Vous êtes sur la page 1sur 39

How Theories Begin: A Historical-Epistemological

Study of Planck’s Black-Body Radiation Theory


Massimiliano Badino
Max-Planck-Institut für Wissenschaftsgeschichte, Boltzmannstrasse 22, 14195 Berlin;
Universitat Autonoma de Barcelona, Department of Philosophy and Centre d’Història de
la Ciència, Cerdanyola del Valles, 08193 Bellaterra (Barcelona).

Abstract
This paper deals with the construction of theories. It is argued that the
conventional wisdom, according to which theories stem from the progressive
enlargement of a central core of assumptions and methods, does not do jus-
tice to the complexity of the historical process. It turns out to be more useful
to think about theories as a less stable and less hierarchized network of con-
ceptual resources historically situated in particular traditions. To show how
this view works, the case of the development of Planck’s theory of radiation
is analyzed. In particular, it is pointed out how the final theory emerges
from the tension between Planck’s intuitive account of irreversibility and the
formal tools used to encapsulate it at various stages of the process of theory
construction.
Keywords:

1. The problem of theory construction


1.1. A neglected issue
Philosophy of science has been mostly concerned with, and at times even
obsessed by, theories. In the last century, a great deal of scholarly work has
been devoted to analyzing how they are accepted, refuted, communicated,
modified, reduced, or generalized. In all these discussions, however, theories
have always been taken as a given. For how theories begin has traditionally
been considered more a psychological or sociological problem than a philo-
sophical one.

Preprint submitted to Studies in History and Philosophy of Science – Part ASeptember 25, 2012
This tendency seems to be a byproduct of the centrality of logic in phi-
losophy of science. A clear example is the philosophical account elaborated
by logical empiricism, to wit the syntactic view. According to the syntactic
view a theory is a set of sentences formalizable, at least in principle, in an
axiomatic system. These sentences can be related to empirical statements by
means of other sentences called correspondence rules.1 What distinguishes
a theory from a bunch of utterances is the tight logical relations among
them and the axiomatic structure. In this framework it is unsurprising that
the construction of a theory, with all its uncertainties, changes of tack, and
deadlocks has hardly attracted some attention. The attitude did not change
substantially with the other grand tradition of philosophy of science, the
so-called semantic view, although this claim needs qualification.
The upholders of the semantic view claim that theories are essentially a
collection of their models. Here models are intended in a Tarskian sense as
set-theoretical structures that make the theoretical sentences true.2 Although
the search of axiomatic order was no longer a central theme, this approach
was still dominated by the logical relations between sentences and models.
However, philosophers grew progressively unsatisfied with this set-theoretical
conception of model. Thus the semantic view gave rise, in the early 1980s,
to new pathways of research that highlighted the representational function
of a model over its logical features. This change of perspective dragged
the philosophical discussion out of the crystalline realm of logic to plunge it
down to the intricacies of working science. Instead of purifying their accounts
from mundane imperfections, philosophers started to realize that the issues
connected with interfacing abstract theories and concrete phenomena were
themselves genuine philosophical issues.
This realization opened up novel reflections on the relation between uni-
versal laws and their applications as well as on the cognitive structure of
theories.3 In this framework models assumed a different role. The focus was
shifted from the logical relations between the various elements of a theory to
the job carried out by the models in mediating between the formal part of a
theory and the details of concrete phenomena.4 As a consequence, philoso-

1
The literature on the syntactic view is immense. For a review see (Suppe, 1977).
2
(Morrison, 2007; Suppe, 1989); Bas van Fraassen has held a slightly different view in
(Van Fraassen, 1980).
3
See for example (Cartwright, 1983; Giere, 1988).
4
(Cartwright et al., 1995; Giere, 1999; Morrison & Morgan, 1999).

2
phers pointed out the localized, problem-oriented character of a model as
well as its being a potential resource for many theories at the same time.
These approaches have provided philosophers and historians of science
with a powerful conceptual reservoir to understand the essence of scientific
theories. However, they still fall short of giving an account of how theo-
ries begin. The reason is that these approaches mainly aim at producing
philosophical narratives embedded into scientific contexts and philosophical
narratives can only deal with theories that have reached a stable level of
consensus among the members of the scientific community. It is possible to
discuss the philosophical meaning of Newtonian mechanics only once there
is such thing as Newtonian mechanics, that is a corpus of concepts, laws,
methods upon which everybody, or almost everybody, agrees. This corpus
is what ought to be connected with natural phenomena and applications.
Philosophical narratives do not usually capture theories in flux.
For instance, in distinguishing between theories and models, Margaret
Morrison argues that theories are characterized by a theoretical core of gen-
eral laws and assumptions, while models include formal techniques and math-
ematical tools that can be shared by different theories.5 This view presup-
poses a stabilized theory because, as we will see in the case of Planck’s theory
of radiation, in the messy phase of theory construction mathematical tech-
niques might play a crucial role in determining the final shape of the theory.
A slightly different case, which deserves independent discussion, is Olivier
Darrigol’s recent “modular” view of theories.6 According to Darrigol theories
are composed of self-contained sub-theories (modules) that can be combined,
reorganized, and exported. Darrigol’s view nicely accounts for the transfer of
knowledge and practices between different theoretical fields with important
bearings on the issue of unity and disunity of science. Darrigol’s modules
are again textbook-like pieces of science. They are sets of procedures stable
enough to work as building-blocks of other theories. This view, I think, suc-
ceeds in capturing an important mechanism of theory construction namely
the use of pre-existing conceptual resources. However, Darrigol’s modules
are deprived of temporal dimension. They appear to remain unaltered in
the process of using them and, most importantly, they are wholly abstracted
from their historical tradition. Also Darrigol’s view, thus, seems more useful

5
(Morrison, 2007).
6
(Darrigol, 2008).

3
to understand the structure of theories, rather than their coming into being.
Once we have established that philosophical narratives concern stable
theories, the argument can be pushed a step farther. For a huge literature,
especially in sociology of science, has showed that a stabilized theory can
be remarkably different from the same theory under construction. In general
stabilization depends on factors as contingent as rhetorical skills, institutional
configurations, academic alliances, personal inclinations, and so on. It is
therefore necessary to construe an analysis, complementary to the social and
cultural analysis, able to make sense of the genuine philosophical problems,
especially concerning the dynamics of knowledge, generated by the process
of theory construction.

1.2. Representations and tools: The deep tension of theorizing


Neither the syntactic view, nor the semantic one, nor their successive de-
velopments exhaust modern philosophy of science. There is a further strand
that has thematized the dynamics of science: the so-called historical episte-
mology. Under the label historical epistemology fall several research lines,
which I will not attempt to unfold here.7 For my goal it suffices to condense
historical epistemology in three very general claims. The first claim is that
historical narratives are inseparable from epistemological problems encapsu-
lated in them. These problems transcend the concrete cases under analysis
and point at a higher, structural dimension. They represent, so to say, the
image of science that we can find even in very localized historical episodes.
The second claim is that the most effective way of dealing with the epis-
temological problems inherent to the development of scientific knowledge
is through the historical investigation of science as a cultural phenomenon.
These problems cannot be treated by the means of formal logic, because
science is embedded into a complex web of cultural values as well as social
practices and interacts with other systems of knowledge. This kind of epis-
temological problems cannot be isolated from the rich context from which
they stem. On the contrary, it is precisely the interaction of the several com-
ponents of this context that represents the epistemological question of the
development of science.

7
See for example (Wartofsky, 1979; Daston, 1994; Renn, 2006); useful overviews of the
debate on the various trends of historical epistemology are (Rheinberger, 2010; Feest &
Sturm, 2011; Stroud, 2011).

4
This does not mean that we have to refute all philosophical investigations
simply because they have assumed theories in a fictitious static state. For,
and this is the third claim, historical epistemology also deploys concepts,
metaphors, and results developed by the traditional philosophical discourse
with the goal of complementing and enlarging them. An additional point,
which is not a claim but rather a state of affairs, is that historical epis-
temology aims at constituting a resource both for the philosopher and for
the historian of science. Its analyses aim at illuminating not only the epis-
temological problems, but also the historical narratives in which they are
embedded.
In this framework, I will explore the problem of theory construction by
discussing Planck’s elaboration of the theory of black-body radiation. The
motivation for this choice is twofold. First, this episode is almost entirely
a purely theoretical story. It concerns conceptual arguments and counter-
arguments with experimental considerations only playing a marginal role.
Second, although we do not possess Planck’s own notes, his published papers
allow us to follow very closely the unfolding of his ideas, including the deep
reorganization of his theory that occurred after Boltzmann’s criticism.
The picture of a physical theory that emerges from this historiographical
analysis is very different from that usually adopted in philosophy of science.
As mentioned above, Margaret Morrison identifies the essence of a theory in
“the notion of a theoretical core: a set of fundamental assumptions that con-
stituted the basic content of the theory, as in the case of Newton’s three laws
and universal gravitation.” 8 This notion is instrumental to her distinction
between theory and model because “these core features are then represented
in the models as in the case of the linear harmonic oscillator which is derived
from the second law.” 9 This view seems quite popular among philosophers of
science. Although the details are obviously different, it appears to be com-
monplace to depict theories according to a principium individuationis, which
usually is conceived as a core of laws, assumptions, principles that define the
essence of the theory. Kuhn’s notion of paradigm as an established set of
practices and techniques to solve specific problems is also inspired by this
view.
I call this the concentric view of theories: it claims that there is such

8
(Morrison, 2007, 197).
9
(Morrison, 2007, 204).

