J. Phys.: Condens. Matter 10 (1998) 601615.

Printed in the UK PII: S0953-8984(98)84320-8

Energy spectra of few-electron quantum dots
E Anisimovas and A Matulis|
Vilnius University, Sauletekio 9, 2054 Vilnius, Lithuania
Semiconductor Physics Institute, Gostauto 11, 2600 Vilnius, Lithuania
Received 20 May 1997, in nal form 16 September 1997
Abstract. We present the renormalized perturbation series for the energy spectra of parabolic
quantum dots with two to ve electrons, taking into consideration the ground and the lowest
excited states. A complete classication of asymptotic energy levels is performed and the
behaviour of energy levels from the quantum to the semiclassical regime is traced. Comparison
between the present results and those from exact numerical Hamiltonian diagonalization indicates
a fair accuracy of the proposed method over the whole range of the electronelectron coupling
constant and magnetic eld values. The results obtained indicate that increasing the number of
electrons in a dot leads to more classical behaviour of the system.
1. Introduction
Quantum dots have been studied extensively as low-dimension nanostructures, interesting
from the intrinsic theoretical point of view and due to their importance to modern electronic
devices [1]. Most of the spectroscopy experiments were carried out on quantum dots with
soft parabolic connement potentials [2]. Such parabolic quantum dots attract attention
as simple model systems for electronelectron interaction studies. Although the parabolic
quantum dots have some disadvantages as indicated by the Kohn theorem [3], which shows
that usually only the centre-of-mass motion can be excited (in absorption and even in
tunnelling), the conning potential in a dot is very often rather close to the parabolic one.
This is why parabolic dots can serve as a good starting point or a reference for investigations
of more sophisticated quantum dots. The small non-parabolicity usually helps to reveal
peculiarities of the relative-electron-motion spectrum, which is sensitive to the electron
electron interaction.
Only in the case of a quantum dot with two electrons can an exact analytical solution of
the eigenvalue problem for the parabolic dot be obtained [4]. Most of the theoretical
investigations were mainly based either on mathematical tricks (which are possible in
the case of parabolic dots with two and three electrons [5, 6]) or on the straightforward
diagonalization of the Coulomb interaction [7]. Although the latter gives an exact solution
of the problem, because the calculations are time consuming it can be applied only in cases
with small numbers of electrons in the dots. This is why the construction of approximate
solutions is of interest. One such was described in [8] where in the case of a high magnetic
eld rather accurate results were obtained using the moving Eckardt frame. Recently it was
shown [9] that an approximate technique based on the renormalized perturbation series can
Present address: Department of Theoretical Physics, University of Lund, S olvegatan 14A, S-223 62 Lund,
Sweden.
| E-mail: [email protected].
0953-8984/98/030601+15$19.50 c _ 1998 IOP Publishing Ltd 601
602 E Anisimovas and A Matulis
be successfully used for energy spectrum estimations over a wide coupling constant range.
Another technique based on Pad e approximants [10] was shown to be very successful in
the case of a parabolic dot with a small number of electrons. Each of these approximations
is based on a technique for interpolating between the perturbation series in the coupling
constant and the asymptotic expansion valid in the large-coupling-constant range.
In the present paper, in order to obtain a complete picture of the energy spectrum,
including the ground and some excited states, we performed a complete classication of
the asymptotic energy levels, paying attention to the symmetry properties of the spin-
wave-function part, and constructed the renormalized perturbation series for the energies of
quantum dots with two to ve electrons on the basis of a rather simple two-term expansion
for both limiting regions of the coupling constant.
The present paper is organized as follows. In section 2 the problem is formulated and the
perturbation series technique is outlined. The asymptotic series are formulated in section 3
and the spin-wave-function part is constructed for the asymptotic region in section 4. The
interpolation technique is described in section 5 and in the last section, section 6, the results
for quantum dots with two to ve electrons are presented and discussed. In appendix A
the algebraic expression for the Coulomb matrix element is given and in appendix B all
coefcients which are necessary for construction of the renormalized perturbation series are
collected together.
2. Perturbation series
Generally the system of N electrons in a quasi-2D quantum dot is described by the following
Hamiltonian:
H =
N

i=1
_

h
2
2m

2
i

m
2
0
2
r
2
i
_

i<j
e
2
[r
i
r
j
[
. (1)
Here m stands for the effective electron mass, is the static dielectric constant and
0
is the characteristic frequency of the parabolic connement potential. For convenience
we shall scale the coordinates: r a
0
r (a
0
=

