- Email: [email protected]
Photoconductivity in tellurium at 77 K
002o.aY91 81 050311.12so2.00 0 Perpmon Press Ltd
PHOTOCONDUCTIVITY
IN TELLURIUM
AT 77 K
I. SHIH and C. H. CHAMPNESS Department of Electrical Engineering. 3480 University Street, Montreal, P. Q.. Canada H3A 2A7 (Received 9 April
1981)
Ahafract-Photoconductivity has been measured in monocrystalline samples of undoped tellurium at 77 K in the wavelength range 1.4 to 4.4 pm. The samples were chemically cut from a Czochralski-grown ingot. The photoresponse decreased by about an order of magnitude as the temperature was increased from 77 to 180 K in the extrinsic range, where holes dominate. At 77 K, the photoresponse was found to be sensitive to sample treatment; it was increased by annealing (at 395°C for 168 hr) and decreased by abrasive polishing. The photoconductive decay, following spark source excitation, was approximately describable in terms of a short time constant I,. measured in microseconds, and a longer time constant T,. measured in tens of microseconds.The ac. photoresponse was found to decrease with frequencyf: particularly asjexceeded I/T, Measurements were made of current noise and spectral detectivity D*. At 77 K, a D.-value of 1.3 x 10” cm Hz”’ W- ’ was observed at i = 3.5 pm on an annealed sample of thickness about 0.2 mm. This appears to be the highest D* yet reported on melt-grown tellurium. An even higher D*. up to the background limited value at I = 3.5 pm, should be possible with further reduction of sample thickness.
INTRODUCTION
Photoconductivity has been known in tellurium for more than half a century. In 1925, Bartlett(‘) first showed that the effect in thin films was larger at liquid oxygen temperature than at room temperature. Some years later Moss (‘) found the photoconductivity maximum to lie at a wavelength near 1.3 pm for cooled thin tellurium films evaporated on glass substrates. However, the first results on bulk crystal samples were reported by Loferskit3’ in 1954. These samples were obtained by cleaving a Bridgman-grown ingot containing randomly oriented crystals, followed by abrasive grinding and polishing with final chemical etching. Edwards et ~1.‘~’published results, in 1961, on samples prepared by vapour phase growth and by cleaving a Czochralski-grown crystal. On one of their vapour-grown samples they reported a detectivity at 77 K, and at a wavelength of 3.4 pm which was essentially background limited. Vis(5) also studied photoconductivity on bulk samples, which were prepared by abrasive cutting from slices cleaved from tellurium Czochralski-grown crystals, followed by chemical etching. In addition he studied the dependence of noise on frequency and temperature. N) In 1966 Grosse and Winzer”’ made a detailed study of photoconductivity in bulk samples, cut’abrasively from Czochralski crystals followed by etching. They observed the decrease in response with decrease of wavelength below 3.5 pm and found this to be more rapid for the electric vector E of the radiation perpendicular to the c-axis of the sample (i.e. Elc) than for that with E parallel to it (i.e. Ellc). In all the above mentioned previously reported work, abrasive action was used in the preparation of the samples and, while chemical etching was often carried out, lattice damage in the material would not have been completely removed. In recent work, improvements have been made in the Czochralski crystal growth method,‘*) and in the development of techniques for ensuring good sample surfaces”’ of tellurium. Because of these material improvements, and an increased awareness of the sensitivity of tellurium to imperfections arising from improper handling, it was considered timely to make a preliminary reassessment of the potential of tellurium as an infrared detector material operating near a wavelength of 3.5 pm. The first part of such a study was reported on in a recent paper, (‘O’where the variation of the photoresponse and the shift of its peak with change of sample thickness was investigated. The second part of the study constitutes the present paper, where the effect of temperature variation and heat treatment are specifically examined and where the results of measurements of noise and responsivity as a 311
312
I. SHIH and C. H. CHAMPNESS
function of frequency, are reported on. In this work all abrasive treatments were avoided in the preparation of the tellurium samples, except in a deliberate experimental case.
EXPERIMENTAL
PROCEDURE
Crystal growth
The samples for study were obtained from a single monocrystalline ingot of tellurium No. CZ-78-11. This crystal was grown by the Czochralski method in our laboratory, according to the procedure described. (s) Briefly, it was grown at a vertical pull rate of 1 cm/hr in the c-direction from high purity material (6-9’s) obtained from Cominco Ltd. The background temperature during growth was 360°C. The ingot was 8 cm in length and 1.6 cm in diameter, and its high quality was established from etching, Laue diffraction patterns and from Hall mobility measurements on samples at 77 K. Sample preparation
The samples for photoconductivity measurements were chemically cut, initially as wafers, from the tellurium ingot with a string saw using a solution consisting of HCl:CrOs: Hz0 in the ratio 1: 1:2 by weight. The wafers were then mounted on aluminium blocks and chemically polished successively on two parallel (lOi0) faces, using a solution of composition HNOs:CrOJ:H20 in the ratio 1:2:4 by weight. The final dimensions were about 8 mm in length (parallel to c-direction), 2 mm in width and the thicknesses from 0.15 to 2.2 mm, as indicated in Table 1, column 2. After the chemical polishing, one of the samples (CZ-78-1 l-AB) was also given an abrasive polish on the two parallel (lOi0) surfaces, using 0.05 m diameter A120J powder on glass for 2 min. Two other samples were chemically cut from an annealed slice. This annealing was done by placing the tellurium slice in a quartz ampoule, which was then evacuated and back-filled with argon at 300 torr pressure and sealed off. The ampoule was then maintained at 395°C for about 170 hr. After slow cooling to room temperature over a period of about 50 hr, the slice was taken out and the samples (78-ll-AN and 78-ll-ANT) were chemically cut from it. Three side-probes for potential measurements were attached to each sample (Fig. 1) by melting in 0.002 in. diameter platinum wires. The ends of the samples were coated with a solder of composition 50% Sn, 47% Bi and 3% Sb by weight. A fine copper wire was attached to one end and the other was soldered to a brass plate, the central part of which was covered with an insulator, blackened to minimize reflection. A thick rectangular paper window served to screen light away from the end contacts, as indicated by the broken line in Fig. 1. Method of measurement The sample holder was mounted inside a dewar with a ZnS (Kodak IRTRAN 2) window. Current from a battery pack was supplied to the sample via a large series resistor. Infrared radiation from a Perkin-Elmer model 13 monochromator was focused on to the sample by an external mirror. The light was chopped at 13 Hz and the monochromatic light intensity was measured with a built-in vacuum thermocouple detector. The relation, between the output of this detector and the incident light energy falling on the sample, was determined by placing a calibrated InSb photovoltaic detector operating at 77 K at the position of the sample and determining its output. The 13 Hz signal from the side-probes of the sample was fed to a lock-in amplifier (Princeton Applied Research model 124A) set at the same reference frequency. The steady current through the sample was set at the maximum value consistent with linearity of a plot of photoconductive signal voltage versus sample current. The steady voltage between the probes was measured with a digital voltmeter. Photoconductive decay was also measured by recording oscillographically the decrease of photoconductivity, following illumination of the sample by a pulse of light. Initially,
CZ-18-I I-ANT
0.2
CZ-18-I I-T
,._
I.25
I .os
2.28
I.51
CZ-18-I I-AB
.~
--
polished None 395’c l68hr
None 395c 168 hr None, Abrasively
1.15
2.2
1.16
Thermal treatment
Width (mm)
CZ-18.1 I-AN
cz-18-I l-l I
Sampk no.
