Theoretical and Mathematical Physics, 167(2): 654667 (2011)

PHASE TRANSITIONS IN REAL GASES AND IDEAL BOSE GASES

V. P. Maslov

Based on number theory, we present a new concept of gas without the particle interaction taken into
account in which there are rst-order phase transitions for T < Tcr on isotherms. We present formulas for
new ideal gases, solving the Gibbs paradox, and also formulas for the transition to real gases based on the
concept of the Zeno line.
Keywords: rst-order phase transition, phase transition of the second kind, Einstein paradox, gas
mixture, cluster, ideal gas, Bose gas
1. Introduction
This paper is a continuation of [1].
According to standard courses of thermodynamics, the ideal gas is a gas satisfying the relation
PV = RT, (1)
where P is the pressure, V is the volume, T is the temperature, and R is the gas constant. In other words,
PV
RT
= 1. (2)
The quantity PV/RT is called the compressibility factor and is denoted by the number Z. This relation is
also well satised for real gases and mixtures of gases under room conditions. And why is the ratio PV/RT
namely unity? This value was adopted for convenience, appropriately choosing the gas constant.
The Boltzmann distribution leads to this expression for any gases. But the so-called Gibbs paradox
arises here. If the answer obtained from the Boltzmann distribution is divided by 1/N!, where N is the
number of particles, then the Gibbs paradox in its original statement can be resolved for any gas consisting
of molecules of the same type. We must note here that the particles of this gas are indistinguishable.
Relation (1) still holds for a mixture of gases, but dividing by 1/N! is already clearly impossible because
the gases are distinguishable.
And what happens if the gases are dierent isotopes of the same matter? If two mixed gases are
absolutely identical, then we must divide by 1/(N
1
+ N
2
)!. And what must we do if they dier only
slightly? We obtain the same jump as in the original Gibbs paradox with a partition between the two gases.
Incidentally, a solution of the original paradox was proposed in the form of a semitransparent partition
(see, e.g., [2][5]). Finally, what happens if we mix Bose gases with dierent masses? This paradox, similar
to the Gibbs paradox, is called the Einstein paradox.
We used purely mathematical theorems [6][18] to show that the value of Z for noninteracting particles
is not equal to unity for each separate gas consisting of molecules of the same type and the principal role
Lomonosov Moscow State University, Moscow, Russia, e-mail: v.p.maslov@mail.ru.
Translated from Teoreticheskaya i Matematicheskaya Fizika, Vol. 167, No. 2, pp. 295310, May, 2011. Original
article submitted February 8, 2011.
654 0040-5779/11/1672-0654
c
is played by the value Z
cr
, i.e., by the compressibility factor expressed in terms of T
cr
, P
cr
, and
cr
, where

cr
is the critical density. This has a number theory meaning, namely,
Z
cr
=
( + 2)
( + 1)
, (3)
where (x) is the Riemann zeta function, = d/2, and d is the fractional dimension of the Bose gas, which
has no meaning as a quantum imperfect Bose gas but can be regarded as the problem of Diophantine
relations (7) and (8) in [1] (see below).
An analogy with number theory permits solving the van der Waals problem of calculating the physical
entropy exactly. In 1908, van der Waals used namely the entropy calculation problem to explain the Gibbs
paradox. He wrote [19]: The entropy itself is not a directly measurable quantity, and we hence cannot
expect that the Gibbs paradox can be explained by measurements.
In [1], we introduced the potential

