Picosecond luminescence approach to vertical transport in GaAs/GaAlAs superlattices

Solid State Communications, Vol. 57, No. 11, pp. 885-889, 1986. Printed in Great Britain.

0038-1098/86 $3.00 + .00 Pergamon Press Ltd.

PICOSECOND LUMINESCENCE APPROACH TO VERTICAL TRANSPORT IN GaAs/GaA1As SUPERLATTICES B. Deveaud, A. Chomette, B. Lambert and A. Regreny Centre National d'Etudes des T616communications (LAB/ICM), 22301 Lannion, France and R. Romestain and P. Edel Laboratoire de spectroscopic physique, Universit6 de Grenoble, BP 68 - 38042 Saint Martin D'Heres Cedex, France

(Received 1 July 1985; in revised form 9 October 1985 by M. Balkanski) Picosecond luminescence of GaAs/GaAIAs superlattices has been measured at 5 K. Asymetrical structures where one larger wellis introduced at 9000 A from the surface are studied. It is then possible to estimate the mean transfer time of photoexcited carriers through 9000 A of superlattice. This time is found to be about 4 nsec in a 40•40 A superlattice and 800 psec in a 30/30A one. This evidences the rather high mobility of small period superlattices in the growth direction.

PICOSECOND LUMINESCENCE of quantum well structures has been studied by different teams [ 1 - 8 ] . Those quantum structures show a decrease of the exciton lifetime as the confinement increases [9], typically the exciton lifetime at 4 K is reduced to about 250 ps for a 50A wide well. If the two dimensional properties of heterojunctions or quantum wells are very promising for their applications in future device technology [10], transport properties along the growth direction (i.e. across the GaA1As barriers) are also very promising [11]. Using standard luminescence, we have demonstrated on short period superlattice (SL) structures modified by the introduction of a few enlarged wells (EW) that motion of the photoexcited carriers across the barriers was possible [12, 13]. When the distance between enlarged wells is rather small (~ 1400 A) the characteristic times on short period SLs correspond to the mean trapping time in the EW. This trapping time has been estimated to be as short as a few ps in a 3 0 - 3 0 SL. It is thus possible to study the diffusion of photoexcited carriers in structures comparable to that used by G6bel et al. [1]: one well is included in the SL at about 9000A from the sample surface. Preliminary results on these structures have already been published [13]. Luminescence and photoluminescence excitation (PLE) studies indeed show that photoexcited carriers are able to move through the structure from the surface where they are created to the enlarged well 9000A away. In this letter, picosecond luminescence results on the same structures are presented.

Our GaAs/GaA1As structures are grown by molecular-beam epitaxy (MBE). Details of the growth conditions have been given elsewhere [14]. Very high quality samples are used, characterized even for very low periods, by the coincidence of the PL and PLE peaks. All structure parameters are checked by X-ray measurements. Picosecond luminescence experiments use a Q-switched YAG: Nd laser and a frequency doubler. Detection is made by a photon counting system with a S 20 photomultiplier followed by a time to amplitude converter and a multichannel analyser. The experimental full width of the whole system is about 350ps. Our fitting procedure takes into account the true response of the system and resolutions lower than lOOps can be obtained. However as many different processes are involved in the experimental curves, the derived times cannot be considered as exact but rather as rough estimates of time constants which are in any case averages over different processes. Let us first describe the luminescence of a regular SL (without introduction of one enlarged well). Luminescence decay curves are presented on Fig. 1 for a 4 0 - 4 0 SL. Two curves are reported: the first one (a) is taken at a higher energy and corresponds to electronhole recombination. The second curve (b) is taken at an energy corresponding to the maximum of the luminescence peak and corresponds to excitonic recombination (in these samples, PL and PLE peak positions coincide, see Fig. 8 in [ 14],in cw excitation conditions). Under picosecond excitation, the luminescence peak is enlarged as a result of larger carrier temperature: typically,

885

886

VERTICAL TRANSPORT IN GaAs/GaA1As SUPERLATTICES

~F-

"

',0-I*0

SL

e-h SLT .

HOT e-h PAIRS ~.