Thickness (dimension parallel to incident light) (mm)
Table 1. Photoconductivity
-_--
2.0
0.5
2.2
0.6
30 130
4
-
10
(L
IO
-
5
(Z)
Photoconductive decay at 77 K
2
I
1.1
I,lL
1.3 x IO”
3.7 X IO’O
1.8 x IO9
Maximum LP at 77 K 9OOHz IHzBW (cm Hz 112W-l)
. I-L-LII--^.“.-“~~-ll__l_l..-“~r^_,-“.----_I_.C-.._.- .-._ l”“-_______,
Max. Aua/&,E,) at 77 K (IO~‘Scm’sec)
results in undoped tellurium samples
5 % =: r:
5. tz
ii=
f 0. 6 t 5. Q 8’
?
I. SHIH and C.
314
H. CHAMPNESS
Fig. 1. Schematic view of the sample with the current leads and the side probes for photoconductivity measurements.
this was done with 3-pm radiation from the monochromator using chopped white light from the Nernst source. Later the source was changed to a spark unit (Xenon Corp. Nanopulser model 437A plus a 2.5 m band-pass filter), where the pulse duration was less than 100 nsec. While the intensity from this source was much greater than that from the Nernst glower, the largest change of conductivity (Aa,+rO)was still less than 0.001. The average noise voltage was also measured between the side probes on three of the samples at 77 K. This was done with the lock-in amplifier at different frequencies and a fixed band width. The noise levels were determined with and without bias current. LOW FREQUENCY
PHOTOCONDUCTIVITY
RESULTS
Variurion wifh remperature The spectral variation of Ao/(aE,) with wavelengths between about 1.4 and 4.4 ,um, measured at three different temperatures for Ellc and ELc, is shown in Fig. 2. Here Ao is photo~onducti~ty increase, no is the dark conductivity and Ep is the photon flux incident on the sample, It is noted that with increase of temperature the photo-response decreases and for Elc the peak values shift towards shorter wavelengths and the peak-to-“plateau” ratios decrease. The variation of Au/(aoE,J at 3.5 pm over a wider temperature range is shown in Fig. 3. The change near 77 K is seen to be relatively flat with a stronger decrease as the temperature is raised. The rapid decrease above about 180 K (i.e. = 5.5) occurs in the intrinsic region, where the Hall coefficient has a negative l~/r sign. Within the extrinsic region (up to 180 K) the photoconductivity changes by about an order of magnitude. The average photocondu~tive decay time T, also plotted in Fig. 3, changes by a similar amount, suggesting that this quantity predominantly controls the temperature variation of photoconductivity. The relative change of conductivity &/a0 at 77 K is plotted against irradiance E at a wavelength of 3.8 pm in Fig. 4. It is seen that up to an E value of 20 ~W/cm’ at least. the variation is linear. This is consistent with the results of Vis”) according to whom non linearity only occurs above Au/a0 value of about 0.01. Eject o~s~mpie tr~#t~n~ and thickness The spectral photoconductive response at 77 K of three samples of similar thickness is shown in Fig. 5, where each has had a different final preparational treatment; one was chemically polished, another annealed and then chemically polished and the third chemically polished and finally abrasively polished as described earlier (p. 2). It is noted that the photoconductivity is much higher for the annealed sample and that the fall-off of photoconductivity with decreasing wavelength, for the abrasively prepared sample, is much stronger than the other two. At shorter wavelengths the difference between the curves for the abrasive and annealed samples amounts almost to two orders of magnitude.
315
Photoconductivity in tellurium at 77 K
I
lt?’ ;
-18 10 _
I
I
I
0°K “lOOK o 130K
"E " 2
3 a
9 9 \ 9
I?CZ-78-11-11
1
8:
3 4 2 Wave1ength.A ,( pm 1
Fig. 2. Relative photon photoconductivity. Au/(aOEJ, vs wavelength for sample CZ-78-I I-l 1 at three different temperatures with EI’C and Elc.
The effect of annealing is also shown for the two thinner samples in Fig. 6. Here the annealed sample ll-ANT has about three times the responsivity of the unannealed sample 11-T at all wavelengths. The effect of sample thickness is evident from the results on the two non-annealed samples in Fig. 7. Here the photo-response for the thinner sample (0.2 mm) lies well above that of the thicker (1.2 mm) sample, except above 4 pm wavelength where the curves cross. This is in accordance with theory as described in the earlier paper.“‘)
TRANSIENT
AND
FREQUENCY
RESPONSE
RESULTS
Photoconductive decay The form of the photoconductive decay at 77 K, following spark source excitation, is shown for one of the samples by the oscilloscope trace in the inset to Fig. 8. Readings from this trace are shown as a semilogarithmic plot in the main part of the figure. The decay clearly does not correspond to a simple exponential but an approximate fit to the experimental results can be made with the sum of two exponentials of the form I~xp( - t/r/) + Z,exp( - t/r,). The solid curve in Fig. 8 represents such a fit. Table 1 lists the parameters obtained by fitting the decay curves of four of the samples. It is noted that the fast decay time constant t/ is about an order of magnitude smaller than the slower time constant t, and that generally I, is greater than I,. During one of the decay measurements, the sample was also illuminated with unchopped white background light, which reduced I, but left Z, essentially unaffected.
