to acceptor) emission, which might be possible at high temperature (T> 100 K). The intensities of the emissions at lower energies (1.41 and 1.36 eV) decrease simultaneously with the high-energy emission when the temperature is increased. This leads to an almost temperature-independent ratio of the intensities of the high- and low-energy emissions up to 100 K (see fig. 3). At higher temperatures an accurate determination of the intensity of the low-energy emission becomes difficult because of the broadening of the emission bands. So the activation energy for quenching could not be determined and no indication about the ionization energies of the donors and acceptors involved in the emissions can be obtained. The observed weak temperature dependence of the intensity ratio together with the varying intensity of the low-energy emission for different samples provide the arguments for the interpretation as an additional DA emission rather than in terms of a phonon wing (see also subsect. 3.6). As is already argued above, the shift of the entire broad band emission over 50 meV to lower energy when going from In-rich to Cu-rich material can be ascribed to a change in the nature of the acceptor involved. From the difference in peak energy between the high- and low-energy emission (37 ± 2 meV, see subsect. 3.1), it can now be deduced that the donor involved in the low-energy emission possesses an ionization energy which is larger than that of the donor belonging to the high-energy emission by the same amount 37 meV. —
—
—
3.5. Emission intensity as a function of excitation intensity
Apart from the peak position as a function of excitation intensity also the emission intensity (see fig. 9 for In-rich CuInS2) has been studied. in the temperature range 4.2 to 150 K over three decades of excitation intensity. The emission intensity I varied as I = CP~with C and x as constants and P1 as the excitation intensity. At 4.2 K, a ranges from 1.20 at lowest excitation intensities to 0.63 at highest excitation intensities. This applied for both Cu-rich and
J.J.M. Binsma et aL
4.0
/ Luminescence of CuInS2. I
47
,74.2K
LogI (au.)
/
//
30
/65K /X
/5
/
/
20
1.0
.
0.0 0.0
7
//
/
/X /128K
(1*10)
/1 /5
I
I
1.0
2.0
I
3.0 Log9 (au.)
Fig. 9. Intensity of the 1.44 eV emission of In-rich CuInS2 at three different temperatures (4.2, 85 and 128 K) as a function of the relative excitation intensity.
In-rich CuInS2. With increasing temperature but constant excitation intensity a increases, the range being 1.5—1.0 at T> 70 K. The temperature and excitation intensity dependence of a observed here is in good agreement with the theory deduced by Maeda [22] for DA transitions (compare figs. 8 and 9). According to this theory, temperature regions with different temperature dependences of I have their characteristic ranges of a values. In the low-temperature region (no temperature dependence of I), the excitation intensity dependence of I should range from linear (low P1) to sublinear (high P1). At high temperatures, where I quenches with an activation energy LIE, the dependence should vary from supra-linear or even quadratic to linear at high P1. This behaviour is reflected in fig. 9. Care has to be taken, however, in the interpretation of the curve at 128 K, because the emission may originate in part from free to bound transitions (see the foregoing subsection). The ratio of the intensities of the high- and low-energy emissions, 11/12, appeared to be nearly independent of the excitation intensity. Only when an appreciable contribution of the deep donor—acceptor transition is present (11/12 <<5), the high-energy emission is slightly favoured with increasing excitation intensity with respect to the low-energy emission (11/12 increases slightly). A saturation of the low-energy emission at high excitation intensities
J.J.M. Binsma et at
48
/
Luminescence of CuInS
2. I
