The use of synchrotron radiation techniques in the characterization of strained semiconductor heterostructures and thin films

Abstract

In the last couple of decades, high-performance electronic and optoelectronic devices based on semiconductor heterostructures have been required to obtain increasingly strict and well-defined performances, needing a detailed control, at the atomic level, of the structural composition of the buried interfaces. This goal has been achieved by an improvement of the epitaxial growth techniques and by the parallel use of increasingly sophisticated characterization techniques. Among them, a leading role has been certainly played by those exploiting synchrotron radiation (SR) sources. In fact synchrotron radiation has distinct advantages as a photon source, notably high brilliance and continuous energy spectrum; by using the latter characteristic atomic selectivity can be obtained and this is of fundamental help to investigate the structural environment of atoms present only in a few angstrom (Å) thick interface layers of heterostructures. The third generation synchrotron radiation sources have allowed to reach the limit of measuring a monolayer of material, corresponding to about 10¹⁴ atoms/cm². Since, in the last decade, the use of intentionally strained heterostructures has greatly enhanced the performance of electrical and electro-optical semiconductor, a particular attention will be devoted to intentionally strained superlattices.

First the effect of strain on the band lineups alignments in strained heterostructures will be discussed deeply. Then the attention will be focused on to review the most important results obtained by several groups in the characterization of semiconductor heterostructures using the following structural SR techniques: (i) X-ray absorption-based techniques such as EXAFS, polarization-dependent EXAFS, surface EXAFS and NEXAFS (or XANES); (ii) X-ray diffraction-based techniques such as high-resolution XRD, grazing incidence XRD, XRD reciprocal space maps, X-ray standing waves and diffraction anomalous fine structure (DAFS); (iii) photoelectron-based techniques.

Introduction

III–V, II–VI and IV–IV heterostructures, single-, multiple-quantum wells (SQW, MQW) and superlattices (SL) are widely used in an always growing number of high technology devices operating from the mid-IR to the UV regions of the electromagnetic spectrum.

III–V heterostructures have played, and still play, a dominant role in the realization of electronic and optoelectronic devices employed in modern optical fiber communication systems at 0.9, 1.3 or 1.55 μm (corresponding to 1.38, 0.95 and 0.80 eV, respectively) [1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17] high electron mobility transistors [18], [19], [20], single-mode tunable lasers [21], [22], [23] and in general for all technological domains requiring remarkable electro-optical characteristics [24], [25], [26], [27], mainly due to the optical non-linearities of quantum structures [11], [28], [29], [30], [31], [32], [33], [34]. This is mainly the domain of ternary and quaternary alloys of the type InGaAlAsP, because the energy gaps of the corresponding binary system, E_g(InAs)=0.41 eV; E_g(InP)=1.42 eV; E_g(GaAs)=1.52 eV and E_g(GaP)=2.87 eV, E_g(AlAs)=2.15 eV and E_g(AlP)=2.45 eV, allow to cover the whole range of interest for optical fiber communication. In this field, the most common systems are GaAlAs [35], [36], [37], [38], [39], [40], [41], InGaAs [35], [36], [38], [41], [42], [43], [44], [45], [46], InGaAlAs [35], [36], [44] and InGaAsP [21], [23], [47], [48], [49], [50], [51].

More recently, also the GaInNAs system has become of interest because the incorporation of very small amounts of nitrogen (a few percent) into III-As materials leads to a dramatic decrease in the band gap energy¹ [52], [53], [54], [55] and has enabled the growth of GaAs-based laser diodes functioning in the 1.3–1.55 μm range, suitable for optical fiber transmission. Meanwhile, the smaller nitrogen atom reduces the lattice parameter of these films, and reduces the strain in epilayers with respect to that observed for systems containing the larger In or Sb atoms. Moreover, the presence of nitrogen strongly improve the electron confinement as a consequence of the increased conduction band discontinuity (ΔE_c)² [54], [56], [57], allowing the devices to work at much higher temperatures with respect to that of conventional systems. As an example, GaInNAs/GaAs vertical-cavity surface-emitting laser (VCSEL) operating at 1.3 μm have been realized [58], [59], [60] with a characteristic temperature (T₀) as high as 215 K [61], to be compared with T₀ values in the 50–100 K range for the more conventional InGaAsP/InP lasers [54].