( h/m
0
)), and measure the energies
in dimensionless units of h
0
. After this transformation, we obtain the dimensionless
Hamiltonian
H =
1
2
N

i=1
(
2
i
r
2
i
)
N

i<j
1
[r
i
r
j
[
= H
0
V (2)
with the single dimensionless coupling constant = a
0
/a
B
(a
B
= h
2
/me
2
is the effective
Bohr radius) characterizing the energy spectrum of the electrons in the parabolic quantum
dot.
The simplest way to solve the eigenvalue problem is just to use the perturbation series
in the coupling constant :
E() = E
0
E
1
. (3)
The rst coefcient is the eigenvalue of the unperturbed Hamiltonian H
0
and can be
presented as a sum of non-interacting electron energies:
E
0
=
N

i=1
(i) (i) = 1 [m
i
[ 2n
i
(4)
where i = (n
i
, m
i
, s
i
) labels the one-electron quantum number set.
Energy spectra of few-electron quantum dots 603
The rst-order correction E
1
can be estimated in a standard way (see, for instance, [11])
and presented as a sum of two-electron Coulomb matrix elements:
ij[V[j
/
i
/
) =
s
i
,s
i
/

s
j
,s
j
/
_
d
2
r
1
_
d
2
r
2
1
[r
i
r
j
[

(i[r
1
)

(j[r
2
)(j
/
[r
2
)(i
/
[r
1
) (5)
where the one-electron wave functions can be expressed in associated Laguerre polynomials:
(n, m[r, ) =
_
2n!
2([m[ n)!
e
im
r
[m[
e
r
2
/2
L
[m[
n
(r
2
). (6)
For numerical estimation of the Coulomb matrix element (5) we used the analogue of the
Girvin and Jach expression [12] which is given in appendix A.
When calculating the rst-order correction, the main problem is the degeneracy of
the zero-order energy levels, because the good quantum numbers of the total angular
momentum M and the total spin S are not sufcient for classication of energy levels.
That is why we used numerical diagonalization of the degenerate blocks of the above rst-
order correction matrix. The results obtainedthe coefcients E
0
and E
1
are collected
together in appendix B.
3. Asymptotic expansion
In the case, the kinetic energy of the electron system is small compared to
the potential energy. It is known that such systems are strongly correlated and show a
tendency to exhibit Wigner crystallization [13]. Monte Carlo results for classical particles
in the parabolic dot [14] show that the particles crystallize in a structure composed of
concentric rings. When the number of electrons in a dot is small, N 5, they fall
into a single ring on which the electrons are located equidistantly, separated by the angle
= 2/N. The quantum dynamics of such a Wigner crystal can be described by means of
the rotation and vibration modes. The general theory of vibration modes is given in [8]. It
is based on introducing the moving Eckardt frame and the diagonalization of the vibrational
Hamiltonian in the harmonic approach. In the case of a small number of electrons located
on a single ring, the vibrational Hamiltonian and the problem of its diagonalization can
be signicantly simplied by representing the electron coordinates as complex variables
z = x iy and turning them into local electron coordinates on the ring by means of the
following transformation:
z
n
(R z
n
)e
in
. (7)
Here n labels the electrons on the ring. We shall follow the simpler method, restricting our
consideration by using the following two-term asymptotic expansion:
E() = c
0

2/3
c
1
. (8)
The rst term of this expansion can be easily obtained by minimizing the potential
in (2). The minimization leads to the equilibrium ring radius R = (A)
1/3
(where
A =

