Home Search Collections Journals About Contact us My IOPscience

The finite quantum grand canonical ensemble and temperature from single-electron statistics

for a mesoscopic device

This article has been downloaded from IOPscience. Please scroll down to see the full text article.

J. Stat. Mech. (2010) P01003

(http://iopscience.iop.org/1742-5468/2010/01/P01003)

The Table of Contents and more related content is available

Download details:
IP Address: 193.205.155.31
The article was downloaded on 23/03/2010 at 09:33

Please note that terms and conditions apply.

J ournal of Statistical Mechanics: Theory and Experiment
An IOP and SISSA journal

The finite quantum grand canonical

ensemble and temperature from
single-electron statistics for a

J. Stat. Mech. (2010) P01003

mesoscopic device
Enrico Prati
Laboratorio Nazionale Materiali e Dispositivi per la Microelettronica, Consiglio
Nazionale delle Ricerche—Istituto Nazionale per la Fisica della Materia,
Via Olivetti 2, I-20041 Agrate Brianza, Italy
E-mail: [email protected]

Received 28 August 2009

Accepted 3 December 2009
Published 13 January 2010

Online at stacks.iop.org/JSTAT/2010/P01003
doi:10.1088/1742-5468/2010/01/P01003

Abstract. I present a theoretical model of a quantum statistical ensemble for

which, unlike in conventional physics, the total number of particles is extremely
small. The thermodynamical quantities are calculated by taking a small N by
virtue of the orthodicity of the canonical ensemble. The finite quantum grand
partition function of a Fermi–Dirac system is calculated. The model is applied
to a quantum dot coupled with a small two-dimensional electron system. Such
a system consists of an alternatively singly and doubly occupied electron system
confined in a quantum dot, which exchanges one electron with a small N two-
dimensional electron reservoir. The analytic determination of the temperature
of a (1 ↔ 2) electron system and the role of ergodicity are discussed. The
generalized temperature expression in the small N regime recovers the usual
temperature expression form on taking the limit of N → ∞ for the electron bath.

Keywords: mesoscopic systems (theory), quantum dots (theory), current

fluctuations

2010
c IOP Publishing Ltd and SISSA 1742-5468/10/P01003+9$30.00
 The finite quantum grand canonical ensemble and temperature

Contents

1. Introduction 2
2. The finite quantum grand canonical ensemble of a (1 ↔ 2) fermionic system 3
3. The single-electron temperature in a quantum dot with a small N electron
reservoir 6
4. Conclusion 8

J. Stat. Mech. (2010) P01003

Acknowledgment 8
References 8

1. Introduction

There is a deep connection between the nature of the elementary objects described by
quantum mechanics and the emerging properties of thermodynamical quantities. Bohr
clarified that the complementarity principle should apply to energy and temperature
measurements. Indeed, the determination of the former is incompatible with knowledge
of the latter when they refer to elementary objects [1]. After the creation of solid state
quantum dots [2], it became possible to explore confined fermionic systems constituted
of just a few electrons [3], down to a single localized electron [4]–[7]. Such systems are
electrically probed by means of accurate charge sensing capable of determining current
fluctuations corresponding to a variation of charge far below the charge unit [8]–[11]. The
experimental determination of the electron temperature in a nanostructure is a difficult
task [12], but it becomes a major issue when the system consists of a (1 ↔ 2) electron
system in particle exchange with a few electron bath. Even if the definition of temperature
T = (δS/δU)−1 in terms of the energy U and the entropy S holds independently of the size
of the ensemble [13], the systems considered here are far from bearing out the conventional
assumptions for deriving thermodynamics from statistical physics. Apparently, it should
not be possible to associate a temperature with such small systems. The reduction of the
size of the system down to one electron in thermal and particle exchange with a finite
electron system (small number of electrons N) implies a twofold change of perspective
for determining the thermodynamical quantities. The first major change is the shift from
space to time ensembles. Statistics can be recovered for a few particle open system only by
considering the average of measurable quantities in the time domain. The second change
consists in the generalization of the thermodynamics by removing the limit of large N,
which implies that the terms at the order 1/N are relevant in the determination of the
physical quantities. The equivalence under the first change of perspective is guaranteed
by the ergodicity, while the second change is granted by the orthodicity. In section 2,
I define the finite quantum grand partition ensemble in the limit of small N. Next, I
determine the probability of occupation of an electronic system capable of containing
either one or two electrons. The state equation is derived and it surprisingly depends
on the heat capacity per area unit cA . In section 3, the determination of the generalized

doi:10.1088/1742-5468/2010/01/P01003 2
 The finite quantum grand canonical ensemble and temperature

temperature expression and its dependence from experimentally measurable quantities is

described. Such evaluation is done by considering the realistic experimental condition of
a single-electron quantum dot coupled with a small electron reservoir. The results are
summarized and briefly discussed in section 4.