may be responsible for this behaviour, as already noted by Verheijen [3], who reported similar observations for the broad band emission of CuInS2. 3.6. Interpretation in terms of defect levels In the foregoing the (high energy) emissions at 1.44 eV for In-rich and at 1.39 eV for Cu-rich materials were correlated to recombinations between the same donor (ionization energy ED1) and two different acceptors, their ionization energies being 100 and 150 meV. The low-energy emissions (1.41 and 1.36 eV) are attributed to donor—acceptor recombinations from a deeper donor (ionization energy ED + 37 meV) to the acceptor levels which participate also in the high-energy emission. Because the separation between high- and low-energy emissions is close to the optical phonon modes of CuInS2, which are lying in the range 35—40 meV [23], the possibility of a phonon-assisted transition should be considered too. Two arguments can be given against this interpretation, however, viz. (i) the strongly varying intensity ratio between the high- and low-energy emissions for different samples, ranging from 0.7 to 5, and (ii) the weak temperature dependence of the intensity ratio (see subsect. 3.4 and fig. 3), which decreases at most 10% from 4.2 to 100 K. The intensity of a phonon-assisted transition involving emission of a phonon to the lattice should quench with respect to the zero-phonon transition according to the factor exp[hw/kT], h~being the phonon energy [24]. The only unknown ionization energy so far is that of the shallower donor, ED. There are strong indications, however, that the correct value of ED is about 35 meV. In the first place, we found by electrical measurements on n-type CuInS2 a donor level of 35 meV [14]. Further, a donor to valence band transition is observed at 1.52 eV which also yields a donor-ionization energy of 35 meV (see Part II). Finally, fitting results of time-resolved measurements by Lahlou and Masse reveal a level of 45 meV [9]. The donor and acceptor levels present in Cu-rich and In-rich CuInS2 are summarized in table 2 together with the peak energies of the donor—acceptor Table 2 Peak energies and donor and acceptor levels for Cu-rich and In-rich CuInS2 (accuracy for E0 is ±5 meV and for EA, ±10meV) Molecularity
E~(eV)a)
ED(meV)
EA(meV)
Cu-rich
1.39 1.36
35 72
150 150
In-rich
1.44 1.41
35 72
100 100
a)
Measured at 4.2 K.
J.J.M. Binsma et a!.
cond. band ~\\~\\\\\
—
/
Luminescence of CuInS
2. I
49
cond. band ~\\
~1\1
3~i.T
T~\j.:oev
f~v
I.5nej
~,,
,,,,,,,,,I~,,
vat, band
val. band
Cu-rich CuInS 2
(a)
In-rich CuInS2 (b)
Fig. 10. Energy level diagrams for (a) Cu-rich CuInS2 and (b) In-rich CuInS2, showing the various radiative transitions and the donor and acceptor levels involved. Also the donor to valence band transition at 1.52 eV, which will be reported in Part II, has been indicated. Energy balance is obtained for the DA emissions when the Coulomb energy ( 0.03 eV) is included. A1 is assigned to either V1,, or Cu1,, and A2 to Vt,,. The nature of D1 and D2 is not known up to now (see Sect. 4).
emissions originating from these levels. In fig. 10, the emission processes involving the donor and acceptor levels are schematically drawn for Cu-rich (a) and In-rich (b) CuInS2. Also the donor to valence band transition found near the band edge [11] has been indicated. When energy balances are2/r constructed for DA for fig. 10, one should take into account the Coulomb energy e emission (see eq. (1)), which corresponds to about 0.03 eV at the emission peak [9]. The nature of the various defects corresponding to the donor and acceptor levels and their dependence on molecularity and stoichiometry will be discussed in the following section.
4. Defect-chemical considerations The experiments on samples which were quenched from the melting point and those on samples which were equilibrated under various conditions strongly indicate that the luminescence properties are governed by intrinsic defects. In a ternary compound, a wide variety of intrinsic defect equilibria is present. These equilibria can be classified into internal equilibria involving only defects in the solid and interphase equilibria between solid and gas phase [7,25]. Compared to binary compounds, additional mechanisms for defect formation originate from deviations from the ideal molecularity and from cation-disorder. In terms of defect-equilibria these mechanisms can be formulated as follows: ~
J.J.M. Binsma et a!.
50
—
—I
a ~Pcu2sL
rv*Cul12F~j*1_L I S J
/ Luminescence of CuInS2. I
‘
Cu~~ + In~~ Cu’r~+ In~~ Kh=[Cu~fl][In~U].
(7)
The equilibrium given by (6) may also be formulated in terms of the indium sulfide pressure and V~.Because the copper and indium sulfide pressures are coupled via the formation reaction, both representations are equivalent. In the equilibria (6) and (7) the point defects have been represented in their neutral state. Charge carriers are not involved in these equilibria, by which comparable amounts of donors and acceptors are formed. This in contrast to the equilibrium which describes deviations from stoichiometry: 2S~v2V~+S2(g) 2p K~ [V~] 5.
