-
Nucleation-related defect-free GaP/Si(100) heteroepitaxy via metal-organic chemicalvapor depositionT. J. Grassman, J. A. Carlin, B. Galiana, L.-M. Yang, F. Yang, M. J. Mills, and S. A. Ringel Citation: Applied Physics Letters 102, 142102 (2013); doi: 10.1063/1.4801498 View online: http://dx.doi.org/10.1063/1.4801498 View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/102/14?ver=pdfcov Published by the AIP Publishing Articles you may be interested in Dislocation reduction through nucleation and growth selectivity of metal-organic chemical vapor deposition GaN J. Appl. Phys. 113, 144908 (2013); 10.1063/1.4799600 Control and elimination of nucleation-related defects in GaP/Si(001) heteroepitaxy Appl. Phys. Lett. 94, 232106 (2009); 10.1063/1.3154548 Oxygen induced strain field homogenization in AlN nucleation layers and its impact on GaN grown by metalorganic vapor phase epitaxy on sapphire: An x-ray diffraction study J. Appl. Phys. 105, 033504 (2009); 10.1063/1.3074095 Ductile relaxation in cracked metal-organic chemical-vapor-deposition-grown AlGaN films on GaN J. Appl. Phys. 97, 123504 (2005); 10.1063/1.1929856 Comparative study of GaN and AlN nucleation layers and their role in growth of GaN on sapphire bymetalorganic chemical vapor deposition Appl. Phys. Lett. 77, 3391 (2000); 10.1063/1.1328091
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
169.228.173.121 On: Wed, 29 Oct 2014 02:45:21
http://scitation.aip.org/content/aip/journal/apl?ver=pdfcovhttp://oasc12039.247realmedia.com/RealMedia/ads/click_lx.ads/www.aip.org/pt/adcenter/pdfcover_test/L-37/615868548/x01/AIP-PT/Asylum_APLArticleDL_102914/AIP-JAD-Cypher.jpg/47344656396c504a5a37344142416b75?xhttp://scitation.aip.org/search?value1=T.+J.+Grassman&option1=authorhttp://scitation.aip.org/search?value1=J.+A.+Carlin&option1=authorhttp://scitation.aip.org/search?value1=B.+Galiana&option1=authorhttp://scitation.aip.org/search?value1=L.-M.+Yang&option1=authorhttp://scitation.aip.org/search?value1=F.+Yang&option1=authorhttp://scitation.aip.org/search?value1=M.+J.+Mills&option1=authorhttp://scitation.aip.org/search?value1=S.+A.+Ringel&option1=authorhttp://scitation.aip.org/content/aip/journal/apl?ver=pdfcovhttp://dx.doi.org/10.1063/1.4801498http://scitation.aip.org/content/aip/journal/apl/102/14?ver=pdfcovhttp://scitation.aip.org/content/aip?ver=pdfcovhttp://scitation.aip.org/content/aip/journal/jap/113/14/10.1063/1.4799600?ver=pdfcovhttp://scitation.aip.org/content/aip/journal/apl/94/23/10.1063/1.3154548?ver=pdfcovhttp://scitation.aip.org/content/aip/journal/jap/105/3/10.1063/1.3074095?ver=pdfcovhttp://scitation.aip.org/content/aip/journal/jap/105/3/10.1063/1.3074095?ver=pdfcovhttp://scitation.aip.org/content/aip/journal/jap/97/12/10.1063/1.1929856?ver=pdfcovhttp://scitation.aip.org/content/aip/journal/apl/77/21/10.1063/1.1328091?ver=pdfcovhttp://scitation.aip.org/content/aip/journal/apl/77/21/10.1063/1.1328091?ver=pdfcov
-
Nucleation-related defect-free GaP/Si(100) heteroepitaxy via metal-organicchemical vapor deposition
T. J. Grassman,1,2 J. A. Carlin,3 B. Galiana,3 L.-M. Yang,2 F. Yang,2 M. J. Mills,2
and S. A. Ringel1,31Department of Electrical and Computer Engineering, The Ohio State University, Columbus, Ohio 43210, USA2Department of Materials Science and Engineering, The Ohio State University, Columbus, Ohio 43210, USA3Institute for Materials Research, The Ohio State University, Columbus, Ohio 43210, USA
(Received 19 February 2013; accepted 28 March 2013; published online 8 April 2013)
GaP/Si heterostructures were grown by metal-organic chemical vapor deposition in which the
formation of all heterovalent nucleation-related defects (antiphase domains, stacking faults, and
microtwins) were fully and simultaneously suppressed, as observed via transmission electron
