Al0.3Ga0.7As(p) double heterostructures

Intrinsic, heterointerface excitonic states in GaAs(n)/Al0.3Ga0.7As(p) double heterostructures G. D. Gilliland, D. J. Wolford, T. F. Kuech, and J. A. Bradley Citation: Journal of Vacuum Science & Technology B 9, 2377 (1991); doi: 10.1116/1.585706 View online: http://dx.doi.org/10.1116/1.585706 View Table of Contents: http://scitation.aip.org/content/avs/journal/jvstb/9/4?ver=pdfcov Published by the AVS: Science & Technology of Materials, Interfaces, and Processing Articles you may be interested in Control of quasibound states by electron Bragg mirrors in GaAs/Al0.3Ga0.7As quantum wells Appl. Phys. Lett. 68, 2720 (1996); 10.1063/1.115576 Electrical characterization of lowtemperature Al0.3Ga0.7As using nin structures Appl. Phys. Lett. 68, 699 (1996); 10.1063/1.116596 Strain relaxation of compositionally graded In x Ga1x As buffer layers for modulationdoped In0.3Ga0.7As/In0.29Al0.71As heterostructures Appl. Phys. Lett. 60, 1129 (1992); 10.1063/1.106429 Lowtemperature CV characteristics of Sidoped Al0.3Ga0.7As and normal nGaAs/NAl0.3Ga0.7As isotype heterojunctions grown via molecular beam epitaxy J. Appl. Phys. 70, 6877 (1991); 10.1063/1.349811 Transient decay of persistent photoconductivity in Al0.3Ga0.7As J. Appl. Phys. 68, 601 (1990); 10.1063/1.346785

Redistribution subject to AVS license or copyright; see http://scitation.aip.org/termsconditions. Download to IP: 129.186.1.55 On: Sun, 24 Aug 2014 19:38:28

Intrinsic, heterointeriace excitonic states in GaAs(n)/ Al o.3 GaO•7 As( p) double heterostructures G. D. Gilliland, D. J. Wolford, T. F. Kuech, a) and J. A. Bradley IBM Research Division, T. J. Watson Research Center, Yorktown Heights. New York 10598

(Received 29 January 1991; accepted 29 January 1991) We have used extensive photoluminescence (PL), PL time-decay measurements, and detailed quantum mechanical modeling to both interpret and quantify the electronic and optical properties of free excitons localized near heterointerfaces. Through detailed spectroscopic measure of the recombination kinetics of the recently observed H-band emission, we find this emission arises from the radiative decay of such weakly bound (=0,5 meV) excitonic species confined to the hole-attractive quantum potentials formed at the p-n heterointerfaces, Detailed measurements in virtually "interface-free" double GaAs(n)! Al o.3 G~J.7As( p) heterostructures shows the effects of GaAs layer thickness upon the H-band kinetics-thus confirming quasi-2D excitons become effectively "shared" by both heterointerfaces for sufficiently thin GaAs layers ( < 0.5 pm). Moreover, we find detailed dynamics of these 2D excitons to be influenced by nonradiative interfacial recombination present in nonideal structures. Lastly, we use a novel, a11optical technique to measure the transport properties of these quasi-2D excitons, and find exceedingly long-range ( > 400 pm) low temperature diffusion.

I. INTRODUCTION

Epitaxial growth by molecular-beam epitaxy (MBE) or metalorganic chemical vapor deposition (MOCVD) has recently produced high-quality semiconductor heterostructures with virtually atomically abrupt heterointerfaces. l - s Such microscopically, and, hopefully, electronically superior, structures may thus provide for careful study of intrinsic heterointerface properties.6-9 Among phenomena possibly traceable to such interfaces is a relatively new photoluminescence (PL) band found in GaAs! AlxGa l _ xAs heterostructures, and referred to as the H-band. 1O Stating briefly, some suggest this PL may be due to intrinsic states within the band bending near the heterointerface, 1l,12 while others insist these recombination processes arise from carriers bound to defects at, or near, the heterointerfaces. 13 - 16 The H-band PL lies energetically between the near-edge free and bound excitons (F,X, D oX, etc.) and the free-tobound transitions (BAe); and has an asymmetric line shape, with a long, low-energy tail. Yuan et al. 10,11 first observed this emission and found that it rapidly quenched with increasing temperature and that its peak position shifted to higher energy [up to the free excitonic (F,X) PL emission] with increasing cw laser power. Most importantly, they found that the H-band emission disappeared after chemically removing the top Alx Gal _x As layer of their structures, and was not evident in structures with graded interfaces. Taken together, these studies 10-12,17 thus suggest this new Hband PL could be traced to radiative recombination of electrons (holes) confined in the electrostatic and heteroepitaxial potential notch of the conduction (valence) band at the heterointerface, with free holes (electrons) in the valence (conduction) band for n-n (p-n) structures. Through additional PL measurements in a magnetic field, it was also found this PL emission is only Zeeman split 12 when the Bfield has a component perpendicular to the heterolayers, again therefore implicating the heterointerface as the PL source. Nonetheless, in addition to the interface being im2377

