This action might not be possible to undo. Are you sure you want to continue?

Welcome to Scribd. Sign in or start your free trial to enjoy unlimited e-books, audiobooks & documents.Find out more

Download

Standard view

Full view

of .

Look up keyword

Like this

Share on social networks

1Activity

×

0 of .

Results for: No results containing your search query

P. 1

PhysRevE.84.041125: Exact and approximate methods of calculating the sum of states for noninteracting classical and quantum particles occupying a finite number of modes Ratings:

(0)|Views: 39|Likes: 0

Published by Agnieszka M-w

Phys. Rev. E 84, 041125 (9/2011)

Phys. Rev. E 84, 041125 (9/2011)

Published by: Agnieszka M-w on Sep 21, 2011

Copyright:Traditional Copyright: All rights reserved

See more

See less

https://www.scribd.com/doc/65816320/PhysRevE-84-041125-Exact-and-approximate-methods-of-calculating-the-sum-of-states-for-noninteracting-classical-and-quantum-particles-occupying-a-fini

11/12/2011

Exact and approximate methods of calculating the sum of states for noninteractingclassical and quantum particles occupying a ﬁnite number of modes

Agnieszka Werpachowska

∗

Department of Physics and Astronomy, University College London,Gower Street, London WC1E 6BT, United Kingdom

We present exact expressions for the sum of states of noninteracting classical and quantum parti-cles occupying a ﬁnite number of modes with arbitrary spacings. Exploiting a probabilistic analogy,we derive an analytic fourth-order approximation to the density of states, which captures its vari-ance and kurtosis, and is superior to the previous, commonly used methods for all three particlestatistics. Our approach employs a simple exact method of calculating the moments of the micro-canonical density of states for quantum particles, which requires less computational eﬀort than thecommonly used saddle-point approximation. We test our methods numerically and discuss theirapplicability to various physical systems.

I. INTRODUCTION

Erwin Schr¨odinger wrote “There is, essentially, onlyone problem in statistical thermodynamics: the distribu-tion of a given amount of energy

E

over

N

identical sys-tems” [1]. As the number of particles in a system growsto inﬁnity, almost all its copies (or, expressing the samenotion diﬀerently, almost all its conﬁgurations) have thesame internal energy. In this limit, the canonical ensem-ble (where we control the temperature) is equivalent tothe microcanonical ensemble (where we control the en-ergy). The opportunity to pass between these diﬀerentdescriptions gives the possibility to choose the smarterway to measure a given quantity. Very often the calcula-tion of thermodynamical quantities in the microcanonicalensemble is an impractical task and thus one is forced toresort to the canonical ensemble. However, their equiva-lence breaks down for small or perfectly isolated systems.One then has to work with the microcanonical ensemble,whose workhorse is the sum of states function, whichcounts the number of microstates realizing a particularvalue of the control parameter. This is the case, for ex-ample, when investigating atoms in microcavities [2] orelongated magneto-optical traps [3], heavy nuclei [4, 5],coupled spins on a lattice [6] or calculating the entropyof a black hole [7, 8].In all physical problems listed above the system iscomposed of

N

classical (distinguishable), bosonic orfermionic particles which can occupy

S

+1 modes, num-bered from 0 to

S

. Each mode can be

g

s

-degenerate andhas the excitation

E

s

>

0, with

E

s

< E

s

+1

and

E

0

= 0.The last assumption does not lead to the loss of gener-ality, because we can always shift the excitations up ordown to ensure that the lowest mode has zero excitation.In many applications, we have

E

s

=

s

(e.g. lattice spinsor harmonic traps) or

E

s

=

s

2

(square potential wells).The control parameter for which we calculate the sum

∗

Electronic address:

a.werpachowska@ucl.ac.uk

of states is deﬁned as

M

=

S

s

=0

E

sg

s

−

1

t

=0

n

st

,

(1)where

n

st

= 0

,

1

,...

is the number of particles occupy-ing mode

s

and (in the case of degenerate modes) itssubmode

t

, subject to the constraint

S

s

=0

g

s

−

1

t

=0

n

st

=

N .