The realization of devices operating in the mid-IR region is important, not only for telecommunication, but also for gas-sensing (environmental physical chemistry) and heat-sensing (military application). It can be achieved either by introducing the large antimony anion into the classical InGaAsP semiconductor alloys or by using the II–VI HgCdTe ternary system [62], [63]. As the energy gaps of antimony-based binary systems given by E_g(InSb)=0.17 eV and E_g(GaSb)=0.73 eV, very low energy gaps can be obtained for the InGaSb ternary and InGaAsSb quaternary systems [64], [65]. InAs/InAsSb heterostructures are the most powerful LED emitters beyond 5 μm (below 0.25 eV) [66] and incoherent LED emission has been observed up to ≈11 μm (down to 0.11 eV) [67], but coherent laser emission has been achieved only in the narrow 3.3–3.8 μm (0.38–0.33 eV) region [68], [69]. InAsSb/InAsSbP double heterostructure resulted in high power LED devices [70], [71].

Coming to the II–VI HgCdTe ternary system [72], HgTe is a semi-metal with a negative energy gap E_g(CdTe)=−0.26 eV, while E_g(CdTe)=1.57 eV, therefore very long wavelength operating devices can be realized [73], [74], [75]. Potentially, the band gap of CdTe/HgTe SL is adjustable from 0 to 1.6 eV depending on the thickness of the CdTe and HgTe layers [73]. Double-heterostructure HgCdTe injection lasers emitting at 2.86 μm (0.43 eV) have been realized [76]. InGaAsSb quantum wells and InAs/InGaSb superlattices are found to be more promising laser candidates than HgCdTe superlattices and InAsSb bulk ternaries. The calculated threshold current densities of InAs/InGaSb superlattices are similar to those of InGaAsSb active layers operating at 2.1 μm (0.59 eV), but are significantly lower at longer wavelengths [63]. As far as low energy lasers are concerned, more recently, high-power quantum cascade lasers based on intersubband transitions in a MQW heterostructure have been realized in the mid-infrared (IR) (λ>4 μm, hν<0.31 eV) region mainly by the Capasso’s group [77], [78], [79], [80], [81], [82], [83].

On the opposite side of the electromagnetic spectrum, wide-gap II–VI compounds such as ZnCdSSe quaternary and ZnSSe ternary compounds [84], [85] and group III nitride semiconductors [86] shear the field of visible and UV light emitters. As for the II–VI ZnCdSSe system CdSe/ZnSSe quantum island laser [87] and ZnSe/ZnCdSe [88] QW lasers [88], [89], have been successfully realized. Yellow-green ZnCdSe/BeZnTe II–VI laser diodes emitting at 0.56 μm II–VI laser diodes were successfully grown on InP substrates [90]. Great progress has recently been made in research and fabrication of optoelectronic devices based on the group III nitride semiconductors [86]. One of the attractive features of the nitride semiconductors is that their direct band gaps span most of the visible spectrum, into the UV, as the energy gaps of the corresponding binary systems is: E_g(InN)=1.89 eV; E_g(GaN)=3.44 eV; E_g(AlN)=6.28 eV. Appropriate alloying allows formation of ternary or quaternary alloys with band gaps intermediate to those of the binary compounds. This feature is widely used in band-structure engineering of nitride-based devices; for instance, most light emitters contain an active region consisting of an InGaN layer with band gap lower than that of GaN [91], [92]. The Al_xGa_1−xN system has been used as cladding layer for LEDs in the UV region of the spectrum [93], [94], [95]. The Ga_xIn_1−xN-based quantum optoelectronic devices play an important role as light emitter and detectors from the UV to the blue/green region of the electromagnetic spectrum [92], [96]. Finally, the Al_xIn_1−xN system is of particular interest, as it is lattice matched to GaN for x=0.83. High-performance AlGaN/GaInN QW lasers have been realized [97], [98]. These are the reasons why, in a remarkably short space of time, the nitrides have caught up with and, in some ways, surpassed the wide band gap II–VI compounds (ZnCdSSe) as materials for short wavelength optoelectronic devices [86].

