Anda di halaman 1dari 12

Home Search Collections Journals About Contact us My IOPscience

Generalized-Lorentzian Thermodynamics

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

1999 Phys. Scr. 59 204

(http://iopscience.iop.org/1402-4896/59/3/004)

View the table of contents for this issue, or go to the journal homepage for more

Download details:
IP Address: 130.183.203.14
The article was downloaded on 17/10/2011 at 04:52

Please note that terms and conditions apply.


Physica Scripta. Vol. 59, 204È214, 1999

Generalized-Lorentzian Thermodynamics
Rudolf A. Treumann
Max-Planck-Institute for extraterrestrial Physics, Postfach 1603, D-85748 Garching, Germany and International Space Science Institute, Hallerstrasse 6,
CH-3012 Bern, Switzerland

Received August 27, 1998 ; accepted October 6, 1998

PACS Ref : 05.20.-y, 05.70.Ce, 51.10. ] y, 52.25.Dg, 52.35.Ra, 52.65.Ff, 94.20.Rr

Abstract been suggested [6] that seemed to have applicability to


We extend the recently developed non-gaussian thermodynamic formalism some range of physical problems. In particular, some kind
[R.A. Treumann, Physica Scripta, 59, 19 (1999)] of a (presumably strongly of thermodynamics was constructed (for the most lucid
turbulent) non-Markovian medium to its most general form that allows for presentation see, e.g., [7]). However, this theory refers to an
the formulation of a consistent thermodynamic theory. All thermodynamic unusual prescription of constructing physical observables
functions, including the deÐnition of the temperature, are shown to be
that is not in agreement with conventional physics.
meaningful. The thermodynamic potential from which all relevant physical
information in equilibrium can be extracted, is deÐned consistently. The Our basic assumption was that the stationary turbulence
most important Ðndings are the following two : (1) The temperature is we were going to describe was not describable by weak
deÐned exactly in the same way as in classical statistical mechanics as the turbulence theory. This assumption implied that any equi-
derivative of the energy with respect to the entropy at constant volume. (2) librium state the system might have achieved could not be
Observables are deÐned in the same way as in Boltzmannian statistics as
constructed by the means of perturbation technique, i.e. no
the linear averages of the new equilibrium distribution function. This lets us
conclude that the new state is a real thermodynamic equilibrium in systems small expansion parameter exists. Systems of this kind are
capable of strong turbulence with the new distribution function replacing subject to non-perturbation techniques. In some of those
the Boltzmann distribution in such systems. We discuss the ideal gas, Ðnd systems the transition from the original state towards turbu-
the equation of state, and derive the speciÐc heat and adiabatic exponent lence proceeds through criticality. Such systems are critical
for such a gas. We also derive the new Gibbsian distribution of states.
and must be treated by renormalisation group methods (cf.,
Finally we discuss the physical reasons for the development of such states
and the observable properties of the new distribution function. e.g., [8,9]). We did not explicate about this point but instead
asked for the properties of the corresponding equilibria,
proposing that the system had actually settled into a critical
and turbulent intermediate equilibrium. We found that these
equilibrium states were describable by a new non-

P
Boltzmannian collision integral
1. Introduction dp
C \ dq T dX G [12], (1)
T dX T
In a recent paper [1] we developed the kinetic theory of a
collisionless (presumably non-Markovian) equilibrium state where p is the (turbulent) collisional cross-section, and
of a system of N particles undergoing strongly turbulent T
G [12] 4 g[ f (1@)]g[ f (2@)] [ g[ f (1)]g[ f (2)] (2)
interactions. Our motivation was to investigate what kind of T
statistical mechanics described an equilibrium state of fully is the correlation functional of the one-particle distribution
developed stationary turbulence, if it existed at all. Evidence functions f after (primed) and before (unprimed) the inter-
of the possibility of such a description comes from obser- action. G [12] itself was found to be the product of func-
T
vations of Levy Ñights and started in the twentieth of this tionals g[ f ] each depending on the distribution function of
century when Richardson found his turbulent spectral law one family of particles only. (The index T indicates that
that was fundamentally di†erent from KolmogorovÏs law of the systems are in a state of about stationary turbulence.)
ordinary spectral behaviour in turbulence (for a review of The turbulent state in that they are found is reached on
the history see, e.g., [2]). passing through critical points after having entered a highly
There had been earlier attempts to construct other than nonlinear phase. The transition to the turbulence is not
Maxwell-Boltzmann probability distributions already in the known in detail, but it can be assumed that during the criti-
past century and continued until today, mostly as attempts cal phase the system evolves into all scales such that the
to describe the thermodynamics of extreme situations [3]. scales cannot be separated anymore. It becomes essentially
These attempts culminated in a purely mathematical exten- scale-invariant, and any perturbation theory breaks down
sion of BoltzmannÏs deÐnition of entropy by Renyi [4,5]. when all scales are highly correlated. In [1] it had been
These extensions, though widely used in chaotic dynamics in demonstrated that the functionals g[ f ] actually consist of
order to infer about multi-fractal behaviour, have not been inÐnite products of correlations suggesting that this inter-
given any physical justiÐcation yet. A certain mathematical pretation is close to the truth.
simpliÐcation of RenyiÏs original proposal has subsequently The breakdown of perturbation theory in the scale-
invariant state suggests that a microscopic approach to the
ÈÈÈ problem will have to refer to renormalization group tech-
E-mail : tre=mpe.mpg.de or treumann=issi.unibe.ch niques. We have shown in [1], however, that it is possible to
Physica Scripta 59 ( Physica Scripta 1999
Generalized-L orentzian T hermodynamics 205