in number theory. According to the physical representations, the

entropy is determined as the derivative of this potential with respect to the temperature. Direct calculations
show that it coincides with the Hartley entropy for the number of solutions of relations (7) and (8) in [1]
(see, e.g., [20]).
The problem of a gas mixture is still relevant. Relation (1) holds for new ideal gases as
(where is the chemical potential). This relation also holds for their mixtures. In contrast to the new ideal
gases, two mixture types are known for real gases: homogeneous and heterogeneous. The homogeneous
mixture is a mixture in which the gas percentage composition does not vary. Air is an example of such a
mixture. A mixture of air and water vapor remains homogeneous in a very narrow temperature range, until
suciently many water clusters form in the vapor. Then the mixture becomes heterogeneous. The initial
percentage composition of gases inside the droplets of fog varies: the mixture becomes heterogeneous. We
consider only mixtures whose homogeneity regions contain the critical points of both mixture components.
Moreover, we assume that the ideal curve (the Bachinskii law) holds in this region, i.e., the curve
Z = 1 on the (, T) diagram is a segment of the straight Zeno line. We can formally continue this curve
to the points T = T
B
(the Boyle temperature) and =
B
(the Boyle density). These points may never be
attained in actuality. We dene them for convenience in calculations.
The experimental data [21] indicate that a mean Zeno line arises for a mixture of two gases. The
method proposed here does not permit calculating this line, although the scattering problem solution in [1]
relates Z
cr
and the Zeno line. It is possible to choose a one-parameter family of Lennard-Jonestype
potentials for dierent values of the parameters uniquely relating dierent values of Z
cr
and the Zeno line.
We believe that the solution of such a problem is simplied because obtaining Z
cr
and the Zeno line using
the Lennard-Jonestype potential requires only the problem skeleton (see [1]), i.e., the set of point pairs
corresponding to the rest points and not a functional of the interaction potential.
Therefore, if we obtain the ideal (pure) gas corresponding to the sum of the pure gases, then we can
use the formulas given below for the transition from an ideal gas to an real gas.
Remark 1. We here replace the standard thermodynamic normalization N/V const as N
with a van der Waalstype statement: T
red
= T/T
cr
, where T
cr
is the critical temperature at the spinodal
point of the liquid phase. This normalization signicantly changes the notion of thermodynamic limit in
statistical physics because in an ideal Bose gas, we have relations (see (7) and (8) in [1]) that are Diophantine
relations in number theory, where N and E . This implies = 1/T 0. The normalization
T
red
= T/T
cr
replaces the balancing of N in the right-hand side of relation (8) in [1] by using the volume
V and retains the problem in the framework of pure number theory. The dierence from the van der Waals
formulation is that the equality T
red
= 1 holds at the spinodal point of the liquid phase. In the ideal gas,
the temperature of that point is uniquely determined by the value Z
cr
.
655
2. Transition from an ideal gas to a real gas in gas, uid, and
liquid phases
The notion of an real gas indeed diers from the new notion of gas, introduced in [1], only in that the
line Z = 1 corresponds to = in ideal gases and to > in real gases. Moreover, the behaviors of
the temperature and pressure densities for Z = 1 are related quite denitely. On the plane (, T), this is a
slanting line called the Zeno line (see [1]). In other words, it suces to know the constants T
B
and
B
.
If the value =
0
corresponds to Z
cr
, then we multiply by a function of V in our distribution (37)
in [1], i.e., in the BoseEinstein distribution, we change
V

(V ),

(V )
V
const as V . (4)
This function is constant for
0
, where
0
= 0.2 is the critical dimension corresponding to the value
Z
cr
= 0.29 [1]. The quantity Z obtained for = 0 therefore has the form

Z =
V

0
(V )

0
(V )

(
0
+ 2)
(
0
+ 1)
= 0.29. (5)
In the gas phase, the constant in (4) is equal to unity for
i
. To obtain the value
P
cr
V
cr
RT
cr
= Z
cr
=
(
0
+ 2)
(
0
+ 1)
for the new ideal gas, we must take
0
= 0.222.
We repeat the calculations performed in [1] in more detail. For any Z < 1, we have
Z =
V

0
(V )

0
(V )

(
0
+ 1)
(
0
+ 2)
__

0

0+1
d
e

1
___

0

0
d
e

1
_
1
=
V

0
(V )

0
(V )
(),
=

T
,

0
(V ) =

V
, V =
ZT
P
,
(6)
where ( ) is the gamma function. For = 0, we obtain (5). Further, we nd (T, V ) as a function of V
from the condition Z = 1:
V

0
(V )

0
(V )
() = 1, = (V ). (7)
On the other hand, the condition on the Zeno line determined by its slope
P =
B
T
_
1
T
T
B
_
(8)
or
P = T
B

_
1

B
_
(9)
and also
T = T
B
_
1

B
_
(10)
is satised for Z = 1. The dependences P(T), T(), and P() are therefore known: P(T) is the Bachinskii
parabola, T() is a straight line, and P() is a parabola.
656
We obtain P from the BoseEinstein distribution, where V is replaced with
0
(V ):
P =

0
(V )T
0+2
(
0
+ 2)
_

0

0+1
d
e

1
,

0
(V ) Li
0+2
(y) =

T
0+1
B
(1 /
B
)
0+1
, =
1
V
,
V

0
(V )

0
(V )
Li
0+2
(y)
Li
0+1
(y)
= 1,
(11)
where Li is the polylogarithm. The initial condition for <
cr
and T < T
cr
is determined by the condition

0
(V )/V 1 as V . The value
1

0
(V ) does not vary for >
0
.
For T > T
cr
as well as for <
0
, the initial condition is
(V )
V
const. (12)
This constant is chosen such that the ideal gas isotherm for T < T
cr
and for Z = 1 falls on the corresponding
point T on the Zeno line. And for T > T
cr
and Z = 1, the isochore issuing from the critical isotherm toward
the plane (Z, P) falls on the corresponding point on the Zeno line.
Of course, the notion of the new ideal gas [1] must generalize the notion of the usual ideal gas and pass
into it for certain values of the parameters (temperature, pressure, density). The new ideal gas related to
number theory, just as the Bose and Fermi gases, must become the old ideal gas as the chemical potential
tends to , and we obtain Z = 1. Because the condition on the Zeno line is satised (see [1]) and
is observed experimentally, it turns out that only one additional number, namely, the ratio of the Boyle
temperature to the critical temperature, is needed to obtain the thermodynamics of real gases from the new
ideal gases by rather simple transformations (solving a rst-order equation and tting its initial condition
by the shooting method).
Remark 2. For T
cr
= 1 and P
cr
= 1, we obtain
cr
= 1/Z
cr
. For the Zeno line, we have

cr
=
B
_
1
1
T
B
_
and therefore
1 = Z
cr

B
_
1
1
T
B
_
.
Hence, on the Zeno line, we obtain
P = Z
cr

B
_
1
T
T
B
_
T = Z
cr
T
B
_
1

B
_

(if T
cr
= 1 and P
cr
= 1).
For T < 1 in the the BoseEinstein distribution of dimension
0
, the volume V is replaced with