1. SHIH and C.
316
500
10;
H. CHAMPNESS
200
100 (K)
CZ-78-11-11
-0 /' Y. -
A=3.5pm / C-m
/
/
_-m -
O2 1
6
I
G; b" H
lOO_ 2
I
4
’
1
I
I
6
8
10
O0
I
12
14
lOOO/T,( K-l) Fig. 3. Relative photoconductivity, Au/(aoE,), and average time constant T at 3.5 pm plotted against the reciprocal of absolute temperature for sample CZ-78-11-11 with E!Ic. The extrinsic region extends up to about 180 K.
This suggests that rS may arise from trapping. Vi@ reported that the decay time decreased with increase of background light, presumably because of the decreased contribution of T,. The effect of annealing on the decay characteristic. is not entirely clear but for the thin samples, Table 1 shows that both decay times 7/ and r, were increased by the thermal treatment.
CZ-78-11-10
2
4
6
810
20
Irradiance.E,( pW/cm2)
I
40
Fig. 4. Relative photoconductivity change, Au/u& plotted against irradiance at 3.8 )rm. It is seen that the photoconductivity increases linearly with the irradiance.
Photoconductivity in tellurium at 77 K
I - 0 CZ-78-11-AN
10”
1
2 Wavelength,
I
317
I
3 4 A , ( pm 1
Fig. 5. Relative photon photoconductivity at 77 K plotted against wavelength for three samples prepared by different treatments with E!jc and Elc. Samples CZ-7%ll-AN annealcd; -11 unanncaled: -AB abrasively polished.
Frequency response
The variation of the normalized photoresponse with chopping frequency measured on the two thin samples is shown in Fig. 9. The signal is seen to decrease by some 7% from the low frequency value to that at 900 Hz for the unannealed sample and by about 207; for the annealed sample. It is easily shown that if the photoconductive decay can be described by a simple exponential decrease with a single time constant r, then the change in the r.m.s. voltage u(j) of the fundamental of the signal, with chopped light, is given by = (1 + 02r2)l/2. Here o is the angular frequency, and v. is the value of ~t,f) uoluo when wr q 1. An attempt to fit this relation to the experimental points in Fig. 9 is shown by the two solid lines. The r values for these curves are 60 and 170psec. which are respectively larger than the T, values of 30 and 130 PC for these two samples, obtained by photoconductive decay. This difference may be due to the decay having a more complex time dependence than that expressible as a two-time constant function. In any case, it may be said that the decrease with frequency only starts to become important when f exceeds l/z,. CALCULATED
LOW
FREQUENCY
RESPONSE
Using equation (1) of reference 10, values of Ao/(aoE,) were calculated with parameters appropriate to the unannealed samples 11 and 11-T, on the one hand, and the annealed samples 1l-AN and 1l-ANT on the other. The variation is shown by the
1. SHIHand C. H. CHAMPNESS
318
0.15mm
I-I/b 0’ 0_ 0 z
5
-
10-18,
T
?$-
m Elc
\I 10 1: _
0 CZ-78-11-ANT
i$I a : q
0
'; a' i? 10-19-
4
A CZ-78-11-T -20, 10 1
4 a :
77 K I
I
I
_
2 3 4 Wavelength,h,( pm1
Fig. 6. Relative photon photoconductivity at 77 K vs wavelength for two thinner samples with (CZ-78-1 I-ANT) and without (-11-T) annealing.
solid lines in Fig. 10 where Ao/(a,,E,) is plotted against sample thickness for the case of Ellc at a wavelength of A = 2 pm. The parameters diffusion coefficient D and extrinsic hole concentration p. were .obtained from transport measurements on the samples at 77 K and the absorption coefficient K value was taken from the work of Tutihasi et al.” ‘) The bulk lifetime r-values used correspond to the fast photoconductive time constant TV, since this quantity made the larger contribution to the photoconductive decay. The s-values were chosen to fit the experimental points for the four samples indicated. Values of s larger by an order of magnitude would have been needed if 7, had been used instead of 7f. The difference in parameters of the calculated curves needed to fit the experimental points, indicates that the increase in photo-response after annealing is not due to a single cause, but to the combined effect of increased mobility and lifetime, and decreased surface recombination velocity. Continuation of the calculated curves to smaller thicknesses (not shown) indicates a maximum photoresponse at a sample thickness of about 6~. EXCESS
NOISE
AND
DETECTIVITY
RESULTS
Measurements of the noise from the sample with the current on and off showed that most of it arises from the passage of the current. Therefore, it is appropriate, following Vis,@)to use an excess noise parameter defined as ( V2 - V:)/( V:w), where V and VT are respectively the noise voltages measured with and without sample current, Vdc is the steady voltage between the measuring probes with current flowing, and Af is the band-
Photoconductivity
319
in tellurium at 77 K
Elc
d
O.?W
5
z 0
‘4 i
-18 10
a’ C 0
% <
10-19
5 az
A CZ-78-11-T
\
o CZ-78-11-11
:: \ a
77 K 10ZC 1
2 3 Wave1ength.h ,(pm)
Fig. 7. Relative photon photoconductivity
,’ .
4
at 77 K vs wavelength for two samples of different thickness.
1
s
aO’ 0.6 2 0
0.4
-t/4.5+ 385
1=3.95e
.
$21
z 0.2
I
I
10
20 Time,
I
30
I
I
40
50
60
t ,( psec)
Fig. 8. Relative magnitude of the photoconductive transient decay following spark source excitation for sample CZ-78-1 l-l 1 at 77 K plotted against time. The solid curve is the result of fitting with the twetime constant equation shown.
