A. Bauer, et al., page 1
Atomic scale interface engineering for strain compensated epitaxially grown InAs/AlSb superlattices A Bauer*, M Dallner, A Herrmann, T Lehnhardt, M Kamp, S Höfling, L Worschech and A Forchel Technische Physik, Physikalisches Institut and Wilhelm-Conrad-Röntgen-Research Center for Complex Material Systems, Julius-Maximilians-Universität Würzburg, Am Hubland, D-97074 Würzburg, Germany *
e-mail:
[email protected], phone: +49-931-31-83480, fax: +49-931-31-85143
Abstract This paper presents a systematic investigation of strain compensation schemes for InAs/AlSb superlattices (SLs) on GaSb substrates. Short growth interruptions (soak times) under varying Arsenic and/or Antimony beam equivalent pressures in InAs/AlSb SLs with exemplary dimensions of about ((2.4/2.4)±0.2) nm were investigated to achieve strain compensation. When using uncracked As4, strain compensation was found to be unaccomplishable unless sub-monolayer AlAs spikes were inserted at the InAs→AlSb interface. In contrast, the supply of cracked As2 dimers leads directly to the formation of strain compensating AlAs-like interfaces. This mechanism allows various growth sequences for strain compensated superlattices, including soak-time-free and Sb-soak-only SL growth. Each of the two latter approaches yields layers with excellent crystal quality and minimal intermixing at the heterointerfaces as verified by high-resolution x-ray diffraction analysis and transmission electron microscopy.
A. Bauer, et al., page 2
1. Introduction InAs/AlSb heterostructures are very attractive for various application prospects like high speed fieldeffect transistors [1] and further novel transport structures [2, 3], particularly due to the small electron effective mass in InAs [4, 5] and the large conduction band offset of 1.35 eV [6]. Thus, systematic investigations of this material combination for use in (opto-)electronic structures have been carried out both theoretically [7] and experimentally [8] already back in the 1980s and research in this area has been continuously conducted since then [9]. Shortly after first epitaxial structures were examined, it became clear that electronic and optical properties were strongly dependent on the type of interface (IF) evolving in alternating InAs and AlSb layers, since, due to changing both, anions and cations at such interfaces, either InSb- or AlAs-like bonds can be forced by implementing according shutter cycles [10]. Numerous research groups have characterized InAs/AlSb structures with different probing techniques including Hall measurements [11-13], scanning tunneling microscopy [14, 15], x-ray scattering and diffraction [16] and Raman scattering [17]. They agree on the basic conclusion that the lower IF with respect to the InAs layer has to be preferably forced to be InSb-like in order to achieve superior (opto-)electronic device performance [18]. However, the issue of lattice strain, introduced by forcing respective bonding types and its dependency on using tetramers or dimers as group V elements, has merely been reported as side effect. While strain introduced at the heterointerfaces indeed generally plays a subsidiary role in structures where only some few ‘X’As/’Y’Sb layers are implemented, small period InAs/AlSb superlattice (SL) structures which are widely used e. g. in Sb-based quantum cascade lasers (QCLs) [19] or interband cascade lasers (ICLs) [20-22] have to be perfectly strain compensated to achieve the required SL thicknesses of some microns, corresponding to several hundreds of SL periods, with high crystal quality. With lattice constants of aInAs=6.0583 Å and aAlSb=6.1355 Å the relative lattice mismatches of InAs and AlSb to the generally used GaSb substrate (aGaSb=6.0959 Å) are -0.62 % and +0.65 %, respectively [23]. On one hand, these relatively small mismatches allow for pseudomorphical growth of single InAs or AlSb layers of some ten nanometer thickness on GaSb without significant defect
A. Bauer, et al., page 3
formation [24, 25]. On the other hand, the strain in more sophisticated and especially in the widely required SL structures rapidly builds up beyond critical limits if no compensating measures are adopted. Since in binary semiconductors strain compensation via stoichiometric alloy adjustments is not an option, this issue has to be addressed differently. Implementing short growth interruptions under arsenic or antimony pressure (so-called soak times) by using appropriate shutter sequences, AlAs- or InSb-like bonds can be formed at the heterointerfaces of InAs/AlSb SLs [10]. Due to these materials’ large offsets in lattice constants of -7.26 % and +6.29 % relative to GaSb [23], these 1-2 monolayers (MLs) represent a highly efficient way of strain engineering in such samples.
