Electric field reduced charging energies and two-electron bound excited states of single donors in silicon

Abstract

We present atomistic simulations of the D0 to D− charging energies of a gated donor in silicon as a function of applied fields and donor depths and find good agreement with experimental measurements. A self-consistent field large-scale tight-binding method is used to compute the D− binding energies with a domain of over 1.4 million atoms, taking into account the full band structure of the host, applied fields, and interfaces. An applied field pulls the loosely bound D− electron toward the interface and reduces the charging energy significantly below the bulk values. This enables formation of bound excited D− states in these gated donors, in contrast to bulk donors. A detailed quantitative comparison of the charging energies with transport spectroscopy measurements with multiple samples of arsenic donors in ultrascaled metal-oxide-semiconductor transistors validates the model results and provides physical insights. We also report measured D− data showing the presence of bound D− excited states under applied fields.

Full text

Page 1

Electric field reduced charging energies and two-electron bound excited states of
single donors in silicon
R. Rahman*,1G. P. Lansbergen,2J. Verduijn,2,3G. C. Tettamanzi,2,3S. H. Park,4
N. Collaert,5S. Biesemans,5G. Klimeck,4L. C. L. Hollenberg,6and S. Rogge2,3
1Advanced Device Technologies, Sandia National Laboratories, Albuquerque, NM 87185, USA
2Kavli Institute of Nanoscience, Delft University of Technology, Lorentzweg 1, 2628 CJ Delft, The Netherlands
3Centre for Quantum Computation and Communication Technology, School of Physics,
University of New South Wales, Sydney, New South Wales 2052, Australia
4Network for Computational Nanotechnology, Purdue University, West Lafayette, IN 47907, USA
5Inter-University Microelectronics Center (IMEC), Kapeldreef 75, 3001 Leuven, Belgium
6Centre for Quantum Computation and Communication Technology,
School of Physics, University of Melbourne, VIC 3010, Australia
(Dated: July 15, 2011)
We present atomistic simulations of the D0to D−charging energies of a gated donor in silicon as
a function of applied fields and donor depths and find good agreement with experimental measure-
ments. A self-consistent field large-scale tight-binding method is used to compute the D−binding
energies with a domain of over 1.4 million atoms, taking into account the full bandstructure of the
host, applied fields, and interfaces. An applied field pulls the loosely bound D−electron towards
the interface and reduces the charging energy significantly below the bulk values. This enables
formation of bound excited D−states in these gated donors, in contrast to bulk donors. A detailed
quantitative comparison of the charging energies with transport spectroscopy measurements with
multiple samples of arsenic donors in ultra-scaled FinFETs validates the model results and provides
physical insights. We also report measured D−data showing for the first time the presence of bound
D−excited states under applied fields.
PACS numbers: 71.55.Cn, 03.67.Lx, 85.35.Gv, 71.70.Ej
I. INTRODUCTION
A single gated donor atom in silicon has attracted
considerable attention over the last decade as a promis-
ing quantum information processing unit in solid-state
[1]. Among other factors, such donor qubits benefit from
exceptionally long spin coherence times [2], compatibil-
ity with the complementary metal-oxide-semiconductor
(CMOS) technology, and accessibility to controllable
atomic physics in the solid-state. Some promising pro-
posals for donor qubits are based on encoding quantum
information in the nuclear [1] or electronic spin [3–5], or
in the molecular charge states of two phosphorus donors
[6]. Besides applications in quantum computing, dis-
crete dopants are becoming increasingly important in
nanoscale electronics, as they strongly affect the sub-
threshold current-voltage characteristics of ultra-scaled
MOSFETs [7, 8]. Other applications of dopants in na-
noelectronics are also emerging, such as the proposal
for classical logic devices based on resonant tunneling
through the donor states [9, 10].
It is well-known that Group V donors in silicon can
bind either 1 electron and form a neutral D0state, or
bind 2 electrons and form a negatively charged D−state.
In bulk donors, the D−ground state is a singlet weakly
bound at 1.7-2.0 meV below the conduction band (CB)
edge, corresponding to a charging energy (CE) of about
43 meV for a phosphorus (P) impurity and 52 meV for
an arsenic (As) impurity [11]. Recently, resonant tunnel-
ing through the D0ground and excited states has been
observed in different experiments [7, 12, 13] with devices
fabricated from both top-down [14] and bottom-up [15]
approaches. Most of the measurements of the D- state
in gated devices show a significantly reduced charging
energy in the range of 25-35 meVs, a conundrum still
largely unresolved although some progress has been made
using effective mass treatments [16–19]. In this work, we
compute the D−charging energy of a donor under ap-
plied fields using self-consistent field tight-binding (TB)
including over 1 million atoms in the spatial simulation
domain. We compare the computed CEs with measured
data on single dopants in seven FinFET devices by sim-
ulating the atomic environment of the devices. Despite
being a mean-field technique, this method enables an ac-
curate quantitative description of the charging energies
as it captures the details of confinement geometries and
valley-orbit interaction from a full bandstructure tech-
nique. The method solves the Poisson equation itera-
tively with the atomistic TB Hamiltonian for charge self-
consistency, and represents an advancement over general
TB as computational hurdles had to be overcome to solve
million atom systems in reasonable time.
Although no excited D−states have ever been ob-
served in bulk samples [11], our measurement shows for
the first time the presence of bound excited states in-
cluding a triplet state, which can be ascertained by the
phenomena of lifetime enhanced transport [20, 21]. This
is a consequence of the applied electric field that lowers
arXiv:1107.2701v1 [cond-mat.mes-hall] 13 Jul 2011

