LUMINESCENCE
Journal of Luminescence 52(1992)17—39
JOURNAL OF
Optical and vibrational properties of doped zinc selenide epitaxial layers Khalid Shahzad, Diego J. Olego and John Petruzzello Philips Laboratories, North American Philips Corporation, Briarcliff Manor, NY 10510, USA
In this review article, we present photoluminescence and/or Raman scattering properties of ZnSe epitaxial layers which were doped with either nitrogen, oxygen, indium, or lithium. In particular, we present photoluminescence data due to 2. recombination at discrete pairs ZnSe layers implanted of with nitrogen with a dose of 1013 cm The experimental pair datadonor—acceptor is explained with the in help of aepitaxial theoretical computation a type-Il spectrum involving In donors substituting for Zn and N acceptors for Se. We also report two-hole transitions involving up to 6S 3/2 states of the nitrogen acceptor. The experimental values of the energy positions of these excited states are in good agreement with those obtained using an effective-mass calculation. A sharp line superimposed on the broad donor—acceptor pair band, whose peak position has a constant separation from the excitation energy, is also observed. This separation is 0.9 meV larger than the 1S—2S energy spacing for the nitrogen acceptor. We show that this line could be either due to resonant inelastic scattering of the exciting photons by the acceptor impurities, or due to selective excitation of the discrete donor—acceptor pairs. We also present photoluminescence and excitation data to show that oxygen substituting isoelectronically in ZnSe gives rise to a pair of transitions A~(2.7895 eV) and B (2.7877 eV) as a result of the exchange interaction between the trapped electron—hole pair. The former is attributed to total angular momentum J = 1, F4 representation and is electric-dipole allowed while the latter is assigned to J = 2 belonging to F3 + F5 representation and is electric-dipole forbidden. Based on this model, we explain several experimental observations including varying PL intensity of the B line from sample to sample, relatively rapid disappearance of the B line as function of increasing temperature and strong longitudinal-optical (LO) couplings of A~and B lines. Raman spectroscopy was used to investigate the coupling mechanisms between LO-phonons and electronic excitations in n-type ZnSe layers. The layers, grown by molecular-beam epitaxy, were intentionally doped below the Mott criterion for the insulator—metal transition. The nature of the electron—phonon interaction is determined by the degree of electron localization, which was effectively changed by temperature and donor concentration. The LO-phonons couple to plasmons when electrons are thermally excited into the conduction band or at donor sites. In both cases, unbound phonons are observed. From the renormalized phonon frequencies at high temperature, values of free electron concentration as a function of temperature were established. They are in excellent agreement with Hall-effect determinations. At low temperatures, the photon Raman profiles are asymmetric and show Fano-type line shapes. The electronic continuum responsible for the phonon self-energies at low temperatures was identified as Raman scattering by bound electrons. In addition, Raman spectroscopy was used to establish the presence of a free hole gas and its coupling to the longitudinal optical phonons in Li-doped ZnSe epilayers. The phonon spectra shift to higher frequencies and broaden with increasing acceptor concentration and temperature, as was the case for indium-doped layers, in accordance with the expectation for coupled phonon—plasmon modes. Values for the concentration and mobility of the holes were obtained from an analysis of the spectral lineshapes. They agree with those determined by other methods. A linear relationship was found between the spectral broadening and the hole concentration. In addition, electronic Raman scattering (ERS) from holes bound to Li acceptors was also studied in ZnSe epilayers and correlated with the net acceptor concentration determined by capacitance versus voltage measurements. The ERS spectra reveal several transitions between the ground is and the excited states of the Li acceptors as well as transitions to a continuum of delocalized valence band states. Values of excitation energies for the bound hydrogenic states and the ionization energy of the acceptors were measured. The strength of the ERS signal normalized to the phonon scattering depends linearly on net acceptor concentration.
1. Introduction Correspondence to: Dr. K. Shahzad Philips Laboratories, North American Philips Corporation, Briarcliff Manor, NY 10510, USA. 0022-2313/92/$05.00 © 1992
—
Following many successes achieved in the case of Ill—V materials and devices produced by
Elsevier Science Publishers B.V. All rights reserved
18
K Shahzad et al.
/ Dopedzinc selenide epitaxial
growth techniques such as molecular-beam epitaxy (MBE) and metal-organic chemical vapor deposition (MOCVD), the Il—VI field also picked up momentum several years ago [1]. It was demonstrated soon afterwards that very high quality intrinsic zinc selenide epilayers could be routinely grown providing the starting material sources were also of very high purity [21. Doping zinc selenide p-type had been at the forefront of the problems for several decades. But now with MOCVD and, particularly, MBE low resistivity p-type ZnSe has been demonstrated using Li and N acceptors [3—71. Both Li and N give rise to shallow acceptor levels with ionization energies E~’ 114 meV8 and E~ 110 meV [9,101 respectively. Li-doped samples tend to show an upper limit to the net carrier concentration (NA ND) attainable (of the order of i0’~cm3) [11]. Theoretical work shows self-compensation to be the main mechanism responsible for this behavior [12]. On the other hand, nitrogen has been shown to dope ZnSe p-type with a net carrier concentration of the order of 1018 cm3 [13,14]. In section 3, we present photoluminescence (PL) data, using above-band-gap excitation as well as selectively excited PL using a CW-dye laser, for the case of ZnSe epitaxial layers grown on [001] GaAs substrates and implanted with nitrogen. Recently, there has also been a report claiming p-type ZnSe using isoelectronic oxygen as a dopant [15]. In section 4, we investigate optical properties of ZnSe epilayers implanted with oxygen and show that our data is consistent with what is expected for an isoelectronic trap but not consistent with an electrical acceptor [16]. In the next section, we discuss Raman scattering results on the n-type ZnSe layers. The interaction between electrons and optical phonons in doped semiconductors has been extensively investigated with inelastic light scattering. Most of these investigations were carried out in Group IV or Ill—V semiconducting materials. In the elemental semiconductors, the phonons couple to interband and intraband excitations of the carriers through the deformation potential mechanism. Striking quantum interference effects with associated changes in phonon self-energies result —
layers
from this coupling in p-type Si and Ge and n-type Si [17,18]. In the polar materials, such as GaAs, the nature of the coupling depends markedly on the transverse or longitudinal character of the phonons. For example, the macroscopic electric field of the longitudinal phonons interacts very strongly with the charge density fluctuation of the free carriers giving rise to the so-called coupled plasmon—phonon modes [17,18]. The renormalization of the phonon frequencies due to this interaction is, in most instances, much larger than those yielded by the self-energy contributions of the deformation potential. In section 5, we report on the mechanisms of the electron—phonon interaction in n-type ZnSe layers, which were intentionally doped with shallow donors in concentrations below the Mott criterion for the metal—insulator transition [19]. Despite the fact that p-type ZnSe with Li or N acceptors has been reproducibly grown, Hall effect measurements still present a challenge. This is due to the large Schottky-barrier height of about 1 eV between common metals such as gold and ZnSe, which makes it difficult to fabricate ohmic contacts to the p-type layers. In section 6, we use an optical technique to measure hole concentration and their mobility [201. As in the case of In-doped layers described in the section 5, we use the Raman scattering technique to detect the changes in the frequency and linewidth of the LO-phonon spectra, which arise in the doped samples due to interaction between the phonon and plasma mode of the hole gas. Electronic Raman scattering (ERS) is also used to measure the excited states of the hole bound to Li acceptor [21].A linear relationship between ERS signal and the net acceptor concentration is found and good agreement with the transport data is achieved. These results are also discussed in section 6.
2. Experimental The ZnSe layers were grown by molecular beam epitaxy (MBE) on (001) surfaces of GaAs wafers. High purity elemental sources of Zn and Se were used to generate the molecular beams
K Shahzad et a!.
/
Doped zinc selenide epitaxial layers
for the ZnSe growth. Before the ZnSe deposition, the GaAs surfaces were chemically cleaned and the native oxide was desorbed following standard MBE procedures [22]. Typical substrate temperatures for growth were around 300°Cand growth rates were on the order of 1 p~mper hour. The photoluminescence spectra were generated with the samples either at 5 K cooled by evaporating liquid helium or at 1.8 K by immersing the sample in the liquid helium and pumping it below the lambda point. The excitation sources were either 351.1—363.8 nm lines of an Ar ion CW laser or continuously variable lasing wavelengths (—~417—480 nm) generated by pumping stilbene III dye by 2 W of CW UV laser. This lasing range is very convenient for ZnSe since the excitonic energies are in the vicinity of 440 nm and the pair spectra fall in the region of 460 nm. The signal was dispersed with an 17.8, 0.85 m Spex double monochromator and detected with an RCA Erma III photomultiplier tube and a lock-in amplifier. The Raman experiments were conducted in backscattering geometry with the exciting and scattered light propagating along the (001) and the (001) directions, respectively. The polarization of the incoming and scattered photons was along (110). This scattering geometry is described by the standard notation z(x +y, x +y)2 [17,18]. Most of the Raman spectra were excited with the 488.0 nm (2.54 eV) visible line of an Ar~ ion laser operating in a continuous wave mode. The energy of such laser photons is smaller than the fundamental band edge of ZnSe, which lies between 2.63 and 2.82 eV going from room temperature down to 4 K [23]. The Raman scattered light is generated in the entire volume of the epitaxial layers and, therefore, it is representative of the doped bulk portion of the ZnSe specimen. Contributions to the total Raman signal from light scattered in depleted space charge regions at the ZnSe—air or ZnSe—GaAs interfaces are expected to be negligible with below band gap excitation [18]. The Raman scattered photons were analyzed with a Spex 1404 double monochromator and detected with photon counting electronics. The reproducibility of Raman peak positions was better than ±0.1cm~. For the ‘~
—
.
.
.
19
temperature dependent measurements, the sampies were mounted on a copper cold finger of a closed cycle He cryostat. The temperature was tuned between 300 and 12 K with an accuracy better than ±0.2 K.