320
I. SHIH and C. H.
1.2
I
I
CHAMPNESS
I
I
1
-
\\
1.0
. CZ-78-11-ANT 77 K
I t=l70 psec
1 +4T?f2r2;“2
-t 0.2
10'
10"
lo4
Frequency , f , ( l-12 1 Fig. 9. Relative photoconductive voltage. [tf),!ro. plotted against frequency. The solid lines rep resent the variation
of (I + tuzrz)~
’’
withffitted
to the experimental
points.
width of the lock-in amplifier-detector. A plot of this excess noise parameter against frequency f on log-log scales is shown in Fig. 11 for two of the samples at 77 K; the results of Vis’6’ are also indicated for comparison. The fall-off with frequency is strong and in one case is steeper than l/t
I
I
I
I
I
1
c= 8 psec D= 125cm2/sec S= 1600 cm/set
D- 52 cm&ec K=
lOOOOcm-'
s = 3000
cm/set
0.154~10'4cm-3
P=
0.' Sample Thickness,d ,(cm) Fig. IO. Calculated variation of photon photoconductivity whh sample thickness using equation (I) of reference IO. The parameters were chosen to correspond to the plotted experimental points of the two annealed and two unannealed samples.
321
Photoconductivity in tellurium at 77 K
-13
10 -
II
I
II
m
I
I
CZ-78-11-T
A CZ-78-11-ANT
Frequency., f,,(Hz) Fig. 1 I. Normalized.noise parameter, (v’ - I+)/( V&f-), plotted against frequency for samples CZ-78-1 I-T and -ANT. together with a curve of Vis. (6).
The normalized spectral detectivity is defined as D* = (A Af)“‘/(NEI’). where A is the illuminated area and NEP is the noise equivalent power of the detector. From the noise voltage and photoconductive responsivity at 900 Hz the D* values were calculated between 1 and 4.2 pm for the two thinner samples 11-T and 1l-ANT, and are plotted against wavelength in Fig. 12. The broken line is a curve for a melt-grown sample of Edwards et a/.,‘4’ which had a maximum D* of 6.4 x 10” cm Hz”*W- ‘. The results for annealed sample ll-ANT are, however, about twice as high as this, with a maximum value near i. = 3.5 pm of 1.3 x 10” cm Hz I’* W- ’ (see Table 1, column 9). This value is in fact the highest D*-value yet reported at 77 K for melt-grown tellurium. Even higher values should be possible for thinner samples and the solid curve shown is an estimate of what should be possible. Here the peak value is that limited by radiation from a 300K background with a field of view of 180’. DISCUSSION
AND
CONCLUSIONS
In the present work it is very clear that even fine abrasive surface polishing of the tellurium surfaces greatly decreases the photoconductive response at 77 K, the effect being particularly large at shorter wavelengths below 2 pm. Conversely, annealing results in increased photoconductivity. This is apparently not due to one cause but to the combined effects of increased mobility, increased lifetime and decreased surface recombination velocity. For tellurium, intrinsic at room temperature, the photoconductivity increases by more than an order of magnitude with decrease of temperature down to 77 K. Most of the change takes place in the extrinsic range and reflects the increase in the decay time constant with decreasing temperature. With spark source illumination, the photoconductive decay appears to involve a fast decay process with a time constant (T/) of a few microseconds, together with a slower one with a time constant (T,) of a few tens of microseconds. This non-exponential decay does not arise from a high level of excess
322
1. SHIHand C.
H. CHAMPNESS
Theoretical Limit \
CZ-78-11-ANT CZ-78-11-T 77K Ellc FOV=180°
1' 4
t i
‘\ L
Te -100-C, Edwards I 1
i et al, ref. t 41’ I I
2 3 Wave1ength.A
4 ,( pm)
5
Fig. 12. Normalized spectral detectivity D* plotted against wavelength for samples CZ-78-11-T and -ANT at 77 K with El~c.The broken curve is from reference 4 on melt-grown Tc. while the solid line is a theoretical curve for background limited conditions in the present case.
carriers, because Aa/ao, resulting from the spark source excitation, amounts only to about 0.001 or less. The fact that the magnitude of the larger time constant contribution (I,) is reduced with steady white background light. suggests that it arises from traps. In this case, the fast decay contribution may correspond to bulk recombination. It is found that the excess current noise parameter falls off with frequency at least as fast as l/c On the other hand, the decrease of responsivity with frequency is much smaller, so that the signal-to-noise ratio improves with increase of frequency. It may be noted that, while annealing increases the responsivity, it also increases the fall-off with frequency. At 900 Hz and a 1 Hz bandwidth, a D* value at i = 3.5 pm has been obtained which is the highest yet reported for melt-grown tellurium. A value up to the background limited detectivity should be possible with an annealed sample having a thickness in the range 1 to 100 pm. Thus, with sufficient effort it would appear that melt-grown tellurium could be shown to be of interest for use as an infrared detector material operating at 77 K near a wavelength of 3.5 pm for application at least up to 1 kHz. However, it could still not compete with InSb at higher frequencies and longer wavelengths. Acknow/edgentenr-The authors wish to acknowledge the support of this work by the National Sciences Engineering Research Council of Canada.
REFERENCES 1. BARTLETT R. S.. Phys. Rec.. 26. 247 (1925). 2. Moss T. S.. Proc. Phrs. Sot.. A62. 264 (1949). 3. LOFERSKI J. J.. Phys. ‘Rec.. 93. 707 (1954). 4. EDWARDS D. F.. C. D. BUTTER & L. D. MCGLAUCHLIN. Sol.-S&r. Elecrron. 3. 5. VIS V. A.. J. Appl. Phys.. 35. 360 (1964). 6. Vrs V. A., J. Appl. Phjx. 35. 365 (1964). 7. Glossy P. & K. WINZIB. Phys. sm. Solidi. 13.269 (l%6). 8. SHIH1. & C. H. CHAMPNES$ J. Crysral Growth, 44.492 (1978).
24 (l%l).