Page 2

2
the charging energy and enables an excited manifold to
form below the CB. A two-electron bound triplet state
may enable easier ways to perform spin readout through
spin blockade between a donor electron and an electron
from a local reservoir (2DEG) state. Previous proposals
of donor based spin readout have relied on using either
Group VI donors [22, 23] or a pair of Group V donors
[17], systems in which two-electron triplet states exist.
Our result shows such spin readout may be feasible with
the more conventional P/As donor species under strong
applied fields. With the recent demonstration of single
shot spin readout of a single electron in silicon [24], this
has become even more significant in the context of quan-
tum computing.
This paper is organized as follows. In Section II, we de-
scribe the theoretical method for computing the charging
energy in detail. We also elaborate on the experimental
technique employed to measure the D−charging ener-
gies and excited states in FinFETs. In Section III, we
discuss the computed results and how they compare to
the measurements. Section IV concludes this work.
II. METHOD
A. Theoretical Model
In the atomistic TB method employed here, the Hamil-
tonian is represented in a basis of 10 localized atomic or-
bitals per atom with the sp3d5s* nearest neighbor model
[25, 26]. The Hamiltonian parameters of the host ma-
terial have been optimized using genetic algorithm with
analytically derived constraints to fit critical features of
the host band structure [27, 28]. Once the TB model
parameters of the host are obtained, they are generally
transferable to a whole range of device simulations, as
benchmarked in a number of earlier works [29–31]. The
full TB Hamiltonian of more than a million silicon atoms
is diagonalized using a parallel Lanczos algorithm to ob-
tain any number of lowest lying eigenvalues and wave-
functions.
The TB Hamiltonian of the host and the donor in an
applied electric field?F is given by,
H = H0− VD+ e?F · (? r − ? r0) + VSCF
where H0 is the TB Hamiltonian of the host material
silicon, and VDis the central-cell corrected Coulomb po-
tential energy of the donor. The 3rd term in eq (1) is that
of a constant electric field, whereas the last term VSCFis
the potential energy due to electron-electron repulsion.
Central-cell correction represents the deviation of the
donor potential from its 1/r form near the donor core
[32], and is responsible for the different D0binding ener-
gies of different Group V donor species [33]. The central-
cell corrected singular-like potential near the donor core
(1)
produces coupling between the six conduction band (CB)
minima Bloch states, and results in a splitting of the or-
bital ground state of the donor from the excited states.
This is commonly known in literature as the ‘valley-orbit’
splitting [34]. The potential energy of the donor to first
order in our model is given by,
VD(? r ?= ? r0) =
e2
4π?Si|? r − ? r0|,VD(? r0) = U0
(2)
where ? r0 is the location of the donor, e the electronic
charge, ?Sithe dielectric constant of silicon, and U0is a
cut-off potential representative of the central-cell correc-
tion to the first order. A detailed model of impurities in
TB can be found in Ref [35].
VSCFis the potential due to a bulk charge density n(? r),
and has to be computed self-consistenly from a reduced
Poisson equation as shown below. In the first iteration,
VSCFis set to zero to obtain the D0states of the donor.
The solution of Hψ = Eψ yields a set of energies E =
{Ei} and wavefunctions ψ = {ψi}, where E1 and ψ1
are the ground state energy and wavefunction of a donor
respectively. For a bulk P donor at zero field, the above
method yields D0binding energy E1of 45.6 meV below
the CB minimum. For an As donor, it is 54 meV [33].
To obtain the D−orbital energy, we assume the ground
state is filled by exactly one electron, described by an
electron density n(? r) = |ψ1(? r)|2. VSCFis then given by,
VSCF(? r) =
?
e2n(?r?)
4π?Si|? r −?r?|d?r?
(3)
With the new VSCFeq (1) is solved for a new set of eigen-
states and vectors. The process is repeated until E1and
n(? r) both converge. Since eq (1) with a converged VSCF
represents an impurity in silicon with a bound electron,
the converged orbital energy E1 represents the binding
energy of the second electron below the CB minimum.
This method is also described in detail in Ref [36].
This technique makes use of a density functional ap-
proximation, as the electron-electron interaction poten-
tial is expressed as a function of the electron density.
Since the D−ground state resides in a closed-shell sin-
glet, there is no exchange energy between the two elec-
trons.The repulsive Coulomb energy evaluated self-
consistently thus offers a good description of the bind-
ing energy. However, the method is still an approximate
one, as higher order exchange-correlation corrections are
ignored.
An exact way to solve this problem is a full config-
uration interaction (CI) method, in which the exact 2e
Hamiltonian is diagonalized in a basis of Slater Determi-
nants describing anti-symmetric 2e configurations built
from a complete set of single particle states [37]. How-
ever, such a method requires a large number D0orbitals