3. Nitrogen acceptors in zinc selenide: donor— acceptor pair lines and two-hole excited states Figure 1 shows PL using above-band-gap exciting radiation, in the excitonic region for three cases. The bottom spectrum is from the as-grown 1.5 ~m ZnSe/GaAs which shows free exciton (FX) transitions which are barely split due to residual tensile strain [24,25], as well as peaks due to excitons recombining at unintentionally present neutral Ga and/or Cl donors, j~0 [26]. For the present discussion, it is important to note that no excitonic transitions are observed at lower
PHOTON ENERGY (eV) 2.81
2.~O
2~79
5)~PHOTOLUMINESCENCE
UVAr -ion excitation
.
.~
xl
2
~
FX X10
I
Li i~
~
w
Unanplanted 500C/30 miii
4 lU
X
x4
22700
I
I
22600
22500
WAVE NUMBERS (cm Fig. 1. 5 K photoluminescence spectra from (bottom) as-grown (— 300°C) 1.5 p.m ZnSe epitaxial layer grown on GaAs; (center) a similar layer annealed at 500 C/30 mm; and (top) 2 and annealed at the same layer implanted with ix 1013 cm 500°C/3Omin.
20
K. Shahzad et al.
/ Doped zinc selenide epitaxial
PHOTON ENERGY (eV) 2.80
2.70
2.60 5K PL
I~-Lo
IJV Ar°EXCITATION xlO
13
FX
-2
N~O~m
H
IDA) 2.
Q
implanted and unimplanted), based on careful energy inter-comparisons, we conclude that the slightly increased intensity in donor—bound-exciton transitions is due to the unintentional incorporation of small amount of indium donors [27] during annealing stages. The most important point is that, in thepeak implanted case, observe very intense at 2.7877 eVwe andalso a weak peaka at 2.7895 eV (strain-split component), identified with excitons bound to Nse acceptors, I ~ [28].
i~-2LO~
I
layers
(DA)
N
10
M ~ ...._.J( I
xl
~
PftJR LINES 2L0
x5 0
I
~‘°
FX’
xiooJ
‘~J
(n.2)j
x250
310
(Note that for the i~peak, the magnification had to be reduced by a factor of ten.) The wider spectral range PL data from a virgin epilayer (fig. 2, bottom spectrum) is dominated by intrinsic transitions. Particularly noteworthy are the longitudinal-optical of the free exciton transitions and a(LO) weakreplicas and broad transition at
22500
22000
21500
1)
21000
WAVE (cm range for as-grown Fig. 2. As fig. 1 but overNUMBERS larger spectral case (bottom) and implanted and annealed case (top).
energies (in the vicinity of 22500 cm i) in this virgin sample, as is also depicted in the bottom spectrum of fig. 2, which covers a much wider spectral range. The PL response from a control sample which was only annealed (but not implanted) at 500°C/30 mm is shown in the center spectrum of fig. 1 which shows that FX and donor-bound exciton transitions have shifted to lower energies and the splittings have increased due to increased tensile strain as a result of annealing. (Note that there is a very weak peak at 22500 cm maybe due to an unintentional incorporation of an acceptor during the annealing stage which is very rarely observed. It is probably not Li since we know from our experiments that Li is very unstable under the annealing conditions described here. Also, we do not observe any pair luminescence in this case.) When the virgin epilayer was implanted with N~=1013 cm2 and annealed at 500°C/30 mm, we obtain the PL response shown in the top spectra in figs. 1 and 2. The positions of the free and donor-bound cxciton components are found to be identical for —
-
both the nitrogen-implanted and annealed cases (fig. 1). In the case of annealed samples (both
2.6 eV (Y
0) whose origin is not well understood. In the(fig. case2,oftop the spectrum) implanted issample, the PL spectrum dominated by a multitude of sharp lines. In particular, we note that the nitrogen—acceptor-bound transition (I~)and its LO-replicas are the dominant features. This strong phonon coupling (Huang— Rhys factor, S 0.05%) is one of the typical characteristics of acceptor—bound-exciton transitions in ZnSe, in contrast with donor—bound-exciton transitions which typically show an order of magnitude weaker coupling to LO-phonons. The top spectrum in fig. 3 shows discrete donor—acceptor (D—A) pair lines involving the nitrogen acceptors. It is extremely rare to observe such transitions in ZnSe [291. To the best of our knowledge, this is the first report involving nitrogen acceptors in this material. The origin of D—A pair transitions is well understood [301.It involves the recombination between an electron trapped at a donor and a hole trapped at an acceptor. Once the electron and the hole recombine, the final transition energy is renormalized due to the Coulombic interaction between the two bare charged impurities left behind. The resultant emission energy is then given by hv R~= E IE + E ‘I + e2 IcR + JI R\ (1 —
~
G
\
D
A/
/
‘.
where E 0 is the free particle band gap, R is the pair separation, c is the static dielectric constant, and e the electronic charge, ED and EA are the
K. Shahzad et al.
/ Doped zinc se!enide epitaxial
2.75 I
2.71
I
I
transitions corresponding to recombination between donor—acceptor pairs of various separations (spectral shape not corrected for the system
(SHELL NO)
response). These are best obtained by strongly exciting with above-band-gap radiation. Several of the discrete lines are broadened due to the inhomogeneous strain which almost always plagues these hetero-epilayers. This broadening may cause several lines to overlap especially for the cases of pairs which are more than 30 apart. With a hindsight knowledge that Nse acceptors and In~~ donors give rise to D—A pair emission, we compute theoretically a type-Il spectrum,
l.6KPL UV Ar’ excitation
~LO~
A
z w ILU
.1 LJ.j Z
o
II
•
hilt
II
IIIIII~
I
II
III I
II~
II
II’
Ill1
II
I
8L L~
II I)~ /\!\...‘~n
RtIlhl.I
I~I
/I~l~I~!V
IUIIVI ‘II
~~ O
21
2.73
PHOTON ENERGY (eV) 2.77 I
layers
~
72-73 85
82 EA=O.1O8 ED~O.O2 0=0.0012
al. [311which also includes the line broadening. basedalso (We upon tried thefitting formalism our data discussed to a type-I by Vink specet
~
trum but with much poorer results.) We use 8.8 [32], donor Bohr radius aB 28 A and a half-width Gaussian u 0.0012 eV. The value of EA is found to be 110 meV from the present analysis of the two-hole and selective pair lumi-
~ I~I29
=
I 27 I128
25
~.
I
22200
if)
=
22000
=
21800
WAVE NUMBERS (cm~) Fig. 3. Top spectrum: FL spectrum showing discrete donor— acceptor pair lines superimposed by LO-phonon replicas of I~and 2s two-hole transitions. Bottom spectrum: a computed spectrum based on type-Il pairs where pairs of the same separation (R) are labeled by shell numbers,
donor and acceptor binding energies. J(R) is the correlation interaction of the overlapping impurity wave functions and is important only for pairs separated by less than the Bohr radius of the shallower impurity (i.e. donor; Ed 30 meV) which is 28 A. Since the donors and acceptors can occupy only discrete lattice positions, this gives rise to the discrete transitions through Rdependent terms in eq. (1) (each value of R is denoted by a shell number). We distinguish here two types of pair spectra. If the donors and acceptors occupy lattice points on a single FCC sublattice of the zincblende structure, then it is referred to as a type-I spectrum. A type-Il spectrum, on the other hand, results if donors and acceptors lie respectively on the alternate FCC sublattices, resulting in different intensities for various pair transitions. Figure 3 (top spectrum) shows experimental data showing a multitude of ‘~
nescence data (presented later on in the text), very close to the previously reported value (111 meV) [9]. The value of E~ is taken to be 28.2 meV [27]. As regards the choice of J(R), we find that the use of the traditional van der Waals’ polarization interaction term (1/R6 dependence) [30] gives a very unsatisfactory result especially in the region of closer pairs (<30 A). Vink et al. [31] used J(R) C/R4 to fit their data; however, we employed an exponential form of J(R) derived by Neumark [32] which is an empirical fit to the discrete pair data. The lower curve in fig. 3 is a computed type-Il spectrum based on the parameters discussed above (we did not attempt to correct for the general shape of the curve). The agreement is fairly satisfactory as far as the peak energy positions of various transitions are concerned (note the presence of strong LO-phonon replicas superimposed on the experimental pair lines). However, it is important to bear in mind that no attempt has been made to take into account the effects of uniaxial strain on the pair transitions which would cause splitting of various lines. Nevertheless, since our data can be fitted by a type-Il spectrum we conclude that nitrogen acceptors on the selenium sites and indium donors
22
/ Doped zinc se!enide epitaxial layers
K Shahzad et a!.
on the zinc sites are responsible for the observed pair spectrum. This is consistent with our conclusion from the analysis of the impurity—bound-exciton lines of the implanted and annealed samples indicating transitions involving excitons bound to indium donors and nitrogen acceptors. We find that the optimum way to observe the two-hole excited states is by resonantly exciting at the peak of the I~line at 2.788 eV [33]. The resulting spectrum is depicted in fig. 4. Two-hole transitions result when an electron—hole pair trapped at the neutral acceptor recombines leaving the impurity hole in one of the excited states. In this case, even the relatively distant (40 A at around 2.73 eV) pair lines are barely visible. However, the very weak peak labeled as 2S in figs. 2 and 3 is replaced by a very sharp (half-width 1.0 meV comparable to I~)and strong peak at 2.7077 eV, shown in fig. 4. This gives a value for 1S—2S of 79.9 meV, consistent with the previous report [9]. In addition, we observe several other peaks associated with other excited states of the nitrogen acceptor as labeled in fig. 4 and summarized in table 1. Note that since we are resonantly exciting at the peak of I~line, its LO-replicas are extremely strong. The identification of these cxcited states is made possible with the help of a theoretical calculation in the effective-mass approximation as given by Baldereschi and Lipari [34]. Using the Luttinger parameters [35] of 71 =
PHOTON ENERGY (eV) 2 77
>~ Z
I
2.75
2.73
2.71
i~~.iI ~1~’
fl
269
2.67
265
I
I
I2S~l~ 11(2W)
9atenaui’~
~
I
I:
(-2249Oco~)
\~P5i2cr~) h~,~tr,)
~) PAP~LI&S 2i~,2
.~
3S
g I 22200
I 22000
I 21800
identification 1S3/2—2P3/2(F~) iS312 2S312 -
experimental 67.8 79.9
theoretical 68.1 79.9
92.8 98.5 102.2 103.4 105.0
94.1 95.4 100.4 103.5 105.3
3/2~~~
lS~/2_3S3/2
1S~/2—2P1/2 1S3/2—4S3/2 1S3/2_5S3/2 iS312 —6S3/2
4.3, y2 0.59, and y~ 1.34 we obtain a spherical coupling factor of 0.48 and a cubic correction factor of 0.174 (the cubic term splits the P5/2 state into F.~ and F~ and renormalizes the P3/2(F~)state). We use the 1S—2S separation to obtain an effective Rydberg, R0 86.6 meV. The nS states (n > 3) are calculated using the relation [361:~ EAn i.76, which is known to hold for the cases of Li and Na acceptors in ZnSe [37]. The resultant theoretical values are compared with experimental ones in table 1. The general agreement is very satisfactory especially for higher lying states, as expected. Note that our experimental value for 1S—3S is 92.8 meV, rather comparable to the theoretical value of 94.1 meV but in marked contrast to a suggested value of 102 =
=
=
=
.