0
(
cr
V ). For T > 1, the volume V is replaced with

(
cr
V ), where >
0
.
The initial condition is replaced with the boundary condition. For an ideal gas T > T
cr
(uids), the
isochoresisodims are constructed (constant density and constant dimension , corresponding to the given
value Z). For T < T
cr
, the isothermsisochores are constructed for the liquid phase. Substituting the
boundary condition for the initial condition also occurs in the case of the liquid phase.
1
There was an unfortunate misprint in [1]: multiplication by the volume V was lost in formulas (26)(29). The condition
(V )/V 1 as V cannot be satised without this factor.
657
3. Normalization to the mercury temperature: Regularization
and correction to the BoseEinstein distribution
We rst perform the necessary renormalization of the BoseEinstein distribution. The integral of the
distribution must be equal to unity in this renormalization. Such an integral can generally diverge. For
example, this is the case for the Pareto-type distributions if the distribution decreases suciently slowly.
We assume that the distribution decreases as 1/ log(|x| + 1). Even its square cannot be integrated.
From the standpoint of Hilbert spaces, this is a generalized function. As an example, we consider the
functions {e
in
} on the interval (, ). If the coecients of the expansion in these functions decrease as
1/ log(|x| + 1), then this is a generalized function in the Hilbert function space L
2
on the interval (, ).
We further assume that its maximum is attained for n = 6 10
23
. This means that from the standpoint of
the Hilbert space L
2
, they oscillate rapidly and hence represent a generalized function equal to zero on the
space of innitely dierentiable nite functions up to 10
23
. Nevertheless, we normalize to this maximum
and study the asymptotic behavior with respect to this maximum as if we were transported into another
world. For the origin (i.e., for the unity), we take the temperature in the hottest tropics, i.e., the tropical
temperature.
Van der Waals, already knowing the standard normalization of the terminological limit N/V const,
noted a new normalization that gave the famous law of corresponding states, which allows classifying gases.
The principal normalization is P
red
= P/P
cr
. We note that if precisely this relation is treated as the
thermodynamic limit in the original statistical physics, then in the original number theory problem of
Diophantine relations (7) and (8) considered in [1], it is already unnecessary to introduce the additional
parameter V (or simply to set V = 1) because N is nite in this remarkable normalization (see (13) below)
and can denote the density , and 1/N is a specic volume.
In the Bose gas, T
cr
is the degeneracy temperature determined as a quantity tending to innity. This
in fact corresponds to the normalization of the distribution to its maximum, very far from zero kelvins.
The number of particles at which the condensate is formed is equal to
N
cr
= (
0
+ 1)(
0
+ 1) (13)
in such a normalization and hence is not too large.
For 0, it is necessary to introduce a correction to the BoseEinstein distribution, namely, to use
formula (37) in [1].
2
We nd the asymptotic form of the relation
N =
_