In this work, various shutter sequences at the interfaces of InAs/AlSb SLs with exemplary dimensions of ((2.4/2.4)±0.2) nm supplying either cracked As2 or uncracked As4 during growth are investigated. Impact on overall SL strain and resulting crystal quality is analyzed in terms of high resolution x-ray diffraction (HR-XRD) scans accompanied by secondary ion mass spectroscopy (SIMS) measurements and scanning transmission electron microscopy (STEM). Optimized shutter cycles allow for perfectly strain compensated SL growth with Sb-soaks-only or no soak times at all when utilizing As2 dimers. On the contrary the reduced arsenic-incorporation due to less chemical reactivity of As4 tetramers had to be compensated by inserting sub-ML AlAs spikes at the InAs→AlSb interfaces. In each case the intermixing zone at the heterointerfaces was found to be less than 1-2 ML wide. Whereas for growth using As4 the interfaces remained InSb-like under all growth conditions and soak times, an additional degree of freedom in small period SL growth is introduced, when operating the arsenic cell in cracking mode, in which the interface can be forced to be either of AlAs- of InSb-type. Moreover, those interfacial bonds were found having a more decisive influence on the emerging mean lattice constant than the respective InAs/AlSb layer widths themselves. Thus, even layer sequences with significantly varying layer thicknesses can be tailored to zero total strain by exploiting the approaches described in this paper.
2. Sample layout and fabrication
A. Bauer, et al., page 4
All samples were grown using an Eiko EV100S molecular beam epitaxy chamber equipped with standard solid-source effusion cells for the group III elements and an antimony cracking cell. The two initially installed solid-source arsenic cells were later-on replaced by an additional cracker to allow for dimeric molecular fluxes of both group V elements. A 200 nm thick GaSb buffer layer was deposited on 2” epi-ready (100) GaSb:Te substrates prior to the growth of 30 SL periods consisting of ((2.4/2.4)±0.2) nm InAs/AlSb layers with varying shutter sequences at the interfaces and growth rates of 500 nm/h each (see table 1 for a summary of the samples). The antimony beam equivalent pressure (BEP) was held constant at about (1.6±0.2)x10-6 torr whereas the As4 and As2 BEPs were chosen to be (1.0±0.1)x10-5 torr and (4.2±0.2)x10-6 torr, respectively. The structures were finally capped with either 20 nm GaSb or InAs. Table 1: Shutter cycles and soak-times of the investigated SL structures as well as resulting overall strain and mean SL period as extracted from corresponding HR-XRD scans. Bold values indicate the use of As2 instead of As4.
shutter cycle Ia Ib Ic IIa IIb IIc IIIa IIIb IV V
nom. AlSb width [nm] 2.4 2.4 2.4 2.4 2.4 2.4 2.3 2.3 2.3 2.3
soak-time [s] Sb2 As4/2Sb2 As4/2 6.0 1.0 6.0 1.0 1.0 6.0 1.0 6.0 6.0 6.0 3.0 6.0 1.0 6.0 6.0 1.0 6.0 2.0 -
nom. InAs soak-time [s] strain period width [nm] As4/2 As4/2Sb2 Sb2 [arcsec] [nm] 2.4 6.0 1.0 6.0 -785 4.89 2.4 6.0 1.0 1.0 -620 4.89 2.4 6.0 -395 4.86 2.4 6.0 -390 4.98 2.4 6.0 0.1 nm AlAs ±0 4.97 2.4 3.0 0.1 nm AlAs ±0 4.97 2.3 6.0 1.0 6.0 -1053 4.62 2.3 6.0 +353 4.60 6.0 1.0 2.3 +110 4.61 2.3 2.0 -217 4.61
The SL dimensions mentioned above were chosen since they are commonly used in the cladding layers of today’s ICL structures. These layers have a total thickness of up to 3 microns [26], corresponding to several hundred layer repetitions, rendering precise and reliable strain compensation obviously most crucial. Additionally, growing InAs and AlSb layers of equal thickness provides direct access to the strain introduced by the interfaces via standard HR-XRD scans, since simulations assuming perfectly abrupt material transitions reveal virtually no resulting strain at all because of the
A. Bauer, et al., page 5
almost identical relative lattice parameter offsets of InAs and AlSb compared to GaSb. Thus, any strain measured in such samples can directly be attributed to the configuration at the interfaces.