.
meV in the previous report [9]. The binding energy of the ground state, i.e. the ionization energy, is calculated to be 110.0 meV. It is interesting to note in fig. 4 that we do not observe the LO-phonon replica of the 2S state at the LO(I’)
servation has been previously reported in detail for the caseasof due acceptors ZnTe and has been interpreted to waveinfunction interference
~s
I 1) 21800
Nitrogen acceptor: EA = 110 meV
=
OS
2P,,2
gies in meV
31.8 meV. Instead, we observe a broadened doublet at 28.8 and 30.7 meV. Similar ob-
l~¼3Lo
w
Table 1 Experimental and theoretical values of the energy difference between the excited states and the ground state of N 5. accep. tor in ZnSe from selective photoluminescence data. All ener-
I 21400
NUI~ERS (cm Fig. 4. FL spectrum WAVE obtained by resonantly exciting at the peak energy of I~ at 2.7877 eV showing various two-hole excited state transitions of N 5e acceptors in ZnSe.
effects arising between bound and unbound LOphonons associated with the neutral acceptor [38]. Information about the excited states of the acceptors may further be obtained by selective excitation of the D—A pairs [35]. The energy of
K Shahzad et al.
/ Doped zinc selenide epitaxia! layers
the pump photon ho~ is chosen so that a donor—acceptor pair with a separation R is cxcited such that, for example, the donor electron is left in the ground state of the impurity while the ~1~f
ground states of donor and acceptor, given by 4is —A hw~ (‘ 25) + i.~J(R), (2) =
PHOTON ENERGY (eV) 2.73 Eexc (eV)
pump and emission photon energies (i.e. hw~ hw1) remains constant at 653 cm’) 81.0 meV). —
,
2.69
2.67
,
1.6 K SPL
2L0
+27776
~
f
.~
+
2.7805
I 2.7817 Z
I
—
where (A1~—A2~)is the energy difference between A1~and A25, and ~J(R) represents the difference in the correlation energies between a (D1~,A2~)and a (D1~,A1~)pair. Hence, a sharp luminescence peak at hw1 will be observed, tracking the pump energy with a separation roughly equal to 1S—2S of the acceptor. As shown in fig. 5, in practice this emission peak is most efficiently observed on the high energy side of the D—A pair continuum background. This is due to the fact that the pump energy hoi~also excites (D15, AAlso, is) pairs competition with D ~ correA 2S) pairs. for in larger pair separations sponding to the lower energy side of the broad pair band, the probability for a pair recombination is low compared with that for the higher energy side and, finally, the lifetime of a largerseparation pair is very long leading to diffusion of the excitation energy. The bottom spectrum in fig. 5 is obtained by resonantly pumping at the peak of the I~at 2.7876 eV giving rise to very strong, resonantly enhanced LO-replicas (only the 2LO replica shown). The dominant two-hole 2S transition is also very evident. As we reduce the excitation energy, the 2S satellite is replaced by a weaker peak labeled hw 1 riding on the broad band PL. In these cases, we can also observe the peak due to inelastic scattering of the pump photons by LO-phonons (i.e. Raman scattering) 2LO. An important aspect of this labeled as L observation is that the separation between the
2.71
N=1a3c,,,2.5o&CI3o~~n ~t
.E
the moderately doped cases energy diffusion between photoexcited pairs remains sufficiently slow, then internal dc-excitation of the hole within a shallow acceptor will take place rapidly and non-radiatively involving phonon emission. The detected radiation hw1 will then be between the
23
2S i~3cm~
(/) L2~LoI
2.7832 2.7848
/
2.7876
I
(1) 21800 I
21600
WAVE NUMBERS (cm~) Fig. 5. FL response as a function of excitation energy ~ from a ZnSe epilayer showing a sharp luminescence 1 .2L0 are I, peak 2L0 (hw1) superimposed on the Raman broad pair band. L laser peaks.
This is in contrast with the case of 1S—25 separation of 646 cm~(—~80.1 meV) from the two-hole data of the previous figure. An inspection of eq. (2) immediately suggests that this difference of 7 cm~ (—~0.9 meV) may reflect the correlation energy difference between D—A pairs in the ground and excited states. Strictly speaking this~ difference should not be constant as pairs of different separations are being excited. However, there is no theoretical framework at this point to show how much difference is expected and it could well be within the experimental accuracy. It is important to note that the shift in the exciting photon is exactly balanced energy shift in energy this luminescence peak by (i.e.thez~hw~ = z~hw 1).Therefore, this process of selective pair luminescence discussed above becomes indistinguishable from resonant Raman scattering pro-
24
K. Shahzad et al.
/ Doped zinc selenide epitaxial
cesses in which the pump photon is inelastically scattered by the acceptor impurity. Nevertheless, from these measurements, we again obtain a value close to 80 meV for the 1S—2S separation.
layers PHOTON ENERGY (eV)
2.80
2.78
2.76 I
D1A~
In summary and conclusion, we reported observations of donor—acceptor discrete pair lines in PL and the data are explained in terms of a type-I spectrum involving In~~donors and N~e acceptors. In addition, two-hole transitions involving up to 6s states of the N57 acceptor have been deciphered using selective excitation spectroscopy. Effective-mass theory gives good agreement with the data. We also reported a sharp line superimposed on the broad D—A band, which has a constant separation from excitation energy, and suggest that it is either due to inelastic scattering of exciting photons by the acceptor impurity or selective excitation of donor—acceptor pairs.
B
FX
F
-
XLO
1
SAM~LE~
4. Isoelectronic oxygen trap in zinc selenide: effeet of exchange electron—hole interaction
4
(SPOT 1)
4 (SPOT2)
Recently, there has been a report describing the achievement of p-type zinc selenide using oxygen as a dopant [15]. In this section, we describe optical properties of zinc selenide epilayers implanted with oxygen ions and show that our data is consistent isoelectronic trap. with what is expected for an In fig. 6 we show photoluminescence (PL) data taken with samples at 5 K using UV lines of an Ar ion laser. The samples were implanted at 110 keV with either ~ 0 + or ~8 0 + at the doses of 7 X lO~~to 7 X 2 and were an10i3 annealing cm nealed by rapid thermal at temperatures indicated in the figure caption. Rapid thermal anneal was carried out in a zinc over-pressure. All the spectra show very strong free exciton related transitions FX~ (2.7980 eV) and FX— (2.8027 eV), the splitting being caused by the tensile strain due to thermal mismatch between the epilayer and the substrate. The transitions due to the recombination of excitons bound to neutral donors are also clearly visible in the form of a strain-split doublet at 2.7935 eV (D ~) and 2.7961 eV (D) [24,25,39]. The novel aspect of these spectra is that we additionally observe the
I°iLo
B x4
A°~0LO I
I
________
22500
22300
1)
WAVEfrom NUMBERS (cm Fig. 6. Photoluminescence three oxygen-implanted ZnSe epitaxial layers. Sample 1 had a dose of 7X 1012 cm~2 and annealed at 700°C for 5 s. Samples 2 and 4 had a dose of 7x iO’~ cm~ the latter was annealed at 700°Cwhile the +
-
former at 750 C for 5 s. A , A and B transitions are identified with the isoelectronic oxygen impurity. D~,D and
FX~,FX
are the strain-split donor—bound- and free-exci-
ton-related transitions. Note the varying intensity of B from sample to sample and within a sample.
two transitions labeled as A~(2.7895 eV) and B (2.7877 eV) as well as a transition appearing as a shoulder on the lower energy side of the D ± line, labeled as A at 2.7919 eV. The line A~ was observed previously [40], but the line B has not been reported. I~at 2.7794 eV is quite often observed in bulk grown or annealed samples and is believed to be related to a copper—zinc vacancy complex [27,41]. An important observation to note
K. Shahzad et a!.
/ Dopedzinc selenide epitaxia! layers
is that in the cases of free and donor—bound-exciton transitions, the higher energy strain-split components are much weaker compared with their lower energy counterparts, due to thermalization effects at 5 K. However, the relative intensity of A~and B lines is not constant and, in fact, A~is stronger than B in most cases. This is seen by comparing the PL spectra from samples 1, 2 and 4 in fig. 6. In addition, the relative intensity of the aforementioned lines is found to vary within a sample, as seen by comparing the bottom two spectra taken at different spots for the sample 4. The full-width-at-half-maximum for the line A~is 1.5 meV whereas that for the B line is 2 meV. A closer look at the two lines reveals that the A + is symmetrically broadened whereas the B line shows asymmetric broadening on the higher energy side. The coupling strengths of A~and B to LO-phonons is 0.03 for both cases, comparable to the cases of typical acceptor—bound-exciton transitions in ZnSe [10]. In fig. 7 we present 1.9 K PL excitation data for sample 2. With the detector tuned at 1LO below the A~line (A1~0),we observe strong PL when the excitation is in resonance with the A~ line, as expected. The signal is weaker when exciting at the B line. The most noteworthy feature is the strong resonant enhancement of the A line. Recall a weak shoulder at 5 Kthat in APL line (fig.appears 6) (see asalso the temperature dependent data presented later). The rest of the higher energy features are due to the usual intrinsic free exciton processes [261. Since our detector is not tuned to either the LO of donor—bound-exciton transitions or their twoelectron satellites, we do not observe D~and D peaks in PLE. The PLE spectrum with the detector tuned at 1LO below the B line (BLO) is shown in the upper curve of fig. 7. The important point to note is that we observe a very strong enhancement of PL at the A~peak. Finally, in fig. 8, we present PL in the excitonic region as a function of the sample temperature. We note that increasing the temperature from 2 to 25 K, the B line becomes weaker much more rapidly than the A~line. At the same time, the A line grows stronger relative to the A~ line, Of course, we also observe the usual thermaliza-
25
PHOTON ENERGY (eV) 2.82
2.81
2.80
I
2.79
I F
SAMPLE 2
DETECTOR POSITION A n2
~.