describe the stationary equilibrium state of the system systems, an interesting property of such states. (Note that, in
without the need to develop the microscopic phase- accord with physical intuition and requirement, S can
T, l
transition theory in detail by the methods of statistical never become sub-additive. Entropy in a single closed
mechanics. This has been done in close analogy to Bolt- system or in a collection of closed systems will always grow
zmannÏs theoretical approach. Detailed balance then as disorder cannot be extinguished by adding other dis-
requires that order.) That this is true can be shown along the same lines
as in [1], for the introduction of the Ðxed number l does not
ln g[ f ] \ [b(e [ k) (3)
p introduce any change in the analysis. Hence, f is an actual
l
where e \ p2/2m is the energy of a particle of momentum p thermodynamic equilibrium distribution that replaces the
p
and mass m, and b, k are two arbitrary constants (playing Boltzmann distribution f under i-conditions. All the dis-
B
the role of Lagrangean multipliers) which have been sug- cussion of [1] can be applied to it.
gested [1] to correspond to the kinetic temperature and In the following we develop the corresponding thermody-
chemical potential of the system, respectively. Below this namics and show that it requires the choice l \ 1. Therefore,
suggestion will be proved in full strength. the correct thermodynamic equilibrium one-particle distribu-
If the system is described by eq. (1) and is in the assumed tion function of the i-gas is given by
highly nonlinear multi-scale turbulent equilibrium, then one
can introduce [1] a control parameter i such that
A
f (e , i) \ 1 [ ] p
B
bk be ~(i`1)
, (8)
p i i
lim g[ f (i, e )] ] f (e ) (4)
i?= p B p where, for convenience, we suppressed the index l \ 1.
reproduces the Boltzmann distribution function f , and G
B T
becomes the ordinary Boltzmann collision functional. In [1] 3. Thermodynamic relations
a particular functional g[ f ] was found and it was demon-
In this section we deÐne the basic thermodynamic functions
strated that this functional actually described a thermal
and demonstrate that a i-gas is a system that though
equilibrium state nicely satisfying an H-theorem and permit-
behaving in a special way is nevertheless in complete accord
ting for a new expression for the turbulent entropy S . One
T with the fundamental thermodynamic relations.
should note that this new mathematical expression for the
entropy does not attach any new physical interpretation to 3.1. T hermodynamic potentials
the entropy. As in the Boltzmann case, entropy describes the All macroscopic thermodynamic information about a
amount of irreversible disorder in the system. The new system in contact with the external world is contained in the
expression for the entropy merely means that in the turbu- thermodynamic potential Q(b, V , k) which is a function of
lent scale-invariant state of the system the increase of dis- the temperature variable b, the volume of the system, V , and
order is calculated in a di†erent way than in the the chemical potential k (or particle number N). We deÐne
conventional Boltzmann state. Q (b, V, k) of the i-gas by
T
Q (b,V , k) \ [
P A
V d3p bk be ~i
1[ ] p
B . (9)
2. Generalisation T b h3 i i
We now generalise the functional g[ f ] to its most general Since we assume that thermodynamics should provide a
form valid description of the macrostate of the i-gas in order to
be in accord with conventional physics, the relation between
g[ f (i, e )] \ expMi[1 [ f ~1@(i`l)(i, e )]N (5)
p p the thermodynamic potential and other thermodynamic
where Mi, lN ½ R, and l is an arbitrary Ðxed real number (R functions is given by
is the space of real numbers). The advantage of introducing l
Q \ F [ kN. (10)
will become clear below. It is then easy to demonstrate that T T
the condition (4) is satisÐed for any arbitrary Ðxed l D O. Here F is the Helmholtz free energy of the i-gas, and Q is
T T
The equilibrium distribution function f (i, e ) can be con- the di†erence between F and the product of the average par-
p
structed using the two equations (3) and (5). The most ticle number and the chemical potential. Hence the free
general distribution function is found to be energy is given by
A bk be ~(i`l)
f (i, e ) \ 1 [ ] p .
B (6) F \ Q ] kN \ [
V P C
d3p bk be ~i
1[ ] p
D
] kN. (11)
l p i i T T b h3 i i
Clearly, it is a function of particle momentum p through the We now show that this deÐnition is consistent with the
particle energy e \ p2/2m and of the two parameters, i and Boltzmann limit. Indeed, taking the limit i ] O in the inte-
p
l, respectively. It is then easy to show in parallel to [1] that gral we immediately Ðnd that
f satisÐes the following new generalized turbulent entropy
l
relation referred to in the above discussion : F\[
V P d3p
exp [ [ b(e [ k)] ] kN,
P
(12)
b h3 p
d3p
S \ [k V f (e , i) ln g[ f (e , i)]. (7) which coincides with the Boltzmann-Helmholtz free energy.
T, l B h3 l p l p
Moreover, this expression identiÐes k with the chemical
This entropy is concave and moreover is super-additive, potential, and b \ 1/k T with the inverse kinetic tem-
B
meaning that the entropy of two independent systems is perature. The latter expression will be proved explicitly and
larger than the sum of the individual entropies of the two in full generality for arbitrary i below.
( Physica Scripta 1999 Physica Scripta 59
206 Rudolf A. T reumann

3.2. Average energy and number density energy F with respect to the average particle number
T
In conventional statistical mechanics, the total mean energy k \ (LF /LN) . (18)
E is calculated from the Helmholtz free energy in the follow- T bV
ing way : This equation is conventionally used to express k through
N.
L
E\ (bF) 3.3. Entropy
Lb

\[
L
V
C P d3p D
exp [ [ b(e [ k)] [ bkN . (13)
In order to Ðnd the entropy relation, we form the following
derivative of the thermodynamic potential i (LQ /Lb~1) .
B T Vk
Lb p
C D
h3 It can be shown that this derivative can be written as
LQ
In the new thermodynamics we replace F with F to obtain
T k T \ [k b[E [ kN [ Q ]
B L(1/b) B T
L Vk
E\ (bF ) 1
Lb T \ (F [ E). (19)

\[
L
V
C P A d3p bk be ~i
1[ ] p
B
[ bkN .
D (14)
T T
The last expression is just the negative of the entropy S
T
Lb h3 i i

P
and coincides with the following deÐnition
Carrying out the partial di†erentiation, this yields the d3p
S \ [k V f (e , i) ln g[ f (e , i)]
following expression for the energy T B h3 p p

E\V
P d3p e
p \ [k iV
P d3p
f (e , i)[1 [ f ~1@(i`1)(e , i)] (20)
h3 (1 [ bk/i ] be /i)i`1 B h3 p p
P
p
d3p when replacing the distribution function inside the integral
\V e f (e , i). (15) with eq. (8) and using the above derived representations for
h3 p p
the average energy (15), particle number (16), and the
Note that the second part of this equation is just the correct thermodynamic potential (9).
physical deÐnition of the average energy of the system as the Clearly this entropy expression has a structure di†erent
integral over phase space of the distribution function, eq. (8), from the ordinary Boltzmann deÐnition. It contains the
as is required by the commonly used deÐnition of an observ- logarithm of the correlation functional g[ f ] in place of the
able in statistical mechanics and kinetic theory. In this logarithm of the distribution function itself. This fact implies
respect, our theory is thus consistent with common sta- that in the i-state of the gas it is the correlations that con-
tistical physics and kinetic theory. It does not require any tribute most to the entropy. While in the Ðnal state of the
di†erent kind of averaging in order to calculate the observ- system when the interactions become purely stochastic and
ables as the physically relevant quantities. This implies that the gas settles into its thermal death, g ] f , and the
B
this theory is also in accord with the fundamental kinetic entropy assumes its classical representation.
BBGKY theory (cf., e.g., [12], chp. 2).
3.4. Consequence I : deÐnition of temperature
In order to be consistent with the previous sections, eq.
(15) suggest that we must identify l \ 1, which justiÐes our As an important application of the above deÐnition of the
previous choice for l. It would of course be possible to entropy S of the i-gas we now derive the expression for
T
choose any arbitrary l in the deÐnition of Q and to adjust the thermodynamic temperature. From classical thermody-
the exponent of f appropriately, but such an action would namics it is known that the only consistent deÐnition of the
introduce some unnecessary arbitrariness that would result temperature is given in the form of a derivative of the

A B
in a simple re-scaling of i. entropy :
The average particle number N is then given as the zeroth 1 LS
order moment of the distribution function f (e , i). This can \ . (21)
p T LE
be shown to be the negative of the partial derivative of the Vk
above thermodynamic potential Q When we use the expression (8) for the distribution function