Ib)

I 7) M 11

a)

Vol. 57, No. 11 SL b)

T8 ~ S

EXCITON

=o i II i

II

\

\

rad

IH',~)

rad

EW

/

2000

'

~.000 TIME ( ps )

0

1000

2000

3000

/*000 TII'IE ( ps )

Fig. 1. Time behaviour of the luminescence of a 40 A GaAs-40A GaA1As superlattice. The time integrated luminescence is shown in the insert; the full line corresponds to continuous excitation and the dashed line to picosecond excitation behaviour. Arrows a and b indicate the energies at which are measured the two time dependences: curve b corresponds to excitonic recombination, curve a to unthermalized electron-hole pair luminescence. Curves a and b are fitted using a very simple model including the set up response function and three times defined on the figure: rl the mean time for exciton formation, r2 the exciton lifetime and ra the electron hole pair radiative lifetime. lOOps after the exciting pulse, the electronic temperature is ~ 100K for usual excitation conditions [8]. As a consequence, we cannot ascertain that we do see excitons recombining at this energy: at 100 K, excitons might dissociate into e-h pairs [15]. Let us nevertheless use the terms exciton, and exciton lifetime in order to differentiate from hot e-h pairs. We have fitted the experimental curves using the very simple model shown in the right insert of Fig. 1: photoexcited electronhole (e-h) pairs created by the picosecond pulse very rapidly thermalize towards the bottom of the conduction band [16, 3] (less than a LO phonon energy) then two processes compete: the formation of excitons with a characteristic time rl and the direct radiative recombination with a characteristic time r3. The radiative lifetime of the excitons is r2. A very good fit to the experimental curves (including their relative intensities) is obtained with the following set of parameters (see Fig. 1): rl = lOOps, r2 = 300ps, ra = 750ps. These times are in good agreement with the results published by other groups [1, 3]. A radiative lifetime of 300 ps for a 4 0 - 4 0 SL seems perhaps a bit low as the excitons can no more be considered as confined in one

Fig. 2. Time behaviour of the luminescence of a 40 A GaAs - 40 A GaAIAs superlattice with one enlarged well 9000 A away from the surface (well width 54 A). Time integrated luminescence is displayed in the insert and is characterized by two peaks: peak b corresponds to excitons in the superlattice and peak c to excitons in the enlarged well. The full curve corresponds to low intensity continuous excitation and the dashed line corresponds to the picosecond set up excitation, a, b and c arrows indicate the positions at which the time dependences are recorded. The fitting model includes the integrated intensities and is shown in the upper insert; eight characteristic times have to be considered (see text).

well in such a structure [14] and we would have expected the radiative lifetime to be longer. However we have observed, as Ryan et al. [3], a shortening of the radiative lifetime at low pump intensity as a result of more efficient trapping by impurities, the excitation density that we have used might be too low to get the true value of r2. Using the same procedure, we have obtained for a 3 0 - 3 0 regular SL the same set of parameters as for the 4 0 - 4 0 sample. If we now introduce one larger well in the structure (9000 A away from the surface, the well being enlarged by 5 a/2 = 14.3 A), the photoexcited carriers will diffuse through the structure and will be trapped by the localized level introduced by this enlarged well (EW). As a matter of fact, two luminescence peaks are observed [13]: one corresponds to excitonic recombination in the SL, and the other one to excitons confined in the EW. PLE experiments have shown that carriers that recombine in the EW have been excited in the SL, near the surface [13]. The experiment is very comparable to the one performed by G6bel et al. [I] excepted that, in our case, the surface layer is not GaA1As but a superlattice. A typical result (for a 4 0 - 4 0 SL + one 54 .& well) is displayed on Fig. 2. As for Fig. 1, different decay curves have been plotted corresponding to different detection energies. Curve 2-a corresponds to e-h recombination (7470A), the detection energy is indicated on the

Vol. 57, No. 11

integrated luminescence curve (see the right insert o f Fig. 2). Curve 2-b corresponds to excitons in the SL (7520A) and curve 2-c to excitons in the EW. As can be guessed without any fitting procedure, a major difference appears as compared with the curves reported by G6bel eta/. [ 1 ] : the onset of luminescence in the EW does not coincide with the disparition of the SL luminescence. It is impossible in our case to use a simple model where photoexcited carriers either move to the EW or recombine. Such a model does not take into account the fact that excitons, once they are formed in a characteristic time of about 300 ps, cannot move as easily as electrons or holes do [13]. So that, in a simplified description, the magnitude of the carrier transfer process depends on the respective values of the exciton formation time and of the e-h trapping time. The resulting model is shown in the insert o f Fig. 2. As many as eight characteristic times have to be considered. 7"1, 7"2 and 7"a have akeady been defined. 7"4 is the e-h diffusion time, rs the e-h radiative lifetime in the EW, 7-6 the formation time of excitons in the EW, 7"7 the radiative lifetime o f EW excitons and 7"8 the diffusion time of SL excitons in the EW. We have assumed equivalent e-h lifetimes and exciton formation times in the SL and in the EW (7"~ = 7-6 and 7-3-----7-5). rf, 7-: and 7"3 are derived from the experiments on the corresponding regular SL. So only three unknown parameters remain. From the fit of the three curves (including their respective intensities), we get 7-7

=

887

VERTICAL TRANSPORT IN GaAs/GaA1As SUPERLATTICES

500ps,

7-4 = 4ns, 7"a = 4ns. Using the same procedure, a 3 0 - 3 0 SL gives:

Ei

50/50 >.