0
t

dt
_
1
(e
t
/z) 1

N
(e
tN
/z
N
) 1
_
= ( + 1)
_
Li
+1
(e

)
Li
+1
(e
N
)
N

_
(14)
at the point N
cr
given by (13). Relation (14) holds near point (13) and renes the BoseEinstein formula.
We note that the kelvin calibration is indenite because absolute zero cannot be reached. In general, it
is more natural to consider log T for the kelvin calibration. Zero then becomes minus innity, which cannot
be reached, and one kelvin becomes zero. The unattainability of minus innity is more obvious than the
unattainability of zero.
It then makes no dierence from what point to measure. Therefore, log T is determined up to a
constant, and T is hence determined up to multiplication by a constant. It is therefore impossible to
determine T
cr
for a single gas in some kind of units (for example, in kelvins). It is therefore more natural
for the ideal gas to consider the normalization to T
cr
just as the Celsius calibration, for example, is related
to the real gas H
2
O [23], [24].
2
This term is obtained from the parastatistics because the number of particles at each energy level cannot exceed N, the
number of particles in the gas. On the other hand, N is suciently large for the EulerMaclaurin estimates [22] to be used in
the parastatistics.
658
Therefore, to resolve the Einstein paradox, we must include the fractal fractional dimensions arising
for a mixture of two BoseEinstein gases for which there are two dierent degeneracy temperatures T
1
and T
2
(let T
1
> T
2
) depending on the percent composition of the gas volumes. We then accept the
normalization T
red
= 1 at the point T
cr
for the gas with the minimal Z
cr
for the mixture components. By
virtue of the equations given below, we nd the temperature T
sum
cr
and the corresponding dimension
sum
0
.
And because we take into account that particles are indistinguishable when counted, this renement just
leads to the solution of the Gibbs paradox.
If we consider the entire gas scale, then we discover that the values of Z
cr
are bounded above. Mercury
has the greatest value of Z
cr
(0.39). The value of Z
cr
is close to it in the van der Waals model. The value
of Z
cr
is bounded below by the value for the gas He3, for which semiclassical eects are apparent (see [1]).
Therefore, the normalization for the gas scale must be performed by setting T/T
cr
as for mercury.
For a particular gas, the normalization T
cr
= 1 and P
cr
= 1 is possible. But if we compare dierent
gases and study their homogeneous mixtures, then it is most natural to consider the normalization on T
cr
for mercury because the value of Z
cr
for it is the maximum among all gases. In this case, for a classical gas
of dimension , the critical temperature has a value less than unity:
T
()
cr
=
_
(2 +
0
)
(2 +)
_
1/(+2)
, (15)
where
0
= 0.33 for mercury. Here, we nd the corresponding spinodal point of the liquid phase and
renormalize with respect to it.
4. New classical ideal gas as an alternative to the Bose condensate
We considered all solutions of Diophantine equations (7) and (8) in [1] equiprobable. As we previously
showed with the example of distributing E bills over N spectators, if N > N
cr
, then N N
cr
spectators
end up with no bills. Of course, we then assumed that N/N
cr
1 + , where > 0 is any xed number
and E and E N as N . As we have written many times, this fact corresponds to the Bose
condensate in physics and can, for example, determine the unemployment rate in economics.
An alternative to this phenomenon in number theory is association into pairs for N = 2N
cr
and into
clusters for greater values of N. As previously noted, this association means a decrease in the dimension,
which corresponds to a decrease in the number of degrees of freedom of the group components.
The question of how to group optimally, i.e., how to decrease the dimension most protably, arises. A
decrease in the dimension (related to the number of degrees of freedom) can occur in economics either
because of grouping or because of restricting market freedom by laws. How to choose the best way?
In the Korovev example,
3
often discussed in our papers, it was stated that it is sucient for the
spectators to unite into groups of ten, and each group then gets at least one banknote with a very large
probability. By dividing the banknotes among the members of the group, everyone obtains a share and
does not die of hunger. This procedure results in a signicant decrease in entropy because the money
is distributed in each group according to some rule, for example, equally among all the members of the
group. But this million can also be distributed equally among the spectators. The entropy then vanishes,
and communism will arrive, as proposed by another person, Sharikov, in Bulgakovs novel. In both cases,
the main goal is achieved: all the spectators obtain money and survive, which in thermodynamic terms
implies that the number of particles is preserved. But the problem is to preserve the number of spectators
(particles) under the condition that the number of spectators (particles) in each group is minimum. This
means that we must obtain the maximum dimension given that the number of particles is preserved. We
3
In a variety show, Korovev used the trick of throwing 1 000 000 banknotes to 10 000 spectators.
659
call this transition to the maximum dimension transition to the liquid state. While we take the dimension
=
0
for the gas branch of the isotherm and do not change this value for >
0
, we see that for the liquid
branch, the value of and of Z

cr
decrease as the temperature decreases.
Obviously, as the temperature decreases, the number of particles in the groups, i.e., the number of
molecules in the clusters, increases. To obtain this increase, we must raise the original temperature (for
=
0
) for the entropy such that the original number of solutions of the Diophantine equations increases.
The number of molecules in the group (cluster) is then also greater. This means that in the original formula
for the entropy, we must increase T
cr
. Hence, as the real temperature of the liquid phase decreases, we must
increase the relative temperature T
cr
in the original formula for the entropy. This implies that the relative
temperature for Z
0
increases and the original entropy becomes
S =
_
T
T
liq
_
+1
_
Li
+2
(e

)( + 2) + Li
+1
(e

)
_
, (16)
where = /T (see [18]).
Remark 3. The polylogarithm Li
+2
(e

) has the form

Li
+2
(e

) =

e
k
k
+2
=

e
k log k
k
2
.
Its derivative with respect to is equal to

Li
+2
(e

) =

log ke
k log k
k
2
.
For = 0, we obtain
S|
=0
=
_
T
T
liq
_
+1
( + 2)( + 2). (17)
We hence obtain the simple condition for maximizing :
dS
d

=0
= 0, log
_
T
T
liq
_
( + 2)( + 2) +

( + 2)( + 2) +( + 2) = 0, (18)

d log ( + 2)
d
= log
_
1
T
liq
_
+
1
+ 2
. (19)
For mercury, just as for the highest value of Z
cr
for all real gases, we set T
cr
= 1. The solution of (18)
then yields = 0.14, and therefore Z
1(0)
= 0.24.
Because N = const, we consequently have
P
1
= const Z
1
on the isotherm. The points Z
1
and P
1
are the spinodal points of the liquid phase, i.e., the points where
the metastable state of the liquid phase terminates.
Remark 4. The gradient descent method for the entropy describes the unstable part, and its trajec-
tory is not observed in experiments. But to avoid discontinuities in the graphs for the gas and liquid phases,
it is more customary to solve the gradient descent equation from
0
to
1
(
0
) from the point =
0
, = 0
up to the point
1
= 0.14.
660
For mercury, we must set T
red
= T/T
cr
, i.e., T
red
= 1 for T = T
cr
. If T < T
cr
, then T
liq
red
= T/T
cr
.
Because the temperature is less than T
cr
at each point of the liquid for <
1
but each point is
associated with a particular value of Z

cr
, it is convenient to normalize each value of Z

cr
by the corresponding
value T

liq
, which means
T
red
=
1
T
cr
,
i.e., the liquid temperature T
0
cr
and also therefore the entropy S increase by the value of the new T
cr
equal
to 1/T
red
,
S
0
= T
+1
cr
_
Li
0+2
(e