The number of physical states of

N

particles realizing agiven

M

is called the sum of states Ω(

N,M,S

). Whenparticles are distinguishable,

M

can be the total magne-tization of a system of lattice spins [6] or the total areaof a black hole [7]. For bosons,

M

is often the total en-ergy of particles in a harmonic oscillator potential [3, 9]or atoms in a microwave cavity [2]. For fermions, it canbe the

z

component of the total angular momentum of aheavy nucleus [4, 5]. For the sake of generality, we call

M

a total excitation.The paper concerns with the calculation of Ω(

N,M,S

)for noninteracting classical, bosonic or fermionic particlesoccupying a ﬁnite number of modes with arbitrary spac-ings. First, we investigate exact (the popular recursiveand proposed iterative) expressions for the marginal orcumulative sum of states, which are useful in practicalcalculations. Exploiting a probabilistic analogy, we thenobtain an analytic fourth-order approximation to the nor-malized sum of states (density of states)

ω

, which cap-tures its mean, variance and kurtosis. The approximationis extended to an analytic-numerical scheme which canalso match the skewness of

ω

and is better suited to han-dle excitation spectra with strongly non-homogeneousdensities. We demonstrate numerically that our analyticapproximation is superior to the previous ones (the Gaus-sian one and its polynomial corrections ﬁtted also to thekurtosis—e.g. [4, 5, 10, 11]) and matches closely the exactsum of states for all three particle statistics. Additionally,our approach employs a simple

exact

method of calculat-ing the moments of the microcanonical density of states

2for quantum particles, which requires less computationaleﬀort than the commonly used saddle-point approxima-tion (see e.g. [5, 12]). We discuss the applications of ourmethods to diﬀerent physical systems pointing out theiradvantages over commonly employed approaches.

II. EXACT EXPRESSIONS FOR

Ω

We begin by reviewing the known recursive exact ex-pressions for the sum of states of classical and quantumnoninteracting particles [13], and proceed later to derivetheir iterative counterparts.

A. Recursive expressions

1. Sum of states

The total excitation of

N

particles distributed among

S

+1 modes is deﬁned by (1). If the mode

S

(degenerateor not) is occupied by

n

particles, than the remainingexcitation

M

−

nE

S

can be redistributed among the

N

−

n

particles in

w

(

N,n

)Ω(

N

−

n,M

−

nE

S

,S

−

1) ways, wherethe factor

w

(

N,n

) =

N n

=

N

!(

N

−

n

)!

n

!

(classical)1 (quantum)accounts for the particle statistics. On the other hand, if the mode

S

is

g

S

-degenerate, the

n

particles occupyingit can be reshuﬄed in

v

(

g

S

,n

) =

(

g

S

)

n

(classical)

n

+

g

S

−

1

g

S

−

1

(bosonic)

g

S

n

1

n

≤

g

S

(fermionic)ways [14]. Hence, we have the recursive expressionΩ(

N,M,S

) =

η

(

g

S

,N

)

n

=0

w

(

N,n

)

v

(

g

S

,n

)Ω(

N

−

n,M

−

nE

S

,S

−

1)(2)where

η

(

g

S

,N

) =

N

(classical/bosonic)min(

g

S

,N

) (fermionic)

.

The boundary condition for Ω isΩ(

N,M,

0) =

δ

M,

0

v

(

g

0

,N

)

.

(3)For many values of

n

the factor Ω(

N

−

n,M

−

nE

S

,S

−

1) will be zero, because

n

will be either too high or toolow. Thus, it makes sense to derive tighter bounds forthe range of summation over

n

. For its given value

M

,the number

n

of particles occupying mode

s

cannot behigher than

Q

S

= min(

⌊

M/E

S

⌋

,η

(

g

S

,N

))(where

⌊

x

⌋

is the highest integer number less than orequal to

x

), otherwise the total excitation would exceed

M

.On the other hand, for classical particles or bosonsthe maximum

M

that

N

−

n

particles occupying modesbelow

S

can realize is (

N

−

n

)

E

S

−

1

, which leads to theinequality

M

≤

nE

S

+(

N

−

n

)

E

S

−

1

=

NE

S

−

1

+

n

(

E

S

−

E

S

−

1

)

.

