Planck's Legacy to Statistical Mechanics
Extracted source text ยท arXiv cond-mat/0101293
\documentclass[12pt]{article}
\usepackage{graphicx}
\usepackage{epsfig}
\newcommand {\bi} {\bibitem}
\newcommand {\be} {\begin{equation}}
\newcommand {\ee} {\end{equation}}
%\newcommand {\be} {$$}
%\newcommand {\ee} {$$}
\newcommand {\bea} {\begin{eqnarray} }
\newcommand {\eea} {\nonumber \end{eqnarray}}
\newcommand {\eps} {\epsilon}
\newcommand {\si} {\sigma}
\newcommand {\de} {\delta}
\newcommand {\De} {\Delta}
\newcommand {\ga} {\gamma}
\newcommand {\la} {\lambda}
\newcommand {\La} {\Lambda}
\newcommand {\al} {\alpha}
\newcommand {\NE }{\not=}
\newcommand {\N} {{\cal N}}
\newcommand {\R} {{\cal R}}
\newcommand {\LL}{{\cal L}}
\newcommand {\ba} {\overline}
\newcommand {\lan} {\langle}
\newcommand {\ran} {\rangle}
\newcommand {\cA} {{\cal A}}
\newcommand {\cC} {{\cal C}}
\newcommand {\cD} {{\cal D}}
\newcommand {\cG} {{\cal G}}
\newcommand {\cH} {{\cal H}}
\newcommand {\cL} {{\cal L}}
\newcommand {\cN} {{\cal N}}
\newcommand {\cM} {{\cal M}}
\newcommand {\cP} {{\cal P}}
\newcommand {\cR} {{\cal R}}
\newcommand {\cS} {{\cal S}}
\newcommand {\cX} {{\cal X}}
\newcommand {\cE} {{\cal E}}
\newcommand {\by} {{\bf y}}
\newcommand {\bx} {{\bf x}}
\newcommand {\ap} {\left(}
\newcommand {\cp} {\right)}
\newcommand {\spa} {\vskip .5cm}
\newcommand {\via} {}
\newcommand {\bc} {\begin{center}}
\newcommand {\ec} {\end{center}}
\newcommand {\bd}{\begin{displaymath}}
\newcommand {\ed}{\end{displaymath}}
\newcommand {\ato} {\left(}
\newcommand {\cto} {\right)}
\newcommand {\tC} { {\tilde{C}} }
\newcommand {\Tr} {\mbox{Tr}}
\newcommand {\sign} {\mbox{sign}}
\newcommand {\for} {\ \ \ \mbox{for}\ \ }
\newcommand {\bra }{\langle 0 |}
\newcommand {\ket} {| 0 \rangle }
\def \form#1 {eq. (\ref{#1}) }
\def \parziale#1#2 {{\partial {#1} \over \partial {#2}}}
\topmargin=-2.5cm \textheight=24.0cm
\textwidth=18. cm\oddsidemargin=-1.cm
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{document}
\title{Planck's Legacy to Statistical Mechanics}
\author{Giorgio Parisi \\ Dipartimento di Fisica, Sezione INFN and unit\`a INFM,\\
Universit\`a di Roma ``La Sapienza'',\\
Piazzale Aldo Moro 2,
I-00185 Roma (Italy)\\
Giorgio.Parisi@Roma1.INFN.it}
\maketitle \abstract{In this talk I will describe the deep influence
Planck had on the development of statistical mechanics. At this end I will first outline the
theoretical situation of statistical mechanics before Planck. I will then describe his main
contributions to this field and the progresses obtained as an immediate consequence of his work. I
will also outline the later evolution of statistical mechanics in relation with Planck's work. I
will finally report on a still unsolved problem in statistical mechanics, historically
related to the properties of black-body radiation. }
\section{Introduction}
In this talk I will try to highlight the role that statistical mechanics played on Planck's discovery
and the consequences of Planck's discovery on statistical mechanics.
For reasons of time I cannot present a complete historical account \cite{KP} and I will select the
points that seem to me more important.
I will first outline the situation in statistical physics at the end of the 19$^{th}$ century. I
will then present Planck's approach to radiation and his two first derivations (1900) of the famous
formula for the black body radiation. I will then outline the short term consequences of Planck's
discovery on statistical mechanics and I will recall a part of the discussions done after it was
realized that it was impossible to derive Planck formula in equilibrium statistical mechanics. I
will go on by describing the discovery of quantum statistics and I will finally report on our
present understanding of approach to equilibrium in weakly non linear systems, a problem that was
already discussed at Planck's time and was relevant in the debate on the black body radiation.
