• Aucun résultat trouvé

Theoretical aspects of magnetic fluctuations

N/A
N/A
Protected

Academic year: 2021

Partager "Theoretical aspects of magnetic fluctuations"

Copied!
7
0
0

Texte intégral

(1)

HAL Id: jpa-00243378

https://hal.archives-ouvertes.fr/jpa-00243378

Submitted on 1 Jan 1970

HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

Theoretical aspects of magnetic fluctuations

Ronald E. Burgess

To cite this version:

Ronald E. Burgess. Theoretical aspects of magnetic fluctuations. Revue de Physique Appliquée,

Société française de physique / EDP, 1970, 5 (1), pp.77-82. �10.1051/rphysap:019700050107700�. �jpa-

00243378�

(2)

THEORETICAL ASPECTS OF MAGNETIC FLUCTUATIONS

By RONALD E. BURGESS,

Department of Physics, University of British Columbia, Vancouver 8, B.C., Canada.

Abstract.

2014

The thermodynamic treatment of equilibrium fluctuations in magnetic systems

is discussed and applied to

a

variety of situations. The spectral density of the magnetic

moment or flux are derived for simply-connected bodies (sphere and cylinder) and for multiply-

connected systems (loop and hollow cylinder). The material is considered to be either

a

normal conductor,

a

superconductor or ferromagnetic. The results are valid in the presence of

a

static

magnetic field and also if the susceptibility is field-dependent. No abrupt change in fluctua-

tions occurs as

a

superconductor is taken through the transition temperature. The introduc-

tion of

a

Josephson junction or

a

resistive link in

a

superconducting loop is shown to increase the flux fluctuations markedly with

a

steady increase as the link is made weaker. At high fre- quencies the fluctuation spectrum for

a

superconductor approaches that for the same system

in the normal state.

PHYSIQUE APPLIQUEZ 5, 1970,

I. Introduction.

-

Fluctuations of the magnetic

moment of a solid body of arbitrary shape can be

calculated for thermal equilibrium by means of the fluctuation-dissipation theorem. Similarly the flllc- tuations of flux and magnetic field within a loop or

other hollow body may be found.

Attention is hère devoted to various configurations

of normal conductors, ferromagnetics and supercon- ductors. The results are valid in the presence of an

arbitrary static magnetic field and the field-dependence

of magnetization (which may be non-linear) enters through the differential susceptibility which is in gene- ral a complex function of frequency.

The inapplicability of the simple form of the equi- partition principle commonly used to evaluate magne- tic fluctuations in superconductors arises from the

neglect of significant components of energy additional

to the magnetic field energy.

A superconducting system customarily exhibits appreciably smaller magnetic fluctuations than a nor-

mal metal system of the same configuration and has

a smaller relaxation time. Nevertheless the fluctua- tions are continuous through the transition température

as is the magnetic polarizability.

The total mean square fluctuation (integrated spec-

trum) of the magnetic moment or flux is simply related

to the polarizabilities at 0)

=

0 and oo and to the

response of the system to an infinitesimal step function

of the external field.

In dealing with superconductors it proves to be

helpful to represent the inertia of the super and normal fluid components in terms of the kinetic inductances.

Similarly the behavior of a Josephson junction can be represented by an effective inductance for small phase

fluctuations. These equivalences are adopted here to provide uniformity of notation.

In practice the fluctuations of relevance may either be the integrated spectrum (in wideband systems) or

the low-frequency portion of the spectrum (in narrow

band systems). Special attention is therefore devoted

to these limiting forms. Depending on the configura-

tion either the fluctuations of magnetic moment (for

a solid body) or of magnetic flux (for a loop or hollow body) may be of primary interest.

Throughout the sinusoidal response functions are in terms of the function exp (icôt) and parameters are expressed in e.m.u.

Spectral densities for fluctuations are defined such that the variance of a quantity is given by the integral :

II. General Relations.

-

We first note some general

relations connecting the spectrum of fluctuations and the integrated variance with the response function of the system to an external magnetic field.