)(
0
+ 2) + Li
0+1
(e

)
_
. (20)
We therefore set T = 1/T
liq
in formulas (18) and (19).
We say that a gas is quasi-ideal if

(V ) = const (see [18]). In this case, the Zeno line appears in

the denition of the spinodal points of the liquid phase, but the liquid remains incompressible. Only the
consideration of the function

(V ) bends the straight lines and takes the compressibility of the liquid
into account. But for T
liq
1, i.e., for suciently small (but still positive) values of , the value of the
function

(V ) already has an insignicant eect. Therefore, a quasi-ideal gas and especially a mixture of
quasi-ideal gases provides a suciently good approximation to the spinodals of the liquid phase.
The geometric locus of the points of the quasi-ideal spinodals
4
is given by
P = T
B
_
1
T
T
B
_
Z

cr
, (21)
where
B
and T
B
are the Boyle density and Boyle temperature. Here,
Z
(T
liq
)
cr
=
(2 +(T
liq
))
(1 +(T
liq
))
, T = T
liq
. (22)
We recall that (T
liq
) can be calculated from algebraic relation (19).
Each point on the quasi-ideal spinodal is joined to the point on the Zeno line corresponding to the
given value of (T
liq
) to the left of the critical line for (T
liq
) = 1. For (T) < 0, we use our previously
introduced correction to the BoseEinstein distribution.
We nd the constants b and from the relations
_

0

_
1
e
b(+)
1

N
e
bN(+1)
_

d = E, (23)
_

0
_
1
e
b(+)
1

N
e
bN(+1)
_

d = N, (24)
where = and b = 1/T.
We set n = E. For = 0, we have
n =
_
d

e
b
1
=
1
b
1+
_

0
d

1
, (25)
where = + 1. Hence,
b =
1
n
1/(1+)
__

0
d

1
_
1/(1+)
. (26)
4
These are the endpoints of the metastable state of the liquid phase.
661
Setting N
cr
= k
0
, we obtain
k
0
=
_

0
_
1
e
b
1

k
0
e
k0b
1
_
d

=
1
b

_

0
_
1
e

1

1

_
d

+
1
b

_

0
_
1

1
(1 + (k
0
/2))
_
d

k
1
0
b

_

0
_
k

0
e
k0
1

k

0
k
0
(1 + (k
0
/2))
_
d

. (27)
We set
c =
_

0
_
1

1
e

1
_

d.
Replacing k
0
= , we obtain
k
1
0
b

_

0
_
k

0
e

1

k

0
(1 +/2)
_
d

=
k
1
0
b

_

0
_
1
e

1

1
(1 +/2)
_
d

=
k
1
0
b

__

0
_
1
e

1

1

_
d

+
_

0
d

2(1 +/2)
_
= c
k
1
0
b

+c
1
k
1
0
b

. (28)
Because
1
(1 +/2)
=
1

1
2(1 +/2)
,
setting
c
1
=
_

0
d

2(1 +/2)
,
we can write
_

0
_
1

1
(1 +k
0
/2)
_
d

=
k
0
2
_

0
d

1 +k
0
/2
=
=
_
k
0
2
_
1
_

0
d

1 +
= c
1
_
k
0
2
_
1
. (29)
Consequently,
k
0
=
1
b

c
1
+
1
b

c
_
k
0
2
_
1

k
1
0
b

_

0
_
1
e

1

1
(1 /2)
_
d

1
2
_
d

1 +/2

k
1
0
b

=
1
b

c +
k
1
0
b

c. (30)
Because k
0
is the number of particles, b = 1/T, and = 1 + , it follows that k
0
b

for > 0 is the

Riemann function (1 + ). Therefore, for < 1, k
+1
0
increases, and we can neglect the rst term in the
right-hand side of (30). Setting T
cr
= 1 in this asymptotic form, we see that it is natural to regard the
662
function
M( + 1) =
_
c()
( + 1)
_
1/(1+)
(31)
as the continuation of the Riemann zeta function to the additional domain of the arguments 0 > 1.
Remark 5. For < 0, the asymptotic form of (23) can be expressed as
_
t