Page 3

3
to represent the spatial extent, symmetry, spin and valley
configuration of the D−states making it both impracti-
cal and computationally intractable. In other words, the
D−state is not well-represented with a basis compris-
ing of the low lying D0states, suggesting the need for
a self-consistent method to iteratively improve the ba-
sis states. Furthermore, the six-fold valley degeneracy in
silicon makes the D0orbital basis set six times as large.
Moreover, evaluating a large number of coulomb and ex-
change integrals with atomistic wavefunctions spanning
over a million atoms becomes a computational challenge.
On the other hand, atomistic wavefunctions are required
for a highly accurate description of the single particle ba-
sis functions to capture details of bandstrucuture, geom-
etry and interfaces, all of which are critical for a proper
description of impurities in silicon [38].
The mean-field method presented here thus serves as
an intermediate theory that combines a highly accurate
single particle basis states with a mean-field description
of the Coulomb interaction, while ignoring exchange-
correlation corrections associated with the 2e system. Al-
though this method cannot capture the exact 2e wave-
functions, it provides a very good description of the D−
binding energies, and the n(? r) associated with the added
electron, as we describe in the Section III.
As a benchmark of the method, we computed the D−
binding energy of a bulk P donor in silicon. Photolumi-
nescence experiments have measured this energy to be
2.0 meV below the CB minimum [11], which corresponds
to a CE of 43.6 meV. In comparison, the SCF TB method
described here yields a binding energy of 3.4 meV for a
bulk P donor, corresponding to a CE of 42.2 meV. The
difference of 1.4 meV from the experimental value can
be regarded as a limitation of the theory due to neglect-
ing exchange-correlation corrections and due to the fact
that the TB method only considers point charges on an
atomistic zincblende lattice.
Typical simulation times of the 1e donor states from
the TB Hamilton for 1.4 million atoms is about 2-3 hours
on 40 processors. The computation of VSCFrequires 0.5
hours on 40 processors. Typical D−energies were found
to converge in between 10 to 20 iterations of eq (1) and
eq (3) [39, 40].
Most other works on the D−donor state have been
based on describing the weakly bound 2e wavefunction
with variational envelope wavefunction in a single val-
ley picture [16–19]. One exception is Ref [41], in which
a Quantum Monte Carlo approach was used to compute
the bulk D−. Two recent works have treated the effect of
screening on the D−energies in the presence of metallic
gates and hetero-interfaces using an effective mass ap-
proach [18, 19], and have provided qualitative trends.
However, the field and depth dependence of the donors
that are directly relevant in measurements need to be in-
vestigated in detail from a more quantitative approach.
B. Experimental Technique
FIG. 1:
b) Coulomb blocked transport in a FinFET transistor. Two
separate charge islands can exist inside the transistor; a quan-
tum dot confined by the triangular potential at the gate in-
terface and residual barriers in the access regions between
source/drain and channel and a donor/well system confined
by the donors Coulomb potential and a well at the interface.
The localized states formed in these charge islands are de-
noted by QDnand D0/D−respectively. Inset: The potential
landscape in a cross-section from the gate to the channel.
The gate electric field induces a triangular potential at the
interface. An (accidental) donor atom in the channel forms a
Coulombic potential on top of the gate potential (green) that
can bind two electrons.
The measurements presented in this work were per-
formed on single As donors in silicon FinFETs, consist-
ing of a silicon nanowire with a gate covering three faces
of the body (Fig. 1(a)). A thin nitrided oxide layer sep-
arates the gate from the channel. In some devices, a
single As donor was found in the channel within 5 nm of
the oxide interface. We measured the low-temperature
sub-threshold current voltage characteristics of about a
100 devices, and selected 7 devices where the transport
characteristics are dominated by a single donor atom in
the channel. The other devices either showed no sub-
threshold signal or a complicated pattern associated with
Coulomb interaction between several donors in the chan-
nel. The experimental technique is detailed in our earlier
works [12, 34, 42].
Fig.1(b) shows the source-drain current of device
a) Colored SEM image of a silicon FinFET device.