5
thing as the essence of a theory that characterizes it univocally and from
which further applications and specializations emanate. This view is a useful
tool to analyze the structure of stable theories, or normal science in Kuhnian
terms. But from the standpoint of the development of scientific knowledge
several arguments can be raised against it. Here I will mention two very
general ones. Firstly, the theoretical core itself comes into being historically.
Newtonian mechanics did not exist before a certain time and existed there-
after. Hence, there is a development of the core. One might respond to this
point by arguing that although it is true that, as a corpus of knowledge, the
theoretical core is codified at a certain time and under contingent conditions,
nonetheless it is self-justifying in an important sense. For the theoretical core
is not only a set of assumptions and laws, but a glimpse on the real struc-
ture of the world. The theoretical core of a fundamental theory is at least
approximately true and this truth can work as a point of accumulation for
the several developments that lead to the codification. Appealing as it may
sound, this reply overlooks a crucial element. The development of scientific
knowledge depends importantly on what scientists believe. As sociology of
science has taught us since the early 1970s, one cannot explain the fact that
a community believes something by appealing to its truth. This move would
immediately introduce an asymmetry between explaining the truth and ex-
plaining the error.10 Other cognitive and epistemological factors are at stake
in believing besides the relation with the external world and it is precisely
the way in which these factors are intertwined with the picture of the world
that determines the historical development of the theoretical core.
Secondly, the concentric view draws an arbitrary line between the con-
struction of a theory as the progressive enlargement of the core, and the
revolutionary change. As a result, scientific revolutions become unintelligi-
ble or paradoxical. How can the development of a theory end up in a radically
new theory characterized by a radically different theoretical core? How can
a set of laws and assumptions generate something completely alien as the
result of problems set in terms of those laws and assumptions and elaborated
according to their own rules? Within the concentric view these questions can
find just the Kuhnian answer based on discontinuity and Gestalt switch.11
This view does not appear to be compatible with a unified picture of radical

10
See for example (Barnes, 1974).
11
On the paradoxical aspect of scientific revolutions see (Renn, 2006, 10-50).

6
change as a legitimate upshot of the usual process of theory construction.
For these reasons I think that the process of constructing a theory calls
for a different picture. I call it here the network view of theories. According
to this picture, there is no stable center that represents the essence of the
theory. Rather, a theory is a network of intersecting procedures, techniques,
practices, theoretical resources each of them with a specific tradition. When a
theory is created these resources, historically situated in traditions, combine
and, over time, might change their specific weight in the general architecture.
Even when the theory is developed by a single author, as in the case of
Planck, one has to look preferably at the interplay of traditions that are
involved and at the way in which their relative importance changes during
the process, rather than at persisting general assumptions. Only in this way
it is possible to move from the pure historical narrative of a very specific
episode to the level of epistemological problems.
However, the interplay of the formal resources sedimented in the theo-
retical traditions does not only respond to the necessity of solving specific
technical problems, but it is also constrained by the intended account. The-
ories are not just problem-solving devices. They are supposed to yield some
sort of understanding of the world as well. All that falls under a repre-
sentation of the essential, qualitative properties of the phenomenon studied,
or equivalently all that cannot be directly understood as a formal resource,
concurs to form the theoretical account. This notion encompasses models,
analogies, general methodological directives, which are, in turn, historically
situated. In his study of irreversibility, for example, Planck looked for a
non-probabilistic account of irreversibility: the state of thermal equilibrium
is reached without exception. This commitment constituted his intended ac-
count of irreversibility. However, the account is not as public as the formal
resources: it cannot be communicated and discussed with the peers. There-
fore it must be cast in terms of concrete problems using the available formal
resources. The tension implicit in the negotiation between the account and
the formal resources to express it is the driving force of theory construc-
tion. As we will see later on, after Boltzmann’s criticism Planck had to give
up his account of irreversibility formulated in terms of an electromagnetic
problem and recast it in terms of Fourier representation of the field. This
move implied a new role for traditional resources as well as a refinement of
the account. Account and formal tools are thus reciprocally constrained and
the interplay between their requirements determines the development of the
theory.

7
The resulting picture of a theory in flux is a complex entanglement of rep-
resentational and problem-solving obligations. The historical-philosophical
analysis of this entanglement is made even harder by two additional consider-
ations. First, contrary to Darrigol’s modules, formal traditions do not come
alone. They are the result of another negotiation between representational
and problem-solving activity and they usually encapsulate prior accounts.
For instance, Planck’s use of the Fourier representation came with a specific
tradition in radiation theory and Planck was effectively guided by the account
of the field implicit in it.12 Second, as I said above, the intended accounts
are not necessarily as public as the formal resources. The latter are custom-
arily coded in papers, textbooks, university courses, thus their loci are easily
recognizable. Instead, to reconstruct the representational work a theory car-
ries out for an author or a school requires to bring in a much broader and
often indirect web of texts and documents. This latter consideration yields a
further reason to focus on Planck’s black-body theory. For Planck was very
outspoken in his view of irreversibility and we possess documents from which
the account he was striving for can be reconstructed fairly precisely.
To sum up the point, the representational activity and the problem-
solving activity of a theory are tightly intertwined. The search for an ade-
quate account influences which resources are used and how, the way in which
problems are set, and the kind of acceptable solutions. In turn, the problems
generated by the formal apparatus of the theory constrain the account and
sometimes force a reconfiguration of it. In some extreme case, the quantum
of action being an apparent example, the formal resources can even generate
theoretical objects that the account cannot incorporate.

1.3. Overview of the paper


In section 2 I discuss the first phase of Planck’s theory of radiation. In this
phase Planck presented his view of irreversibility as a universal law of nature
and put forward his first argument to prove that the cavity radiation reaches
equilibrium irreversibly. This dream was wiped out by Boltzmann’s criticism
(section 3). As a consequence, Planck changed his argument for irreversibility
and refined his account. The reorganization of his theory went hand in hand

12
It may happen, of course, that a theoretical resource is picked up as a mere compu-
tational device, totally stripped of its historical meaning. This is the case of Planck’s use
of combinatorics. However, as we will see, even this instrumental choice has an epistemo-
logical consequence on the kind of theory eventually constructed.

8
with an extended use of the Fourier representation of the field. This formal
resource allowed him to create the conceptual space for his hypothesis of
natural radiation. I analyze this transformation in section 4. In section 5 I
tackle the second big crisis of Planck’s theory of radiation, the one that led
to the introduction of the quantum of action. This section will suggest some
considerations on the problem of revolutionary change. I argue that Planck
was not able to integrate the quantum in his account. For that reason this
theoretical object remained in epistemic isolation.

2. The first phase of Planck’s theory


2.1. A deterministic view of irreversibility
Irreversibility was the great underlying theme of Planck’s work, since the
time of his doctoral dissertation in 1879.13 In that work Planck claimed that
the second law of thermodynamics, according to which in thermal processes
entropy increases until it reaches a maximum value corresponding to the state
of thermal equilibrium, is a law of nature which holds without exceptions.
Planck also considered entropy as a universal quantity expressing nature’s
degree of preference toward equilibrium. When in the 1880s Planck moved
to research in physical chemistry, entropy became his main theoretical tool.
Planck’s predilection for entropy and general thermodynamics went to-
gether with his rejection of kinetic theory. Albeit not an opponent of the
atomistic hypothesis as such, he strongly disapproved kinetic theory for its
use of statistical arguments. Since the early 1870s Ludwig Boltzmann and
James Clerk Maxwell had been arguing that a foundation of thermodynamics
can be provided by combining the mechanics of microscopic particles with
statistical considerations. This view implied that the law of the increase of
entropy expressed just a very probable process, but by no means a deter-
ministic one. To this approach Planck moved two kinds of objections. First
he found that the mathematical difficulties involved in blending together
mechanics and statistics were disproportionate to the results that can be ob-
tained.14 Second, Planck thought that the statistical view made a wrong use
of probability. According to Boltzmann and Maxwell thermal processes tend
to equilibrium merely because it is the most probable state. In other words,

13
(Planck, 1879, 1958, I, 1-61).
14
(Planck, 1891, 1958, I, 372-381).

9
nature moves from less probable to more probable configurations. Planck
thought differently and his view on probability was summarized in a letter
to his friend and colleague Leo Graetz:15
Probability calculus can serve if nothing is known in advance,
to determine the most probable state. But it cannot serve, if an
improbable state is given, to compute the following [state]. That
is determined not by probability, but by mechanics. To maintain
that change in nature always proceeds from [states of] lower to
higher probability would be totally without foundation.
Planck’s point is that probability only comes in handy when one needs
to draw some inferences from an insufficient amount of information, but it
cannot be the deep-seated reason for the behavior of the system. This reason
comes from a deterministic dynamics, not from probabilistic considerations.
After years spent on thermodynamics and physical chemistry, in the mid-
1890s Planck suddenly started a completely new series of studies on elec-
tromagnetic heat radiation. Physical bodies are able to absorb and to emit
an electromagnetic radiation that depends only on the temperature. This
is called heat radiation and it is normally described by a distribution func-
tion that, for a given temperature, provides the amount of energy allocated
on each frequency component of the radiation. In general, this function de-
pends on the amount of energy absorbed and emitted in the process. To
simplify the analysis, in 1860 Gustav Kirchhoff introduced an ideal entity
called black-body, which is able to absorb entirely the radiation impinging
upon it. Kirchhoff also showed that, once the thermal equilibrium has been
reached, the radiation contained in a empty cavity is a good approximation
to a black-body.16
In the ensuing years several results contributed to narrow down the form
of the distribution function. On the experimental side, the careful observa-
tion of sunlight by means of interferometric techniques allowed physicists to
conclude that the wavelength carrying the maximum energy changed with
temperature by a constant value. Of this result, commonly called ‘displace-
ment law’, Wilhelm Wien gave a theoretical derivation in 1893.17 Further,

15
Planck to Leo Graetz, 23 May 1897.
16
(Kirchhoff, 1860, 277). For a history of the experimental side of the heat radiation see
(Kangro, 1970).
17
(Crova, 1880; Langley, 1886a,b; Wien, 1893).