dt
k
2
e
k1
e
t/k2
1
( + 1) Li
+1
(e

), (32)
and for k
1
, k
2
, and 0, it becomes c().
The compressibility factor Z

= ( + 2)/M( + 1) changes by a jump

5
from = 0 to < 0.
We see that the metastable region of negative pressures terminates. This implies the paradoxical eect of
increasing density as the temperature decreases, which was observed experimentally (see [26] and also the
video on the site http://www.iem.ac.ru//staff/kiril, where this phenomenon is observed).
In the considered approach, we nd that the trajectory satises the relation
dN = 0. (33)
Hence, the new ideal gas corresponds to discontinuous isotherms composed of the gas branch and the
incompressible liquid branch.
From the standpoint of economics, this means that the decrease in freedom is a result of rst
association and then strengthening of laws such that, for example, there would be no unemployment and
the number N would be preserved. From the physical standpoint, this means that the increase in N for
T > 1 leads to a cascade increase of associations into clusters, and the incompressibility phenomenon then
arises, i.e., an increase in the pressure no longer leads to a decrease in the volume.
The trajectory, for T > 1 consisting of two nonintersecting straight lines on the plane (Z, P), following
from number theory, is the skeleton of thermodynamics and corresponds to the new ideal gas. The smoothing
(enveloping by soft tissue) of this skeleton reects the eects of the interaction.
Remark 6. The rst-order phase transition on the plane (P, Z) at T < T
cr
occurs for P = P
cr
and density and chemical potential variations, and the rst-order phase transition at T < T
cr
occurs on
the isotherm for dierent pressures and dierent chemical potentials. This transition corresponds to the
transition on the van der Waals diagram from the gas branch to the uid branch on the entire interval of
metastable states: from the spinodal point on the gas branch to the spinodal point on the liquid branch.
To obtain a phase transition of standard form for T < T
cr
, we must nd a point of equal chemical
potentials on the two isotherms (the Maxwell rule). This is a separate condition of projection on the plane
(T, P) of two Lagrangian manifolds with an edge (the edge corresponds to the spinodal) under which the
Gibbs potential takes the maximum value. In other words, the chemical potential along the isotherm is
minimized.
The so-called geometric quantization of these two manifolds can be performed by the tunnel canonical
operator, where 1/N is a small parameter. This is the only small parameter in number theory. According
to [1], it is natural to take the gas diusion as a small parameter for real gases (see Sec. 6).
5
The jump of the spinodal as it passes through the point P = 0 from positive pressures (compression) to negative pressures
(stretching) is natural for a liquid because like rubber with small pores for which Hookes law for strain and stress undergoes
a jump, the derivative of pressure with respect to density also changes by a jump. Of course, this jump is smoothed for a real
gas [25].
663
5. Einstein paradox and ideal gas of a mixture
The notion of a mixture of ideal gases according to the old concept of an ideal gas corresponds to the
Gibbs paradox. We acted radically: we rejected the old notion of an ideal gas and accepted a new notion
of it and the corresponding new notion of a mixture of ideal gases.
Describing a real gas, we can construct a potential that diers somewhat from the Bose gas of
fractional dimensions in the gas phase and diers signicantly for uids and the liquid phase [1].
We present the formulation of the Bose-statistics paradox, which is often compared to the Gibbs
paradox [27]. Einstein wrote ([28]): In conclusion, I want to draw attention to a paradox that I could not
explain. Using the method presented here, we can also easily consider a mixture of two dierent gases. In
this case, each type of molecule has its own specic cell, whence the additivity of the entropy of the mixture
components follows. Therefore, each component behaves in the sense of the molecular energy, pressure, and
statistical distribution as if the mixture volume contained only it alone. But for a given temperature, the
mixture of n
1
molecules of one type and n
2
molecules of another type in which the dierence between the
molecules of the rst and the second types is arbitrarily small (especially with respect to the molecular
masses m
1
and m
2
) has a dierent pressure and a dierent state distribution than the simple gas with
the n
1
+n
2
molecules of practically the same molecular mass occupying the same volume. But this seems
impossible.
If we consider the quantum Bose gas for which the initial function is symmetric with respect to n
1
particles of the rst type and n
2
particles of the second type, then the solution is obvious. Our Hamiltonian
is not symmetric with respect to permutations of p
(1)
i
, q
(1)
i
and p
(2)
i
, q
(2)
i
, where the superscript distinguishes
particles of the rst and second types. Even if we symmetrize the initial function , this symmetry
immediately disappears for t > 0. And the symmetrization itself is a jump of the symmetry operator.
As we saw in [1], the Bose-gas statistics corresponding to the number theory problem is a quite
another matter. If we deal with calculations of the Bose-condensate point, then the calculation of the sum
dimension also corresponds to the Bose condensate at the same Z
sum
cr
, as follows from [1]. We hence obtain
the Einstein rule of addition of Bose gases of any fractional dimension with the corresponding values of the
Bose condensates (i.e., critical points). Each of the masses m
1
and m
2
is associated with its own values T
(1)
cr
and T
(2)
cr
or N
(1)
cr
and N
(2)
cr
according to the van der Waals law of corresponding states and the normalization
P
cr
= 1: gases with the same value of Z
cr
are equivalent.
We speak about a mixture of gases with dierent values of the critical quantities if we dene the density
of this mixture. This again reduces to a number theory problem and to Diophantine relations. Our problem
is to introduce a new notion, a new ideal gas of a mixture of gases. It follows from the mass conservation
law for N
(1)
cr
, N
(2)
cr
, and N
()
cr
that
(m
1
+m
2
)
+1
(T
()
cr
)
+1
( + 1) = m
1+1
1
(T
(1)
cr
)
1+1
(
1
+ 1) +
+m
2+1
2
(T
(2)
cr
)
2+1
(
2
+ 1), (34)
where + = 1. It follows from the additivity of entropy that
((m
1
+m
2
)
+1
( + 2)( + 2)(T
()
cr
)
+1
= m
+1
1
(
1
+ 2)(
1
+ 2)(T
(1)
cr
)
1+1
+
+m
+2
2
(
2
+ 2)(
2
+ 2)(T
(2)
cr
)
2+1
(35)
(both identities are in a nonequilibrium state). An analogue of the Zeno line and the function
0
(V ) will
be considered in detail in the next paper.
664
6. Critical exponents, the scaling hypothesis, the tunnel canonical
operator, and the Maxwell rule
Not only the considerations presented above but also the arguments involving the Lennard-Jones
interaction potential are based on the role of attraction between particles or, equivalently, on the formation
of combinations (dimers and clusters) in the denition of the new ideal gas. The role of repulsion was only
auxiliary when we spoke of the value of small viscosity for a dimer captured in a trap (potential well).
Namely, the dimer must eventually descend to the bottom of the well because of small viscosity.
In the notion of ideal gas thus introduced, the states that are metastable on the liquid and gas branches
of the ideal gas do not dier from the stable state corresponding to the Maxwell rule for which the chemical
potentials on these branches must be equal. This remarkable fact is explained by repulsion and thus by
viscosity. In our opinion, Widoms scaling hypothesis is related to this point.
Because the data of the classical theory of critical exponents did not coincide with the experimental
data and Widom (1965) developed a hypothesis to some extent contradicting the basic postulates of ther-
modynamics, it would seem natural that the solution of the problem concerning the Gibbs paradox would
also abolish this contradiction. Indeed, the nonanalyticity arises in the new theory concerning the transition
from the new ideal gas to the imperfect one. Namely, the derivative of