FX
E
0~
~.
B
~
(22234.1cm~)
A
I B
~ x2
I
A (22247.4 ~~i~’) I xl
I I
I
I
22700
I
I
22500
1)
WAVEexcitation NUMBERSdata (cmfor the oxygen-imFig. 7. Photoluminescence planted ZnSe epilayer 2. With the detector at AL 0, a strong resonance at A is observed. With the detector at BLO, a strong resonance at A~is observed.
tion effects between the donor—bound- and freeexciton strain-split components [25,39]. The most important point is that the B line thermalizes out at a much faster rate than the D~or FX~lines. For example, at 40 K, the B line is hardly observable whereas the other two lines can be clearly seen. We will now discuss the implications of the data presented in figs. 6 to 8. First of all, since the lines A~ and B are observed only in those samples which are implanted with oxygen ions, we conclude that they originate as result of recombination of an exciton trapped to this isoelectronic impurity. It is clear from the varying inten-
26
K. Shahzad et a!.
/ Doped zinc selenide
the rapid disappearance of the B line compared with D~and FX~lines again points to the fact that A~and B are not a strain-split doublet.
PHOTON ENERGY (eV) 2.81 I PL
2.79
ZnSe:O
C
T
.~
40K
>-
2.77
1 I
However, the splitting between A~ and A increases with increasing anneal temperature
FX+
~
F~-
enhancement addition, PLEatoftheA A shows line. These a strong dataresonant strongly suggest that A~ and A are a strain-split doublet. Although A~and B are not strain-split cornponents and their ratio of intensities varies from ~
I
I-
0 z
A line than the of A~ 5 of K but becomes from 2.42isstronger tomuch 2.60 weaker meV. as the Also, temperature intensity isatraised. the In
III
LU Lii
I
B
1/ )~,,
z
25K
LU 0
Ui -J 0
epitaxial layers
appear to be related to the oxygen impurity. This
sample is clear to from and PLE even dataline within which a shows sample, strong they detected enhancement atsample thethe ofBLO the position. A~ Wewhen will now the discuss PL is a possible origin of A~and B lines. Oxygen in ZnSe forms an isoelectronic trap by replacing the selenium atom. We postulate that
15K
I-
0 I a-
2K -
I I
22600
22400
WAVE NUMBERS
(cm~”1)
Fig. 8. Temperature dependent photoluminescence from oxygen-implanted ZnSe epilayer 2. Note the relatively fast disappearance of the B line compared with D’ and FX~ lines and increase in the intensity of A line relative to A~line.
sity ratio of A~to B, that these two lines are not a strain-split doublet. The cases of strain-split doublets can be seen for the D~ and D lines and FX~and FX lines where the higher energy components are much weaker than their lower energy counterparts at 5 K and grow in intensity as the sample temperature is raised. This is further affirmed by noticing that as the anneal temperature is raised from 700 to 750°C,the splitting (D~—D), for example, increases from 2.66 to 2.76 meV, as a consequence of the increasing tensile strain between the epilayer and the substrate. On the contrary, the energy separation between A~ and B decreases from 1.80 to 1.67 meV for the same anneal temperatures. With increase in the lattice temperature of the sample,
the lines A~and arisecenter from of thetherecombination of direct exciton atB the zone, bound to the oxygen impurity. The mechanism of binding an electron—hole can be understood in the following way [42].pair Oxygen, being more dcctronegative than selenium, would bind an dcctron by a strong short-range interaction [42,43]. The resulting charged center then binds a hole through long-range Coulomb interaction, producing a center which behaves like an isoelectronic “acceptor” [44—47].The angular momentum of the trapped electron and hole, j = and I = combine to form states with total angular momentum J = 1 and J = 2, which are split by the Coulomb interaction. Theoretically the state with the highest degeneracy (J = 2) will be the lowest in energy. The transitions from the upper level (J = 1) to the J = 0 ground state are electric dipole allowed whereas transitions from the J = 2 are electric dipole forbidden. In the present case, the J = 1 transition corresponds to the A~ line whereas the J = 2 corresponds to the B line. The varying intensity of the B line is simply a manifestation of the break-down of this ideal selection rule. This is expected to be particularly evident in the cases of samples which have varying degrees ~-
K Shahzad et al.
/ Doped zinc selenide epitaxial
layers
27
of implantation damage which is not fully annealed out. Furthermore, 0-doped MBE layers [15] show only the A line pointing to more homogeneous quality of the samples. Thus, the varia-
LO-phonon replicas of D~and D- are not visible on this magnification. This is to be contrasted with the case of an isoelectronic “donor” (i.e. a trapped hole) [44]. In this case F1, F1~and 1’~~
tion in the structural quality from sample to sampIe or inhomogeneities within a sample would lead to varying degrees of the relaxation of the selection rule and, hence, varying intensities of the B line, The fact that the A~ line is found to be homogeneously broadened is consistent with the fact that this transition, being an excited state, is expected to suffer strong life-time broadening effects [48]. The inhomogeneous broadening in the B line is understood in the followingway. The electron and hole belong to representation F6 and F8. The direct product F6 X F8 gives F4 + F3 + 1T~.The J = 1 corresponds to F4 representation while J = 2 belongs to a reducible representation F3 + F5. Hence, the J = 2 level is further split by the crystal field into F3 and F5 with the latter lying lowest in energy [43,48]. In zero external magnetic field this splitting is expected to be very small. Instead it gives rise to the inhomogeneously broadened B line. When the temperature of the sample is increased, the B level, being lower in energy, depopulates. This, coupled with the fact that this transition is electric-dipole forbidden, causes much more rapid disappearance of the B line compared with D~and FX~lines, The coupling of the A~ and B lines to LOphonons is found to be almost identical and the PL and PLE spectra are found to be almost mirror images of each other. This is consistent with what is expected for an isoelectronic “acceptor” [44]. For the case of a trapped electron (i.e. isoelectronic “acceptor”), coupling to only F1 phonons are important. These phonons do not mix different electron or hole spin states, and the entire PL spectrum from B will be the same as that from A~except shifted by the splitting At—B. In this case LO-coupling is expected to be F~ type and to be strong. This is exactly what is observed for the present case, i.e. a Huang—Rhys coupling factor of 0.03 for A~and B lines is strong compared with typical values expected for donor—bound-exciton transitions. This is clearly evident from the bottom spectrum of fig. 6 where
phonons can strongly couple. This leads to crosscoupling between A~and B states which, in turn, causes very different coupling of A~and B lines to the lattice modes. Such cases have been observed, for example, for Bi in GaP [49] and Te in CdS [50]. In summary, we showed that oxygen forms an isoelectronic trap giving rise to a pair of transitions A~and B. The former is attributed to J = 1 and F4 representation and is electric-dipole allowed while the latter is assigned to J = 2 with F3 + F5 representation and is electric dipole forbidden. Based on this model, we explained all the experimental data presented here. For example, the varying intensity of the B line is attributed to the relaxation of this selection rule as a consequence of inhomogeneities in the implanted samples and also explains why mainly the A~line is observed in MBE in situ doped layers. Strong and comparable couplings of A~and B lines to LOphonons is shown to be consistent with the case of an isoelectronic “acceptor” where only F1 phonons strongly couple to this state [45]. However, we did not find any evidence of acceptor-like behavior from the electrical measurements, or, any clear indication of donor—acceptor pair bands pointing to oxygen as a shallow acceptor.
5. Indium donors in zinc selenide: phonon coupling to free and localized electronic excitations In this section, we discuss the mechanisms of the electron—phonon interaction in n-type ZnSe layers, which were intentionally doped with shallow donors in concentrations below the Mott criterion for the metal—insulator transition. The coupling mechanisms between the longitudinal optical phonons and the electronic excitations were established using Raman scattering experiments and by correlating them with measurements of transport properties carried out on the same samples. We have monitored the Raman
28
K. Shahzad et a!.
/ Doped zinc selenide epitaxial
T (K) 300 100
50
20
15
108
10
Hall Effect Raman
17
10
10 •
~
•
•
.
~
1016
.