10
experiments with a bolometer also established that the distribution function
decreased at both extremities of the spectrum, so suggesting that the curve
had a Gaussian-like form. Theoreticians also contributed to the advancement
of research on this topic. In 1884 Boltzmann gave a rigorous deduction of a
law inferred by Josef Stefan from experimental data, that is that the total
energy emitted by bodies is proportional to the fourth power of the temper-
ature.18 Finally in 1896 Wien obtained an expression for the distribution
function that satisfied the experimental data available.19 Wien’s argument
assumed that the microscopic particles, responsible for the emission of heat
radiation were distributed according to the Maxwell law and from it he could
obtain the Gaussian exponential factor that fitted the observations.
Thus, when Planck entered the field, some of the major problems seemed
already solved. Most importantly, the solution made ample use of the tech-
niques and the assumptions of kinetic theory. Why was then Planck inter-
ested in this problem in the first place?

2.2. Planck’s first argument


Unsurprisingly, Planck was not interested in the problem of heat radi-
ation as such. He wanted to use this case to bring fresh support to his
view of irreversibility. Being only temperature dependent, heat radiation is
essentially a thermodynamic phenomenon. From the 1880s it had become
customary to treat radiation as an electromagnetic wave, but the final result,
that is the distribution law, was independent of the underlying nature of the
phenomenon. The problem of black-body radiation had many qualitative el-
ements that defined the issue without the necessity of a commitment on the
non-observable microscopic processes. Planck was not so much interested in
the characteristic of radiation at equilibrium (i.e. the distribution law) as
he was in the process of reaching the equilibrium. He thought that he had
a way to prove that this process was strictly irreversible. Since radiation
is described by electromagnetic theory and since electromagnetic equations
are reversible, Planck concluded that it could have been possible “to trace
one-directional processes back to conservative effects.” 20
Planck’s theory of black-body radiation was developed in a series of five

18
(Stefan, 1879; Boltzmann, 1884).
19
(Wien, 1896).
20
(Planck, 1897b, 1958, I, 493).

11
papers that, for brevity, I will call here the Pentalogy.21 The Pentalogy
can be divided into two phases. The first three papers try to respond to
Boltzmann’s criticisms and put forward what I call Planck’s first argument for
irreversibility. The fourth and fifth papers follow Planck’s partial acceptance
of Boltzmann’s argument and contain the reorganization of his radiation
theory.
In the first paper of the collection Planck outlined his account of the
phenomenon. An empty cavity containing only electromagnetic radiation
maintains its initial state forever. The radiation waves expand from the
center to the walls, get reflected toward the center and so on. The state can
be changed only if we introduce in the cavity a device able to interact with the
field and thereby to modify the configuration of the system. There is no need
to introduce complicated assumptions on this device: it just needs to be an
oscillating object capable of resonating with waves and having “a certain one-
sided, equilibrating tendency.” 22 This object is called by Planck a resonator.
In studies carried out in the previous years Planck had already concluded that
the resonator changes the field because it cancels out the spatial fluctuation
in the energy intensity.23 Thus, Planck’s account of heat radiation consists
in the interaction between a microscopic, roughly determined resonator and
the field in form of electromagnetic waves; this interaction modifies the field
in the sense that it eliminates the uneven spatial distribution of the radiation
intensity so that, at the end of the process, the system will be in a perfectly
uniform state. Most importantly, this change is strictly irreversible.
To prove that the equilibration of cavity radiation has the character of
strict irreversibility is a threefold task whose steps Planck outlined in the
third installment of the Pentalogy.24 Firstly “a direct reverse of the process
must be excluded.” This means that if one makes the time-reversal of the
process of equilibration (i.e. one replaces t with −t in the equations describing
the process and also changes the direction of the magnetic field) the result
must be physically meaningless. Secondly, it also must be excluded that the
system comes back to a previously encountered state as an effect of a periodic
evolution. Finally, the final state of the process must have the character of
stationarity, that is it must remain forever.

21
(Planck, 1897b,c, 1898a,b, 1899)
22
(Planck, 1897b, 1958, I, 495).
23
(Planck, 1896, 1897a).
24
(Planck, 1898a, 1958, I, 509).

12
The third condition has been already discussed. In the final state of the
process described by Planck the energy intensity is uniformly distributed over
the cavity and it is inconceivable that the system breaks spontaneously this
uniform state. The second condition has to do with Zermelo’s recurrence
objection: for brevity’s sake I will skip the discussion of this point. For the
time being, I would like to focus upon the first condition that represents the
core of Planck’s account.
Planck’s most demanding challenge is to express formally, using the theo-
retical tools of electromagnetism, the exclusion of the time-reverse. Without
the resonator the description of the system appears very simple. As men-
tioned above, in this case the solution of the Maxwell equations is a superpo-
sition of spherical waves that expand from the center of the cavity (assumed
spherical for simplicity’s sake) and contract on the center again. One sets the
boundary conditions according to which the vector potential describing the
waves is zero at the walls and at the center and the problem is completely
defined. The introduction of the resonator changes the boundary conditions.
An electromagnetic resonator is similar to a mechanical mass attached
to a spring and it is described by essentially the same equation. The only
difference is that the resonator interacts with the field waves and thus its
behavior depends on an electric quantity called dipole moment. A resonator
vibrates at a certain characteristic frequency and its response is stronger
when the frequency of the impinging wave is very close to it. When stimulated
by the field the resonator emits a wave at the characteristic frequency. Thus,
after the interaction the total field is a superposition of the original wave and
the wave emitted by the resonator.25 Planck described the resulting field by
writing down explicit forms for the vector potential of the field and for the
dipole moment of the resonator. In both cases these forms are Fourier series
involving harmonics of the characteristic frequencies. I will come back to the
significance of the use of Fourier series in section 4. The point to highlight
here is how Planck concluded his argument for irreversibility from this formal
description.
In the process picture by Planck, a wave impinges on the resonator at the
center, gets diffused in the cavity and superposed with the secondary wave
emitted by the resonator. As we have seen, this secondary wave has the
property of canceling out the irregularity in the spatial distribution of the

25
The two waves have a phase difference related to the response time of the resonator.

13
intensity of the first wave. Planck wrote down the new boundary conditions
for this problem and calculated the supposed time-reversal of the process
above described. To do that he merely changed the signs in the series and in
the arguments of the trigonometric functions for the potential and the dipole
moment. It was very easy for Planck to show that this new process did not
satisfy the boundary conditions and therefore it could not be an acceptable
solution of the problem.
The upshot of this argument was the following: it is possible to set up
a perfectly defined electromagnetic problem (given by the electromagnetic
equations and suitable boundary conditions) involving the interaction be-
tween the radiation field and the resonator, such that the time-reversal of
a solution of this problem is not in turn a solution of the problem because
it does not fulfill the boundary conditions. Therefore, although the time-
reversal of the process can be formally written down, it has no physical
meaning. This is Planck’s first argument for irreversibility. This argument
is very strong: it rules out completely the possibility of a time-reversal. As a
consequence, the process described by Planck is a genuinely irreversible one.
Isn’t it?

3. Boltzmann’s criticism
3.1. The human side
The relation between Planck and Boltzmann had been complicated from
the very beginning. Since the end of the 1860s Boltzmann had been a trail-
blazer in the field of kinetic theory and statistical mechanics. His most impor-
tant results were an equation expressing the time behavior of the distribution
function of a gas (now called Boltzmann equation) and the H-theorem, both
dating back to 1872.26 The H-theorem shows that a special functional H of
the distribution function exists such that it increases until it reaches a maxi-
mum corresponding to the state of thermal equilibrium. Since the beginning
of his work in kinetic theory, Boltzmann had made use of probabilistic con-
cepts. Over the 1870s, he came to claim overtly what later became known as
the statistical view of irreversibility.27 According to this view, the increase of

26
(Boltzmann, 1872, 1909, I, 316-402).
27
The evolution of Boltzmann’s statistical ideas has been at the center of a heated
debate that lasts until today. For a more extended discussion see (Klein, 1973; Uffink,
2007; Brown et al., 2009; Badino, 2011).

14
entropy, and hence the reach of the thermal equilibrium, is not a strictly irre-
versible process, but just an overwhelmingly probable one. The fact that we
observe constantly entropy increasing processes and never entropy decreasing
ones is due to the high improbability of the latter.
I have already mentioned Planck’s negative evaluation of this line of
thought in 1891. Two more serious clashes between him and Boltzmann
occurred about the time Planck was starting out on his radiation theory.
The first occasion was the publication of Kirchhoff’s lectures on heat theory
edited by Planck. In a short note Boltzmann accused Planck of having com-
pletely misunderstood the meaning of the Stosszahlansatz with the outcome
that Kirchhoff’s arguments resulted misrepresented.28 Planck replied defend-
ing convincingly his interpretation, but eventually — perhaps voluntarily —
overlooking the point that Boltzmann was eager to make, to wit that some
sort of probabilistic assumption on the “disorder” of the system was implicit
in irreversible behavior. Before closing the dispute, Boltzmann made an-
other attempt to get his message across: “the assumption is necessary that
the state of the gas is disordered and remains so, namely [. . . ] the frequency
of each kind of collision can be derived from the laws of probability.” 29 This
small episode would become unexpectedly relevant some years later, as we
will see in section 4.2.
The tension between the two escalated again in 1896 when Ernst Zermelo
deployed a theorem recently proved by Poincaré to argue against Boltzmann’s
H-theorem.30 A bitter polemic ensued and Boltzmann had the impression
that he had to defend himself not only from Zermelo, but also from Planck,
Zermelo’s doctoral advisor, who was at the time occupying the influential
position of editor of Annalen der Physik. He wrote to Eilhard Wiedermann,
former student of his and son on Gustav Wiedermann chief editor of Annalen
to demand “(1) that Herr Planck does not delay the publication of my essay
[in response to Zermelo’s], (2) that not a word will be changed of it, (3) that
a reply to it does not appear on the same issue.” 31 Apparently, Boltzmann’s

28
(Boltzmann, 1894, 1909, III, 528-531); the Stosszahlansatz is a fundamental proba-
bilistic assumption of kinetic theory. More on this point on section 4.1.
29
(Planck, 1895, 1958, I, 442-444) and (Boltzmann, 1895b, 1909, III, 532-533).
30
(Zermelo, 1896). Zermelo’s famous objection was that a confined dynamical system
such as the gas must come back, sooner or later, to a state previously passed through.
According to Zermelo this amounted to a blatant violation of the H-theorem.
31
Boltzmann to Eilhard Wiedermann, 20 March 1896, (Höflechner, 1994, doc. 427).