P C D
T in the deÐnition of the entropy (20) we recover that
d3p bk be ~(i`1)
N\V 1[ ] p S \ k b(E [ kN), (22)
h3 i i T B

\[
A BLQ
T . (16)
which is nothing else than a rearranged version of (19). Then
taking the partial derivative with respect to the average
Lk energy E we immediately identify the temperature T as in
bB
conventional thermodynamics with the inverse of the Lag-
The average density of the i-gas is correspondingly
rangean multiplier b
obtained as

N
n4 \
P d3p
f (e , i). (17)
1
b
\k T .
B
(23)
V h3 p
This very important relation proves that the temperature
On the other hand, it must be required that the chemical T of the i-gas is deÐned exactly in the same manner as the
potential of the i-gas is the partial derivative of the free temperature of the classical Boltzmann gas. In this way it
Physica Scripta 59 ( Physica Scripta 1999
Generalized-L orentzian T hermodynamics 207

renders all other deÐnitions of the temperature of the i-gas 4.1. Chemical potential
used in the literature invalid. Those deÐnitions still con- We start with the average particle number N of an ideal
tained dependencies on i (cf., e.g., [13È16] and elsewhere). i-gas. This number is given as the phase-space integral over
These i-dependencies turn out to be unphysical. They are the ideal gas distribution function f (e , i) eq. (8).
the result of a naive use of the so-called experimentally
determined “i-distributionsÏ in calculating a formal expres- N\V
A B A
C(i [ 1/2) 2nmi 3@2
1[
Bp
bk ~(i~1@2)
. (26)
sion for the temperature. The physically correct distribution C(i ] 1) h2b i
function eq. (8) resembles the i-distribution, but it contains It is assumed throughout thermodynamics that the average
the non-vanishing chemical potential. It is only this com- number density n \ N/V would be known. Hence, eq. (26) is
plete distribution function that leads to a correct thermody- basically an equation for the chemical potential k of the
namics, with the temperature T being deÐned as a physical ideal i-gas. Inverting (26) and introducing the “quantum i-
quantity by its thermodynamic deÐnition, eqs (21) and (23). densityÏ
The temperature in this sense is a measure of the state of the
system. It is a parameter that characterises the gas. It is not
n \
A B
2nmi 3@2
(27)
a measure of the mean energy. Only in the Boltzmann gas qi h2b
the two, mean energy and temperature, measured in energy
we obtain for the chemical potential
units, are related in a simple way. Below we are going to
derive the relation valid in the i-gas.
The above derivation of the temperature thus justiÐes the 1[
bk C D
n C(i [ 1/2) 1@(i~1@2)
\ qi . (28)
i n C(i ] 1)
use of the constant b as the unambiguous and only measure
of the thermodynamic temperature of any (turbulent) i- The right-hand side of this expression is always positive,
system and shows that in such a system the above deÐnition and hence either the chemical potential is negative k \ 0 or
of the entropy consistently replaces the Boltzmann deÐni- bk/i \ 1. It is not too difficult to demonstrate [1] that in
tion. the limit i ] O this expression reproduces the classical
chemical potential of an ideal gas [10] with n \
q
3.5. Consequence II : Equation of State (2nm/h2b)3@2 the “quantum densityÏ of the classical gas.
In order to complete the set of thermodynamic relations we 4.2. Mean energy
may now construct the equation of state taking the partial Also, carrying out the integration in eq. (15), the average
derivative of Q with respect to the volume V . This pro- energy of the ideal i-gas follows as

LQ
T
cedure yields the pressure
A B E\
3N i
1[
Abk B
2b i [ 3/2 i
C D
P\[ T
LV

\
1 P A
d3p
bk
bk be ~i
1[ ] p .
B (24)
\
3N i n C(i [ 1/2) 1@(i~1@2)
qi
2b i [ 3/2 n C(i ] 1)
. (29)
b h3 i i
With the above identiÐcation of b \ 1/i T one immediately
B
Rewriting this expression with the help of eq. (9) we Ðnd the realises that the limit i ] O reproduces the classical well-
following important relation known result (cf., e.g., [10], p. 77)
PV \ [Q (b, V , k). (25) E
T \ 3k T (30)
This is the fundamental equation of state of a (turbulent) N 2 B
i-gas. It shows that such gases possess complicated equa- for the energy per particle. As is obvious from eq. (29), in the
tions of state. Such behaviour has been expected from the i-gas this last relation between the mean energy and the
very beginning, because the presence of the long-range temperature becomes much more involved. Moreover, it is
correlations should become manifest in the average proper- also obvious that the above deÐnition of the energy requires
ties of the gas as well. Even an ideal i-gas turns out to have that
a non-simple equation of state, as will be demonstrated
below. i [ i \ 3, (31)
min 2
The remaining Ðrst and second order thermodynamic a condition that is consistent with a similar one derived in
relations can all be obtained from the previous relations and our previous publication [1].
will not be given here.
4.3. Ideal gas equation of state
We now calculate the pressure of the ideal i-gas in order to
determine its equation of state. From eq. (25) we obtain
A B
4. Review of ideal j-gas properties
In our previous paper [1] we derived some of the relations N i bk
PV \ 1[
b i [ 3/2 i
C D
for an ideal gas. Here, because of the precise deÐnition of
the thermodynamic potential and the above consistency N i n C(i [ 1/2) 1@(i~1@2)
proof of the thermodynamic relations we can considerably \ qi . (32)
b i [ 3/2 n C(i ] 1)
simplify the expressions given before. Moreover, the formu-
las given below correct them for thermodynamic consis- This equation of state becomes the ordinary ideal gas equa-
tency. tion of state PV \ Nk T only in the Boltzmann limit
B
( Physica Scripta 1999 Physica Scripta 59
208 Rudolf A. T reumann