1015
II
0.01
1015
III
0.03
0.05
1/T(K
10~
0.07
0.09
1
Fig. 9. Hall effect results as a function of temperature for a 17 donors cm3 typical n-type doped ZnSe layer with 1.7x i0 grown by molecular beam epitaxy. The dip in the inverse of the Hall coefficient RH at around 50 K establishes the presence of a “two carrier” system. Temperatures larger than 50 K activate electrons into conduction band states. The free electron concentration n(T) is then given by (R 0eY* At temperatures dominates. Thebelow optically 50 K,established conduction carrier in anconcentration impurity band is plotted for comparison. Good agreement is obtained with the transport determination,
frequencies and lineshapes for will different degrees of electron localization and we show that the longitudinal optical phonons of ZnSe can either form a coupled system with the plasmons or display self-energy effects typical of discrete—continuum interactions. Temperature is a very effective way to control localization of electron wave functions in ZnSe because the ionization energy ED of donors ranges between 25 and 28 meV [27] so that at room temperature most of the donor atoms have contributed their electrons to delocalized conduction band states [51]. Conversely, when kBT ~szED strong localization of the dcctrons is expected either at the donor sites or, as the concentration approaches the metal—insulator transition, in an impurity band [52]. We will begin by summarizing typical results of the transport properties of these In-doped layers. The dots in fig. 9 represent results of Hall effect measurements as a function of temperature T for a typical n-type doped ZnSe layer. One can see that the values of (RHeY where R 11 is display the Halla coefficient and e is the electron charge, ~,
layers
pronounced dip in the neighborhood of 50 K. The presence of this dip and the overall dependence on T correspond to a “two-carrier system” in which impurity band conduction occurs along with transport in the conduction band [52]. For T above the one corresponding to the dip in (R~e)~,transport is dominated by free carriers populating delocalized states of the conduction band. For T < 50 K, conduction proceeds through localized states of an impurity band. These qualitative conclusions can be rigorously obtained by appropriate modeling of the Hall effect results [52]. In K the
particular, it can be shown thatn(T) for Tin 100 free carrier concentration the conduction band is given by n(T) = (R~eY The decrease of n(T) with decreasing T from 300 to 100 K is due to the progressive freeze out of the electrons. A model fit to n(T) in this range of T with the standard expressions of semiconductor statistics yields 3 witha donor a compensation concentration ratioND of of 1.7 >< 10i7 cm 0.08 and an activation energy of 14 meV to promote carriers into the delocalized conduction states [51]. The so-determined ND is smaller than the predicted donor concentration of 4 X 3 for the Mott insulator—metal transition10i7 in cm [52]. Therefore, the impurity band comes ZnSe about from partial overlap of donor wave functions or broadening of donor orbitals by potential fluctuations. Conductivity and mobility measurements reinforce the interpretation of the Hall effect data; in particular the presence of localized carriers at low T. Between 300 and 100 K the functional dependence of the conductivity a’ on T takes the form In a’ a T which is the expected behavior for free carriers. Below 50 K a dependence of the type ln a’ a T 1/4 is measured, which identifies the conduction mechanism in the impurity band as variable range hopping [521.The mobility as a function of T also shows the transition from one conduction regime to another. Room ternperature mobility for the sample of fig. 1 is 400 cm2/V s, which is a state of the art value for a doped sample in the low 3 range. 10i7 donors cm steadily With decreasing T the mobility increases 2/V s at and maximum valueofof the 800 cm T reaches 100 K. aFurther cooling sample de—
~,
—
K. Shahzad et al.
/ Doped zinc selenide epitaxia! !ayers
creases the mobility, but at a smaller rate than the one expected if free carriers are subject to impurity scattering after phonons have been quenched [51]. Below 50 K, the mobility settles slowly to a more or less con~tant value of 40 cm2/V s. This is much larger than any calculated mobility value for non-degenerate ZnSe with free carrier concentration determined by semiconductor statistics and impurity scattering as the main process inferred from the conductivity data for scattering process. It is in line with the transport localized electrons. Next, we will discuss Raman scattering results
~j ~
undoped
layers excited for selected values ofthe T.wavelength These spectra modes, were arise from which scattering with are allowed 488.0 bynm by long laser scattering radiation selecLOand tion rules for the geometry described in section 2 [18]. The undoped layer is semi-insulating and its free carrier concentration at 300 K is less than 1014 electrons cm3. This doped sample is the one whose transport properti~swere described above and given in fig. 9. The striking feature of the data in fig. 10 is the shift to higher frequencies of the spectra corresponding to the doped layer when compared with those of the reference sample. In addition, these spectra are substantially broader. At the lowest T, they display a pronounced asymmetry towards the high energy side. The measured peak positions for the doped and The control samples are of plotted against Tsample, in fig. 11. LO-frequencies the undoped which we will label ~Lo in this discussion, show a clear softening with increasing T. This effect, which has been seen already in ZnSe and other semiconductors, arises from anharmonic contributions in the phonon—phonon interaction [53— 55]. In the plot of the frequencies of the doped layer, Q~Sed, we can distinguish a kink at around 60 K. Contrary to the well established softening of ~LO with T, QdoPed increases monotonically LO between 60 and 300 K. Below 60 K, it does not change with T within the experimental uncertainty although it remains slightly higher than ~Lo’ We note that the kink in the temperature
I\/
ZnSe
\
\doped
/
300 K
/
C’)
z
th
\
/
—
/ CI)
z LU z
on theRaman order same spectra layers. ofFigure doped and 10 compares undoped ZnSe first
29
z —
~iI /
N
113K! —
240
250
260
270
STOKES SHIFTS (cm1) Fig. 10. Raman spectra of longitudinal optical phonons for doped (i.7x io’~donors cm3) and undoped ZnSe layers excited with 488.0 nm laser radiation. The scattering geometry is given by z(x + y,x + y)~and this notation is described in section 2. The peaks for the doped layer are shifted to higher frequencies The line shape for the doped sampleatatall13temperatures. K displays asymmetrical broadening.
dependence of 12~ed occurs at1comparable temshown in fig. 9. peratures thejust dip as in the (RHeY We argue as that, dip in (R 11eY mdicates a transition from localization to delocalization in the electron system, the kink in the frequency plot of ~~Cd establishes a transition between two different modes of electron—phonon coupling in the doped layers. We postulate that above 100 K the upward renormalization of (2~’~is due to the coupling of the phonons with the charge density fluctuations of the free carriers, which populate the conduction band with a density n(T). The Raman modes of the doped sample in fig. 10 for 100 < T <300 K are then identified as the phonon-like component of the plasmon—phonon coupled
30
K. Shahzad et a!.
/ Doped zinc selenide epitaxial
“optical values” of n(T) and by comparing them with the transport determination. With 12LO and fl~o~ed taken from fig. 11 for T> 100 K, measured values of ~TO as a function of T (not shown here, but we have established that the
ZnSe 258
x
E >—
0
xxx
splitting is 46.5 cm~ and is independent of T), ~ = 6.1 and m* = 0.16 [23], we obtamed the values of n(T) represented by the open triangles in fig. 9. The optical and transport determinations corn~Lo~~To
256
0 Lii U-
+
254 .
undoped doped
•
252 100
200
300
TEMPERATURE (K) Fig. 11. Peak frequencies of the Raman spectra measured in fig. 10 as a function of temperature. The frequencies of the undoped layer are labeled 17L0 in the text. In the case of the doped sample the notation Q~7d is being used. Notice the kink in the plot of the doped frequencies. It happens at around the same temperature as the dip in (R 0eY seen in fig. 1. Between 13 and 50 K a discrete-continuum interaction renormalizes the doped frequencies. Above 100 K the plasmon—phonon coupling is responsible for their substantial upward shift,
modes Under this assumption, given by[17,18]. the relationship p
layers
— fldoPed2 = Qdoped2 LO La LO ~2 — fldoPed2’ TO LO
11~ed
is (3)
where ~TO is the temperature dependent frequency of the transverse optical TO-phonons in the undoped sample and ui~is the plasma frequency of the charge density excitations of the free electrons in the doped sample. The plasma frequency is related to the temperature dependent electron concentration n(T) through the expression n(T)e2 = (4) ~) e,~m* in which e~ and m* are the high frequency dielectric constant and the conduction band effective mass, respectively. Therefore, one way to test the validity of the above stated description is by determining with the help of eqs. (1) and (2)
pare very favorably, thus justifying the assumptions about the phonon—plasmon coupling. Minor quantitative disagreements (less than 10%) are probably the result of not having considered any dependence of m* on n(T). The overall good agreement in fig. 9 is also an indication that is not mobility limited for ND up to low 1017 donors cm3 and that wave vector non-conservation does not play a leading role in the light scattering process of these doped layers [18]. The
~Lc~’
increasing splitting ~ with T above 100 K in fig. 11 is a clear manifestation in the optical properties of the ionization of the donors with the associated population of conduction band states by electrons. We have also studied samples with lower concentrations (lots between
K. Shahzad et al.
/ Doped zinc se!enide epitaxial !ayers
discrete state of the LO-phonon couples to a continuum of excitations of the now localized electrons. Insight into the nature of the electronic excitations producing the continuum was gained by performing Raman experiments with laser energy below but close to the fundamental gap of ZnSe. We present in fig. 12 Raman results obtamed with laser light at 457.9 nm (2.71 eV) which is in near resonance with the low energy side of a weak and featureless luminescence background below the gap. In addition to the slightly upward shifted LO-phonon line at 256.8 cm a broad band appears in the Raman spectrum. This extra Raman signal was not observed with the 488.0 nm laser excitation of fig. 10. It corresponds to electronic Raman scattering from transitions between bound donor states. This process, which is depicted in the inset to fig. 12, provides the electronic continuum that interacts with the phonon. The peak riding on the broad ~,
1CB
I
band at around 210 cm_i arises from scattering by the TO-phonons due probably to some relaxation in the selection rules. The two lower energy features at around 175 and 197 cm’ can be identified with the is—2p and ls—3s transitions of the donors [27]. Scattering to the right-hand side of the TO line spreading all the way up to the LO-peak is due to transitions from the is fundamental state to levels at or above the ionization threshold of the donors. For an isolated In donor ED is 28.2 meV (225 cm I) [27]. The electronic transitions that in isolated donors will be of discrete nature are broadened in the doped samples for the same reasons as that the impurity band is formed. It is intriguing that spectroscopical signatures of individual donors are observed though the transport proceeds through an impurity band. The high energy side of the electronic continuum overlaps with the LO-phonon line and a weak antiresonance is observed just before the LO-peak maximum. The asymmetrical broadening of the phonon line already seen in the data of fig. 2 is apparent in fig. 12. In the framework of the discrete—continuum interaction the renormalization of the phonon frequency in the doped sample Q~~Cd is given by
DOPED ZnSe
I ~—
31
DVELS
[17,181:
2R(E)
(5)
Q~edu1LO+V where V is the matrix element for the coupling
between the discrete phonon state and the continuum of electronic excitations and R(E) is the Hilbert transform of the density p(E) of continuum states. To perform a quantitative compariI
100
I
200
300
Stokes Shifts (cm-1) Fig. 12. Raman spectrum of a doped ZnSe layer (1.7x iOi7 donors cm3) taken at 13 K with laser photons at 457.9 nm in the scattering geometry of fig. 2. The longitudinal optical phonon peak at 256.8 cm is prominently seen and towards lower frequencies the broad band is resonant electronic Raman scattering by bound electrons. The inset depicts such a process, The asymmetry in the phonon line shape seen in fig. 2 is also observed here. A weak antiresonance on the low energy side of the phonon peak can ,be distinguished. The weaker structure at around 295 cm is due to the longitudinal optical phonon of the GaAs substrate. It is observed because the ZnSe layer is transparent to the incoming laser radiation,
son between theory and experiment we need a microscopic model for V and R(E) which we lack presently. Renormalization of phonon frequencies due to interactions with bound electron or hole excitations of the type ls—2s have been reported in doped semiconductors [10,17,56,57]. .