N1
n=1
[4 sin(n/2)[
1
) which corresponds to the main term c
0

2/3
= 1.5NR
2
in
the asymptotic energy expansion. The next term in expansion (8) corresponds to the ring
oscillations. To obtain it, we insert (7) into Hamiltonian (2) and expand the potential into
a z-series. This leads to the following second-order correction to the equilibrium potential:
V
2
=
1
2
N

n=1
[z
n
[
2

N
32A
N1

n=1
(3
2
n
3
2
n
2[
n
[
2
)
604 E Anisimovas and A Matulis
where

nm
=
z
n
e
in
z
m
e
im
e
in
e
im
=
z
n
[1 e
i(nm)
] z
m
[1 e
i(mn)
]
4 sin
2
[(n m)/2]

nm
.
Now let us apply the Fourier transform
z
n
=
1

k
e
ikn
z
k
(9)
where k = 0, 1, . . . , N1 denotes the vibration modes. Inserting it into the above potential
expression and subsequently into (2), after simple but laborious transformation we obtain
the following expression for the asymptotic Hamiltonian with the second-order potential
corrections:
H =

k
H
k
=

k
_
2

2
z
k
z

k
A
k
z
k
z

k
B
k
(z
k
z
k
z

k
z

k
)
_
(10)
where
A
k
=
1
2

1
16A
N1

n=1
[sin(n/2)[
3
sin
2
[(k 1)n/2]
B
k
=
3
32A
N1

n=1
[sin(n/2)[
3
[sin
2
(n/2) sin
2
(kn/2)].
The Hamiltonian is composed of parts H
k
which describe the behaviour of both
polarizations of the k-modes as a system of four coupled harmonic oscillators. They
can be diagonalized in a standard way, just introducing the appropriate variables. Let us
separate the real and imaginary parts of the initial variables, and introduce the new variables
in the following way:
z
k
= x iy z
k
= u iv. (11)
In these variables every part H
k
of Hamiltonian (10) splits into two terms. The rst term is
H
(1)
k
=
1
2
_

2
x
2


2
u
2
_
A
k
x
2
A
k
u
2
2(B
k
B
k
)xu (12)
while the other part H
(2)
k
can be obtained simply by making the replacements x y and
u v.
H
(1,2)
k
both have the same eigenvalues, which can be found by the diagonalization of
the symmetric matrix
_
A
k
B
k
B
k
B
k
B
k
A
k
_
which denes the quadratic form of the potential term, and leads to the following expression
for the eigenfrequencies:

()
k
=
_
A
k
A
k

_
[A
k
A
k
]
2
4[B
k
B
k
]
2
. (13)
Thus, since the transformed Hamiltonian corresponds to the set of independent harmonic
oscillators, the corresponding vibration wave function can be constructed in a standard
way, introducing the phonon creation (a

k,
) and annihilation (a
k,
) operators:
= a

k
1
,
1
a

k
p
,
p
[0). (14)
Energy spectra of few-electron quantum dots 605
The vacuum state [0) corresponds to the systems ground state and the symbol = 1, 2
stands for the polarization.
For interpolation purposes it is important to know the wave functions symmetry
properties with respect to the rotation of the electronic ring as a whole. Let us denote
this operation by the operator P. It is evident that the ground state is invariant under
this operation, i.e. P[0) = [0). As the above rotation is equivalent to the substitution
n n 1, it follows from expression (9) that Pz
k
= exp(ik)z
k
. Due to the mixing of
the two k-modes, the operators a

k,
and wave function (14) obey a more complicated
symmetry condition. However, because of the degeneracy of the vibration states obtained,
it can be signicantly simplied using the appropriate superposition of the eigenstates of
the two Hamiltonian parts H
(1,2)
k
. So, we shall assume that a wave-function superposition
of this sort is constructed, and consequently that the phonon operators obey the symmetry
condition
Pa

k,
= e
ik
a

k,
. (15)
We shall use the above symmetry condition for the classication of the asymptotic
vibration modes obtained, whose frequencies
()
k
are collected together in table 1.
Table 1. The ring vibration frequencies and the asymptotic expansion coefcients.
N: 2 3 4 5
k : 1 2 1 2 1 2 1 2
0 1.7321 0.0 1.7321 0.0 1.7321 0.0 1.7321 0.0
1 1.0 1.0 1.2247 1.0 1.3186 1.0 1.3719 1.0
2 1.2247 1.0 1.4888 0.8852 1.6890 0.7274
3 1.3186 1.0 1.6890 0.7274
4 1.3719 1.0
c
0
1.1905 3.1201 5.8272 9.2801
c
(0)
1
1.8660 3.0908 4.3716 5.6544
Two modes with eigenfrequencies
()
k
= 1 correspond to the centre-of-mass motion.
We do not take them into account as we consider the electron-relative-motion energy
spectrum only. The mode k = 0 corresponds to the breathing mode for = 1 and
to the rotation of the ring as a whole for = 2. The latter case should be considered
separately. Actually, in the harmonic approximation used, the rotation mode is separated
from the vibration modes. Its eigenfunction is exp(iM) ( is an angle determining the
rotation of the ring as a whole) and the corresponding energy is M
2
/2NR
2