If the magnetic moment of a body is m exp (Í6Jt) in

response to a small sinusoidal magnetic field H exp (Í6Jt)

their ratio is denoted by m/H

=

rl(6J) where ce(m) is

the complex polarizability. This function is analytic

in the lower half of the 6J-plane and it follows by

contour integration that :

where oc

=

a’

-

ia". The dissipative part ce"(ce) is

a non-negative odd function of frequency which va-

nishes at ce

=

0 and cc.

From the fluctuation-dissipation theorem (Landau

and Lifshitz, 1958), it may be deduced that the ma-

gnetic moment fluctuation spectrum is :

If the significant frequencies in the dispersive beha-

vior of (X( (0) lie below kT/~, the spectrum simplifies to

the classical form :

The total variance of the moment fluctuations inte- grates to :

Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/rphysap:019700050107700

(3)

78

Now the function ce(m) is related to the response function for an abrupt change of magnetic field. If

the applied magnetic field is increased by an infini-

tesimal step 8H at t

=

0 and 3m(t) is the resulting

time variation of the magnetic moment, the incre- mental response function is defined as :

This is in general a tensor but we shall only consider

cases where it is a scalar by making the coordinate

axes coincide with the principal axes of the system.

The Fourier transform relation connecting r(t) and et. ( (0) is :

The prompt and the final steady state response functions are related to ce(ce) by :

The variance of a Cartesian component of the ma- gnetic moment is thus :

where r(t) is function applicable to the component of the moment considered. This result can be further

generalized to give the autocorrelation function for the moment component :

For a normal conductor r( oo) = 0 if [J.

=

1 and

r( oo) > 0 if y > 1 while for a superconductor r( oo) 0 corresponding to diamagnetic behavior.

For all systems r(0+) 0 and is simply related to

the geometry of the system.

One situation of practical interest is that of a long

solenoid enclosing the magnetic body which can be

of arbitrary shape. The magnetic moment of the body m in the direction of the solenoid axis sets up in the solenoid a flux :

where no is the number of turns per unit length of the

solenoid. This flux is independent of the dimensions of the solenoid so long as its length is large compared

with its radius.

The observable would be the EMF induced in the solenoid by variations of the magnetic moment. If

the body is in thermal equilibrium at temperature T the spectrum of the induced EMF is :

This is just (kT/03C0) times the resistance induced in the solenoid by the presence of the body, as is indeed required by the general thermodynamic validity of Nyquist’s theorem.

III. Simply-Connected System.

-

III.1. SPHERE.

-

The sphere exemplifies the case of a simply-connec-

ted body with an isotropic demagnetization coefficient

so the polarizability tensor is particularly simple.

A sphere of radius a, normal conductivity a and permeability [L = 1 has polarizability :

where z

=

Ka

=

a(47ti(ù(j)1/2 and K is the wave

constant.

Thus oc 0

=

0 and cc(oo) == - 1 a3 gives :

The low-frequency moment spectrum is :

If however the sphere is ferromagnetic with permea-

bility p,., these results become modified to :

where f ( z) = z-1 coth z

-

z- 2 and now :

The limiting values are :

As expected the final steady state response :

is not zero but has the positive value appropriate to

a polarizable sphere of permeability f1. taking account

of the demagnetization effect. The variance and the low frequency spectrum of the magnetic moment are

greater than those for a non-ferromagnetic sphere :

We discuss the variance of the magnetic moment

instead of its mean square value since for a ferromagne-

tic system there may be some steady magnetization

upon which the fluctuations are superimposed. The permeability 03BC here must be interpreted at the static

differential permeability (dB/dH) taken at the value

of the steady magnetization.

Some ambiguous results can arise if the sphere

with 03BC > 1 be assumed to have zero conductivity

i.e. an ideal ferromagnet. The value of x(0) is unchanged but the evaluation of z(oo) and hence

of tl( (0) for zero conductivity is indeterminate. In this situation and others of similar character it must

be recognized that the conductivity of a real ferro- magnet may be very small but positive for eu > 0.