i ] O. The ideal i-gas behaves di†erently, possessing a the i-gas is formally identical with that of the ordinary
much more complicate equation of state. This seems reason- Boltzmann gas.
able because, as mentioned above, the fact that the i-gas is Making extensive use of the equations of state (32), energy
a highly correlated many-body system, which we assume is (29), and speciÐc heat at constant volume (35) we obtain for
in an evolved turbulent state, must in the Ðrst place become the di†erence (37) of the two speciÐc heats
obvious in its equation of state. Such a gas should exhibit a
behaviour that di†ers from that of an ordinary laminar ideal C [C \k N
i C D
n C(i ] 1) 1@(i~1@2)
. (38)
P V B i [ 3/2 n C(i [ 1/2)
gas. Nevertheless, however, it is then both surprising and qi
satisfactory that, comparing eqs (29) and (32), one recovers Inspection of this equation suggests that in a dilute i-gas
the basic thermodynamic relation between the pressure and where the density n \ N/V > n is much less than the
average energy, qi
quantum density, the di†erence between the two speciÐc
PV \ 2 E, (33) heats will become small. On the other hand, close to
3 maximum correlation (at i ] 3/2) this di†erence can
as is valid also in ordinary ideal gases. The unbroken valid- become large. These are two further distinctions between
ity of this relation even in the i-gas is interesting and is an ordinary and i-gases, respectively.
important Ðnding. Its physical meaning is that it relates the The most interesting quantity is, however, the index of
mean energy in a simple geometrical way to the pressure of adiabacity, c \ C /C . Dividing by C in (38) one Ðnds that
the i-gas. This retains the physical interpretation that the P V V
in the i-gas this ratio becomes
pressure is the geometrical e†ect of the average motion of
the constituents of the gas. The temperature assumes quite a c \1]
C
1 n C(i ] 1) 1@(i~1@2) D
i 3 n C(i [ 1/2)
GC D H
di†erent meaning then as an independent parameter mea- qi
suring the irreversibility of the system in place of its thermal n C(i [ 1/2) 1@(i~1@2) 1 ~1
mean energy. The average energy of the i-gas is thus not a ] qi [ . (39)
n !(i ] 1) 2
simple measure of the temperature T of the i-gas. The
average energy per particle E/N in a i-gas is not anymore a Indeed, for i ] O one can show that
simple fraction of the kinetic temperature. Only in the lim c \ 5 , (40)
Boltzmann gas become both, mean energy and temperature, i?= i 3
simple equivalents. which is in accord with the thermodynamics of an ordinary
(three-dimensional) gas. But, for i \ O, the adiabatic index
4.4. SpeciÐc heat and adiabatic index may deviate considerably from this value. In particular,
In order to complete the discussion of the ideal gas, we cal- when the term in the brackets containing the quantum

C D
culate the speciÐc heat C at constant volume V from density is sufficiently large compared to 1/2,

C \ [b2k
A B
LE
.
V
(34) c B1]
i
1 n !(i ] 1) 2@(i~1@2)
3 n C(i [ 1/2)
. (41)
V B Lb qi
Vk
Since E is known, it is simple matter to Ðnd This value is clearly smaller than 5/3 and in the limit of a

i A B very dilute gas approaches unity. In particular, taking i \


i \ 3/2.
AB
C \ 3k N min
V B i [ 3/2

]
GC D
n C(i [ 1/2) 1@(i~1@2) 1
qi [ .
H (35)
c B1]
3@2
n n 2
18 n
q
. (42)
n C(i ] 1) 2
This is very close to unity for small ratios n/n . For classical
q
Again, in the limit i ] O, this expression converges to the i-gases we can therefore conclude that the adiabatic index
ideal gas value (3/2)k N as the quantity in the curly falls into the interval
B
brackets becomes just
1\c Oc\ 5 (43)
i 3
lim (nm/h2b)3@(2i~1) ] 1. (36)
i?= with the right-hand side holding for the ordinary gas. On
However, as expected, C in the i-gas is not a constant (or the other hand, there is the possibility that in very dense
V i-gases with n D n the second term in eqn (39) becomes
otherwise an extensive function of the particle number). It qk
negative. In such a situation, values c \ 1 would even arise.
depends on both, density and temperature. This com- i
plication suggests that the adiabatic exponent c of the For instance, assuming n ] n , one Ðnds that
i q
i-gas will as well not be a simple constant, as is the case for J2n
lim c \ 1 [ \0 (44)
the ideal Boltzmann gas. 3@2 3J3 [ 2J2
Calculation of the speciÐc heat C at constant pressure n?nq
P
turns out to be more involved. We can, however, take becomes negative. Of course, this extreme case is physically
advantage of the general thermodynamic formula impossible, and is a result of using the minimum value for i,
and n \ n at the same time. This discussion shows,
C [C \T
(LP/Lb)
V, q
P V
(37) however, that for sufficiently high densities and small i the
(Lb/LV )
P adiabatic index becomes small. Investigation of these cases
relating C and C (cf., e.g., [10]). This is possible because requires the development of the quantum theoretical exten-
P V
we have already demonstrated that the thermodynamics of sion of the present theory. This will be given elsewhere [11].
Physica Scripta 59 ( Physica Scripta 1999
Generalized-L orentzian T hermodynamics 209

The above result on the reduction of the adiabatic index It is remarkable that these conditions exactly coincide with
is exciting. Reduction of c physically implies that the those for the isentropic processes in a classical Boltzmann
number of degrees of freedom increases. Remember that, gas. We thus learn from this agreement that isentropy
classically, c is a measure of the internal freedom of the results in simple geometric behaviour of both Boltzmann
system and is microscopically related to the degrees d of and i-gases. Since the thermodynamic di†erential equations
freedom by remain valid for the i-gas, they can in principle be used to
express these exponents through c . In a i-gas, however, the
c \ (d ] 2)/d. (45) i
ratio of speciÐc heat has no direct relation to the isentropic
Equation (43) then suggests that for i \ i \ O, the processes as is the case in an ordinary ideal gas.
min
number d increases by a large amount. For d ? 2, the index
c quickly approaches the value c ] 1. In the i-gas, c may 4.6. V olume coefficients
i
not necessarily be related to d in the same simple way (45) as In order to conclude this section we Ðnally derive the coeffi-
in the classical gas. It is possible that the relation between c cients of thermal expansion and compressibility, respec-
i
and d will be more involved. In order to obtain the actual tively. The Ðrst is given by the well-known formula
dependence, one must develop the microscopic theory of the
interactions in a multi-scale medium which is outside our a \
1 A B LV
. (51)
present reaches. Nevertheless, the decrease in c provides a ex V LT
Pk
strong argument for the validity of the initial assumption It can be easily calculated from the equation of state (32) of
underlying our theory. This assumption was that the i-gas the i-gas. Replacing the chemical potential in the Ðnal
is in a multi-scale state with an enormously large number of expression after having performed the relevant di†erentia-
degrees of freedom, and that these scales are all correlated in tions, we Ðnd that
a way as is expected for fully developed turbulence. Media
like this are believed to be scale-invariant and should micro- a \
1 C D
n C(i ] 1) 1@(i~1@2)
(52)
scopically be treated by renormalisation-group methods [9]. ex, i T n C(i [ 1/2)
qi
retains its inverse proportionality to the temperature T
4.5. Isentropic process while otherwise being small for dilute media. The thermal
compressibility
A B
With the adiabatic index at hand we can investigate the
isentropic (adiabatic) evolution of the i-gas. To this end we 1 LV
need the explicit expression for the ideal gas entropy K \[ (53)