Finally in the conclusions I will present some personal considerations.
\section{Statistical Physics at the end of the 19$^{th}$ century}
At the end of the 19$^{th}$ century thermodynamics was well established. One of the most impressive
results is the purely thermodynamical derivation of Wien's displacement law for the black body
radiation using as input the first two laws of thermodynamics.
If we call $E(\nu,T)$ the intensity of the black body radiation at frequency $\nu$, Wien's
displacement law \cite{WDISP} states that
\be
E(\nu,T)=\nu^{3} \phi(\nu/T) \ , \label{WIEN}
\ee
where $\phi(\nu/T)$ is a function that cannot be determined on purely thermodynamics grounds. The
proof, which I have not checked, is rather sophisticated; it involves monochromatic filters, the
analysis of the effect of the radiation pressure on a moving piston and the variation of the
radiation spectrum as a consequence of the movement of the piston, using the Doppler effect.
However we could say that at that time there was no statistical mechanics. The name was practically
not present in the literature: what really existed was ``kinetic gas theory.'' Statements like ``the
probability distribution of a system at equilibrium is $\exp (-\beta H)$'' were not so interesting.
The Hamiltonian of $N$ particles could be written as
\be H=\sum_{i}{p^{2}_{i}\over 2m} +V(x)\ ,
\ee
but the form of $V(x)$ was completely unknown (e.g. in its first two papers Einstein 1902 supposes
that $V(x)$ is Coulomb like, but he was not able to find a good agreement with the experiments).
The real interesting problem was the computation of the distribution of the momenta at equilibrium
and of the related transport coefficients. Planck made the first successful attempt to go beyond
kinetic gas theory and to apply statistical mechanics to a completely different systems.
\section{Planck's approach to radiation}
Planck started his carrier as physicist with thermodynamics, a subject that fascinated him. He was
particularly interested to the second law. For
example, while
Clausius usually wrote the second law as:
\be \oint {\delta Q \over T}\le 0\ .
\ee
Planck preferred to use the following mathematical equivalent formulation (which was not unknown to
Clausius):
\be
S(final)\ge S(initial)\ .
\ee
In this way there was a symmetry among the two laws: the first implies the conservation of energy,
the second the increase of entropy.
At the beginning Planck had difficult relations with Boltzmann. He considered Boltzmann's theory
more or less a useless effort, with not enough results. When his young assistant Zermelo attacked
Boltzmann \cite{ZERMELO}, claiming that the Poincar\'e recurrence theorem implies that one cannot
obtain irreversible effects using a reversible Hamiltonian, he agreed with Zermelo from the
technical point of view: he seemed not to like that his beloved second law was not exact, but only a
probabilistic statement. Contrary to Zermelo, he believed that the second law could be proved by a
microscopic theory where continuous degrees of freedom were present (later on, he took the point of
view of Boltzmann on the use of probabilistic methods). According to Planck, the interaction of
matter with the radiation is the origine of irreversibility: he started to study the black body
radiation in order to construct a mechanical proof of the second law.
Why radiation should be the origine of irreversibility? We know now that the equation of motion of
matter and radiations are obviously invariant under time reversal, so radiation cannot decide the
arrow of time. Which was Planck's approach? He started \cite{PRIS} by deriving the equation of
motion of a radiating dipole $f$:
\be
{d^{2}f \over dt^{2}}+ \alpha {df \over dt} +\omega^{2}f=E(t)\ ,
\ee
where the dumping term (proportional to $\alpha$) is due to the emission of radiation and the
forcing term ($E(t)$) is proportional to the external electric field. Under this form the equations
are no more invariant under time reversal, however, in order to derive the previous equation, it is
necessary to assume the electromagnetic equivalent of Boltzmann molecular chaos (e.g. a random
distribution of the phases of the different modes in a cavity). His equation for the dumped
radiating dipole was in reality derived with the same logical steps of Boltzmann, and when Planck
realized this, he was able to understand much better Boltzmann's point of view, which he
publicly joined.
\section{Planck's derivation of his formula for the black body radiation}
The main argument of Planck on black body radiation was based on the interaction of the radiation
field with matter. In his approach, the simplified model for matter was a set of resonators of
frequency $\nu$. Resonators are essentially radiation dumped harmonic oscillators (radiating
dipoles), with the simplification that they can absorb and emit only radiation with the frequency $\nu$.