microscopy (TEM). This was achieved through a combination of intentional Si(100) substrate
misorientation, Si homoepitaxy prior to GaP growth, and GaP nucleation by Ga-initiated atomic
layer epitaxy. Unintentional (311) Si surface faceting due to biatomic step-bunching during Si
homoepitaxy was observed by atomic force microscopy and TEM and was found to also yield
defect-free GaP/Si interfaces. VC 2013 AIP Publishing LLC. [http://dx.doi.org/10.1063/1.4801498]
The ability to grow high quality epitaxial GaP layers on
Si has been of interest for decades, largely due to the desire
to use a GaP/Si heterostructure, with its small but non-
negligible lattice constant mismatch (0.37% at 300 K), as a
method to achieve III-V/Si integration. Integration is of
interest as an enabling pathway for the production of high-
efficiency, monolithically integrated III-V/Si photovoltaics,
as well as Si-based photonics, optoelectronics, and high-
speed microelectronics, since a virtual GaP (on Si) substrate
would support subsequent metamorphic epitaxy based on
GaAsP and GaInP alloys. However, the integration of GaP
with Si is not trivial, and incompatibilities of intrinsic mate-
rials properties, such as lattice mismatch, interfacial hetero-
valency (polar/non-polar), thermal expansion mismatch, and
surface/interfacial chemistry, have provided substantial bar-
riers to success and a general inability to sufficiently sup-
press the resulting electrically active crystalline defects,
including anti-phase domains (APDs), stacking faults (SFs)
and microtwins (MTs), and dislocations.1,2
However, recent molecular beam epitaxy (MBE) based
work focusing on the application of surface and interface
chemistry knowledge to the careful control and design of the
Si substrate preparation and GaP nucleation has demon-
strated that these heterovalent nucleation related defects
(APDs, SFs, MTs) can indeed be simultaneously and entirely
suppressed, and nucleation-defect free GaP/Si has indeed
been achieved.3 For this, a combination of Si(100) substrate
vicinality and preparation to yield a pristine, biatomic step
reconstructed surface,4 well-controlled GaP nucleation via
migration enhanced epitaxy,5 and an overall consideration of
epitaxial film mechanics and heterovalent interface proper-
ties in epitaxial structure design was used to produce high-
quality heteroepitaxial GaP/Si layers capable of supporting
further III-V epitaxy for device applications, particularly
photovoltaics.6,7
Similarly, much recent progress has also been made via
metal-organic chemical vapor deposition (MOCVD) growth
methods for such GaP/Si heteroepitaxy, wherein an insightful
surface preparation procedure on nominally exact-oriented
Si(100) substrates, along with an atomic layer epitaxy (ALE)
nucleation approach, has been shown to yield GaP (and sub-
sequent high-quality Si-matched BGaP and GaNP) films free
of SF and MT defects, although with a relatively sparse popu-
lation of internally self-annihilating APD defects still
present.8,9
Nonetheless, III-V alloys that are lattice-matched to Si
are inadequate for use in most technological applications
where an integrated III-V/Si materials system would be de-
sirable. In many cases, such as where metamorphic (lattice-
mismatched) III-V epitaxy is needed to provide access to tar-
get materials, the integrated GaP/Si system is expected to
serve as a growth template or virtual substrate. However, the
existence of interfacial APD disorder is likely to inhibit
dislocation glide, preventing the efficient relaxation of the
misfit strain between the GaP and Si;10 any such interference
with interfacial dislocation glide will likely result in the
generation of excess dislocation density, which can be detri-
mental to ensuing integrated devices.