15
demands were satisfied.
3.2. Boltzmann’s ultimate argument
Eventually, Planck behaved correctly in the dispute between Zermelo and
Boltzmann (and we know from the letter to Leo Graetz that he supported
Zermelo’s position, though not acritically). However, in the light of these
difficult relations, Planck’s theory of radiation must have appeared to Boltz-
mann as a glaring provocation. The account outlined in the first paper of the
Pentalogy was a head-on challenge to the statistical view of irreversibility and
Boltzmann, as a champion of that view, could not help reacting. Initially, he
made the general point that, since the equations of electromagnetism were
invariant under time-reversal, Planck could not hope to get irreversible be-
havior without introducing some special probabilistic assumption.32
Planck dismissed these objections and went on with the third paper of the
Pentalogy. He thought that the best reply was simply to state his argument
fully; hence he laid down the equations and the boundary conditions of the
problem, found a general solution that satisfied them in terms of expanding
wave and oscillating dipole moment of the resonator, determined the time-
reversal of this solution, and finally showed that this solution did not satisfy
the boundary conditions. Boltzmann was anything but impressed. He im-
mediately noticed a flaw in the argument and commented on it in a letter to
Felix Klein:33
Perhaps you have quickly glimpsed at the dispute I have pur-
sued with Herr Planck in the Berlin Academy under the heading
‘irreversible processes.’ Now in a communication of the 16 De-
cember, Herr Planck has reversed the exciting wave for a specific
case, but he has completely forgotten that the wave previously
emitted by the resonator must be reversed as well. From the
circumstances that he has obtained a totally counterintuitive for-
mula, he has not concluded that he was wrong, but rather that he
had found out a process whose reversal is not possible. I have sent
him directly my considerations, a move that will not necessarily
shorten the dispute; I’m curious to hear his response.

32
(Boltzmann, 1897a,b, III, 616 and 618-621). With a touch of irony, Boltzmann took
also the opportunity of pointing out to Planck that his theory fell under the argument of
Zermelo exactly like kinetic theory.
33
Boltzmann to Felix Klein, 12 February 1898, (Höflechner, 1994, Doc. 462).

16
The considerations that Boltzmann sent off privately to Planck were sub-
sequently published some weeks later.34 They amounted to an ultimate ar-
gument that crushed Planck’s first theory of radiation. As alluded to in the
letter to Klein, Planck had committed a gross mistake in the calculation of
the time-reversal of the solution.35 But Boltzmann went well beyond. He
accepted both the equations and the boundary conditions defined by Planck
and showed by a simple calculation that the correct time-reversal necessarily
satisfies them.
Thus, Planck was defeated with his own weapons. But the interpreta-
tions of this defeat were still quite divergent. For Boltzmann this argu-
ment proved beyond doubt that some sort of statistical assumption was
necessary and therefore that Planck’s account of irreversibility had to be
abandoned: irreversibility is an essentially statistical business. For Planck,
instead, Boltzmann’s argument simply indicated that it was impossible to
frame his intended account of irreversibility in terms of a purportedly chosen
electromagnetic problem and therefore that the tools used so far to express
that account had to be integrated with other resources. Only distinguishing
between the account of the theory and the tools to formulate it in terms of
solutions to specific problems is it possible to understand Planck’s dismissal
of Boltzmann’s argument and the reorganization of his theory that took place
in the fourth and fifth installment of the Pentalogy.

4. The reorganization of Planck’s theory


4.1. The mathematical paraphernalia
Although the general architecture of the theory was still standing, Planck
understood that he had to enlarge the arsenal of theoretical resources to in-
clude some of the concepts and techniques of kinetic theory. Kinetic the-
ory starts out from the analysis of the mechanical behavior of microscopic
particles and tries to draw consequences for macroscopic parameters (such
as temperature or thermal conduction) in terms of averages of microscopic
quantities. To link up individual quantities with averages, kinetic theory
deploys probability together with some suitable assumptions. One notable

34
(Boltzmann, 1898b, 1909, III, 622-628).
35
In modern parlance one can say that Planck reversed the free field solution but com-
bined it with the retarded solution of the wave emitted by the resonator. He should have
combined it with the advanced solution.

17
example is the already mentioned Stosszahlansatz. It states that the prob-
ability that a randomly chosen pair of molecules have certain velocities is
the product of the probabilities that each molecule has the respective veloc-
ities independently. In other words, molecules are supposed so disordered
that their velocities are uncorrelated. Essentially, the Stosszahlansatz selects
a particular way to connect individual probabilities with the corresponding
joint probability among the infinitely many a priori available. This is pre-
cisely the kind of job that the probabilistic assumptions are called to do. As
Planck had noted in his letter to Leo Graetz, they must allow us to overcome
an information deficit, in this case the gap existing between the microscopic
unknowable world and the macroscopic one.
To handle a similar gap in radiation theory, Planck relied on a specific
mathematical resource: the representation of electromagnetic quantities in
terms of Fourier series. Fourier series are a technique to represent periodic
behavior in terms of an infinite summation of trigonometric functions. They
represent a trajectory as a superposition of infinitely many harmonic mo-
tions. The amplitudes of these motions are called Fourier coefficients and
can be harmonic functions themselves. The arguments of the trigonometric
functions contain the integer multiples of a fundamental frequency, which
Planck called ‘partial vibrations’. The first important point to realize is
that, whereas Fourier series were extensively used in mechanics and acous-
tics, they were not so common in the theory of heat radiation. Wien, for
instance, never used them.
The second important point is that the Fourier series as a mathematical
technique also conveyed a very specific account of the electromagnetic field.
This account stemmed from the debate on the nature of white light at the
end of the nineteenth century. Many physicists thought that the frequencies
that made up white light were disposed with some sort of regularity. The
main argument for this claim was that the light components did not display
any mutual interference in optical experiments with gratings. However, in
1889 Louis Gouy challenged this conclusion. He argued that the observed
regularity could derive from the grating itself and that the radiation con-
sisted of irregularly distributed impulses representable by a Fourier series.36
This idea pleased very much Lord Rayleigh, who pointed out that, contrary
to what happens in acoustics, the light vibrations had to be ‘inextricably

36
(Gouy, 1886).

18
blended’, to the extent that the irregular pulses making up the radiation
could be represented by a law of error exactly as in the case of molecular
velocities in kinetic theory.37
Thus, built in the Fourier series there was a representation of field as
an irregular disposition of partial vibrations. The analogy between this ac-
count of field and the description of a gas given by kinetic theory is apparent.
Planck was aware of this similarity and he elaborated on it in a seldom quoted
paper published some years later: “the amplitudes and phases of the partial
vibrations are arranged in a completely irregular fashion, in particular two
neighbor partial vibrations are in no relation whatsoever, just as it is the case
for the velocities of two neighbor gas molecules.” 38 Partial vibrations are as
disordered and as uncorrelated as gas molecules. But there is a further ele-
ment of interest. The individual partial vibrations have no physical meaning
outside the Fourier series. The tradition of Gouy and Rayleigh maintained
that a perfectly monochromatic radiation, that is a radiation described by
a single-frequency component, is just a mathematical abstraction. In the
physical world there is no such thing as purely monochromatic light. Hence,
it is the totality of the Fourier series, but not the individual partial vibra-
tion, that is endowed with physical meaning. Planck made this point and its
consequences explicit in the forth paper of the Pentalogy: “a single member
of the Fourier series has absolutely not independent meaning as long as it
cannot be physically isolated and measured.” 39
This intertwining between the mathematical technique and the physical
account allowed Planck to give an important symbolic meaning to the Fourier
series. In representing the measurable quantities of his theory (the average
energy of the resonator and the field intensity) by this formal tool, he treated
the macroscopic quantity as the combination of very many microscopical en-
tities (partial vibrations) that have no independent physical meaning. The
ensuing macro/micro divide automatically black-boxes the microscopic level
and places the ‘physical meaningfulness’ on the macroscopic level. This re-
distribution of the ontological weight is key to the reorganization of Planck’s
theory. For it allowed for a definition a new technical problem that led to
the introduction of the hypothesis of natural radiation.

37
(Rayleigh, 1889, 462-464). This view was further developed in (Schuster, 1894); for a
general account see (Wien, 1909, 344-345).
38
(Planck, 1902, 1958, I, 763-773, citation on page 766).
39
(Planck, 1898b, 1958, I, 534).