He firstly studied a resonator with frequency $\nu$.
Using the equations of motion for the dumped harmonic oscillator, he obtained \cite{PWIEN} that
\be
E_{Radiation}(\nu,T)={8 \pi \nu^{2} \over c^{3}} U_{Matter}(\nu,T) \ ,
\ee
where $U_{Matter}(\nu,T)$ is the energy of a resonator of frequency $\nu$ at temperature $T$. In
the same framework he had already derived an electromagnetic equivalent of Boltzmann's $H$-theorem.
For Planck the energy $U(\nu,T)$ is fixed by the knowledge of the entropy $S(\nu,U)$, which,
contrary to any logical argument, is not the usual entropy of the harmonic oscillator. In other
words Planck believed that it was consistent to treat the resonators as harmonic oscillators as far
as the interaction with the radiation was concerned and to suppose that the form of the entropy
$S(\nu,U)$ was determined by the unknown internal structure of the resonators. The complete
ignorance of the structure of matter lead Planck to think that $S(\nu,U)$ could be any function,
provided that it is in agreement with Wien's displacement law. Indeed Wien's displacement law is
satisfied only if
\be
S(\nu,U)=s\left({U \over \nu }\right) \ .
\ee
Planck believed that $S(\nu,U)$ should be a simple expression. In his first attempt to explain the
black body radiation \cite{PWIEN} he obtained the Wien distribution \cite{WDIST}:
\be E(\nu) \propto \nu^{3} \exp(-a \nu/T) \ ,
\ee
assuming that the entropy was given by the expression
\be
S(\nu,U)={-U \over a \nu }\ln{U \over b\nu} \ ,\ee
that seemed particular simple to Planck (he liked the presence of a logarithm).
It may be interesting to notice that the previous equation implies that
\be
{\partial ^{2}S \over \partial U^{2}} \propto U^{-1} \label{W} \ ,
\ee
while for a classical harmonic oscillator we have
\be
{\partial ^{2}S \over \partial U^{2}} \propto U^{-2} \label{C}\ .
\ee
In the early afternoon of the Sunday October $7^{th}$ of 1900 he learned from Rubens that the data
for the black body radiation in the far infrared were proportional to the temperature, as follows
from the classical expression eq. \ref{C}. He tried to interpolate among the previous formulae:
the first was {\sl experimentally} correct at small $U$ and the second was correct at large $U$.
The simplest interpolation among the two formulae was
\be
{\partial ^{2}S \over \partial U^{2}} ={-\alpha(\nu) \over U (U+\beta(\nu))} \ .
\ee
Starting from it, he derived in late afternoon of the same day Planck's law for the black body
radiation and its expression was sent to Rubens on a postcard the same evening.
Planck's law was presented to the Academy \cite{PDIS1} two weeks later, nearly simultaneously with
the presentation of the experimental data of Rubens and Kurlbaum.
In the second paper \cite{PDIS2} he computed the entropy of the resonator using the formula
\be
E_{n}= n h \nu \ , \label{Q}
\ee
where this formula applies only to the resonators ({\sl not} to the
electromagnetic fields).
However he did not assume that the energy of the resonators is quantized, i.e. it is given by the
previous formula.
If one reads carefully the paper, one discovers that he did something very different: he
defined $n$ as the integer nearest to $E/(h \nu)$:
\be
n \approx {E \over h \nu} \ .
\ee
He assumed also that the entropy of the oscillator can be computed by summing over $n$ instead
of integrating over the energy. All values of the energies are allowed. The quantization of the
energy enters only in the computation of the entropy \footnote{The formula
$ S= k \ln W $ is written on Boltzmann's grave but it
was not written by Boltzmann.
Planck introduced $k$ in the in its first paper on the black body radiation. He was the first to
write later $S= k \ln W +const $. At the beginning $k$, not $h$, was called Planck constant.}.
\section{The development of statistical mechanics}
Nearly simultaneously to Planck paper, Gibbs and Einstein \cite{GIBBS, ESTAT}, put forward the modern
statistical mechanics: canonical ensembles, microcanonical ensembles, thermal fluctuations and all
that. In this way physicists had the theoretical tools needed to apply statistical mechanics to
many different problems.