A B
T V LP
Tk
2nim 3@2 C(i [ 3/2)C(i) can most easily be obtained from eqn (38) exploiting the
S \k V
T B h2b C(i)
G C D H
relation
1 n C(i [ 1/2) 1@(i~1@2) a
] ] qi . (46) C [ C \ T V exp . (54)
2 n C(i ] 1) P V K
T
Isentropic processes leave the entropy constant. To Ðrst This relation together with eq. (52) leads to the following
approximation this requires that useful result
V T 3@2 \ const, (47) V i [ 3/2
K \
Ti Nk T i
C D
a condition that is identical to the condition for adiabacity B
in the thermodynamics of ordinary gases (cf., e.g., [17]). For n C(i ] 1) 1@(i~1@2)
i-gases, the second term in the curly brackets in eq. (46) ] . (55)
n C(i [ 1/2)
seems to introduce a further complication. However, in the qi
particular case when the term in the square brackets on This completes our account of the stationary properties of
the right-hand side of eq. (46) is large, we still arrive at the ideal i-gas.
the surprising result that the condition eq. (47) is exactly
satisÐed. Hence, only in the intermediate regime when
5. Fluctuations
n C(i ] 1)
qi D (48) Because it is a system in thermal equilibrium the i-gas is
n 2i~1@2C(i [ 1/2)
capable of thermal Ñuctuations as well. Such Ñuctuations
the adiabatic relation between volume and temperature are known to be the root-mean-square amplitudes of the
changes. This practical independence of the condition of oscillations of the various macroscopic quantities around
isentropy on the value of i is a purely geometrical e†ect that the thermal equilibrium state. Independent of the very par-
tells that a sufficiently fast expansion of any medium should ticular form of the equilibrium distribution function the
cause cooling. It is only reasonable that this behaviour does Ñuctuation amplitudes average out when averaged over suf-
not halt in front of turbulent or scale-invariant systems. Ðciently long times or spatial scales. Under i-conditions this
With the help of (47) the two remaining isentropic relations restriction poses caution on the deÐnition of Ñuctuations.
are obtained from the equation of state as Times can be no longer than the inverse binary collision
time as it is our philosophy that the collisionless turbulent
PT ~5@2 \ const, (49)
scale-invariant state refers to times shorter than the binary
PV 5@3 \ const. (50) collision time 1/l (cf., [1]). Hence, very slow oscillations
c
( Physica Scripta 1999 Physica Scripta 59
210 Rudolf A. T reumann

may not average out even linearly and may survive the general thermodynamic expression for the energy Ñuctua-
entire scale-invariant regime until they enter the binary col- tion in order to obtain the average Ñuctuation of the inter-
lisional state when they become ultimately depleted. More- nal mean energy
over, the assumption of scale-invariance (or self-similarity)
also implies that very long oscillations may survive. With S*ET2 \
3 iA B
N(k T )2. (61)
rms 2 i [ 3/2 B
these restrictions in mind the rms Ñuctuation amplitude of a
quantity A is deÐned in the ordinary way as The rms Ñuctuation amplitude is proportional to the
S*AT2 \ SA2T [ SAT2 thermal energy k T and increases as the root of the particle
rms
(56) B
number. Its absolute value also increases with i approach-
where the average is understood as the ensemble average or ing i . The smaller i the higher is the absolute level of the
&
expectation value. Our discussion presented in the previous Ñuctuations in energy. On the other hand, when dividing the
sections has shown that calculating expectation values by rms amplitude by the mean energy eq. (29) one Ðnds that at
linear averaging over the distribution function is the appro- the same time the relative energy Ñuctuation
priate way to stay in accord with the requirement of being
S*ET2 1 i [ 3/2
able to deÐne a consistent thermodynamic theory in the i- rms P (62)
regime of the evolution of the system. Here we are interested E2 N i
in providing the expressions for the most fundamental rms decreases because the mean energy increases faster than the
Ñuctuation amplitudes. Ñuctuation. This can be easily understood as the increasing
e†ect of storage of energy in the high-energy tail of the dis-
5.1. Particle number Ñuctuations tribution near i . The relative amount of energy stored in
&
the Ñuctuations produced by the tail becomes less than the
The particle number Ñuctuation cannot be calculated
energy itself. Nevertheless, though the energy diverges faster
directly from the zeroth order moment of the distribution
than the Ñuctuation energy, the gas contains Ñuctuations of
function. Remembering, however, that the particle number
large amplitude, which is an expression of its turbulent
is closely related to the chemical potential the general ther-
nature and the appearance of many scales.
modynamic expression for the number Ñuctuation is (cf.,
e.g., [10] p. 152)
L2P
S*NT2 \ k T V . (57) 6. Properties of the equilibrium distribution
rms B Lk2
In this section we return to the equilibrium distribution
The second-order derivative of the pressure on the right- function f (e , i) and investigate its most obvious properties.
hand side of this equation can be shown [10] to be equiva- p
Because distribution functions nowadays have become mea-
lent to
C D
surable quantities (though not observables in the strict
L2P N3 LP ~1 physical sense) their shape and behaviour provide a direct
\[ . (58) link to the physical properties of the medium. Examples of
Lk2 V 3 L(V /N)
so-called i-distributions observed in space plasmas have
Using the ideal gas equation (32) for the pressure, we obtain been given continuously for about thirty years (cf., [19È23]),
for the rms particle number Ñuctuation with the most extended observations being obtained in the
EarthÏs magnetospheric tail [21,22]. Similar distributions
S*NT2 i [ 3/2
rms \ are believed to be related to Levy Ñight dynamics [2] and
N2 Ni

]
C D
n C(i ] 1) 1@(i~1@2)
. (59)
have been used in an attempt to interpret cross-Ðeld di†u-
sion [24]. But the physics of such i-distributions has not
been understood for a long time. Attempts of their explana-
n C(i [ 1/2)
qi tion trace back to various types of solutions of the Fokker-
The rms number Ñuctuation increases with N while the rela- Planck equation for particular interactions [25È27]. The
tive Ñuctuation in particle number decreases as N increases. general thermodynamic theory developed in [1] and in the
Note that if we would have used the average expressions to present paper is the Ðrst consistent physical theory of these
calculate the number Ñuctuation, we would have found a families of distribution functions, aside from another
zero Ñuctuation in number. Hence the e†ect is small in a approximate attempt [28] that was based on so-called non-
dilute gas, as it should be. Moreover, comparing the Ñuctu- extensive thermodynamics [6].
tion with the compressibility (55) one observes that the In the following we discuss the information that can be
number Ñuctuation is proportional to K as has been extracted from observation of generalised-Lorentzian dis-
Ti
expected since this is a particular case of the Ñuctuation- tribution functions.
dissipation theorem holding for small Ñuctuations (cf., e.g.,
[18], chapter 8).
6.1. V elocity distribution
The distribution function f (e , i), eq. (8) has been in terms
5.2. Energy Ñuctuations p
of the particle momentum p. It belongs to the family of
It is more interesting to determine the Ñuctuation of the generalized-Lorentzian functions. Distribution functions like
mean energy. It is most convenient to simply calculate the this one possess the property that not all inÐnitely many
two expectation values in eq. (56) with A 4 E and to take moments of the distribution can exist. For given i, the
their di†erence. It is, however, much simpler to use the moments of order r [ 2i [ 1 start diverging. This fact
Physica Scripta 59 ( Physica Scripta 1999
Generalized-L orentzian T hermodynamics 211

