- Email: [email protected]
Time-resolved photoluminescence in amorphous P2Se
~
Solld State Communications, Vol. 69, No. 2, pp.163-166, 1989. ~rlnted in Great Britain.
0038-1098/89 $3.00 + .00 Pergamon Press plc
TIME-RESOLVED PHOTOLUMINESCENCE IN AMORPHOUS P2Se D. Wolverson and R.T. Phillips Department of Physics, University of Exeter, Exeter EX4 4QL, Great Britain , (Received
13 May 1988 by
G.
Lonzarich )
The first measurements of time-resolved photolnminescence in g-P2Se are presented, and reveal a mid-gap Gaussian emission band typical of a chalcogenide glass, together with high-energy lnmlnescence extending at least up to 1.95eV, which is attributed to band-to-band recombination. Results are also given for luminescence in crystalline P4Se3, and comparisons are made between glass and crystal phosphorus selenides, and between these and published data on g-As,2S3. It is seen that neither glJ:~s nor crystal show a significant time-resolved shift of the mid-gap PL band, implying a similan'ty between the PL processes involved in the two materials.
has any consequences for the defects believed to be involved in PL, since the more familiar(As) chalcogenides are thought to have a much more homogeneous microstructure.W e present here the firstmeasurements of time-resolved PL in g-P2Se, which is a stable and reproducible glass of the P-rich glass forming region, whose basic physical properties are well describedby Borisova (6).The dependence of the PL spectrum on excitationenergy is discussed, and the PL is compared to thatof g-As2S 3 and c-P4Sc3.
Inl~luction Experimental studies of photoluminescence (PL) in amorphous semiconductors generally reveal broad Gaussian bands, of full width at half m a x i m u m ( F W H M ) of around 0.2eV, peaking at about half the band-gap energy (I).The excitation spectra do not in general have a simple shape, but commonly peak at the energy where the absorption coefficient is approximately 102 era- 1(2). Time-resolved measurements after pulsed excitation reveal a wide range of lifetimes, from nanoseconds to milliseconds (3), whilst the shift in energy of the emission band with time shows a connection between the lifetime and the energy of the states between which radiative recombination occurs. Similarities in the energy range and shape of the PL emission spectra of glassy and crystalline ehalcogenides have been taken to indicate that similar radiative recombination mechanisms are involved (4), though excitation spectra and time dependences are in general different. Recent work has pointed to differences between the recombination processes for glassy and crystalline As2Se3 on the basis of the proportion of geminate recombination (5). The mechanisms proposed include recombination of distant carriers trapped in band-tail states, possibly with spatial fluctuations in the band-gap energy, recombination of carriers or excitons trapped at coordination defects or valence alternation pairs, or of self-trapped exeitons. Models based on intrinsic defect centres are widely applied to the glasses, and the large Stokes shift of the luminescence is then explained by analogy with that of colour centres in the alkali halides, as being due to structural rearrangement of the defect site after optical transitions. It is of interest, therefore, to study new chalcogenidc systems, in an effort to test the generality of the PL phenomena mentioned, and to attempt to distinguish between the proposed models. Systems which offer the possibility of comparison between glassy and crystalline materials are particularly desirable; the phosphorus selenides are useful in this respect since glasses PxSel.x are easily prepared for 0
Exverirnental technioues - Samples were synthesised by fusion of crushed red a-P (suppliers M.C.P., purity 99.999%) at a temperature of 600°C, just above the melting point of red a-P (we confirmed that our red a-P melted at 590°C). Two batches were prepared, in fused silica ampoules which were baked under vacuum to -10 -6 torr at 200°C before being filled in a dry N 2 atmosphere, and re-evacuated. Slow quenching of the melt was used, to reduce the cracking of the glass under thermal stresses, and so no fictive temperature can be specified. Both batches produced initially clear transparent red glass, and the second batch remained so. The first batch, which was cooled at a faster rate, showed some spontaneous crystallization of c-P4Se3, over a few days at room temperature. X-ray powder photographs of several samples of the second batch were taken to check that no crystallization had occurred; only two broad rings due to the glass were observed. A large piece from the second batch (10 x 4 x 4ram) was polished on one face with a diamond paste and was mounted in a continuous-flow He cryostat; all measurements were made at 4.2K. An Nd-YAG pumped dye laser was used for excitation, providing tunable light pulses of 12ns duration. Intensities of around 200ttj were used, focussed onto an area of about 3ram diameter. The PL was dispersed by a single-grating spectrometer of 9nm bandpass, and detected using cooled S 1 or $20 photomultipliers. Edge filters were used to block the excitation light, and for order-so~Jng, and an IR polariser was used to remove the effect of the grating anomaly. The signals were measured with a boxcar averager using a 2ns gate width. The spectra shown here are normalised to the peak intensity, after fitting with Gaussian lineshapes where appropriate, and after correcting for the spectral response of the complete detection system. Spectra are plotted in arbitrary units proportional to the number of photons per unit energy bandwidth. Where spectra obtained with the two detectors are overlaid, a scaling factor was determined from the zero-delay 163
164
TIME-RESOLVED PHOTOLUMINESCENCE
spectra SO as to give a smooth join, and the same factor was applied to spectra at longer delays. The agreement between the spectra in the region of overlap gives confidence in the accuracy of the response correction procedure.