One of the main consequences of Planck work was the possibility of computing the specific heat of
solids and of diatomic gases at low temperature. The first step was done by Einstein in 1908 and it
was followed by a large number of papers. The subject was very important and it was a crucial
ingredient for computing the absolute value of the entropy, which had important practical relevance
in the study of chemical equilibrium \cite{NERST}. One could say that the construction of models in
statistical mechanics started in order to reach a better understanding of the specific heat of
solids. In this way statistical mechanics was applied to many different cases and the original
kinetic gas theory becomes only a restricted field.
However, these new developments could happen only after a clear understanding of Planck's original
proofs. It was crucial to establish which was the correct way to apply statistical mechanics to
black body radiation and to other quantum systems. There was a strong debate and eventually it was
clear that Planck's proofs were inconsistent. The conclusions was that according to classical
statistical mechanics the properties of the resonators are irrelevant: the equipartition law should
be valid for the electromagnetic field, which is a nearly perfect harmonic oscillator,
irrespectively of any contrived form of the entropy of the resonator. The entropy of the resonator
could be extremely complex due to non linear effects, without affecting in any way the black-body
radiation \cite{ERE}.
It was also realized that, at thermal equilibrium in classical mechanics,
the Rayleigh-Einstein-Jean law \cite{REL,ERJEJ,JEANS} must be valid:
\be
E_{Radiation}(\nu)={8 \pi \nu^{2} \over c^{3}} kT \ ,
\ee
This law coincide with Planck's law,
\be
E_{Radiation}(\nu)={8 \pi \nu^{2} \over c^{3}} {h \nu \over \exp (h \nu/kT) -1}\ ,
\ee
only in the region of small frequencies.
It is evident that the Rayleigh-Einstein-Jean law is not experimentally correct, because it corresponds to an
infinite electromagnetic energy \footnote{The fact that classical mechanics leads
to the so called ultraviolet catastrofe was discovered only a few years after Planck's paper and not
viceversa, as suggested by the usual folklore.}.
The solution to this paradox was not evident. Rayleigh, Jeans, Einstein and Planck had very different
opinions on the direction which should be taken.
\subsection{Rayleigh's solution}
The point of view of Rayleigh \cite{JEANS} was that the equipartition law (and consequently
statistical mechanics) is not correct in general. It is valid for the low frequency modes, but not
at high frequencies. One could paradoxically ascribe this opinion to Boltzmann: he was the first to
notice that the oscillatory degrees of freedom, that can be observed in the absorption and emission
spectrum, do not contribute to the specific heat of a solid or of a gas. Equipartition gave always
the wrong result at high frequency. A wrong hypothesis should be present in its derivation: where
else the error could lurk?
\subsection{Jean's solution} Roughly speaking, Jeans believed that at thermal equilibrium
the Rayleigh-Einstein-Jean law should be valid, bur the black body radiation could take an extremely long
time to reach equilibrium \cite{JEANS}. In real experiments the radiation should have not yet reached the
equilibrium. The Planck distribution was only an approximate law, where $h$ could be a time dependent
decreasing function.
However, if black body radiation is not at equilibrium, the thermodynamical derivation of the Wien
law (eq. \ref{WIEN}) was no more valid. People were reluctant to abandon one of the best verified
law on black body radiation and Jean's proposal was not very successful.
\subsection{Einstein's solution}
Einstein's solution was the photon, i.e. he supposed that the energy of the electromagnetic field is
quantized. In a beautiful paper in 1909 \cite{EPHOT} he observed that if the Wien law were correct,
the entropy of the electromagnetic field would be the same as that of $N$ independent pointlike quanta of
energy given by eq. \ref{Q}. Einstein used this argument for jumping to the correct conclusion that the
energy of the electromagnetic field is quantized and each of the $N$ quanta (the photons) may be
absorbed independently from the other (a position he had already taken in the explanation of the
photoelectric effect in 1905).
As far I understand, Einstein did not realize that in the regime where the Wien law is correct, the average
value of $N$ is very small and it can be computed assuming that $N$, may takes the values 0 or 1.
Indeed, Planck's law does not correspond to a gas of classical non interacting photons. We know now
that only in the deep quantum regime the correct expression for the quantum photon gas coincides with
that of a classical photon gas: the deep quantum regime for the wave is the classical regime for
particles.
The first modern proof of Planck law's, where one assumes that the energy of the
electromagnetic field is quantized (i.e. satisfies eq. \ref{Q}), appeared in 1910 \cite{DEB}. It is
interesting to note that many physicists think that this is the original proof presented by
Planck.