3. Results and discussion 3.1. Superlattice growth using As4 Since no cracker source for arsenic was available at the beginning of our SL growth studies, As4 tetramers were used during growth while antimony was provided in cracked form to ensure reasonable crystal quality in antimony-containing bulk layers. In a first approach, shutter sequence Ia (see table 1) was used at both heterointerfaces, implementing temporally symmetric soaks with durations of 6s/1s/6s As4(Sb2)/As4Sb2/Sb2(As4) when switching from InAs (AlSb) to AlSb (InAs) layer growth. None of the two group V elements is favored in this sequence in terms of temporal exposure time of the substrate. Since the mismatch of InAs and AlSb with respect to GaSb is almost identical, the mean lattice parameter of the SL should be very close to that of the substrate if the exchange of As4 for Sb2 and Sb2 for As4 at the interfaces was equally efficient. However, due to the much higher chemical reactivity of the Sb2 dimers, InSb-like interfaces are predominantly developing and the SL is strongly compressively strained by –785 arcsec, as depicted in terms of an HR-XRD scan (red line) in figure 1.
Figure 1: (color online) HR-XRD spectra of three nominally identical SL structures with 30 repetitions of (2.4/2.4) nm InAs/AlSb using cracked Sb2 and uncracked As4 as group V sources each, but with different shutter cycles Ia (red), Ib (green) and Ic (blue) at the heterointerfaces (for details see table 1). Even without Sb2-soaks at the interfaces (cycle Ic) the SL still remains compressively strained by -395 arcsec, despite the identical layer widths.
A. Bauer, et al., page 6
To reduce the strain in the SL structures to reasonable values (within ±200 arcsec), some fraction of the InSb bonds at the interfaces has to be replaced by AlAs-type bonds in order to achieve a smaller overall lattice constant. Therefore, growth interruptions under Sb2 flux were first shortened to 1 s at each side (cycle Ib). Since this resulted only in a slight decrease of the strain to -620 arcsec, both Sb2 and Sb2As4 soak times at the InAs layer surface were totally omitted during the growth of a third sample (cycle Ic). Nevertheless, -395 arcsec of strain still persist in this case, as can be seen from the position of the corresponding 0th order SL peak in figure 1 (blue). Without the possibility to provide cracked As2 molecules, the means to achieve lattice matching are rather limited. Theoretically, increasing the time of growth interruptions under As4 flux should lead to a higher As4 ↔ Sb2 exchange rate at the interfaces but at the same time raises the background pressure in the MBE chamber. Thus, disadvantageous incorporation of residual arsenic in the adjacent AlSb layers would emerge and the same would happen when using a higher As4 flux. Lowering the Sb2 flux during growth to minimize the residual antimony pressure after the shutter has already been closed is also undesirable, since the Sb2 BEP of (1.4-1.6)x10-6 is already very close to our MBE systems’ lower limit for high quality epitaxial antimony layer growth with the chosen growth rates.
The solution to this issue was finally found by inserting an intentionally grown AlAs layer of sub-ML thickness (about 0.1 nm) right before the AlSb growth (cycle IIb). This way, the InAs→AlSb interface is obviously forced to be almost entirely AlAs-like and the mean SL period is perfectly lattice matched to the underlying GaSb substrate (figure 2, red).