(V ) with respect to vanishes for

0
and T < T
cr
, where
0
corresponds to the critical point, while this derivative is nonzero for >
0
.
But the independence of the critical exponents evaluated according to the scaling hypothesis for six
critical values of gas leads us to suggest that the principal point of contradiction involves an additional factor.
For a homogeneous mixture, no experimental conrmations of the critical parameters evaluated using the
scaling hypothesis are known to us, but we have no doubt that these parameters will be conrmed.
After constructing a rigorous mathematical conception of thermodynamics, it is necessary to give an
answer concerning this disagreement. In this section, we answer this question, although the appropriate
derivations and numerical coincidences are still not available.
The relation between viscosity and diusion is well known in molecular physics. The diusion parameter
and the parameter ih, where h denotes the Planck constant, are closely related (see, e.g., the formulas for
instantons in [29] and [30]).
We have also written several times that the so-called thermodynamic potentials in the phase space
T, P, V, S form a two-dimensional surface, which we called a Lagrangian manifold in the phase space
(q
1
, q
2
, p
1
, p
2
) in [31], and the thermodynamic potentials are a representation of the action S =
_
p dq in
dierent local charts of this manifold.
As noted above, the thermodynamics of an imperfect gas that we have constructed is not analytic at
the critical point. But the leading term of the tunnel canonical operator [32], where the role of a small
parameter is played by diusion, is characterized in the Maxwell phase transition by the jump of the gas
branch to the liquid branch for the same value of the chemical potential. Diusion erodes this density
jump in the same way that viscosity in thermodynamics smooths the shock wave, although this is not so
clear in experiments. But at the critical point, as at the instant of generating a shock wave (for example,
for the Burgers equation), the focus formed at the point is more signicant the stronger the inuence of
the canonical operator and diusion. Therefore, the contribution near the focal point (in our case, near the
critical point) gives an essential correction to the critical index obtained from the asymptotic behavior of
the canonical operator. This contribution is independent of the value of the critical point.
The universal property, i.e., the independence from diusion and dependence only on the defect of the
focal point [33], is related to the fact that the distance from the critical point is also a small parameter that
is much less than the diusion parameter. The scaling rule holds for the simplest focal points expressed
using the Airy function or the Weber function. The defect of a focal point [33] is an invariant related to the
projection of this point on the coordinate plane (T, P). Similarly, the projection to other coordinate planes
665
gives other values of the critical parameters. In mathematics, this procedure (in the case of an oscillating
nontunnel canonical operator) is known as geometric quantization (see [34]). Accordingly, the creation
operators for dimers and other clusters also generalize the notion of secondary quantization, extend the area
of application of thermodynamics to the semiclassical case (see [35]), and, in particular, allow obtaining the
superuidity of classical gases (e.g., H
2
O) in nanotubes [36][38], which was conrmed experimentally.
Acknowledgments. The author deeply thanks V. S. Vorobiev for the useful discussions.
REFERENCES
1. V. P. Maslov, Theor. Math. Phys., 165, 16991720 (2010).