Einstein's proposal, i.e. the quantization of the electromagnetic field, was treated as heretical:
in 1911 Planck wrote a letter where he said he supported Einstein candidature for a chair in
spite of the fact that he has done the mistake of proposing the photon. The general opinion was
that the resonator could behave in a strange way: its equation of motion were not known, however the
electromagnetic field was described by the Maxwell equations and there was no place for a quantized
electromagnetic field. The Maxwell equations were in very good agreement with
experiments and there was no reason to assume that they were not correct. In other words, many
physicists believed for a long time that quantization was the non-linear effect of a classical dynamics
and that there was no place for it, if the equations of motions were linear.
\subsection{Planck's solution}
Planck was rejecting the quantization of electromagnetic energy. It was also rejecting the
possibility that the resonator could absorb the energy from the electromagnetic field in a discontinuous
way. He was assuming that the microscopic system were evolving in a continuous way. He could accept
the quantization in the emission of radiation, as a very fast process in which the radiation goes
from the resonator to the electromagnetic field; the process should be so fast that it could be considered
discontinuous. He could not accept the quantization in the absorption of radiation, because, as an effect
of local energy conservation, the absorption of a very energetic quantum should take a
very large amount of time \footnote{The time should exponentially diverge when the energy increases
because the amount of available energy in the high frequency modes is exponentially small.} .
He put forward the following rather strange theory: the resonators adsorb energy in a continuous way
from the electromagnetic field and they change their energy in a continuous way. The resonator may
emit a radiation equal to $E_{n}\equiv n h \nu$ only when its energy is exactly equal to $E_{n}$ and
in this way it makes a discontinuous transition to the state of zero energy.
This proposal has nothing to do with reality, but it contains many new ideas which had a deep
influence on later developments:
\begin{itemize}
\item The different orbits ($H=E_{n}$) in the two dimensional space divide the phase space in
regions of area $h$. We could call each region a quantum state. In this way, each quantum state
occupies an area in phase space equal to $h$: in this respect $h$ appears as a quantum of
action. The similarities at technical level with the later work of Bohr and Sommerfeld are striking.
\item The average energy of a quantum state is
\be
{E_{n}+E_{n+1}\over 2}=(n+\frac12) h \nu \ .
\ee
The minimal energy of a resonator is $\frac12 h \nu $. The zero point energy enters in physics
for the first time \footnote{It is amusing to note that the proposal of a zero point energy
soon became popular: it was applied it to the specific heat of diatomic gas, showing that
the experimental data presented a strong support for a zero point energy in the rotational
spectrum \cite{ZERO} (nowadays we know that zero point energy in the rotational spectrum is
not present: in the $s$-wave the rotational energy is zero!)}.
\item The emission of the
radiation (when the oscillator crosses the level $E_{n}$) is a probabilistic process, which
happens with a given probability. This was the first case where a transition
probability between quantum states was introduced .
Planck's intuition that the emission of radiation should
not be a deterministic process is remarkable.
\end{itemize}
\section{Quantum Statistical Mechanics}
Of course, the final solution of the previous dilemma was quantum statistical mechanics: Einstein was
right and the energies of both matter and the electromagnetic fields are quantized (in a first
approximation according to the Bohr-Sommerfeld quantization formula). This solution took long time
to be completely accepted. Still in 1924 Bohr did not believed in the reality of the photon.
According to him, the electrons were jumping in discontinuous way, but they were emitting and
adsorbing energy in continuous way; this was possible due to the fact that energy and momentum were
not conserved in individual quantum events \cite{BOHR}. The experimental observation of the Comtpon
effect, where it was possible to check the conservation of energy and momentum in a given
individual process, was the definitive and needed proof of the quantization of the energy of the
electromagnetic field and it convinced nearly everybody.
However, quantum statistical mechanics for many particles was not easy to establish. The Planck
formula did not correspond to a gas of classical photons.
The way to attack this problem was found by Bose in the year 1924 \cite{BOSE}. According to classical
mechanics, if we have $N$ particles which can stay in three different states and we denote with
$k_{i}$ the number of particles in the $i^{th}$state ($k_{1}+k_{2}+k_{3}=N$), the probability has a
combinatorial prefactor which is given by:
\be
P(k_{1},k_{2},k_{3})={
N! \over k_{1}!k_{2}!k_{3}!}\ , \label{CLASSICA}
\ee
i.e. by the number of combinations in which the $N$ particles may be divided in three sets. This
formula is sometimes called the Boltzmann counting, because it was used by Boltzmann.
