1:
2: \documentstyle[twocolumn,aps,prb]{revtex}
3: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4: %TCIDATA{OutputFilter=LATEX.DLL}
5: %TCIDATA{LastRevised=Tuesday, September 02, 2003 19:18:22}
6: %TCIDATA{<META NAME="GraphicsSave" CONTENT="32">}
7: %TCIDATA{Language=American English}
8:
9: \begin{document}
10: \title{{\bf Nitrogen local electronic structure in Ga(In)AsN alloys}\\
11: {\bf by soft-X-ray absorption and emission: Implications for optical
12: properties }}
13: \author{V.N. Strocov,$^{1,\ast }$ P.O. Nilsson,$^{2}$ T. Schmitt,$^{3}$ A.
14: Augustsson,$^{3}$ L. Gridneva,$^{3}$ D. Debowska-Nilsson,$^{2}$ R. Claessen,$%
15: ^{1}$A.Yu. Egorov,$^{4}$ V.M. Ustinov,$^{4}$ Zh.I. Alferov$^{4}$}
16: \address{$^{1}$Experimentalphysik II, Universit\"{a}t Augsburg,
17: D-86135Augsburg, Germany}
18: \address{$^{2}$Department of Physics, Chalmers University of Technology and
19: G\"{o}teborg University, SE-41296 G\"{o}teborg, Sweden}
20: \address{$^{3}$Department of Physics, Uppsala University, \AA ngstr\"{o}m
21: Laboratory, Box 530, S-75121 Uppsala, Sweden}
22: \address{$^{4}$A.F. Ioffe Physico-Technical Institute, 194021 St.Petersburg,
23: Russia }
24: \date{\today }
25: \maketitle
26:
27: \begin{abstract}
28: Soft-X-ray emission and absorption spectroscopies with their elemental
29: specificity are used to determine the local electronic structure of N atoms
30: in Ga(In)AsN diluted semiconductor alloys (N concentrations about 3\%) in
31: view of applications of such materials in optoelectronics. Deviations of the
32: N local electronic structure in Ga(In)AsN from the crystalline state in GaN
33: are dramatic in both valence and conduction bands. In particular, a
34: depletion of the valence band maximum in the N local charge, taking place at
35: the N impurities, appears as one of the fundamental origins of reduced
36: optical efficiency of Ga(In)AsN. Incorporation of In in large concentrations
37: forms In-rich N local environments such as In$_{4}$N whose the electronic
38: structure evolves towards improved efficiency. Furthermore, a {\bf k}%
39: -character of some valence and conduction states, despite the random alloy
40: nature of Ga(In)AsN, manifests itself in resonant inelastic X-ray scattering.
41: \end{abstract}
42:
43: \pacs{78.70.En, 78.70.Dm, 71.55.-i, 78.55.Cr}
44:
45: \section{Introduction}
46:
47: Ga(In)AsN semiconductor alloys are new promising optoelectronic materials,
48: whose potential applications range from efficient solar cells to laser
49: diodes operating in the long wavelength range ($\lambda \sim $1.3 $\mu $m)
50: which fits the transparency window ofoptofibers used in local networks. A
51: remarkable property of the Ga(In)AsN alloys is an extremely strong
52: dependence of the band gap width $E_{g}$ on the N content, characterized by
53: a giant bowing coefficient with $dE_{g}/dx$=15-20 eV (see, e.g., Ref.\cite%
54: {Kent01}). This figure is more than one order of magnitude larger compared
55: to the conventional III-V alloys, which suggests that physical mechanisms to
56: narrow the band gap are quite different. A disadvantage of the Ga(In)AsN
57: alloys is however their low optical efficiency compared to conventional
58: III-V alloys such as GaAs and AlAs.
59:
60: Physics of Ga(In)AsN and related alloys has been under intense study during
61: the last few years (see Refs.\cite{Kent01,Mattila99,Yacoub00,Yu00,Buyanova01}
62: and references therein). Due to a strong difference in the N and As
63: scattering potentials, insertion of N atoms into the host lattice results in
64: a giant perturbation of the electronic structure and formation of
65: fundamentally new electronic states such as resonant impurity states. Their
66: hybridization with the host states in the conduction band strongly perturbs
67: and shifts these states to lower energies, which narrows the band gap.
68: Different local environments of N atoms such as isolated impurities, N-N
69: pairs and various clusters form different states hybridizing with each
70: other. Because of the immense complexity of such a system no exhaustive
71: theoretical treatment exists up to now. Different approaches such as the
72: empirical pseudopotential supercell method,\cite{Kent01,Mattila99}
73: first-principles pseudopotential method\cite{Yacoub00} and band anticrossing
74: model\cite{Yu00} often give conflicting predictions. Moreover, their
75: experimental verification is complicated by significant scatter in the
76: experimental results depending on the sample preparation. Despite
77: significant advances in understanding of the band gap narrowing in Ga(In)AsN
78: alloys, mechanisms responsible for degradation of their optical efficiency
79: are still not completely clear.