J. Vat::. ScI. Technol. B 9 (4), Jull Aug 1991

portant, other studies I3 - 16,18 have concluded that this emission is impurity-induced, with possibly shallow defect charge states from within the Alx Gal _x As layer, but next to the heterointerface, being the source. In contrast to most of these earlier conclusions, however, our own recent detailed PL and PL time-decay measurements and calculations, 19 prove this emission instead arises not from impurities, or even free carriers, but from the "intrinsic" recombination of free-excitons, quantum-confined to the electrostatic potential notches at the heterointerfaces. Before our H-band study, we fully characterized the nonradiative interface recombination. We observe room-temperature, band-to-band recombination lifetimes of 2.5 j.ls, and through detailed thickness-dependence measurements, interface recombination velocities of :$ 40 em/s. 20 This extremely low, nonradiative decay rate at the heterointerfaces thus allows for the possibility that interfacial charged carriers making up the H-band excitation might live sufficiently long for its intrinsic properties to be properly assessed, unimpeded by spurious unrelated extrinsic decay, In addition, we have used a novel, all-optical PL imaging technique to measure the spatial transport properties of the interfacial bound carriers which, upon recombination, give rise to H-band emission. II. EXPERIMENT

We have chosen to assess the effects that heterointerfaces have on the carrier dynamics by studying, in detail, a series of samples identical in aU respects except one, the thickness of the optically active region (equivalently the distance between heterointerfaces). Our samples were high-quality, undoped MOCVD-prepared GaAs (n) / Ala. 3 GaO. 7 As (p) double heterostructures with background GaAs n-type doping of - I X 10 15 cm - 3, and unintentional Ala.3 Gao.7 As ptype doping of 3 X 1016 cm - J. All samples were prepared at 750°C on SI substrates with growth interruptions at each interface. GaAs thicknesses range from 0.1 to 2.0 pm,

0734-211X/91/042377-Q7$01.00

@ 1991 American Vacuum Society

2377

Redistribution subject to AVS license or copyright; see http://scitation.aip.org/termsconditions. Download to IP: 129.186.1.55 On: Sun, 24 Aug 2014 19:38:28 · ••••-.-.·.·.·..-.-o·.'.·.-.·.·.--;,••.•.• •.•• " ' n ' .....-,. __ ;.~

0

• • __ " • • • > . . . . . . . . . . . . . . ~ • • • • • • • •

2378

Gilliland et al.: GaAs(n)1Alo.3 Ga o.7As( p) double heterostructures

whereas all Alo.3 Gao.7 As layers were 0.5 pm thick. Photoluminescence was excited by.a synchronously pumped, cavity-dumped Ar +- pumped dye laser, tunable from 6400-7000 A, with a pulse width of 1 ps, and a repetition rate variable from ~ 1 kHz to 76 MHz. Thus, optical excitation was below the Al o.3 Gao.7 As band gap and provided uniform excitation throughout the GaAs layer only. Samples were mounted in a variable temperature (1.8-300 K) optical cryostat, and all data-PL spectra, time decays, and time-resolved spectra-were obtained using the technique of time-correlated single photon counting.

2378

TEMPERATURE 50

~

....,...,Z

20 15

20

~ 10

\

,...4

0...

a.... ,..., 50 f.lS with decreasing emission energy; (2) all decay kinetics are highly nonexponential; (3) lifetimes in thick structures ( ;::; C,Sltm) are all identical; (4) lifetime saturation at low energies is however evident in thinner structures; and (:;) all results for all structures are identical on the highenergy side of the H-band emission profile. In addition, we find drastically shorter H-band lifetimes in other, less ideal structures showing higher interfacial recombination velocities (as determined through 300 K band-to-band recombination lifetime measurements). Evidently, the nonradiative decay associated with these nonideal heterointerfaces also affect H-band kinetics, thus precluding for imperfect interfaces meaningful studies of intrinsic H-band dynamics. Our thorough characterization of the interface recombination velocities of our samples, with the ultimate selection of the most perfect structures, was therefore essential for the current work. No microscopic model proposed, thus far, for H-band emission adequately describes all observed features. For example, the "free-carrier" picture (with at least one particle bound at the interfacial band bending), originally proposed