2. The finite quantum grand canonical ensemble of a (1 ↔ 2) fermionic system

The grand canonical ensemble consists of a large open ensemble made of identical systems,
which is in thermal equilibrium with a reservoir at a given temperature. The ensemble
under investigation and the thermal reservoir may exchange energy and particles. In this

J. Stat. Mech. (2010) P01003

section the statistical physics of a similar system is investigated. The model is derived for
a generic system of fermions and it applies for electrons in a quantum dot. In contrast to
the grand canonical ensemble case, here the open system under investigation may contain
only one or two fermions. The reservoir is a low dimensional system of a few fermions.
According to the physics and technology of semiconductor nanodevices, for which the
present study is relevant, the reservoir is well approximated by a two-dimensional electron
system (2DES), with a negligible extent in the direction perpendicular to the plane of the
electrons [7, 14]. To achieve uniformity of the notation, the island capable of capturing
one extra electron is exemplified with a trap spatially extended along two dimensions,
without any loss of generality. Such an assumption is realistic, the confinement of the
electron wavefunction being symmetric in the x–y plane both in defects at the Si/SiO2
interface and in a quantum dot which is defined either lithographically or with a split gate
in a heterostructure [14]. Therefore, all the definitions in the following refer to surfaces
Σi instead of volumes Vi where i = 1 indicates the quantum dot and i = 2 a thermal bath
made of N2 electrons.
A system has orthodicity if the averages of the physical quantities over a given
distribution recover the laws of thermodynamics. The canonical ensemble has native
orthodicity, so such properties are present for the finite N grand canonical ensemble by
construction.
A given H(N), the Hamiltonian of the system constituted by N electrons distributed
between the island and the 2DEG, can be separated as H = H1 (N1 ) + H2 (N2 ) where
N1 = 1, 2.
Consequently one defines the finite quantum grand partition function for identical
particles as follows:

2
Z(Σ, N, T ) = Z(Σ1 , N1 , T )Z(Σ2 , N2 , T ) (1)
N1 =1


2
= tr(e−βH1 ) tr(e−βH2 ) (2)
N1 =1

where N = N1 + N2 and Σ = Σ1 + Σ2 . We have introduced the function

Z(Σ2 , N2 , T )
ρ(N1 ) = e−βH1 (N1 ) (3)
Z(Σ1 + Σ2 , N1 + N2 , T )

doi:10.1088/1742-5468/2010/01/P01003 3
 The finite quantum grand canonical ensemble and temperature

where β = (kT )−1 , which by definition satisfies


2
tr(ρ(N1 )) = 1. (4)
N1 =1

Since tr(ρ) is the canonical partition function times Z(Σ2 , N2 , T )/Z(Σ, N, T ), in the
following such a ratio is calculated by means of the Helmholtz potential Ψ = −β −1 log Z,
which gives
Z(Σ2 , N2 , T )
= eβΨ(Σ1 +Σ2 ,N1 +N2 ,T )−βΨ(Σ2 ,N2 ,T ) . (5)

J. Stat. Mech. (2010) P01003

Z(Σ, N, T )
In order to explicitly evaluate Ψ, the internal energy U is now calculated, since
Ψ = U − T S = U + T (δΨ/δT ) where S = −(δΨ/δT ). Temperature T is defined as
usual using T = (δS/δU)−1 . For a Fermi 2DES of M electrons in a surface with area A,
in the limit of small temperature T , it holds that
 μF 
π2 2 
U(M, A) = A · u = A Eg(E) dE + (kB T ) [μg (μ) + g(μ)] (6)
0 6
 μF 
π2 2 
M =A·n=A g(E) dE + (kB T ) [g (μ)] (7)
0 6
where the density of states per surface unit is g(E) dE = (m/π 2 2 ) dE for a two-
dimensional system with d = 2 and μF is the Fermi energy. Since g  (E) = 0 at d = 2,

g π2
U(M, A) = Aμ2F + gA(kB T )2 (8)
2 6
M = gAμF. (9)
The inversion of equation (9) gives the chemical potential
M
μ(M, A) = (10)
gA
so the internal energy U can be equivalently written as
1 M 2 π2
U(M) = + gA(kB T )2 . (11)
2 gA 6
The above relations are valid for M = N2 electrons with A = Σ2 and for M = N1 +N2
with A = Σ1 + Σ2 , while U(N1 ) is treated separately because of the dependence of the
nature of confinement when N1 = 1, 2. The polynomial shape of U(M) = aM,A + bA T 2 ,
where aM,A = 12 (M 2 /gA) and bA = (π 2 /6)gA(kB)2 , as a function of the temperature T
implies that Ψ(M, A) = aM,A − bA T 2 , so
1 M 2 π2
Ψ(M, A) = − gA(kB T )2 . (12)
2 gA 6

doi:10.1088/1742-5468/2010/01/P01003 4
 The finite quantum grand canonical ensemble and temperature