80:
81: A vast amount of the experimental data on the Ga(In)AsN and similar alloys
82: has been obtained using optical spectroscopies such as photoluminescence
83: (PL) and electroreflectance (see, e.g., a compilation of references in Ref.%
84: \cite{Kent01}). However, they are largely restricted to the band gap region,
85: and give in general only bare positions of the energy levels without any
86: direct information about spatial localization or orbital character of
87: wavefunctions. Such an information can be achieved by soft-X-ray emission
88: (SXE) and absorption (SXA) spectroscopies with their specificity on the
89: chemical element and orbital character (see, e.g., a recent review in Ref.%
90: \cite{Kotani01}). Although their energy resolution, intrinsically limited by
91: the core hole lifetime, never matches that of the optical spectroscopies,
92: they give an overall picture of the electronic structure on the energy scale
93: of the whole valence band (VB) and condiction band (CB). Moreover, as the
94: orbital selection rules involve the core state and thus engage the VB and CB
95: states different from those engaged in optical transitions, the soft-X-ray
96: spectroscopies give a complementary view of the electronic structure.
97: Because of the small atomic concentrations in diluted alloys such as
98: Ga(In)AsN, and small crossection of the SXA and SXE processes, these
99: experiments require the use of 3-generation synchrotron radiation sources,
100: providing soft X-rays at high intensity and brilliance, and high-resolution
101: SXE spectrometers with multichannel detection.\cite{Nordgren00}
102:
103: Extending our pilot work,\cite{Strocov02} we here present experimental SXE
104: and SXA data on Ga(In)AsN diluted alloys, which unveil the local electronic
105: structure of N impurities through the whole VB and CB. This yields new
106: fundamental physics of Ga(In)AsN and related alloys, in particular,
107: electronic structure origins of their limited optical efficiency, effect of
108: In on the N\ local environments and electronic structure, and {\bf k}%
109: -character of some VB and CB states coupled by resonant inelastic X-ray
110: scattering (RIXS).
111:
112: \section{Experimental procedure and results}
113:
114: \subsection{Sample growth}
115:
116: The Ga(In)AsN samples were grown by molecular beam epitaxy (MBE) at an
117: EP-1203 machine (Russia) equipped by solid-phase Ga, In and As sources and a
118: radio-frequency plasma N source. Details of the growth procedure and sample
119: characterization are given elsewhere.\cite{Egorov01} Briefly, the growth was
120: performed on a GaAs(001) substrate at 430$^{\circ }$C in As-rich conditions.
121: The active layer in our GaAsN and GaInAsN samples was grown, respectively,
122: as GaAs$_{0.97}$N$_{0.03}$ with a thickness of 200 \AA , and In$_{0.07}$Ga$%
123: _{0.93}$As$_{0.97}$N$_{0.03}$ with a thickness of 240 \AA\ (growth of
124: thicker layers was hindered by phase segregation). The concentrations of In
125: and N were checked by high-resolution X-ray rocking curves. A buffer layer
126: between the substrate and the Ga(In)AsN active layer, and a cap layer on top
127: of it were grown each as a 50 \AA\ thick AlAs layer sandwiched between two
128: 50 \AA\ thick GaAs layers. Such an insertion of wide band gap AlAs is a
129: usual method to increase the PL intensity by confining the carriers in the
130: Ga(In)AsN layer. Moreover, a high-temperature annealing of the grown
131: structure can be performed after deposition of AlAs in the cap layer without
132: desorption of GaAs. Such an annealing lasts about 10 minutes at 700-750$%
133: ^{\circ }$C. The resulting improvement of the Ga(In)AsN layer crystal
134: quality typically increases the PL intensity by a factor of 10-20.
135:
136: The annealing effect on the local environments of N atoms is less clear. The
137: N impurities are known to interact with each other due to long-range lattice
138: relaxation and long tails of their wavefunctions down to N concentrations of
139: 0.1\%,\cite{Kent01} which translates to a characteristic interaction length
140: of 60 \AA . We expect that on this length scale the annealing can promote
141: energetically favourable N local environments. In GaAsN such envoronments
142: are, for example, (100)-oriented N pairs.\cite{Kent01} In the GaInAsN
143: quaternary alloy the situation is more complicated: whereas as-grown samples
144: have nearly random distributions of In and N atoms with a significant
145: fraction of InAs clusters having small chemical bond energy, the annealing
146: should promote formation of In-N bonds, providing better lattice match to
147: the GaAs substrate and thus mimimizing the strain energy.\cite{Kim01} In any
148: case, on a length scale larger than the N interaction length the annealing
149: should improve homogeneity of the N concentration. This is of paramount
150: importance, in particular, for our experiment because due to the giant
151: bowing coefficient of Ga(In)AsN any fluctuations of the N concentration
152: should result in significant fluctuations of the electronic structure\cite%
153: {Mintairov01,Matsuda01} and therefore in smearing of spectral structures.
154:
155: \subsection{SXE/SXA measurements}
156:
157: The SXE/SXA experiments were performed in MAX-lab, Sweden, at the undulator
158: beamline I511-3 equipped with a modified SX-700 plane grating monochromator
159: and a high-resolution Rowland-mount grazing incidence spectrometer.\cite%
160: {Nordgren89} SXE/SXA measurements employed the N 1$s$ core level at
161: approximately 400 eV.