Finally, group IV materials, in sp³ hybridization, exhibit a band structure that moves from the insulating diamond to the extremely narrow energy gap of tin: E_g(C)=5.33 eV; E_g(Si)=1.14 eV; E_g(Ge)=0.67 eV; E_g(Sn)=0.08 eV. The indirect nature of the band gap in group IV materials makes such materials unsuitable for large-scale use in optoelectronic technologies up to now. However, several attempts have been made to modify the band gap structure of group IV semiconductors in order to obtain direct gap materials [99], [100], [101], [102], [103], [104], [105], [106]. It has been shown that alloying silicon with a few percent of carbon can render the band gap direct with strong optical absorption, provided the carbon atoms are ordered [104]. The addition of carbon introduces a significant s character into the conduction band minimum, resulting in a large dipole matrix element. The direct energy gap has been measured for coherently strained Sn_xGe_1−x alloys on Ge(0 0 1) substrates with 0.035<x<0.115 and film thickness 50–200 nm [103], [105]. The energy gap for coherently strained Sn_xGe_1−x alloys indicates a large alloy contribution and a small strain contribution to the decrease in direct energy gap with increasing Sn composition. These results are consistent with a deformation potential model for changes in the valence and conduction band density of states (DOS) with coherency strain for this alloy system [105]. Yang and co-workers have very recently made great progresses in the synthesis and characterization of one-dimensional Si/SiGe SLs exhibiting great potential for applications in electronics because they could function as a transistor, light-emitting diode, biochemical sensor, and heat-pumping thermoelectric device simultaneously [107]. The advantages of SiGeC/Si for optoelectronic devices are adjustable strain, band gap, and band offsets; all of which come from tailoring the Si_1−x−yGe_xC_y composition. It felt that SiGeC would offer the same band offsets as SiGe/Si with less strain. Substitution of C atoms for Ge atoms in SiGe/Si compensates for Ge-induced strain because of carbon’s smaller atomic size. Perfect compensation (zero net strain) occurs when the x–y choice gives a lattice match to Si [103].

On an other hand, with the rapid development of very large-scale integration technology, the number of components per integrated circuit chip is increasing considerably and the power density increases accordingly. Device performance and reliability degrade significantly when devices are overheated. Heat generation and thermal management are becoming one of the barriers to further increases in clock speed and decreases of feature size. As SiGe is a good thermoelectric material for high temperature applications [108], SiGe/Si and SiGe/SiGe SLs structures are used to enhance the cooler performance by reducing the thermal conductivity between the hot and the cold junctions [109], [110] and by selective emission of hot carriers above the barrier layers in the thermionic emission process [111]. Si and SiGe devices can integrate directly with these coolers to achieve an high device performance [100], [112], [113]. The SiGe technology has also found applications in the field of solar cells [114], [115].

Finally, the development of SiC research is mainly supported and justified by the needs to elaborate a new generation of devices, which can work in very extreme conditions. Because of this need, SiC-based devices have been considered as an alternative to the silicon and silicon-based devices, which are currently playing an important role in modern semiconductor electronic, and optoelectronic device. The heterostructure SiC/GaN/AlN is also guessed for laser devices applications emitting in the blue/ultraviolet spectrum which are characterized by high thermal conductivity, high breakdown electric field, high forward current density, high saturated electron drift velocity, high electronic mobility and high blocking voltage [116].

High quality devices can be realized only if the layers forming the heterostructure exhibit a perfect crystallinity, a request that can in turn be fulfilled only if the lattice parameter of the epitaxial layer (a) is almost equal to that of the substrate (a_s). By “almost” we mean that |a−a_s|/a should be in the order of few units in 10⁻⁴. In such a case we are under lattice match conditions and very thick films can be grown under pseudomorphic regime. Conversely, for |a−a_s|/a in the order of some units in 10⁻³, we are under lattice mismatch conditions and only thin epitaxial film can be deposited on the substrate under pseudomorphic regime. The maximum growth thickness for a pseudomorphic film on a given substrate is called thickness T_c and its value decreases dramatically by increasing |a−a_s|/a. Such films are strained and the cell of their lattice is distorted to fit with the substrate cell. Notwithstanding the great technological problems related to the growth of pseudomorphic strained films, the use of intentionally strained heterostructures has greatly enhanced the performances of electronic and optoelectronic devices [117], [118]. In fact, strain-based heterostructures offer further advantages in that the energy band lineups can be shifted by the strain (vide infra Section 2.2), giving an added flexibility in the design of the devices [119]. The presence of strain reduces the crystal symmetry and modifies the energy band lineups [15], [117], [118], [120], [121], [122], [123], [124], [125] (vide infra Section 2.2). This is the reason why the effects of strain have been employed in the field of: InGaAs/InP [43], [46], [49], [126], InGaAs/GaAs [127], InAsP/InP [128], GaAs/AlGaAs [37], [129], InGaAs/AlAs [130], InGaAs/InGaAlAs [35], InGaAs/InGaAsP [48] QW lasers; compressive-strained InGaAsP QW lasers [44], [131], [132], [133]; tensile-strained GaAs_1−yP_y/Al_0.35Ga_0.65As QW lasers [39]; alternating tensile/compressive (i.e. zero net strain) InGaAsP QW lasers [47], [134]; InGaAs/InP avalanche photodetectors [135] (see also the recent review by Brennanm and Haralson [136]); InGaAs/GaAs high-transconductance p-channel field-effect transistors [137], [138]; AlGaN/GaN field-effect transistors [139]; AlInAs/InGaAs [140], [141], [142] and AlGaAs/GaAs [19] high electron mobility transistors; strained InAs/InAsSb [64] and InGaAsSb/AlGaAsSb [65] mid-IR lasers; strain-compensated InAsP/InGaP electroabsorption modulators [143]; high-performance AlGaN/GaInN strain-compensated QW lasers [97], [98]; InGaAsSb/AlGaAsSb QW mid-IR lasers [65], [144]; and high power LEDs (mid-IR emitting) InAs/InAsSb strained SL [145]. Coming to IV–IV-based heterostructures, Si_1−xGe_x/Si strained alloys and SL results particularly attracting for the achievement of novel microdevices directly integrable in the Si-based technology [100], [112], [113], vide supra.