9. EL-AZABM., C. R. MCLAUGHLIN & C. H. CHAMPNES.S. J. Crysrdl Growth, 2& I (1975). 10. SHYAMPRASAD N. G.. C. H. CHAMPNE~~ & 1. SHIH.I#-ared Phys., 21.45 (1981). Il. TU~IHASIS.. G. G. ROBERTR. C. KEEZER & R. E. DREW&Ph~s. Rec., 177. 1143 (1969)
and
PHOTOCONDUCTIVITY
IN TELLURIUM
AT 77 K
I. SHIH and C. H. CHAMPNESS Department of Electrical Engineering. 3480 University Street, Montreal, P. Q.. Canada H3A 2A7 (Received 9 April
1981)
Ahafract-Photoconductivity has been measured in monocrystalline samples of undoped tellurium at 77 K in the wavelength range 1.4 to 4.4 pm. The samples were chemically cut from a Czochralski-grown ingot. The photoresponse decreased by about an order of magnitude as the temperature was increased from 77 to 180 K in the extrinsic range, where holes dominate. At 77 K, the photoresponse was found to be sensitive to sample treatment; it was increased by annealing (at 395°C for 168 hr) and decreased by abrasive polishing. The photoconductive decay, following spark source excitation, was approximately describable in terms of a short time constant I,. measured in microseconds, and a longer time constant T,. measured in tens of microseconds.The ac. photoresponse was found to decrease with frequencyf: particularly asjexceeded I/T, Measurements were made of current noise and spectral detectivity D*. At 77 K, a D.-value of 1.3 x 10” cm Hz”’ W- ’ was observed at i = 3.5 pm on an annealed sample of thickness about 0.2 mm. This appears to be the highest D* yet reported on melt-grown tellurium. An even higher D*. up to the background limited value at I = 3.5 pm, should be possible with further reduction of sample thickness.
INTRODUCTION
Photoconductivity has been known in tellurium for more than half a century. In 1925, Bartlett(‘) first showed that the effect in thin films was larger at liquid oxygen temperature than at room temperature. Some years later Moss (‘) found the photoconductivity maximum to lie at a wavelength near 1.3 pm for cooled thin tellurium films evaporated on glass substrates. However, the first results on bulk crystal samples were reported by Loferskit3’ in 1954. These samples were obtained by cleaving a Bridgman-grown ingot containing randomly oriented crystals, followed by abrasive grinding and polishing with final chemical etching. Edwards et ~1.‘~’published results, in 1961, on samples prepared by vapour phase growth and by cleaving a Czochralski-grown crystal. On one of their vapour-grown samples they reported a detectivity at 77 K, and at a wavelength of 3.4 pm which was essentially background limited. Vis(5) also studied photoconductivity on bulk samples, which were prepared by abrasive cutting from slices cleaved from tellurium Czochralski-grown crystals, followed by chemical etching. In addition he studied the dependence of noise on frequency and temperature. N) In 1966 Grosse and Winzer”’ made a detailed study of photoconductivity in bulk samples, cut’abrasively from Czochralski crystals followed by etching. They observed the decrease in response with decrease of wavelength below 3.5 pm and found this to be more rapid for the electric vector E of the radiation perpendicular to the c-axis of the sample (i.e. Elc) than for that with E parallel to it (i.e. Ellc). In all the above mentioned previously reported work, abrasive action was used in the preparation of the samples and, while chemical etching was often carried out, lattice damage in the material would not have been completely removed. In recent work, improvements have been made in the Czochralski crystal growth method,‘*) and in the development of techniques for ensuring good sample surfaces”’ of tellurium. Because of these material improvements, and an increased awareness of the sensitivity of tellurium to imperfections arising from improper handling, it was considered timely to make a preliminary reassessment of the potential of tellurium as an infrared detector material operating near a wavelength of 3.5 pm. The first part of such a study was reported on in a recent paper, (‘O’where the variation of the photoresponse and the shift of its peak with change of sample thickness was investigated. The second part of the study constitutes the present paper, where the effect of temperature variation and heat treatment are specifically examined and where the results of measurements of noise and responsivity as a 311
312
I. SHIH and C. H. CHAMPNESS
function of frequency, are reported on. In this work all abrasive treatments were avoided in the preparation of the tellurium samples, except in a deliberate experimental case.
EXPERIMENTAL
PROCEDURE
Crystal growth
The samples for study were obtained from a single monocrystalline ingot of tellurium No. CZ-78-11. This crystal was grown by the Czochralski method in our laboratory, according to the procedure described. (s) Briefly, it was grown at a vertical pull rate of 1 cm/hr in the c-direction from high purity material (6-9’s) obtained from Cominco Ltd. The background temperature during growth was 360°C. The ingot was 8 cm in length and 1.6 cm in diameter, and its high quality was established from etching, Laue diffraction patterns and from Hall mobility measurements on samples at 77 K. Sample preparation
The samples for photoconductivity measurements were chemically cut, initially as wafers, from the tellurium ingot with a string saw using a solution consisting of HCl:CrOs: Hz0 in the ratio 1: 1:2 by weight. The wafers were then mounted on aluminium blocks and chemically polished successively on two parallel (lOi0) faces, using a solution of composition HNOs:CrOJ:H20 in the ratio 1:2:4 by weight. The final dimensions were about 8 mm in length (parallel to c-direction), 2 mm in width and the thicknesses from 0.15 to 2.2 mm, as indicated in Table 1, column 2. After the chemical polishing, one of the samples (CZ-78-1 l-AB) was also given an abrasive polish on the two parallel (lOi0) surfaces, using 0.05 m diameter A120J powder on glass for 2 min. Two other samples were chemically cut from an annealed slice. This annealing was done by placing the tellurium slice in a quartz ampoule, which was then evacuated and back-filled with argon at 300 torr pressure and sealed off. The ampoule was then maintained at 395°C for about 170 hr. After slow cooling to room temperature over a period of about 50 hr, the slice was taken out and the samples (78-ll-AN and 78-ll-ANT) were chemically cut from it. Three side-probes for potential measurements were attached to each sample (Fig. 1) by melting in 0.002 in. diameter platinum wires. The ends of the samples were coated with a solder of composition 50% Sn, 47% Bi and 3% Sb by weight. A fine copper wire was attached to one end and the other was soldered to a brass plate, the central part of which was covered with an insulator, blackened to minimize reflection. A thick rectangular paper window served to screen light away from the end contacts, as indicated by the broken line in Fig. 1. Method of measurement The sample holder was mounted inside a dewar with a ZnS (Kodak IRTRAN 2) window. Current from a battery pack was supplied to the sample via a large series resistor. Infrared radiation from a Perkin-Elmer model 13 monochromator was focused on to the sample by an external mirror. The light was chopped at 13 Hz and the monochromatic light intensity was measured with a built-in vacuum thermocouple detector. The relation, between the output of this detector and the incident light energy falling on the sample, was determined by placing a calibrated InSb photovoltaic detector operating at 77 K at the position of the sample and determining its output. The 13 Hz signal from the side-probes of the sample was fed to a lock-in amplifier (Princeton Applied Research model 124A) set at the same reference frequency. The steady current through the sample was set at the maximum value consistent with linearity of a plot of photoconductive signal voltage versus sample current. The steady voltage between the probes was measured with a digital voltmeter. Photoconductive decay was also measured by recording oscillographically the decrease of photoconductivity, following illumination of the sample by a pulse of light. Initially,
CZ-18-I I-ANT
0.2
CZ-18-I I-T
,._
I.25
I .os
2.28
I.51
CZ-18-I I-AB
.~
--
polished None 395’c l68hr
None 395c 168 hr None, Abrasively
1.15
2.2
1.16
Thermal treatment
Width (mm)
CZ-18.1 I-AN
cz-18-I l-l I
Sampk no.