2/3
. It is
seen that in the case of large angular momenta M this term becomes crucial and thus it
should be included into the main asymptotic energy term, and consequently it leads to the
equilibrium ring radius dependence on the total angular momentum M (see [8], for instance).
However, the above term is of order
2/3
and lies beyond the approximation (8). Hence
considering the states with small M only, we neglect this rotation mode correction (and the
dependence of the ring radius on M as well) together with the anharmonicity corrections
which are of the same order of magnitude. We should keep in mind, however, that there
are many states labelled by different M-values corresponding to every asymptotic vibration
term.
606 E Anisimovas and A Matulis
Thus, the rst-order correction is given by
c
1
= c
(0)
1

k,
n
k,

()
k
where c
(0)
1
is the contribution of the ground-vibration-state energy and the symbol n
k,
stands
for the lling factor of the corresponding phonon mode. These coefcients together with
the rst-term coefcients c
0
are given in the last two lines of table 1.
Separating the rotation mode, we shall represent the total wave function in the asymptotic
region as a product of rotation, vibration and spin parts:
= e
iM
. (16)
According to Paulis principle the total wave function should be antisymmetric. In the
asymptotic region the classical electron equilibrium positions are separated by high potential
barriers. That is why we will restrict ourselves to the physically important permutation
the rotation P. As the rotation corresponds to an odd permutation in the case of an even
number of electrons (and vice versa), the total wave function obeys the following symmetry
condition:
P = (1)
N1
. (17)
According to (14)(16), the rotation and vibration parts obey the condition
Pe
iM
= e
i(M )
e
iM
=

k,
/
kn
k,
. (18)
The prime on the summation indicates that the mode corresponding to the ring rotation is
excluded. The different factors in the two above conditions should be compensated by the
appropriate choice of the spin-function part.
4. The spin-wave-function part
Following [8], we diagonalize the operator P in the S
z
(total spin z-projection operator)
eigenfunction space and construct the spin-function part obeying the symmetry condition
P = e
i
(19)
which, together with (17) and (18), leads to the following selection rule:
M = Np. (20)
In the case of an odd number N of electrons in a dot, the symbol p stands for an arbitrary
integer, while in the case of even N, it should be replaced by p 1/2. The integer para-
meter which characterizes the spin-function-part symmetry properties should be dened
for every particular case. In the case in which the above procedure leads to the degeneracy
of spin states, the total spin operator S
2
should be diagonalized additionally.
The case of two electrons is a trivial one. There is the triplet function [) with
S
z
= 1 corresponding to = 0 and the singlet function {[) [)]/

2 with S
z
= 0
corresponding to = 1. We shall not consider the functions with S
z
< 0 as they give no
extra information.
The case of three electrons is considered in detail in [8]. There is a quartet state
corresponding to = 0 and two doublet states with = 1.
Applying the same treatment to the systems with four and ve electrons we obtain the
spin-function parts with the symmetry properties of (19) type. All of the necessary -values
are collected together in table 2.
Energy spectra of few-electron quantum dots 607
Table 2. -values for the various multiplets and the ground-state terms.
N S = 0 S = 1/2 S = 1 S = 3/2 S = 2 S = 5/2 Terms
2 1 0
1
S,
3
P,
1
D,
3
F, . . .
3 1, 2 0
4
S,
2
P,
2
D,
4
F, . . .
4 0, 2 1, 2, 3 0
1,3
S,
3
P,
1,5
D,
3
F,
1,3
G, . . .
5 0, 1, 2, 3, 4 1, 2, 3, 4 0
2,6
S,
2,4
P,
2,4
D,
2,4
F,
2,4
G, . . .
The above selection rule gives us the possible terms for every vibration state
characterized by a -value. The terms for the ground vibration state ( = 0) are collected
together in the last column of table 2. For the sake of brevity we use notation analogous
to that adopted in atomic spectroscopy. Thus the capital letters S, P, D, . . . indicate the
corresponding M = 0, 1, 2, . . . values, and the left-hand superscript symbol stands for the
multiplicity (2S 1). The terms corresponding to the excited vibration states with ,= 0
can be easily obtained from those presented in table 2 by applying the shift M M .
Furthermore, we shall label those excited terms with the additional right-hand subscript
symbol k, indicating the modes of the phonons present in that state.
5. Interpolation
In constructing the interpolating expression we shall follow [9], where the perturbation
series renormalization procedure is given in detail. The main idea is the replacing of (2) by
the generalized Hamiltonian
H =
1
2
N