2. M. Planck, Einf uhrung in die theoretische Physik, Vol. 5, Einf uhrung in die Theorie der Warme zum Gebrauch
bei Vortragen, sowie zum Selbstunterricht, S. Hirzel, Leipzig (1930).
3. O. Wiedeburg, Ann. Phys., 289, 684697 (1894).
4. A. Byk, Z. Phys. Chem., 43/45, 465495 (1903).
5. G. Lorentz, Lectures on Theoretical Physics, Vol. 2, Thermodynamics, Macmillan, London (1927).
6. V. P. Maslov, Russ. J. Math. Phys., 18 (2011, to appear).
7. V. P. Maslov, Math. Notes, 86, 522529 (2009).
8. V. P. Maslov, Math. Notes, 86, 605611 (2009).
9. V. P. Maslov, Math. Notes, 87, 384391 (2010).
10. V. P. Maslov, Russ. J. Math. Phys., 17, 240250 (2010).
11. V. P. Maslov, Math. Notes, 87, 728737 (2010).
12. V. P. Maslov, J. Fixed Point Theory Appl., 8, 81111 (2010).
13. V. P. Maslov, Math. Notes, 88, 5766 (2010).
14. V. P. Maslov, Russ. J. Math. Phys., 17, 288306 (2010).
15. V. P. Maslov, Zeno-line, binodal, T diagram, and clusters as a new Bose-condensate bases on new global
distributions in number theory, arXiv:1007.4182v3 [math-ph] (2010).
16. V. P. Maslov, Math. Notes, 88, 516523 (2010).
17. V. P. Maslov, Math. Notes, 88, 723731 (2010).
18. V. P. Maslov, Math. Notes, 89, 266276 (2011).
19. J. D. van der Waals and P. Kohnstamm, Lehrbuch der Thermodynamik, Vol. 1, Maas and von Suchtelen, Leipzig
(1908).
20. I. A. Molotkov, Russ. J. Math. Phys., 17, 476485 (2010).
21. V. V. Sychev, A. A. Vasserman, A. D. Kozlov, G. A. Spiridonov, and V. A. Tsymarny, Thermodynamic Prop-
erties of Air [in Russian], Standards Press, Moscow (1978); English transl.: Thermodynamic Properties of
Oxygen, Hemisphere, New York (1987).
22. V. P. Maslov and V. E. Nazaikinskii, Math. Notes, 83, 211237 (2008).
23. V. P. Maslov, Theor. Math. Phys., 160, 11441145 (2009).
24. V. P. Maslov, Math. Notes, 83, 716722 (2008).
25. V. P. Maslov and P. P. Mosolov, Nonlinear wave equations perturbed by viscous terms, in: De Gruyter
Expositions in Mathematics, Vol. 31, de Gruyter, Berlin (2000).
26. K. I. Shmulovich and L. Mercury, Geochemical processes at negative pressures [in Russian], http://web.ru/
conf/khitariada/1-2006/informbul-1 2006/hydroterm-16.pdf (2006).
27. Ya. M. Gelfer, V. L. Lyuboshits, and M. I. Podgoretsky, Gibbs Paradox and Identity of Particles in Quantum
Mechanics [in Russian], Nauka, Moscow (1975).
28. A. Einstein, The Collected Papers of Albert Einstein, Vol. 4, Quantum Theory of a Monoatomic Ideal Gas,
Princeton Univ. Press, Princeton, N. J. (1996).
29. M. Dykman and L. Pryadko,, Lectures of Theory of Dissipative Tunneling (http://www.pa.msu.edu/dykman/
PHY972/instanton lectures.pdf) (1998).
30. V. P. Maslov, Trudy MIAN, 163, 150180 (1984).
666
31. V. P. Maslov, Theorie des Perturbations et Methodes Asymptotiques [in Russian], Moscow State Univ. Press,
Moscow (1965); French transl., Dunod, Paris (1972).
32. V. P. Maslov,, Russian J. Math. Phys., 17, 400408 (2010).
33. V. P. Maslov, Asymptotic Methods and Perturbation Theory [in Russian], Nauka, Moscow (1988).
34. N. E. Hart, Geometric Quantization in Action, D. Reidel, Dordrecht (1983).
35. V. P. Maslov, Quantization of Thermodynamics and Ultrasecond Quantization [in Russian], Inst. Kompyuter-
nykh Issledovanii, Moscow (2001).
36. V. P. Maslov, Russian J. Math. Phys., 14, 304318 (2007).
37. V. P. Maslov, Russian J. Math. Phys., 14, 401412 (2007).
38. V. P. Maslov, Russian J. Math. Phys., 15, 6165 (2008).
667
Copyright of Theoretical & Mathematical Physics is the property of Springer Science & Business Media B.V.
and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright
holder's express written permission. However, users may print, download, or email articles for individual use.