.
A convenient way to observe these interactions is in the phonon replica of photoluminescence cither from recombinations of neutral donor and acceptor bound excitons or from the so-called “two-electron” or “two-hole”transitions. In most cases, the phonon energy is less than the ls—2s energy and the result of the interaction is the creation of bound phonons with smaller frequen-
32
‘
K. Shahzad et a!.
/ Doped zinc selenide epitaxial
cies in the neighborhood of the impurity. The situation considered here is quite different as the LO-phonons remain unbound and their frequencies increase. Another important point is that the electronic continuum is not provided by intraband transitions of free carriers at the zone center, because symmetry considerations preclude in such case coupling with zone center phonons [181. The kind of discrete—continuum interaction considered here can lead to the asymmetries measured in the LO-lineshapes in figs. 10 and 12 because of quantum-mechanical interference. Indeed, the Fano—Breit—Wigner profile provides a good fit to these asymmetrical lineshapes. The dependence of the Fano fitting parameters and behavior of the electronic scattering on laser wavelength and ND go beyond the scope of the present work and will be dealt with separately, The interaction with the electrons not only renormalizes the frequency of the phonons but changes their lifetimes as well. The latter effect shows up in broader Raman lines for the doped samples. The breadths at half maximum of the Raman peaks in fig. 10 corrected for spectrometer resolution and those of samples with smaller ND are plotted against T in fig. 13. The correlation between ND and the symbols used in fig. 13 is explained in the figure caption. The increase with T of the LO-phonon line width, FLO, in the undoped sample is due to anharmonic decay into acoustic phonons [53—55].In the doped samples the line width, ~ also increases with T and for a given T with ND. The contribution to F~gP~ of the electron—phonon interaction is determined by the broadening p~~d FLO’ which LO excludes the thermal effects. An inspection of the behavior of p~ed in fig. 13 also indicates the presence of two regimes depending on the degree of electron localization. When the electrons are localized, Fd0P~~ FLO remains more or less conLO stant with increasing T. On the other hand, when free electrons start to populate the conduction band as evidenced by Q~J~~dand Hall effect, F~ij7d deviates markedly from FLO and FLO doped FLO substantially increases. The value of T at which this change in behavior takes place increases with decreasing ND. In fig. 13 we observe that it occurs at around 50, 100 and 150 K for —
—
—
layers
ZnSe 15
• undoped oAX doped
‘~E -~-
i
10
a
6 A
LU Z
A A
5
A
X
0
x~
A
A A A 0 000
•
S
o•
•‘
S
I
I
100
200
300
TEMPERATURE (K) Fig. 13. Line widths of the Raman peaks of ZnSe layers measured with 488.0 nm laser excitation as a function of temperature. The data have been corrected for instrumental resolution. In the text, the linewidths of the undoped layer are TLQ and those of the doped samples ~~f°’1.The labeledconcentration ND of the doped samples is the followdonor ing: crosses
2.1~x J~dIIpcd circles
—
LO
I.7x iO’~,open triangles 3.6X i0~’ and open iO’~donors cm3, respectively. The broadening increases with temperature when electrons pop-
LO
ulate conduction band states and remains constant when the
electrons are subject to localization.
ND = 1.7 x 1017, 3.6 x lO~~ and 2.1 x 1016 donors cm~3,respectively. The increasing pdOp~ T with T in the LO ~LO regime of delocalized electrons can be attributed to plasmon damping by electron collisions [18]. The collision rates become larger as more dcctrons are thermally promoted to the conduction band, thus producing a larger phenomenological plasmon damping with T. This damping smears the plasma response of the electron gas which in turn broadens the coupled phonon—plasmon mode. When the electrons are localized, the low T broadening is the imaginary part of the phonon self energy due to the interaction with the continuum. The discrete-continuum interaction theory predicts for the broadening [17,181, F(~0Ped F = -~V2p( E). (6) —
—
LO
LO
We have found an empirical correlation between the broadening F~p7d FLQ at 13 K and the electron concentration n(300 K) at room temperature. The plot in fig. 14 establishes an almost —
K Shahzad et a!.
/ Doped zinc se!enide epitaxia! layers
3
doped ZnSe
0
Z
•
2
LU
a
• I—
33
them are the evolution of the electron—phonon interaction as the insulator—metal transition is approached and the electron—electron scattering in the electron gas. Also, comparisons of activation energies for transport processes and the spectroscopical determinations of bound dcctronic transitions should help in the understanding of the formation of impurity bands.
S
o • I
1016
6. Li acceptors in zinc selenide: electronic Ra-
I
man scattering and hole—phonon coupling
1017
3)
ROOM TEMPERATURE n (cm’ Fig. 14. Low temperature (13 K) broadening of the Raman lines of doped ZnSe layers plotted against the free electron density at room temperature. The Raman data corresponds to the 488.0 mm laser excitation and have been corrected for instrumental resolution. An empirical correlation is found.
linear relationship in the semilogarithmic graph. If we assume that V does not change with doping and consider that in lightly compensated samples n(300 K) ND, the plot in fig. 14 will indicate that p(E) is not linearly proportional to ND. The departure from a linear dependence particularly at the highest ND can be taken as additional indication of wave function overlap [17]. To summarize this section, we focused on the general characteristics of the coupling between electrons and longitudinal optical phonons in ntype ZnSe layers, which were doped below the
states Mott We nates showed criterion when in thethat free conduction for phonon—plasmon electrons the metal—insulator band. populate Acoupling discrete—condelocalized transition. domilocalized in donor states and gives rise to asymtinuum interaction prevails when electrons are metrical Fano lineshapes in the phonon spectra. In addition, we found empirical correlations which may be of value for future characterization of these materials. We have determined carrier concentrations, established a relationship between phonon line widths at low temperature and carncr concentration and observed Raman scattering from donors. These results open possibilities for further investigations of fundamental properties of wide band gap semiconductors. Among
Low temperature Stokes—Raman spectra for selected samples with two different Li doping levels are displayed in fig. 15. The spectra were excited with laser photons at 4880 A, whose energy of 2.54 eV is smaller than the band gap of ZnSe of 2.82 eV at 12 K. The spectral range shown here lies above the region of the first and second order phonon Raman scattering in ZnSe. With the exception of the peak at 512 cm~1 labelled 2L0 all other features are unique to the Li-doped samples. They correspond, as discussed ___________ _____________________________ Li DOPED ZnSe 2L0
~
XL 4880 A
T 12K
A1S2.
~ ls-3s
500
I
600
I
700
EA
I
I
800
STOKES SHIFTS
900
7.4 x 10~
I
1000
(cm-i)
Fig. 15. Electronic Raman scattering spectra from transitions of holes bound to Li acceptors displayed for two values of the net acceptor concentration NA — N The transition energies are summarized in table 2 and the0.individual processes are identified in the text. The 2L0 scattering calibration,
serves
as internal
34
K Shahzad et a!.
/ Doped zinc se!enide epitaxia! layers
below, to ERS from holes bound to substitutional Li acceptors. Their intensity and complexity of the spectra are a very strong function of the doping level. The 2LO peak arises from overtone scattering by two LO-phonons at the zone center. In the first-order spectra, LO-phonon peak is measured at 256.2 cm~ The 2L0 intensity remains constant when the measurements are done under similar conditions and laser powers. It can therefore be used as an internal calibration for the ERS intensities, The Raman nature of the signal in fig. 15 was verified by the unchanged frequency position of the peaks with respect to the excitation energy upon excitation with different laser frequencies. However, the scattering strength of the ERS peaks depends on the exciting photon energy, in line with the expectation that the ERS is a strongly resonant process [17].the Asintensity the lasertends approaches the bandgap energy, ~.