Here, GaP/Si(100) heterostructures grown by MOCVD
that are simultaneously devoid of SFs, MTs, and APDs are
reported. The approach is generally modeled after the authors’
previously developed MBE-based methodology.3 A similar
combination of intentional Si(100) substrate misorientation to
yield a biatomic surface step reconstruction, Si epitaxy prior to
GaP nucleation to provide a pristine growth surface, and the
use of a Ga-initiated ALE nucleation process was found to pro-
duce GaP films free of all nucleation-related defects (APDs,
SFs, MTs). Additionally, the Si epitaxy conditions used in this
work yielded unintentional step-bunching, leading to the
formation of (311) surface facets, which are also found to be
benign with respect to GaP nucleation.
All GaP and Si growth reported herein was performed in
a 3� 2 in. Aixtron MOCVD system with a close-coupledshowerhead (130 mm inject area diameter). Precursors used
for the associated growths were silane (SiH4), triethylgallium
(C6H15Ga, “TEGa”), and tert-butylphosphine (C4H11P,“TBP”). The typical working pressure for all growth modes
reported here was 150 millibars. In-situ monitoring of the
0003-6951/2013/102(14)/142102/4/$30.00 VC 2013 AIP Publishing LLC102, 142102-1
APPLIED PHYSICS LETTERS 102, 142102 (2013)
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
169.228.173.121 On: Wed, 29 Oct 2014 02:45:21
http://dx.doi.org/10.1063/1.4801498http://dx.doi.org/10.1063/1.4801498http://dx.doi.org/10.1063/1.4801498http://crossmark.crossref.org/dialog/?doi=10.1063/1.4801498&domain=pdf&date_stamp=2013-04-08
-
growths was performed via reflectance/pyrometry, while
post-growth analysis included atomic force microscopy
(AFM), high-resolution triple-axis X-ray diffractometry
(XRD), transmission electron microscopy (TEM), and scan-
ning transmission electron microscopy (STEM).
A variety of (100)-oriented Si substrates were used for
the growths described herein, including wafers with
50.8 mm and 76.2 mm diameters and intentional offcuts of
4� and 6� in the [011] direction, all with equivalent results.Prior to loading into the reactor, the substrates were etched
in a 1:100 HF:DI H2O solution to remove the native oxide
and provide a H-passivated surface. As in the authors’ previ-
ous MBE-based GaP/Si work,3 a thin (�90 nm here) Sihomoepitaxial cap layer was grown in-situ on the Si sub-strates at �760 �C, immediately prior to any III-V nuclea-tion, as a means to bury any persistent and likely
problematic, surface contaminants (particularly carbon); this
temperature is also sufficient to yield the biatomic step sur-
face reconstruction.4
Also following the previous MBE-based work,3 GaP
nucleation on the pristine Si surface was achieved via low-
temperature (450 �C) ALE. Low-temperature reactive TEGaand TBP precursors were used for the ALE nucleation, with
the TEGa pulses carefully calibrated (via AFM) to give exactly
full monolayer coverage. To prevent adverse Si-P reactions,
the ALE GaP nucleation was Ga-initiated.3 Following each
TEGa dose and purge, the surface was exposed to a flow of
TBP, followed by a purge and short pause to allow for reaction
equilibration. A total of 25 GaP ALE cycles was used for the
growths reported here. After ALE-GaP nucleation, the sub-
strate temperature was increased under TBP flow to �580 �C,where a 250 nm thick, but otherwise unoptimized, homoepitax-
ial GaP cap layer was grown via standard MOCVD process to
provide a film with sufficient bulk cohesion to survive subse-
quent high-temperature exposure, thereby producing a true
GaP/Si virtual substrate.
Figure 1(a) presents AFM scans of a homoepitaxial
(�90 nm) Si surface on a 6� offcut Si(100) substrate (with noGaP growth). Evident on the Si surface is a quasi-periodic
“ripple” structure, with vertical and lateral length scales
larger than expected for biatomic stepping on a vicinal sub-
strate, suggesting some sort of step bunching; the surface
appears, otherwise, to be atomically smooth, with a maxi-
mum calculated root-mean-square (RMS) roughness (from a
500� 500 nm2 image to somewhat minimize the effect of theripple) of 0.39 nm, equal to that of the pre-epitaxy Si surface.