On the other hand, even for intentionally unstrained heterostructures, several studies [146], [147], [148], [149], [150], [151], [152], [153], [154], [155], [156], [157], [158], [159], [160], [161], [162], [163] have shown that undesirable strained monolayers (MLs) are present at the heterointerfaces. The presence of strained interface layers is related to the impossibility to realize, during the growth, instantaneous switches between wells and barriers. This compositional interface chemical gradient, spreading over a distance of some MLs has been observed by several groups for different systems [40], [45], [127], [146], [158], [159], [160], [161], [162], [163], [164], [165], [166], [167]. It is well known that the performances of the processed devices are influenced by the quality of the interfaces [42], [168]. In fact, interface crystalline imperfections—such as planarity and compositional grading [147]—cause scattering processes, yielding to a reduction of the exciton decay time [169], [170], to a limitation on the electron mobility [171] and to an increase of the non-radiative recombination [127], [172]. It is hence evident that a detailed characterization of the interfacial layers is very important for their optimization. This implies again the need of studying the properties of strained layers. This is the reason why the consequences of strain on II–VI, III–V and IV–IV semiconductor heterostructures have been investigated so widely on both experimental and theoretical grounds [12], [15], [17], [38], [117], [142], [160], [162], [163], [164], [165], [166], [167], [173], [174], [175], [176], [177], [178], [179], [180], [181], [182], [183], [184], [185], [186], [187], [188], [189], [190], [191], [192], [193], [194], [195], [196], [197], [198], [199], [200], [201], [202], [203].

Improvements in the realization of the devices above mentioned have been realized by a strict interplay among the progress achieved on three grounds: (A) theoretical solid-state physics (or quantum chemistry) aimed to predict the characteristic of an ideal heterostructure [162], [163], [164], [194], [200], [204], [205], [206], [207], [208], [209], [210], [211], [212], [213], [214], [215], [216], [217], [218], [219], [220], [221], [222], [223], [224], [225], [226], [227], [228], [229], [230], [231], [232], [233], [234], [235], [236], [237], [238], [239], [240], [241], [242], [243], [244], [245], [246], [247], [248], [249], [250], [251], [252], [253], [254], [255], [256], [257], [258], [259], [260], [261], [262], [263], [264], [265], [266], [267], [268], [269], [270], [271], [272], [273], [274], [275]; (B) epitaxial growth techniques aimed to realize heterostructure as close as possible to the desired target [43], [84], [85], [92], [116], [118], [131], [147], [149], [152], [155], [158], [159], [161], [276], [277], [278], [279], [280], [281], [282], [283], [284], [285], [286], [287], [288], [289], [290], [291], [292], [293], [294] and (C) sophisticated characterization techniques aimed to verify the closeness between the target and the real heterostructure [147], [158], [159], [160], [161], [164], [165], [166], [167], [188], [189], [190], [191], [279], [281], [287], [288], [291], [292], [293], [294], [295], [296], [297], [298], [299], [300], [301], [302], [303], [304], [305], [306], [307], [308], [309], [310], [311], [312], [313]. Of particular interest are the in situ characterization techniques performed during heterostructure growth and thus combining directly topics (B) and (C) [287], [288], [292], [293], [294], [296], [297]. In this field, beside the classical electron and ion surface probes, used in laboratory, synchrotron radiation (SR) techniques have shown tremendous potentialities [287], [292], [293], [294], [296], [297].