IN AMORPHOUS P2Se 28
Vol. 69, No. 2
arb. units
24 Results ,~ delay, ns Figure I shows spectra obtained with an excitation energy 20 of 2.175eV, and gate delays from zero to 50ns, where zero delay implies measurement during the excitation pulse. The noise in the data results in a quite broad scatter of peak 16 200 positions of the fits to the data. The excitation energy used corresponds to the peak of the cw PL excitation spectrum, ~, 100 which appears to show a slight sensitivityto the quenching rate. 12 The cw PL band, excited at this energy, lies at about I.IeV. The firstcomment to make is that for delays greater than 30ns, ~ h ~'~" 40 the spectra at lower energies arc reasonably well described by a 8 \ ~ ~ ' - - ao Gaussian shape. The F W H M is of the order of 0.2cV, a typical value for a chalcogcnide glass. The peak of the spectrum for the ,~ ~ ~'~ 20 second batch is at 1.23eV :i: 10meV. Though no accurate \h ~ 10 measurements have been made of the absorption edge of this ~ ~,~o material (10),by the fact that it is red and transparent, one can 0 I ~.~ o , estimate its optical gap, for instm~c¢, E04, to lie higher than 1"2 ' 1"4 1"6 1.8 2.2cV. The cw PL excitation spectrum suggests that the e¥ absorption coefficient is about I02cm -I at 2.2eV. The c w PL band thus conforms to the 'mid-gap' rule, and the zero-delay band is some 130meV higher in energy. N o significant shiftof Figure 2. PL spectra of c-P4Se 3 excited at 2.175eV, normalised to the band occurs from 10ns to 100ns, when the scatterin the fits the sarnc peak height, and verticallydisplaced by 2 units. is taken into account. This suggests that recombination mechanisms involving a Coulomb interaction term in the P L carrier trapped at a single defect centre, would involve no energy are unlikely. These include recombination between a Coulomb shift,and is a possible mechanism, ie: distant pair of carders, both trapped at defects, ie, the reaction: e" + (h+ + D ' ) --+ D" + h ~ , L 2D°--->D + + D- + hDpL The second feature of interest is that at delays less than in the notation of Mott, Davis, and Street (11), for which a shift 50ns, PL is sccn at higher energies, extending up to the upper to lower energies with time of more than 100meV is anticipated, limit of the detection system ( ~1.95eV, determined by the since closer pairs, with stronger Coulomb interaction in the cutoff of the edge filter used to block the excitation beam). No final state, have a greater rate of recombination. Recombination clear peak is sccn in the higher energy PL excited at 2.175cV; between a free carder or a carder trapped in a band-tail, and a there is a slight maximum in the signal at about 1.75eV, but that is all. Because of the poor stray light rejection of the single-grating spectrometer, it was not possible to dispense arb. units with the edge filter,and so extend the energy range towards the /4 . . . . . . . . illumination energy any further. The room temperature PL spectrum, measured with a more sensitive, slow detection 12 ~ delay, ns system, showed only a very weak band at about 1.SeV, which one can ascribe to hot carrierPL, as observed in g-As2S 3 (12). The time-dependence of this band could not be studied easily. 10 The absolute quantum efficienciesfor the room tcmperatu~ and 4 K PL signals at this energy were not measured, but that at ,~ AA 295K was considerably smallc~, hot carder P L may be expe_~_ed C to have a small quantum efficiency, of order 10-1a-~,or :; \~ tt ,% thereabouts,(12) so that the low temperature, high-energy signal 0 is much larger than would be expected if it were due to hot carder PL. 10 The simplest explanation for the higher energy PL is that it is due to crystalline contamination of the sample, probably by c-P4Se 3, as was known to exist in the firstbatch of material. 2O One might expect PL of a shorter lifetime, and higher energy, from band-to-band recombination in a crystalline material. 30Therefore, a pure polycrystalline sample of c-P4Se 3 was
4i
50"
prepared in the cc phase (13), by sublimation of c-P4Se 3 in vacuum, the starting material lacing synthesized from the elements as described for the glasses. PL spectra w c m recorded under the same conditions as for the glass, and are shown in Figure 2. It is clear that the high energy PL is not due to Figure 1. PL spectra of g-P2Sc excited at 2.175eV, normalised to contamination of the glass; in the crystallinematerial, no signal the sarnc height and vertically shifted by 1.5 units for clarity. The was observed between 1.95eV and the usual mid-gap band large ticks indicate the baseline for each spectrum. Circles represent near 1.2eV. Also, the high energy PL observed in the second, the $20 dam, and triangles the $1 data. Two zero delay spectra arc apparently homogeneous, batch of g-P2Se was similar to tl~nthe firstbatch. A PL mechanism intrinsic to the glass, and not shown. ~
|
|
"1.3
i
=
.