restricts the value of i to those values that satisfy the condi- most probable velocity v6 (momentum) of the particles. This
tion velocity is given by
i [ (r ] 1)/2. (63)
v6 \
A B C
2k T 1@2 n C(i [ 1/2) 1@(2i~1)
B qi .
D (71)
m n C(i ] 1)
For the mean energy to be a physically measurable quantity
this expression sets the limit for i [ 3/2. At the current state On the other hand, the rms speed is deÐned as
it is difficult to discuss what will happen to the higher
moments and in which way the theory can be “renormalisedÏ v \
1 4nV C P D
d3pv2f (¿, i)
1@2
. (72)
if at all. Probably, the high energy tails will break the scale- rms N h3
invariance at some stage and will close the system of
The value obtained is
A B A B C D
moments by self-limitation of the tail of the distribution
function. The more important requirement here is that f (e ,
p 3k T 1@2 i 1@2 n C(i [ 1/2) 1@(2i~1)
i) is a probability function and must therefore be positive v \ B qi . (73)
rms m i [ 3/2 n C(i ] 1)
and real. This property renders
It is not surprising that this value di†ers from v6 as this is
b/i \ 1, 0¹e \O (64)
p also true for the Maxwell-Boltzmann distribution of veloci-
ties. The ratio of both speeds is

SA B
a condition that has been shown to be satisÐed for ideal
gases and in the absence of external potential Ðelds. Other-
v6 2 i [ 3/2 1@2
wise one requires that \ . (74)
v 3 i
rms
e [ k6 [ i/b [ 0, (65)
p The most probable speed is smaller than the rms velocity.
where k6 is the total chemical potential including the external The value of these expressions is mainly that measurement
potential Ðeld. In the second case, a certain range of low of both velocities provides an independent possibility to
particle energies (or low particle velocities) is not covered by determine the value of i and, subsequently, the value of the
the distribution. This implies that for sufficiently large posi- thermodynamic temperature T .
tive total chemical potentials the moment integrals may
become non-analytic. Such cases provide enormous diffi-
6.2. Energy distribution
culties for physical interpretation. In the following we will
exclude this case from discussion. We brieÑy turn to a discussion of the energy distribution as
In terms of the velocity ¿ of the particles the isotropic this in practice is sometimes more important than the veloc-
velocity distribution function reads ity distribution itself. Similar to the Boltzmann case the

A
f (¿, i) \ 1 [ ]
B
bk mbv2 ~(i`1)
. (66)
energy distribution is deÐned as
f (e, i) \ e1@2 f (¿, i). (75)
i 2i
Since 1 [ bk/i [ 0, there exists a range of velocities for At small particle energies e it increases as f P e1@2 and
which reaches maximum at

mbv2/2i \ 1 [ bk/i. (67) be6 1 [ bk/i


\ , (76)
i 2i ] 1
In this range of velocities the velocity distribution function
and at high energies, becomes power-law and drops as
f (¿, i) D (1 [ bk/i)~(i`1) \ const, v\v (68) ei`1@2. Again, from its asymptotic behaviour one deter-
c
mines, i, while the value of the most probable energy
is practically Ñat, a property that has been frequently
increases with temperature T . The maximum value of the
observed but was barely understood so far. The present
distribution itself is
A B
theory provides a simple straightforward interpretation of
this Ñatness problem. (2im3)1@2 2i ] 1 i`1@2
On the other hand, for high velocities v ? v the distribu- f (e6 , i) \ . (77)
c m [2(i ] 1)]i`1 1 [ bk/i
tion function becomes power law
This value is always smaller than Jm3/b. Hence this value
f (¿, i) P v~2(i`1). (69) increases at about JT . Its determination allows for the
Any Ðt at high velocities with exponent a will provide a measurement of the thermodynamic temperature of the gas
measurement of i \ a/2 [ 1. The break point of the dis- and in combination with the above formula also for the
tribution is at velocity measurement of the chemical potential.

v B [2i(1 [ bk/i)/mb]1@2, (70)


b 6.3. Gases in external potentials
allowing for the determination of b \ 1/k T , i.e., the deter- An important frequently realised case is that a gas is
B
mination of the temperature. brought into an external potential Ðeld /. Then the distribu-
A general important property of the (isotropic) velocity tion becomes
distribution is that the value 4np2f (p, i) is the probability of
Ðnding any particles in the interval of momentum [p, C
bk b(e [ /) ~(i`1)
f (e [ /, i) \ 1 [ ] p .
D (78)
p ] dp]. The maximum of this expression thus deÐned the p i i
( Physica Scripta 1999 Physica Scripta 59
212 Rudolf A. T reumann

The external potential simply changes the value of the Debye-lengths becomes practically inÐnite,
chemical potential
lim
A B A
j
D, i ]
41@33nk T 3@4
B
B
. (84)
k]k]/ (79) j n h2
i?3@3 DB 0
as in an ordinary gas as well. Depending on the sign of the The e†ect of such large screening lengths on the behaviour
external potential this may have enormous consequences on of i-plasmas is of considerable interest. It may possibly
the analyticity of the distribution function, as has been men- support electric Ðelds in plasmas stronger than ordinarily
tioned in the introduction to this section. As long as this expected to exist. Also, because the level of electron thermal
Ñuctuations in Langmuir waves is proportional to 1/n j3 ,
analyticity is guaranteed, further integration of the distribu- 0 D, i
tion function provides no difficulties. Langmuir Ñuctuations in i-plasmas should be strongly
As for a Ðrst and simple application we calculate the reduced. It may be speculated that such plasmas would be
density of an electron plasma immersed into an external even less collisional and exhibit lower levels of spontaneous
electric Ðeld. In this case the average external density in the emission as well as lower levels of thermal radiation from
absence of the Ðeld is n , and / ] [ e/ in eq. (78) is under- the plasma. However, no Ðrm conclusion can be drawn at
0 this level of investigation as long as the microscopic pro-
stood as the potential energy of the electron in the electric
potential Ðeld /. Carrying out the integration of the zeroth cesses acting in i-plasmas have not been clariÐed.
order moment of 78 one Ðnds for the electron density