Thickness (dimension parallel to incident light) (mm)
Table 1. Photoconductivity
-_--
2.0
0.5
2.2
0.6
30 130
4
-
10
(L
IO
-
5
(Z)
Photoconductive decay at 77 K
2
I
1.1
I,lL
1.3 x IO”
3.7 X IO’O
1.8 x IO9
Maximum LP at 77 K 9OOHz IHzBW (cm Hz 112W-l)
. I-L-LII--^.“.-“~~-ll__l_l..-“~r^_,-“.----_I_.C-.._.- .-._ l”“-_______,
Max. Aua/&,E,) at 77 K (IO~‘Scm’sec)
results in undoped tellurium samples
5 % =: r:
5. tz
ii=
f 0. 6 t 5. Q 8’
?
I. SHIH and C.
314
H. CHAMPNESS
Fig. 1. Schematic view of the sample with the current leads and the side probes for photoconductivity measurements.
this was done with 3-pm radiation from the monochromator using chopped white light from the Nernst source. Later the source was changed to a spark unit (Xenon Corp. Nanopulser model 437A plus a 2.5 m band-pass filter), where the pulse duration was less than 100 nsec. While the intensity from this source was much greater than that from the Nernst glower, the largest change of conductivity (Aa,+rO)was still less than 0.001. The average noise voltage was also measured between the side probes on three of the samples at 77 K. This was done with the lock-in amplifier at different frequencies and a fixed band width. The noise levels were determined with and without bias current. LOW FREQUENCY
PHOTOCONDUCTIVITY
RESULTS
Variurion wifh remperature The spectral variation of Ao/(aE,) with wavelengths between about 1.4 and 4.4 ,um, measured at three different temperatures for Ellc and ELc, is shown in Fig. 2. Here Ao is photo~onducti~ty increase, no is the dark conductivity and Ep is the photon flux incident on the sample, It is noted that with increase of temperature the photo-response decreases and for Elc the peak values shift towards shorter wavelengths and the peak-to-“plateau” ratios decrease. The variation of Au/(aoE,J at 3.5 pm over a wider temperature range is shown in Fig. 3. The change near 77 K is seen to be relatively flat with a stronger decrease as the temperature is raised. The rapid decrease above about 180 K (i.e. = 5.5) occurs in the intrinsic region, where the Hall coefficient has a negative l~/r sign. Within the extrinsic region (up to 180 K) the photoconductivity changes by about an order of magnitude. The average photocondu~tive decay time T, also plotted in Fig. 3, changes by a similar amount, suggesting that this quantity predominantly controls the temperature variation of photoconductivity. The relative change of conductivity &/a0 at 77 K is plotted against irradiance E at a wavelength of 3.8 pm in Fig. 4. It is seen that up to an E value of 20 ~W/cm’ at least. the variation is linear. This is consistent with the results of Vis”) according to whom non linearity only occurs above Au/a0 value of about 0.01. Eject o~s~mpie tr~#t~n~ and thickness The spectral photoconductive response at 77 K of three samples of similar thickness is shown in Fig. 5, where each has had a different final preparational treatment; one was chemically polished, another annealed and then chemically polished and the third chemically polished and finally abrasively polished as described earlier (p. 2). It is noted that the photoconductivity is much higher for the annealed sample and that the fall-off of photoconductivity with decreasing wavelength, for the abrasively prepared sample, is much stronger than the other two. At shorter wavelengths the difference between the curves for the abrasive and annealed samples amounts almost to two orders of magnitude.
315
Photoconductivity in tellurium at 77 K
I
lt?’ ;
-18 10 _
I
I
I
0°K “lOOK o 130K
"E " 2
3 a
9 9 \ 9
I?CZ-78-11-11
1
8:
3 4 2 Wave1ength.A ,( pm 1
Fig. 2. Relative photon photoconductivity. Au/(aOEJ, vs wavelength for sample CZ-78-I I-l 1 at three different temperatures with EI’C and Elc.
The effect of annealing is also shown for the two thinner samples in Fig. 6. Here the annealed sample ll-ANT has about three times the responsivity of the unannealed sample 11-T at all wavelengths. The effect of sample thickness is evident from the results on the two non-annealed samples in Fig. 7. Here the photo-response for the thinner sample (0.2 mm) lies well above that of the thicker (1.2 mm) sample, except above 4 pm wavelength where the curves cross. This is in accordance with theory as described in the earlier paper.“‘)
TRANSIENT
AND
FREQUENCY
RESPONSE
RESULTS
Photoconductive decay The form of the photoconductive decay at 77 K, following spark source excitation, is shown for one of the samples by the oscilloscope trace in the inset to Fig. 8. Readings from this trace are shown as a semilogarithmic plot in the main part of the figure. The decay clearly does not correspond to a simple exponential but an approximate fit to the experimental results can be made with the sum of two exponentials of the form I~xp( - t/r/) + Z,exp( - t/r,). The solid curve in Fig. 8 represents such a fit. Table 1 lists the parameters obtained by fitting the decay curves of four of the samples. It is noted that the fast decay time constant t/ is about an order of magnitude smaller than the slower time constant t, and that generally I, is greater than I,. During one of the decay measurements, the sample was also illuminated with unchopped white background light, which reduced I, but left Z, essentially unaffected.
