You are on page 1of 8

Home Search Collections Journals About Contact us My IOPscience

Discovering by analogy: the case of Schrödinger's equation

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

1998 Eur. J. Phys. 19 69

(http://iopscience.iop.org/0143-0807/19/1/010)

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

Download details:

IP Address: 131.151.244.7
This content was downloaded on 16/08/2014 at 23:43

Please note that terms and conditions apply.


Eur. J. Phys. 19 (1998) 69–75. Printed in the UK PII: S0143-0807(98)85847-9

Discovering by analogy: the case of


Schrödinger’s equation

Constantinos Tzanakis
Department of Education, University of Crete, 74100 Rethymnon, Crete, Greece

Received 10 July 1997, in final form 6 October 1997

Abstract. Schrödinger’s equation is derived by using Hamilton’s mathematically unified


treatment of geometrical optics and analytical mechanics. This is done by a didactically
appropriate reconstruction of Schrödinger’s approach, in which the formal similarity of the two
theories is considered in the strict sense of the proportionality of the corresponding quantities in
the two theories. This is taken as an example emphasizing the crucial role analogy arguments
can play in the formulation of new physical ideas or theories and illustrating the role the
knowledge of the historical development of a subject can play in its presentation.

1. Introduction

In mathematics and physics we can distinguish at least three different types of reasoning (see
Polya’s distinction between demonstrative and plausible reasoning (Polya 1954)):
(i) Deductive reasoning, on the basis of which complete mathematical proofs are given, or
the foundations of a theory are laid. In this way, the validity of scientific results and their
subsequent acceptance by the scientific community is ensured; moreover, a logically clear
presentation of a scientific domain can be given, which thus appears as a firm edifice to any
one coming in contact with it.
(ii) Inductive reasoning, on the basis of which statistical inferences are based; conclusions
from empirical facts are drawn; conjectures concerning the generality of certain theoretical
results are formulated, on the basis of the examination of many (similar) cases; or, confidence
is acquired to the validity of a hypothesis, which is verified in many different cases, or its
long-time use has not led to contradictions.
(iii) Analogy arguments, on the basis of which new concepts or theories are formulated, or
already existing ones are further extended. Here, analogy is meant in the sense of a loose (or
sometimes, even strict) similarity of structures, with which a priori different sets of objects are
equipped. Besides the formulation of conjectures, in this way, concepts, theories or methods
can be generalized, so that they become applicable to classes of objects which are not of the same
kind as those to which these concepts, theories or methods originally apply. Consequently, new
research domains may emerge, by naturally establishing appropriate methodological and/or
conceptual frameworks (see the examples below).
It is clear, although sometimes forgotten (especially while teaching or writing textbooks),
that a deductive presentation of a mathematical or physical discipline is most suited to revealing
its logical structure and completeness and to secure the validity of the results obtained.

