Ab initio electronic properties of dual phosphorus monolayers in silicon
- Daniel W Drumm^{1, 2}Email author,
- Manolo C Per^{1, 3},
- Akin Budi^{2, 4},
- Lloyd CL Hollenberg^{2} and
- Salvy P Russo^{1}
DOI: 10.1186/1556-276X-9-443
© Drumm et al.; licensee Springer. 2014
Received: 11 June 2014
Accepted: 15 August 2014
Published: 28 August 2014
Abstract
In the midst of the epitaxial circuitry revolution in silicon technology, we look ahead to the next paradigm shift: effective use of the third dimension - in particular, its combination with epitaxial technology. We perform ab initio calculations of atomically thin epitaxial bilayers in silicon, investigating the fundamental electronic properties of monolayer pairs. Quantitative band splittings and the electronic density are presented, along with effects of the layers’ relative alignment and comments on disordered systems, and for the first time, the effective electronic widths of such device components are calculated.
Keywords
Ab initio Density functional theory Bilayers Phosphorus in siliconBackground
We are currently living through a transition in electronic circuitry from the classical to the quantum domain. With Moore’s Law on the way out, thanks to the recent unveiling of ohmic 2 nm epitaxial nanowires[1] and epitaxially gated single-atom quantum transistors[2], the challenge for scientists becomes finding new ways to increase the density and speed of devices as we can no longer rely on being able to shrink their components.
Far-sighted speculation has already been abundant for many years regarding efficient use of the third dimension in device architecture[3–6], breaking the two-dimensional paradigm of current electronics manufacturing techniques. Recent germanium-based works[7, 8] illustrated fundamental physics required for full 3D device implementation and heralded the creation of multiple stacked δ-layers of dopants within a semiconductor. Each of these layers could potentially display atomically abrupt doped regions for in-plane device function and control. Multiple layers of this nature have indeed been created in Ge[9]. The P in Ge atomic layer deposition technique parallels phosphorus in silicon 1/4 monolayer (ML) doping (Si: δ P), created using scanning tunnelling microscope lithography, with a few minor technological improvements (annealling temperatures, amongst others)[8]. In contrast, one major advantage of improvements to silicon technology is that uptake may be far easier, given the ubiquity of silicon architecture in the present everyday life. We may therefore expect to see, in the near future, Si: δ P systems of similar construction.
The time is thus ripe to attend to possible three-dimensional architectures built from phosphorus in silicon. Although Si:P single-donor physics is well understood, and several studies have been completed on single-structure epitaxial Si: δ P circuit components (such as infinite single monolayers[10–17], single thicker layers[18, 19], epitaxial dots[20], and nanowires[1, 21]), a true extension studying interactions between device building blocks in the third dimension is currently missing.
The description of experimental devices is a thorny problem involving the trade-off between describing quantum systems with enough rigour and yet taking sufficient account of the disorder inherent to manufacturing processes. A first approach might therefore be to study the simplest case of interacting device components, namely two P-doped single monolayers (bilayers)[22, 23]. Given the computational limitations of ab initio modelling it is currently not possible to treat the disordered multi-layer system in full. Two approaches suggest themselves. In[23] the approach was to simplify the description of the delta-layer in order to study disorder through a mixed atom pseudopotentials approach. Here, we instead develop a rigorous model of an idealised, perfectly ordered multi-layer system in order to make connections to an understanding gleaned from both the mixed-atom approach and from other idealised models. The two approaches are complementary: alone, neither achieves a complete description, but together, they offer good comparisons from which one may draw the firmest conclusions available regarding experimental devices. The second approach, dwelt upon in this work, also offers descriptions of systems that should become available with improvements to the manufacturing processes mentioned above. As such, this is the focus of our discussion.
