Photoluminescence dynamics in chalcogenide glasses and crystals

Photoluminescence dynamics in chalcogenide glasses and crystals

Journal of Non-Crystalline Solids 40 (1980) 587-594 © North-Holland Publishing Company PHOTOLUMINESCENCE DYNAMICS IN CHALCOGENIDE GLASSES AND CRYSTAL...

414KB Sizes 0 Downloads 10 Views

Journal of Non-Crystalline Solids 40 (1980) 587-594 © North-Holland Publishing Company

PHOTOLUMINESCENCE DYNAMICS IN CHALCOGENIDE GLASSES AND CRYSTALS M.A. BOSCH, R.W. EPWORTH Bell Laboratories, Holmdel, NJ 07733, USA and D. EMIN * Sandia Laboratories t, Albuquerque, NM 87185, USA

Time-resolved photoluminescence studies reveal distinct differences between the recombination processes in a chalcogenide glass and in its crystalline counterpart. Here the three luminescence bands of a-As2S3 are interpreted in terms of the recombination of an excition, a selftrapped exciton, and a pair of electron- and hole-like small polarons. The two luminescence bands observed in the crystal are attributed to the recombination of two types of excitons composed respectively of a hole bound to a self-trapped electron, and a hole which is induced to self-trap in the presence of a self-trapped electron.

1. Introduction Steady state photoluminescence measurements on amorphous and crystalline arsenic chalcogenides revealed no drastic differences between the spectra o f the two states [1,2]. These photoluminescence (PL) spectra are characterized b y a broad peak at half the band gap energy. The half band gap PL has been attributed to the existence o f localized states in the gap either charged [3] or neutral [4] or due to the formation o f small polaron states [5]. It is however difficult to find experimental evidence which would enable one to distinguish between the proposed models. Time resolved PL measurements possess the virtue o f yielding a more detailed understanding o f the carrier kinetics in b o t h amorphous and crystalline chalcogenides. They even enable us to distinguish clearly between the amorphous and the crystalline state. The transient PL behavior in a-As2 $3 revealed the existence o f several recombination processes [6,7]. The dynamics o f these processes are manifested in the temporal dependence of the intensities and spectral positions o f the PL bands as well as their temperature and excitation energy dependence. Similarly * Supported by the US Department of Energy, under Contract DE-AC04-76-DP00789. "["USA Department of Energy facility. 587

588

M.A. Bosch et al. / PL dynamics

the temporal behavior of the PL from crystalline (c) As2S3 exhibits new detailed information on the recombination kinetics of a charge pair which was generated by the absorption of a photon. The two luminescence bands observed in the crystal are attributed to the recombination of two types of excitons composed of either a hole bound to a self-trapped electron or a hole which was induced to self-trap by the presence of the self-trapped electron. The delay of the decay of the latter (lowest energy) PL band is attributable to the presence of a barrier associated with the deep polaronic trapping of the hole at the self-trapped electron. Consistent with the small polaron picture is the observation that the lowest energy PL band of the a-As2S3 shifts to lower energy with time while that of the crystal does not.

2. Experimental Several different samples of glassy and crystalline arsenic trisulfide have been measured. The glass was prepared by either mixing stochiometric amounts of the elements (6N) or using the high purity compound (5N) and melting the batch in an evacuated and sealed silica tube which was kept for at least 12 h in a rocking-furnace at 750°C. Air or forced-air quenching was found to be satisfactory for this easy glass-forming system. Large bulk samples of a-As2S3 were purchased from Servo Corporation; Orpiment single crystals which were mined in Peru or in Nevada [8] have been selected due to their exceptional quality. The samples were illuminated by a train of 11 ns Ar*-laser pulses separated by 30 #s with an average intensity of 1 W cm -2. The emitted radiation was detected during a time window comparable with the laser pulse width by means of a coincidence technique. Coincidence gating (3 ns electronic gate time capability) allowed us to select any delay time z after the laser pulse with cable delay lines for a delay of up to 64 ns or with a digital delay generator for longer delay times. The photon counter and the step motor driven double grating monochromator were controlled by an on-line computer. This arrangement was extremely helpful since the spectra were immediately corrected for the system response.

3. Experimental results In fig. 1 we show the PL spectra of a-As2S3 for steady state [2] and pulse observation. The laser pulse width and gate time were approximately I 1 ns with no delay of the gate with respect to the laser pulse. The three processes measured under pulse-observation have different temporal decay characteristics. The peak at 5700 A decays faster than the laser pulse time. The peak at 7800 A has a decay time of tens of nanoseconds whereas the peak at 9000 A reveals a long decay time but more important, this process shifts with time towards the steady state spectrum

[6].