162:
163: The SXA spectra were recorded in the fluorescence yield (FY), because due to
164: the thick cap layer the electron yield did not show any N 1$s$ absorption
165: structure. The measurements were performed in partial FY using the SXE
166: spectrometer operated slitless. It was adjusted at a photon energy window
167: centered at the N $K$-emission line and covering an interval, in the 1st
168: order of diffraction, from some 320 to 470 eV. The signal was detected with
169: the spectrometer position-sensitive detector as the integral fluorescence
170: within this energy window. Interestingly, usual measurements in the total FY
171: (detected with a microchannel plate detector in front of the sample)
172: returned considerably different spectra. This is possibly because the total
173: FY is more susceptible to irrelevant contributions due to higher-order
174: incident light and low-energy photoelectron bremsstrahlung fluorescence,
175: significant with our low N concentrations in the host material. As the
176: partial FY measurements are characterized by significant intensity loss due
177: to smaller acceptance angle of the spectrometer, we operated the
178: monochromator at an energy resolution of 0.45 eV FWHM (the N 1$s$ lifetime
179: broadening is about 0.1 eV\cite{Lawniczak00}).
180:
181: The synchrotron radiation excited SXE spectra were measured, in view of the
182: low crossection of the SXE process and small N concentration, with the
183: monochromator resolution lowered to $\sim $1.5 eV and to $\sim $0.5 eV for
184: the off-resonance and resonant spectra, respectively. The spectrometer was
185: operated with a spherical grating of 5 m radius and 400 lines/mm groove
186: density in the 1st order of diffraction, providing a resolution of $\sim $%
187: 1.2 eV. The signal from the position-sensitive detector was aberration
188: corrected using 3rd-order polynomial fitting and normalized to the total
189: illuminated area in each channel on the detector. Normal data acquisition
190: time was 2-5 hours per spectrum. Despite the cap layer we could also see a N
191: signal under 3.5 keV electron beam excitation, although on top of strong
192: bremsstrahlung background, but this was not suitable for resonant
193: measurements.
194:
195: Energy calibration of the spectrometer was performed in absolute photon
196: energies employing the Ni $L_{l}$, $L_{\alpha _{1,2}}$ and $L_{\beta _{1}}$
197: lines seen in the 2nd order of diffraction. Abberations in the dispersion
198: direction of the position-sensitive detector were corrected by setting an
199: energy scale as a function of the channel number using 2nd order polynomial
200: fitting. Based on the elastic peaks in SXE spectra, the monochromator was
201: then calibrated in the same absolute energy scale with an accuracy about $%
202: \pm $0.15 eV.
203:
204: \subsection{Experimental results}
205:
206: Our experimental N 1$s$ SXA spectra (measured in the partial FY)\cite%
207: {TotalFY} and off-resonant SXE spectra (excitation energy of 420 eV, well
208: above the absorption threshold) of the GaAs$_{0.97}$N$_{0.03}$ and Ga$_{0.93}
209: $In$_{0.07}$As$_{0.97}$N$_{0.03}$ samples are shown in Fig.1 ({\it upper
210: panel}). The binding energy scale is set relative to the VB maximum (VBM)
211: determined, roughly, by linear extrapolation of the SXE spectral leading
212: edge. We intentionally give the spectra without denoising to facilitate
213: judgement the significance of the spectral structures compared to the noise
214: level. Recent supercell calculations by Persson and Zunger\cite{Persson03}
215: are in good agreement with our experimental results.
216:
217: Recent SXA data on GaAs$_{0.97}$N$_{0.03}$ by Lordi {\it et al},\cite%
218: {Lordi03} which appeared after initial submission of this paper, are
219: consistent with our results (apart from some energy shift which is
220: presumably because their energy scale was affected by the monochromator
221: calibration). Previous SXA data by Soo {\it et al}\cite{Soo99} suffer from
222: worse experimental resolution and sample quality.
223:
224: Local environments of the N atoms in Ga(In)AsN are polymorphic,
225: corresponding to isolated impurities and various clusters.\cite{Kent01}
226: Applying random statistics, the concentration ratio of the pair and
227: higher-order N clusters to the total number of N atoms is given by 1-(1-$x$)$%
228: ^{m}$, where $x$ is the N concentration and $m$=4 the number of the nearest
229: anions in the zinc-blende lattice. With our N concentrations of 3\% this
230: ratio is only 11.5\%. Therefore, our SXE/SXA spectra characterize mainly the
231: isolated N impurities.
232:
233: \section{Discussion}
234:
235: \subsection{Overall picture of the electronic structure}
236:
237: The experimental SXA and off-resonance SXE spectra in Fig.1 reflect, by the
238: dipole selection rules requiring that the orbital quantum number $l$ is
239: changed by $\pm 1$, the $p$-component of the DOS locally in the N core
240: region. The $p$-component, by analogy with crystalline GaN,\cite{Lawniczak00}
241: should in fact dominate the total DOS through the whole VB and CB region.
242: Core excitonic effects are presumably less significant because the direct
243: recombination peak\cite{Agui99} does not show up in our SXE spectra.
244: Splitting of the VBM into the light and heavy hole subbands due to a strain
245: imposed by the GaAs substrate,\cite{Zhang00} being about a few tenths of eV,
246: is below our experimental resolution.