n(/)
\1[
e/ i [ 1/2 C 0
D
n C(i ] 1) 1@(i~1@2)
. (80)
6.4. Relativistic distribution
In this last subsection we brieÑy investigate the relativistic
n k T i n C(i [ 1/2)
0 B qi version of the Lorentzian distribution function. In rela-
This expression replaces the well-known commonly used tivistic gases, e \ mc c2, with c the relativistic Gamma-
p rel rel
Boltzmann formula factor. The phase space volume element transforms as
4nd3xp2dp \ 4nV m3c3(c2 [ 1)1@2dc , (85)
n (/) \ n exp (e//k T ). (81) 0 rel rel
B 0 B
where dV \ dV /c , p \ mc \ mc(c2 )1@2. This leads to
Clearly, this expression is the limiting form of eq. (80) for 0 rel rel rel
the following expression for the thermodynamic potential
i ] O. An exactly equivalent formula may be found, e.g.,
for gases embedded in an external gravitational Ðeld of Q\[
4nV m3c3
0
P = dc (c2 [ 1)1@2
rel rel . (86)
acceleration [g (the so-called barometric law). In order to h3b [1 [ bk/i ] (bmc2/i)c ]i
1 rel
obtain the corresponding expression one replaces With the help of this expression we can calculate the particle
e//k T ] [ mgH/k T , where H is the height above level
B B number
H \ 0, where n(H \ 0) \ n . The interesting point about
0
both these density distributions is that, in a i-gas, the
N\[
A B
LQ
density reacts algebraically to the presence of the external Lk
potential. For the barometric case this implies that it decays
less than exponentially. \ 0
bV0
4nV m3c3 = P dc (c2 [ 1)1@2
rel rel . (87)
A simple application of the density formula eqn (80) is to h3 [1 [ bk/i ] (bmc2/i)c ]i`1
1 rel
the calculation of the screening distance of an ion in a quasi- From this formula we identify the relativistic distribution
neutral i-plasma, the so-called Debye length. This length function as
may be found, expanding the right-hand side of eq. (80) for
small potentials and using PoissonÏs law. For illustration we 4nV m3c3
f (c , i) \ 0 (c2 [ 1)1@2
rel rel h3 rel
C D
show only the one-dimensional case. Then, from

d2/ e*n bk bmc2 ~(i`1)


\[ ] 1[ ] c , i [ 2. (88)
(82) i i rel
dx2 ½
0
(with ½ the vacuum dielectric constant) we Ðnd that the The restriction on i results from the requirement that the
0 relativistic energy as a moment of the distribution function
new Debye length j can be expressed through the
D, i must exist. For small but still relativistic energies, i.e., for
Boltzmann-Debye length j as

C D
DB
1 \ c \ (i/bmc2)(1 [ bk/i) (89)
i n C(i [ 1/2) 1@(i~1@2) rel
j2 \ j 2 qi . (83)
D, i DB i [ 3/2 n C(i ] 1) the relativistic distribution function varies according to
0
f (c , i) D J2(c [ 1)1@2. (90)
This value always exceeds the Boltzmann-Debye length. rel rel rel
Intuitively this e†ect is clear and very satisfactory in as far as This increase with energy reproduces the Je low energy
p
one may easily convince oneself that the presence of an increase of the non-relativistic energy distribution function
excess of electrons in the tail of the distribution function found above. On the other hand, in the ultra-relativistic
over the Maxwell-Boltzmann distribution will worsen the domain c ? 1,
rel
conditions for screening. The faster electrons are less
f (c , i) P c~i, (91)
e†ected by the potential of the test ion and, hence, this rel rel rel
potential will reach further out into space. Actually, at the implying a Ñatter decay than exhibited by the non-
absolute limiting value i \ 3/2 the ratio of the two relativistic energy distribution, the latter evolving as f (e,
&
Physica Scripta 59 ( Physica Scripta 1999
Generalized-L orentzian T hermodynamics 213

i) P e~(i`1@2). The maximum of this distribution is at Though it is obvious that it maximises S thereby leading
T
energy to the state of maximum disorder, this does not mean that

c6 \
1 [ bk/i C A
1] 1]
i ] 1 4b2m2c4 1@2
.
B D (92)
this state is achieved in a simple probabilistic way. There are
non-obvious underlying processes that speed up this dis-
rel 2bmc2 i 1 [ bk/i tribution. These processes will have to be investigated in
This is the most probable relativistic energy in a i-gas. future.

8. Conclusions
7. Discrete energy levels : Gibbsian distribution In the present paper we extended our initial theory of the
The theory developed in [1] and explicated in the former statistical mechanics of the i-state to develop the thermody-
sections of this paper assumes that the energy levels in the namics of such a state. We found that it is indeed possible
i-gas are continuously distributed. If, on the other hand, the and in asymptotic agreement with classical statistical
energy levels are discrete, the formalism must be replaced by mechanics to construct the full thermodynamics of the i-
a discrete formalism. This can be achieved by replacing the state. It was possible to deÐne the thermodynamic potential
Q that contains all the physics of the equilibrium state. One
integrals in the most fundamental formulas by sums over T
the discrete states of the system. The most important of the most important Ðndings was that the deÐnition of
replacement is to be done in the thermodynamic potential Q temperature can be retained in its full validity also in the
of eq. (9). When summing over discrete states, the volume i-state of a system. Another not less important conclusion
factor V disappears, and the restriction on momentum space was that the deÐnition of observables in the i-state is
volume elements contained in the di†erential d3p/h3 exactly the same as in classical statistical mechanics. This
becomes unnecessary. Thus, the discrete partition function fact is most satisfactory because it does not violate physical
becomes intuition. Moreover, the entropy can only grow in states of

1 N
Q \[ ; 1[ ] r
A bk bE ~i B . (93)
the kind corresponding to i-states. This is satisfactory as
well : addition of disorder does not diminish disorder, it can
T b i i only increase it. There is no process in closed systems, inde-
r/1
pendent on their internal dynamics that could minimise
Here E is the r-th discrete energy level, and the summation
r entropy. To this end one must do external work on the
is over all dynamically possible levels E . In a i-system, this
r system. The claim that instability would generate order in a
expression replaces (up to a classically unimportant factor
closed system and reduce entropy is misleading. Entropy
containing the chemical potential) the classical (Gibbsian)
can only be reduced locally on the expense of entropy
partition function
increase in other locations in the system. Instability can do
N nothing else but redistribute free energy until it makes it
Q \ ; exp ([bE ), (94)
r available to other dissipative processes.
r/1 A large number of questions has been left open for future
which is the sum over Gibbs factors, as can be shown by investigation. The most important is not that concerning the
forming the limit value for i ] O. (Note that the above
mentioned factor [b~1 exp ([bk) that remains afterwards
drops out by simple normalisation in the classical case,
while it is important to be retained in the i-system.) The
partition function (93) contains all the information about
the equilibrium state of the i-system. In analogy to our
derivation of the distribution function it also suggests that
the most probable Gibbs distribution of discrete states can
be written as
A
bk bE ~(i`1)
w \ 1[ ] r .
B (95)
r i i
Here w is a probability (or else the most probable
r
occupation number of state r). Accordingly, the entropy of
the discrete system becomes
N
S \ [k i ; w (1 [ w~1@(i`1)), (96) Fig. 1. Schematic evolution of 1/i for a di†erent case. The solid line refers
T B r r to the model used in the present paper. i is zero at times t \ t and t [
r/1 nl
t \ //nu , respectively. There are two critical transitions in this case where
an expression that is of similar though not identical form as c c
i jumps from zero to a Ðnite value i [ i at these transitions (the Ðgure
the one given in [6]. All thermodynamic relations apply to &
shows the marginal case i \ i ). The dotted line shows a possible non-
&
Q . It is important to note that there is no simple probabil- critical transition where i D 0 already in the interval t \ t \ t . Such a
T l nl
transition may avoid criticality and would be describable by ordinary
istic way of distributing N particles on arbitrary states that
strong turbulence theory (nonlinear wave-wave and wave-particle
would reproduce the above Lorentz-Gibbsian probability
interactions). In addition, i can depend on time (or otherwise temperature
distribution. The occupation number w is not a simple most
r if accounting for transitional heating in turbulence. Then either the cases of
probable probability distribution in the ordinary meaning of the dotted (retaining a late critical phase transition) or thin solid lines (no
the word. It cannot be achieved at by throwing the dice. Ðnal transition) apply, respectively.