(8)
The formation of n-type material as a result of sulfur deficiency is directly given by (8), because V~’acts as a donor. When the equilibrium is formulated in terms of sulfur interstitials (which are acceptors) it will be clear that a sulfur-excess gives rise to p-type material. Here charge carriers are thus indirectly formed. The occurrence of disorder corresponding to (6) and (7) may explain why for the chalcopyrites usually high compensation is found in contrast to the binary compounds, e.g. Il—VI compounds [25]. The attribution of specific defects to the observed levels (see table 2) will require further measurements (e.g. of the electrical properties), but some conclusions can already be drawn. As the nature of the acceptor level changes with annealing or quenching it may be attributed to an intrinsic majority defect. Table 3 gives a survey of the various defects which can act as acceptors in CuInS2 together with their conditions for the deviation from molecularity (LIx>0: Cu-rich; LIx<0: In-rich). These conditions can be deduced on the basis of the relations between exact composition and defect contents given in ref. [7]. For In-rich CuInS2 the acceptor at 0.10 eV may correspond to V1~ when we exclude the energetically unfavourable 5,. The acceptor at 0.15 eV is Table 3 Possible intrinsic majority defects acting as acceptors and the conditions for their occurrence in terms of the deviation from molecularity, ~x = [Cu]/[Inl — Defect Vc~
<0
V1,, Si Cu1,,
>0 <,=,>0 >0
J.J.M. Binsma et a!. / Luminescence of CuInS
2. I
51
correlated to Cu-rich CuInS2 as revealed by this work and that of other authors (see subsect. 3.4). Table 3 gives two possible defects for this level, viz. V1,, and Cu ~ when S, is again excluded. It is difficult to discriminate between the remaining possibilities, but that the acceptors are of intrinsic character and not due to impurities is indicated by the fact that an acceptor with an ionization energy of about 0.15 eV is found for various Cu-containing chalcopyrites [26] independent of their band gap. The energy levels of a cation vacancy (surrounded by four anions for the chalcopyrites) are solely related to the valence band as shown by Lang et al. [27]. Changing the type of cation (e.g. in Al~GaI _~As)does not alter the environment of the cation vacancy, resulting in an acceptor level which is fixed relative to the valence band (condition of unchanged environment). Impurity levels on the other hand are related to both the valence band and the conduction band and are thus shifting with the band gap when varying the nature of the cations. For the same reasons the ionization energy of the CuM acceptor (corresponding to a copper atom surrounded by four anions) will also not be affected by the band gap. It is clear from the defect-chemical considerations summarized in table 3 that the attribution of the 0.15 eV acceptor to either V1,, or Cu In is much more plausible than that to Va,, as given in ref. [26]. In addition, V1~and Cu1~fulfil the condition of unchanged environment in the CuM111 X2 compounds even better than Vc~ does, if second nearest neighbours are included. For the donor levels it is not clear up to now whether they correspond to majority or minority defects and a definite attribution is thus not yet possible for these levels. Various defects are possible such as V5, Cu,, In,, Inc~ or foreign elements incorporated in the lattice. Especially the undesired incorporation of iron may be significant as has been demonstrated by EPR measurements for CuInS2 and other chalcopyrite compounds [28—30].Spectrochemical analysis reveals the presence of 3Fefor as both by farmelt-grown the most important impurityCuInS up to and CVT-grown a concentration of 5 X 1018 cm 2 [14]. This concentration is equal to the measured majority defect concentrations, which means that a defect-chemical model of CuInS2 is not complete when the role of iron is ignored. Measurements which should lead to a better understanding of this role are under way. Finally, it may be noticed that the interpretation of the levels at 45 and 90 meV as given by Masse et al. [10] needs to be reconsidered. They ascribed the 45 meV level to the VCU acceptor on the basis of the analogy with some other chalcopyrites. From our optical and electrical experiments it can be deduced that the 90 meV level corresponds to ~ whereas the 45 meV level originates from a donor. The nature of this donor has still to be revealed (see above).
52
J.J.M. Binsma et a!.
/ Luminescence of CuInS2. I
5. Conclusion By studying the broad band emission as a function of the exact composition, intrinsic defects were found to play an important role in the defect-chemistry of CuInS2. On the one hand, deviations from molecularity (Cu-excess or In-excess) determine the type of broad band emission. This shows that the ternary character of the compound is an important factor in fixing the defect-chemical state of the compound. The influence of deviations from stoichiometry, on the other hand, is reflected in the electrical properties, sulfur-rich material being p type and metal-rich being n type. The various types of CuInS2 can be converted into each other by annealing in specific atmospheres corresponding to the possible three-phase equilibria. Finally, it should be mentioned that the melt-grown and CVT-grown crystals showed no fundamental differences in emission.