247:
248: It is instructive to compare our Ga(In)AsN spectra to the corresponding
249: spectra of crystalline GaN. They are reproduced in Fig.1 ({\it lower panel})
250: in the binding energy scale determined in the same way as for Ga(In)AsN. The
251: spectra of GaN in the metastable zinc-blende structure, which has the same N
252: coordination as Ga(In)AsN, were measured by Lawniczak {\it et al},\cite%
253: {Lawniczak00} and those of wurtzite GaN by Stagarescu {\it et al}.\cite%
254: {Stagarescu96} Apart from the CB shift, the spectra of the two crystalline
255: forms are similar in overall shape. They are well understood in terms of the
256: local orbital-projected DOS and band structure.\cite{Lawniczak00,Lambrecht97}
257:
258: Comparison of the SXE/SXA data on Ga(In)AsN to those on the two GaN
259: crystalline structures shows:
260:
261: (1) In the VB, the overall shape of the SXE signal for Ga(In)AsN is similar
262: to crystalline GaN. However, the spectral maximum is strongly shifted
263: towards the VB interior, with the leading edge at the VBM being much less
264: steep (which has important implications for optical efficiency, see below).
265: This is not a resolution effect, because the reference spectra of
266: crystalline GaN were taken at close resolution figures (around 0.8 eV for
267: zinc-blende GaN and 1.1 eV for wurtzite GaN). Our experimental data
268: demonstrate thus that the VB electronic structure undergoes, contrary to the
269: common point of view, significant changes upon incorporation of N atoms into
270: GaAs. Interestingly, our SXE spectrum did not show any structure due to
271: hybridization with the Ga 3{\it d} states at $\sim $19 eV below the VBM,
272: found in wurtzite GaN;\cite{Stagarescu96,Duda98}
273:
274: (2) In the CB, the differences are radical. The leading peak of the SXA
275: spectrum for Ga(In)AsN rises immediately at the CB minimum (CBM) and has
276: much larger amplitude compared to the leading shoulder-like structure in the
277: spectra of crystalline GaN. The energy separation between the VB and CB
278: states for Ga(In)AsN is much smaller, which correlates with smaller
279: fundamental band gap.
280:
281: On the whole, the observed differences of the Ga(In)AsN spectra to
282: crystalline GaN manifest that the local electronic structure of the N atoms
283: in the Ga(In)AsN random alloy is radically different from that in the
284: regular GaN lattice.
285:
286: Although further theoretical analysis is required to interpret our
287: experimental data in detail, we can tentatively assign the leading SXA peak
288: to the $t_{2}(L_{1c})$ derived perturbed host state which, according to the
289: calculations by Kent and Zunger on GaAsN,\cite{Kent01} has the strongest N
290: localization in the CBM\ region. This assignment is corroborated by the
291: resonant SXE data (see below) which reveals the $L$-character of the leading
292: SXA peak.
293:
294: It should be noted that the dipole selection rules in SXE/SXA, inherently
295: involving transitions from and to the core level, project out the states
296: from the VB and CB, which can differ from those projected out by the optical
297: transitions between the VB and CB states themselves. For example,
298: delocalized states can give only a small contribution to the SXE/SXA signal
299: due to relatively small overlap with the core state, but they can strongly
300: overlap with each other and give a strong PL signal. Our SXA data give
301: explicit examples of this: The $a_{1}(\Gamma _{1c})$ derived states near the
302: CBM (see Ref.\cite{Kent01}) are not seen in the SXA spectrum due to the
303: weaker N localization compared to the $t_{2}(L_{1c})$ states, but in optical
304: spectroscopies they manifest themselves as the intense $E^{-}$ transitions.
305: On the other hand, the $t_{2}(L_{1c})$ states are not seen in the optical
306: spectra due to unfavorable matrix elements, but show up as a prominent SXA
307: peak. Moreover, the energy separation between the VB and CB states in the
308: SXE/SXA spectra gives only an upper estimate for the fundamental band gap,
309: because weakly localized N states as well as Ga and As derived states are
310: not seen. Therefore, the SXE/SXA spectroscopies give a view of the VB and CB
311: complementary to that by optical spectroscopies.
312:
313: \subsection{Charge depletion in the VBM: Origin of reduced optical efficiency%
314: }
315:
316: The vast body of optical spectroscopy data on Ga(In)AsN evidences that the
317: optical efficiency sharply drops upon incorporation of the smallest N
318: concentrations into GaAs, and then decreases further with increase of the N
319: molar fraction (see, e.g., a compilation in Ref. \cite{Buyanova01}). This is
320: most pronounced for GaAsN, where the PL intensity loss towards N
321: concentrations of 5\% is at least 50 times as compared to GaAs. Exact
322: origins of such a dramatic efficiency degradation are not completely clear.
323: Supercell calculations in Ref.\cite{Bellaiche97} suggest that about 30\% of
324: the GaAs efficiency is lost due to gradual smearing of VBM and CBM in their $%
325: \Gamma $-character, which results in reduction of the optical transition
326: matrix element. However, this effect is by far weaker compared to the
327: experimental degradation. Another known origin is relatively poor structural
328: quality of Ga(In)AsN layers epitaxially grown on GaAs. This is due to,
329: firstly, low growth temperatures which are used with large N concentrations
330: to promote high N uptake and, secondly, some lattice mismatch between
331: Ga(In)AsN and GaAs. However, the first problem can be alleviated by
332: post-growth high-temperature annealing, and the second by tuning the In
333: concentration in GaInAsN which allows matching the GaAs lattice constant.