.
"1"6
.
.
.
.
1-9
2"2 eV
TIME-RESOLVED
Vol. 69, No. 2
PHOTOLUMINESCENCE
involving any Stokes shift is therefore suggested, which competes with the Stokes-shifted, lower-energy emission process. The former could be due to geminate recombination between carriers trapped in band-tail states, or to self-trapped excitons, as discussed in reference (14), in which it was concluded, on the basis of the polarization memory, that the similar high-energy PL in g-As2S 3 was due to self-trapped excitons. We next consider the dependence of the spectra of the glass on excitation energy. Figure 3 shows spectra at four delays in g-P2Se, with excitation energy 2.43eV. Some points should be noted; the spectra are extremely broad (FWHM .q3.4eV at zero delay), high in energy, and shift downward in energy rapidly; also, the PL does not extend to such high energies as for the lower excitation energy. These spectra are consistent with the idea that the higher energy emission now dominates the spectrum, possibly with a contribution from the lower band. The downward shift of the combined band could then be partly due to the shorter lifetime of the higher energy process. In other words, the h~gher energy illumination preferentially excites the band-edge or excitonic luminescence. If this is the case, one must explain why the 2.43eV excitation generates band-edge PL at a lower mean energy than the 2.175eV illumination, around 1.6 to 1.7eV, rather than 1.8eV and above. It may be that the lower-energy light predominantly excites the defects, whilst the higher-energy light excites band-edge states; this distinction would be accentuated if band-edge carriers do not trap rapidly into defect states. This situation is represented in Figure 4. Alternatively, carriers excited at higher energies may be more mobile, and hence may sample lower energy recombination centres. The fast downward shift in energy of the high energy PL band cannot be accounted for completely by a thermalization model, assuming successive phonon emission in an exponential band tail. In such a model, the luminescence energy E at time t obeys: E = E o - 3 k T o In ( I n ( "Oot )) where TOis the band-tail characteristic temperature, and ~0 is a phonon frequency (15). The observed shift is too large, and occurs over too long a time to be fitted by any reasonable parameters; most of the energy loss in a thermallzation model 10 ar.b. u n i t s . . . . . . . . .
o~
~
delay, ns
/:,.,\\ 8
J teQro oOo%° Zo ~ x~.
o° °c
•e
o
oo. 2
°8
Ooo
"~.
:
", ,~
•
0'
• 1
-
o
• "~
% '~30
so
. . . . . . 1.3
1-6
1-9
2"2 eV
Figure 3. PL spectra of g-P2Se excited at 2.43eV, normalised to the same peak height and vertically displaced by 1.5 units.