Page 4

4
GLG14 [12] at low bias as a function of gate voltage.
At any gate voltage where a localized state in the chan-
nel is within the bias window defined by source/drain, it
gives a contribution to the transport and a peak in the
current occurs. As such, we can perform spectroscopy,
with the gate voltage being a measure for the energy of
the level (E = αVGwhere α is the electrostatic coupling
between the gate and the level). Based on the aforemen-
tioned criteria, we identify the first two resonances as the
D0and D−charge states of a single As donor. The res-
onances indicated by QDnare due to a localized state
which is confined by the gate electric field and two bar-
riers in the access regions between source/drain and the
channel [43–45].
The inset of Fig. 1(b) shows a 1D schematic of the
potential of a donor close to the interface at a high E-
field. The field transforms the confining potential to a
(hybridized) mix between the donors Coulomb potential
and a triangular well at the interface. This system is
thus essentially a gated donor where the donor-bound
electrons are partly pulled toward the Si/SiO2interface.
Due to the proximity of the donor to the interface (< 5
nm), it is possible to apply a high E-field without fully
ionizing the donor electron to the interface well. As the
E-field is increased, the donor electron hybridizes with
interface states, and makes a smooth transition to the
interface well, in contrast to a bulk donor, where the ion-
ization process is rather abrupt and occurs at much lower
fields [38]. In our previous work, we combined data from
excited state spectroscopy of the D0state with a large-
scale tight-binding analysis, to verify the species of the
donor [46], their locations and the E-fields they experi-
enced [12]. In this work, we focus on the D−charging
energies for those six devices, as well as a new device
sample, to show the trends in charging energies, as well
as to show bound excited 2e states in the spectrum.
III. RESULTS AND DISCUSSIONS
A donor located close to the oxide-silicon interface
[23, 38, 47, 48] has been proposed as an important vari-
ant of the Kane qubit architecture [1] based on deeply
buried donors. In contrast to bulk donors, it is possible
to adiabatically pull the donor electron to the interface,
and hence to perform precise quantum control and wave-
function engineering by means of an applied gate bias.
Hybrid architectures have also been proposed in which
electrons from surface bound quantum dots can be se-
lectively shuttled to nearby donor to preserve their spin
coherence for longer time-scales [49].
In Fig. 2, we show the D0and D−regions of the sta-
bility diagram of a newly measured device with a gated
As donor. Although the CE of a bulk As donor is 52
meV, the measured CE in this sample is 30 meV. We
will show that this is a consequence of the applied field
FIG. 2: Measured Coulomb diamonds of a device (A18G17)
sample showing the D0and D−regions. Traces of conduc-
tance near the D−diamond show existence of bound excited
states in the two electron spectrum.
that pulls the electron cloud away from each other. Con-
ductance traces through excited states can be observed
in both the D0and D−regions. While the existence
of bound excited states of the D0are well established
[12], bound D−excited states are a novel phenomena re-
ported here for the first time. This a consequence of the
reduced CE that cause excited manifolds to form below
the conduction band.
Fig. 3(a) shows the 1D potential schematic of the sys-
tem for various field strengths. As shown in Fig. 1(b), the
potential in this system is a superposition of the Coulomb
potential of the donor and a linear potential due to the
applied E-field.At low field strengths (Fig.
the Coulomb potential dominates, and both electrons are
bound in the Coulomb well of the donor. This is a bulk-
like system with high CE with both electrons confined
to a very small region of space. Hence, the electron den-
sity around the donor core is high, and the electrostatic
repulsion between the electrons stronger, resulting in a
higher CE. As the E-field is increased, a triangular well
forms at the interface, and the electrons are gradually
pulled towards it.At one point, one electron resides
in a strongly hybridized state relative to the other, a
state which is more delocalized in space, as shown in
Fig. 3(a)(ii). The reduced electron densities result in a
reduced Coulomb interaction, and the CE decreases. At
larger E-fields, both electrons are pulled towards the in-
terface (Fig. 3(a)(iii)), and spreads out more laterally in
space.
Fig. 3(b) shows the binding energies of D0and the
D−states as a function of the E-field for an As donor
at 3.8 nm from the oxide interface. The energies are ex-
pressed relative to the Stark shifted conduction band at
the donor site. The difference between the two binding
energies is the charging energy, CE. At low fields, corre-
sponding to regime (i) of Fig. 3(a), both electrons are
at the donor, and the D−state is loosely bound below
the CB, reminiscent of a bulk-like system. However, the
electronic wavefunctions are influenced by the extra con-
finement due to the nearby interface, and the states are
3(a)(i)),