(4)Since

E

S

> E

S

−

1

,

n

is greater or equal to

P

S

deﬁned as

P

S

= max

M

−

NE

S

−

1

E

S

−

E

S

−

1

,

0

(class./bosonic) (5)where

⌈

x

⌉

is the lowest integer number greater than orequal to

x

. Equation (2) is modiﬁed toΩ(

N,M,S

) =

Q

S

n

=

P

S

w

(

N,n

)

v

(

g

S

,n

)Ω(

N

−

n,M

−

nE

S

,S

−

1)

.

(6)The remaining question is the deﬁnition of

P

S

forfermions, for which it should be higher due to the Pauliexclusion principle. In principle, we can derive an in-equality similar to (4), but it would be tedious to use,as it would depend on the excitation values of multiplemodes

s < S

. However, we can safely use the

P

S

asdeﬁned for bosons and classical particles (5) when deal-ing with fermions, at the cost of summing over a largernumber of zero values.For classical particles, we can also use recursion in thenumber of particles, not in the number of modes, writingΩ(

N,M,S

) =

S

s

=0

g

s

Ω(

N

−

1

,M

−

E

s

,S

)

,

(7)with the same boundary condition as before. This for-mula holds an advantage over (6), because it does notrequire the evaluation of factorials.

2. Cumulative sum of states

We deﬁne the cumulative sum of states as the numberof states with total excitation less than or equal to

M

,Σ(

N,M,S

). To distinguish between the two quantities,the previously deﬁned sum of states Ω(

N,M,S

) can alsobe called the

marginal

sum of states. Σ(

N,M,S

) satis-ﬁes a recursive equation similar to (6), but without thelower limit

P

S

(which ensured that the total excitationconstraint (1) was satisﬁed),Σ(

N,M,S

) =

Q

S

n

=0

w

(

N,n

)

v

(

g

S

,n

)Σ(

N

−

n,M

−

nE

S

,S

−

1)

,

3with a boundary condition also somewhat diﬀerentfrom (3),Σ(

N,M,

0) =

v

(

g

0

,N

)

.

(8)Analogously to (7), we obtain for classical particlesonlyΣ(

N,M,S

) =

S

s

=0

g

s

Σ(

N

−

1

,M

−

E

s

,S

)

,

with the same boundary condition (8).In numerical computations, the recursive formulas (6)and (7) require the storage of the intermediate values of Ω or Σ. This means that the excitation values

E

s

and

M

must be discretized, which for irregular excitation valuesleads to large memory requirements. On the other hand,they do not iterate over every conﬁguration of the system,avoiding the problem of the exponential growth of theirtotal number.

B. Iterative expressions

1. Sum of states

In certain applications, it is necessary to have accessto each of the system conﬁgurations counted by the sumof states. For this purpose, the recursive formulas can beconverted to an explicit iterative expression for the sumof states:Ω(

N,M,S

) =

Q

S

n

S

=

P

S

···

Q

1

n

1

=

P

1

u

(

{

n

s

}

)

,

(9)where

n

s

=

g

s

−

1

s

′

=0

n

ss

′

,

P

s

= max

M

−

E

s

−

1

N

−

St

=

s

+1

(

E

t

−

E

s

−

1

)

n

t

E

s

−

E

s

−

1

,

0

,Q

s

= min

M

−

St

=

s

+1

E

t

n

t

E

s

,η

(

g

s

,N

−

S

t

=

s

+1

n

t

)

,

and

n

0

=

N

−

St

=1

n

t

.The factor

u

(

{

n

s

}

) =

S

s

=0

w

N

−

S

t

=

s

+1

n

s

,n

s

v

(

g

s

,n

s

)is the number of particle states realizing a given combi-nation

{

n

s

}

. Equation (9) provides a simple way to sumover all combinations of

{

n

s

}

, subject to the constraints

Ss

=0

n

s

=

N

and

Ss

=0

n

s

E

s

=

M

, and takes into ac-count the mode degeneracies. Again, the diﬀerence be-tween the formulas for bosons and fermions amounts toexcluding, in the latter case, all combinations where atleast one

n

s

> g

s

.