0143-0807/98/010069+07$19.50 © 1998 IOP Publishing Ltd 69


70 C Tzanakis

However, the motivation for the introduction of new concepts, theories or proofs is not
usually created deductively; first the basic idea is somehow grasped by some other means
(e.g., by induction or analogy), then one persuades oneself as to its validity, or at least that it
is worthwhile to investigate its consequences, (e.g., inductively), and then one tries to put it
on a firm basis deductively. Therefore, a strictly deductive (e.g., axiomatic) presentation of a
subject may be helpful to those knowing the subject, or at least having enough acquaintance
with it or other related subjects. Actually, in such a presentation the beginner ‘. . .is faced with
a closed system not knowing how [the] definitions were arrived at and absolutely nothing is
presented to his imagination. . .; the [presentation] has the drawback that it does not stimulate
thought’ (Klein 1979, page 316).
On the other hand, inductive reasoning is at the heart of all statistical inferences drawn
for example, from observations or experimental results. It is also important in theoretical
investigations, in the form of not absolutely rigorous or safe, but highly probable, arguments;
for instance, when many consequences of a certain hypothesis are known to be true, or when
long-time efforts of a scientific community to discover a counter-example to a given conjecture
have failed, people can convincingly argue that the hypothesis or conjecture is probably true
(e.g., Polya 1968). For example, prior to their proof, this was the case with Fermat’s last
theorem, and the theorem on the distribution of primes in number theory, or the positive-energy
theorem(s) in General Relativity; today, this seems to be the case with the cosmic censorship
hypothesis (in some of its various forms) in relativistic theory of gravitational collapse (‘[a]part
from the absence of known counter-examples, there is virtually no evidence for or against the
validity of this. . .conjecture’ (Wald 1984, page 305, see also Hawking and Penrose 1996).
Apparently, the role of analogy as a key pattern of reasoning in formulating and
understanding fundamentally new ideas and concepts, or in obtaining new results, has not been
sufficiently emphasized, though it is a common tool in scientific research. Think, for instance,
of the identification of isomorphic structures in mathematics research, as a motivation for the
subsequent development of a common theory, in which analogy arguments play a prominent
role (for a historically important example analysed in detail see Tzanakis 1995); or, think of
the way one manipulates concepts referring to new objects, the properties of which, however,
one knows from other more familiar cases, e.g., the integral and derivative concept in a Banach
space (e.g., Choquet-Bruhat et al 1982), or integration theory in measure spaces (e.g., Halmos
1950). In addition, there are very important examples of new theories or results in the history of
physics, that have been obtained using analogy as a basic methodological tool. The following
examples fall into this category:

• The matrix operations in matrix mechanics conceived by Heisenberg in 1925, on the basis
of Ritz’s combination principle and in analogy with the way operations with Fourier series
are carried out in classical wave theory (Heisenberg 1949, pages 109–13, van der Waerden
1967)†.
• Black-hole thermodynamics, developed on the basis of the analogy between certain
theorems in black-hole physics and the laws of classical (equilibrium) thermodynamics
(Bekenstein 1973, 1974, Hawking 1975, Wald 1984, section 12.5)‡.
• The formulation and proof of ‘singularity theorems’ in general relativity (Hawking and
Ellis 1973, O’ Neill 1983) which is often based on appropriate translations of concepts,
methods or theorems of Riemannian geometry, in the Lorentzian case (which constitutes
the geometric framework of general relativity), though in the Riemannian case, the
corresponding theorems have nothing to do with the existence of singularities (Beem and
Ehrlich 1981, Naber 1987); in other words, although these concepts, methods, or theorems

† ‘We shall now sketch the deduction of the fundamental equations of the new quantum mechanics. . . It should be
distinctly understood. . .that this cannot be a deduction in the mathematical sense of the word. . .’ (Heisenberg 1949,
page 108; my italics).
‡ ‘All the analogies we have mentioned are suggestive of a connection between thermodynamics and blackhole
physics. . .’ (Bekenstein 1973, page 2335).
Discovering by analogy 71

can sometimes technically be treated in the same way in the two cases, conceptually they
are very different, leading to completely new ideas.
• The formulation of wave mechanics by Schrödinger (Schrödinger 1982), using the formal
similarity between geometrical optics and analytical mechanics elaborated by Hamilton
(e.g., Mehra and Rechenberg 1987, pages 506–14, Dugas 1988)† and noted in the context
of quantum physics by de Broglie (de Broglie 1927, for the role of de Broglie see Mehra
and Rechenberg 1987, pages 502–6, Jammer 1965, Kragh 1982, pages 155–57, 169, Pais
1991, Brush 1983)‡.
In the present paper, we analyse the fourth example in detail and, in particular, we
reconstruct Schrödinger’s procedure that led him to the equation which bears his name. Our
purpose is twofold:
(i) to illustrate in some detail the use of analogy as a pattern of reasoning in active scientific
research;
(ii) to give an example of the way history can be useful in explaining the motivation behind the
introduction of basic ideas, concepts or theories, which remain totally unintelligible when
introduced deductively (for further examples in mathematics and/or physics see Tzanakis
1993, 1995, 1996, 1997a, b, see also Toeplitz 1963, Tomonaga 1962, 1966, Hairer and
Wanner 1996, Edwards 1977 for textbooks at least implicitly based on this idea).
It should be noted at this point that the historical development is not (and usually should not)
be respected in any strict sense, given that this development is almost never straightforward and
cummulative; on the contrary, it is rather often complicated, involving periods of stagnation and
confusion, full of prejudices and misconceptions, in which, new concepts or theories are not
introduced in the simplest and most transparent way. In the present case for instance, following
a strictly historical presentation, one had to go through Schrödinger’s various derivations of
his equation (including his relativistic considerations), given in his papers and unpublished
notebooks (Schrödinger 1982 papers I, II, Mehra and Rechenberg 1987, sections III.3, III.4,
Kragh 1982, especially sections 3, 4, 6, Kragh 1990; think, for example, of the rather ad hoc
and badly justified approach that Schrödinger adopted in his first paper on wave mechanics,
Schrödinger 1982, paper I).
Instead, the key steps of the historical development should be identified and used to motivate
its didactically appropriate reconstruction (see Brousseau 1983, Schubring 1977, Sierpinska
1991). An example is provided in section 2 (see also the comments in section 3).
The Schrödinger equation (SE) in such a genetic perspective has been treated somewhat
differently in Tzanakis and Coutsomitros (1988) and references therein. Here, the analogy
argument implicit in Schrödinger’s approach will be made transparent, showing that this
equation can be understood mathematically in the context of classical (analytical) mechanics,
though it could not have been anticipated by Hamilton as a physically meaningful equation for
reasons explained in the next section.