589

M.A. Bosch et al. / PL dynamics 2.2 2.0 "1,8 4.6 40

i

i

i

4.4

i

2.2 2.0 t.8

-1.2 (eV)

I

i

a-AS2oS3 ~

tO

/"~

1

I

4.6

4.4

4.2

I

i

I

(eV) .....

c- As 2 S3 4880/~

8

I

,-4

SSt-Obs~

// XlOO

6

z tlJ

~ 4 z

/

_J o.

/

6

/

x ~450 \

4

11 ns Pulse-Obs

2~y S

\

2

SSt-Obs I

6000

.L

I

i

0

I

8000 40000 WAVELENGTH (A)

`12000

t

6000

I

I

i

I

8000 iO000 WAVELENGTH (A)

~

"12000

Fig. 1. Steady state and time resolved photoluminescence spectra of amorphous arsenic trisulfide. Fig. 2. Steady state and time resolved photoluminescence spectra of orpiment.

The spectral dependence of the PL from c-As2Sa is given in fig. 2 for the steady state [2] and also for pulse observation. In contrast to a-As2Sa the single crystalline Orpiment samples exhibit only two radiative recombination processes. The peak positions for the PL processes at 7600 A varies slightly (200 A) for crystals with different origins. (It should be noted that Kolomiets et al. [1] found the steady state peak position at 9800 A). The PL intensity of the process at 5600 A is larger for the Nevada Orpiment * than for the crystal from Peru shown in fig. 2. The temporal behavior of the PL processes of c-As2Sa is shown in fig. 3. The radiative recombination process at higher energy decays rapidly in time. The low energy process manifests a very different temporal behavior. First the intensity increases with delay time and after a comparatively long time (200 ns) the process decays. This delay time characteristics is displayed in fig. 4 for the peak wavelength (~ ~ 7600 A). Data points are given for two Orpiment crystals, both from Peru. Fig. 4 indicates that the population of the luminescence centres has to be built up and decays slowly afterwards. Furthermore no spectral shift with time occurs for this process in contrast to the low energy PL in the glass which experiences a very large spectral shift towards lower energy with increasing delay time.

* One Orpiment (Nevada) sample was kindly supplied to us by R. ZaUen [10].

M.A. Bosch et al. / PL dynamics

590 2.4

2.2

2.0

t.8

t .6

[

i

i

i

i

, c - A s 2 S 3 ' PERU" 2K

(eV)

r"\170ns / \ /20ns

\

,88o ; . n ,

// OnsX\

vo ),,. t-

i

i

i

i

1

i

i

i

i

]

i

c-As2 S 3

o lo

i

i

i

"PERU"

v

z w iz / 0.

i/;,,,ooo.,,~

DELAY:

z I.u ~ 4 z

[I ;

Ons io.,

// /';;);ns/ I// / 5000

i

Oi

6 oo

~

~2 i

70oo

i

8000

WAVELENGTH X (~,)

L

I

I

I

i

L

I

1000 2000 DELAY T (ns)

L

L

I

L

3000

Fig. 3. Transient photoluminescence spectra of orpiment (Peru) for pulsed observation with various delay times. Fig. 4. Photoluminescence intensity for the peak at h ~- 7600 A as a function of observation delay time. Pulse and gate duration were 11 ns.

4. Interpretation 4.1. Self-trapping and small formation The interpretation of the transient behavior of the photoluminescence of both glassy and crystalline chalcogenides is outlined below on the basis of exciton and small polaron formation. The absorption of a photon with sufficient energy is presumed to create either a pair o f separated charges or an exciton. Since electronic adjustment and transfer times (~10 -is s), are orders of magnitude shorter than lattice relaxation times (<~10 -12 s), a carrier which is initially localized, spreads out or diffuses from its initial location before the atoms can respond to the change. Self-trapping occurs only when the atoms in the vicinity of the charge adjust to a configuration corresponding to a potential well which confmes the excitation sufficiently. Further localization and ultimate small-polaron formation readily ensues. The time delay for self-trapping involved in this process is associated with a barrier to self-trapping, which is depicted in fig. 5 [9]. This figure shows three plots of the (adiabatic) energy of the system comprizing an electronic charge in a three-dimensional deformable continuum characterized by a local electron-lattice interaction, against a normalized variable proportional to the spatial extent (radius) of the carrier. These three curves successively correspond to either an increasing carrier mass (M a =

M.A. Bosch et al. / PL dynamics 8

I

I

I

'

.

,

I

I

I

591

E !