Methods
δ layers of P are created on Si (001) terraces before being epitaxially coated with further Si[24–27]. It is easy to envision this coating process being monitored and halted at a desired buffer thickness, before a new δ layer of P is created (and/or patterned). Single δ layer findings[16] suggest that layers interact when less than 80 monolayers (approximately 10.9 nm) of silicon separate them, and that at 80 ML, their properties converge with respect to silicon cladding depth. In that model, periodic replications of the layers were identical by construction, with no possibility of any deviation. Here, we explicitly allow for such differences by including a second layer in the model.
c(2×2) cells including two δ-layers at N ML separation and 80 ML of Si cladding were built (N ∈ {4,8,16,40,60,80}). Doping into a new layer can be accomplished at several locations[19]. For N mod(4) = 0 systems, this can occur in three ways (Figure1b): directly above the original dopant (type A), at either position nearest A in the plane (type B), or at maximal in-plane separation (type C). Note that B is not the nearest neighbour of either A or C in the silicon lattice (see Figure1b).
We note that N mod(4) ∈ {1,2,3} systems exhibit new position types, requiring further modelling. Although such investigation would greatly inform the ongoing discussion of disorder in δ-doped systems, due to computational resource constraints, they are not considered here.
Models were replicated as A_{ N }, B_{ N }, C_{ N }, and undoped (for bulk properties comparison without band-folding complication) structures. Electronic relaxation was undertaken, with opposite donor spins initialised for each layer and various properties calculated. The general method of[16] using SIESTA[28], and energy convergence of 10^{-6} eV, was used with two exceptions: an optimised double- ζ with polarisation (DZP) basis[19] (rather than the default) was employed for all calculations, and the C_{80} model was only converged to 2 × 10^{-4} in density (and 10^{-6} eV in energy) due to intractability. Band structures had at least 25 points between high-symmetry locations.
The choice of a DZP basis over a single- ζ with polarisation (SZP) basis was discussed in[16], where it was found for single δ layers to give valley splittings in far better agreement with those calculated via plane-wave methods. In the recent study by Carter et al.[23], less resource-intensive methods were employed to approximate the disordered-bilayer system, however, here we employ the DZP basis to model the completely ordered system.
Results and discussion
Benchmarking of N = 80 model
Although we used the general method of[16], as we used the optimised basis of[19], we benchmark our A_{80} model with their 80 ML single- δ-layer (δ_{1}) calculation rather than those of[16]. (Lee et al.[18] also used the same general method.) Our supercell being precisely twice theirs, apart from having spin freedom between layers, results should be near identical.
Band structures and splittings
Bilayer models’ band minima energies, Fermi levels, and differences between band minima
Model (type_{ N }) | Band minima (at Γ, meV) | Differences (meV) | E_{F} (meV) | ||||||
---|---|---|---|---|---|---|---|---|---|
Type 1 | Type 2 | ||||||||
1 | 2 | 3 | 4 | 2 -1 | 4 -3 | 3 -1 | 4 -2 | ||
A _{80} | 397 | 397 | 515 | 515 | 0 | 0 | 119 | 119 | 720 |
A _{60} | 397 | 397 | 516 | 516 | 0 | 0 | 119 | 119 | 720 |
A _{40} | 397 | 397 | 516 | 516 | 0 | 0 | 119 | 119 | 721 |
A _{16} | 403 | 421 | 524 | 533 | 18 | 9 | 121 | 112 | 758 |
A _{8} | 377 | 417 | 498 | 605 | 40 | 107 | 122 | 188 | 761 |
A _{4} | 323 | 371 | 615 | 652 | 48 | 37 | 291 | 281 | 771 |
B _{80} | 396 | 396 | 515 | 515 | 0 | 0 | 119 | 119 | 720 |
B _{60} | 397 | 397 | 516 | 516 | 0 | 0 | 119 | 119 | 720 |
B _{40} | 397 | 397 | 516 | 516 | 0 | 0 | 119 | 119 | 721 |
B _{16} | 410 | 410 | 522 | 532 | 0 | 10 | 112 | 122 | 758 |
B _{8} | 374 | 460 | 505 | 604 | 86 | 99 | 131 | 144 | 765 |
B _{4} | 340 | 357 | 602 | 649 | 17 | 47 | 262 | 292 | 772 |
C _{80} | 396 | 397 | 515 | 515 | 0 | 0 | 119 | 119 | 720 |
C _{60} | 397 | 397 | 516 | 516 | 0 | 0 | 119 | 119 | 720 |
C _{40} | 397 | 397 | 516 | 516 | 0 | 0 | 119 | 119 | 721 |
C _{16} | 411 | 414 | 524 | 535 | 3 | 11 | 113 | 121 | 758 |
C _{8} | 375 | 438 | 488 | 591 | 62 | 103 | 112 | 153 | 758 |
C _{4} | 180 | 413 | 608 | 710 | 233^{a} | 102 | 428^{b} | 299 | 774 |
As has been noted before[14, 16], the specific ordering of donors and symmetries inherent in (or broken by) their placement have great effect upon band energies. Whilst for single layers, valley splitting was paramount[15, 16]; here, we introduce the additional possibilities of Coulombic interaction with far-away dopants and quantum interactions with near-neighbour dopants.