Page 5

5
FIG. 3: (a) 1D schematic of the total potential due to donor
and the applied E-field. (i) At low E-field, both the elec-
trons are donor bound. (ii) At increased E-field, one electron
is pulled towards the interface, while another remains donor
bound. (iii) At higher E-fields, both electrons reside at the
interface. (b) Binding energies of the D0and D−states as
a function of electric field for an As donor at 3.8 nm depth.
The difference between the two energies is the charging energy
(CE). (c) CE as a function of electric field for three different
donor depths.
pushed up in energy. As a result, the D−binding energy
in this case is even less than the bulk value of 2 meV
below the CB.
As the E-field is increased, the system gradually moves
to the regime (ii) of Fig. 3(a). The electrons begin to
hybridize with the interface well states, and both the
D0and D−binding energies are pushed downwards with
applied field, as shown in Fig. 3(b). This orbital Stark
effect for the D0electron has been described in detail in
Ref [38]. In the 2e system, the added electron hybridizes
with the interface states sooner than the other electron.
As the electron density spreads out more in space, the
CE decreases as shown by the decreasing gap between
the two energies in Fig 3(b).
At higher fields, the interface well is occupied by both
electrons, as described by regime (iii) of Fig. 3(a). The
electronic densities are more laterally extended than be-
fore, resulting in a further decrease in CE, which is rep-
resented by a constant energy gap between D0and D−
states, as shown in the region (iii) of Fig. 3(b). However,
the electrons are still laterally bounded by the Coulomb
potential of the donor, which prevents them from forming
a two-dimensional electron gas (2DEG). The lateral con-
finement potential is stronger for shallower donors. This
makes the charge density larger, and the electronic repul-
sion stronger. Therefore, the CE is expected to increase
with decreasing donor depths, as the electrons are pulled
into the interface well. In regime (iii), any increase in the
vertical field does not influence the electronic wavefunc-
tions significantly, and the CE becomes insensitive to the
applied field.
Fig. 3(c) shows the charging energy as a function
of field for three different donor depths. For a specific
depth, the CE makes a smooth transition from a bulk-
like value of above 50 meV to about 20-30 meV repre-
sentative of an interfacial confinement regime. Once the
electrons are interface bound, the CE becomes field in-
dependent.However, due to the lateral Coulomb po-
tential of the donor, the CE values still increase as the
donor depth decreases.In the intermediate field regime,
any CE value from about 50 to 30 meV is possible due to
hybridization of the donor states with the interface well
states [38]. For smaller donor depths, the transition from
the bulk-like CE value to the interface-like CE value is
smoother because the tunnel coupling between the two
wells is stronger. For a bulk donor, an abrupt step-like
transition is expected, as the electrons are ionized with-
out any further confinement.
Although we have ignored the exact effect of screening
in this work, we can qualitatively understand its effect
on the D−binding energies and CEs from Fig.
and 3(c). Screening essentially modifies the value of the
applied E-field due to dielectric mismatch at the silicon-
oxide and the gate-oxide interfaces, which results in im-
age charges. If the net screening is dominated by the
gate electrode, then the induced image charges would be
of opposite polarity to the negatively charged D−system.
As a result, the effective field will be reduced. Since we
have used the E-field as a free parameter to investigate
the D−energy over a wide range of field values, the effect
of screening can be deduced to first order. For example,
if a sample is at 25 MV/m field in Fig. 3(b) and (c),
a metallic screening will essentially move the sample to
the left of this point in the curve. If the net screening
is dominated by the insulating oxide layer, then the in-
duced charges are of the same polarity as the D−state
[38]. This results in an increased field, which shifts the
sample to the right in Fig. 3(b) and 3(c).
However, to obtain an exact quantitative description
of screening, the full Poisson equation has to be solved
numerically in 3D including the oxide layer and the metal
gate, and also self-consistently with the TB Hamiltonian
over the whole domain. However, this poses a problem
because the atomistic TB Hamiltonian has to be solved
over silicon only, as the oxide layer is amorphous in na-
ture and lacks a regular structure. One way to side-step
this is to assume a virtual crystal (VC) model of the ox-
3(b)