Rc_

/c

Fig. 5. Adiabatic small polaron energy versus extent of lattice distortion R for three carrier effective masses. Mb/2 =Mc/3 ) or an increasing electron-lattice coupling strength. The small R cutoff at R = Rc occurs at the small-polaron state, where the carrier is confined within a region of interatomic dimensions. Curves a, b and c, respectively, correspond to self-trapping at R = R c being energetically unstable, metastable andstable. With the situation depicted in curve c, a carrier inserted (optically excited) into an undeformed system (R ~ oo) will not be driven to localize unless atoms in its vicinity assume a configuration corresponding to a potential well which confines the carrier within a radius less than Rrnax. Having achieved this condition, the system will then fred it energetically favorable to shrink to the smaU-polaron situation, R =Rc in ~10 -12 s. A barrier for deep small-polaronic trapping of excitons and charge carriers can also occur for excitations initially localized by defects and disorder [9]. Here, the barrier and time delay decrease with increasing initial (non-polaronic) localization. If one or both of the charge carriers of the unrelaxed system have mobilities comparable with those in a crystal (an itinerant picture), the absorption of a superband-gap photon produces an electron and hole which tend to diffuse far apart from one another before relaxing to form small polaronic carriers. When the energy imparted to the relative motion of the oppositely charged itinerant carriers is sufficiently great, the carriers typically move quite far apart before relaxing. For example, if the energy associated with the relative motion of the carrier is 0.1 eV in the case of carriers with free- electron masses, they will separate by about 3000 A in a picosecond. This is typically very much greater than the coulomb capture radius (r c = e2/ekT, where e is the electronic charge, e is the dielectric constant and k T is the thermal energy) at all but very low temperatures. Thus, if the unrelaxed carriers are itinerant, optical excitation produce electron-hole pairs which typically escape from each other's coulombic field and contribute to the photoconductivity. The mobility associated with photoconductivity is the average of the relatively high non-polaronic mobility and the much lower small-polaronic mobility weighted by

592

M.A. Bosch et aL /PL dynamics

the fraction of the photoconductivity lifetime during which the carrier is in each of the two states. Thus with a sufficiently long time delay for self-trapping, the photoconductivity may even manifest high mobility itinerant behavior while the dc transport indicates low-mobility small-polaron hopping. Alternatively, both carriers of the unrelaxed system may be localized or have their motion severely impeded. For example, this can occur if the material possesses structural localization which renders it similar to a molecular crystal. Then optically generated charge carriers may often not separate further than their coulomb capture radius before relaxing and the system will manifest geminate recombination: charge carriers recombine with their nascent partners and thereby avoid contributing to the dc photoconductivity [5]. Furthermore, since localization tends to reduce the barrier and time delay for smaU-polaron formation [9], the photoconductivity will tend to be dominated by the smaU-polaron component. In addition, super-band-gap absorption can produce localized excition-like excitations which, as in molecular crystals, may relax and degrade in energy very little compared with itinerant carriers (~1012 eV s -1 for hot electrons). The observation of geminate recombination [10], small- polaronic dark transport and phototransport [11-13], and long-lived un-Stokes-shifted super-band-gap luminescence [7] in chalcogenide glasses suggests that these materials are examples of this localized variety. In this localized picture, the unrelaxed electronic excitation generated by a super-band-gap photon is taken to be characterized by a distribution of charge-separation lengths. If the resulting electron and hole remain sufficiently close so that their wavefunctions overlap substantially, they are regarded as a neutral entity, an exciton. Otherwise, whether or not they geminately recombine, charges whose wavefunctions do not overlap stibstantially are viewed as separated, or metastably separated charges. Since the interaction of an exciton with the atoms in its immediate vicinity is typically much weaker than that of a single carrier, the exciton is taken to be characterized by a much longer time delay toward small-polaronic relaxation than are separated charges. In this interpretation, the self-trapping of separated charges in the chalcogenide glasses appears to take place in about a picosecond. 4.2. Dynamics in the glassy state

The kinetics of the hopping of separated electron-like and hole-like small polarons from their positions at creation to those from which they recombine reflects itself in the time dependence of the luminescence and induced absorption. To understand the dynamics of the recombination of the electron- and hole-like small polarons involves considering the energy of the system as a function of their pair separation. In addition to their Coulomb attraction, they interact via the overlap of their atomic displacement patterns. The coulomb, polaronic, and net potential energies of the pair as functions of their separation, r, in units of the interatomic spacing, are plotted in fig. 6. The pair-potential minima corresponding to the forma-

593

M.A. Bosch et al. / PL dynamics > o|

...-'

. . . . . '- - - - -c-6~?c?fiff - -

-

f/

~-11 z

, ,, IEI

"