A. Bauer, et al., page 7
Figure 2: (color online) HR-XRD scan of a 30x (2.4/2.4)nm InAs/AlSb SL structure with applied shutter cycles IIb (red) and IIc (green). The implemented sub-ML AlAs spikes render the SL strain independent from the time of As4 soak in-between the layers, thus increasing the SL reproducibility.
Additionally, the SL still remains strain compensated when lowering the As4 soak time to 3 s only (figure 2, green) according to cycle IIc. This indicates that the AlSb→InAs interfaces remain almost completely InSb-like regardless of As4 soak time length or chosen BEP (since both layers are of the same thickness and the InAs→AlSb interface is forced to be AlAs-like because of the grown AlAsspike), which improves the reproducibility of the SL strain compensation significantly. Well-defined heterointerfaces with remarkable crystal quality, as indicated by the sharp and intense SL peaks in the XRD spectra could be verified using scanning electron microscopy (SEM) and scanning transmission electron microscopy (STEM). Complementary SIMS measurements applying low sputtering rates for increased depth resolution were carried out, to assess a higher statistical reliability due to examining a relatively large sample area of typically 100x100 µm2 (figure 3). These indicate a measurable but almost negligible increase in overall arsenic level in the samples grown with cycle IIb containing AlAs-spikes (dark brown) compared to the ones grown with cycle IIa without AlAs-spikes (light brown), once more affirming the suitability of this approach for producing high quality epitaxial layers when lacking an arsenic cracking source.
A. Bauer, et al., page 8
Figure 3: (color online) Upper part: SIMS depth profile of a (2.4/2.4) nm InAs/AlSb SL sample using shutter cycle IIa (without AlAs-spikes), showing Indium2 (blue) and Aluminium2 (red) signals, overlaid on top of a STEM image of a typical SL section. Lower part: Arsenic ion count rate of the same sample (light brown) and a nominally identical second one but grown with AlAs-spikes (dark brown) according to cycle IIb, indicating a traceable but very minor increased overall arsenic content in the latter.
3.2. Superlattice growth using As2 As soon as one utilizes an As2 source instead, the arsenic for antimony exchange at the heterointerfaces becomes extremely efficient. This is verified by HR-XRD spectra (figure 4) of two nominally identical samples with 30 repetitions of a (2.3/2.3) nm InAs/AlSb SL. During the growth of these samples, the valve positions of both group V cells were left unchanged and identical shutter cycles IIIa and IIIb in-between layer growth have been used. The only difference is the use of uncracked As4 (cycle IIIa, red) and mostly As2 dimers (cycle IIIb, green). Solely due to the increased chemical reactivity of dimeric As2, a shift of the SL strain from -1053 arcsec compressive to +353 arcsec tensile can be observed.
A. Bauer, et al., page 9
Figure 4: (color online) The overall strain in nominally identical samples containing a 30x (2.3/2.3) nm InAs/AlSb SL changes from -1053 arcsec compressive to +353 arcsec tensile strain when switching from uncracked (red) to cracked (green) arsenic flux.
Apparently, with either group V element in dimeric form, an additional degree of freedom during SL growth is introduced, allowing an engineering of the resulting strain over a rather large scale when applying proper soak times and adequate BEPs. In particular, strain compensated soak-time free and Sb2-soak only SL growth becomes feasible when applying proper arsenic and antimony BEPs (table 1, cycles IV and V, respectively). Although one might expect increased interface roughness if no soak times are applied to smooth the respective semiconductor surface before switching to the next material, HR-XRD measurements and high resolution STEM images reveal defect-free SLs with close to perfectly straight and parallel layers. Figure 5 depicts the InAs→AlSb heterointerface of an according STEM sample, which was prepared using a FEI dual-beam system (Helios nanolab) and then imaged in a FEI Titan 80-300 TEM system operated at 300 kV. Detection of electrons scattered under large angles (high angle annular dark field) gives rise to a signal that is roughly proportional to the square of the atomic number. Therefore, the indium-columns show up bright in the InAs dumbbells whereas the antimony-columns appear brighter in the AlSb dumbbells. Thus, it can directly be seen, that the transition region between the two binaries is smaller than a single lattice period. Moreover, ICL devices with claddings grown using exactly this shutter cycle IV showed pulsed operation at room temperature and above without any noticeable degradation in device performance
A. Bauer, et al., page 10
compared to structures grown with soaked SL claddings. The benefit of a soak-free growth is a 10-12 % reduction of the growth time due to the omission of growth interruptions, thus avoiding unnecessary group III material consumption and integration of residual gas molecules during group V-only soak times.
