A 1986 summary abstract by Uppal and Kroemer reported molecular-beam-epitaxial (MBE) growth of GaAs and (Al,Ga)As on precisely (211)-oriented Si substrates, nucleated through a ten-period GaAs/(Al,Ga)As superlattice buffer only 0.1 μm thick: the layers were free of antiphase domains and stacking faults, showed photoluminescence comparable to (211)B GaAs homoepitaxy, and hosted 1-μm-gate heterostructure field-effect transistors (HFETs) with room-temperature transconductance up to 150 mS/mm and source resistances of 0.5–1.0 Ω·mm. The n-type background doping (~2×10162\times10^{16} cm−3^{-3}, attributed to Ge contamination in the MBE system) was believed, but not shown, to limit this performance. The work left open: (i) whether reducing the background doping to ~101510^{15} cm−3^{-3} or below yields substantially higher transconductance and lower source resistance, i.e. whether the performance limit is extrinsic (contamination) or intrinsic (threading dislocations); (ii) the optimum superlattice buffer thickness, including whether less than 0.1 μm suffices; (iii) the threading-dislocation density and minority-carrier lifetime of the layers; and (iv) the demonstration of other device types (lasers, solar cells) on this material system. Later literature partially resolved (iii): transmission-electron-microscopy studies of GaAs on Si(211) mapped the dislocation structure and showed strained-layer superlattices markedly reduce the density above the buffer, and short-period superlattices recover near-homoepitaxial luminescence. However, the field abandoned (211) Si in favor of (100)/(001) Si, where FETs, lasers, solar cells, and lifetime and dislocation engineering were all demonstrated; no post-1986 FET, laser, or solar-cell result on GaAs/Si(211) exists. Items (i), (ii), and (iv), and the minority-carrier-lifetime measurement on (211) material, therefore remain open — a precise but dormant open core surviving because the (211) program was dropped, not because attempts failed.

Background

Growing a III-V compound semiconductor on silicon promises to combine III-V optoelectronic performance with Si low cost, large wafer area, and mechanical robustness. This is called polar-on-nonpolar heteroepitaxy: GaAs is a polar crystal with two distinct sublattices (Ga and As), while Si is a nonpolar elemental crystal. Two obstacles dominate. First, the ~4% lattice mismatch between GaAs and Si generates misfit dislocations at the interface and threading dislocations — line defects that run up through the growing layer, scatter carriers, and act as nonradiative-recombination centers. Second, antiphase domains (APDs) arise because the growing polar film must choose which sublattice (Ga or As) sits on which site of the nonpolar Si surface; on the conventional (100) surface, single-atomic-height steps force a sublattice swap, creating antiphase boundaries that degrade optical and electronic quality unless special tricks (e.g. offcut substrates, wafers deliberately cut a few degrees away from the exact crystallographic plane) are used. Molecular beam epitaxy (MBE) is an ultrahigh-vacuum technique in which elemental beams are deposited layer by layer, giving atomic-scale control of layer sequences and doping. A superlattice buffer — many thin alternating periods of GaAs and (Al,Ga)As — inserted before the main layer, is a standard device for trapping and filtering dislocations. Kroemer's group at UC Santa Barbara argued from theory that the high-index (211) surface orientation is particularly favorable for polar-on-nonpolar growth, because it suppresses APD formation regardless of surface-step structure, and demonstrated this experimentally: Uppal and Kroemer, J. Appl. Phys. 58, 2195 (1985), showed that MBE GaAs on Si(211), nucleated with one monolayer of gallium and Ga-shutter cycling (periodically opening and closing the gallium source shutter during nucleation) and initiated through a ten-period (5 nm + 5 nm) GaAs/(Al,Ga)As superlattice buffer of only 0.1 μm total thickness, gives morphology rivaling GaAs homoepitaxy (growth of GaAs on a GaAs substrate), (211)B polarity (the B-terminated variant of the two inequivalent (211) surfaces), and no antiphase domains by etch-pit microscopy (in which defects are revealed as pits by preferential chemical etching) and transmission-electron-microscopy (TEM) studies. Not-intentionally-doped layers were n-type with about 2×10162\times10^{16} electrons per cm3^{3} — attributed by the authors to germanium contamination from another project in the shared MBE system, with Si carryover from the substrate below that level. A companion (211)-orientation homoepitaxy study (Subbanna, Kroemer, and Merz, 1986) established the different growth, doping, and luminescence behavior of the two inequivalent (211)A and (211)B GaAs surfaces. In this context, the 1986 summary abstract (the source of the present question, a one-page account of an AVS-symposium presentation) added device results: heterostructure field-effect transistors (HFETs — field-effect transistors whose channel sits at a GaAs/(Al,Ga)As heterointerface) with 1 μm gate lengths, both normally-on and normally-off, exhibited room-temperature transconductance up to 150 mS/mm and source resistances of 0.5–1.0 Ω·mm. Transconductance (the change of channel current per unit change of gate voltage, normalized per mm of gate width) and source resistance (the parasitic access resistance between the source contact and the channel, in Ω·mm) are the standard figures of merit: high unintentional background doping in and around the channel region degrades gate charge control and access resistance, while threading dislocations degrade mobility and leakage. For comparison, contemporaneous modulation-doped FETs (field-effect transistors whose channel carriers are supplied by dopants placed in the adjacent wider-bandgap (Al,Ga)As barrier layer rather than in the channel) on Si substrates, grown by the mainstream (non-(211)) route (Fischer et al., 1986), reached 170 mS/mm. The UCSB authors believed their numbers were depressed by the Ge-related background doping rather than by the intrinsic (dislocation-limited) quality of the GaAs-on-Si(211) material, but could not demonstrate this with the contaminated system, and deferred dislocation-density and minority-carrier-lifetime characterization (the lifetime of minority carriers — the less abundant carrier type in a doped semiconductor, e.g. holes in n-type GaAs — is a sensitive probe of dislocation-induced recombination and a prerequisite for lasers and solar cells) and the demonstration of other device types to future work. What has happened since: a follow-up by the original growth authors (Ahearn, Uppal, Liu, and Kroemer, by then at Martin Marietta Laboratories in collaboration with UCSB; 1987) mapped the dislocation structure of GaAs on Si(211) and the luminescence response of superlattices in their dislocated GaAs-on-Si material (Liu, Petroff, Kroemer, 1988); the field's center of gravity moved to (100)/(001)-oriented Si, and by 1990 the major review by Fang et al. documented FETs, lasers, and solar cells in GaAs-on-Si on that route, alongside independent milestones such as the first GaAs/Si diode lasers (Windhorn et al., 1984, on Ge-coated Si) and room-temperature laser operation on Si (Sugo et al., 1990). Modern III-V-on-Si monolithic integration is pursued on (001) Si with antiphase-domain control on on-axis substrates (wafers cut exactly along the (001) plane, without deliberate miscut; Gilbert et al., 2023), and the (211) orientation reappears only sporadically, e.g. antiphase-boundary-free GaSb on Si(211) (Hosseini Vajargah et al., 2012), without any devices. No device results on GaAs/Si(211) beyond the 1986 summary abstract are known.