Page 6

6
ide, and to extend the TB Hamiltonian to include this
region, however VC models of SiO2 are still not well-
established in literature [50, 51]. Although a truncated
TB Hamiltonian over the silicon region can be iterated
with the Poisson equation over the whole domain, there
could be issues relating to charge inconsistencies which
would affect the convergence.
glected an exact quantitative description of gate screen-
ing, and have used the field as the free parameter to
investigate its effects approximately. Although some of
these limitations can be overcome in principle, this is out
of the scope of the present work.
We have therefore ne-
FIG. 4: Colormap of the modeled charging energy as a func-
tion of gate electric field (F) and donor depth (d). The black
traces indicate the region where the modeled charging energy
is between 29 and 35 meV, as we also find experimentally.
The black data points indicate the positions of the samples
in the F-d plane, as determined in a previous publication [12]
from their D0level spectrum.
In Fig. 4, we show a 2D colormap of the CEs over a
range of donor depths (horizontal axis) and E-fields (ver-
tical axis). At high E-fields, the CEs are between 20-25
meV, as indicated by the blue region. This corresponds
to interfacial confinement. At low E-fields, the CEs are
about 50 meVs, marked by the red region, corresponding
to donor bound D−states. The green and the yellow
regions show the intermediate field hybridized regimes.
If the donor depth from the interface is large, it takes a
smaller field to detune the two wells. Hence, ionization
to the interface well takes effect at a lower field value.
This is why the blue region of interface-like CEs grow in
area from left to right of the plot.
We now compare the measured CEs of six device sam-
ples with the computed values. The fields (F) and the
depths (d) of these samples were determined from the D0
excited state transport spectroscopy in our earlier work
[12]. We can map these samples on Fig. 4 based on
these extracted F and d. These data points are marked
as black squares. This shows that the D−energies of
these samples were close to the border between the hy-
bridized regime (Fig. 3(a)(ii)) and the interface-bound
regime (Fig. 3(a)(iii)), which explains their CEs in the
range 29-35 meV. Ideally, all samples should lie between
the two black lines plotted in Fig. 4. We attribute the
small discrepancy to the fact that we have neglected the
exact nature of screening for the D−state, as discussed
above.
IV. CONCLUSION
We have shown through transport spectroscopy mea-
surements on gated As donors in silicon that the charg-
ing energies of these donors can be significantly reduced
below the bulk values in the presence of an applied E-
field. As a consequence, bound excited states are ob-
served for the first time in the D−spectrum of the donors.
This opens up the prospect of performing spin-readout
through spin-to-charge conversion between interface and
donor bound states using the same Group V donors as
the qubits. We present a large-scale self-consistent tight-
binding method to compute the charging energies in these
nanostructures, taking into account the atomistic details
and potentials. The simulations show that the charging
energies of the donors are reduced with applied fields,
as the electrons hybridize with interface states and de-
localize. At low fields, high CEs of above 50 meVs are
expected, while at high ionizing fields, the CEs can de-
crease to about 20-30 meVs. As single donor devices are
being fabricated from both top-down and bottom-up ap-
proaches, our technique can be used to model the D−
binding energies in a variety of realistic devices.
We acknowledge financial support from the EC FP7
FET-proactive NanoICT projects MOLOC (215750) and
AFSiD (214989), the Dutch Fundamenteel Onderzoek
der Materie FOM, the Australian Research Council,
the Australian Government, the U.S. National Security
Agency (NSA) and the Army Research Office (ARO)
under Contract No. W911NF-04-1-0290. This research
was conducted by the Australian Research Council Cen-
tre of Excellence for Quantum Computation and Com-
munication Technology (Project number CE110001027).
NEMO 3D was initially developed at JPL, Caltech un-
der a contract with the NASA. Sandia is a multiprogram
laboratory operated by Sandia Corporation, a Lockheed
Martin Company, for the United States Department of
Energy’s National Nuclear Security Administration un-
der Contract No. DE-AC04-94AL85000. NSF funded
NCN/nanoHUB.org computational resources were used.
Electronic address: rrahman@sandia.gov
[1] B. E. Kane, Nature, 393, 133 (1998).
[2] A. M. Tyryshkin, S. A. Lyon, A. V. Astashkin, and A.
M. Raitsimring, Phys. Rev. B 68, 193207 (2003).