19
Let us see more in detail how the Fourier series created the conceptual
space for natural radiation. After Boltzmann’s criticism, Planck decided to
go deeper in the analysis of the interaction between resonator and field. At
the same time, he decided to describe this interaction in terms of macroscopic
quantities only, that is the average energy of the resonator and the radiation
intensity. Now, a resonator interacts only with a radiation component very
close to its characteristic frequency, while intensity is composed of very many
different frequencies. It is therefore necessary to decompose the Fourier spec-
trum of the radiation intensity into components of different frequencies. But
according to the view of the field inherent to the Fourier series, it follows
that one cannot couple the resonator energy with single partial vibrations
because the latter have no physical meaning. Hence, Planck had to find a
new physically acceptable decomposition. To do so he introduced a sort of
ideal measurement process through an “analyzing resonator”. The analyzing
resonator works similarly to a tuning fork: it is tuned on a certain frequency
and Planck defined the spectral component of the radiation intensity at a fre-
quency as the energy emitted by the analyzing resonator at that frequency
when coupled with the field.40
Because of the way in which the spectral decomposition of the radiation
intensity has been conceived, the component at one frequency is not a sim-
ple radiation but it is “represented by a large number of neighbor partial
vibration.” 41 This means that it can in turn be represented by a Fourier
series. In other words, this supposedly simple radiation component is made
of very many microscopic vibrations, whose amplitudes and phases cannot
be known: “these [quantities] are not determined by the radiation intensity
at a certain frequency because in general many partial vibrations of the wave
furnish a contribution to the radiation intensity at a certain frequency.” 42
Instead of coupling the resonator energy, a macroscopic quantity, with the
radiation intensity, a macroscopic quantity, through a microscopic mecha-
nism, Planck used another macroscopic quantity, the spectral intensity, with
the result that the microscopic amplitudes and phase remained completely
undetermined. When then Planck tried to couple the resonator energy with

40
Technically, an analyzing resonator is characterized by its damping constant: it must
be small enough to respond to an infinitesimal interval around the chosen frequency, but
large enough to stop the vibration very soon.
41
(Planck, 1898b, 1958, I, 534).
42
(Planck, 1898b, 1958, I, 549).

20
the spectral intensity, he found himself in an unsurprising deadlock because
it turned out to be impossible to establish a relation between amplitudes
and phases of the two systems. Thus a fundamental underdetermination had
been set up: there were infinitely many ways to choose the relation between
microscopic amplitudes and phases. The tight functional relation between
this underdetermination and the hypothesis of natural radiation was candidly
admitted by Planck himself: “precisely in the gaps left by it, the hypothe-
sis of natural radiation finds its place; were these gaps not there, then the
hypothesis would be either superfluous or impossible because the process
would be completely determined without it.” 43 The gaps left by the coupling
between totally macroscopic quantities defined the conceptual space Planck
was looking for.

4.2. Natural radiation and molecular chaos


By coupling resonator energy and spectral intensity Planck had arrived
at an equation between the respective amplitudes and phases that contained
additional parameters. Depending on their values one had different couplings
compatible with the macroscopic constraints. To overcome this underdeter-
mination Planck chose all parameters equal to zero. He called this choice the
hypothesis of natural radiation (henceforth HNR). From a physical point of
view, the HNR is the electromagnetic translation of a conceptual resource
of Boltzmann’s kinetic theory: the hypothesis of molecular chaos. But, on a
closer look, it appears to be a very peculiar translation. To understand why
it is necessary to recapitulate the origin of the molecular chaos.
The concept of molecular chaos was the upshot of an intricate debate
that took place in 1894-95 between Boltzmann and some English experts
of kinetic theory. In 1877 Boltzmann had made clear the statistical mean-
ing of the H-theorem by arguing that equilibrium is a very probable state
because the processes resulting in it outnumber those resulting in a nonequi-
librium state. At that time, nobody seemed willing, or able, to challenge
Boltzmann’s account until, in a letter to Nature on 25 October 1894, E. P.
Culverwell moved a straightforward objection: if the laws of mechanics still
hold for microscopic particles, then the number of processes approaching the
equilibrium must be equal to the number of those going in the opposite direc-
tion, the latter being just the time-reversals of the former. Then one cannot

43
(Planck, 1958, I, 475).

21
conclude that the former outnumber the latter. But then, if the preference
for equilibrium cannot be conceived in terms of a larger number of processes,
he concluded, “will some one say exactly what the H-theorem proves?’44 Cul-
verwell’s apparently naive argument was received with some embarrassment.
The ensuing debate showed that, Boltzmann’s authority notwithstanding,
there was little consensus on this topic among the experts. Four kinds of
answer to Culverwell’s riddle were proposed.
First there was the argument that the time-reversal must be rejected
because the H-theorem requires the Stosszahlansatz and this assumption does
not hold for reverse collisions.45 Such an argument would reduce the H-
theorem to a purely mathematical statement: the assumptions it relies on
do not work, by definition, for the time-reversal. The physical meaning
of the theorem, which was what Culverwell’s question was about, would
disappear from the scene. Furthermore, it seemed inappropriate to explain
irreversibility for mechanical system by claiming that ‘natural’ phenomena
follow unmechanical assumptions.46 Finally, this argument implies that if a
state is disordered, then its time-reversal is ordered and therefore it must
be discarded.47 In this case, Culverwell’s objection that there are as many
ordered states as disordered ones would still stand because it is a completely
arbitrary choice between one state and its reverse.
A second popular argument was based on mechanical instability. Gas
collisions are such a complicate business that the precise time-reversal of a
state required for the H-theorem to be violated would be like riding a bicycle
backward: theoretically possible, but practically unfeasible.48 The smallest
variation from the exact sequence of collisions would bring the system back
to the normal, entropy-increasing evolution. This idea was at times used by
Boltzmann himself but never on a regular basis.
A third argument had a more mathematical character, but it can be
clarified with a metaphor. The function H behaves like a tree that extends
from a common trunk to different branches. It is true that for each path
that climbs from the trunk to the tip, one can make the opposite path and
this is precisely Culverwell’s objection. However, when one is at some stage

44
(Culverwell, 1894).
45
(Burbury, 1894; Bryan, 1894a).
46
(Culverwell, 1895a).
47
(Jeans, 1903).
48
(Bryan, 1894b,c, 1895; Burbury, 1894).

22
of the climbing, there are still many branches going up, but only one way
to go down (the reverse of the path hitherto made). The H-theorem, thus,
refers to the behavior of the system at a certain arbitrary nonequilibrium
state, not to the whole spectrum of possible evolutions. Culverwell found
this argument, construed by Boltzmann, very convincing.49
But Boltzmann also developed a fourth, more physical, argument hinging
upon the concept of molecular chaos. He anticipated this concept during
the above mentioned polemic with Planck on Kirchhoff’s lectures and it is
likely on that occasion that Planck started reflecting on chaos assumptions.50
However, the concept was officially introduced in the first volume of the
Gastheorie and never used again since.51 Boltzmann’s contemporaries found
it extremely puzzling.52 This is unsurprising if we consider that, to elaborate
this concept, Boltzmann selected and reorganized in a personal way elements
of the complex debate previously analyzed.
For Boltzmann molecular chaos (or molecular disorder) holds in a gas
when molecules are not arranged so as to collide in a specific preconceived
way:53

[I]f we choose the initial configuration on the basis of a pre-


vious calculation of the path of each molecule, so as to violate
intentionally the laws of probability, then of course we can con-
struct a persistent regularity or an almost molecular disordered
distribution, which will become molecular ordered at a particular
time.

The key is the concept of ‘conspiracy’ implicit in the notion of microscopic


order. Molecular chaos occurs when it is impossible to guess beforehand the
result of the collision process and we have to rely on probability. As the quo-
tation shows, we can turn any state into an ordered one. Therefore, a time-
reversal is an ordered state not in virtue of the arrangement of the molecules,

49
(Boltzmann, 1895d,c,a; Culverwell, 1895b). Subsequently Boltzmann formalized this
argument with the notion of H-curve (Boltzmann, 1898a, 1909, III, 629-637).
50
(Boltzmann, 1895b).
51
(Boltzmann, 1896, sect. 3 and 6).
52
Very telling in this respect is James Jeans’ reaction, who distinguished between
a working version of molecular chaos (that is a probabilistic assumption such as the
Stosszahlansatz) and its physical meaning, which remained mysterious (Jeans, 1903).
53
(Boltzmann, 1896, sect. 3, italics added).

23
but because the operation of time-reversing introduces a ‘preparation’ in the
state. From that moment on, molecules will collide in a predetermined way.
As a result, Boltzmann stresses that the reversal only affects the observed
evolution:54

Consider any motion for which H decreases from time t0 to


time t1 . When one reverses all the velocities at time t0 , he would
by no means arrive at a motion for which H must increase; on
the contrary, H would probably still decrease. It is only when
one reverses the velocities at time t1 that he obtains a motion for
which H must increase during the time interval t1 − t0 and even
then H would probably decrease again after that, so that motions
for which H continually remains very near to its minimum value
are by far the most probable.

Boltzmann’s crucial point is that the distinction between chaos and or-
der is not inherent in the state. This distinction must rather emerge from
the general evolution of the system. The consequence of this point is the
following. Since an ordered state is not prohibited in itself, it can result
from a purely probabilistic evolution. An example will clarify this point. We
can throw a coin one thousand times with heads up and without making it
flip. We will obtain a sequence of one thousand heads, but the process ob-
viously violates the hypothesis of molecular chaos because we arranged the
system to get a predetermined result. However, it is perfectly possible, albeit
highly improbable, to obtain the same sequence of one thousand consecutive
heads by flipping the coin. In other words, Boltzmann’s molecular chaos does
not cancel out the reversed state, it prohibits the time-reversal as a ordering
operation. This trait is essential to understanding the difference between
Boltzmann’s molecular chaos and Planck’s HNR.
Let us now come back to Planck’s original problem. We have seen that
Planck had constructed a precise formal problem, the underdetermination of
the microscopic amplitudes and phases, by means of the resources of the the-
ory. To this problem he had given a solution in terms of a formal hypothesis
that he had called HNR. However, this formal solution is tightly connected
with a specific account of chaos that Planck has in mind and that, in turn,
is functional to his account of irreversibility.