1. SHIH and C.
316
500
10;
H. CHAMPNESS
200
100 (K)
CZ-78-11-11
-0 /' Y. -
A=3.5pm / C-m
/
/
_-m -
O2 1
6
I
G; b" H
lOO_ 2
I
4
’
1
I
I
6
8
10
O0
I
12
14
lOOO/T,( K-l) Fig. 3. Relative photoconductivity, Au/(aoE,), and average time constant T at 3.5 pm plotted against the reciprocal of absolute temperature for sample CZ-78-11-11 with E!Ic. The extrinsic region extends up to about 180 K.
This suggests that rS may arise from trapping. Vi@ reported that the decay time decreased with increase of background light, presumably because of the decreased contribution of T,. The effect of annealing on the decay characteristic. is not entirely clear but for the thin samples, Table 1 shows that both decay times 7/ and r, were increased by the thermal treatment.
CZ-78-11-10
2
4
6
810
20
Irradiance.E,( pW/cm2)
I
40
Fig. 4. Relative photoconductivity change, Au/u& plotted against irradiance at 3.8 )rm. It is seen that the photoconductivity increases linearly with the irradiance.
Photoconductivity in tellurium at 77 K
I - 0 CZ-78-11-AN
10”
1
2 Wavelength,
I
317
I
3 4 A , ( pm 1
Fig. 5. Relative photon photoconductivity at 77 K plotted against wavelength for three samples prepared by different treatments with E!jc and Elc. Samples CZ-7%ll-AN annealcd; -11 unanncaled: -AB abrasively polished.
Frequency response
The variation of the normalized photoresponse with chopping frequency measured on the two thin samples is shown in Fig. 9. The signal is seen to decrease by some 7% from the low frequency value to that at 900 Hz for the unannealed sample and by about 207; for the annealed sample. It is easily shown that if the photoconductive decay can be described by a simple exponential decrease with a single time constant r, then the change in the r.m.s. voltage u(j) of the fundamental of the signal, with chopped light, is given by = (1 + 02r2)l/2. Here o is the angular frequency, and v. is the value of ~t,f) uoluo when wr q 1. An attempt to fit this relation to the experimental points in Fig. 9 is shown by the two solid lines. The r values for these curves are 60 and 170psec. which are respectively larger than the T, values of 30 and 130 PC for these two samples, obtained by photoconductive decay. This difference may be due to the decay having a more complex time dependence than that expressible as a two-time constant function. In any case, it may be said that the decrease with frequency only starts to become important when f exceeds l/z,. CALCULATED
LOW
FREQUENCY
RESPONSE
Using equation (1) of reference 10, values of Ao/(aoE,) were calculated with parameters appropriate to the unannealed samples 11 and 11-T, on the one hand, and the annealed samples 1l-AN and 1l-ANT on the other. The variation is shown by the
1. SHIHand C. H. CHAMPNESS
318
0.15mm
I-I/b 0’ 0_ 0 z
5
-
10-18,
T
?$-
m Elc
\I 10 1: _
0 CZ-78-11-ANT
i$I a : q
0
'; a' i? 10-19-
4
A CZ-78-11-T -20, 10 1
4 a :
77 K I
I
I
_
2 3 4 Wavelength,h,( pm1
Fig. 6. Relative photon photoconductivity at 77 K vs wavelength for two thinner samples with (CZ-78-1 I-ANT) and without (-11-T) annealing.
solid lines in Fig. 10 where Ao/(a,,E,) is plotted against sample thickness for the case of Ellc at a wavelength of A = 2 pm. The parameters diffusion coefficient D and extrinsic hole concentration p. were .obtained from transport measurements on the samples at 77 K and the absorption coefficient K value was taken from the work of Tutihasi et al.” ‘) The bulk lifetime r-values used correspond to the fast photoconductive time constant TV, since this quantity made the larger contribution to the photoconductive decay. The s-values were chosen to fit the experimental points for the four samples indicated. Values of s larger by an order of magnitude would have been needed if 7, had been used instead of 7f. The difference in parameters of the calculated curves needed to fit the experimental points, indicates that the increase in photo-response after annealing is not due to a single cause, but to the combined effect of increased mobility and lifetime, and decreased surface recombination velocity. Continuation of the calculated curves to smaller thicknesses (not shown) indicates a maximum photoresponse at a sample thickness of about 6~. EXCESS
NOISE
AND
DETECTIVITY
RESULTS
Measurements of the noise from the sample with the current on and off showed that most of it arises from the passage of the current. Therefore, it is appropriate, following Vis,@)to use an excess noise parameter defined as ( V2 - V:)/( V:w), where V and VT are respectively the noise voltages measured with and without sample current, Vdc is the steady voltage between the measuring probes with current flowing, and Af is the band-
Photoconductivity
319
in tellurium at 77 K
Elc
d
O.?W
5
z 0
‘4 i
-18 10
a’ C 0
% <
10-19
5 az
A CZ-78-11-T
\
o CZ-78-11-11
:: \ a
77 K 10ZC 1
2 3 Wave1ength.h ,(pm)
Fig. 7. Relative photon photoconductivity
,’ .
4
at 77 K vs wavelength for two samples of different thickness.
1
s
aO’ 0.6 2 0
0.4
-t/4.5+ 385
1=3.95e
.
$21
z 0.2
I
I
10
20 Time,
I
30
I
I
40
50
60
t ,( psec)
Fig. 8. Relative magnitude of the photoconductive transient decay following spark source excitation for sample CZ-78-1 l-l 1 at 77 K plotted against time. The solid curve is the result of fitting with the twetime constant equation shown.