- 2 ~ \ . -3 I 0

x I 1

s P-BINDING GSI 2

1 3

r

Fig. 6. Coulomb, small polaron, and net energy of the polaron pair as a function of the pair separation r in units of the interatomic spacing. tion of an exciton (E), a self-trapped exciton (STE), and the groundstate of separated electron and hole small polarons (GS) are indicated on the figure. Those carriers which do not separate (rinitia I ~ l) after excitation are viewed as forming excitons which ultimately self-trap to STE. Those charges which separate (/'initial ~ 1) rapidly (~10 -12 s) relax to form small polarons at r > 1. They then tend to hop toward their groundstate before recombining. Those that radiatively recombine before reaching GS will luminesce at a higher energy than those which reach GS. Furthermore, since the overlap and the matrix elements for recombination decrease with increasing separation, the luminescence will tend to shift to lower energy with time as the smaU-r pairs recombine first. We now summarize the results for the three PL bands in a-As2Sa. At the highest energy lies the simple exciton luminescence. Since there is little lattice relaxation about the localized exciton it manifests a minimal Stokes shift. A lower-energy luminescence, manifesting a moderate Stokes shift, is due to the self-trapped exciton. Finally, the lowest-energy luminescence is associated with the recombination of separated small polarons. As indicated above, this Stokes shift increases in time as more luminescing carriers converge on their minimum energy separation. Furthermore, calculations [15] show that as a result of reduced electronic energy separation and enhanced electron-lattice coupling strength, non-radiative recombination dominates at the ground state and in its vicinity. 4.3. Dynamics in the crystalline state

From transport measurements it is known that only one carrier, presumably the electron, will form a small polaron in a chalcogenide crystal. With this in mind we expect that the low energy PL band observed in the glass has no analog in the crystalline state. It is assumed that the absorption of a photon with energy in excess of the absorption edge creates an exciton much like those of other crystals in which one carrier self-traps; KC1, NaC1 and KBr are examples. The electron wavefunction will

594

M.A. Bosch et al. / PL dynamics

localize rapidly as the electron self-traps. The hole will then be readily bound to the electron, a coulombic center. This self-trapped exciton can radiatively recombine with a moderate Stokes shift resulting from the electron's self-trapping, The low energy PL process in c-As2S3 manifests the stronger relaxation of a charge pair. To this end we consider a more extreme localization mechanism. Namely, since the hole is localized in the coulomb field of the self-trapped electron, it too may subsequently self-trap. This different type of severely localized selftrapped exciton is associated with a large relaxation, as in the non-crystalline case. These two types of self-trapped excitons can be used to explain the observed spectral occurrence of the two PL processes. The temporal behavior depicted in fig. 4 can be understood readily by the fact that the first STE is a precursor of the second. Therefore the population of the second STE is fed by the first STE. The rise at short times (N100 ns) is due to the increase in the density of severely localized self-trapped excitons as less-localized excitons are converted to the morelocalized variety. The subsequent fall, at much longer times, reflects the usual exciton decay due to recombination.

References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15]

B.T. Kolomiets, T.N. Mamontova and A.A. Babaev, J. Non-Cryst. Solids 4 (1970) 289. R.A. Street, T.M. Searle and I.G. Austin, J. Phys. C: Solid St. Phys. 6 (1973) 1830. R.A. Street and N.F. Mott, Phys. Rev. Letters 35 (1975) 1293. M. Kastner, D. Adler and H. Fritzsche, Phys. Rev. Letters 37 (1976) 1504. D. Emin, Amorphous and Liquid Semiconductors, (University of Edinburgh, Edinburgh, 1977) p. 261. M.A. B6sch and J. Shah, Phys. Rev. Letters 42 (1979) 118. J. Shah and M.A. B6sch, Phys. Rev. Letters 42 (1979) 1420. R. Zallen, Xerox Res. Center, Webster, New York, USA. D. Emin and T. Holstein, Phys. Rev. Letters 36 (1976) 323. P.M. Pai and R.C. Enck, Phys. Rev. Bll (1975) 5163. D. Emin, Amorphous and Liquid Semiconductors, (University of Edinburgh, Edinburgh, 1977) p. 249. D. Emin, C.H. Seager and R.K. Quinn, Phys. Rev. Letters 28 (1972) 813. T.D. Moustakas and K. Weiser, Phys. Rev. B12 (1975) 2448. R.L. Fork, C.V. Shank, A.M. Glass, A. Migus, M.A. B6sch and J. Shah, Phys. Rev. Letters 43 (1979) 394. D. Emin, Adv. Phys. 27 (1973) 57.