Upon closer inspection, holding too closely to single-layer valley splitting proves to be a somewhat naïve way of discussing of the band structures of Figure3. When all models are compared from N = 80 down, it is easily seen that bands come in pairs in the bilayer models, and therefore, at N = 80, the equivalent of single-layer valley splitting is the gap between bands one and three (type 2 in Table1). Due to their large spatial separation, electrons inhabiting bands one and two will overlap only to a negligible extent and, hence, share the same energy here. (This type 1 separation corresponds to interlayer effects - see ‘Consideration of disorder’ section for further discussion.)
As N →4, however, the layers approach and interact; for the C-type model, bands two and three quite clearly cross each other, and it is possible that some mixing of states occurs - which might well be utilised for information transfer between circuit components in a three-dimensional device design; consider two wires crossing at close distance (N < 16) in order to share a state between them.
In fact, the differences columns of Table1 show that the valley splitting is not particularly perturbed until the layers are quite close to each other (A_{4}, B_{8}, and C_{4}), whilst bands which are effectively degenerate at N = 80 are not for N ≤ 16. The layers are interacting, affecting the multi-electronic wavefunction under these close-approach conditions. At N = 4, it is currently impossible to say which contributes more to the band structure.
Within the approximate treatment in[23] it was concluded that the valley splitting in the interacting delta-layers is the same as that for the individual delta-layer. Here we find that in the DZP approach the valley splitting of 119 meV for the interacting delta-layers is about 30% larger than for the individual delta-layer[19]. Of course, Carter et al. themselves acknowledge that their reduced basis functions are not complete enough to represent the ideal system; the SZP results on disordered systems could not have predicted such a difference. We therefore suggest that their estimate of splitting of 63 meV be revised upwards somewhat; the 30% difference seen between ideal single and double layers may be thought of as an upper bound, since the influence of disorder may well counter that of introducing the second layer.
Density of states and conduction
The perpendicular electronic cross-section
Figure6b depicts the worst-case overlap of the gap-states’ wavefunction (modulus). By N = 40, we see (for quantum information purposes) non-negligible overlap (>2%) between the layers. Conversely, N ≥ 80 models show that |Ψ_{gap}| falls off to less than e^{-5}. By N = 8, |Ψ_{gap}| ≥ e^{-2} between the layers. This information will be crucial in assessing future quantum device designs.
In-plane density maps
To focus on bilayer-specific effects, N = 4 and 80 models were rescaled, and their differences are shown in Figure8b. The electronic density reorganises as the layers approach, in a type-dependent manner. The magnitude of the rearrangement is ≤ 20% of the single-layer density.