SL

o) Ts= 6800C b) Ts =695°C

l

a)

SL

I

I

I

0.80

0.79

0.78

I

I

I

0.77 0.76 0.75 O.?l, WAVELENGTH (pm)

Fig. 3. Comparison of the luminescence spectra of two 5 0 - 5 0 SL, one enlarged well is introduced every 1400 A. The two luminescence peaks respectively correspond to the SL and the EW. Sample No. 286 (curve a) has been grown at 680°C, as the temperature is not optimized the linewidth i : q urte " large, sample No. 293 (curve b) is grown at 695 °C`and shows a much smaller linewidth showing a better interface quality. This change in interface quality induces a change in vertical transport properties evidenced by the change in intensity ratio between EW and SL luminescence. Vertical transport is favoured in sample b (better interfaces).

7"7 = 400ps, 7"4 = 800 ps, 7"s

=

2.5ns.

The determination of 7"8 is quite unprecise and the previous values must be considered as minimum values. The radiative lifetime of excitons in the EW is found to be about 500 ps in both cases, larger than the radiative lifetime in the SL. The time of greatest interest as far as transport properties are concerned is 7"4. This characteristic time corresponds to the mean value o f the diffusion times of photo-excited electron-hole pairs, from the points where they have been created, down to the enlarged well in which they are very efficiently trapped, r4 is one of the times that govern the onset of luminescence in the enlarged well. This diffusion time is found to be

approximately 4 n s (+ 300ps) in the 4 0 - 4 0 S L and 800ps (-+ lOOps) in the 3 0 - 3 0 L. These values confirm our previous interpretation [12, 13]: the occurrence of very rapid diffusion of photo-excited carriers in a superlattice (velocities of about 10 s cm s -1 are estimated). Preliminary results on a 2 0 - 2 0 SL indicate that diffusion would be even faster in that case. Such an easy diffusion is in our opinion only possible if Bloch type conduction occurs in the growth direction. This is in apparent contradiction with some experimental results [17, 18] indicating that conduction perpendicular to the layers proceeds by phonon assisted tunneling between adjacent wells (hopping conduction). Localization that gives rise to hopping conduction would be induced by intralayer and interlayer thickness fluctuations producing fluctuations in the confined

888

VERTICAL TRANSPORT IN GaAs/GaA1As SUPERLATTICES

level energies of the wells. However, the samples that we used in [17] were not grown in optimized conditions. After the realization of these samples, the growth conditions have been further optimized so that very high quality is now achieved [ 14]. This quality is characterized by a very small number of intralayer thickness fluctuations and in negligible interlayer fluctuations [14]. Luminescence excitation linewidth is used to characterize thickness fluctuations [19] the most striking result that we have obtained is a halfwidth of 3.8 meV for a 11.2-11.2 SL (4 GaAs monolayers - 4 GaA1As monolayers). Our interpretation in terms of Bloch condition is then not in contradiction with the results of Palmier et al. [17] as we are not studying the same samples. A typical example of the influence of growth optimization is shown on Fig. 3. Two 5 0 - 5 0 Sis are compared, each includes one enlarged well (58 A, Lz + 3a/2) every 1400A. The upper curve (a) corresponds to a growth temperature of 680°C and the lower one (b) to a growth temperature of 695°C. The better conditions in that second case are evidenced by the smaller luminescence line width. The important result is that a better quality of the interfaces induces a better efficiency of carrier transfer to the enlarged wells evidenced by an increased luminescence intensity [ 12]. The main argument against hopping conduction in our case, aside from the very high velocity of photoexcited carriers, is the fact that diffusion is observed at very low temperatures (1.5 K under cw operation where the photo excited carrier temperature is also 1.5 K) where phonon assisted hopping cannot be activated. Furthermore temperature dependent luminescence experiments (under cw operation) in our samples do not show any activation of the diffusion between 2 and 80K. Above 80K, the luminescence efficiency of the trapping wells decreases, not as a result of slower diffusion but as a result of carrier reemission from the enlarged wells. Further modelization of vertical transport has to include electric field effects in order to give the value of mobility, such experiments are under progress. In our experiments diffusion occurs without electric field and involves both electrons and holes by ambipolar diffusion so that both mobilities are involved. In the case of a 3 0 - 3 0 superlattice, the electron mini-band width is not very much affected by the band offsets (84meV for A B C = 8 0 % , 81meV for 70% and 68meV for 60%, ABC being the conduction band offset). On the contrary, the width of the mini-band for holes strongly depends on the offset: 7 meV if ABC = 60% 2 meV if ABC = 80%. In principle, one might use the Einstein relation to deduce some ambipolar mobility however this. relation includes the carrier temperature which is