i=1
(
2

2
i
r
2
i
)

i<j
1
[r
i
r
j
[
(21)
and using the scaling relation for the generalized eigenvalue:
E(, ) = E(1,
3/2
). (22)
The eigenvalue of the basic problem with Hamiltonian (2) is given as E() = E(1, ).
The scaling relation denes the trajectory =
0

3/2
in the -plane, as is shown in
gure 1. That trajectory enables us to map the vertical dashed line = 1 (where we need
to obtain the eigenvalue of the basic problem) onto the tilted line KP which we dene as
= K K = (1 tan )/2
= P(1 ) P = (1 cot )/2.
(23)
For instance, the point
0
is mapped to the point (, ).
Unlike the line = 1, the line KP is located in the nite region of the -plane
(0 < < 1), and that is why the energy E((), ()) can be successfully expanded into
a -series. Adjusting the above series to the -series (3) in the region 0 and to the
asymptotic expansion (8) in the region , we obtain the nal expression for the
renormalized series:
E = (b
0
b
1
b
2

2
b
3

3
)/(1 )
=
K
P
3/2
(1 )
3/2
.
(24)
608 E Anisimovas and A Matulis
Figure 1. The mapping in the -parameter plane.
The expansion coefcients can be expressed in terms of those in (3) and (8) as
b
0
= E
0
b
1
= KP
3/2
E
1
E
0
b
2
= (7/3)K
2/3
P
1
c
0
c
1
E
0
2KP
3/2
E
1
b
3
= (4/3)K
2/3
P
1
c
0
c
1
E
0
KP
3/2
E
1
.
(25)
As we restricted ourselves to the two-term approximation in both expansions (3) and (8),
we introduced the additional adjustable parameter (as proposed in [15]) for improving the
interpolation accuracy. The geometric meaning of that parameter is clear from gure 1it
indicates the rotation angle of the mapping line KP around the point M to which the energy
E(

2) is always mapped. The rotation changes the relative contributions of the -series
and the asymptotic expansion to the renormalized series. The value = 0 corresponds to
the full control of the renormalization procedure by the -series and the value = /2 to
that of the asymptotic expansion.
It was shown in [15] that usually the energy exhibits some plateau as a function of the
parameter controlling the mapping line ( in our case). When the number of terms in the
-series is increased, that plateau shows a tendency to increase, too, thus giving a good
basis for the choice of the controlling parameter. In gure 2 the energy at the xed point
=

2 as a function of the parameter is shown by the solid line in the case of the
rst excited triplet state of the two-electron system. As our -series are rather short, we
nd no prominent plateau. However, some curve twist is present, and we hope that it is a
rather good idea to choose the controlling parameter somewhere between the points Max
and Min. The difference of energy values corresponding to those points can serve as an
estimation of the interpolation accuracy. We see that the accuracy is of about 1%.
We used the inection point of the E(