2. The Schrödinger equation

Although a cornerstone of quantum theory, the SE is often introduced to the beginner as an ad


hoc basic axiom, the relevance of which is to be decided a posteriori by its agreement with
† ‘The inner connection between Hamilton’s theory and the process of wave propagation is anything but a new
idea. . .Hamilton’s variational principle. . .correspond[s] to Fermat’s Principle for a wave propagation in configuration
space. . .and the Hamilton–Jacobi equation expresses Huygen’s Principle for this wave propagation. . .we must regard
[this] analogy as one between mechanics and geometrical optics and not. . .undulatory optics’ (Schrödinger 1982,
pages 13, 17).
‡ ‘Guidé par l’idée d’une identité profonde du principe de moindre action et de celui de Fermat, J’ai été conduit. . .à
admettre [que] les trajectoires dynamiquement possibles de l’un coı̈ncidaient avec les rayons possibles de l’autre’ (de
Broglie 1927, page 45).
72 C Tzanakis

experiment (Tzanakis and Coutsomitros 1988 and references therein). Roughly speaking, the
presentation is based on:
• a brief, verbal description of de Broglie’s idea on the wave properties of matter and the
use of his famous relations in the wave equation, to yield the SE for the free particle;
• the often ad hoc generalization of this equation to a system with many degrees of freedom
in the presence of external fields, thus leaving a logical gap while making the crucial step
of replacing the classical Hamiltonian function by the Hamiltonian operator.
The drawbacks of such an approach have been discussed (Tzanakis and Coutsomitros 1988,
Rüdinger 1976). Here we obtain the SE by exploiting Hamilton’s mathematically unified
treatment of geometrical optics and analytical mechanics and Schrödinger’s analogy argument
that as geometrical optics is an approximation to wave optics, so classical mechanics is an
approximation to some (new) wave mechanics†.
Geometrical optics are based on Fermat’s principle of least time for the light rays in any
optical medium, and classical mechanics of conservative systems on Maupertuis’ least action
principle. They are expressed as variational principles as (e.g., Goldstein 1980, section 8-6)
Z s2
c
δL ≡ δ n ds = 0 n≡ (1)
s1 v
Z  2
ρ2 √ 1 1 dρ
δW ≡ δ 2T dρ = 0 T = mij q̇ i q̇ j ≡ . (2)
ρ1 2 2 dt
Here L is the optical length, n the index of refraction, c, v the velocities of light in a vacuum
and in the medium considered, q i are generalized coordinates, W is Hamilton’s characteristic
function, and the kinetic energy T is assumed to be a positive-definite quadratic form of the
q i (the summation convention is used), so that dρ represents a Riemannian line element in
configuration space, analogous to ds in ordinary space. It can be shown (e.g., Pauli 1973,
and, more generally, Courant and Hilbert 1962), that (1), (2) lead respectively to the so-called
eikonal and Hamilton–Jacobi equations:
k∇Lk2 = n2 , k∇W k2 = 2(E − V ) (3)
E = T + V , V respectively being the total and potential energies of the system and in (3)
k∇W k2 = mij (∂W/∂q i )∂W/∂q j . Thus equations (1)–(3) become formally identical via the
correspondence