334: Althought the PL intensity from lattice-matched GaInAsN layers does increase
335: by a factor about 5 compared to GaAsN, this still remains by far low
336: compared to GaAs. Moreover, the structural quality does not explain the
337: efficiency drop at the smallest N concentrations.
338:
339: Our SXE/SXA results unveil another origin of the optical efficiency
340: degradation in the very electronic structure. By virtue of the N
341: localization of the CBM wavefunction\cite{Kent01} the N\ impurities act as
342: the main recombination centers in Ga(In)AsN. At the same time, the local
343: valence charge at the N impurities is shifted off the VBM. This appears
344: immediately from comparison of our experimental SXE spectra of Ga(In)AsN
345: with those of crystalline GaN, which are in fact representative of GaAs by
346: virtue of qualitatively similar valence DOS of these materials\cite%
347: {Chelikowski89} (direct measurements on GaAs are hindered by very low
348: fluorescence yield of As in the soft-X-ray region). Such a charge depletion
349: in the VBM, equivalent to reduction of the VBM wavefunction amplitude,
350: results in a weak overlap of the CBM and VBM wavefunctions at the N\
351: impurities, which immediately reduces efficiency of the N impurities as
352: radiative recombination centers. This VBM depletion effect, characteristic
353: of isolated N impurities, is one of fundamental origins of the reduced
354: optical efficiency of Ga(In)AsN. Being in play already at the smallest N
355: concentrations, it immediately explains the initial efficiency drop, whereas
356: further efficiency degradation with increase of N concentration is
357: presumably through the structural quality effects.
358:
359: To explain the observed VBM charge depletion, in Ref.\cite{Strocov02} we
360: suggested a VBM charge transfer off the N atoms in Ga(In)AsN compared to GaN
361: (in Ref.\cite{Persson03} this our statement was misinterpreted as a charge
362: transfer to As, but N has larger electronegativity). In fact, the charge
363: transfer is more likely to take place not in space but in energy towards
364: deeper valence states, which is supported by recent computational analysis
365: of Persson and Zunger.\cite{Persson03} Physically, the local electronic
366: structure of the N impurities in the GaAs lattice appears somewhere in
367: between that of the crystalline state,\ and that of isolated atoms. The
368: observed DOS peaked near the VB center can therefore be viewed as a
369: transitional case between the DOS of extended band states piling up near the
370: band edges, and the singularity-like DOS of isolated atoms at the VB center.
371:
372: Formation of N local environments different from the isolated impurities can
373: be suggested as a way to increase the optical efficiency of Ga(In)AsN. For
374: example, in multiatomic N local environments such as clusters of
375: Ga-separated N atoms the wavefunctions may become closer to crystalline GaN
376: with its DOS piling up at the VBM. Based on the random statistics, the
377: cluster concentration should increase with the total N concentration.
378: Alternatively, the N local environments can be changed by replacing some
379: neighbour Ga atoms by different cations.
380:
381: \subsection{Effect of In}
382:
383: Quaternary GaInAsN alloys, where some Ga atoms are replaced by In, allow
384: improvement of the optical efficiency by a factor about 5. This is
385: predominantly due to two factors: a better lattice match of GaInAsN layers
386: to the GaAs substrate, which improves their structural quality, and electron
387: confinement effects connected with concentration fluctuations.\cite%
388: {Mintairov01} We here endevoured investigation whether the incorporation of
389: In also causes any favourable changes in the electronic structure.
390:
391: At relatively low In concentrations, evolution of the N local electronic
392: structure is evidenced by comparison of the Ga$_{0.93}$In$_{0.07}$As$_{0.97}$%
393: N$_{0.03}$ and GaAs$_{0.97}$N$_{0.03}$ experimental spectra in Fig.1. The
394: SXE spectra zoomed in the VB\ region are also shown in Fig.2. Surprisingly,
395: the comparison shows no notable changes within the experimental statistics,
396: nor in the spectral shapes, neither in energies of the spectral structures.
397: This evidences that despite the high-temperature annealing the N atoms
398: reside mostly in In-depleted local environments such as Ga$_{4}$N and
399: possibly\cite{Kurtz01} In$_{1}$Ga$_{3}$N where the presence of only one In
400: atom in 4 nearest neighbours should not change the N local electronic
401: structure dramatically. This experimental finding seriously questions
402: results of recent Monte Carlo simulations\cite{Kim01} which predict
403: predominance of In-rich N local environments such as Ga$_{1}$In$_{3}$N and In%
404: $_{4}$N, at least with low In concentrations. Any effects connected with
405: insufficient annealing of our samples can be ruled out, as evidenced by
406: stabilization of PL spectra already after 5 min of annealing. The absence of
407: any significant electronic structure changes in Ga$_{0.93}$In$_{0.07}$As$%
408: _{0.97}$N$_{0.03}$ compared to GaAs$_{0.97}$N$_{0.03}$ suggests that at low
409: In concentrations the optical efficiency improvement is exclusively due to
410: the structural and electron confinement effects.