Thus in evaluatingf( (0) to obtain ce (0) we let m - aJ

and 03C3 ~ 0+ which again leads to the value found above for oc ( oo ) .

A further factor is that the permeability itself will

in general be a complex function of frequency due to

the inevitable inertia of the spin motion in a varying magnetic field. Thus in evaluating ce oo) it is relevant that not only is f (z) frequency dependent but so also

is [.L. These considerations are without effect if f ( 00 )

is taken as 0, yielding again the above expression for

var m even when there is dispersion of 03BC.

It is important to remember that the conductivity 03C3(03C9) is complex and bounded at high frequency. For

a normal conductor in which the electrons can be

considered to have a momentum relaxation time Tn :

(4)

where nn is the (normal) electron density. Thus as

(ù -+ oo the wave constant :

where 11n

=

m/nne2.

Thus K and z do not tend to infinity in this limit,

and the corresponding characteristic length :

for a typical metal is about 2 X 10-6 cm. Thus except for very small particles we may take z(co) > 1

as m - oo and thus f( (0)

=

0 as assumed above.

This limit is equally valid for a ferromagnetic con-

ductor.

If however our concern is with very small spheres

the earlier expressions for the variance of the magnetic

moment must be modified to take account of the finite value of K( oo) and becomes :

In discussing superconducting bodies we shall use

the approximation of London (London, 1961).

The results for a superconducting sphere can then be

derived from those of a normal conductor of the same

configuration by a simple procedure using the two-

fluid approach.

The wave constant K used above for steady-state periodic excitation at frequency m in a normal conduc-

tor with unity permeability is :

For a superconductor we modify this to include the contributions of both the super and normal fluids :

where 11s

=

m/ns e2 with ns

=

superfluid number den-

sityand À is the London penetration depth

=

(As/41t)1/2.

The limiting values of the wave constant K(to) are

then :

where n = ns --+- nn and A -1

=

As’ + An’.

The essential difference between the normal and

superconducting spheres is that for the latter z(O) is

not zero but 2/À.

Thus in calculating the low frequency spectrum

care is needed in approximating since z(w) is not necessarily small (as for the normal sphère). Indeed

if the radius of the sphere is large compared with the

London penetration depth :

so the low-frequency moment spectrum is :

which is smaller by the factor (45/2) (03BB/a)3 than the

low frequency moment fluctuations in the corres-

ponding normal sphere.

If a/03BB is small compared with unity we recover the

result found for the normal sphere at low frequency :

The presence of the superfluid does not therefore

manifest itselfin the magnetic fluctuations for particles

small compared with h.

Since the penetration depth 03BB increases as T in-

creases to Tc, it is seen that the fluctuation spectrum increases rapidly and steadily with temperature and joins smoothly to the value for a normal conductor

at T

=

T,. This is a general inference for an arbi- trary shape of the conductor since K((ù) varies conti- nuously with temperature through T,.

The integrated spectrum of the moment fluctuations in a superconducting sphere also has two limiting

forms depending on whether the radius is large or

small compared with À.

For a » h the moment variance is smaller by the

factor 3(h - h’) la compared with the same sphere

in the normal state :

where 03BB’

_

(A/4n) . For a ~ 03BB’ :

which is determined by the normal fluid and corres-

ponds to the result for the normal sphere with p.

=

1.

It is a general result that for superconducting bodies

with dimensions small compared with À the relevant electron density is that of the normal fluid in deter-

mining the magnetic moment fluctuations.

III.2. LoVG CYLINDER.

-

A long cylinder of a

normal conductor with unity permeability has pola- rizability in the axial direction given by :

where a is the radius and 1 the length of the cylinder

and z

=

a( 4ni(j(ù) 1/2 as for the sphere.

Thus (x(0)=0 and oc ( oo) _ - a2 lj4 for the cylinder radius large compared with the skin depth :

The low frequency spectrum of the moment fluc- tuations is :

If the cylinder is ferromagnetic the polarizability

becomes :

where the first term represents the induced magnetiza-

tion and the second term the diamagnetic effect of the

eddy currents.