Indeed, cross-sectional TEM/STEM (X-TEM/STEM) analy-
sis of such structures capped with subsequent GaP growth
(shown in Figure 2 for a similar 6� offcut substrate) reveal theexistence of tall (�4 nm), step-like structures with wide(�50 nm), flat terraces. The smooth step sidewall faces arefound to lie at an angle of about 25� from the (100) surface,consistent with the formation of (311) facets via biatomic
step bunching, as reported elsewhere.11,12 These (311) facets
are the result of anisotropies in the Si adatom surface and step
edge diffusion kinetics and are found to manifest under cer-
tain combinations of growth and/or annealing conditions
(temperature, growth rate, substrate misorientation).13,14 For
the sake of simplicity, and because results obtained from
4� offcut substrates were effectively identical to those from
6� offcut, with only slight differences in step periodicity andheight, only 6� offcut results are presented.
Fig. 1(b) presents an AFM scan of a GaP nucleation
layer grown by ALE on a 6� offcut Si(100) substrate identi-cal to that shown in Fig. 1(a). The image indicates a smooth
(0.60 nm RMS surface roughness), continuous film with no
Volmer-Weber type island formation15 and which maintains
much of the underlying Si ripple structure. Fig. 1(c) presents
AFM data from a 250 nm thick GaP layer grown via low-
temperature (580 �C) MOCVD on the ALE-grown GaP/Sinucleation layer. These as-grown films present flat and
FIG. 1. Atomic force microscopy images of the surfaces of (a) homoepitax-
ial Si (90 nm thick), (b) ALE-grown heterovalent nucleation layer, and (c)
as-grown heteroepitaxial GaP (250 nm total GaP thickness) surfaces grown
on Si(100) substrates misoriented 6� toward [011].
142102-2 Grassman et al. Appl. Phys. Lett. 102, 142102 (2013)
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
169.228.173.121 On: Wed, 29 Oct 2014 02:45:21
-
relatively smooth surfaces, with typical RMS surface rough-
ness less than 1.0 nm (e.g., 0.87 nm for the 20� 20 lm2 imagein Fig. 1(c)) and no evidence of large-scale APDs.16 A few
dark spots can be seen in the image due to the presence of
small surface voids (�5 nm deep by AFM), most likely due tothe lack of optimization of the growth conditions and layer
thickness of the low-temperature MOCVD GaP cap layer;
such features are neither observed on the post-ALE GaP sur-
face (Fig. 1(b)) nor after subsequent high-temperature (725 �C)GaP MOCVD growth (not shown here)17 and are thus unre-
lated to the GaP/Si interface.
Figure 2 presents X-(S)TEM micrographs, at a range of
magnifications, from an example 6� offcut GaP/Si(100)sample prepared in the methods described above. Multiple
samples were imaged via both TEM and STEM in zone-axis
and two-beam g(200) and g(220) [not shown] conditions to
provide a thorough analysis of the heterovalent interface. As
displayed in Fig. 2(a), and effectively identical to the results
obtained previously via MBE,3 the only defects evident in
any such sample were the expected misfit/threading disloca-
tions due to growing beyond the critical thickness (approxi-
mately 40–90 nm at growth temp),18 indicating that the
heteroepitaxial nucleation and growth process used was
indeed successful at fully preventing the formation of all
nucleation related defects, including SFs, MTs, and even
small-scale APDs. However, not observed in the MBE work
are the periodic, large-scale step features that can clearly be
seen via zone-axis STEM imaging, such as in Fig. 2(b);
high-resolution STEM imaging of these steps, provided in
Fig. 2(c), shows the 25� inclined (311) step-edge faceting.While the (311) faceted step-bunching was an unin-
tended result of the epitaxial Si growth conditions used for
this work, its presence, and any potential impact (or lack
thereof) on the GaP/Si heterovalent interface, is nonetheless
interesting and worth consideration. Because the (311) facet-
ing is formed via the bunching of biatomic steps, the result-
ing large-scale steps would be expected to possess an even
number of atomic layers (i.e., hstep¼ 2n � aSi/4), a necessityfor the preclusion of APD nucleation due to shifted III-V lat-
tice registry. As such, STEM image based measurements
reveal even-numbered step heights on all such faceted steps
measured, with a median value of 8aSi on the 6� samples.