2. Cumulative sum of states

The diﬀerence between Ω(

N,M,S

) and Σ(

N,M,S

) isthe relaxation of constraint (1) when calculating the lat-ter. Hence, we can easily write an iterative exact formulafor it, similar to (9),Σ(

N,M,S

) =

Q

S

n

S

=0

···

Q

1

n

1

=0

u

(

{

n

s

}

)

.

The other conditions and deﬁnitions remain unchanged.

III. ANALYTIC APPROXIMATIONS FOR

Ω

Formulas for the sum of states (6), (7) and (9), thoughaccurate, are not very convenient for analytic calcula-tions. We will now derive two analytic approximationsfor Ω(

N,M,S

) (more precisely, for its version

ω

(

N,M,S

)normalized to unity), by exploiting an analogy betweena system of classical/quantum independent particles anda set of independent/correlated random variables.

A. Classical particles

When dealing with classical particles, one can associatethe excitation state of each particle with an independent,uniformly distributed random variableˆ

M

j

taking realvalues

E

0

,...,E

S

, each having a probability 1

/

Ss

=0

g

s

,so that its expected value

E

[ˆ

M

j

] =

Ss

=0

g

s

E

s

/

Ss

=0

g

s

and varianceVar[ˆ

M

j

] =

E

[(ˆ

M

j

−

E

[ˆ

M

j

])

2

] =

Ss

=0

g

s

(

E

s

−

E

[ˆ

M

j

])

2

Ss

=0

g

s

.

The total excitation

M

corresponds to the sum of thesevariables,ˆ

M

=

N j

=1

ˆ

M

j

and the probability thatˆ

M

=

M

is given by

P

(ˆ

M

=

M

) =

ω

(

N,M,S

). The varianceof ˆ

M

is thusVar[ˆ

M

] =

N

j

=1

Var[ˆ

M

j

] (10)and the mean

E

[ˆ

M

] =

N

j

=1

E

[ˆ

M

j

]

.

(11)In the case of equal mode spacings,

E

s

=

s

, and

g

s

≡

1we have

E

[ˆ

M

] =

NS

2and Var[ˆ

M

] =

NS

(

S

+ 1)12

.

The numerical examples presented below correspond tothis case, but the presented method will be applicable to

Reduced Operator Approximation for Modelling Open Quantum Systems [OSID, 22, 1550008 (2015)]

QMM_researchgateQuasi-Markov model for extrapolating observed discrete
distributions

Reading the Mind in the Eyes

Reading the Mind in the Eyes - Answers

Michael Nyman - The Heart Asks Pleasure First

Michael Nyman - The Heart Asks Pleasure First

PhysRevLett.104.106601

Chopin - Zwei Polonaises, Op.40

PhD thesis

[Accepted for publication in ICPS30 Proceedings, subm. 7/2010] Theory of spin waves in ferromagnetic (Ga,Mn)As

Loewdin calculus applied to half-ﬁlled Hubbard model

RKKY coupling in dilute magnetic semiconductors

Spintronics in semiconductors (review)

Nature A

Loewdin calculus for multiband Hamiltonians

FEL laser induced heating of the electron gas in metal

Chaotic movement of electrons in an antidot lattice subject to the magnetic field (PL)

Electron billiards, Delta magazine (04/2004) (PL)

PhysRevB.82.085204

PhysRevLett.104.106601

DMRG (PL)

- Read and print without ads
- Download to keep your version
- Edit, email or read offline

Loewdin calculus applied to half-ﬁlled Hubbard model

Spintronics in semiconductors (review)

Reduced Operator Approximation for Modelling Open Quantum Systems [OSID, 22, 1550008 (2015)]

PhysRevB.82.085204

PhD thesis

Loewdin calculus for multiband Hamiltonians

RKKY coupling in dilute magnetic semiconductors

[Accepted for publication in ICPS30 Proceedings, subm. 7/2010] Theory of spin waves in ferromagnetic (Ga,Mn)As

PhysRevLett.104.106601

FEL laser induced heating of the electron gas in metal

Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

CANCEL

OK

You've been reading!

NO, THANKS

OK

scribd