According to quantum mechanics, if the particles are indistinguishable and they are Bosons, the wave
function is symmetric and there is only one quantum state for each set of occupation numbers.
Therefore in this case the correct formula is simply given by:
\be
P(k_{1},k_{2},k_{3})=1 \ . \label{CORRECT}
\ee
The way in which this formula was found is remarkable. Bose started by introducing the quantity
$n_{k}$ that are equal to the number of states with $k$ particles (i.e. $\sum_{i}
\delta_{k,k_{i}}=n_{k}$)\footnote{For example in the case where we have 4 states, 8 particles and
$k_{1}=3,\ k_{2}=2,\ k_{3}=2,\ k_{4}=1$, we have $n_{3}=n_{1}=1,\ n_{2}=2,\
n_{0}=0$. One obviously has that $\sum_{k} k \ n_{k}= N$.}. The main step by Bose was to suppose
that
\be
P(n_{0}, n_{1}, n_{2}) \propto {N! \over n_{0}! n_{1}! n_{2}! \ldots} \ . \label{MAGICA}
\ee
The amazing point is that Bose did not give any explanation for his formula (which has no classical
justification) and he declared that it is evident! According to Pais \cite{KP} it is reasonable to suppose
that Bose has written eq. \ref{MAGICA} in analogy with the classical Boltzmann counting (eq.
\ref{CLASSICA}), without realizing that the $n$'s and the $k$'s are very different objects. In other
words, his formula was based on a misunderstanding and his derivation was not correct. On the other
hand, it was impossible to derive the Bose formula (eq. \ref{MAGICA}) from what was known at that time.
Only two years later, Fermi extended the argument to the case of Fermions, and Dirac presented a
microscopical derivation of both statistics from the assumptions of symmetry (or antisymmetry) of
the wave functions. He was also the first to point out that the mysterious Bose's formula (eq.
\ref{MAGICA}) could have been written in a simple way as eq. \ref{CORRECT}.
\section{Were Rayleigh and Jeans correct?}
Statistical mechanics developed at a steady pace for the rest of the century.
One of the main
problems, the existence of phase transition in the canonical ensemble was a puzzle for a long time.
In a finite volume everything (e.g. the energy) is an analytic function of the temperature and
a phase transition cannot exist.
The modern solution (i.e. a phase transition is present only in the infinite volume limit) appeared quite
slowly: at the van der Waals conference (1937) there was a public discussion and 50\% of the
physicists expressed the opinion that a phase transition cannot be present in a formalism based on
the partition function \cite{WDW}. Fortunately the situation changed very fast; in 1943 Onsager
obtained its solution of the two dimensional Ising model, where a phase transition is present, and
in 1951 Lee and Yang presented a clear analysis of the mathematical mechanism at the basis of the
phase transitions.
In the later years we had good and bad news on the foundation of statistical mechanics.
The good new was that
Sinai proved that for a small number of hard spheres the microcanonical distribution is correct
(actually he proved much more).
The bad new was that Fermi, Pasta and Ulam took a chain of anharmonic oscillators, studied their
evolution on a computer and found that the microcanonical distribution is not reached. The
Hamiltonian they were considering was
\be
H=\sum_{i=1,N}\left({p_{i}^{2}\over 2}+ {(\phi_{i}-\phi_{i+1})^{2}\over 2 }
+g {(\phi_{i}-\phi_{i+1})^{4}\over 4}\right),
\ee
The corresponding equation of motions in the continuum limit are
\be
{\partial ^{2} \phi \over \partial t^{2}}={\partial ^{2} \phi \over \partial x^{2}}+g
{\partial \over \partial x}\left({\partial \phi \over \partial x}\right)^{3}
\ee
At the same time KAM theorem was proved. Very roughly speaking, KAM theorem states that in a
systems of $N$ weakly coupled oscillators, when the anharmonicity $g$ is small, there are $N$ constant
of motions, in the same way as in the uncoupled case at $g=0$. Only for large values of $g$ the
energy (and the momentum) is the unique conserved quantity.
Both papers show that for finite $N$ and not too large anharmonicity Rayleigh was correct and
equipartition law is not satisfied.
Many numerical simulations \cite{GALGANI} imply that
for finite $N$ there is an ergodicity threshold in $g$.