IN AMORPHOUS
165
P2Se
E
g(E)
cd
a
b
f
\ Figure 4. Possible excitation and radiative transitions in g-P2Se. The arrows represent: (a) direct excitation of defects; (b) excitation from band-edge; (c) and (d) band-to-band recombination with emission of PL. The shaded areas represent the photoexcited carrier densities. occurs within lns. However, several factors probably combine to produce the observed shift, and thermalization may be one. Some possible excitation and emission processes are summarised in Figure 5, where the D O to D- transition is arbitrarily chosen as the low energy radiative transition, and in which band-edge, rather than excitonic, high-energy luminescence is represented. A comparison of these results with published data on g-As2S 3 is extremely interesting. A great deal of work has been done on many aspects of PL in this material, with detailed arguments being developed to explain the results.Time-resolved PL phenomena have proved sensitive to the excitation energy(3), sample stoichiometry(16), and also to the accuracy of the detection system response correction(17), so that quite different behaviour has been reported by various groups. However, it was agreed that there appear to be two types of mid-gap PL process in g-As2S 3 at short times, of which the higher energy process has a shorter lifetime(18). It was proposed to shift with the excitation energy, whereas the lower energy process was not (16). In the view of Higashi and Kastner (19), these two processes arose from a single band of PL centres. The situation in g-P2Se is simpler; only one PL band is seen in the mid-gap region, which does not shift in energy over 0-100ns. This band can be ascribed to conventionaldefect PL, as discussed above. The higher-energy PL is shorter-lived than this band, and shows an almost fiat spectrum. In g-As2S 3, the high energy PL was measured in coincidence with the pulsed illumination, as it decayed within the excitation pulse duration of 10ns. It was found to be insensitive to temperature up to 200K. Initially, PL in this region of the spectrum was atmbuted to band-to-band recombination (20), though a later study of the same spectral region interpreted it as hot carrier PL from self-trapped excitons, on the basis of the polarization memory(14). In g-P2Se, the lifetime of the high energy PL is longer, as the data here show, it can be detected even at SOns. One would expect a lifetime for hot carrier PL of the order of the relaxation time of the excited state, around lps, which
166
TIME-RESOLVED PHOTOLUMINESCENCE IN AMORPHOUS P2Se
supports the interpretation of high energy PL in g-As2S 3 as hot PL. On the other hand, in g-P2Se, the lifetime of the high energy PL seems too long for this explanation to hold; here, band-to-band recombination appears more plausible. The apparent absence of high energy (but sub-band gap) PL from crystalline P4Se 3 is then easily explained, since only the glass would be expected to possess a wide band-tail from which band-to-band recombination might occur. Conclusions The fh'st measurements of time-resolved PL in g-P2Se and c-P4Se 3 are presented, and demonstrate the presence of a typical chalcogenide luminescence band of about mid-gap energy and Gaussian shape. The crystal and glass show similar PL spectra in this respect, as is observed in other chalcogenides. Neither shows a significant shift of this band
Vol. 69, No. 2
over lOOns after excitation, suggesting a similarity in the PL process.It seems, then, that the unusual mierostructure expected for g-P2Se still permits defects of the kind often proposed for amorphous chalcogertides. A less commonly observed feature of broad, short-lived high energy emission is also seen in the glass, but not in the crystal, over the accessible energy range. This is here attributed to band-edge luminescence, and is reminiscent of PL observed in g-As2S 3 under similar conditions. Direct recombination of carders trapped in band-tail states has been studied relatively little in chalgogenide glasses, and so the appearance of this process in a new material is of interest for further work. Acknowledgements We thank the SERC for supporting this work, and one of us (DW) acknowledges the SERC for a research studentship during part of this work.
References (1) R.A. Street, Adv. in Phys. 25, 397, (1976) (2) J. Robertson, Phys. Chem. Glasses, 23, 1, (1982) (3) G.S. Higashi and M.A. Kasmer, Phil. Mag. B, 47, 83, (1983) (4) S.P. Depinna, B.C. Cavenett, and W.E. Lamb, Phil. Mag. B, 47, 99, (1983) (5) J. Ristein and G. Weiser, Sol. St. Commun., 57, 639, (1986) (6) Z.U. Borisova, "Glassy Semiconductors", Plenum (New York), (1981) (7) A.A. Babaev, J. Appl. Spectrosc., 33, 937, (1980) (8) G.R. Burns, J. Phys. Chem. Sols., 47, 681, (1986) (9) D.L. Price, M. Mlsawa, S. Susman, T.I. Morrison, G.K. Shenoy, and M.Grimsditch, J. Non-Crystalline Sols., 66, 443, (1984) (10) Simple preliminary measurements of ours give an estimate of E02 for g-P2Se of 2.26eV, and approximately 2.5eV for E04.
(11) N.F. Mott, E.A. Davis, and R.A. Street, Phil. Mag., 32, 961, (1975 (12) K. Murayama, J. Non-Crystalline Sols., 97 & 98, 1147, (1987) (13) Y. Monteil and H. Vincent, Can. J. Chem., 52, 2190, (1974) (14) K. Murayama and M.A. B6sch, Phys. Rev. B, 25, 6542, (1982) (15) M.A. Kastner, J. Non-Crystalline Sols., 35 & 36, 807, (1980) (16) K. Murayama, G.S. I-hgashi and M.A. Kasmer, Phil. Mag. B, 48, 277, (1983) (17) F. Mollot, Sol. St. Commun., 43, 641, (1982) (18) M.A. BOsch and J. Shah, Phys. Rev. Letts., 42, 118, (1979) (19) G.S.Higashi and M.A.Kasmer, Phys. Rev. Letts., 47, 124, (1981) (20) J. Shah and M.A. B6sch, Phys. Rev. Letts., 42, 1420, (1979)
Solld State Communications, Vol. 69, No. 2, pp.163-166, 1989. ~rlnted in Great Britain.