54
(Boltzmann, 1896, sect. 6).

24
Planck’s view of chaos is best expressed in the fifth paper of the Pen-
talogy.55 He distinguished two realms of physical reality. On the one side,
the microscopic realm made of quantities that vary very rapidly, such as
amplitudes and phases of partial vibrations. These quantities are beyond
our empirical control because they change value many times in the typical
time scale of our experiments. Their behavior is as chaotic as the molecular
motion: they escape our knowledge and our practical interest. What we can
have access to are the quantities that populate the macroscopic realm, such
as radiation intensity, which change comparatively slowly over time. Given
these circumstances, we have to introduce a simplifying assumption such as
the Stosszahlansatz, that is the HNR. The HNR, Planck argued, amounts
to saying that since the amplitudes and phases change so rapidly and so
randomly, at the time scale of our interest they can be replaced by average
values, to wit macroscopic quantities. We can simply put aside the very
complicate and unknowable evolution of amplitudes and phases. Behind this
view is the belief that macroscopic quantities are the object of physics, while
the microscopic realm is little more than a useful support. As a consequence,
the HNR is meant to warrant that the microscopic quantities ‘behave well’
with respect to the observable reality. They must simplify our calculations
and allow us to derive an accurate description of the observed nature. Ul-
timately, we are not interested in the microscopic realm and the HNR is an
effective way to black-box it.

4.3. Planck’s second argument


Thus the HNR is supposed to cancel out all ‘odd’ behaviors. It says that
the variation of amplitudes and phases is so complicate that they might as
well be replaced by their averages. But averages are macroscopic quanti-
ties and macroscopic quantities display strict irreversibility. Hence, Planck
concluded, the action of the HNR is to eliminate all those weird ‘anti-
thermodynamic’ behaviors that in reality never occur at macroscopic level.56
Planck’s understanding of the role of chaos assumptions is thus profoundly
different from Boltzmann’s. As we have seen, the molecular chaos is compati-
ble with the probabilistic appearance of low entropy states. On the contrary,

55
(Planck, 1899, 1958, I, 560-600).
56
Planck will be explicit on this point later, in the first edition of his lectures on heat
radiation (Planck, 1906, 134-135).

25
for Planck the HNR must conciliate the microscopic and the macroscopic
realms by wiping out of the map the unnatural entropy-decreasing processes.
From the problem-solving perspective the HNR does an excellent job. It
provides a simple relation between the spectral intensity and the resonator
energy which describes completely the interaction process. Here Planck put
forward his second argument for irreversibility. This argument, differently
from the first one, hinged on the concept of entropy. To prove the irre-
versibility of the process Planck showed that a function could be defined
that increases monotonically until the equilibrium is reached. This function
represented the electromagnetic analogue of thermodynamic entropy. How-
ever, the analogy was purely formal. Planck had no condition to apply in
order to get a concrete expression for entropy. He therefore decided to anchor
the concept to the best empirically confirmed result of heat radiation: the
distribution law obtained by Wien.
Working backwards from the Wien distribution, Planck obtained a log-
arithmic expression involving energy and frequency. Applying this formula
to resonator and field he showed that the total entropy is a monotonically
increasing function. Until this point his analysis was completely independent
of the temperature of the cavity. Since temperature is defined for the equi-
librium state, Planck considered the maximum of entropy and, after defining
the temperature, he found the Wien law. Once again the solution depended
essentially on the way in which the problem was defined. However this proce-
dure was very problematic.57 However, the point I want to highlight concerns
the tight relation between Planck’s second argument and the Wien law. At
that point of the argument, this latter was the only support to Planck’s en-
tropy concept. Without Wien’s law, there is no entropy and therefore no
proof of irreversibility. The fate of Planck’s account of irreversibility was
bound to the fate of that particular formula.

57
For example, in evaluating the temperature Planck considered a ‘virtual’ displacement
of the field energies among the different frequencies. But the resonator only interacts with
the field component vibrating at the characteristic frequency, hence Planck’s theory gave
no physical mechanism to explain how the energy could be redistributed over different
frequencies.

26
5. After the Pentalogy: The quantum hypothesis
5.1. The March paper
Planck’s theory of radiation had established a strong connection between
Wien’s law and the entropy function, but this connection was still shaky.
Planck needed some proof of uniqueness for the entropy function. After
all, Wien’s law was compatible with infinitely many other possible entropy
functions. This issue was tackled in a paper published in March 1900.58 In
the introduction, Planck returned to the fundamental distinction between
microscopic and macroscopic realms. He pointed out that the resonator
field had to be considered a macroscopic entity because “it consists of a
superposition of very many individual oscillations with almost equal period
and constant amplitudes and phases.” 59 For this reason the notion of disorder
can be applied to an individual resonator. This preliminary consideration set
the stage for his new, completely thermodynamical argument.
Planck assumed a system of resonators and field in equilibrium. If the
resonator energy is slightly changed, the system will tend to come back to the
equilibrium value, so triggering an entropy-increasing process. In the evalu-
ation of the entropy variation Planck made a surprising move. He expanded
in Taylor series the entropy function and found the second derivative of the
resonator entropy with respect to the energy, a quantity that never appears
in thermodynamics. After expressing the entropy variation in terms of the
second derivative, Planck found another expression for the same variation
and showed that the second derivative must be a simple function of the en-
ergy.60 A double integration yielded immediately the formula already found
in the Pentalogy.
Thus Planck had derived the needed entropy function using only thermo-
dynamic ingredients and a pinch of analytical techniques. From his point of
view this argument was everything he could possibly desire: it was general,
cogent, and completely independent of microscopic assumptions of any sort.
However, it also implied an even stronger commitment. The entire formal
representation of Planck’s account of irreversibility depended now not only

58
(Planck, 1900a, 1958, I, 668-686).
59
(Planck, 1900a, 1958, I, 674).
60
More precisely, the second derivative of the entropy must have the form −c/U , where
c is a constant and U the resonator energy.

27
on the validity of Wien’s law, but on a precise expression for the second
derivative of the entropy with respect to the energy.

5.2. Combinatorics
Experimenters started to raise doubts about Wien’s law when Planck’s
theory was still developing. In a series of measurements carried out in Febru-
ary 1899, Otto Lummer and Ernst Pringsheim found minor deviations from
the predicted values of energy. These deviations became more and more sys-
tematic at higher temperature and larger wavelength. By the time Planck
was elaborating his March paper, Lummer and Pringsheim presented new
outcomes from which they concluded that “the Wien-Planck spectral equa-
tion does not represent the black radiation measured by us.” 61
At the beginning, Planck was not too worried by these results because
none of them seemed to imply a radical change in the radiation law. More-
over, the small modifications that could have been envisioned, were still com-
patible with the expression for the second derivative of the entropy he had
found in the March paper. The situation changed radically in early October,
when Planck was privately informed by Heinrich Rubens that his new exper-
iments at large wavelength showed a remarkable deviation from Wien’s law.
Upon examining the data, Planck realized that they called for a revision of
the second derivative of the entropy. This result was a disaster for Planck’s
theory. His general and straightforward March paper had eventually made his
argument for irreversibility too rigid to tolerate the abandonment of Wien’s
law. The bond with the second derivative was too strong to allow Planck’s
argument to be patched up with some piecemeal measure. An injection of
new conceptual resources was needed.
To save his view of irreversibility Planck had to find a derivation for the
new radiation law. It was again a matter of integrating a problem-solving
task with a general account. It is well known that he found the tools for his
new derivation in Boltzmann’s 1877 combinatorial procedure.62 The relation
between Boltzmann’s combinatorial procedure and Planck’s appropriation of
it has been at the center of a very complex debate, which is still open today.
I will not enter into this debate here.63 For my point it is sufficient to look

61
(Lummer & Pringsheim, 1900, 171).
62
(Boltzmann, 1877, 1909, II, 164-223).
63
The origin of the debate is (Klein, 1962; Kuhn, 1978; Needell, 1980; Kuhn, 1984); a
modern perspective can be found in (Darrigol, 2001; Gearhart, 2002; Badino, 2009).

28
at the impact that this new set of problem-solving tools had on Planck’s
account.
Boltzmann’s combinatorial procedure can be described briefly. It amounts
to leaving aside all difficulties related to the collision mechanism and to focus
on the gas molecules simply as receptacles of energy. The total energy of the
system gets partitioned among individual molecules. Using combinatorics we
can calculate the number of ways in which this partition can occur. However,
an energy distribution does not concern individual molecules, but how many
molecules have the same amount of energy. It is thus natural to define
the probability of a distribution as the number (or the relative frequency)
of individual energy allocations compatible with it. This number can be
found by combinatorial techniques and Boltzmann managed to show that
the equilibrium state was by far the most probable in this sense.
In his derivation of the new distribution law, Planck followed Boltzmann’s
procedure as close as possible.64 With an important difference: while Boltz-
mann meant this procedure as an illustration of the statistical meaning of
the entropy, Planck used it simply as a computational device to express the
resonator entropy as a function of the energy. He considered a system formed
by the radiation field and many resonators with different characteristic fre-
quencies. We are interested in how much energy is ascribed to each frequency,
or collectively to the resonators working at that frequency. To calculate the
number of ways of distributing the energy over the resonators one has to
make two moves. First, one has to assume that each individual allocation of
energy is as probable as any other. No resonator is privileged in receiving
energy. Secondly, one has to divide up the energy into an integer number
of finite, elementary quantities. For reasons related to some conditions on
the entropy function, Planck chose to define this elementary quantum of en-
ergy as  = hν, where ν is the frequency of the resonator and h a universal
constant that would come to be called Planck’s constant.
This procedure provided Planck with an expression for probability con-
taining the resonator energy. To relate it to the entropy Planck condensed
the essence of Boltzmann’s combinatorial technique in what he called ‘the
Boltzmann principle’: S = k log W , where S is the entropy, W the prob-

64
(Planck, 1900b, 1958, I, 698-706); see also (Planck, 1901, 1958, I, 717-727). A com-
parison between Boltzmann’s and Planck’s use of combinatorics can be found in (Badino,
2009).