Acknowledgements The authors would like to thank Mr. H.W.J.M. Hoekstra, who participated in the emission measurements. The investigations were performed as part of the research program of the Stichting Scheikundig Onderzoek in Nederland (SON) with financial support from the Nederlandse Organisatie voor Zuiver Wetenschappelijk Onderzoek (ZWO).
References [I] B. Tell, J.L. Shay and H.M. Kasper. Phys. Rev. B4 (1971) 2463. [2] D.C. Look and J.C. Manthuruthil, J. Phys. Chem. Solids 37 (1976) 173. [3] A.W. Verheijen, Thesis, University of Nijmegen, Nijmegen (1979). [41B. Tell, J.L. Shay and H.M. Kasper, J. Appi. Phys. 43 (1972) 2469. [5] L.L. Kazmerski, M.S. Ayyagari and G.A. Sanborn, J. AppI. Phys. 46 (1975) 4865. [6] H. Neumann, B. Schumann, D. Peters, A. Tempel and G. Kuhn, Kristall und Technik 14 (1979) 379. [7] J.A. Groenink and PH. Janse, Z. Phys. Chem. N.F. 110 (1978) 17. [8] J.M. Meese, J.C. Manthuruthil and D.R. Locker, Bull. Am. Phys. Soc. 20 (1975) 696. [9] N. Lahlou and G. Masse, J. AppI. Phys. 52 (1981) 978. [10] G. Masse, N. Lahlou and C. Butti, J. Phys. Chem. Solids 42 (1981) 449. [Ill J.J.M. Binsma, L.J. Giling and J. Bloem, J. Lumin. 27(1982)55. [12] C. Paorici, L. Zanotti and 0. Zucalli, J. Crystal Growth 43 (1978) 705. [13] JiM. Binsma, W.J.P. van Enckevort and G.W.M. Staarink, J. Crystal Growth, in press. published. [14] J.J.M. Binsma, to be published. [15] J.J.M. Binsma, L.J. Giling and J. Bloem, J. Crystal Growth 50(1980) 429. [16] P.J. Dean, in: Progress in Solid State Chemistry, Vol. 8. Eds. JO. McCaldin and G. Somorjai (Pergamon, New York, 1973) p. 1.
J.J.M. Binsma et a!.
/
Luminescence of CuInS
2. 1
53
[17] J.I. Pankove, Optical Processes in Semiconductors (Dover, New York, 1975). [18] MI. Nathan and T.N. Morgan, in: Physics of Quantum Electronics, Eds. P.L. Kelley, B. Lax and P.R Tannenwald (McGraw-Hill, New York, 1966) p. 478. [19] E. Zacks and A. Halperin, Phys. Rev. B6 (1972) 72. [20] D. Curie, Luminescence Cristalline (Dunod, Paris, 1961). [21] S.S. Strelchenko, S.A. Bondar, A.D. Molodyk, LI. Berger and A.E. Balanevskaya, Izv. Akad. Nauk. SSSR, Neorgan. Mater. 5 (1969) 593. [22] K. Maeda, J. Phys. Chem. Solids 26 (1965) 595. [23] W.H. Koschel and M. Bettini, Phys. Stat. Sol. (b) 72 (1975) 729. [24] H. Barry Webb and E.W. Williams, in: Semiconductors and Semimetals, Vol. 8, Eds. R.K. Williardson and A.C. Beer (Academic Press, New York, 1972) p. 182. [25] F.A. Kroger, Chemistry of Imperfect Crystals (Norh-Holland, Amsterdam, 1973). [26] H. Neumann, D. Peters, B. Schumann and G. Kuhn, Phys. Stat. Sol. (a) 52 (1979) 559. [27] D.V. Lang, R.A. Logan and L.C. Kimerling, Phys. Rev. B 15 (1977) 4874. [28] G. Brandt, A. Rauber and J. Schneider, Solid State Commun, 12 (1973) 481. [29] H.J. von Bardeleben, A. Goltzené, C. Schwab and R.S. Feigelson, AppI. Phys. Lett. 32 (1978) 741.
[30] Hi. von Bardeleben and R.D. Tomlinson, J. Phys. Cl3 (1980) Ll097.