It is useful to calculate the heat capacity per area unit at constant surface area:
 
1 ∂U(M, A) π2
cA = = gkB2 T, (13)
A ∂T A 3
and the pressure:
 
∂Ψ(M, A) 1 M2 π2
P (M, A) = − = + g(kB T )2 . (14)
∂A T 2 gA2 6
Since the island which traps the electron has a very small spatial extent, the
approximation Σ ∼= Σ2 holds, which simplifies the analysis. It is now possible to evaluate

J. Stat. Mech. (2010) P01003

the Helmholtz energy variation between N1 + N2 and N2 electrons:

ΔΨ = Ψ(N1 + N2 , Σ1 + Σ2 ) − Ψ(N2 , Σ2 ) (15)

(N1 + N2 )2 N22 π2 π2
= − − g(Σ1 + Σ2 )(kB T )2 + gΣ2 (kB T )2 (16)
2g(Σ1 + Σ2 ) 2gΣ2 6 6

∼ N12 + 2N1 N2 + N22 N22 π2

= − − gΣ1 (kB T )2 (17)
2gΣ2 2gΣ2 6
 
N1 1
= N1 μ(N2 , Σ2 ) 1 + − cA Σ1 T. (18)
2N2 2

Putting ΔΨ in equation (5) gives

Z(Σ2 , N − N1 , T )
= eβμ(N2 )N1 (1+(N1 /2N2 ))−(cA Σ1 /2kB ) . (19)
Z(Σ, N, T )
The finite quantum grand canonical ensemble can therefore be defined via the pair
(E, ρ) with E = Γ and
ρ(N1 ) = z N1 (1+N1 /2N2 ) e−(cA Σ1 /2kB )−βH(N1 ) (20)
where z = eβμ(N2 ) , while Γ indicates all the states of the system over which the summations
are performed and consists of the discrete quantum analogue of the Gibbs Γ space [15].
The probability of occupation of the subsystem 1 with L fermions is therefore
tr ρ(L) z L(1+L/2N2 ) tr(e−βH1 (L) )
p(L) = 2 = 2 (21)
N1 (1+N1 /2N2 ) tr(e−βH1 (N1 ) )
N1 =1 tr(ρ(N1 )) N1 =1 z

z L(1+L/2N2 ) tr(e−βH1 (L) )

= (22)
ZQG
where

2
ZQG = z N1 (1+N1 /2N2 ) tr(e−βH1 (N1 ) ). (23)
N1 =1

The present section is concluded by expressing the state equation

 and the relationship
between the temperature T and the occupation statistics. Since 2N1 =1 tr(ρ(N1 )) = 1, it

doi:10.1088/1742-5468/2010/01/P01003 5
 The finite quantum grand canonical ensemble and temperature

holds that
ZQG · e−(cA Σ1 /2kB ) = 1 (24)
or, equivalently, if one considers the logarithm of both sides,
cA Σ1
log ZQG = + . (25)
2kB
The generalized temperature of a (1 ↔ 2) system is therefore given by inversion:
p(1) z 1+(1/2N2 ) tr(e−βH1 (1) )
= 2(1+(1/N2 )) . (26)
p(2) z tr(e−βH1 (2) )

J. Stat. Mech. (2010) P01003

In section 3 such a ratio is explicitly evaluated by considering a realistic value of a
possible solid state quantum device.