A relevant question is to establish what happens for $N$ going to infinity. One can present
hand waving arguments, suggesting that in the limit $N \to \infty$ at non-zero $g$ the system will
reach thermal equilibrium and an upper limit on the equilibration time ($\tau(g)$) is proportional to
$\exp(A/g)$; for times much larger that $\tau(g)$ an infinite system is at thermal equilibrium
\cite{PARISI1}.
The conclusion is that in the case of a generic weakly anharmonic system Jeans was correct:
the equipartition law is correct, but the time to reach equilibrium may be very large. This effect,
due to the difficulty in transferring energy from the low frequency modes to the high
frequency modes, could be observed experimentally (not only in numerical simulations) by heating a
crystal by pumping the acoustic phonons and looking for the energy transfer to optical phonons.
An interesting and open question is the precise behaviour of $\tau(g)$ for small $g$. There are no
doubts that $\tau(g)$ is strongly divergent for small $g$, however its precise form is not evident.
A careful numerical experiment has been done for the slightly different Hamiltonian \cite{PARISI2}:
\be
H=\sum_{i=1,N}\left({p_{i}^{2}\over 2}+{\phi_{i}^{2}\over 2} +{(\phi_{i}-\phi_{i+1})^{2}\over 2 }
+g {\phi_{i}^{4}\over 4}\right)\ .
\ee
The corresponding equation of motion in the continuum limit are
\be
{\partial ^{2} \phi\over \partial t^{2}}={\partial ^{2} \phi\over \partial x^{2}}+ m^{2} \phi +g
\phi(x)^{3}\ .
\ee
In this case the numerical data suggest a behaviour of the form
\be \tau (g)\propto g^{-4}\ee \ .
In
the original Fermi Pasta Ulam model other numerical simulations suggest \cite{FPU2} the
behaviour $\tau (g)\propto g^{-3}$.
Which is the correct behaviour? Numerical simulations may hint in some directions but they cannot be
conclusive if a theoretical understanding is lacking.
If the behaviour of $\tau(g)$ is a power law (as it looks like) it should be possible to control it in
perturbation theory. Unfortunately such an analysis has not be done and
the answer is not known.
\section{Conclusions}
In my opinion one of the most impressive fact is the large number of wrong statements,
wrong beliefs and incomplete proofs that can be found in these historical papers. Let me summarize
some of the
most impressive:
\begin{itemize}
\item Planck believed that the interaction of matter with the radiation was the key to explain
irreversibility without using molecular chaos.
\item
Planck did not realize that at equilibrium in classical mechanics the properties of the resonators
have no influence on the black body radiation.
\item Einstein deduced the existence of the photon from the properties of the black body
radiation showing that Wien's law implies that the radiation behaves as $N$ independent {\sl photons}
without noticing that Planck's law implies that as soon $N>1$ (i.e. in the whole non trivial
region) the photons are not independent.
\item
Bose wrote his main formula stating that it was evident and it did not realize that there were no
reasons for its validity.
\end{itemize}
It is quite likely that quantum mechanics could be discovered only in this illogical way. The gap
among classical and quantum mechanics is too large to be crossed in one step only. If people were
consistent and correct at all time, they would never move and they would make no progress, because
their aim, to find a classical explanation of quantum phenomena, was an impossible task. For
example, it is not surprising that Planck presented a wrong classical proof of the black body
radiation formula, because a correct proof of this statement does not exist. The influence of
classical mechanics was so strong that at the beginning people always tried to wrongly derive
quantum formulae in a classical contest.
These great men had the ability (conscious or unconscious, who knows?) to isolate the physical
relevant points, where progress could be done, not paying attention to the possible contradictions,
not exploring in their first papers the dangerous points where inconsistencies may arise. Only
after that the experimental correctness of the theory was verified, these points could be discussed
without destroying the theory.
It is difficult to predict what could have been the history of physics if Planck had make the
correct classical equilibrium statement, i.e. the Raleigh-Einstein-Jeans law. As stated by
Einstein, it is quite likely that Planck would not have discovered Planck's law. It is possible
that the efforts would have concentrated on one (or both) of the directions suggested by Raleigh and
Jeans. In this case, physicists would have tried to solve a problem, which a century later we have
not solved, moreover without having any ideas of the forces involved. In spite of the difficulties
this program could have been partially successfully and when Planck's law would be later discovered
it could be considered a nice fit, to be explained in a non-quantum context. It is likely that
quantum mechanics would eventually be discovered, but it would have been much more difficult.