Figure 5: (color online) STEM image of the InAs→AlSb interface of a 2.3/2.3 nm InAs/AlSb SL grown without any soak times at all (cycle V). A schematic projection of the InAs and AlSb lattice along the (110) axis is superimposed on the STEM image. Intermixing at the interface is confined to no more than about one lattice constant.
Although, in the initial studies characterizing further the structural, optical and electric properties of InAs/AlSb SLs, the combinations As4 and Sb4 [1, 10, 11] as well as As2 and Sb2 [14, 15] were used, potential differences when switching from tetramers to dimers have not been addressed. Therefore, it should be quite interesting to see to what extent the huge differences in resulting strain, when switching from As4 to As2, reflect themselves in similar ways there, when using analogue probing
techniques to further investigate this issue. Especially the occurrence of reported growth asymmetries [14, 15], extractable via scanning tunneling microscopy, might be directly dependent on the type of utilized group V molecules. However, although for quantitative conclusions further
systematic studies on additional samples would be required, shown HR-XRD and STEM analysis already suggest a reasonable SL quality in any of the investigated approaches. Nevertheless, for optimal device performance, forcing InSb-like interfaces at either layer side is ultimately preferential due to the improved structural, optical [18] and transport properties [10] of such SLs. Consequently, it is especially beneficial in the ICLs’ optically active injector regions, consisting
A. Bauer, et al., page 11
of about ten InAs/AlSb layering pairs with thicknesses ranging from (2.4/1.5) nm up to (12.0/1.2) nm of InAs/AlSb [27]. Particularly for high ratio InAs/AlSb layering thicknesses, long antimony soak times on either interface might already be sufficient for achieving perfect strain compensation instead of having to use stronger strain compensating but less confining AlxIn1-xSb as barrier material [22].
4. Conclusion In summary, the impact of different shutter cycles at the interfaces of widely used ((2.4/2.4)±0.2) nm InAs/AlSb SL structures on resulting strain has been investigated. Arsenic molecules were provided either in tetrameric (As4) or dimeric (As2) form. For the growth with As4, the insertion of sub-ML AlAs layers was found to be essential for strain compensation due to the reduced chemical reactivity of As4 compared to cracked As2. Using this approach, high quality short period SLs can be grown without an arsenic cracker. The supply of As2 dimers instead allows efficiently engineering the ratio of interfacial AlAs- and InSb-like bonds via soak-times at the heterointerfaces to achieve totally strain compensated InAs/AlSb SLs. Any of the investigated approaches, including soak-time free and even Sb2-soak-only growth provided excellent crystal quality, verified by high resolution XRD, SIMS and STEM measurements, the latter proving an intermixing layer of no more than 1-2 ML thickness at the heterointerfaces even in the soak-time-free approaches. Thus, by purposely engineering the heterointerfaces in InAs/AlSb interfaces to preferentially develop either AlAs- or InSb-like or mixed atomic bonds, an additional degree of freedom in epitaxial growth of such structures is introduced, allowing for virtually perfectly strain compensated layer growth, being highly beneficial for ‘X’As/’Y’Sb (opto-)electronic device approaches.
Acknowledgments
Expert technical support in sample preparation by M. Emmerling, S. Handel, S. Kuhn, M. Wagenbrenner and A. Wolf is gratefully acknowledged as are the fruitful discussions with S. Hein and D. Bisping. This work has been funded by the European Commission in the frame of the FP7 project 'SensHy' (grant no. 223998) and by the State of Bavaria.