Consideration of disorder
Listing of ab initio works in this field covering systems with 1/4 ML phosphorus density
Model type | SZP | DZP | |
---|---|---|---|
System | Arrangement | ||
Bulk | bulk-DZP[16] | ||
Ordered | |||
δ | Disordered | δ-DZP-dis[23] | |
Mixed-pseudo | δ-DZP-mix[23] | ||
δ _{n∈{2..5}} | Ordered | δ_{ n }-DZP-ord[19] | |
Ordered | δ δ-SZP-ord[23] | δ δ-DZP-ord^{a} | |
δ δ | Disordered | δ δ-SZP-dis[23] | Intractable |
Mixed-pseudo | δ δ-SZP-mix[23] | ||
δ-wire | Ordered | δ-wire-DZP-ord[21] | |
Staggered | δ-wire-DZP-stag[21] |
Interacting δ layers have recently been studied from the point of view that current experimental systems involve some inherent level of disorder[23]. Whilst it is recognised that a complete DZP model of interacting quasi-disordered bilayers is currently intractable (let alone incorporating disorder on any realistic scale), they offered the rational approach of contrasting a DZP model of a single quasi-disordered δ layer against an SZP model and then extending the SZP model to cover a quasi-disordered bilayer. The reasonable assumption there was that the differences between SZP and DZP models should be similar in both cases, and the valley splittings of the (missing) DZP model of a quasi-disordered bilayer could thus be inferred. They also considered the approach of using a DZP basis and mixed pseudopotential to describe the disorder; this approach is vastly cheaper computationally and purports to inform us about the splittings due to the presence of the second layer. It is supported by SZP mixed and explicit pseudopotential results in which these interlayer splittings are preserved.
The approach taken in this paper, of calculating the properties of an explicitly ordered bilayer system using a DZP basis, complements that previous work. We can equivalently make comparisons between the ordered single-layer systems of[19] (δ-DZP-ord) and ordered double-layer systems as calculated with DZP bases here (δ δ-DZP-ord), and between the δ-DZP-ord systems of[19] and the (DZP) quasi-disordered single-layer system (δ-DZP-dis) presented in[23], in order to draw inferences about the (intractable, missing) δ δ-DZP-dis model, without at any stage compromising the accuracy of the results by using a less-complete basis set. (We shall now proceed to drop the ‘DZP’ from the labels, since it is ubiquitous here.)
One important point in the consideration of disorder from these ideal models is that, at the lowest separation distances, the crystalline order and alignment of the layers is greatly influencing their band structure. In a disordered system, the alignment effects would largely be negated, or averaged out, since one would expect to encounter all possible arrangements. We therefore limit ourselves to discussing averages of splittings.
Model properties and prediction of disordered splittings
Separation | VS (meV) | VS (meV) | ILS (Γ, meV) |
---|---|---|---|
(ML) | (ord-δδ, avg.) | (dis-δδ, est.) | (ord-δδ, avg.) |
80 | 119 | 81 | 0 |
60 | 119 | 81 | 0 |
40 | 119 | 81 | 0 |
16 | 117 | 80 | 9 |
8 | 142 | 97 | 83 |
4 | 309^{a} | 211^{a} | 81^{a} |
We can estimate the interlayer splitting by taking the differences between bands 1 and 2 and bands 3 and 4 (except at low separation). Averaged values for these are also presented in Table3. Unfortunately, beyond the SZP models, we have no further information as to the likely behaviour of the δ δ-dis model at the DZP level in this regard, as there can be no interlayer splitting in the isolate single-layer models to compare against.
It is clear from Table3 that the estimated values for the valley splitting differ from those predicted by the SZP approach (63 meV for all but ‘extremely close separations’). We are in agreement with the finding that narrow separations affect the value greatly. Even allowing for the possibility of overestimation of the valley splitting here (the δ-ord value was 92 meV) only adjusts the estimated δ δ-ord value by 8 meV, not the 20 required to match the values obtained using the SZP approach.
Obviously, the extension to a full DZP model has brought to light behaviours at small separation not evident from the SZP approach, and further work is required to elucidate these as computational resources improve.
Conclusions
We have modelled Si: δ P bilayers, varying their separation and in-plane alignment. Whilst layers behave independently at large separations (above 40 ML), they interact when brought close together: band structures are affected considerably; variation in the energy splitting between the first two occupied bands for N = 4 is considerable, and this variation must be taken into account in any future models of disorder in such closely spaced layers; in-plane charge densities shift by ≤20%. Out-of-plane charge densites overlap to varying extent; wavefunction moduli are more sensitive. For 8 ≤ N ≤ 16, four new conduction channels open, making eight in total. Consequences for device design will depend heavily on the desired purpose; detailed information has been presented for several possible issues to facilitate successful design and operation of future three-dimensional devices, be they classical or quantum in nature. Finally, despite single- ζ with polarisation results indicating that valley splittings are the same in single- and double- δ-layered systems, our results indicate otherwise at double- ζ with polarisation level (previously shown to be adequately complete), with implications for the ongoing discussion of disordered systems of this type.