However, slope (i.e., height/width) analysis of the faceted
steps indicates that, in the case of the 6� offcut samples, only�3.8� worth of misorientation is actually accounted for bythe large-scale steps, suggesting that the wide terraces them-
selves still possess �2.2� worth of misorientation, which isstill sufficient to ensure full biatomic stepping to prevent
APD nucleation.19
Of particular interest is the GaP/Si interface at the (311)
facets. Since there are no observable nucleation-related
defects (APDs, SFs, MTs) at these facets, they must be con-
sidered benign with respect to the heterovalent nucleation.
Indeed, the atomic configuration of the Si(311) surface, with
alternating rows of 2- and 3-fold coordinated Si surface
atoms (i.e., with two and one dangling bonds, respectively,
as depicted in Figure 3),20 should yield a site-selective chem-
istry conducive to APD-free heterovalent epitaxy at these
facets, consistent with growth on Si(211) and Si(311) surfa-
ces as previously reported.15,21 Therefore, the serendipitous
combination of both biatomic stepping and (311) faceting on
the vicinal Si(100) surface, which on their own should be
sufficient for the prevention of APDs, appears to work in a
complementary (or at least non-conflicting) manner, ena-
bling the growth of GaP on Si(100) free of all nucleation-
related defects.
The results presented herein demonstrate the growth, by
MOCVD, of GaP on Si(100), free of all heterovalent
nucleation-related defects, including APDs, SFs, and MTs.
The same surface science guided approach previously dem-
onstrated via MBE for the suppression of nucleation-related
defects3—the combination of substrate vicinality, pristine
surface preparation by homoepitaxial Si epitaxy, and
Ga-initiated ALE nucleation—was shown to be equally
FIG. 2. Cross-sectional (scanning) transmission electron micrographs of a
nucleation-related defect-free heteroepitaxial GaP/Si sample grown
by MOCVD on a 6�-[011] misoriented Si(100) substrate: (a) bright-fieldtwo-beam g(200) TEM, (b) bright-field ½0�11� zone-axis STEM, (c) andatomic-resolution ½0�11� zone-axis high angle annular dark field STEM, not-ing the 25� angle of inclination of the (311) step-bunch facet from the (100)surface (155� declination from the associated terrace top).
142102-3 Grassman et al. Appl. Phys. Lett. 102, 142102 (2013)
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
169.228.173.121 On: Wed, 29 Oct 2014 02:45:21
-
effective for MOCVD-based growth. A particularly interest-
ing aspect of the MOCVD-based growths not seen in the
MBE-based work was the formation of (311) faceted step-
bunches which were found to yield defect-free GaP/Si
interfaces, providing an unanticipated simultaneous demon-
stration of an alternative heterovalent integration mecha-
nism. This work represents a critical validation of
methodology transition between the MBE and MOCVD
growth techniques, providing a pathway for the production
of heteroepitaxially integrated III-V/Si structures, with
applicability toward a range of important device technolo-
gies, via standard commercial fabrication approaches.
This work was fully or partially supported by Department
of Energy under the FPACE program (DE-EE0005398), Army
Research Office (DAAD 19-01-0588), and the Ohio Office of
Technology Investments’ Third Frontier program.
1V. Narayanan, S. Mahajan, K. J. Bachmann, V. Woods, and N. Dietz,
Philos. Mag. A 82(4), 685 (2002).2V. Narayanan, S. Mahajan, K. J. Bachmann, V. Woods, and N. Dietz, Acta
Mater. 50(6), 1275 (2002).3T. J. Grassman, M. R. Brenner, S. Rajagopalan, R. Unocic, R. Dehoff, M.
Mills, H. Fraser, and S. A. Ringel, Appl. Phys. Lett. 94(23), 232106(2009).
4D. J. Chadi, Phys. Rev. Lett. 59(15), 1691 (1987).5Y. Takagi, H. Yonezu, K. Samonji, T. Tsuji, and N. Ohshima, J. Cryst.
Growth 187(1), 42 (1998).6T. J. Grassman, M. R. Brenner, M. Gonzalez, A. M. Carlin, R. R. Unocic,
R. R. Dehoff, M. J. Mills, and S. A. Ringel, IEEE Trans. Electron Devices
57(12), 3361 (2010).7T. J. Grassman, A. M. Carlin, K. Swaminathan, C. Ratcliff, J. Grandal,
L.-M. Yang, M. J. Mills, and S. A. Ringel, in 37th IEEE PhotovoltiacSpecialists Conf. (2011), p. 003375.