Problem Statement

For GaAs and (Al,Ga)As layers grown by molecular beam epitaxy (MBE, ultrahigh-vacuum layer-by-layer deposition from elemental beams) on precisely (211)-oriented Si substrates, nucleated through a GaAs/(Al,Ga)As superlattice buffer layer (a stack of thin alternating GaAs and (Al,Ga)As periods; 0.1 μm thick, ten periods of 5+5 nm in the source work): (1) when the n-type background doping is reduced from the 2×10162\times10^{16} cm−3^{-3} level attributed to Ge contamination to typical MBE background levels (101510^{15} cm−3^{-3} or below), do 1-μm-gate heterostructure field-effect transistors in this material system achieve substantially higher room-temperature transconductances (above the 150 mS/mm reported in 1986; transconductance is the change of channel current per unit change of gate voltage, per mm of gate width) and lower source resistances (below 0.5–1.0 Ω·mm; the parasitic access resistance between the source contact and the channel), i.e. is the device performance limit set by extrinsic background doping or by the intrinsic (threading-dislocation-limited) material quality? (2) What is the optimum superlattice buffer thickness for dislocation filtering — the thickness minimising the threading-dislocation density above the buffer while sustaining the 1986 transconductance — and in particular, does a buffer thinner than 0.1 μm retain the antiphase-domain-free morphology and the room-temperature transconductance of the 0.1 μm buffer, as the authors suggested without systematic exploration? (3) What are the threading-dislocation density (the areal density of the line defects threading from the GaAs/Si interface up through the layer) — taken from the cited 1987–1988 TEM studies of this material — and the minority-carrier lifetime (the recombination lifetime of the less abundant carrier type in the layer) of these layers, where only the lifetime requires new determination? (4) Are other device types — lasers and solar cells — feasible on this material system, and what material-quality evidence would decide it? The unifying objective of these four open items is to establish the device quality of GaAs/(Al,Ga)As grown on (211)-oriented Si with thin superlattice buffers — covering device performance, buffer optimization, material characterization, and device feasibility. Each item is an independently answerable open sub-question: items (1)–(3) are resolved by a well-argued quantitative result and item (4) by a well-argued feasibility assessment. The task is fully resolved only when all four items are answered; a partial answer resolves only the item or items it addresses.