Page 7

7
[3] Rutger Vrijen, Eli Yablonovitch, Kang Wang, Hong Wen
Jiang, Alex Balandin, Vwani Roychowdhury, Tal Mor,
and David DiVincenzo, Phys. Rev. A 62, 012306 (2000).
[4] C. D. Hill, L. C. L. Hollenberg, A. G. Fowler, C. J.
Wellard, A. D. Greentree, and H.-S. Goan, Phys. Rev.
B 72, 045350 (2005).
[5] L. C. L. Hollenberg, A. D. Greentree, A. G. Fowler, and
C. J. Wellard, Phys. Rev. B 74, 045311 (2006).
[6] L. C. L. Hollenberg, A. S. Dzurak, C. Wellard, A. R.
Hamilton, D. J. Reilly, G. J. Milburn, and R. G. Clark,
Phys. Rev. B 69, 113301 (2004).
[7] M. Pierre, R. Wacquez, X. Jehl, M. Sanquer, M. Vinet,
and O. Cueto, Nature Nanotechnology 5, 133 (2010).
[8] Scott Roy and Asen Asenov, Science Vol. 309, 388-390
(2005).
[9] M. Klein, J. A. Mol, J. Verduijn, G. P. Lansbergen, S.
Rogge, R. D. Levine, and F. Remacle, Appl. Phys. Lett.
96, 043107 (2010).
[10] Yonghong Yan, J. A. Mol, J. Verduijn, S. Rogge, R. D.
Levine, and F. Remacle, J. Phys. Chem. C, 114, 20380
(2010).
[11] M. Taniguchi and S. Narita, Solid State Communica-
tions, Vol. 20, Issue 2, Pages 131-133 (1976).
[12] G. P. Lansbergen, R. Rahman, C. J. Wellard, I. Woo, J.
Caro, N. Collaert, S. Biesemans, G. Klimeck, L. C. L.
Hollenberg, and S. Rogge, Nature Physics 4, 656 (2008).
[13] Kuan Yen Tan, Kok Wai Chan, Mikko Mottonen, Andrea
Morello, Changyi Yang, Jessica van Donkelaar, Andrew
Alves, Juha-Matti Pirkkalainen, David N. Jamieson,
Robert G. Clark, and Andrew S. Dzurak, Nano Lett. 10,
11 (2010).
[14] D. N. Jamieson, C. Yang, T. Hopf, S. M. Hearne, C. I.
Pakes, S. Prawer, M. Mitic, E. Gauja, S. E. Andresen, F.
E. Hudson, A. S. Dzurak, and R. G. Clark, Appl. Phys.
Lett. 86, 202101 (2005).
[15] S. R. Schofield, N. J. Curson, M. Y. Simmons, F. J. Rue,
T. Hallam, L. Oberbeck, and R. G. Clark, Phys. Rev.
Lett. 91, 136104 (2003).
[16] A. Fang, Y. C. Chang, and J. Tucker, Phys. Rev. B 66,
155331 (2002).
[17] L. C. L. Hollenberg, C. J. Wellard, C. I. Pakes, and A.
G. Fowler, Phys. Rev. B 69, 233301 (2004).
[18] M. J. Calderon, J. Verduijn, G. P. Lansbergen, G. C.
Tettamanzi, S. Rogge, and Belita Koiller, Phys. Rev. B
82, 075317 (2010).
[19] Y. L. Hao, A. P. Djotyan, A. A. Avetisyan, and F. M.
Peeters, Phys. Rev. B 80, 035329 (2009).
[20] G. P. Lansbergen, R. Rahman, J. Verduijn, G.C. Tet-
tamanzi, N. Collaert, S. Biesemans, G. Klimeck, L.C.L.
Hollenberg, S. Rogge, arXiv: 1008.1381v1 (2010).
[21] Nakul Shaji, C. B. Simmons, Madhu Thalakulam, Lev-
ente J. Klein, Hua Qin, H. Luo, D. E. Savage, M. G. La-
gally, A. J. Rimberg, R. Joynt, M. Friesen, R. H. Blick,
S. N. Coppersmith, and M. A. Eriksson, Nature Physics
4, 540 - 544 (2008).
[22] B. E. Kane, N. S. McAlpine, A. S. Dzurak, R. G. Clark,
G. J. Milburn, He Bi Sun, and Howard Wiseman, Phys.
Rev. B 61, 2961 (2000).
[23] M. J. Calderon, Belita Koiller, and S. Das Sarma, Phys.
Rev. B 75 161304 (2007).
[24] Andrea Morello, Jarryd J. Pla, Floris A. Zwanenburg,
Kok W. Chan, Kuan Y. Tan, Hans Huebl, Mikko Mt-
tnen, Christopher D. Nugroho, Changyi Yang, Jessica A.
van Donkelaar, Andrew D. C. Alves, David N. Jamieson,
Christopher C. Escott, Lloyd C. L. Hollenberg, Robert G.
Clark, and Andrew S. Dzurak, Nature, 467, 687 (2010).
[25] W. A. Harrison, Electronic structure and the properties
of solids: the physics of the chemical bond (Dover Publi-
cations Inc. , 1989).
[26] J. C. Slater and G. F. Koster, Phys. Rev. 94, 1498 (1954).
[27] Gerhard Klimeck, Fabiano Oyafuso, Timothy B. Boykin,
R. Chris Bowen, and Paul von Allmen, Computer Mod-
eling in Engineering and Science (CMES) Volume 3, No.
5 pp 601-642 (2002), ISSN: 1526-1492.
[28] B. Boykin, G. Klimeck, and F. Oyafuso, Phys. Rev. B
69, 115201 (2004).
[29] Rajib Rahman, Cameron J. Wellard, Forrest R. Brad-
bury, Marta Prada, Jared H. Cole, Gerhard Klimeck,
and Lloyd C. L. Hollenberg, Phys Rev. Lett. 99, 036403
(2007).
[30] Gerhard Klimeck, Shaikh Ahmed, Hansang Bae, Neerav
Kharche, Steve Clark, Benjamin Haley, Sunhee Lee,
Maxim Naumov, Hoon Ryu, Faisal Saied, Marta Prada,
Marek Korkusinski, Timothy B. Boykin, Rajib Rahman,
IEEE Trans. Electron Dev. 54, 2079-2089 (2007).
[31] N. Kharche, M. Prada, T. B. Boykin, and G. Klimeck,
Appl. Phys. Lett. 90, 092109 (2007).
[32] W. Kohn and J. M. Luttinger, Phys. Rev. 98, 915 (1955).
[33] A. K. Ramdas and S. Rodriguez, Rep. Prog. Phys., Vol.
44 (1981).
[34] R. Rahman, J. Verduijn, N. Kharche, G. P. Lansbergen,
G. Klimeck, L. C. L. Hollenberg, and S. Rogge, Phys.
Rev. B 83, 195323 (2011).
[35] S. Ahmed, N. Kharche, R. Rahman, M. Usman, S. Lee,
H. Ryu, H. Bae, S. Clark, B. Haley, M. Naumov, F. Saied,
M. Korkusinski, R. Kennel, M. McLennan, T. B. Boykin,
and G. Klimeck, Multimillion Atom Simulations with
NEMO 3-D, Springer Encyclopedia of Complexity and
Systems Science, edited by Robert A. Meyers Springer-
Verlag GmbH, Heidelberg, 2009, pp. 5745, ISBN: 978-0-
387-75888-6.
[36] S. Datta, Quantum Transport :
(Cambridge University Press, 2005).
[37] Szabo A., and Ostlund, N.S. in Modern Quantum
Chemistry-Introduction to Advanced Electronic Structure
Theory (Dover Publications Inc.,1989).
[38] Rajib Rahman, G. P. Lansbergen, Seung H. Park, J. Ver-
duijn, Gerhard Klimeck, S. Rogge, and Lloyd C. L. Hol-
lenberg, Phys. Rev. B, 80 165314 (2009).
[39] Hoon Ryu, Sunhee Lee, and Gerhard Klimeck, IEEE pro-
ceedings of the 13th International Workshop on Compu-
tational Electronics (IWCE), Tsinghua University, Bei-
jing, May 27-29, 2009, pg 1-4, ISBN: 978-1-4244-3925-6.
DOI: 10.1109/IWCE.2009.5091082.
[40] Sunhee Lee, Hoon Ryu, Zhengping Jiang, and Ger-
hard Klimeck,IEEE proceedings of the 13th In-
ternational Workshopon
ics (IWCE), Tsinghua University, Beijing, May 27-
29, 2009, pg 1-4, ISBN: 978-1-4244-3925-6. DOI:
10.1109/IWCE.2009.5091117.
[41] Jun-ichi Inoue, Jun Nakamura, and Akiko Natori, Phys.
Rev. B 77, 125213 (2008).
[42] H. Sellier, G. P. Lansbergen, J. Caro, and S. Rogge, N.
Collaert, I. Ferain, M. Jurczak, and S. Biesemans, Phys
Rev. Lett. 97, 206805 (2006).
[43] F. Boeuf, X. Jehl, M. Sanquer, and T. Skotnicki, IEEE
Trans. Nanotechnology 2, 144 (2003).
[44] X. Jehl, M. Sanquer, G. Bertrand, G.Guegan, S.
Atom to Transistor
Computational Electron-