The variance and low frequency spectrum are now :

where fL === fL’ - ifL".

If the cylinder is superconducting we make the same

modification to the wave constant K as we did for the sphere. For a » h the variance of the magnetic

moment is :

(5)

80

which is 2 (h

-

À’) la times the variance for the cylinder

in the normal state whereas for a radius small compared

with the London penetration depth :

A generalization of the moment variance for a solid

superconducting body of arbitrary shape and having

all dimensions large compared with X is :

where S. is the surface area of the body and D is the demagnetization coefficient in the direction of the observed component of the moment (e. g. for a cylin-

der D

=

0, for a sphere D

=

1/3).

III .3. MEDIUM COMPOSED OF A LATTICE OF MAGNE-

TIC PARTICLES.

-

Here we consider a medium with

macroscopic polarizable particles distributed in a cubic array and for simplicity of isotropy we consider the

particles to be spherical. The magnitude and orien-

tation of the particle moment is determined by the

local field.

The relation between the local field H1 and magne- tization M assumed to be given by the Lorentz relation :

where Ho is the external field.

If 03B11(03C9) is the polarizability of each particle and

n is the particle density :

A long cylindrical sample of the medium (of vo-

lume V) thus has polarizability :

Hence if the particles have static differential per-

meability lL :

where F

==

4na3/3 is the particle filling factor.

The variance of the axial magnetic moment of the cylinder is : -.

For a given filling factor F the variance increases

by the factor 3/ (1 - F) as p. goes from 1 to cc. It may be inferred that the filling factor is of more

significance than the particle permeability in deter- mining the fluctuations.

IV. Ring or Hollow Cylinder.

-

IV.1. NORMAL

CONDUCTOR. - The elementary multiply-connected

systems of interest are the planar loop and the long

hollow cylinder. In both cases the magnetic pro-

perties are simple if the thickness of the conductor is small compared with the other dimensions and the field penetration depth.

We first consider a planar loop of normal conductor with impedance Z(co) and area A. Then the pola- rizability in the direction normal to the plane of the loop is :

The fluctuation spectra of the loop flux (D is :

where L is the inductance of the loop.

We shall express the results for hollow systems in

terms of the flux spectrum since this would usually be

the observable of most direct interest.

Now the impedance is composed of the magnetic

and the electron contributions :

where Ln

=

(mlnn e2) (piS) and Rn = Ln/Tn while p

and S are the perimeter and cross-sectional area of the loop conductor. Provided that this representation

is valid up to frequencies of the order of Rn/L so that

skin effect and radiation impedance are not important

the fluctuation of the flux is :

and the-mean total energy is :

,

which signifies that the magnetic energy and the electron kinetic energy (which both depend on 12) together correspond to one degree of freedom in formulating the equipartition result.

In practice Ln will customarily be very small com-

pared with L. Thus for a circular loop of radius a

and circular cross-section radius b :

which will be negligible for a metallic loop unless b

is less than about 10-6 cm.

The long thin hollow cylinder is very similar to the

loop for equilibrium fluctuation properties. The flux

fluctuation spectrum is still (kTL2j1t) Re (11Z) where

Z is now the impedance for a current sheet with peri- meter p = 2a and S = 1 d :

where a is the radius of the cylinder, l its length and

d the wall thickness which is assumed to be small.

Accordingly the mean square flux fluctuation is

k TL2/ (L + L.) as for the loop. The ratio LJL of

kinetic to magnetic inductance is :

If this is small compared with unity the mean square axial magnetic field at any point within the hollow

region is given by : -.

which is the value corresponding to equating the ave-

rage field energy to (1/2)kT.

IV. 2. SUPERCONDUCTOR.

-

For a superconducting loop or hollow cylinder the impedance Z(cù) can be

written in the first approximation for the two-fluid model :

where Lg

=

(mlns e2) (piS) is the effective inductance associated with the kinetic energy of the superfluid component while Rn

=

L.1-r. is the resistance asso-

ciated with the normal fluid. Here Tn is the momen-

tum relaxation time of the normal electrons.