411:
412: To force formation of In-rich N local environments, we have grown a sample
413: of Ga$_{0.69}$In$_{0.31}$As$_{0.98}$N$_{0.02}$ (170 \AA\ thick active layer)
414: where the In/N concentration ratio is much increased (the decrease in bare N
415: concentration is presumably less important because interaction of the
416: isolated N impurities in such diluted alloys should be weak). The
417: experimental SXE spectrum of Ga$_{0.69}$In$_{0.31}$As$_{0.98}$N$_{0.02}$,
418: measured under the same off-resonance conditions as in Fig.1, is also shown
419: in Fig.2. Now the spectral maximum is shifted by some 0.25 eV to higher
420: energies compared to GaAs$_{0.97}$N$_{0.03}$, indicating changes in the N
421: local electronic structure caused by In-rich N environments. Interestingly,
422: the XAS data for Ga$_{0.7}$In$_{0.3}$As$_{0.97}$N$_{0.03}$ from Ref.\cite%
423: {Lordi03} demonstrates the CBM simultaneously shifts to lower energies.
424:
425: The observed VB and CB shifts towards each other suggest that the In-rich N
426: environments become the main recombination centers in GaInAsN. Moreover,
427: both holes and electrons become confined in In-reach regions formed by
428: statistical fluctuations of In concentration on $\mu m$-scale. This effect
429: increases the optical efficiency of GaInAsN.\cite{Mintairov01}
430:
431: To see whether the observed changes in the VB affect the optical efficiency
432: within the above VBM\ depletion mechanism, we examined closely the VBM
433: region (insert in Fig.2). The shift of the spectral maximum is definitely
434: larger than that of the VBM (although its exact location requires better
435: statistics). This indicates certain charge accumulation at the VBM compared
436: to GaAs$_{0.97}$N$_{0.03}$, and thus increase of the optical efficiency of
437: In-rich N local environments compared to Ga$_{4}$N. The observed
438: accumulation seems though rather subtle to explain the increase in GaInAsN
439: wholly, and the most of it still resides with the structural and electron
440: confinement effects.
441:
442: \subsection{{\bf k}-conservation in the RIXS process}
443:
444: Resonant phenomena were investigated on the GaAsN prototype alloy. Fig.3
445: shows resonant SXE spectra measured with excitation energies near the two
446: dominant SXA structures in Fig.1 compared to that measured well above the
447: absorption threshold. The spectra are normalized to the integral excitation
448: flux, which was registered from the photocurrent at a gold mesh inserted
449: after the refocussing mirror.
450:
451: Intriguingly, not only does the intensity of the resonant spectra increase
452: in this RIXS process, but also the shoulder at the VB bottom scales up and
453: becomes a distinct narrow peak at a binding energy of $\sim $7.4 eV. Such a
454: behavior reveals states near the VB bottom which effectively overlap with
455: states near the CB\ bottom into which the core electron is excited.
456:
457: Despite the random alloy nature of Ga(In)AsN, the observed effect can be
458: interpreted in terms of momentum conservation which appears in the RIXS
459: process due to coupling of absorption and emission in one single event (see,
460: e.g., Refs.\cite{Kotani01,Eisebitt98,Carlisle99} and references therein). At
461: first glance, this should not occur in a random alloy, because the very
462: concept of momentum, strictly speaking, collapses due to the lack of
463: translational invariance. However, the description in terms of wavevectors
464: {\bf k} can be revived using a spectral decomposition%
465: \[
466: \psi ^{N}({\bf r})=\sum\limits_{{\bf k}}C_{{\bf k}}\phi _{{\bf k}}^{GaAs}(%
467: {\bf r})
468: \]%
469: of the N-localized wavefunction $\psi ^{N}({\bf r})$ over the Bloch waves $%
470: \phi _{{\bf k}}^{GaAs}({\bf r})$ of the unperturbed GaAs lattice, each
471: having a well-defined {\bf k}.\cite{Kent01,Wang98} Then the first SXA peak
472: is due to the $t_{2}(L_{1c})$ state, whose decomposition is dominated by
473: {\bf k} from the $L$-point in the Brillouin zone of GaAs.\cite{Kent01} The
474: VB bottom, by analogy with the zinc-blende GaN band structure,\cite{Lewis01}
475: should be dominated by the same $L$-point. The RIXS process will then couple
476: these points in the CB and VB, blowing up the SXE signal in the VB bottom as
477: observed in the experiment. Our resonant data demonstrate thus, to our
478: knowledge for the first time, a possibility for the {\bf k}-conserving RIXS
479: phenomenon in random alloys.
480:
481: \section{Conclusion}
482:
483: Local electronic structure of N atoms in Ga(In)AsN diluted semiconductor
484: alloys (N concentrations about 3\%) has been determined using SXE/SXA
485: spectroscopies with their elemental specificity. The experimental N 1$s$
486: off-resonance SXE spectra and SXA spectra yield the local $p$-DOS of N
487: impurities through the whole VB\ and CB, complementing information about the
488: band gap region achieved by optical spectroscopies. The experimental results
489: demonstrate dramatic differences of the N\ local electronic structure in
490: Ga(In)AsN from that in the crystalline GaN state. A few peculiarities have
491: immediate implications for optical properties: (1) The N impurities are
492: characterized by depletion of the the local charge in the VBM due to charge
493: transfer towards deeper valence states, which reduces overlap with the CBM
494: states. This is one of the fundamental origins of the reduced optical
495: efficiency of Ga(In)AsN. Formation of different N\ local environments can
496: improve the efficiency; (2) Whereas incorporation of In in small
497: concentrations has an insignificant effect on the N local electronic
498: structure, large In concentrations result in formation of In-rich N local
499: environments whose electronic structure evolves towards improved optical
500: efficiency. Furthermore, the experimental resonant SXE spectra reveal,
501: despite the random alloy nature of Ga(In)AsN, a {\bf k}-conserving RIXS
502: process which couples valence and conduction states having the same $L$%
503: -character.