3. The single-electron temperature in a quantum dot with a small N electron

reservoir

In this section the physical parameters involved in the electron occupation probability of
a realistic quantum dot and the consequent experimental determination of the generalized
temperature are discussed. The study of the electron occupation of a quantum dot can
be realized by measuring its charge state via the current in a channel electrically coupled
to the electron charges confined to the dot [8, 16]. In the case of a natural quantum dot
like a donor or a lattice point defect close to the Si/SiO2 interface, the channel is provided
by the two-dimensional gas formed at the interface by applying a gate voltage [9, 11]. In
the case of lithographically defined quantum dots, a current flows in the proximity of the
island which confines the localized charges [14]. Let us consider as simple a system as we
can, like a point defect close to the Si/SiO2 interface. Typically the 2DES is confined in
(50–300) × (50–300) nm2 , while the electron wavefunction is spread along a few nm in
the direction perpendicular to the 2DES [18]. The point defect can only accept one extra
electron. When the defect is paramagnetic (it switches from N1 = 1 to 2 and vice versa),
the first electron fills the ground state of a hydrogen-like shell, at energy E(1) = ET .
Indeed, the high extraction energy of the unpaired electron makes it impossible to achieve
the ionization of the first localized electron, unless a metal gate is used to manually
modify the charge state from N1 = 1 to N1 = 0, which is not the case for the situation
considered here. Because of the spin degeneracy, a second √ electron can be captured at the
same energy level ET , to constitute a singlet state 1/ 2(|↑|↓ − |↓|↑), with two extra
contributions due to the Coulomb charging energy ΔEC and to the lattice relaxation ΔEL .
While the origin of the first contribution is straightforward, the second requires a short
discussion. The presence of the second electron involves in fact a rearrangement of the
lattice [17]. At the low temperature considered here, the 2DES is weakly coupled with
the crystal [19]. Phonons are involved in the emission and capture of one electron from
the defect and the relaxation of the crystal implies a change of energy of SHR ω where
SHR is the Huang–Rhys factor and ω is the average phonon frequency in the configuration
coordinate picture [20]. In the following we consider ΔEL = SHR ω, the energy gain of
the lattice when an electron is captured. It is not possible for a natural quantum dot to
capture a third electron because the energy of the system would exceed the conduction
band edge energy.

doi:10.1088/1742-5468/2010/01/P01003 6
 The finite quantum grand canonical ensemble and temperature

J. Stat. Mech. (2010) P01003

Figure 1. The change of the charge occupation, for both an artificial quantum
dot and a natural point defect, is monitored by recording the sudden changes
of an ultraweak channel current electrically coupled with it. The capture and
the emission phenomena are generally governed by a Poisson process which
determines the average occupation time. N1 is the number of electrons in the dot.
In this example the high current state is associated with the N1 = 1 occupation.
The two states could also be reversed (N1 = 2 would refer to the high current
state), depending on the microscopic nature of the electrostatic coupling of the
island with the two-dimensional system.

The total energy of the two electrons is therefore

E(2) = 2ET + ΔEC + ΔEL . (27)
Experimentally, the time resolved traces of the current are analysed by means of fast
digitizers and they typically appear as random telegraph signals for both the quantum
dots [16] and the point defects [10]. A typical switching current trace is shown in figure 1.
The capture time τ (1) and the emission time τ (2) are obtained from averaging over
thousands of switching events and they generally obey a Poissonian statistics. Deviations
from such statistics are quantified by a non-unit Fano factor and they reflect quantum
coherence typical of peculiarly coupled quantum dots [21]. Their ratio is governed by the
ratio of the two occupation probabilities:
τ (1) p(1)
= . (28)
τ (2) p(2)
It is consequently appropriate to calculate the relative occupation probability for the
two states with one or two electrons, since their ratio can be experimentally accessed. For
the particular case discussed here, when the subsystem 1 is a paramagnetic point defect,
p(1) −βμ(1+(3/2N2 )) e−βET
= 2e (29)
p(2) e−β2ET +ΔEC +ΔEL

= 2eβ(ET +ΔEL +ΔEC −μ(N2 )(1+(3/2N2 ))) . (30)

doi:10.1088/1742-5468/2010/01/P01003 7
 The finite quantum grand canonical ensemble and temperature

The factor 2 takes into account the spin degeneracy of the ground state. The values of
p(1) and p(2) are experimentally obtained as the average occupation times for the states
1 and 2 respectively. Such identification is possible by virtue of the ergodic hypothesis.
The complete experimental determination of the parameters involved in equation (30)
provides the generalized temperature T shared by the electrons in the quantum dot and
the electrons in the 2DES. The generalized temperature of the electron(s) localized in the
dots in thermal equilibrium with the small N2 -electron bath may consequently be defined:
  
2τ (2) 3

J. Stat. Mech. (2010) P01003

T = kB−1 ln ET + ΔEL + ΔEC − μ(N2 ) 1 + . (31)
τ (1) 2N2
It is remarkable that such a definition of the temperature of a time ensemble of a few
electrons coincides with the usual one for a space ensemble just on taking N2 → ∞. For
this reason it can be considered a meaningful extension of the temperature definition for a
small quantum system of electrons of the kind treated in the present paper. Such a result
can be easily adapted to the case of the confinement induced in a artificial quantum dot,
which does not require a lattice relaxation term.