(6)

The spectrum and variance of the flux fluctuations

are :

where Q

=

L(LS + Ln) + Ls Ln.

It is seen that the flux fluctuation for a given loop

or hollow cylinder is much smaller in the superconduc- ting state than in the normal state if L is large compared

with L, and Ln. Then for the superconducting loop :

Nevertheless it must be stressed that the fluctuations do not undergo a discontinuous change at the tran-

sition temperature, since as this temperature is approa- ched from below n, tends to zero, Ls tends to infinity

and (D2 uniformly tends to the value kTL2/(L + Ln)

for the normal state.

The reason for the large decrease of D2 below Tc is

that the kinetic energy terms (1/2)Ln I2n and (1/2)Ls I2s

for the normal and superfluids considerably exceed

the magnetic term ( 1 /2) L12

=

03A62/2L. From the two-

fluid model it is readily found that the energy ratios

are :

and :

The field fluctuations inside a hollow superconduc- ting cylinder thus decrease rapidly as the temperature is reduced; for instance in going from t

==

0.9 to

t

==

0.5 the RMS field is reduced approximately

sevenfold (t

=

T/Tc).

It should be remarked that the equations derived

here incorporate the conservation and quantization

of fluxoid in the form (D + L, I,

=

const.

=

Nhl2e.

Quantization is not however relevant in the discussion of fluctuations unless one is concerned with transitions from one fluxoid state to another but this topic lies

outside the scope of the present paper.

IV. 3. SUPERCONDUCTING LOOP WITH JOSEPHSON JUNCTION.

-

Since this problem has been discussed earlier (Burgess, 1967, 1968), it will be outlined here and the essential results presented. The Josephson junction can be conveniently represented as a parallel

combination of its capacitance C, the conductance G due to dielectric loss and quasi-particle tunneling and

thé effective inductance Li

=

n/2eI1 representing the Josephson relations between voltage, phase cp and the

junction current 7j whose maximum or critical value is 11.

The conservation and quantization conditions are :

where IJ is the Josephson tunneling current in the junction. If the system is in thermal equilibrium the

fluctuation sources are given by the Nyquist relation applied to the Rn of the normal fluid in the super-

conductor and to the conductance G of the junction.

The spectral density of the flux fluctuations is then :

The mean square flux fluctuation is thence :

which is always greater than the flux fluctuation in the superconducting loop without a junction.

If L is large compared with Zg and Ln this simplifies

to :

If the junction conductance G is not too great the system can be approximated by the loop inductance L connected to the junction reactance composed of C

and L J in parallel. Then the fluctuation spectrum of the current has its maximum at the frequency :

and has its minimum at the (slightly) lower frequency equal to Josephson’s "plasma frequency" :

Thus the spectrum has its most conspicuous features

of a minimum at 0153j and a maximum at 03C9J(1 -p LJIL) 1/2

in a relatively narrow range of frequency.

For an extremely weak junction with very small Il

such that Li exceeded L the flux fluctuations approach

the value for a superconducting loop with a resistive junction discussed in the next section. If the Josephson junction is not too weak so that L is large compared

with Li we must revert to the unapproximated (D2

where we see that Lj, Ls and Ln are competing for

control of the current fluctuations which are then very much smaller than kT/L.

IV.4. SUPERCONDUCTING LOOP WITH RESISTIVE

JUNCTION.

-

If a superconducting loop is interrupted by a junction (e. g. of dielectric or normal metal)

which cannot support a supercurrent the current fluctuations are markedly enhanced.

First we consider a narrow resistive region of resis-

tance R which is too wide however to permit tunneling

of a supercurrent. In this case we do not invoke the

Josephson conditions and no quantization condition

for flux and junction phase is applicable. The system behaves like a simply-connected superconductor to

which the potential difference

-

IR is applied. The

two-fluid model is used for the superconductor to treat

time varying currents. There are two fluctuation

sources

-

the resistance Rn associated with the normal

(7)

82

fluid in the superconductor and the R of the resistive

junction.