320
I. SHIH and C. H.
1.2
I
I
CHAMPNESS
I
I
1
-
\\
1.0
. CZ-78-11-ANT 77 K
I t=l70 psec
1 +4T?f2r2;“2
-t 0.2
10'
10"
lo4
Frequency , f , ( l-12 1 Fig. 9. Relative photoconductive voltage. [tf),!ro. plotted against frequency. The solid lines rep resent the variation
of (I + tuzrz)~
’’
withffitted
to the experimental
points.
width of the lock-in amplifier-detector. A plot of this excess noise parameter against frequency f on log-log scales is shown in Fig. 11 for two of the samples at 77 K; the results of Vis’6’ are also indicated for comparison. The fall-off with frequency is strong and in one case is steeper than l/t
I
I
I
I
I
1
c= 8 psec D= 125cm2/sec S= 1600 cm/set
D- 52 cm&ec K=
lOOOOcm-'
s = 3000
cm/set
0.154~10'4cm-3
P=
0.' Sample Thickness,d ,(cm) Fig. IO. Calculated variation of photon photoconductivity whh sample thickness using equation (I) of reference IO. The parameters were chosen to correspond to the plotted experimental points of the two annealed and two unannealed samples.
321
Photoconductivity in tellurium at 77 K
-13
10 -
II
I
II
m
I
I
CZ-78-11-T
A CZ-78-11-ANT
Frequency., f,,(Hz) Fig. 1 I. Normalized.noise parameter, (v’ - I+)/( V&f-), plotted against frequency for samples CZ-78-1 I-T and -ANT. together with a curve of Vis. (6).
The normalized spectral detectivity is defined as D* = (A Af)“‘/(NEI’). where A is the illuminated area and NEP is the noise equivalent power of the detector. From the noise voltage and photoconductive responsivity at 900 Hz the D* values were calculated between 1 and 4.2 pm for the two thinner samples 11-T and 1l-ANT, and are plotted against wavelength in Fig. 12. The broken line is a curve for a melt-grown sample of Edwards et a/.,‘4’ which had a maximum D* of 6.4 x 10” cm Hz”*W- ‘. The results for annealed sample ll-ANT are, however, about twice as high as this, with a maximum value near i. = 3.5 pm of 1.3 x 10” cm Hz I’* W- ’ (see Table 1, column 9). This value is in fact the highest D*-value yet reported at 77 K for melt-grown tellurium. Even higher values should be possible for thinner samples and the solid curve shown is an estimate of what should be possible. Here the peak value is that limited by radiation from a 300K background with a field of view of 180’. DISCUSSION
AND
CONCLUSIONS
In the present work it is very clear that even fine abrasive surface polishing of the tellurium surfaces greatly decreases the photoconductive response at 77 K, the effect being particularly large at shorter wavelengths below 2 pm. Conversely, annealing results in increased photoconductivity. This is apparently not due to one cause but to the combined effects of increased mobility, increased lifetime and decreased surface recombination velocity. For tellurium, intrinsic at room temperature, the photoconductivity increases by more than an order of magnitude with decrease of temperature down to 77 K. Most of the change takes place in the extrinsic range and reflects the increase in the decay time constant with decreasing temperature. With spark source illumination, the photoconductive decay appears to involve a fast decay process with a time constant (T/) of a few microseconds, together with a slower one with a time constant (T,) of a few tens of microseconds. This non-exponential decay does not arise from a high level of excess
322
1. SHIHand C.
H. CHAMPNESS
Theoretical Limit \
CZ-78-11-ANT CZ-78-11-T 77K Ellc FOV=180°
1' 4
t i
‘\ L
Te -100-C, Edwards I 1
i et al, ref. t 41’ I I
2 3 Wave1ength.A
4 ,( pm)
5
Fig. 12. Normalized spectral detectivity D* plotted against wavelength for samples CZ-78-11-T and -ANT at 77 K with El~c.The broken curve is from reference 4 on melt-grown Tc. while the solid line is a theoretical curve for background limited conditions in the present case.
carriers, because Aa/ao, resulting from the spark source excitation, amounts only to about 0.001 or less. The fact that the magnitude of the larger time constant contribution (I,) is reduced with steady white background light. suggests that it arises from traps. In this case, the fast decay contribution may correspond to bulk recombination. It is found that the excess current noise parameter falls off with frequency at least as fast as l/c On the other hand, the decrease of responsivity with frequency is much smaller, so that the signal-to-noise ratio improves with increase of frequency. It may be noted that, while annealing increases the responsivity, it also increases the fall-off with frequency. At 900 Hz and a 1 Hz bandwidth, a D* value at i = 3.5 pm has been obtained which is the highest yet reported for melt-grown tellurium. A value up to the background limited detectivity should be possible with an annealed sample having a thickness in the range 1 to 100 pm. Thus, with sufficient effort it would appear that melt-grown tellurium could be shown to be of interest for use as an infrared detector material operating at 77 K near a wavelength of 3.5 pm for application at least up to 1 kHz. However, it could still not compete with InSb at higher frequencies and longer wavelengths. Acknow/edgentenr-The authors wish to acknowledge the support of this work by the National Sciences Engineering Research Council of Canada.
REFERENCES 1. BARTLETT R. S.. Phys. Rec.. 26. 247 (1925). 2. Moss T. S.. Proc. Phrs. Sot.. A62. 264 (1949). 3. LOFERSKI J. J.. Phys. ‘Rec.. 93. 707 (1954). 4. EDWARDS D. F.. C. D. BUTTER & L. D. MCGLAUCHLIN. Sol.-S&r. Elecrron. 3. 5. VIS V. A.. J. Appl. Phys.. 35. 360 (1964). 6. Vrs V. A., J. Appl. Phjx. 35. 365 (1964). 7. Glossy P. & K. WINZIB. Phys. sm. Solidi. 13.269 (l%6). 8. SHIH1. & C. H. CHAMPNES$ J. Crysral Growth, 44.492 (1978).
24 (l%l).
9. EL-AZABM., C. R. MCLAUGHLIN & C. H. CHAMPNES.S. J. Crysrdl Growth, 2& I (1975). 10. SHYAMPRASAD N. G.. C. H. CHAMPNE~~ & 1. SHIH.I#-ared Phys., 21.45 (1981). Il. TU~IHASIS.. G. G. ROBERTR. C. KEEZER & R. E. DREW&Ph~s. Rec., 177. 1143 (1969)
and