0038-1098/89 $3.00 + .00 Pergamon Press plc
TIME-RESOLVED PHOTOLUMINESCENCE IN AMORPHOUS P2Se D. Wolverson and R.T. Phillips Department of Physics, University of Exeter, Exeter EX4 4QL, Great Britain , (Received
13 May 1988 by
G.
Lonzarich )
The first measurements of time-resolved photolnminescence in g-P2Se are presented, and reveal a mid-gap Gaussian emission band typical of a chalcogenide glass, together with high-energy lnmlnescence extending at least up to 1.95eV, which is attributed to band-to-band recombination. Results are also given for luminescence in crystalline P4Se3, and comparisons are made between glass and crystal phosphorus selenides, and between these and published data on g-As,2S3. It is seen that neither glJ:~s nor crystal show a significant time-resolved shift of the mid-gap PL band, implying a similan'ty between the PL processes involved in the two materials.
has any consequences for the defects believed to be involved in PL, since the more familiar(As) chalcogenides are thought to have a much more homogeneous microstructure.W e present here the firstmeasurements of time-resolved PL in g-P2Se, which is a stable and reproducible glass of the P-rich glass forming region, whose basic physical properties are well describedby Borisova (6).The dependence of the PL spectrum on excitationenergy is discussed, and the PL is compared to thatof g-As2S 3 and c-P4Sc3.
Inl~luction Experimental studies of photoluminescence (PL) in amorphous semiconductors generally reveal broad Gaussian bands, of full width at half m a x i m u m ( F W H M ) of around 0.2eV, peaking at about half the band-gap energy (I).The excitation spectra do not in general have a simple shape, but commonly peak at the energy where the absorption coefficient is approximately 102 era- 1(2). Time-resolved measurements after pulsed excitation reveal a wide range of lifetimes, from nanoseconds to milliseconds (3), whilst the shift in energy of the emission band with time shows a connection between the lifetime and the energy of the states between which radiative recombination occurs. Similarities in the energy range and shape of the PL emission spectra of glassy and crystalline ehalcogenides have been taken to indicate that similar radiative recombination mechanisms are involved (4), though excitation spectra and time dependences are in general different. Recent work has pointed to differences between the recombination processes for glassy and crystalline As2Se3 on the basis of the proportion of geminate recombination (5). The mechanisms proposed include recombination of distant carriers trapped in band-tail states, possibly with spatial fluctuations in the band-gap energy, recombination of carriers or excitons trapped at coordination defects or valence alternation pairs, or of self-trapped exeitons. Models based on intrinsic defect centres are widely applied to the glasses, and the large Stokes shift of the luminescence is then explained by analogy with that of colour centres in the alkali halides, as being due to structural rearrangement of the defect site after optical transitions. It is of interest, therefore, to study new chalcogenidc systems, in an effort to test the generality of the PL phenomena mentioned, and to attempt to distinguish between the proposed models. Systems which offer the possibility of comparison between glassy and crystalline materials are particularly desirable; the phosphorus selenides are useful in this respect since glasses PxSel.x are easily prepared for 0
Exverirnental technioues - Samples were synthesised by fusion of crushed red a-P (suppliers M.C.P., purity 99.999%) at a temperature of 600°C, just above the melting point of red a-P (we confirmed that our red a-P melted at 590°C). Two batches were prepared, in fused silica ampoules which were baked under vacuum to -10 -6 torr at 200°C before being filled in a dry N 2 atmosphere, and re-evacuated. Slow quenching of the melt was used, to reduce the cracking of the glass under thermal stresses, and so no fictive temperature can be specified. Both batches produced initially clear transparent red glass, and the second batch remained so. The first batch, which was cooled at a faster rate, showed some spontaneous crystallization of c-P4Se3, over a few days at room temperature. X-ray powder photographs of several samples of the second batch were taken to check that no crystallization had occurred; only two broad rings due to the glass were observed. A large piece from the second batch (10 x 4 x 4ram) was polished on one face with a diamond paste and was mounted in a continuous-flow He cryostat; all measurements were made at 4.2K. An Nd-YAG pumped dye laser was used for excitation, providing tunable light pulses of 12ns duration. Intensities of around 200ttj were used, focussed onto an area of about 3ram diameter. The PL was dispersed by a single-grating spectrometer of 9nm bandpass, and detected using cooled S 1 or $20 photomultipliers. Edge filters were used to block the excitation light, and for order-so~Jng, and an IR polariser was used to remove the effect of the grating anomaly. The signals were measured with a boxcar averager using a 2ns gate width. The spectra shown here are normalised to the peak intensity, after fitting with Gaussian lineshapes where appropriate, and after correcting for the spectral response of the complete detection system. Spectra are plotted in arbitrary units proportional to the number of photons per unit energy bandwidth. Where spectra obtained with the two detectors are overlaid, a scaling factor was determined from the zero-delay 163
164
TIME-RESOLVED PHOTOLUMINESCENCE
spectra SO as to give a smooth join, and the same factor was applied to spectra at longer delays. The agreement between the spectra in the region of overlap gives confidence in the accuracy of the response correction procedure.