29
ability and k a proportionality constant nowadays known as Boltzmann’s
constant. By maximizing the probability and bringing in the temperature,
Planck arrived easily at the correct distribution law. But for his general ac-
count the important question was to incorporate the three key ingredients
that came with combinatorics: the Boltzmann principle, the equiprobability
of the allocations, and the quantum of energy hν.
The Boltzmann principle essentially expressed a relation between prob-
ability and entropy that is mediated by the fundamental disorder of the
microscopic realm. But disorder had been already introduced in the Pental-
ogy:65

Entropy means disorder, and I thought that one should find


this disorder in the irregularity with which even in a completely
stationary radiation field the vibrations of the resonator change
their amplitude and phase [. . . ]. The constant energy of the sta-
tionary vibrating resonator can thus only be considered to be a
time average, or, put differently, to be an instantaneous average
of the energies of a large number of identical resonators which are
in the same stationary radiation field, but far enough from one
another not to influence each other directly. Since the entropy
of a resonator is thus determined by the way in which the en-
ergy is distributed at one time over many resonators, I suspected
that one should evaluate this quantity by introducing probability
considerations into the electromagnetic theory of radiation.

The Boltzmann principle is therefore another tool to deal with the cru-
cial distinction between microscopic and macroscopic realms. Analogously,
Planck argued that the condition of equiprobability “can be understood as a
more detailed definition of the hypothesis of natural radiation.” 66 However,
the quantum of energy is a completely different matter. It comes with the
combinatorial procedure, but it was not in Boltzmann’s original version of it.
In the second part of his 1877 paper, Boltzmann systematically replaced the
summations over discrete elements of energy with integrals over infinitesimal
quantities. Planck could not do the same. The expression of the quantum of

65
(Planck, 1900b, 1958, I, 698-699).
66
(Planck, 1900b, 1958, I, 704).

30
energy is determined by dimensional considerations as well as by the necessity
of a second universal constant besides Boltzmann’s constant.
In other words, the quantum of energy was an object generated by the
problem-solving activity implicit in Planck’s theory but it remained “epis-
temically isolated” in his account.67 The interplay of account and formal
tools that constituted Planck’s system of knowledge in radiation theory was
able to generate the quantum as a way to treat a particular problem, but
it was incapable to integrate it with the remainder of the knowledge. From
this point of view it becomes easy to understand the profound discontent
that Einstein, Lorentz, Ehrenfest, and other experts of statistical mechanics
felt with Planck’s theory. Incapable to find a place for the quantum and
unwilling to be wedded to the intricacies of the microscopic realm, Planck
considered the combinatorial procedure as a mere formal trick to derive the
distribution law.68 His main target remained the argument for irreversibility,
which was still safe with the new definition of entropy. For the rest of the
community, though, the quantum was a riddle, an element of tension that
called for a new system of knowledge.69

6. Epilogue: Planck’s theory as a network


Planck’s theory of radiation is constituted by a complex network of re-
sources belonging to electrodynamics, thermodynamics, and kinetic theory
of gas. Each of these resources come with its own particular tradition, do-
main of application, and implicit account. The construction of the theory
consists in deploying these resources within the tension between the account
of irreversibility Planck strives to reach and the concrete physical problems
in terms of which this account ought to be cast. The notion of a theoretical
core is here useless to understand the construction process. Both the account
and the formal arsenal deployed by Planck changed importantly during the
elaboration of the Pentalogy. We cannot point at any set of specific as-
sumptions or formal tools that steadfastly drove the development of Planck’s
ideas. One can appeal, to be sure, to Planck’s predilection for thermody-
namic arguments and a certain view of irreversibility. But this predilection

67
On the concept of epistemic isolation applied to the quantum hypothesis see (Büttner
et al., 2003).
68
(Needell, 1980; Badino, 2009).
69
For an analysis of Einstein’s reaction to this riddle see (Büttner et al., 2003).

31
falls way short of representing a viable theoretical core. It represents, at
best, some vague methodological directive rooted in Planck’s specific train-
ing and research style. It helps explain some contingent choices made by
Planck, but it tells us very little about the epistemological aspects of theory
construction. The network view, I think, captures better the messy process
of gathering, re-elaborating, and reconfiguring the resources inherent in the
theory construction. It starts from a weaker notion of theory, but it is more
amenable to providing a picture of the epistemological tensions going on in
theory construction.
Moreover, in relying on the consideration that when theories begin vir-
tually everything is in flux, the network view highlights three important
points. First, that in the historical-epistemological analysis of a theory we
ought to unfold the entwining of cultural traditions that contribute to it,
trying to maintain their historical dimensions. For this dimension plays a
role in shaping the content of the theory. Second, these traditions are not
arranged into a settled hierarchy, but form a network in which their roles
and specific weights can change at any time. Third, precisely because the
development is not driven by a central theoretical core, the interplay of the
resources might generate objects that cannot be immediately integrated in
the account. The quantum is an example of a theoretical entity produced by
Planck’s theory, but not integrable in it. The case of the quantum does not
need to be the rule. The larger point here is that unintended consequences
or unexpected interpretations are possible when the dynamics is no longer
dominated by a theoretical core, but it follows the tension between account
and formalization.

Acknowledgments

This paper has been written as part of the History of Quantum Physics
Project of the Max-Planck-Institut für Wissenschaftsgeschichte and the Fritz-
Haber-Institut der Max-Planck-Gesellschaft. A preliminary version has been
presented at the conference “Continuity and Discontinuity in the Physical
Sciences Since the Enlightenment”, American Institute of Physics, July 2011.
I am indebted to Guido Bacciagaluppi, Stefano Bordoni, Marta Jordi, and
Jürgen Renn for reading and commenting the manuscript.

32
References
Badino, M. (2009). The Odd Couple: Boltzmann, Planck and the Application
of Statistics to Physics (1900-1913). Annalen der Physik , 18 , 81–101.

Badino, M. (2011). Mechanistic Slumber vs. Statistical Insomnia: The Early


Phase of Boltzmann’s H-theorem (1868-1877). European Physical Journal
H , 36 , 353–378.

Barnes, B. (1974). Scientific Knowledge and Sociological Theory. Boston:


Routledge.

Boltzmann, L. (1872). Weitere Studien über das Wärmegleichgewicht unter


Gasmolekülen. Sitzungsberichte der Akademie der Wissenschaften zu
Wien, 66 , 275–370.

Boltzmann, L. (1877). Über die Beziehung zwischen dem zweiten Hauptsatze


der mechanischen Wärmetheorie und der Wahrscheinlichkeitsrechnung re-
spective den Sätzen über das Wärmegleichgewicht. Sitzungsberichte der
Akademie der Wissenschaften zu Wien, 76 , 373–435.

Boltzmann, L. (1884). Ableitung des Stefan’schen Gesetzes betreffend die


Abhängigkeit der Wärmestrahlung von der Temperatur aus der elektro-
magnetiscen Lichttheorie. Annalen der Physik , 22 , 291–294.

Boltzmann, L. (1894). Über den Beweis des Maxwellschen


Geschwindigkeitsverteilungsgesetzes unter Gasmolekülen. Annalen
der Physik , 53 , 955–958.

Boltzmann, L. (1895a). Erwiderung an Culverwell. Nature, 51 , 581.

Boltzmann, L. (1895b). Nochmals das Maxwellsche Verteilungsgesetz der


Geschwindigkeiten. Annalen der Physik , 55 , 223–224.

Boltzmann, L. (1895c). On Certain Questions of the Theory of Gases. Nature,


51 , 413–415.

Boltzmann, L. (1895d). On the Minimum Theorem in the Theory of Gases.


Nature, 52 , 211.

Boltzmann, L. (1896). Vorlesungen über Gastheorie volume I. Leipzig: Barth.

33
Boltzmann, L. (1897a). Über irreversible Strahlungsvorgänge I. Sitzungs-
berichte der Preussischen Akademie der Wissenschaften, 2 , 660–662.

Boltzmann, L. (1897b). Über irreversible Strahlungsvorgänge II. Sitzungs-


berichte der Preussischen Akademie der Wissenschaften, 2 , 1013–1018.

Boltzmann, L. (1898a). Über die sogenannte H-Kurve. Mathematische An-


nalen, 50 , 325–332.

Boltzmann, L. (1898b). Über vermeintlich irreversible Strahlungsvorgänge.


Sitzungsberichte der Preussischen Akademie der Wissenschaften, 1 , 182–
187.

Boltzmann, L. (1909). Wissenschaftliche Abhandlungen. Leipzig: Barth.

Brown, H. R., Myrvold, W., & Uffink, J. (2009). Boltzmann’s H-theorem,


its Discontents, and the Birth of Statistical Mechanics. Studies in History
and Philosophy of Modern Physics, 40 , 174–191.

Bryan, G. H. (1894a). The Kinetic Theory of Gases I. Nature, 51 , 176.

Bryan, G. H. (1894b). The Kinetic Theory of Gases II. Nature, 51 , 152.

Bryan, G. H. (1894c). The Kinetic Theory of Gases III. Nature, 51 , 176.