2) versus curve as the choice for the controlling

parameter . The energy value obtained with that choice is shown by the dashed line in
gure 2. Comparing it with the exact result which is known for the two-electron case
Energy spectra of few-electron quantum dots 609
Figure 2. The rst excited triplet state for two electrons: (1) the energy versus plot; (2) the
exact result; (3) the energy value at the inection point.
(dotted line), we see that the accuracy is even better than 1%. The same energy versus
behaviour and the same accuracy were obtained for all other terms considered. The
inection point values of for all of the terms considered are given in appendix B in
tables A1 and A2. They, together with the coefcients E
0
, E
1
and expressions (23)(25),
determine the renormalized perturbation series and enable one to obtain the energies over
the whole range of -values.
We should note that according to [9] the same parameter values can be used for
constructing the renormalized series in the case in which the additional magnetic eld
B is applied in the direction perpendicular to the quantum dot plane. In that case we have
E(, B) =
_
1 (
c
/2
0
)
2
E(
/
) M
c
/2
0
g

B
BS
z
/ h
0
(26)
and

/
= [1 (
c
/2
0
)
2
]
1/4
(27)
where the symbol
c
= eB/mc stands for cyclotron frequency. Note that there is a misprint
in (5.4) of [9]. The last term added in (26) corresponds to the Zeeman energy, where g

is
the effective g-factor and
B
is the Bohr magneton.
6. Results and discussion
As an illustration of the renormalization technique introduced, we present in gures 36 the
excitation energy versus plot for the relative electron motion for quantum dots with two
to ve electrons. The inclusion of the centre-of-mass-motion energy is a trivial one. One
needs just to shift up the spectra presented by the some integer value, as the centre-of-mass
motion is not affected by the electron interaction.
610 E Anisimovas and A Matulis
Figure 3. The spectrum for two electrons.
Figure 4. The spectrum for three electrons.
The main problem is that of deciding how the terms in the two regions ( 0 and
) should be interconnected. Taking into account the fact known from quantum
mechanics that two terms with the same good quantum numbers never cross each other, we
used the simplest term connection scheme. Starting from the ground state we connected
successively each term in the region 0 to the lowest possible unconnected term with
the same quantum numbers M and S in the asymptotic region.
In gures 3 and 4 the energy plots for the systems of two and three electrons are
presented. The terms with the zero energy exceeding the ground-state energy by 3 and
2, correspondingly, are depicted. In the cases of four and ve electrons, the energy plots
are given in gures 5 and 6. The degeneracy of the states grows rapidly with the number
Energy spectra of few-electron quantum dots 611
Figure 5. The spectrum for four electrons.
Figure 6. The spectrum for ve electrons.
of electrons and thus here we present only the two lowest terms tending to each of the
asymptotic vibration states. The asymptotic terms in the above gures are labelled by
the numbers in parentheses along the right-hand axis, which indicate the k-values of the
phonons present in the state. The phonon state with polarization corresponding to the larger
frequency is indicated by a prime.
It can be clearly seen how the term structure transforms from that of the weakly
interacting quantum system, where the energy levels are governed by the lling of non-
interacting single-electron states, to the semiclassical strongly correlated system, where
the energy levels correspond to the electronic ring vibration frequencies. Such switching
of the dot behaviour can be explained by means of simple dimensional considerations.
Namely, the three terms in Hamiltonian (1) can be estimated as h
2
N/mR
2
, m
2
0
R
2
N and
612 E Anisimovas and A Matulis
Figure 7. The ground-state energy of a quantum dot in a magnetic eld: (a) two electrons;
(b) three electrons; (c) four electrons; (d) ve electrons. The single-electron ground-state energies
E(1) are subtracted.
e
2
N(N 1)/R. Equating them, we dene the critical -value

1/(N 1) where the

above-mentioned switching occurs. This is in agreement with our energy dependencies. As
the dimensionless coupling constant is proportional to a
0
1/