Vol. 57, No. 11

different for electrons and for holes and varies during the movement of the carriers. It is therefore difficult to get the mobility. However it is possible to make comparisons with the same kind of experiments performed by G6bel et" al. [1, 20] on GaA1As layers. They use the same layer thickness than we do and find a mean diffusion time equal to 400 ps. This value, when compared with our 800 ps for a 3 0 - 3 0 SL implies that it is possible to get close to bulk mobility values when the SL period is small enough. As a conclusion, we have studied under picosecond excitation conditions the luminescence decay curves of superlattices. Comparison of the results on regular SLs with those obtained on samples modified by one enlarged well 9000)k away from the surface allows to determine a mean diffusion time of the photoexcited carriers across the SL. The diffusion times are found to a 4ns in a 4 0 - 4 0 S L and 800ps in a 3 0 - 3 0 S L . They can be explained by Bloch type conduction in the SL.

Acknowledgements - We wish to thank J.Y. Emery and G. Dupas for samples growth, P. Auvray and M. Baudet for X-ray characterisations and G. Bastard for helpful discussions. REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13.

14.

E.O. G6bel, H. Jung, J. Kuhl & K. Ploog, Phys. Rev. Lett. 51, 1588 (1983). Y. Masumoto, S. Shionoya & H. Kawaguchi, Phys. Rev. B29, 2324 (1984). J.F. Ryan, R.A. Taylor, A.J. Turberfield, A. Maciel, J.M. Worlock, A.C. Gossard & W. Wiegmann, Phy~ Rev. Lett. 53, 1841 (1984). P. Dawson, G. Duggan, M.I. Ralph & K. Woodbridge, Phys. Rev. B28, 7381 (1983). J. Christen, D. Bimberg, A. Steckenborn & G. Weimann, Appl. Phys. Lett. 44, 84 (1984). D. Bimberg, J. Christen, A. Steckenborn, G. Weimann & W. Schlapp, J. Lumin, 30, 562 (1985). J.E. Fouquet & A.E. Siegman, Appl. Phys. Lett. 46,280 (1985). J.F. Ryan, Physica 127B, 343 (1984). E.O. G6bel, J. Kuhl & R. Hoger, J. Lumin. 30, 541 (1985). See for example A.C. Gossard, Treatise Mat. Sci. Technol. 54,537 (1982). L.L. Chang, 1st lnt. Workshop on future Electron Devices, Tokyo, (1984). A. Chomette, B. Deveaud, J.Y. Emery, A. Regreny & B. Lambert, Solid State Commun. 54, 75 (1985). B. Lambert, R. Romestain, D.Miller, A. Chomette, A. Regreny & B. Deveaud, Proc. 15th Int. Conf. on GaAs and related compounds, Biarritz, The Institute of Physics. Conference Series No. 74; 357 (1984). B. Deveaud, J.Y. Emery, A. Chomette, B. Lambert & M. Baudet, Appl. Phys. Lett. 45, 1078(1984);B.

Vol. 57, No. 11

15. 16. 17.

VERTICAL TRANSPORT IN GaAs/GaA1As SUPERLATTICES

Deveaud, J.Y. Emery, A. Regreny & A. Chomette, to be published in J. Appl. Phys. PLE studies on 30.30 and 40.40SLs show that the excitonic resonances disappear at about 100K (B. Deveaud, private comm.). C.V. Shank, R.L. Fork, R. Yen, J. Shah, B.I. Greene, A.C. Gossard & C. Weisbuch - Solid State Commun. 47, 981 (1983). J.F. Palmier, M. Leperson, C. Minot, A. Chomette,

18. 19. 20.

889

A. Regreny & D. Calecki, Superlattices andMicrostructures 1, 67 (1985). D. Calecki, J.F. Palmier & A. Chomette, J. Phys. C. 17, 5017 (1984). C. Weisbuch, R. Dingle, A.C. Gossard & W. Wiegmann, Solid State Commun. 38, 709 (1981). E.O. G6bel, M.R.S. Europe Spring Meeting, Strasbourg (1985) to be published in Journal de Physique - Colloque.