504:
505: \section{Acknowledgements}
506:
507: We are grateful to A. Zunger and C. Persson for valuable comments and
508: communicating their computational results before publication. We thank S.
509: Butorin for his advise on SXE data processing, J. Guo for help with
510: preliminary experiments, and G. Cirlin for valuable discussions. The work in
511: the Ioffe institute is supported by the NATO Science for Peace Program
512: (SfP-972484) and Russian Foundation for Basic Research (project 02-02-17677).
513:
514: \begin{references}
515: \bibitem[*]{Contact} Contact author, email strocov@physik.uni-augsburg.de
516:
517: \bibitem{Kent01} P.R.C. Kent and A. Zunger, Phys. Rev. Lett. {\bf 86} (2001)
518: 2613; Phys. Rev. B {\bf 64}, 115208 (2001)
519:
520: \bibitem{Mattila99} T. Mattila, S.-H. Wei and A. Zunger, Phys. Rev. B {\bf 60%
521: }, R11245 (1999)
522:
523: \bibitem{Yacoub00} A. Al-Yacoub and L. Bellaiche, Phys. Rev. B {\bf 62},
524: 10847 (2000)
525:
526: \bibitem{Yu00} K.M. Yu, W. Walukiewicz, W. Shan, J.W. Ager III, J. Wu, E.E.
527: Haller, J.F. Giesz, D.J. Friedman and J.M. Olson, Phys. Rev. B {\bf 61},
528: R13337 (2000)
529:
530: \bibitem{Buyanova01} I.A. Buyanova, W.M. Chen and B. Monemar, MRS Internet
531: J. Nitride Semicond. Res. {\bf 6}, 2 (2001)
532:
533: \bibitem{Kotani01} A. Kotani and S. Shin, Rev. Mod. Phys. {\bf 73}, 203
534: (2001)
535:
536: \bibitem{Nordgren00} J. Nordgren and J. Guo, J. Electron Spectrosc. and
537: Relat. Phen. {\bf 110--111}, 1 (2000)
538:
539: \bibitem{Strocov02} V.N. Strocov, P.O. Nilsson, A. Augustsson, T. Schmitt,
540: D. Debowska-Nilsson, R. Claessen, A.Yu. Egorov, V.M. Ustinov and Zh.I.
541: Alferov, phys. stat. sol. (b) {\bf 233}, R1--R3 (2002)
542:
543: \bibitem{Egorov01} A.Yu. Egorov, D. Bernklau, B. Borchert, S. Illek, D.
544: Livshits, A. Rucki, M. Schuster, A. Kaschner, A. Hoffmann, Gh. Dumitras,
545: M.C. Amann and H. Riechert, J. of Cryst. Growth {\bf 227-228}, 545 (2001)
546:
547: \bibitem{Kim01} K. Kim and A. Zunger, Phys. Rev. Lett. {\bf 86}, 2609 (2001)
548:
549: \bibitem{Mintairov01} A.M. Mintairov, T.H. Kosel, J.L. Merz, P.A. Blagnov,
550: A.S. Vlasov, V.M. Ustinov and R.E. Cook, Phys. Rev. Lett. {\bf 87}, 277401
551: (2001)
552:
553: \bibitem{Matsuda01} K. Matsuda, T. Saiki, M. Takahashi, A. Moto and S.
554: Takagishi, Appl. Phys. Lett. {\bf 78}, 1508 (2001)
555:
556: \bibitem{Nordgren89} J. Nordgren, G. Bray, S. Cramm, R. Nyholm, J.-E.
557: Rubensson and N. Wassdahl. Rev. Sci. Instrum. {\bf 60}, 1690 (1989)
558:
559: \bibitem{TotalFY} SXA spectra measured in the total FY (see our Ref.\cite%
560: {Strocov02}) reproducibly show up a pronounced peak in some 2 eV above the
561: leading absoption peak, which may be due to structure in the low-energy
562: bremsstrahlung.
563:
564: \bibitem{Persson03} C. Persson and A. Zunger, Phys. Rev. B, in press (2003)
565:
566: \bibitem{Lordi03} V. Lordi, V. Gambin, S. Friedrich, T. Funk, T. Takizawa,
567: K. Uno and J.S. Harris, Phys. Rev. Lett. {\bf 90}, 145505 (2003)
568:
569: \bibitem{Soo99} Y.L. Soo, S. Huang, Y.H. Kao, J.G. Chen, S.L. Hulbert, J.F.
570: Geisz, S. Kurtz, J.M. Olson, S.R. Kurtz, E.D. Jones and A.A. Allerman, Phys.