0
, the above estimation
leads to some condition for the connement potential frequency showing whether the dot
behaves like a quantum system or a classical one. When the connement potential is mainly
caused by some positive neutralizing charge located at the rather distant gate, the increase
Energy spectra of few-electron quantum dots 613
in the connement potential frequency
0
does not exceed N [7]. In that case, increasing
the number of electrons in a dot leads to a more classical system.
As an illustration of the renormalized series technique presented, for the case with a
magnetic eld (according to (26)) we calculated the ground-state energies versus magnetic
eld for two to ve electrons; these are displayed in gures 7(a)7(d). The curve in (a)
for the two-electron system is in good agreement with that presented in [7] (we choose
the parameters as in [7] and also include the ctitious gate charge, but do not subtract the
zero-point motion of the centre of mass). It is seen from gures 7(b)7(d) that on increasing
the number of electrons the structure of the dependencies becomes less well resolved and
the peaks lose their sharpness. Thus we miss a tiny additional peak in the position marked
by an arrow in gure 7(b) which is present in the data from the exact diagonalization.
Comparing our results with those obtained in [8], say, for N = 3, B = 20 T, S = 3/2 and
M = 3, we obtained the ground-state energy value 77.5 meV, and nd a good coincidence,
which proves that the approximate separation of rotation and vibration modes works fairly
well for small angular momenta.
In concluding, we would like to mention that the parabolic quantum dot is, due to the
softness of the connement potential, a rather favourable object as regards constructing the
interpolation expressions, in contrast to a hard-wall dot [17], for which the accuracy is much
less good.
Acknowledgment
One of us (EA) is grateful to Professor Per Hedeg ard (rsted Laboratory, University of
Copenhagen) for useful comments.
Appendix A. Coulomb matrix elements
In [12], the algebraic expression for calculation of the two-particle matrix element of the
Coulomb interaction of electrons in the lowest Landau level (corresponding to n
i
= 0 in
our case) was used (see the comments in [16]). In a more general (n
i
,= 0) case, the
analogous Girvin and Jach expression can be derived, employing the same centre-of-mass
and relative-motion complex-coordinate technique after expanding the associated Laguerre
polynomials in sums and representing the terms in the form
(r
1
e
i
1
)

1
(r
2
e
i
2
)

2
(r
2
e
i
2
)

3
(r
1
e
i
1
)

4
e
(r
2
1
r
2
2
)
[r
1
r
2
[
.
Here the symbols
i
stand for numbers:

1
= j
1
j
4
([m
1
[ m
1
)/2 ([m
4
[ m
4
)/2

4
= j
1
j
4
([m
1
[ m
1
)/2 ([m
4
[ m
4
)/2.
(A1)
The symbols j
i
are integer summation indices running from 0 to n
i
.
2
and
3
can be
obtained from (A1) by performing the following replacement of indices: 1 2 and 4 3.
The nal expression reads
12[V[34) =
s
1
,s
4

s
2
,s
3

m
1
m
2
,m
3
m
4
_
4

i=1
n
i
!
([m
i
[ n
i
)!
_
1/2
n

(4)j=0
(1)
j
1
j
2
j
3
j
4
j
1
!j
2
!j
3
!j
4
!

_
4

i=1
_
n
i
[m
i
[
n
i
j
i
_
_
1
2
(G1)/2

(4)l=0
(1)

3
l
2
l
3
614 E Anisimovas and A Matulis
Table A1. Coefcients for quantum dots with N = 2 and N = 3 electrons.
N = 2 N = 3
Term E
0
E
1
Term E
0
E
1

1
S 2 1.2533 0.3571
2
P 4 2.8200 0.5871
3
P 3 0.6267 1.2211
4
S 5 1.8800 1.0722
1
D 4 0.4700 1.3993
2
D 5 2.2325 0.9358
1
S
0
4 0.9400 0.7204
2
S
1
5 2.5850 0.5836
3
F 5 0.3917 1.4651
4
F 6 1.7037 1.1984
3
P
0
5 0.5483 1.3065
4
P
1
6 1.8212 1.0783
2
P
1
6 1.8413 1.0698
2
F
1
6 2.2618 0.8991
2
P
0
6 2.5942 0.7067
Table A2. Coefcients for quantum dots with N = 4 and N = 5 electrons.
N = 4 N = 5
Term E
0
E
1
Term E
0
E
1