A. Bauer, et al., page 12
REFERENCES
[1]
G Tuttle and H Kroemer 1987 IEEE Trans. Electron Devices 34 2358
[2]
Li J, Yang W and Chang K 2009 Phys. Rev. B 80 035303
[3]
Liu C, Hughes T L, Qi X-L, Wang K and Zhang S-C 2008 Phys. Rev. Lett. 100 236601
[4]
S Ideshita, A Furukawa, Y Mochizuki and M Mizuta 1992 Appl. Phys. Lett. 60 2549
[5]
M J Yang,
P J Lin-Chung,
R J Wagner,
J R Waterman,
W J Moore
and
B V Shanabrook 1993 Semicond. Sci. Technol. 8 S129-S132 [6]
A Nakagaws, H Kroemer and J H English 1989 Appl. Phys. Lett. 54 1893
[7]
L Esaki, L L Chang and E E Mendez 1981 Jpn. J. Appl. Phys. 20 L529-L532
[8]
C Chang, L L Chang, E E Mendez, M S Christie and L Esaki 1984 J. Vac. Sci. Technol. B 2 214
[9]
B R Bennett, R Magno, J B Boos, W Kruppa and M G Ancona 2005 Solid-State Electron. 49 1875–1895
[10]
G Tuttle, H Kroemer and J H English 1990 J. Appl. Phys. 67 3032-3037
[11]
G Tuttle, H Kroemer and J H English 1989 J. Appl. Phys. 65 5239-5242
[12]
C R Bolognesi, H Kroemer and J H English 1992 J. Vac. ScL Technol. B 10 877879
[13]
C R Bolognesi, H Kroemer and J H English 1992 Appl. Phys. Lett. 61 213-215
[14]
J Steinshnider, M Weimer, R Kaspi and G W Turner 2000 Phys. Rev. Lett. 85 2953-2956
[15]
J Steinshnider, J Harper, M Weimer, C H Lin, S S Pei and D H Chow 2000 Phys. Rev. Lett. 85 4562-4565
[16]
B Jenichen, S A Stepanov, B Brar and H Kroemer 1995 J. Appl. Phys. 79 120
A. Bauer, et al., page 13
[17]
J Wagner, J Schmitz, M Maier, J D Ralston and P Kiodl 1994 Solid-State Electron. 37 1037-1040
[18]
J Spitzer, A Hijpner, M Kuball, M Cardona, B Jenichen, H Neuroth, B Brar and H Kroemer 1995 J. Appl. Phys. 77 811-820
[19]
Vurgaftman I, Meyer J R, Julien F H, Ram-Mohan L R 1998 Appl. Phys. Lett. 73 711-713
[20]
Yang R Q 1995 Superlatt. Microstruct. 17 77-83
[21]
Meyer J R, Hoffman C A, Bartoli F J and Ram-Mohan L R 1995 Appl. Phys. Lett. 67 757
[22]
C L Canedy, W W Bewley, J R Lindle, C S Kim, M Kim, I Vurgaftman
and
J R Meyer 2006 Appl. Phys. Lett. 88 161103 [23]
Vurgaftman I, Meyer J R and Ram-Mohan L R 2001 J. Appl. Phys. 89 5815-5875
[24]
Gossmann H J, Davidson B A, Gualtieri G J, Schwartz G P, Macrander A T, Slusky S E, Grabow M H and Sunder W A 1989 J. Appl. Phys. 66 1687-1694
[25]
Bennett B R 1998 Appl. Phys. Lett. 73 3736-3738
[26]
Hill C J and Yang R Q 2005 J. Crys. Growth 278 167-172
[27]
Bauer A,
Langer F,
Dallner M,
Kamp M,
Motyka M,
Sek G,
Misiewicz J, Höfling S and Forchel A 2009 Appl. Phys. Lett. 95 251103
Ryczko K,