IN AMORPHOUS P2Se 28
Vol. 69, No. 2
arb. units
24 Results ,~ delay, ns Figure I shows spectra obtained with an excitation energy 20 of 2.175eV, and gate delays from zero to 50ns, where zero delay implies measurement during the excitation pulse. The noise in the data results in a quite broad scatter of peak 16 200 positions of the fits to the data. The excitation energy used corresponds to the peak of the cw PL excitation spectrum, ~, 100 which appears to show a slight sensitivityto the quenching rate. 12 The cw PL band, excited at this energy, lies at about I.IeV. The firstcomment to make is that for delays greater than 30ns, ~ h ~'~" 40 the spectra at lower energies arc reasonably well described by a 8 \ ~ ~ ' - - ao Gaussian shape. The F W H M is of the order of 0.2cV, a typical value for a chalcogcnide glass. The peak of the spectrum for the ,~ ~ ~'~ 20 second batch is at 1.23eV :i: 10meV. Though no accurate \h ~ 10 measurements have been made of the absorption edge of this ~ ~,~o material (10),by the fact that it is red and transparent, one can 0 I ~.~ o , estimate its optical gap, for instm~c¢, E04, to lie higher than 1"2 ' 1"4 1"6 1.8 2.2cV. The cw PL excitation spectrum suggests that the e¥ absorption coefficient is about I02cm -I at 2.2eV. The c w PL band thus conforms to the 'mid-gap' rule, and the zero-delay band is some 130meV higher in energy. N o significant shiftof Figure 2. PL spectra of c-P4Se 3 excited at 2.175eV, normalised to the band occurs from 10ns to 100ns, when the scatterin the fits the sarnc peak height, and verticallydisplaced by 2 units. is taken into account. This suggests that recombination mechanisms involving a Coulomb interaction term in the P L carrier trapped at a single defect centre, would involve no energy are unlikely. These include recombination between a Coulomb shift,and is a possible mechanism, ie: distant pair of carders, both trapped at defects, ie, the reaction: e" + (h+ + D ' ) --+ D" + h ~ , L 2D°--->D + + D- + hDpL The second feature of interest is that at delays less than in the notation of Mott, Davis, and Street (11), for which a shift 50ns, PL is sccn at higher energies, extending up to the upper to lower energies with time of more than 100meV is anticipated, limit of the detection system ( ~1.95eV, determined by the since closer pairs, with stronger Coulomb interaction in the cutoff of the edge filter used to block the excitation beam). No final state, have a greater rate of recombination. Recombination clear peak is sccn in the higher energy PL excited at 2.175cV; between a free carder or a carder trapped in a band-tail, and a there is a slight maximum in the signal at about 1.75eV, but that is all. Because of the poor stray light rejection of the single-grating spectrometer, it was not possible to dispense arb. units with the edge filter,and so extend the energy range towards the /4 . . . . . . . . illumination energy any further. The room temperature PL spectrum, measured with a more sensitive, slow detection 12 ~ delay, ns system, showed only a very weak band at about 1.SeV, which one can ascribe to hot carrierPL, as observed in g-As2S 3 (12). The time-dependence of this band could not be studied easily. 10 The absolute quantum efficienciesfor the room tcmperatu~ and 4 K PL signals at this energy were not measured, but that at ,~ AA 295K was considerably smallc~, hot carder P L may be expe_~_ed C to have a small quantum efficiency, of order 10-1a-~,or :; \~ tt ,% thereabouts,(12) so that the low temperature, high-energy signal 0 is much larger than would be expected if it were due to hot carder PL. 10 The simplest explanation for the higher energy PL is that it is due to crystalline contamination of the sample, probably by c-P4Se 3, as was known to exist in the firstbatch of material. 2O One might expect PL of a shorter lifetime, and higher energy, from band-to-band recombination in a crystalline material. 30Therefore, a pure polycrystalline sample of c-P4Se 3 was
4i
50"
prepared in the cc phase (13), by sublimation of c-P4Se 3 in vacuum, the starting material lacing synthesized from the elements as described for the glasses. PL spectra w c m recorded under the same conditions as for the glass, and are shown in Figure 2. It is clear that the high energy PL is not due to Figure 1. PL spectra of g-P2Sc excited at 2.175eV, normalised to contamination of the glass; in the crystallinematerial, no signal the sarnc height and vertically shifted by 1.5 units for clarity. The was observed between 1.95eV and the usual mid-gap band large ticks indicate the baseline for each spectrum. Circles represent near 1.2eV. Also, the high energy PL observed in the second, the $20 dam, and triangles the $1 data. Two zero delay spectra arc apparently homogeneous, batch of g-P2Se was similar to tl~nthe firstbatch. A PL mechanism intrinsic to the glass, and not shown. ~
|
|
"1.3
i
=
.