Declarations
Acknowledgements
The authors acknowledge funding by the ARC Discovery grant DP0986635. This research was undertaken on the NCI National Facility, Canberra, Australia, supported by the Australian Commonwealth Government.
Authors’ Affiliations
References
- Weber B, Mahapatra S, Ryu H, Lee S, Fuhrer A, Reusch TCG, Thompson DL, Lee WCT, Klimeck G, Hollenberg LCL, Simmons MY: Ohm’s law survives to the atomic scale. Science 2012, 335: 64–67. 10.1126/science.1214319View ArticleGoogle Scholar
- Fuechsle M, Miwa JA, Mahapatra S, Ryu H, Lee S, Warschkow O, Hollenberg LCL, Klimeck G, Simmons MY: A single-atom transistor. Nat Nanotechnol 2012, 7: 242–246. 10.1038/nnano.2012.21View ArticleGoogle Scholar
- Eisele I: Delta-type doping profiles in silicon. Appl Surf Sci 1989, 36: 39–51. 10.1016/0169-4332(89)90897-0View ArticleGoogle Scholar
- Kiunke W, Hammerl E, Eisele I: Electrical transport between delta layers in silicon. J Appl Phys 1992, 72(8):3602. 10.1063/1.352300View ArticleGoogle Scholar
- Rodriguez-Vargas I, Gaggero-Sager LM: Subband and transport calculations in double n-type δ -doped quantum wells in Si. J Appl Phys 2006, 99: 033702. 10.1063/1.2168024View ArticleGoogle Scholar
- Cartwright J: Intel enters the third dimension. Nature News 2011. doi:10.1038/news.2011.274 doi:10.1038/news.2011.274Google Scholar
- Scappucci G, Capellini G, Klesse WM, Simmons MY: Dual-temperature encapsulation of phosphorus in germanium δ -layers toward ultra-shallow junctions. J Cryst Growth 2011, 316: 81–84. 10.1016/j.jcrysgro.2010.12.046View ArticleGoogle Scholar
- Scappucci G, Cappellini G, Johnston B, Klesse WM, Miwa JA, Simmons MY: A complete fabrication route for atomic-scale, donor-based devices in single-crystal germanium. Nano Lett 2011, 11: 2272–2279. 10.1021/nl200449vView ArticleGoogle Scholar
- Scappucci G, Capellini G, Klesse WM, Simmons MY: Phosphorus atomic layer doping of germanium by the stacking of multiple δ layers. Nanotechnology 2011, 22: 375203. 10.1088/0957-4484/22/37/375203View ArticleGoogle Scholar
- Qian G, Chang Y-C, Tucker JR: Theoretical study of phosphorous δ -doped silicon for quantum computing. Phys Rev B 2005, 71: 045309.View ArticleGoogle Scholar
- Carter DJ, Warschkow O, Marks NA, McKenzie DR: Electronic structure models of phosphorus δ -doped silicon. Phys Rev B 2009, 79: 033204.View ArticleGoogle Scholar
- Ryu H, Lee S, Klimeck G: A study of temperature-dependent properties of n-type δ -doped si band-structures in equilibrium. In Proceedings of the 13th International Workshop on Computational Electronics. Beijing: Tsinghua University; 2009:1–4. arXiv:1003.4926v1 [cond-mat.mtrl-sci] arXiv:1003.4926v1 [cond-mat.mtrl-sci]Google Scholar
- Ryu H, Lee S, Weber B, Mahapatra S, Simmons MY, Hollenberg LCL, Klimeck G: Quantum transport in ultra-scaled phosphorus-doped silicon nanowires. Silicon Nanoelectronics Workshop 2010 2010, 1–2.View ArticleGoogle Scholar
- Carter DJ, Marks NA, Warschkow O, McKenzie DR: Phosphorus δ -doped silicon: mixed-atom pseudopotentials and dopant disorder effects. Nanotechnology 2011, 22: 065701. 10.1088/0957-4484/22/6/065701View ArticleGoogle Scholar