to rise but stronger PL backgrounds severely affect the line shapes. With laser photons well above or below the gap, only the strongest of the ERS features in fig. 15 were observed for the samples with the largest doping levels. The situation depicted in fig. 15, and at other excitation energies within 100 meV of the 4880 A photons represents a good compromise between resonant enhancement and the unwanted effects of competing PL processes. This situation resembles closely the one in GaAs, in which below-band-gap laser light of selected energies is preferred for investigation of ERS from acceptors [58—611. The excited states of the Li acceptors in ZnSe have been investigated theoretically within the framework of the effective-mass approximation [34] and measured experimentally with selective pair PL in bulk samples [35]. We used these results to identify the individual ERS processes giving rise to the peaks in fig. 15. Following the notation in ref. [34], the following assignments are made: the strongest peak appearing at 670 cm~ and labeled ls—2s is due to transitions from the 1s3/2 ground state to the 2s3/2 excited state. The shoulder at 698 cm~ corresponds to excita2P tions of the ground state to the 5/2 level. The ls—3s feature arises from scattering between the and 3s3/2 levels. The ground and excited
states mentioned above are made of Bloch functions from the F8 valence bands multiplied by the corresponding hydrogenic envelope functions. The EA structure peaking at 915 cm~ is ascribed to transitions from the is 3/2 state to the continuum of delocalized states at the top of the valence band. Its peak energy measures directly the ionization energy of the Li acceptors. The p’eaks at 540 and 575 cm~are very close in energy to overtone scattering by two transverse and two longitudinal phonons of GaAs, respectively. One could attribute them to phonon scattering from the substrate, which is seen because the incoming photons are not absorbed by the ZnSe. However, the intensities of these peaks clearly increase with Li doping. Also, we failed to observe such strong scattering in bare GaAs wafers. Therefore we attribute these two features 1can to the presence be ascribed to of theLilSas well. 2P The 575 cm~ 3/2~ 3/2 transition. The assignment of the peak at 540 cm~ is more complicated, because its frequency does not conrespond to any of the excited states of Li or transitions between excited states. One possibility is that this peak is due to scattering by the local vibrational mode of Li. A simple estimate predicts the local mode of Li at around 615 cm fairly close to the measured frequency. The ERS results with the exception of the 540 cm peak are summarized in table 2 where they are cornpared to the theoretical and experimental values obtained by selective PL [34]. Good agreement is found between the different sets of values for the bound transition energies and the ionization en-
Table Excitation energies in meV of holes bound to Li acceptors in ZnSe. The transitions are from the ground iS7 2 state to shallower S and P states and to the valence bands giving the ionization energy EA. The electronic Raman scattering (ERS) values result from this work. The photoluminescence (PL) and theoretical values were taken from ref. [151 Transition ERS PL Theory iS1/2~2S1/2 83 82.6 82.7 1S3/2_2P5/2 86.5 85.8 84 1S3/2—3S3/2 1S3/2_2P3/2 EA
98.5 71.3 113.4
97.8 72.9 114
97.5 71.9
/ Doped zinc se!enide epitaxial
K Shahzad et a!.
Li-DOPED ZnSe T~i2K
•
—
0.45
of about 5.2 x 1016 cm~3 using the standard semiconductor statistics [65], with 114 meV for the ionization energy of the Li acceptors [8,211. The salient feature of the data is that, at 300 K, the Raman peak of the ZnSe : Li sample, when compared to the intrinsic reference sample, ap-
•
‘I ~“
35
Figure 17 displays the Stokes spectra near the LO frequency for intrinsic (undoped) and p-type doped ZnSe : Li layers. In this ZnSe : Li sample, 3from C—V, which NA ND wasto1.2 X 1017 corresponds a free holecm~ concentration at 300 K
0.55 0.50
layers
0.40
j,
0.35 0.30 025
III
5
10I
I
I
15
NA- ND (1016cm3) Fig, 16. Normalized intensities of the ls—2s peaks with respect to the one of the 2L0-phonon scattering as a function of NA — N 0. A linear dependence is obtained,
ergy of the Li acceptors directly measured by ERS. The peak intensity of the ls—2s transition was normalized to the intensity of the 2LO line. Figure 16 displays this normalized value plotted against NA N0. A linear dependence of the normalized intensities is found, which confirms theoretical predictions for the concentration dependence of the ERS cross-section [17]. In addition, it reproduces similar findings realized in GaAs [59], and supports the assertion that the scattering cross-section depends on the net acceptor concentration [58,611. The observation of ERS and its relationship to NA ND lead to the conclusion that Li impurities are mainly incorporated as substitutional acceptors. If comparable or larger amounts of interstitial Li were present and behaved as donors then the Li acceptors would be neutralized and ERS would not be observed. Contrary to the GaAs case in which the holes were produced by photoexcitation [58—61], in the present case, the holes are present by virtue of doping and are bound to acceptors at low temperatures. Increasing the sample temperature produces a reduction of the ERS intensity because the thermal ionization of the Li acceptor frees the bound holes. In the heaviest doped sample, we observed the ls—2s peak up to ternperatures of 150 to 200 K. —
pears slightly broadened. At shifted 12 K, on to the higher otherfrequencies hand, no such and differences are observed. The complete dependence on temperature T of the frequency shift ~w and broadening ~y for this ZnSe: Li layer is shown in fig. 18. The decrease of z~wand ~y with temperature is gradual and both are zero below 100 K. Also, at a constant T the position and line width of the ZnSe : Li Raman peak are a function of NA ND. Figure 19 shows this dependence of ~w and z~yat 300 K for a series of ZnSe : Li layers. Both parameters go smoothly to zero with decreasing Li doping and similar to the dependence on temperature the effect appears more pronounced for ~y. The renormalization of the phonon properties shown in figs. 18 and 19 is due to the interaction of the LO-phonon with the plasmon of the free holes. The underlying physical aspects of such coupling have been extensively discussed in the literature [18]. The plasma frequency w~= —
—
~ 300K
12K
~ *ITR6ISIC
~
Li-DOPED I
240
250
260
6240
I
250
280
STOKES SHIFTS (cm-i) Fig. 17. Raman spectra due to LO-modes in undoped 3). and Li-doped ZnSe:Li layers(N~— ND i2~iOi7 cm
36
K. Shahzad et a!.
/ Doped zinc selenide epitaxial U ~
ZnSe:Li
0.9
!ayers
Li-DOPED ZnSe ROO
EXPE
0.7 — 5 E°~
5
_
>-
U
•
0.3
•
z
~
b,m
U 0.1
U
.
i•
-0.1 0 .
.
‘S
I
I
240
I
100 200 300 Temperature (K)
Fig. 18. Frequency shift ~a and broadening ~.y versus ternperature for the ZnSe : Li spectra shown in fig. 17.
(4 2p/m*)i/2, where e is the electronic 11e m* is the effective mass of the holes and charge, = 6.1 is the high frequency dielectric constant [51], is for our samples smaller than the LO-phonon frequency in undoped ZnSe as shown below, Therefore, the ZnSe : Li Raman peaks correspond to the so-called L~ component of the plasmon—phonon coupled modes [18]. Under the conditions of our experiment the frequency distni-
250
260
STOKES SHIFTS (cm1) Fig. 20. Line shape fit to the 300 K ZnSe : Li spectrum shown in fig. 1.
bution of the scattered light 1(w) by the L~mode near w(LO) is described by [18]: 1(w) =Im[—1/(w)], (7) where ~(w) is the frequency dependent dielectric function. It includes the phonon and plasmon contributions according to the relationship (w)
[
=e
1~ 1 +
w~ 1 2 + iwy~,]’
w~ w~ —
—
—
iwy1
—
w (8)
where w~is the frequency of the transverse optical phonon (not affected by the plasmon), is the intrinsic line width of the LO-phonon in the undoped material and h/ye is a phenomenological scattering time in the hole gas [18]. Figure 20 displays the fitting to the 300 K ZnSe : Li line-
I
~
ZnSe:Li 0.8
300 K
Ar
shape shown in fig. 17, which was performed with (1) and (2) using w~,and as adjustable parameters with w 1 = 252 cm~, w1 = 205 cm_i and Yi = 4.7 cm_i, respectively [19]. The calcu-
0.6
eqs.
—
0.4
.
~.
. 0.2
S
U
•
U
0 0
2
4
lated curve reproduces the overall shape for = 35 cm~’and y0 = 400 cm From3,this value which is of w~,we calculate p = 4.9 x lO°~ cm in excellent agreement with the value inferred from C—V. On the other hand, the estimate yields a hole mobility of 38 cm2/V s, which also agrees with experimental and theoretical values quoted recently for p-type ZnSe [51,62—64]. Similar line shape analysis for other T or Li doping level have also reproduced the experimen-
6
I
I
8
10
12
14
NA-No(i0’6cm~)
Fig. 19. Dependence of ~w and ~y at 300 K on NA— ND.
K Shahzad et a!. I
/ Doped zinc selenide epitaxial
layers
37
.
which can be used for contactless characterizalion of p-type ZnSe. The measured values for
A
ZnSe:Li
bound excitations and ionization energies of Li acceptors in ZnSe are in good agreement with previous reports in bulk samples. In addition, we also performed a Raman scattering characterization of the phonon—hole—plasmon interaction in p-type ZnSe layers. A clear signature of the effect was found in the renormalization of the phonon properties. We have shown that values for con-
0.8
. 0.6
A
• 4,
0.4 0 6.
A
•
0.2
centration and mobility of the holes can be obtained with an all optical technique.