It seems to me that science advances by small steps and therefore, if we have to construct a
completely new theory, at some stage we must go through inconsistencies. Only great men, that have
a strong physical intuition for what should be right and what should be wrong, could successfully
navigate in these dangerous waters.
\begin{thebibliography}{99}
\bibitem{KP} Two very good historical accounts are: T.S. Kuhn, {\sl Black Body Theory and the
Quantum Discontinuity 1894-1912}, Oxford University
Press (Oxford 1978) and
A. Pais {\sl `Subtle is the Lord. ' The Science and the Life of Albert Einstein}, Oxford University
Press (Oxford 1982). The interested reader should read them.
\bibitem{WDISP}W. Wien, Berl. Ber, 55 (1893).
\bibitem{ZERMELO} E.F.F. Zermelo, Ann. d. Phys., {\bf 57}, 485 (1896).
\bibitem{PRIS}M. Planck, Berl. Ber, 453 (1899).
\bibitem{PWIEN}M. Planck, Berl. Ber, 550 (1899).
\bibitem{WDIST}W. Wien, Ann. d. Phys., {\bf 58}, 662 (1896).
\bibitem{PDIS1}M. Planck, Verh. d. D. Phys. Ges, {\bf 2}, 202 (1900).
\bibitem{PDIS2}M. Planck, Verh. d. D. Phys. Ges, {\bf 2}, 237 (1900).
\bibitem{GIBBS} J.W Gibbs {\sl Elementary Principles in Statistical Mechanics. Developed with Special
Reference to the Rational Foundations of Thermodynamics} Scribner's (New York 1902).
\bibitem{ESTAT}A. Einstein, Ann. d. Phys., {\bf 9}, 417 (1902); Ann. d. Phys., {\bf 11}, 170
(1903);
Ann. d. Phys., {\bf 14}, 354 (1904).
\bibitem{ERE}P. Ehrenfest, Phys. ZS, {\bf 7}, 528 (1906).
\bibitem{NERST} W. Nerst, G\"{o}ttinger Nacht., 1 (1906).
\bibitem{REL} Lord Rayleigh, Phil. Mag, {\bf 49}, 539 (1900).
\bibitem{ERJEJ}A. Einstein, Ann. d. Phys., {\bf 17}, 132 (1905).
\bibitem{JEANS} J.H. Jeans, Nature, {\bf 72}, 101 (1905), Phil. Mag., {\bf 10}, 91 (1905).
\bibitem{DEB}P. Debye, Ann. d. Phys., {\bf 33}, 1427 (1896).
\bibitem{EPHOT}A. Einstein, Phys. ZS., {\bf 10}, 185 (1909).
\bibitem{PDIS3}M. Planck, Verh. d. D. Phys. Ges, {\bf 13}, 138 (1911).
\bibitem{ZERO}A. Einstein and O. Stern, Ann. d. Phys., {\bf 40}, 551 (1913).
\bibitem{BOHR} N. Bohr, H.A. Kramers and J.C. Slater Phil. Mag., {\bf 47}, 785 (1924).
\bibitem{BOSE}S.N. Bose, Phys. ZS., {\bf 26}, 178 (1924).
\bibitem{WDW}G.E.Uhlenbeck, {\sl Proceedings Einstein Centennial conference 1979} Addison-Wesley
(Reading 1980).
\bibitem{FPU} E. Fermi, J. Pasta, and S. Ulam,
Los Alamos Report LA-1940 (1955), in {\it Collected papers of Enrico
Fermi}, edited by E. Segr\'e, (University of Chicago, Chicago,
1965), Vol. 2, p. 978.
\bibitem{GALGANI} G. Benettin, L. Galgani and J.-M. Strelcyn, Phys, Rev. {\bf A 14} 2338 (1976);
Mechanica {\bf 15} 1 (1980).
\bibitem{PARISI1}F.~Fucito, F.~Marchesoni, E.~Marinari, G. Parisi, L.~Peliti, S.~Ruffo and
A.~Vulpiani, J.~Physique {\bf 43} (1982) 707.
\bibitem{PARISI2} G. Parisi. Europhys. Lett., {\bf 40}, 357 (1997).
\bibitem{FPU2}L. Casetti, M. Cerruti-Sola, M. Pettini and E.G.D. Cohen, Phys. Rev. {\bf E 55,} 6566 (1997).
\end{thebibliography}
\end{document}