- Drumm DW, Hollenberg LCL, Simmons MY, Friesen M: Effective mass theory of monolayer δ doping in the high-density limit. Phys Rev B 2012, 85(15):155419. arXiv:1201.3750v1 [cond-mat.mtrl-sci] arXiv:1201.3750v1 [cond-mat.mtrl-sci]View ArticleGoogle Scholar
- Drumm DW, Smith JS, Budi A, Per MC, Russo SP, Hollenberg LCL: Ab initio electronic properties of monolayer phosphorus nanowires in silicon. Phys Rev Lett 2013, 110: 126802.View ArticleGoogle Scholar
- Smith JS, Cole JH, Russo SP: Electronic properties of δ -doped Si:P and Ge:P layers in the high-density limit using a Thomas-Fermi method. Phys Rev B 2014, 89: 035306.View ArticleGoogle Scholar
- Lee S, Ryu H, Campbell H, Hollenberg LCL, Simmons MY, Klimeck G: Electronic structure of realistically extended atomistically resolved disordered Si:P δ -doped layers. Phys Rev B 2011, 84: 205309.View ArticleGoogle Scholar
- Budi A, Drumm DW, Per MC, Tregonning A, Russo SP, Hollenberg LCL: Electronic properties of multiple adjacent δ -doped Si:P layers: the approach to monolayer confinement. Phys Rev B 2012, 86(16):165123.View ArticleGoogle Scholar
- Fuechsle M, Mahapatra S, Zwanenburg FA, Friesen M, Eriksson MA, Simmons MY: Spectroscopy of few-electron single-crystal silicon quantum dots. Nat Nanotechnol 2010, 5: 502–505. 10.1038/nnano.2010.95View ArticleGoogle Scholar
- Drumm DW, Budi A, Per MC, Russo SP, Hollenberg LCL: Ab initio calculation of valley splitting in monolayer δ -doped phosphorus in silicon. Nanoscale Research Letters 2013, 8: arXiv:1201.3751v1 [cond-mat.mtrl-sci].View ArticleGoogle Scholar
- Drumm DW: Physics of low-dimensional nano structures. PhD thesis, The University of Melbourne, 2012 PhD thesis, The University of Melbourne, 2012
- Carter DJ, Warschkow O, Marks NA, Mackenzie DR: Electronic structure of two interacting phosphorus δ -doped layers in silicon. Phys Rev B 2013, 87: 045204.View ArticleGoogle Scholar
- Tucker JR, Shen T-C: Prospects for atomically ordered device structures based on STM lithography. Solid State Electron 1998, 42(7–8):1061–1067.View ArticleGoogle Scholar
- O’Brien JL, Schofield SR, Simmons MY, Clark RG, Dzurak AS, Curson NJ, Kane BE, McAlpine NS, Hawley ME, Brown GW: Towards the fabrication of phosphorus qubits for a silicon quantum computer. Phys Rev B 2001, 64: 161401(R).View ArticleGoogle Scholar
- Shen T-C, Ji J-Y, Zudov MA, Du R-R, Kline JS, Tucker JR: Ultradense phosphorous delta layers grown into silicon from PH_{3} molecular precursors. Appl Phys Lett 2002, 80(9):1580–1582. 10.1063/1.1456949View ArticleGoogle Scholar
- Fuechsle M, Ruess FJ, Reusch TCG, Mitic M, Simmons MY: Surface gate and contact alignment for buried, atomically precise scanning tunneling microscopy-ppatterned devices. J Vac Sci Tech B 2007, 25(6):2562–2567. 10.1116/1.2781512View ArticleGoogle Scholar
- Artacho E, Anglada E, Diéguez O, Gale JD, Garciá A, Junquera J, Martin P, Ordejón RM, Pruneda JM, Sánchez-Portal D, Soler JM: The SIESTA method; developments and applicability. J Phys Condens Matter 2008, 20: 064208. 10.1088/0953-8984/20/6/064208View ArticleGoogle Scholar
Copyright
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.