0~ 0
I
I
I
I
1
2
3
4
5
6
3
Hole Concentration p C 1016 cm Fig. 21. Hole concentration dependence of ~y. Symbols are explained in the text.
tal spectra. In general we find that > wi,, which indicates that the hole plasmon is overdamped and explains why the stronger effect of the coupling is on ~y [18]. The decreasing z~wand ~y with T or NA ND are understood in terms of a decreasing p either by carrier freeze-out or doping, respectively [19]: both phenomena produce the same changes in the Raman spectra. This can be seen in fig. 21, where the dependence of ~Y on p is illustrated. The triangles correspond to the values obtained from the NA ND dependence and the circles from the behavior on T. The solid line through the experimental points was drawn with the expression z~y= 0.17 x —
—
10i6~
This empirical relationship can be applied for fast quantitative characterization of the non-degenerately doped p-type ZnSe. In summary, we have characterized Li doped MBE layers with ERS. Using below-band-gap laser excitation, very detailed ERS spectra could be measured without the interfering effects PL backgrounds. The ERS results were normalized and related to transport properties. A linear relationship was found between the net acceptor concentration and the strength of the ERS signal,
References [11 T. Yao, in: The Technology and Physics of Molecular Beam Epitaxy, ed. E.H.C. Parker (Plenum, London, 1985) p. 313. [21 J.M. DePuydt, H. Cheng, i.E. Potts and 5K. Mohapatra, J. Appl. Phys. 62 (1987) 4756. [3] H. Cheng, J.M. DePuydt, i.E. Potts and T.L. Smith, AppI. Phys. Lett. 52 (1988) 147; J.M. DePuydt, MA. Haase, H. Cheng and i.E. Potts, App!. Phys. Lett. 55 (1989) 1103. [4] T. Yasuda, I. Mitsuishi and H. Kukimoto, Appl. Phys. Lett, 52 (1988) 57. [5] M. Migita, A. Taike and H. Yamamoto, J. Appl. Phys. 68 (1990) 880. [61A. Ohki, N. Shibata, K. Ando and A. Katsui, J. Cryst. Growth 99 (1990) 413. [7] K. Ohkawa and T. Mitsuyu, J. App!. Phys. 70 (1991). [8] J.L. Merz, K. Nassau and i.W. Shiever, Phys. Rev. B 8 (1973) 1444. [9] P.J. Dean, W. Stutius, G.F. Neumark, B.J. Fitzpatrick and RN. Bhargava, Phys. Rev. B 27 (1983) 2419. [10] K. Shahzad, B.A. Khan, Di. Olego and D.A. Cammack, Phys. Rev. B 42 (1990) 11240. [ii] MA. Haase, H. Cheng, J.M. DePuydt and J.E. Potts, i. App!. Phys. 67 (1990) 448. [12] C.G. van de Wa!le and D.B. Laks, J. Lumin. 52 (1992) 1. [13] R.M. Park, M.B. Troffer, C.M. Rouleau, J.M. Depuydt and MA. Haase, App!. Phys. Lett. 57 (1990) 21. [14] MA. Haase, J. Qiu, J.M. Depuydt and H. Cheng, App!. Phys. Lett,, to be published. [15] K. Akimoto, T. Miyajima and Y. Mori, Phys. Rev. B 39 (1989) 3138. [16] K. Shahzad, KS. Jones, PD. Lowen and R.M. Park, Phys. Rev, B 43 (1991) 9247. [17] M.V. Klein, in: Light Scattering in Solids, ed. M. Cardona (Springer, Heidelberg, 1975) p. 147, and references therein.
38
K. Shahzad et a!.
/ Doped zinc se!enide epitaxia! layers
[18] G. Abstreiter, M. Cardona and A. Pinczuk, in: Light Scattering in Solids IV, eds. M. Cardona and G. Guntherodt (Springer, Heidelberg, 1984) p. 5, and references therein, [19] Di. Olego, T. Marshall, J. Gaines and K. Shahzad, Phys. Rev. B 42 (1990) 9067. [20] D.i. Olego, J. Petruzzello, T. Marshall and D.A. CamApp!, T. Phys. Lett., to be Cammack, published, K. Shahzad and [21] mack, D.J. Olego, Marshall, D.A.
[40] In order to compare our value of A~(2.7895 eV) for the epilayer with that of the oxygen-doped layer in ref. [15], we must add a 3.7 rneV correction factor to take into account the shift due to increased thermal mismatch as a result of higher temperature annealing. This factor can be readily obtained from subtracting the measured value of I~ (2.7794 eV) from the bu!k (unstrained) value 4’, and (2.7831 Weby find to a first A are eV). shifted thisthat, amount fromorder, their FX~,D bulk values.
J. Petruzze!lo, App!. Phys. Lett, 58 (1991) 2654. [22] A.Y. Cho, in: The Technology and Physics of Molecular Beam Epitaxy, ed. E.H.C. Parker (Plenum, London, 1985) p. 1. [23] H. Hartmann, R. Mach and B. Selle, in: Current Topics in Materials Science, Vol. 9, ed. E. Kaldis (North-Ho!land, Amsterdam, 1982) p. 1, and references therein. [24] K. Shahzad and D.C. Cammack, Appl, Phys. Lett, 56 (1990) 180. [25] K. Shahzad, Phys. Rev. B 38 (1989) 8309, [26] K. Shahzad, D.J. Olego and D.A. Cammack. Phys. Rev, B 39 (1989) 13016. [27] P.J. Dean, D.C. Herbert, C.J. Werkhoven, B.J. Fitzpatrick and R.N. Bhargava, Phys. Rev. B 23 (1981) 4888. [28] Z.L. Wu, J.L. Merz, C.J. Werkhoven, B.J. Fitzpatrick and RN. Bhargava, AppI. Phys, Lett. 40 (1982) 345. [29] Discrete donor—acceptor pair lines associated with Li and Na acceptors have been reported. See, for example, ref. [8] as well as: B.J. Fitzpatrick, Ci. Werkhoven, T.F. McGee, P.M. Harnack, S,P. Herko, RN. Bhargava and P.J. Dean, IEEE Trans. Electron Devices ED-28 (1981) 440, for Li-related pair lines; and G.F. Neumark, S.P. Herko, T.F. McGee and B.J. Fitzpatrick, Phys. Rev. Lett. 53 (1984) 604, for Na-related
[41] RN. Bhargava, J. Cryst. Growth 59 (1982) 15. [42] Y. Yafet and D.G. Thomas, Phys. Rev. 131 (1963) 2405. [43] J.L. Merz, Phys. Rev. 176 (1968) 961; J. App!. Phys. 42 (1971) 2463. [44] J.J. Hopfield, D.G. Thomas and R.T. Lynch, Phys. Rev. Lett. 17 (1966) 312; D.G. Thomas, J. Phys Soc. ipn. Suppl. 21(1966) 265. [45] The term isoelectronic “acceptor” was originally coined to describe a model in which a hole is bound to the resulting negative ion through Coulomb potential in an acceptor-like wave function. Likewise, an isoelectronic hole trap [46] will produce an isoelectronic “donor”. With the exception ofa claim made in ref. [15], the donor and acceptor nature of the isoelectronic impurity has not been substantiated by electrical measurements. However, they may provide trapping states as, for example, has been reported by Aten et a!. [47] who showed that CdS:Te acts as a hole trap resulting in n-type photoconductivity. In all the implanted samples that we studied, we did not find any evidence of acceptor-like behavior from the electrical measurements. In addition, we did not observe any clear indication of donor_acceptor pair bands pointing to oxygen behaving like a typical acceptor in ZnSe. [46] See, for example: Q. Fu, D. Lee, A.V. Nurmiko, L.A. Kolodziejski and R.L. Gunshor, Phys. Rev, B 39 (1989) 3173. [47] AC. Aten, J.H. Haanstra and H. de Vries, Philips Res. Rept. 20 (1965) 395. [48] R.E. Dietz, D.G. Thomas and J.J. Hopfield, Phys. Rev. Lett. 8 (1962) 391. [49] F.A. Trumbore, M. Gershenzon and D.G. Thomas, App!. Phys. Lett. 9 (1966) 4. [50] J.D. Cuthbert and D.G. Thomas, J. App!. Phys. 39 (1968) 1573. [51] H.E. Ruda, J. Appi. Phys. 59 (1986) 1220. [52] T. Marshall and J. Gaines, AppI. Phys. Lett. 56 (1990) 2669. [53] J.L. La Combe and J.C. Irwin, Solid State Commun. 8 (1970) 1427. [54] M. Balkanski, R.F. Wallis and E. Haro, Phys. Rev. B 28 (1983) 1928. [55] J. Menendez and M. Cardona, Phys. Rev. B 29 (1984) 2051. [561 C.H. Henry and J.J. Hopfield, Phys. Rev. B 6 (1972) 2233.
pair lines.
[30] For a review, see, for example: P.J. Dean, in: Progress in Solid State Chemistry, Vol. 8, eds. JO. McCaldin and G. Somarjai (Pergamon, New York, 1973) p. 1. [31] A.T. Vink, R.L.A. van der Heyden and JAW. van der Does de Bye, J. Lumin. 8 (1973) 105. [32] G.F. Neumark, Phys. Rev, 29 (1984) 1050. [33] This is in contrast to what the authors in ref. [9] found for their ZnSe sample grown by liquid-phase epitaxy i.e. intense above-band-gap excitation yielded 2S satellite optimally.
[34] NO. Lipari and A. Baldereschi, Phys. Rev. B 8 (1973) 2696; B 9 (1974) 1525. [35] H. Tews, H. Venghaus and P.J. Dean, Phys. Rev. B 19 (1979) 5178. [36] D.C. Herbert, P.J. Dean, H. Venghaus and J.C. Pfister, i. Phys. C 11(1978) 3641. [37] H. Venghaus, Phys. Rev. B 19 (1979) 3071. [38] P.J. Dean, H. Venghaus, J.C. Pfister, B. Schaub and i. Marine, J. Lumin. 16 (1978) 363. [39] K. Ohkawa, T. Mitsuyu and 0. Yamazaki, Phys, Rev. B 38 (1988) 12465.
K Shahzad et a!.
/ Doped zinc se!enide epitaxia! layers
[57] P.i. Dean, H. Venghaus, J.C. Pfister, B. Schaub and i. Marine, i. Lumin, 16 (1978) 363. [58] K. Wan and R. Bray, Phys. Rev. 32 (1985) 5265; R. Bray, K. Wan and J. Parker, Phys. Rev. Lett. 57(1986) 2434. [59] J. Wagner, M. Ramsteiner, H. Seelewind and J. Clark, J. App!. Phys. 64 (1988) 802. [60] i. Wagner, H. Seelewind and U. Kaufmann, AppI. Phys. Lett. 48 (1986) 1054. [61] T.D. Harris, M. Lamont Schnoes and L. Seibles, Anal. Chem. 61(1989) 998.
39
[62] A. Yahata, H. Mitsuhashi, K. Hirahara and T. Beppu, Jpn. i. App!. Phys. 29 (1990) L4. [63] A. Taike, M. Migita and H. Yamamoto, AppI. Phys. Lett. 56 (1990) 1989. [64] J. Ren, K.A. Bowers, B. Sneed, DL. Dreifus, J.W. Cook, iF. Schetzina and R.M. Kolbas, App!. Phys, Lett. 57 (1990). [65] T. Marshall and D.A. Cammack, J. App!, Phys, 69 (1991) 4149.