2379

by Yuan et al.,IO,11 fails to predict the obvious and strong temperature dependence (and small activation energy) we find in Fig. 2 for the H-band. It does, however, appear to qualitatively explain the spectral peak shift, versus time and laser power, of the PL. As a second example, the impurityinduced model proposed by Alferov et al. 13 appears incapable of explaining the observed temperature dependence, the spectral position, or the structural dependence of the decay kinetics. (Moreover, we find no evidence in our data for shallow interlace impurities in the Ala.3 Gao.? As layers.) Finally, the existence of H -band emission in both p-n and nn structures precludes any model based on impurities at the heterointerface in the Alo.3 GaO,7 As layer, since such a model requires donor and acceptor-like impurities, both with small virtually identical binding energies of SO.75 meV. We have thus resorted to a new model similar to Balslev,22 but differing in detail. Here, we interpret the H-band recombination as arising fromJree-excitons which have drifted to the high-field region of the heterostructures, near the heterointerfaces. These excitons become confined in the electrostatic and heterointerfacial potential notch in the valence or conduction band, with only one principal particle being quantum mechanically bound, and, consequently, acquiring two-dimensional excitonic character. This model must necessarily include the entire structure, i.e., both heterointerfaces, since our data indicates a GaAs thickness dependence to the dynamics. In addition, we also find that electron-hole Coulomb interaction, together with inclusion of hole bound states at both interfaces sharing an excitonic electron, is necessary to accurately describe aU H-band dynamics. We thus solve numerically Schrodinger's equation, including the electron-hole interaction, and the time-dependent, electrostatic conduction and valence band potentials for the ground-state electron and hole envelope wave functions making up the quasi-2D exciton. Herein, the time-dependent band-edge potentials result from the time-dependence of the carrier and exciton densities, which may screen electrostatic fields at the heterointerface, thereby reducing the band-bending there. We have simulated such dynamics through a simple, yet realistic parametrization of these timedependent potentials, rather than directly solving the extremely complex, coupled Poisson's, Schrodinger's, and Boltzmann's equations. Thus, the potentials were parametrized as according to do ( ) _ D'd - ([z[/dan) (1) 'l'v Z - £1 aBe ,

ifJc (z)

=

¢v (z)

+ Eg ,

(2)

where F represents the maximum field at the heterointerface, a B is the bulk 3D Bohr radius, and d is the parametrized screening length. (Here d .... O corresponds to early times after pulsed excitation, when the bands are flat due to screening; d -+ de,! corresponds to the long-time regime, or the static case, in the absence of additional photo excited carriers. ) Results of our numerical calculations agree, in detail, with all experimental observations typified in Figs. 1-4. In particular, the observed time-dependent H-band shift [Figs. 3 and 4(b)] is reproduced in the calculation and the result shown in Fig. 5. This theoretical plot shows the H-band transition energy might be expected to vary all the way from the GaAs

J. Vac_ Sci. Techno!. B, Vol. 9, No.4, Jull Aug 1991

Redistribution subject to AVS license or copyright; see http://scitation.aip.org/termsconditions. Download to IP: 129.186.1.55 On: Sun, 24 Aug 2014 19:38:28

Gilliland et 81.: GaAs(n)! Al o.3Ga O•7As( p) double heterostructures

2380

25

I

I

I

I

I

I

,

lOll

1.8 K

20 $' 15 CJ

-5

r-
< u c.::: p.:J Z

u.l

::10

_

"

]'" 2.0 '-LI

o

o 000000000

2

4

6 8 d (aB)

10

12

6. Binding energy, E" vs screening parameter, dB (in excitonic Bohr radii), obtained from model calculation for a 0.2,um double heterostructure. Hatched region indicates region where model calculation is not strictly valid, but displays the correet asymptotic form.

FIG.

""

1000

,,

.../

....

-

"-

/

2000

, "-

50 d

0

2000

= 4.0

-----.-.

1000

a

2000

10ol_l~i)\11

100

l· / "'

50-d = 2.0

3.0

'-

--------

100

,'.,

band gap (1.5194 eV at 1.8 K) for a compact, single heterointerface-bound 2D exciton (resembling most in spatial dimensions, a free exciton), to more than 20 meV below the gap for a diffuse, highly elongated 2D exciton (aligned perpendicular to the interface) shared equally between the front and rear heterointerfaces. In agreement with this calculation, the data of Fig. 4 indeed show a linear energy shift of the PL, down to 28 meV below the band gap. Figure 6 shows the calculated binding energy of the quasi2D exciton versus the parameterized screening length. This shows that initially, as d -> 0 (equivalent to t -+ 0 in the pulsed experiments), the quasi-2D exciton binding energy approaches that of bulk 3D excitons, or 4.2 meV. In the contrasting limit of increasing d (more band bending at increasing times) the exciton binding energy decreases to a fraction of a meV and saturates there for all large d (i.e., for long times after excitation). Taking the results together, we may

"

0

°trHO d 1000

....

- --r -r--'

0

.

"

....

....

50

E ':0 ~ v,,~/,,] 1

>

o

FIG. 5. Transition energy, ET vs screening parameter, dB (in exciton Bohr radii), obtained from model calculation for a 0.2 P.ill structure. Er is measured from the band gap.

....

....

\