Page 8

8
Deleonibus, and D. Fraboulet, IEEE Trans. Nanotech-
nology 2, 308 (2003).
[45] H. Sellier, G. P. Lansbergen, J. Caro, S. Rogge, N. Col-
laert, I. Ferain, M. Jurczak, and S. Biesemans, Appl.
Phys. Lett. 90, 073502 (2007).
[46] G. P. Lansbergen, R. Rahman, C. J. Wellard, J. Caro, N.
Collaert, S. Biesemans, G. Klimeck, L. C. L. Hollenberg,
and S. Rogge, Transport-based dopant metrology in ad-
vanced FinFETs, IEEE IEDM, San Francisco, December
1517, 2008.
[47] G. D. J. Smit, S. Rogge, J. Caro, and T. M. Klapwijk,
Phys. Rev. B 70, 035206 (2004).
[48] A. S. Martins, R. B. Capaz, and Belita Koiller, Phys.
Rev. B 69, 085320 (2004).
[49] M. J. Calderon, Belita Koiller, Xuedong Hu, and S. Das
Sarma, Phys. Rev. Lett. 96, 096802 (2006).
[50] S. Kim, A. Paul, M. Luisier, T. B. Boykin, G. Klimeck,
IEEE Transactions on Electron Devices Vol: PP, Issue
99, 1-10 (2011).
[51] A. L. Saraiva, Belita Koiller, and Mark Friesen, Phys.
Rev. B 82, 245314 (2010).