3
S 6 5.0133 0.6105
2
P 8 8.3032 0.4913
1
D 6 5.2483 0.5800
4
D 9 7.4416 0.6649
1
S 6 5.4832 0.5527
2
S 9 7.4812 0.6607
3
P 7 4.3608 0.8333
4
S
2
9 7.6765 0.6409
3
F 7 4.6412 0.7828
2
S
2
9 7.7647 0.6325
1
P
1
7 4.6999 0.6087
2
G 9 7.8234 0.6270
3
P
2
7 4.8628 0.6482
2
D 9 7.8284 0.6265
5
D 8 3.5641 1.0635
2
D
2
9 8.1710 0.5969
5
S
2
8 3.6816 0.9920
2
S
2
9 8.3125 0.5263
3
S
1
8 3.9491 0.8956
4
P 10 6.7149 0.8303
1
S
2
8 3.9822 0.9218
4
F 10 6.8596 0.8110
3
D
2
8 4.0215 0.9130
4
P
2
10 6.9609 0.7980
1
G 8 4.0919 0.9543
2
F 10 6.9615 0.7980
3
G 8 4.1320 0.9466
2
P
2
10 7.1355 0.7768
1
D
2
8 4.3311 0.8487
4
F
2
10 7.2010 0.7224
3
D
1
8 4.3783 0.8004
2
P
2
10 7.2489 0.7168
1
S
2
/ 8 4.3866 0.7810
2
H 10 7.2543 0.7631
3
S
1
8 4.5628 0.7656
2
F
2
10 7.3103 0.7568
3
D
1
8 4.5642 0.7654
4
P
2
10 7.3563 0.7046
1
G
2
8 4.6678 0.7885
2
P
1
10 7.4043 0.6993
1
D
2
/ 8 4.7087 0.7220
2
P
1
10 7.5375 0.6783
3
S
0
8 4.7653 0.6836
2
F
2
10 7.6303 0.6757
1
S
0
8 4.8182 0.6751
2
F
1
10 7.6670 0.6720
1
D
0
8 5.0062 0.6472
2
P
2,2
10 7.9283 0.6407
1
S
2,2
8 5.2186 0.6140
2
P
2,2
10 8.0142 0.6331

l
1
l
2
,l
3
l
4
_
4

i=1
_

i
l
i
_
_

_
1

2
_

_
G1
2
_
(A2)
where
n

(4)j=0

n
1

j
1
=0
n
2

j
2
=0
n
3

j
3
=0
n
4

j
4
=0
G =

i
=

i
l
i
_
n
m
_
=
n!
m!(n m)!
.
Energy spectra of few-electron quantum dots 615
Although the above expression involves terms with alternating signs and leads to
numerical difculties in the case of large quantum number values [16], we found it rather
convenient and accurate in the cases of quantum dots with small numbers of electrons.
Appendix B. Expansion coefcients
Here, in tables A1 and A2, the coefcients E
0
and E
1
of the perturbation series (3), and
the inection point angle are presented for all terms under consideration. These values,
together with the asymptotic expansion (8) given in table 1 and expressions (23)(25), are
sufcient for constructing the renormalized perturbation series.
References
[1] Johnson N F 1995 J. Phys.: Condens. Matter 7 965
[2] Meurer B, Heitmann D and Ploog K 1993 Phys. Rev. B 48 11 488
[3] Maksym P A and Chakraborty T 1990 Phys. Rev. Lett. 65 108
[4] Taut M 1993 Phys. Rev. A 48 3561
[5] Merkt U, Huser J and Wagner M 1991 Phys. Rev. B 43 7320
[6] Pfannkuche D and Ulloa S E 1995 Phys. Rev. Lett. 74 1194
[7] Hawrylak P 1993 Phys. Rev. Lett. 71 3347
[8] Maksym P A 1996 Phys. Rev. B 53 10 871
[9] Matulis A and Peeters F M 1994 J. Phys.: Condens. Matter 6 7751
[10] Gonzalez A 1997 J. Phys.: Condens. Matter 9 4643
[11] Fjrestad J O, Matulis A and Chao K A 1997 Phys. Scr. T 69 138
[12] Girvin S M and Jach T 1983 Phys. Rev. B 28 4506
[13] Maksym P A 1993 Physica B 184 385
[14] Bedanov V M and Peeters F M 1994 Phys. Rev. B 184 2667
[15] Fernand ez F M, Arteca G A, Maluendes S A and Castro E 1984 Phys. Lett. 103A 19
[16] Stone M, Wyld H W and Schult R L 1992 Phys. Rev. B 45 14 156
[17] Matulis A, Fjrestad J O and Chao K A 1997 Phys. Scr. T 69 85