The interplay among the fields (A), (B) and (C) can be basically schematized in the following flow chard: (i) theoretical solid-state physics predicts the physical properties of a given heterostructure; (ii) epitaxial growth techniques try to realize it; (iii) structural characterization techniques check whether the actually realized heterostructure corresponds to the desired one or not; (iiia) if not the growth parameters have to be optimized and step (ii) has to be repeated; (iiib) if yes, then optical, electrical and electronic characterization techniques check weather the desired heterostructure has actually the foreseen physical properties; (iva) if not then the level of theory used in step (i) has to be improved and the game has to restart again from the beginning; (ivb) if yes then, end of the story. Point (ivb) represents the final point of the scientific work and the future of the device lies now on an engineering/economical level where the production rate, the realization costs and the demand of the device are the main driving forces. Of course, the interplay can also move in the opposite direction, e.g. when theoretical models help in the interpretation of previous non-understood (or wrongly interpreted) experimental results.

Stimulation for the improvement of each of the three branches (A–C) originates from the requests coming from the remaining two. A short and non-exhaustive list of examples follows: the need of realizing short period SL has improved the realization of fast switches apparatus in the growth chambers; the need of having flat interfaces for testing theoretical models and innovative characterization techniques on an almost ideal heterostructure has stimulated the born of atomic layer epitaxy (ALE, which is useless on an industrial scale due to the too low growth rate); the need of characterizing ML-thick interfaces has stimulated the development of sophisticated and powerful characterization techniques (such as those based on synchrotron radiation); high quality experimental results have given important check values for ab initio models, while unexpected experimental results have asked for an improvement of the level of theory used so far, etc.

It is evident that the global problem just mentioned in Section 1.3 is so large to need several books to be faced. The aim of this review is then focused on a subset of topic (C), mainly synchrotron radiation-based techniques, which still remain a pretty large topic. In this domain we find both structural and electronic characterization techniques i.e. what needed to face points (iii) and (iv) of the ideal flow chard discussed in the previous section. Of course, when needed the interplay between results obtained from synchrotron radiation techniques and other characterization techniques, theoretical models and growth techniques will be underlined.

The important role played by synchrotron radiation techniques in the characterization of semiconductor heterostructures is due to the fact that synchrotron radiation has distinct advantages as a photon source, notably high brilliance and continuous energy spectrum [314], [315], [316], [317], [318], [319], [320], [321], [322], [323], [324], [325], [326], see Fig. 1. By using the latter characteristic atomic selectivity can be obtained and this is of fundamental help to investigate the structural environment of atoms present only in a few angstrom (Å) thick interface layers of heterostructures. The third generation synchrotron radiation sources have allowed reaching the limit of measuring a monolayer of material, corresponding to about 10¹⁴ atoms/cm².

The materials of interest are the semiconductor heterostructures (QW, MQW and SL) mentioned in 1.1 The technological impact of semiconductor heterostructures: a brief overview, 1.2 Strained heterostructures but, when needed, also thin films are discussed, because they represent a less complex systems with respect to the heterostructures used in the devices, from which the fundamental physics of epitaxy and lattice strain can be learnt easier.

This review begins with a brief overview on the theoretical fundamentals of band alignment in strained heterostructures (Sections 2.1–2.3), followed by a detailed analysis of band profiles in ideal semiconductor heterostructures characterized by chemically abrupt interfaces (2.4 Band profiles in unstrained or nearly unstrained SQWs with abrupt interfaces, 2.5 Band profiles in strained SQWs with abrupt interfaces). In Section 2, the effect that the parallel (ε_∥) and the perpendicular (ε_⊥) strain have on the band alignment is defined. In Section 3 the problems related with the growth of a real heterostructure are briefly discussed together with the consequences that the actual (non-ideal) composition of the heterostructure has on its bands profile. The conventional characterization techniques used to investigate the interface quality is discussed in Section 3.1. The following five sections are devoted to describe five structural characterization techniques that give direct information on ε_∥, ε_⊥ and on the interface mixing of the heterostructures: extended X-ray absorption fine structure (XAFS; Section 4); high-resolution XRD (Section 5); diffraction anomalous fine structure (DAFS; Section 6); X-ray standing waves (XSW; Section 7) and X-ray reflectivity (Section 8). Finally, Section 9, discuss the results obtained with X-ray photoemission spectroscopies and closes the loop, giving access to the experimental values for the band offsets, predicted in accord to the methods described in Section 2 for a given strain (which can be measured with the techniques described in 5 Application of high-resolution XRD, 6 Application of DAFS, 7 Application of XSW, 8 Application of X-ray reflectivity: basic concepts).

Each section is divided into subsection, whose content is briefly summarized at the begin of the section. The different sections can be read separately, this allows the reader interested on a given characterization technique to focus his attention on the corresponding section only. There are several topics that are discussed along the review in the different sections because different characterization techniques have provided important information. In such cases, in all sections where the topic has been treated reminds to the other pertinent sections are made.