In terms of the notation used previously :

This spectrum as well as the response of the system

to an abrupt change of external magnetic field corres- ponds to two relaxation times 03C41 and T2 whose recipro-

cals are the roots 81 and S2 of the quadratic equation : S2 Q - 8[Rn(L + LS) + R(Ln + Ls)] + RRn

=

0.

These relaxation times will be approximately (L + Ls) IR and (Ln + Ls) I Rn if Ls is small compared

with L or Ln that is for temperatures not too close to Tc.

On the other hand the superconducting loop without

a junction has a single relaxation time :

which is much smaller than the value (Ln + L)/Rn

for the normal loop but always larger than in

=

Ln/Rn.

The relaxation time is continuous at T,.

The flux fluctuation is :

which is very close to that for the normal loop. Thus

when only a very small segment of a superconducting loop is resistive the flux fluctuations are greatly enhan-

ced over the value for a completely superconducting loop and approach the limiting value kTL2J(L + Ln) .

No discontinuity occurs in the flux fluctuations when the temperature is taken through Tc because :

is a constant and is equal to the value of Ln when the loop is in the normal state.

The results for the loop presented above show that there is a steady increase in the flux fluctuations pro-

ceeding in the sequence : superconducting loop

~

superconducting loop with Josephson junction

~

superconducting loop with resistive junction

~

nor- mal loop. The popularly cited value of k T L for the flux variance is a good approximation for the last two

of these situations but a gross overestimate for the first two situations. This results from the inhibiting

influence of fluxoid conservation on flux fluctuations in the superconducting loop with or without a Joseph-

son junction. With a resistive junction there is no

fluxoid conservation or quantization. One criterion

for the existence of fluxoid conservation in a loop is

that the limit of iú) 1 z (ú)) as 03C9 ~ 0 is non-zero.

REFERENCES BURGESS (R. E.), Quantization and Fluctuations in Su-

perconductors. Proc. Symposium on Physics of Super- conducting Devices, Charlottesville, 1967, pp. H1-H17.

BURGESS (R. E.), Thermodynamic Fluctuations in Super-

conductors. Proc. Conference

on

Fluctuations in Su-

perconductors, Asilomar, 1968, pp. 47-79.

LANDAU (L. D.) and LIFSHITZ (E. M.), Statistical Physics (Pergamon, 1958), sections 119-122.

LONDON (F.), Superfluids, 1, Macroscopic Theory of

Superconductivity (Dover, 1961), chapter B.

Références

Documents relatifs

The use of our new cost function has allowed to increase the performances of the pre-network as compared to the standard mean squared error : the false alarm is reduced by 20 % (on

In the open circuit condition, the air-gap flux varies as a function of the field current with minimum and max- imum values exist in the flux weakening and flux en- hancing

ءاصحلإا بولسلأا سفن اتمدختسا انهوك في تيسارد عم ةساردلا هذى يقتلت ( يزييمتلا ليلحتلا بولسأ - ةوطبخ ةوطخ - ) نافلتبزو ،جذومنلا ءانب في ةمدختسلدا

Here we consider a Josephson junction consisting of two non-centrosymmetric su- perconductors (NCSs) connected by a uniaxial ferromagnet, and demonstrate that it exhibits a

Herschel and SOFIA observations have shown that it is best characterized through the far- IR fine structure lines of ionized carbon, [CII], oxygen, [OI], atomic carbon, [CI],

The comparison shows that the exploration of the program’s design space – as distinct from transforming source code to execute design moves – is a critical activity in redesign.. J

A possible solution to this problem might be to raise yearly medical tuition to the level of the cost in the United States ($50 000 per year) with the off er of a

Does circumcision reduce the risk of penile human papillomavirus (HPV) infection in a man and cervical cancer in his female partner.. Type of article and design