.
"1"6
.
.
.
.
1-9
2"2 eV
TIME-RESOLVED
Vol. 69, No. 2
PHOTOLUMINESCENCE
involving any Stokes shift is therefore suggested, which competes with the Stokes-shifted, lower-energy emission process. The former could be due to geminate recombination between carriers trapped in band-tail states, or to self-trapped excitons, as discussed in reference (14), in which it was concluded, on the basis of the polarization memory, that the similar high-energy PL in g-As2S 3 was due to self-trapped excitons. We next consider the dependence of the spectra of the glass on excitation energy. Figure 3 shows spectra at four delays in g-P2Se, with excitation energy 2.43eV. Some points should be noted; the spectra are extremely broad (FWHM .q3.4eV at zero delay), high in energy, and shift downward in energy rapidly; also, the PL does not extend to such high energies as for the lower excitation energy. These spectra are consistent with the idea that the higher energy emission now dominates the spectrum, possibly with a contribution from the lower band. The downward shift of the combined band could then be partly due to the shorter lifetime of the higher energy process. In other words, the h~gher energy illumination preferentially excites the band-edge or excitonic luminescence. If this is the case, one must explain why the 2.43eV excitation generates band-edge PL at a lower mean energy than the 2.175eV illumination, around 1.6 to 1.7eV, rather than 1.8eV and above. It may be that the lower-energy light predominantly excites the defects, whilst the higher-energy light excites band-edge states; this distinction would be accentuated if band-edge carriers do not trap rapidly into defect states. This situation is represented in Figure 4. Alternatively, carriers excited at higher energies may be more mobile, and hence may sample lower energy recombination centres. The fast downward shift in energy of the high energy PL band cannot be accounted for completely by a thermalization model, assuming successive phonon emission in an exponential band tail. In such a model, the luminescence energy E at time t obeys: E = E o - 3 k T o In ( I n ( "Oot )) where TOis the band-tail characteristic temperature, and ~0 is a phonon frequency (15). The observed shift is too large, and occurs over too long a time to be fitted by any reasonable parameters; most of the energy loss in a thermallzation model 10 ar.b. u n i t s . . . . . . . . .
o~
~
delay, ns
/:,.,\\ 8
J teQro oOo%° Zo ~ x~.
o° °c
•e
o
oo. 2
°8
Ooo
"~.
:
", ,~
•
0'
• 1
-
o
• "~
% '~30
so
. . . . . . 1.3
1-6
1-9
2"2 eV
Figure 3. PL spectra of g-P2Se excited at 2.43eV, normalised to the same peak height and vertically displaced by 1.5 units.