4. Conclusion

The finite quantum grand canonical ensemble (E, ρ) has been defined for a fermionic
system constituted by a cell capable of confining either one or two electrons, and a
thermal bath of finite and small size made from a two-dimensional system of a few
electrons. The state equation is governed by the heat capacity per unit surface cA
rather than the pressure. Averaging over a time ensemble replaces the average over
space ensembles for the determination of the thermodynamical quantities. The ensemble
approach has been applied for a quantum dot constituted by a natural point defect
at the Si/SiO2 interface and it holds for a general (1 ↔ 2) system. The ratio of the
characteristic times monitored via the current two-state fluctuations is given by the ratio
of the occupation probabilities calculated with the approach presented. Therefore, the
generalized temperature of such a small time ensemble can be defined and extracted from
the ratio of the average characteristic times of the two current states. Such a definition
of temperature returns the usual temperature on taking the limit N → ∞ of the electron
bath.
Acknowledgment

The author would like to thank Sergio Servadio (Universita di Pisa) for a careful reading
of the manuscript and useful suggestions and criticisms.
References
[1] Rosenfeld L, Foundations of quantum theory and complementarity, 1961 Nature 190 384
[2] Beenakker C W J, 1991 Phys. Rev. B 44 1646
[3] Sanquer M, Specht M, Ghenim L, Deleonibus S and Guegan G, 2000 Phys. Rev. B 61 7249
[4] Sellier H, Lansbergen G P, Caro J, Rogge S, Collaert N, Ferain I, Jurczak M and Biesemans S, 2006 Phys.
Rev. Lett. 97 206805

doi:10.1088/1742-5468/2010/01/P01003 8
 The finite quantum grand canonical ensemble and temperature

[5] Koppens F H L, Buizert C, Tielrooij K J, Vink I T, Nowack K C, Meunier T, Kouwenhoven L P and

Vandersypen L M K, 2006 Nature 442 766
[6] Prati E, Latempa R and Fanciulli M, 2009 Topics in Applied Physics vol 115 (Berlin: Springer) pp 241–58
[7] Prati E, Latempa R and Fanciulli M, 2009 Phys. Rev. B 80 165331
[8] Prati E, Fanciulli M, Calderoni A, Ferrari G and Sampietro M, 2008 J. Appl. Phys. 103 104502
[9] Ralls K S, Skocpol W J, Jackel L D, Howard R E, Fetter L A, Epworth R W and Tennant D M, 1984 Phys.
Rev. Lett. 52 228
[10] Prati E, Fanciulli M, Ferrari G and Sampietro M, 2006 Phys. Rev. B 74 033309
[11] Prati E, Fanciulli M, Ferrari G and Sampietro M, 2008 J. Appl. Phys. 103 123707
[12] Giazotto F, Heikkilä T T, Luukanen A, Savin A M and Pekola J P, 2006 Rev. Mod. Phys. 78 217
[13] Landau L and Lifshitz L, 1980 Statistical Physics (Course of Theoretical Physics vol 5) 3rd edn (Oxford:
Pergamon)

J. Stat. Mech. (2010) P01003

[14] Simmons C B, Thalakulam M, Shaji N, Klein L J, Qin H, Blick R H, Savage D E, Lagally M G,
Coppersmith S N and Eriksson M A, 2007 Appl. Phys. Lett. 91 213103
[15] Huang K, 1963 Statistical Mechanics Int. edn (New York: Wiley) p 189
[16] Fricke C, Hohls F, Wegscheider W and Haug R J, 2007 Phys. Rev. B 76 155307
[17] Henry C H and Lang D V, 1977 Phys. Rev. B 15 989
[18] Palma A, Godoy A, Jiménez-Tejada J A, Carceller J E and Lopez-Villanueva J A, 1997 Phys. Rev. B
56 9565
[19] Kivinen P, Savin A, Zgirski M, Torma P, Pekola J P, Prunnila M and Ahopelto J, 2003 J. Appl. Phys.
94 3201
[20] Goguenheim D and Lannoo M, 1990 J. Appl. Phys. 68 1059
[21] Kießlich G, Schöll E, Brandes T, Hohls F and Haug R J, 2007 Phys. Rev. Lett. 99 206602

doi:10.1088/1742-5468/2010/01/P01003 9