Bryan, G. H. (1895). The Assumption in Boltzmann’s Minimum Theorem.


Nature, 52 , 29–30.

Burbury, S. H. (1894). Boltzmann’s Minimum Function. Nature, 51 , 78–79.

Büttner, J., Renn, J., & Schemmel, M. (2003). Exploring the Limits of Classi-
cal Physics: Planck, Einstein, and the Structure of a Scientific Revolution.
Studies in History and Philosophy of Modern Physics, 34 , 37–59.

Cartwright, N. (1983). How the Laws of Physics Lie. Oxford: Oxford Uni-
versity Press.

Cartwright, N., Suarez, M., & Shomar, T. (1995). The Tool Box of Science. In
W. E. Herlfel, W. Krajewski, I. Niiniluoto, & R. Wojcicki (Eds.), Theories
and Models in Scientific Processes (pp. 137–149). Amsterdam: Rodopi.

34
Crova, A. (1880). Etude des radiations emises par les corps incandes-
cents. Mesure optique des haute temperatures. Annales de Chimie et de
Physique, 19 , 472–550.

Culverwell, E. P. (1894). Dr. Watson’s Proof of Boltzmann’s Theorem on


Permanence of Distributions. Nature, 50 , 617.

Culverwell, E. P. (1895a). Boltzmann’s Minimum Theorem. Nature, 51 , 246.

Culverwell, E. P. (1895b). Professor Boltzmann’s Letter on the Kinetic The-


ory of Gases. Nature, 51 , 581.

Darrigol, O. (2001). The Historians’ Disagreements over the Meaning of


Planck’s Quantum. Centaurus, 43 , 219–239.

Darrigol, O. (2008). The Modular Structure of Physical Theories. Synthese,


162 , 195–223.

Daston, L. (1994). Historical Epistemology. In J. Chandler, A. I. Davidson, &


H. D. Harootunian (Eds.), Questions of Evidence (pp. 282–289). Chicago:
University of Chicago Press.

Feest, U., & Sturm, T. (2011). What (Good) is Historical Epistemology?


Editor’s Introduction. Erkenntnis, 75 , 285–302.

Gearhart, C. A. (2002). Planck, the Quantum, and the Historians. Physics


in Perspective, 4 , 170–215.

Giere, R. (1988). Explaining Science: A Cognitive Approach. Chicago: Uni-


versity of Chicago Press.

Giere, R. (1999). Science without Laws. Chicago: University of Chicago


Press.

Gouy, L. G. (1886). Sur le mouvement lumineux. Journal de Physique


Théorique et Appliquée, 5 , 354–362.

Höflechner, W. (1994). Ludwig Boltzmann. Leben und Briefe. Graz:


Akademische Druck und Verlaganstalt.

Jeans, J. H. (1903). The Kinetic Theory of Gases Developed from a new


Standpoint. Philosophical Magazine, 5 , 597–620.

35
Kangro, H. (1970). Vorgeschichte des Planckschen Strahlungsgesetzes. Wies-
baden: Steiner.

Kirchhoff, G. R. (1860). Über die Verhältnis zwischen dem Emissionsver-


mögen und dem Absorptionsvermögen der Körper für Wärme und Licht.
Annalen der Physik , 109 , 275–301.

Klein, M. J. (1962). Max Planck and the Beginnings of the Quantum Theory.
Archive for History of Exact Sciences, 1 , 459–479.

Klein, M. J. (1973). The Development of Boltzmann’s Statistical Ideas. Acta


Physica Austriaca, Supplementum, 10 , 53–106.

Kuhn, T. S. (1978). Black-Body Theory and the Quantum Discontinuity,


1894-1912 . Oxford: Oxford University Press.

Kuhn, T. S. (1984). Revisiting Planck. Historical Studies in the Physical


Science, 14 , 232–252.

Langley, S. P. (1886a). Observations on Invisible Heat-Spectra and the


Recognition of hitherto Unmeasured Wavelength, Made at the Allegheny
Observatory. Philosophical Magazine, 21 , 394–409.

Langley, S. P. (1886b). Sur les spectres invisibles. Annales de Chimie et de


Physique, 9 , 433–506.

Lummer, O., & Pringsheim, E. (1900). Über die Strahlung des schwarzen
Körpers für lange Wellen. Verhandlungen der Deutschen Physikalische
Gesellschaft, 2 , 163–180.

Morrison, M. (2007). Where Have All the Theories Gone? Philosophy of


Science, 74 , 195–228.

Morrison, M., & Morgan, M. S. (Eds.) (1999). Models as Mediators. Cam-


bridge: Cambridge University Press.

Needell, A. (1980). Irreversibility and the Failure of Classical Dynamics:


Max Planck’s Work on the Quantum Theory, 1900-1915 . Ph.D. thesis
University of Michigan Ann Arbor.

Planck, M. (1879). Über den zweiten Hauptsatz der mechanischen Wärmethe-


orie. Munich: Ackermann.

36
Planck, M. (1891). Allgemeines zur neueren Entwicklung der Wärmetheorie.
Zeitschrift für physikalische Chemie, 8 , 647–656.

Planck, M. (1895). Über den Beweis des Maxwellschen


Geschwindigkeitsverteilungsgesetzes unter den Gasmolekülen. Annalen
der Physik , 55 , 220–222.

Planck, M. (1896). Absorption und Emission electrischer Wellen durch Res-


onanz. Annalen der Physik , 57 , 1–14.

Planck, M. (1897a). Über electrische Schwingungen, welche durch Resonanz


erregt und durch Strahlung gedämpft werden. Annalen der Physik , 60 ,
577–599.

Planck, M. (1897b). Über irreversible Strahlungsvorgänge. 1. Mitteilung.


Sitzungsberichte der Preussischen Akademie der Wissenschaften, 1 , 57–
68.

Planck, M. (1897c). Über irreversible Strahlungsvorgänge. 2. Mitteilung.


Sitzungsberichte der Preussischen Akademie der Wissenschaften, 2 , 715–
717.

Planck, M. (1898a). Über irreversible Strahlungsvorgänge. 3. Mitteilung.


Sitzungsberichte der Preussischen Akademie der Wissenschaften, 1 , 1122–
1145.

Planck, M. (1898b). Über irreversible Strahlungsvorgänge. 4. Mitteilung.


Sitzungsberichte der Preussischen Akademie der Wissenschaften, 2 , 449–
476.

Planck, M. (1899). Über irreversible Strahlungsvorgänge. 5. Mitteilung.


Sitzungsberichte der Preussischen Akademie der Wissenschaften, 1 , 440–
480.

Planck, M. (1900a). Entropie und Temperatur strahlender Wärme. Annalen


der Physik , 4 , 719–737.

Planck, M. (1900b). Zur Theorie des Gesetzes der Energieverteilung im Nor-


malspektrum. Verhandlungen der Deutschen Physikalische Gesellschaft,
2 , 237–245.

37
Planck, M. (1901). Über das Gesetz der Energieverteilung im Normalspek-
trum. Annalen der Physik , 4 , 553–563.

Planck, M. (1902). Über die Natur des weissen Lichtes. Annalen der Physik ,
7 , 390–400.

Planck, M. (1906). Vorlesungen über die Theorie der Wärmestrahlung.


Leipzig: Barth.

Planck, M. (1958). Physikalische Abhandlungen und Vorträge. Braunschweig:


Vieweg u. Sohn.

Rayleigh, J. W. S. (1889). On the Character of the Complete Radiation at


a given Temperature. Philosophical Magazine, 27 , 460–469.

Renn, J. (2006). Auf den Schultern von Riesen und Zwergen. Weinheim:
Wiley-VCH.

Rheinberger, H.-J. (2010). On Historicizing Epistemology. Stanford: Stan-


ford University Press.

Schuster, A. (1894). On Interference Phenomena. Philosophical Magazine,


37 , 509–545.

Stefan, J. (1879). Über die Beziehung zwischen der Wärmestrahlung und der
Temperatur. Sitzungsberichte der Akademie der Wissenschaften zu Wien,
79 , 391–428.

Stroud, B. (2011). Epistemology, the History of Epistemology, Historical


Epistemology. Erkenntnis, 75 , 495–503.

Suppe, F. (1977). The Search for Philosophical Understanding of Scientific


Theories. In F. Suppe (Ed.), The Structure of Scientific Theories (pp.
3–232). Chicago: University of Illinois Press.

Suppe, F. (1989). The Semantic Conception of Theories and Scientific Re-


alism. Chicago: University of Illinois Press.

Uffink, J. (2007). Compendium of the Foundations of Classical Statistical


Mechanics. In J. Butterfield, & J. Earman (Eds.), Philosophy of Physics
(pp. 923–1074). Amsterdam: North Holland volume 2.

38
Van Fraassen, B. (1980). The Scientific Image. Oxford: Oxford University
Press.

Wartofsky, M. (1979). Models. Dordrecht: Reidel.

Wien, W. (1893). Eine neue Beziehung der Strahlung schwarzer Körper zum
zweiten Hauptsatz der Wärmetheorie. Sitzungsberichte der Preussischen
Akademie der Wissenschaften, 1 , 55–62.

Wien, W. (1896). Über die Energievertheilung in Emissionsspectrum eines


schwarzen Körpers. Annalen der Physik , 58 , 662–669.

Wien, W. (1909). Theorie der Strahlung. In A. Sommerfeld (Ed.), En-


cyklopädie der mathematischen Wissenschaften (pp. 282–357). Leipzig:
Teubner volume V.3.

Zermelo, E. (1896). Über einen Satz der Dynamik und die mechanische
Wärmetheorie. Annalen der Physik , 57 , 485–494.

39

Vous aimerez peut-être aussi