IN AMORPHOUS
165
P2Se
E
g(E)
cd
a
b
f
\ Figure 4. Possible excitation and radiative transitions in g-P2Se. The arrows represent: (a) direct excitation of defects; (b) excitation from band-edge; (c) and (d) band-to-band recombination with emission of PL. The shaded areas represent the photoexcited carrier densities. occurs within lns. However, several factors probably combine to produce the observed shift, and thermalization may be one. Some possible excitation and emission processes are summarised in Figure 5, where the D O to D- transition is arbitrarily chosen as the low energy radiative transition, and in which band-edge, rather than excitonic, high-energy luminescence is represented. A comparison of these results with published data on g-As2S 3 is extremely interesting. A great deal of work has been done on many aspects of PL in this material, with detailed arguments being developed to explain the results.Time-resolved PL phenomena have proved sensitive to the excitation energy(3), sample stoichiometry(16), and also to the accuracy of the detection system response correction(17), so that quite different behaviour has been reported by various groups. However, it was agreed that there appear to be two types of mid-gap PL process in g-As2S 3 at short times, of which the higher energy process has a shorter lifetime(18). It was proposed to shift with the excitation energy, whereas the lower energy process was not (16). In the view of Higashi and Kastner (19), these two processes arose from a single band of PL centres. The situation in g-P2Se is simpler; only one PL band is seen in the mid-gap region, which does not shift in energy over 0-100ns. This band can be ascribed to conventionaldefect PL, as discussed above. The higher-energy PL is shorter-lived than this band, and shows an almost fiat spectrum. In g-As2S 3, the high energy PL was measured in coincidence with the pulsed illumination, as it decayed within the excitation pulse duration of 10ns. It was found to be insensitive to temperature up to 200K. Initially, PL in this region of the spectrum was atmbuted to band-to-band recombination (20), though a later study of the same spectral region interpreted it as hot carrier PL from self-trapped excitons, on the basis of the polarization memory(14). In g-P2Se, the lifetime of the high energy PL is longer, as the data here show, it can be detected even at SOns. One would expect a lifetime for hot carrier PL of the order of the relaxation time of the excited state, around lps, which
166
TIME-RESOLVED PHOTOLUMINESCENCE IN AMORPHOUS P2Se
supports the interpretation of high energy PL in g-As2S 3 as hot PL. On the other hand, in g-P2Se, the lifetime of the high energy PL seems too long for this explanation to hold; here, band-to-band recombination appears more plausible. The apparent absence of high energy (but sub-band gap) PL from crystalline P4Se 3 is then easily explained, since only the glass would be expected to possess a wide band-tail from which band-to-band recombination might occur. Conclusions The fh'st measurements of time-resolved PL in g-P2Se and c-P4Se 3 are presented, and demonstrate the presence of a typical chalcogenide luminescence band of about mid-gap energy and Gaussian shape. The crystal and glass show similar PL spectra in this respect, as is observed in other chalcogenides. Neither shows a significant shift of this band
Vol. 69, No. 2
over lOOns after excitation, suggesting a similarity in the PL process.It seems, then, that the unusual mierostructure expected for g-P2Se still permits defects of the kind often proposed for amorphous chalcogertides. A less commonly observed feature of broad, short-lived high energy emission is also seen in the glass, but not in the crystal, over the accessible energy range. This is here attributed to band-edge luminescence, and is reminiscent of PL observed in g-As2S 3 under similar conditions. Direct recombination of carders trapped in band-tail states has been studied relatively little in chalgogenide glasses, and so the appearance of this process in a new material is of interest for further work. Acknowledgements We thank the SERC for supporting this work, and one of us (DW) acknowledges the SERC for a research studentship during part of this work.
References (1) R.A. Street, Adv. in Phys. 25, 397, (1976) (2) J. Robertson, Phys. Chem. Glasses, 23, 1, (1982) (3) G.S. Higashi and M.A. Kasmer, Phil. Mag. B, 47, 83, (1983) (4) S.P. Depinna, B.C. Cavenett, and W.E. Lamb, Phil. Mag. B, 47, 99, (1983) (5) J. Ristein and G. Weiser, Sol. St. Commun., 57, 639, (1986) (6) Z.U. Borisova, "Glassy Semiconductors", Plenum (New York), (1981) (7) A.A. Babaev, J. Appl. Spectrosc., 33, 937, (1980) (8) G.R. Burns, J. Phys. Chem. Sols., 47, 681, (1986) (9) D.L. Price, M. Mlsawa, S. Susman, T.I. Morrison, G.K. Shenoy, and M.Grimsditch, J. Non-Crystalline Sols., 66, 443, (1984) (10) Simple preliminary measurements of ours give an estimate of E02 for g-P2Se of 2.26eV, and approximately 2.5eV for E04.
(11) N.F. Mott, E.A. Davis, and R.A. Street, Phil. Mag., 32, 961, (1975 (12) K. Murayama, J. Non-Crystalline Sols., 97 & 98, 1147, (1987) (13) Y. Monteil and H. Vincent, Can. J. Chem., 52, 2190, (1974) (14) K. Murayama and M.A. B6sch, Phys. Rev. B, 25, 6542, (1982) (15) M.A. Kastner, J. Non-Crystalline Sols., 35 & 36, 807, (1980) (16) K. Murayama, G.S. I-hgashi and M.A. Kasmer, Phil. Mag. B, 48, 277, (1983) (17) F. Mollot, Sol. St. Commun., 43, 641, (1982) (18) M.A. BOsch and J. Shah, Phys. Rev. Letts., 42, 118, (1979) (19) G.S.Higashi and M.A.Kasmer, Phys. Rev. Letts., 47, 124, (1981) (20) J. Shah and M.A. B6sch, Phys. Rev. Letts., 42, 1420, (1979)