( Physica Scripta 1999 Physica Scripta 59


214 Rudolf A. T reumann

reality of processes like the once discussed here. Obser- We will have to develop the non-equilibrium thermodyna-
vation of Levy Ñights and increasing frequency of measure- mics for i-systems before these questions can be answered.
ment of so-called “i-distributionsÏ (after those we have This goal in addition requires to solve the problem of diver-
modelled our distribution function by the use of the control gencies appearing in the higher moments of the distribution
parameter i) in large collisionless systems like the near- function.
Earth space environment provide sufficient justiÐcation for
an expedition into basic physics. But the question remains
of what is the physical relevance of the index i ? In an inter- Acknowledgements
esting paper [25], Hasegawa and co-workers found a It is a great pleasure to thank J. Geiss, B. Hultqvist (both ISSI) and G.
similar distribution function solving a Fokker-Planck equa- MorÐll (MPE) for their continuous interest and moral support during the
tion under the assumption of a very particular weakly turb- period of writing this report. Discussions with T. Chang, J. Geiss, M.
ulent interaction in a plasma. There, i turned out to depend Hoshino, A. Kull, G. MorÐll, M. Scholer, J. Scudder, and T. Terasawa are
on the dispersion of plasma waves. deeply acknowledged. Most of this work has been performed at the ISSI
Bern, after it had originally been initiated during a visiting professorship at
A solution of this kind may remain a particular case of STEL, Nagoya University, Japan. The hospitality of Y. Kamide and S.
weak turbulence. It rather seems that the i state of a system Kokubun at STEL, Toyokawa, as well as of Nagoya University is also
refers to an intermediate strongly turbulent state when the gratefully acknowledged.
system has settled into quasi-equilibrium for some time
before binary collisions set on to destroy the state and to
dissipate the energy stored in the turbulent motion, as has References
been argued in [1]. Equilibria usually imply the dissipative
1. Treumann, R. A., Physica Scripta, 59, 19 (1998).
destruction of the free energy. In i-equilibrium it seems that 2. Shlesinger, M. F., Zaslavsky, G. M. and Klafter, J., Nature 363, 31
this is not necessarily the case. But the observation that i (1993).
appears only in this intermediate state, suggests that there 3. Lavenda, B. H., ““Thermodynamics of ExtremesÏÏ (Albion Publ., West-
must be discontinuous transitions between the initial and ergate, Chichester 1995).
4. Balatoni, J. and Renyi, A., Pub. Math. Inst. Hungarian Acad. Sci. 1, 9
the i-state and possibly also to the Boltzmann regime.
(1956).
Figure 1 schematically illustrates this behaviour of i for 5. Renyi, A., ““Probability TheoryÏÏ (North Holland, Amsterdam 1970).
the same scenario as in [1]. In this Ðgure we plotted the 6. Tsallis, C., J. Stat. Phys. 52, 479 (1988).
value of i~1. This value must be zero in the respective linear 7. Boghosian, B. M., Phys. Rev. E53, 4754 (1996).
and in the Ðnal Boltzmann regimes, because these regimes 8. Wilson, K. G., Rev. Mod. Phys. 55, 583 (1983).
9. Chang, T., Vvedensky, D. D. and Nicoll, J. F., Phys. Rep. 217, 279
are classical (i ] O). In the turbulent quasi-equilibrium
(1992).
state i~1 suddenly assumes a value between zero and 1/i
& 10. Huang, K., ““Statistical MechanicsÏÏ (Wiley, New York 1987).
(the Ðgure shows the marginal case i \ const \ i ). The 11. Treumann, R. A., ““Lorentzian quantum statistical mechanicsÏÏ, in
&
sudden increase from i~1 \ 0 to i~1 D 0 is understood as a preparation.
critical phase transition. A possible non-critical transition is 12. Libo†, R. L., ““Introduction to the Theory of Kinetic EquationsÏÏ
(Krieger, Huntington, N.Y. 1979).
shown as the dotted line in the non-linear violent relaxation
13. Meyer-Vernet, N. and Perche, C., J. Geophys. Res. 94, 2405 (1989).
regime where our theory does not apply. 14. Summers, D. A. and Thorne, R. M., Phys. Fluids B3, 1835 (1991).
Physically spoken, there is no need for i being constant 15. Mace, R. L. and Hellberg, M. A., Phys. Plasmas 2, 2098 (1995).
throughout the entire quasi-equilibrium. If i changes suffi- 16. Mace, R. L., Hellberg, M. A. and Treumann, R. A., J. Plasma Phys. 59,
ciently slowly compared to the length of this quasi- 393 (1998).
17. Kittel, C. and Kroemer, H., ““Thermal PhysicsÏÏ (Freeman, New York,
equilibrium the equilibrium theory can still be applied.
1980).
There is reason to believe that such a situation is closer to 18. de Groot, S. R. and Mazur, P., ““Non-equilibrium ThermodynamicsÏÏ
reality than i \ const. Dissipative non-collisional irrevers- (Dover, New York 1984).
ible processes will necessarily accompany the turbulent 19. Vasyliunas, V. M. J. Geophys. Res. 73, 2839 (1968).
state. Because the heating causes T to increase with time, 20. Scudder, J. D. and Olbert, S., J. Geophys. Res. 84, 6603 (1979).
21. Christon, S. P., et al., J. Geophys. Res. 93, 2562 (1988).
i(T ) will itself become a function of time. Two such cases
22. Christon, S. P., et al., J. Geophys. Res. 96, 1 (1991).
have been included in Fig. 1. In the Ðrst one (thin solid line), 23. Lin, R. P., et al., Geophys. Res. Lett. 23, 1211 (1995).
the Ðnal critical phase transition does entirely disappear. 24. Treumann, R. A., Geophys. Res. Lett. 24, 1727 (1997).
i(T ) tends to inÐnity right at the end of the intermediate 25. Hasegawa, A., Mima, K. and Duong-van, M., Phys. Rev. Lett. 54,
stationary state. In the second case (dotted line), i remains 2608 (1985).
26. Collier, M. R., Geophys. Res. Lett. 22, 303, 2673 (1995).
Ðnite towards the end of the quasi-equilibrium, and another
27. Reynolds, M. A., Fried, B. D. and Morales, G. J., Phys. Plasmas 4,
critical phase transition is needed when suddenly binary col- 1286 (1998).
lisions take over. Which one of these cases will be realised, 28. Tsallis, C., Levy, S. V. F., Sousa, A. M. C. and Maynard, R., Phys.
depends on the properties of the system during its evolution. Rev. Lett. 75, 3589 (1995).

Physica Scripta 59 ( Physica Scripta 1999

Anda mungkin juga menyukai