571: Rev. B {\bf 60}, 13605 (1999)
572:
573: \bibitem{Lawniczak00} K. Lawniczak-Jablonska, T. Suski, I. Gorczyca, N.E.
574: Christensen, K.E. Attenkofer, R.C.C. Perera, E.M. Gullikson, J.H. Underwood,
575: D.L. Ederer and Z. Liliental Weber, Phys. Rev. B {\bf 61}, 16623 (2000)
576:
577: \bibitem{Stagarescu96} C.B. Stagarescu, L.-C. Duda, K.E. Smith, J.H. Guo, J.
578: Nordgren, R. Singh and T.D. Moustakas, Phys. Rev. B {\bf 54}, R17335 (1996)
579:
580: \bibitem{Duda98} L.-C. Duda, C.B. Stagarescu, J. Downes, K.E. Smith, D.
581: Korakakis, T.D. Moustakas, J.H. Guo and J. Nordgren, Phys. Rev. B {\bf 58},
582: 1928 (1998)
583:
584: \bibitem{Agui99} A. Agui, S. Shin, C. Wu, K. Shiba and K. Inoue, Phys. Rev.
585: B {\bf 59}, 10792 (1999)
586:
587: \bibitem{Zhang00} Y. Zhang, A. Mascarenhas, H.P. Xin and C.W. Tu, Phys. Rev.
588: B {\bf 61}, 4433 (2000)
589:
590: \bibitem{Lambrecht97} W.R.L. Lambrecht, S.N. Rashkeev, B. Segall, K.
591: Lawniczak-Jablonska, T. Suski, E.M. Gullikson, J.H. Underwood, R.C.C.
592: Perera, J.C. Rife, I. Grzegory, S. Porowski and D. K. Wickenden, Phys. Rev.
593: B {\bf 55}, 2612 (1997)
594:
595: \bibitem{Bellaiche97} L. Bellaiche, S.-H. Wei and A. Zunger, Phys. Rev. B
596: {\bf 56}, 10233 (1997)
597:
598: \bibitem{Chelikowski89} J. R. Chelikowski, T. J. Wagener, J. H. Weaver, and
599: A. Jin, Phys. Rev. B {\bf 40}, 9644 (1989)
600:
601: \bibitem{Kurtz01} S. Kurtz, J. Webb, L. Gedvilas, D. Friedman, J. Geisz, J.
602: Olson, R. King, D. Joslin and N. Karam, Appl. Phys. Lett. {\bf 78}, 748
603: (2001)
604:
605: \bibitem{Eisebitt98} S. Eisebitt, J. L\"{u}ning, J.-E. Rubensson, A.
606: Settels, P.H. Dederichs, W. Eberhardt, S.N. Patitsas and T. Tiedje, J.
607: Electron Spectr. and Rel. Phen. {\bf 93}, 245 (1998)
608:
609: \bibitem{Carlisle99} J.A. Carlisle, E.L. Shirley, L.J. Terminello, J.J. Jia,
610: T.A. Callcott, D.L. Ederer, R.C.C. Perera and F.J. Himpsel, Phys. Rev. B
611: {\bf 59}, 7433 (1999)
612:
613: \bibitem{Wang98} L.-W. Wang, L. Bellaiche, S.-H. Wei and A. Zunger, Phys.
614: Rev. Lett. {\bf 80}, 4725 (1998)
615:
616: \bibitem{Lewis01} J.P. Lewis, K.R. Glaesemann, G.A. Voth, J. Fritsch, A.A.
617: Demkov, J. Ortega and O.F. Sankey, Phys. Rev. B {\bf 64}, 19510 (2001)
618: \end{references}
619:
620: \begin{figure}[tbp]
621: \caption{({\it upper panel}) Experimental N 1$s$ off-resonant SXE
622: (excitation energy 420 eV) and SXA spectra of GaAs$_{0.97}$N$_{0.03}$ and Ga$%
623: _{0.93}$In$_{0.07}$As$_{0.93}$N$_{0.03}$, reflecting the N local $p$-DOS
624: through the VB and CB; ({\it lower panel}) The corresponding spectra of
625: crystalline GaN in the zinc-blende \protect\cite{Lawniczak00} and wurtzite
626: structures \protect\cite{Stagarescu96} shown as a reference. The SXE
627: spectral maximum for Ga(In)AsN is strongly shifted to lower energies
628: compared to the crystalline state, resulting in depletion of the N local
629: charge in the VBM which reduces optical efficiency.}
630: \end{figure}
631: \begin{figure}[tbp]
632: \caption{Experimental off-resonant N1s SXE spectrum of In-rich Ga$_{0.69}$In$%
633: _{0.31}$As$_{0.98}$N$_{0.02}$ compared to GaAs$_{0.93}$N$_{0.03}$. The
634: insert details the spectral leading edges (Gaussian smoothed with FWHM of
635: 0.6 eV) which suggest improved optical efficiency of In-rich N local
636: environments. }
637: \end{figure}
638:
639: \begin{figure}[tbp]
640: \caption{Resonant SXE spectra with the indicated excitation energies
641: compared to an off-resonant spectrum. The elastic peaks are marked by
642: vertical ticks. The resonant intensity enhancement in the VB bottom
643: manifests a {\bf k}-conserving RIXS process. }
644: \end{figure}
645:
646: \end{document}
647: