GENERAL ARTICLES

Phenomenologism vs fundamentalism:
The case of superconductivity
Towfic L. Shomar
This article argues that phenomenological treatment of physical problems is more powerful than
fundamental treatment. Developments in the field of superconductivity present us with a clear
example of such superiority. The BCS (Bardeen, Cooper and Schrieffer) was accepted as the fundamental theory of superconductivity for a long time. Nevertheless, Landau and Ginzburg phenomenological model has so far proven to be a more fruitful theoretical representation to understand
and to predict the features of superconductivity and superconductive materials.
Keywords:

BCS theory, fundamentalism, Landau and Ginzberg, phenomenalism, superconductivity.

TWO distinct types of theoretical work can be found in

physics. The main difference between them is the way in
which they are built. One starts from abstract entities to
accommodate experimental observation and the other
proceeds in the opposite direction1.
The first is expressed in fundamental theories. I claim
that fundamental theories are merely tools for constructing
both models and other fundamental theories. This article
will show that previous fundamental theories play the
role of tools in constructing new fundamental theories
(note 1). They do not express truth about nature. They are
theories that:
(1) Give a coherent story about the studied phenomena.
(2) Can explain existing empirical findings in the field
and can predict new aspects of the phenomena under
study that can be tested empirically.
(3) Are consistent with previous theories.
(4) Are either derivable from previous accepted fundamental theories; or at least correspond at a certain
limit to these previous theories; or must be able, at
least, to provide a mathematical basis for a net of
theoretical models that can provide good explanations to the properties of the field of study.
(5) Are in principle able to be unified with each other.
Hence, the basis for fundamentalism is mathematical in
structure. How a fundamental theory is constructed is
given later in the article.
The second type of theoretical representation is that
which does not have a deductive relation to previous established fundamental theories. This kind of theoretical
Towfic L. Shomar is in the Department of Human and Social Sciences,
Philadelphia University, Jordan and the Centre for the Philosophy of
Human and Social Science, London School of Economics, London.
e-mail: [email protected]
1256

work is closely concerned with experimental work and

takes as a point of departure a descriptive account of experimental activity. Models constructed in this way are
phenomenological models and they have the following
features:
(1) Are a special kind of theoretical representation that
departs from the phenomenological level.
(2) Describe the phenomena, including all the factors
that would allow them to exhibit themselves, with
whatever experimental boundary conditions deemed
necessary.
(3) Represent natural phenomena in as far as the model
is able to mediate between the experimental and the
theoretical. They do so by presenting a story that can
relate the mathematical parameters of the models to
the properties known about the phenomenon.
A simple description of what physicists do will illustrate
what a phenomenological model is. If a new phenomenon
is discovered (or built), physicists try to understand it. A
group of usual procedures is applied. The first of these is
to know the conditions in which the phenomenon occurs.
In accordance with these conditions, physicists suggest a
set of experiments that would help in generating new
data. At this point our knowledge of the great variety of
theories available in physics will be important. Physicists
will search within all possible theoretical schemes to find
a mathematical structure that can represent the data. They
do this by plotting the data; they then begin thinking
about which of the known mathematical forms used
in physics can mimic the plotted data. These mathematical forms need not be related to the field of the phenomenon.
Now, it would not be enough to find a mathematical
form that expresses the data pattern formally. The physical
intuition of the physicists would lead them to relate the
CURRENT SCIENCE, VOL. 94, NO. 10, 25 MAY 2008

GENERAL ARTICLES
parameters of the mathematical form to the properties of
the phenomena. The established relations between the parameters and the properties need, in order to adhere to
coherency, to be consistent with our previous established
pictures of the related fields. In order to do so the model
would need a descriptive body that would present a coherent picture, consistent with our previously established
models.
At this stage we have only a scattered descriptive
measures, with a set of mathematical forms and a set of
possible connections with established models in physics.
Physicists would start to present an overall model that will
account for all the known features of the studied phenomenon. If the outcome scenario succeeds in relating the
properties to the parameters, then the mathematical form
together with the descriptive body and the story would
constitute a phenomenological model.
However, the physical intuition does not mean relating the data to an existing theory already agreed to cover
the phenomenon, but possibly changing a tool from an
understood theory, in another domain, to adapt it to what
is studied. Hence, the physical intuition plays a role in
presenting a story which can relate the data and the ontological world of the phenomenon with the expected
mathematical structure. Construction of a phenomenological model is presented later in the article, showing
how the phenomenological aspects and different tools
from different fundamental theories help in this construction (note 2).
It is important to mention that knowing the mathematical structure, which might be used to illustrate the data,
will not provide enough information to build a model. I
do not accept, for example, that a phenomenological
model is merely a mathematical structure (note 3). At the
same time, it is not naked data, there is more to it. The
story which provides the basis for how to deal with the
mathematics in relation with the phenomenon, is as crucial
as finding a mathematical structure. Hence a phenomenological model in physics consists of a mathematical
structure plus a descriptive level, depending on the human
activity, that is driven by our experience and by the empirical data themselves and has two levels: the objective
description of the experimental set-up or the environmental
set-up and a kind of explanation that provides a deeper
understanding to the phenomenon. Hence, a phenomenological model is a type of theoretical explanation that departs from the description of the environmental set-up of
the phenomenon to give a structural account of its relations. However, the theoretical explanation that the phenomenological model provides is not as abstract as the
fundamental theory explanation.
It should be noted here that I am changing the term experimental set-up with environmental set-up because
phenomenological models try to capture nature within its
boundaries and not to impose purified boundaries such as
those imposed by the experimental set-up. That is to say,
CURRENT SCIENCE, VOL. 94, NO. 10, 25 MAY 2008

the experimental set-up usually takes the natural phenomenon from its natural setting and tries to eliminate all
possible distractive elements before conducting the experiment. Phenomenological models, because of the way
they are built (i.e. from a bottom-up approach), try to
capture as much of the natural environment as possible,
regardless of the ability to understand all the used factors.
As long as the model is able to represent the overall features of the phenomenon, it would be accepted as a phenomenological model.
Although the mathematical form of the phenomenological model is similar in nature to fundamental theories,
it is usually associated with some parameters that cannot
be specified theoretically, but are found through experimentation and fed into the model at certain points during
the building process. By virtue of such parameters, the
phenomenological model might be counted as low-level
theorization. Hence, this would be a good reason to think
of the explanation provided by the phenomenological
model as a whole, as being less abstract.
Therefore, the main differences between a fundamental
theory and a phenomenological model are:
(1) Fundamental theories are constructed using a topdown approach, while phenomenological models are
constructed using a bottom-up approach.
(2) Phenomenological models tend to be more flexible.
(3) The story presented by the phenomenological model
is not as abstract as that presented in the fundamental theory.
(4) The phenomenological models need not tell us why
certain phenomena behave the way they do; it is sufficient that they give good predictions about any
phenomenon.
Furthermore, with the help of different tools from the so
called fundamental theories, physicists can construct
new phenomenological models to account for any new
observed properties. It renders them with higher flexibility than fundamental theories.
This entails a deeper thesis: fundamental theories need
not play a representative role. They only serve as tools in
developing more theories that are fundamental and for
constructing phenomenological models; the latter, from a
realist point of view, can be accepted as representative of
nature.
Before indulging in the Bardeen, Cooper and Schrieffer
(BCS) vs Landau and Ginzburg (LG) discussion, it might
be useful to look at a simpler example. Let us consider
Newtons second law. If we are thinking of the Newtonian framework, then we ought to interpret the law in accordance with moving bodies in a three-dimensional
world in an infinite space and infinite time. Also, in the
fundamental level of the law we present it in the most
general possible form:
1257

GENERAL ARTICLES

F = a M ,

(1)

where F is the force, M the mass and a the acceleration.

However, when we want to apply such a general law
on a particular setting, we start to add certain concepts
and elements to account for the specific feature of such a
setting. Let us consider, for example, the case of a box
sliding on a rough surface with an angle with the horizon. In such a case the exact model that accounts for this
particular case would be:
F = mg cos R = ma,

(2)

where g is the gravitational constant and R the resistant.

In this specific case, we are dealing with a specific model
that can be interpreted as the outcome of a series of observations of moving bodies over slopes with known friction. Although it depends on our previous knowledge of
trigonometry and a generalized model, Newtons second
law, it might also be inductively inferred from direct observations and mathematical knowledge. This model is
still true, whether we accept the Newtonian absolute
space and absolute time, or the relativistic spacetime
curvature. Hence, such a model with its connotations and
associated story would be considered as a low-level theoretical representation and would be counted as a phenomenological model. Here in such a case, we are
differentiating between the high-level generalization and
low-level representative model. The first expresses a general law that can be generic of any force in certain space,
while the latter expresses a specific case with specific
features.
The more complex the theoretical representation,
clearer is the distinction between fundamental theories
and phenomenological models. In case of Maxwells
equations, for example, some of the terms in the theoretical representation of the fundamental theory do not have
a direct counterpart in nature, such as the electric displacement D, which has no clear physical meaning5.
The article studies the contrast between a fundamental
theory, say the BCS theory of superconductivity and the
LG-phenomenological model of superconductivity, a
model accepted now as the best representative of superconductivity. It then discusses the similarities and differences between these two theoretical works, and later
argues that new discoveries bring into question the very
idea of having a fundamental theory for all kinds of superconductivity.

Superconductivity
Superconductivity was first discovered by the Dutch
physicist Onnes in 1911. He detected that metals when
cooled to a very low temperature, inside liquid helium
(under 4 K), exhibit a strange phenomenon: the total dis1258

appearance of resistance under a critical transition temperature Tc. Later, in 1933, Meissner and Ochsenfeld
discovered that the magnetic field is expelled inside the
superconductor under a certain transition magnetic field
Hc (the Meissner effect)6.
The discovered experimental constrains shaped the
theoretical understanding of superconductivity from the
beginning. In 1935, London and London7 (note 4) suggested the first theoretical explanation of superconductivity.
Using their knowledge from diamagnetism, they constructed
a phenomenological model expressed by the London
equations:
curl js + H = 0,

( js ) E = 0,
t

(3)

where js is the superconducting current density, H the

magnetic field, E the electrical field, c the speed of light
and is an experimental constant = m/ne2, where m is the
mass of the electron, n the number of electrons and e the
charge of the electron.
Later, a number of experimental activities tested the
correctness of these equations. The results show that
Londons equations could give good predictions for the
superconducting current if the following conditions are
satisfied:
(1) H << Hc, otherwise at the boundary zone where H is
near the critical magnetic field, the model failed to
predict what happens experimentally.
(2) The superconducting electron density must be constant.
(3) The penetration depth is less than the thickness of
the sample.
The difficulties facing London equations are as follows.
First, when the temperature T Tc, the penetration depth
becomes larger and the thickness of the walls separating
the normal and the superconducting states also become
larger.
The second difficulty is related to the change in the free
energy between the normal and superconducting states.
As pointed out by Ginzburg9, if we restrict ourselves to
the case of a steady field, then (Londons equations), together with Maxwells equations, are sufficient for determining the density js of the superconducting current
and the field H in the superconductor. However, if we
want to have a broader understanding of the change from
the normal to the superconducting state, the London approach is not sufficient.
Any new theory or model should also aim at explaining
additional empirical facts:
(1) The transition temperature is proportional to the
isotopic mass M of the metal nuclei (the isotope
effect)10.
CURRENT SCIENCE, VOL. 94, NO. 10, 25 MAY 2008

GENERAL ARTICLES
(2) The phase transition at the critical temperature
would transfer the electrons from a disordered bunch
to an ordered one.
(3) The thermal vibration of the atoms is the principal
cause of electrical resistance in metals at ordinary
temperature10.
(4) The energy gap.
Two schemes were developed to overcome the difficulties which the London model faces: the LG phenomenological model and the BCS fundamental theory.

LG model
In 1950, LG developed an extension of the London phenomenological model to take into account a space variation of the order parameter11. They suggested the following
equations:

+ | |2 +

j=

1
2m

2eA

i= c = 0,

ei=
e2
* A,
( * *)
2m
mc

(4)

where , are experimental constants, a pseudo wave

function and A the local vector potential.
How did they arrive at their conclusions? Seeing how
they did so will help make clear why such a model is
called phenomenological.
When London and London suggested their solution for
the Meissner effect, they ended their paper by stating that
a more general solution for superconductivity might be
inspired by studying Gordons formulae for electric current:

story can be told? At that time it was an experimental fact

that superconductivity exhibited some type of thermal
fluctuation. The thermal vibration is known to be the cause
of electrical resistivity. Hence the no-resistivity region
should be related to what could happen for such thermal
vibration.
Landau had been working on a theory for the phase
transition in solid-state physics. He thought that it could
be of aid in their derivation. Experimentally, it was expected that transition at the critical temperature is one
from disorder to order. However, there is more than one
way to represent the shift from disorder to order. Nevertheless, the experimental evidence helped Landau suggest
that the transition from the normal to the superconducting
state is a second-order transition (as in the transition from
ferromagnetic to paramagnetic), where we can neglect the
higher terms. This means that the thermal vibration of the
electrons that cause electrical resistivity will be ordered
under the transition from the normal to the superconducting
state. No thermal vibration will occur anymore, which
implies that no resistivity will occur as well.
London and London suggested that the flow of electrons through the sample can be considered as two fluids:
superconducting and normal. Hence, it was normal to assume that the superconducting state would be presented
thermodynamically. Then to relate the transition at the
critical temperature to thermodynamics, LG offered a
guess for the free energy:

F (; T , E )d r,
3

where is an order parameter which is function of the

distance r (Schrieffer12). Expanding this function in terms
of power series in 2 we find that
F = E + g0 +

J =

he
e2
( * * )
* A,
4 im
mc

*
e2

* .

t
t mc 2
4 imc 2
he

(5)

This is in effect the mathematical form that London and

London expected would represent the experimental data
of superconductivity. However, London and London
failed to present a story that could identify the parameters
in the equation with properties of superconductivity. So,
LG started from this remark in addition to the experimental evidence. The idea was to try to find tools from existing theories and to borrow forms that might end up with a
similar equation to that of Gordons, but which can be directly related to the field of superconductivity.
What kind of information can be provided by the experimental and phenomenological facts? And what kind of
CURRENT SCIENCE, VOL. 94, NO. 10, 25 MAY 2008

1
1
g 2 + g 4 4 + ....
2 2
4

The standard text by Tilley and Tilley13 describes this

saying: Landaus general theory of second-order phase
transitions is based on the idea that a phase transition
could be characterized by some kind of order parameter,
and a simple postulated form for the dependence of the
free energy on the order parameter.
Consider the non-vanishing terms. Because it is a second-order phase transition, the term g4 is positive and the
higher-order terms can be neglected (note 5). Think about
the transition from the normal to the superconducting
states as one from disorder to order, and take the normal
free energy to have the same form as in thermodynamics,
which is:
F = U T,
where U is the initial energy.
1259

GENERAL ARTICLES
In the case of a variable magnetic field, the relation between the free energy and the order parameter is:
F = Fn + a 2 +

b 4
+C
2

2eA
B2

i =c + 8 .

(6)

And in order to keep the invariance, there is an added

vector potential A, which is related to the field B
(B = curl A).
Tilley and Tilley13 continue: the next crucial insight in
LG was that for a superconductor the order parameter
must be identified with the macroscopic wave function
. As I said, LG were inspired by Gordons equations.
To correspond their equation to that of Gordon, they
ought to identify a parameter in their equation with the
wave function. The only parameter which can be thought
of as conveying the same properties as is the order parameter. So, in this sense it was a must for LG to identify
the order parameter with the macroscopic wave function .
Then by setting the correct values of the coefficients,
depending on the experimental results we get:
F = Fn + 2 +

4 +

1
2m

2eA

i= c

B2
.
8
(7)

It may seem that F has just been modified from eq. (6) to
eq. (7), but this is far from correct. Two features of the
construction illustrate my thesis about the use of theories
as a tool for constructing phenomenological models: first,
eq. (6) itself had been constructed using tools from different theoretical models which are not connected to superconductivity. These tools are: the phase transition, the
assumption that the free energy is a function of the order
parameter and the Gordons equations. Second, eq. (7) is
not derived from quantum mechanics, rather a nonquantum mechanical eq. (6) is reformed in eq. (7). This
illustrates a theoretical influence that functions merely as
a way to express the experimental results. Physicists see
that this construction of eq. (7), independent of any detailed theory of the superconducting state, represented a
tour de force of physical intuition14.
Now from the modification of the free energy in eq.
(7), LG set F = 0, to obtain the following equations:

+ | |2 +

j=

1
2eA
= 0,
i =
c
2m

e *i =
e *2
* A.
( * *)
2m*
m*c

(8)

As we can see, these equations have the same form as

Gordons equations even if there is no direct relation in a
1260

deductive or physical sense (note 6). Gordons equations

are formulated from the relativistic Schrdinger equation,
and as we have seen the LG equations were the result of
reasoning about the phase transition in fluids in relation
to the free energy.
Therefore, the LG model would be described as a phenomenological model because it consists of: (1) Mathematical equations similar in form to a previous mathematical
form. (2) A set of identifications that would relate parameters in the mathematical form, like and , with
properties in superconductivity like coherence length and
the penetration depth. Another important example is that
of identifying the order parameter with the pseudo wave
function. (3) A story that gives a description of the environmental set-up of the phenomenon of superconductivity; relates the different properties of superconductivity
and displays how the mathematical forms might represent
these properties. Furthermore, the LG model had been
constructed by departing from the phenomenon.

The BCS theory

In 1957, the BCS theory was proposed taking into consideration the empirical facts stated earlier (note 7) and
aiming to build a microscopic theory of superconductivity. While LG started from the experimental facts and
tried to see how they could identify some data-outcome
plot with Gordons theoretical equation, the BCS theory
was derived from quantum field theory. The belief that
quantum field theory could establish a basis for understanding the properties of superconductivity came into
place after an accumulation of microscopic models accounting for any aspect of superconductivity. Hence, the standpoint for BCS was a theoretical one, whereas for LG it
was an experimental one.
The BCS derivation is long and complicated. It depends on a great number of theoretical tools to fulfil the
constraints of the phenomenon. These tools are as follows.
(1) Previous physical theories: quantum field theory,
electrodynamics, diamagnetism, Maxwells equations and
thermodynamics. These theories are what give legitimacy
to the mathematical forms constructed by BCS to account
for superconductivity. If BCS fail to provide a good and
tidy derivation, using these accepted theories their theory
will not qualify as a fundamental one. Though this may
not be the only criterion for being a fundamental theory,
it is a necessary one.
(2) Mathematical tools: the theory uses the quantum
field theoretical technique of second quantization, perturbation theory, and different types of special approximations like HartreeFock-type. These abstract tools can help
in investigating the existing facts so that they can be seen
as similar to some part of an existing fundamental theory.
(3) A story: Here also we find a story is important to
connect the previous tools with each other, and to connect
CURRENT SCIENCE, VOL. 94, NO. 10, 25 MAY 2008

GENERAL ARTICLES
the mathematical parts of the theory to the natural phenomena. A metaphor here might be a good aid: while the
story that the phenomenological model presents is
strongly associated with the empirical findings, the story
which the fundamental theory gives is like a piece of art
(a sculpture or a painting) which the artist meant to be realist but, usually, turns out to be surrealist or even abstract.
The point here is that the story given by the fundamental
theory would use accepted theoretical models that might
be justified only if we accept the underlying theoretical
concepts carried on from previous fundamental theories.
As we will see, in the case of the BCS theory, it depended
on accepting ideas such as Fermi surface, the Cooper
pair, electronphonon interaction and Fermi sea. These
theoretical models are all justified only if we accept
quantum field theory as a point of departure.
Although generally physically motivated, identification
of the mathematical terms with different sorts of real entities in the superconducting phenomenon does not usually
tend to give a good description of the phenomenon, on
the theoretical level. This point is important because such
identification gives the mathematical models and tools a
physics body. In our case, BCS runs the following story.
The BCS theory depends on a group of theoretical models
that can associate superconductivity with quantum field
theory. In 1955, Cooper suggested a way to understand
the fact that the charge quanta in superconductivity are
2e. He suggested that the electrons in the superconducting state occur in correlated pairs (Cooper-pairs) a theoretical model that have the same quantum state. The
importance of this point is obvious once we recall that
electrons are fermions, which means they are unlikely to
coexist in the same quantum state. But we can model a
pair of electrons as a quasi-bosons. These pairs of electrons can be created through the electronphonon interaction a theoretical assumption which considers that
electrons interact with a lattice giving phonons. This idea
was supported when the isotopic effect was observed
materials in nature are a combination of many isotopes,
and the isotopic effect is the fact that the transition temperature depends on the isotopic nuclear mass. The idea
that electronphonon interactions are primarily responsible for superconductivity seems reasonable, because it
indicates that the vibrational motion of heavy nuclei
plays an essential role in the formation of pairs of electrons. Let us remember that the relation between thermal
vibration and conductivity was an established fact by that
time.
The BCS theory also used the two-fluid model a phenomenological model which assumes that we can imagine the superconducting material as if it consists of two
kinds of overlapping fluids, one of which is responsible
for the normal state and the other for the superconducting
state. This model can also help, in addition to the Cooper
pairs, in understanding the use of another theoretical
model, that of Fermi surface, which is an imaginary surCURRENT SCIENCE, VOL. 94, NO. 10, 25 MAY 2008

face in k-space (spin-vector space) that separates the occupied energy levels from unoccupied energy levels and
will define the first empty level.
We can see that BCS used some important assumptions
(which I have underlined). These assumptions cannot be
neglected in the theory; otherwise what will be left are
merely the mathematical tools and bits and pieces from
previous fundamental theories. However, right now all of
these assumptions are being challenged.
It is important for any theory of superconductivity to
derive the relation between the current density and both
the potential and the momentum, because out of these
two equations all the other known mathematical descriptions of the properties of the superconductors can be derived.
For this derivation, using a quantum field theoretical
framework, the BCS theory needs to employ all the mentioned models and tools from fundamental theories.
They start by suggesting a Hamiltonian for the electrons in the superconducting state. Then they add the isotopic mass, Ms, and its relation to the phononelectron
interaction, to see its effect on the non-diagonalized terms
in the Hamiltonian. Then the diagonalized part is renormalized using Bloch energies. Introducing the idea of
the Fermi surface will allow them to use FermiDirac statistics on matrix elements. The story goes on using annihilation and creation operators, then HartreeFock-like
approximations, etc. They arrived in the end at a derivation for a special kind of wave function which, by defining the correct Hamiltonian and accepting a certain gauge
where A = 0, gives us a derivation for the paramagnetic
and diamagnetic current densities:
j = j p + jd ,
e 2 = 2 (2 )3/ 2
(2k + q)k

2m 2 c 2 i 0 k , q , k ,q,

iq r

( 0 (T ) | c*k + q, ck , |
a (q)e

j p (r ) =

*
i (t )) ( i (T ) | c k + q , ck , | 0 (T )) W W ,
0
i

+ complex conjugate

and
ne 2
jd (r ) =
A( r ).
mc

This, of course, can be accepted as a straightforward

derivation from fundamental theory, namely quantum
field theory.
It should be stated at this point that both the BCS and
LG equations can account for the most important properties of superconductors, such as: (i) The penetration
depth, i.e. the depth that a magnetic field can penetrate in
a superconducting sample; the depth where the sample
1261

GENERAL ARTICLES
does not exhibit Meissner effect. (ii) The coherence length,
which is a measure of the distance within which the superconducting electron concentration cannot change drastically in a spatially-varying magnetic field16. (iii) The
energy gap. In the case of the BCS theory, it needs a series of approximations to account for each of these three
properties.
Hence, the BCS theory fulfils the criterion suggested
earlier in the article. It is consistent with and derivable
from previous fundamental theory (quantum field theory),
gives a coherent story about type-one superconductors,
and can explain existing empirical findings.

Discussion
The previous sections showed that the LG model was
built using a bottom-up approach, while the BCS theory
was built using a top-bottom approach. Hence, the standpoint of the LG model was experimental evidence and that
of the BCS model was quantum field theory. Let me discuss the differences between these two approaches. Bardeen11 stated, way back in 1956, that:
Anything approaching a rigorous deduction of superconductivity from the basic equations of quantum
theory is a truly formidable task. The energy difference between normal and superconducting phases at
absolute zero is only of the order of 108 eV per atom.
This is far smaller than errors involved in the most
exacting calculations of the energy of either phase.
One must neglect terms or make approximations
which introduce errors which are many orders of
magnitude larger than the small energy difference one
is looking for. One can only hope to isolate the physically
significant factors which distinguish the two phases.
For this, considerable reliance must be placed on experimental findings and the inductive approach.
So Bardeen, with whom Cooper and Schrieffer put
forward the BCS theory of superconductivity, himself
admits that any theory that departs from the quantum theory would need to neglect terms or make approximations
which introduce errors which are many orders of magnitude larger than the quantities one is looking for. The
BCS theory needs exactly these approximations to account for practical situations.
On the contrary, the LG model relays on experimental
findings and the inductive approach and is able to present a mathematical structure that can be consistent with
a representation of the phenomenon, trying to relate different bits and pieces from the shattered information provided through years of experimentation.
The BCS theory was accepted for a long time as the
fundamental theory of superconductivity. There were different factors that contributed toward calling it fundamental, the most one being its use of quantum field
1262

theoretical grounds, i.e. a microscopic base for understanding a macroscopic phenomenon. The LG model also
depended partly on microscopic factors the Gordons
formula and LG employed their knowledge about fundamental theories to construct their model; yet nobody
considered their model as fundamental. That was because
their derivation did not give a clear reason for taking the
order parameter to be a wave function; and because their
derivation was not considered as a straightforward one
from a previous fundamental theory.
Interestingly, the LG model has proved to be capable
of adapting to new properties of superconductivity,
whereas the BCS model failed to be modified to account
for these new discoveries, as we shall see next. In this
sense, LG can give an example of the way phenomenological models can prove more fruitful than fundamental
theories.
Nevertheless, why did the BCS theory, in spite of all
its success, fail to maintain its position as the fundamental theory of superconductivity? Up to a certain point, the
BCS theory is reliable in giving an understanding of superconductivity. This is especially so if we are dealing
with type-one superconductors. The BCS theory also
managed, using further assumptions, to account for typetwo superconductors. However, other kinds of superconductors, especially high-temperature superconductors which
were discovered in 1986 by Bednorz and Mller, prove
more problematic. It is important to mention here that in
all the interpretations of the BCS theory concerning the
critical temperature, the most optimistic one suggests
30 K to be the highest possible critical temperature. Now
we have superconductors with (125 K) Tc (note 8). So the
BCS theory cannot be seen as valid for all kinds of superconductivity.
In a discussion between Anderson and Schrieffer19 on
the difficulties facing a theory for high-temperature
superconductivity, Anderson mentions: I think few people
realize that we now know of at least six different classes
of electron superconductors, and two other BCS fluids as
well. Out of these only one obeys the so-called conventional theory that is, BCS with phonons that fit unmodified versions of Eliashbergs equations.
Anderson continues by stating that it is crazy to think
that the new high-temperature superconductors can fit the
BCS theory, since even most of the simpler ones do not.
He states that: Back in the 1960s we may have created
the abomination, a theory that has become nonfalsifiable in the Popperian sense in that people insist on inventing more and more ingenious ways to make it fit any
anomaly!.
In fact, that was quite right on the theoretical level; even
a great physicist like Pippard mentioned in 1964 about
the success of the BCS theory: This success is so remarkable that I almost believe you would forgive me if I
were to say there now remain no problems in superconductivity18. Nevertheless, most physicists were reluctant
CURRENT SCIENCE, VOL. 94, NO. 10, 25 MAY 2008

GENERAL ARTICLES
to continue using the BCS techniques, especially after
they found that the LG model could give them the same
predictions with simpler mathematics. A survey of the
textbooks on superconductivity can tell us about the role
of the BCS theory. One of the most widely read textbooks
was Michael Tinkhams Introduction to Superconductivity20, where he writes: The emphasis is on the rich array
of phenomena and how they may be understood in the
simplest possible way. Consequently, the use of thermal
Green Functions has been completely avoided, despite
their fashionability and undeniable power in the hands of
skilled theorists. Rather the power of phenomenological
theory in giving insight is emphasized, and microscopic
theory is often narrowly directed to the task of computing
the coefficients in phenomenological equations.
The BCS theory, as already stated, cannot be accepted
as a genuine theory unless certain assumptions are associated with it. The new theoretical work is questioning
each of these assumptions. In addition, the BCS theory
(note 9) does not speak about important factors in hightemperature superconductivity, like the chemical structure
of the materials and their normal state. Another reason for
not accepting the BCS theory is its disagreement with
new experimental outcomes: the highest critical temperature
predicted by BCS is 30 K; the value of the energy gap
that the BCS accepts is less than 3.5 kTc, while the new
superconductors exceed that limit to twice the value, etc.
Hence, while the fundamental theory of superconductivity failed to accommodate high-temperature superconductors and other kinds of superconductors, the LG
phenomenological model of superconductivity proved to
be more fruitful in representing all kinds of superconductors. The main difference between the BCS theory and
the LGs model was the point of departure. Both of them
used tools from previous fundamental theories. Both of
them need a story to relate the mathematics with the
properties of the phenomenon. Moreover, both of them
hold a theoretical explanation. However, LG departs from
the experimental level, while the BCS theory departs
from the quantum field theory.
This important difference plays a major role in the type
of story associated with the theoretical explanation. We
saw that the BCS story needs certain assumptions to be
consistent with the quantum field theory and to be able to
derive the needed mathematical form. As shown above,
all these assumptions are being challenged. The LG story
depends on a well-tested set of assumptions, related to the
empirical findings only. This gives the model the advantage of being attached to experimental evidence.
The association of the phenomenological model with
experimental evidence, and the liberty which the model
provides to some parameters to be measured experimentally, gives it a better stand toward representing new kinds
of superconductors. Hence, the phenomenological model
proved to be more fruitful than the fundamental theory of
superconductivity.
CURRENT SCIENCE, VOL. 94, NO. 10, 25 MAY 2008

A side issue: Is there a fundamental theory of

superconductivity?
The striking thing about superconductivity is that it is a
phenomenon that has two distinguishing properties, zeroresistance and Meissner effect (or magnetic vertex penetration), but these two do not have the same known origin. There are, as already mentioned, six known types of
superconductors and each of them has different normalstate properties. The chemical properties of some are
complicated and give rise to many contradictory results.
Theory has so far failed to give a single generalized account for these different kinds of superconductivity. Many
factors have been investigated in an attempt to account
for superconductivity, but until now all of these factors
appear to have experimental evidence against them21.
It is important for any fundamental theory not to contradict any of the experimental observations that cannot
be accepted as exceptions. Anderson21 urged this kind of
position in addressing the BCS assumption that all superconductors are Fermi liquid-type materials: Here I must
appeal to a point of logic. The common response, when
one makes a firm statement that none of these materials
are Fermi liquids because of one or another observation,
is to say that the observation encounters exceptions
among these many materials. However, that is not the
point: if they are all at the same fixed point and they
clearly are it will be non-Fermi liquid for all if it is not
for any one: it is necessary only to prove the negative in
one instance. Exceptions are logically irrelevant.
There is no fundamental theory for superconductivity.
Physicists in the field still use the generalized LG phenomenological theory of superconductivity. This theory can
equip physicists with effective mathematical techniques
to predict the behaviour of a superconducting material or
design superconducting devices. The major factor that
makes the phenomenological models and theories so
powerful is the fact that they are a first-level abstraction,
departing from the experimental level.
The experimental observations now seem to indicate
that it is highly probable that we will not be able to arrive
at a fundamental theory of superconductivity that starts
from existing fundamental theories. This is because the
essential assumptions for the candidate theories, for such
starting point, have proven to be in contradiction with
experiments involving any of superconductors. That leads
us, if we want to continue to search for fundamental theories, to one of two options.
The first option is to consider that the candidate theories are no good for superconductivity, but some other
theory will emerge that can account for all the aspects of
superconductivity. Of course, such a point of view does
not tell us a lot, because fundamental theories should be,
by definition, compatible with previous fundamental
theories in the same domain of applicability. Einsteins
theory, for example, should conflict with Newtons the1263

GENERAL ARTICLES
ory in the domain of low velocities and small masses.
Hence, the predictions of the new theory should not be in
contradiction with the well-confirmed predictions of the
previous theories. So this option will require a whole new
theoretical approach, not just for superconductivity, but
for other domains as well.
The other option is to consider that there is more than
one fundamental theory for superconductivity: one each
for conventional superconductors, high-temperature superconductors, organic superconductors, etc. This option
might work, though it would end up by contradicting the
unification assumption of fundamental theories, i.e. fundamental theories are compatible with each other and
there will be a way in which all these theories can be unified in a theory of everything.
None of these two options is necessary. If we accept
that phenomenological models are representative of nature, then our theories will eventually be merely tools to
help us in constructing new theoretical tools and new
phenomenological models. Science and scientists will
have more freedom by doing that, and this will help them
to go beyond the theoretical limitations. After all, if all
the physicists had accepted the BCS theory, we would
have never been able to discover high-temperature superconductors.

Notes
1.

2.

3.
4.
5.

6.

7.

In a previous paper with Nancy Cartwright and Mauricio Surez2,

we argued that theories are tools to construct phenomenological
models. The idea presented here is a step further.
There are, however, some cases where another practice can be
found in physics: a theory predicts, due to certain hypothesis, the
occurrence of a phenomenon under such and such circumstances.
The theory also provides a theoretical model to describe the idealized situation predicted by it. If the technological developments
can permit applying the prediction, a set of experiments would be
conducted to test the hypothesis. During the course of testing, a
procedure of approximations and modifications is applied on the
theoretical model and most of the time we end up having a new
model which has many elements that are not relevant to the original theory. I will call such a model also a phenomenological model.
It is important to notice here that the new model, although the phenomenon was predicted by the theory, cannot at most times be
deduced from the postulates of the theory because of introducing
new irrelevant factors to the theory and is not deducible from it3.
As Margaret Morrison claims in her reply to Ian Hacking4.
For a detailed discussion of London and London model see
Suarez8.
Second-order transition means that g4 is positive in contrast with
the first-order transition, where g4 is negative and we cannot neglect the rest of the terms.
This kind of associating a form from relativistic Schrdinger equation with a derived form from another field is form correspondence.
For details about form correspondence and its role in physics see
Shomar 1.
The BCS paper starts as follows: The main facts which a theory of
superconductivity must explain are (1) a second-order phase transition at the critical temperature, Tc, (2) an electronic specific heat

1264

8.
9.

varying as exp(T0/T) near T = 0 K and other evidence for an energy gap for individual particle-like excitations, (3) the Meissner
Ochsenfeld effect (B = 0), (4) effects associated with infinite conductivity (E = 0), and (5) the dependence of Tc on isotopic mass,
TcvM = const15.
For a historical account of the developments in the field of superconductivity see Schechter17 and Vidali18.
The BCS theory was accepted as a fundamental theory because of
its ability, according to the known features of superconductivity at
the time, to represent a coherent explanation of the phenomenon.
Nonetheless, it failed to account for many elements and was not
able to explain new features of superconductivity. In this regard,
although the LG model does not have a rigorous derivational origin, it was open to changes and was able to account for the newly
discovered high-temperature superconductors.

1. Shomar, T., Phenomenological realism, superconductivity and

quantum mechanics, Ph D thesis, London School of Economics,
London, 1998.
2. Cartwright, N., Shomar, T. and Surez, M., The tool-box of science. In Theories and Models in Scientific Processes (eds Herfel,
W. et al.), 1995, vol. 44, pp. 137149.
3. Shomar, T., Modelling in applied physics: The case of polymers.
Dirasat, Pure Science, 2006, 33, 241250.
4. Morrison, M., Unifying Theories: Physical Theories and Mathematical Structures, Cambridge University Press, Cambridge (in
press).
5. Grant, I. S. and Philips, W. R., Electromagnetism, John Wiley,
New York, 1982, p. 73.
6. Burns, G., High-Temperature Superconductivity: An Introduction,
Academic Press, San Diego, 1992, p. 10.
7. London, F. and London, H., The electromagnetic equations of the
supra-conductor. Proc. R. Soc. London, Ser. A, 1935, 149, 71.
8. Suarez, M., The role of models in application of scientific theories: Epistemological implications. In Modals as Mediators (eds
Morrison and Morgan, M.), Cambridge University Press, Cambridge, 1999.
9. Ginzburg, V. L., Some remarks concerning the macroscopic theory of
superconductivity. Sov. Phys. JETP, 1956, vol. 2, pp. 589600.
10. Sproull, R. and Phillips, W., Modern Physics: The Quantum Physics of Atoms, Solids and Nuclei, John Wiley, New York, 1976.
11. Bardeen, J., Theory of superconductivity. Encyclopaedia of Physics, Springer-Verlag, Berlin, 1956.
12. Schrieffer, J. R., Theory of Superconductivity, W.A. Benjamin
Inc., New York, 1964, p. 19.
13. Tilley, D. and Tilley, J., Superfluidity and Superconductivity,
Adam Hilger Ltd, Bristol, 1986.
14. De Gennes, P. G. and Pincus, P. A., Superconductivity of Metals
and Alloys, W. A. Benjamin Inc, New York, 1966, p. 176.
15. Bardeen, J., Cooper, L. N. and Schrieffer, J. R., Theory of superconductivity. Phys. Rev., 1957, 108.
16. Kittel, C., Introduction to Solid State Physics, John Wiley, New
York, 1986, p. 336.
17. Schechter, B., The Path of No Resistance, Simon Schuster, New
York, 1989.
18. Vidali, G., Superconductivity the Next Revolution?, Cambridge
University Press, Cambridge, 1993.
19. Anderson, P. and Schrieffer, R., A discussion of superconductivity. Physics Today, June 1991, p. 54.
20. Tinkham, M., Introduction to Superconductivity, McGraw-Hill,
New York, 1975.
21. Anderson, P., Science, 12 June 1992, 256.
Received 9 October 2007; revised accepted 28 March 2008

CURRENT SCIENCE, VOL. 94, NO. 10, 25 MAY 2008