References

  1. P. N. Uppal and H. Kroemer, 'Summary Abstract: Growth of device-quality GaAs and (Al,Ga)As on (211)-oriented silicon substrates, with thin (0.1 μm) superlattice buffer layers', Journal of Vacuum Science & Technology B 4(2), 641 (1986). DOI: 10.1116/1.583584, https://doi.org/10.1116/1.583584
  2. P. N. Uppal and H. Kroemer, 'Molecular beam epitaxial growth of GaAs on Si(211)', Journal of Applied Physics 58(6), 2195–2203 (1985). DOI: 10.1063/1.335987, https://doi.org/10.1063/1.335987
  3. P. N. Uppal and H. Kroemer, 'Summary Abstract: MBE growth of GaAs and GaP on Si(211)', Journal of Vacuum Science & Technology B 3(2), 603 (1985). DOI: 10.1116/1.583145, https://doi.org/10.1116/1.583145
  4. S. Subbanna, H. Kroemer, and J. L. Merz, 'Molecular-beam-epitaxial growth and selected properties of GaAs layers and GaAs/(Al,Ga)As superlattices with the (211) orientation', Journal of Applied Physics 59(2), 488–494 (1986). DOI: 10.1063/1.336658, https://doi.org/10.1063/1.336658
  5. H. Kroemer, 'Polar-on-nonpolar epitaxy', Journal of Crystal Growth 81(1–4), 193–204 (1987). DOI: 10.1016/0022-0248(87)90391-5, https://doi.org/10.1016/0022-0248(87)90391-5
  6. J. S. Ahearn, P. N. Uppal, T. K. Liu, and H. Kroemer, 'Control of dislocations in GaAs grown on Si(211) by molecular beam epitaxy', Journal of Vacuum Science & Technology B 5(4), 1156–1161 (1987). DOI: 10.1116/1.583703, https://doi.org/10.1116/1.583703
  7. T. Y. Liu, P. M. Petroff, and H. Kroemer, 'Luminescence of GaAs/(Al,Ga)As superlattices grown on Si substrates, containing a high density of threading dislocations: Strong effect of the superlattice period', Journal of Applied Physics 64(12), 6810–6814 (1988). DOI: 10.1063/1.341994, https://doi.org/10.1063/1.341994
  8. L. Fotiadis and R. Kaplan, 'Initial stages of growth of GaAs on silicon (211) substrates by migration-enhanced molecular beam epitaxy', Applied Physics Letters 55(24), 2538–2540 (1989). DOI: 10.1063/1.101974, https://doi.org/10.1063/1.101974
  9. S. F. Fang, K. Adomi, S. Iyer, H. Morkoç, H. Zabel, C. Choi, and N. Otsuka, 'Gallium arsenide and other compound semiconductors on silicon', Journal of Applied Physics 68(7), R31–R58 (1990). DOI: 10.1063/1.346284, https://doi.org/10.1063/1.346284
  10. R. J. Fischer, W. F. Kopp, J. S. Gedymin, and H. Morkoç, 'Properties of MODFET's grown on Si substrates at DC and microwave frequencies', IEEE Transactions on Electron Devices 33(10), 1407–1412 (1986). DOI: 10.1109/T-ED.1986.22687, https://doi.org/10.1109/T-ED.1986.22687
  11. T. H. Windhorn, G. M. Metze, B.-Y. Tsaur, and J. C. C. Fan, 'AlGaAs double-heterostructure diode lasers fabricated on a monolithic GaAs/Si substrate', Applied Physics Letters 45(4), 309–311 (1984). DOI: 10.1063/1.95273, https://doi.org/10.1063/1.95273
  12. M. Sugo, H. Mori, M. Tachikawa, Y. Itoh, and M. Yamamoto, 'Room-temperature operation of an InGaAsP double-heterostructure laser emitting at 1.55 μm on a Si substrate', Applied Physics Letters 57(6), 593–595 (1990). DOI: 10.1063/1.103608, https://doi.org/10.1063/1.103608
  13. S. Hosseini Vajargah, S. Ghanad-Tavakoli, J. S. Preston, G. A. Botton, and R. N. Kleiman, 'Lattice-registered growth of GaSb on Si (211) with molecular beam epitaxy', Journal of Applied Physics 112(9), 093103 (2012). DOI: 10.1063/1.4761970, https://doi.org/10.1063/1.4761970
  14. A. Gilbert, M. Ramonda, L. Cerutti, C. Cornet, G. Patriarche, É. Tournié, and J.-B. Rodriguez, 'Epitaxial Growth of III-Vs on On-Axis Si: Breaking the Symmetry for Antiphase Domains Control and Burying', Advanced Optical Materials 11(15), 2203050 (2023). DOI: 10.1002/adom.202203050, https://doi.org/10.1002/adom.202203050