8S. Liebich, M. Zimprich, A. Beyer, C. Lange, D. J. Franzbach, S.
Chatterjee, N. Hossain, S. J. Sweeney, K. Volz, B. Kunert, and W. Stolz,
Appl. Phys. Lett. 99(7), 071109 (2011).9K. Volz, A. Beyer, W. Witte, J. Ohlmann, I. N�emeth, B. Kunert, and W.Stolz, J. Cryst. Growth 315(1), 37 (2011).
10E. Kvam, J. Electron. Mater. 23(10), 1021 (1994).11A. Oshiyama, Phys. Rev. Lett. 74(1), 130 (1995).12W. Weiss, D. Schmeisser, and W. G€opel, Phys. Rev. Lett. 60(13), 1326
(1988).13J. Mysliveček, C. Schelling, F. Sch€affler, G. Springholz, P. �Smilauer,
J. Krug, and B. Voigtl€ander, Surf. Sci. 520(3), 193 (2002).14A. Pascale, I. Berbezier, A. Ronda, A. Videcoq, and A. Pimpinelli, Appl.
Phys. Lett. 89(10), 104108 (2006).15V. Narayanan, S. Mahajan, N. Sukidi, K. J. Bachmann, V. Woods, and N.
Dietz, Philos. Mag. A 80(3), 555 (2000).16Q. Xu, J. W. P. Hsu, S. M. Ting, E. A. Fitzgerald, R. M. Sieg, and S. A.
Ringel, J. Electron. Mater. 27(9), 1010 (1998).17C. Ratcliff, T. J. Grassman, J. A. Carlin, and S. A. Ringel, Appl. Phys.
Lett. 99, 141905 (2011).18Y. Takagi, Y. Furukawa, A. Wakahara, and H. Kan, J. Appl. Phys. 107(6),
063506 (2010).19O. L. Alerhand, A. N. Berker, J. D. Joannopoulos, D. Vanderbilt, R. J.
Hamers, and J. E. Demuth, Phys. Rev. Lett. 64(20), 2406 (1990).20D. J. Chadi, Phys. Rev. B 29(2), 785 (1984).21H. Kroemer, J. Cryst. Growth 81(1-4), 193 (1987).
FIG. 3. Atomic structures and GaP/Si interfacial bonding configurations at
the (a) Si(311) and (b) Si(211) surface/interface planes. Si dangling bonds,
and thus subsequent GaP/Si interfacial bonds, are indicated by the open
ovals.
142102-4 Grassman et al. Appl. Phys. Lett. 102, 142102 (2013)
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
169.228.173.121 On: Wed, 29 Oct 2014 02:45:21
http://dx.doi.org/10.1080/01418610208243196http://dx.doi.org/10.1016/S1359-6454(01)00408-6http://dx.doi.org/10.1016/S1359-6454(01)00408-6http://dx.doi.org/10.1063/1.3154548http://dx.doi.org/10.1103/PhysRevLett.59.1691http://dx.doi.org/10.1016/S0022-0248(97)00862-2http://dx.doi.org/10.1016/S0022-0248(97)00862-2http://dx.doi.org/10.1109/TED.2010.2082310http://dx.doi.org/10.1063/1.3624927http://dx.doi.org/10.1016/j.jcrysgro.2010.10.036http://dx.doi.org/10.1007/BF02650370http://dx.doi.org/10.1103/PhysRevLett.74.130http://dx.doi.org/10.1103/PhysRevLett.60.1326http://dx.doi.org/10.1016/S0039-6028(02)02273-2http://dx.doi.org/10.1063/1.2345223http://dx.doi.org/10.1063/1.2345223http://dx.doi.org/10.1080/01418610008212068http://dx.doi.org/10.1007/s11664-998-0154-8http://dx.doi.org/10.1063/1.3644956http://dx.doi.org/10.1063/1.3644956http://dx.doi.org/10.1063/1.3310479http://dx.doi.org/10.1103/PhysRevLett.64.2406http://dx.doi.org/10.1103/PhysRevB.29.785http://dx.doi.org/10.1016/0022-0248(87)90391-5