s↔ρ n ↔ 2T L ↔ W.
Actually in the geometrical optics approximation of wave optics, a light wave of frequency ω
and wavevector k is given by the wavefunction (e.g., Landau and Lifshitz 1969)
ω
ψ(x, t) = ψ̂(x)eiφ φ(x) ≡ k0 L(x) − ωt k0 = (4)
c
where k gives the direction of the light rays
∂φ ω
k = ∇φ ω=− v= k = kkk (5)
∂t k
and L satisfies the first expression in (3), which follows from the wave equation, as shown by
Sommerfeld and Runge in 1911 (Mehra and Rechenberg 1987, pages 519–20, Sommerfeld
1964, Goldstein 1980, section 10-8) and acknowledged by Schrödinger (Schrödinger 1982,
† ‘. . .our classical mechanics is the complete analogy of geometrical optics. . . Then it becomes a question of
searching for an undulatory mechanics. . .[by] working out. . .the Hamiltonian analogy on the lines of undulatory
optics.’ (Schrödinger 1982, paper II, page 18). This analogy was clear to Schrödinger, as early as 1918, as can be seen
from his notebooks, where it is mathematically worked out in detail and applied to other fields as well, like general
relativity and thermodynamics (Mehra and Rechenberg 1987, pages 522–32).
Discovering by analogy 73

footnote 1 to page 18). On the other hand, in classical mechanics, E and the generalized
momenta p ≡ (pi ) are similarly given by
∂S
p = ∇S E=− S = W − Et (6)
∂t
S being Hamilton’s principal function, playing the role of φ in (6). Clearly, equations (1), (2),
(4)–(6) imply the isomorphism of the two theories under the correspondence
p↔k E↔ω S↔φ W ↔ L. (7)
Therefore, mechanical motion may be viewed as taking place along rays given by the normal
to the ‘wave fronts’ defined by S = constant, an idea fully appreciated and elaborated by
Hamilton.
Now Schrödinger’s approach may be reconstructed in the following way. The analogy
(isomorphism) (7) can be taken in the strict sense of proportionality
p = αk E = βω S = σφ W = ζL (8)
where the factors of proportionality in general depend on E. This is very close to de Broglie’s
original ideas that led him to the generalization of the Planck–Einstein relations to matter (de
Broglie 1927, chapter II and page 45,(quoted in the second footnote to page 71 of this paper).
Inserting equations (8) in (6) and using (3), (4), (5), we obtain
E
α=β=σ ζ = (9)
c
which in turn imply that the phase speed of the mechanical waves is
|E| |E|
v= =√ . (10)
k∇W k 2(E − V )
Thus, there is only one independent parameter, say σ . To clarify its meaning, we insert ψ in
(4), in the wave equation ∇ 2 ψ − (1/v 2 )∂ 2 ψ/∂t 2 and taking account of (6), (8), (10), we obtain
∂ψ σ2
iσ = − ∇ 2 ψ + V ψ. (11)
∂t 2
Provided σ is independent of E, this is formally identical to the SE, although it has been derived
entirely within the mathematical framework of classical mechanics. Note that equation (11)
is independent of the number of degrees of freedom (in general ∇ 2 is the Laplacian in
configuration space, considered as a Riemannian space with line element dρ given by (2)).
Now it becomes clear why Hamilton could not have formulated wave mechanics, though
in principle he could have derived equation (11):
(i) At that time, there was no deep physical reason for considering the formal similarity between
geometrical optics and classical mechanics as being anything more than a mathematical
isomorphism, in as much as the derivation of the eikonal equation from the wave equation
came much later (see the discussion following (5) and Kragh 1982, pages 161–2); actually,
Hamilton never attributed a deeper significance to this similarity (Mehra and Rechenberg 1987,
page 513). As emphasized by Schrödinger, it did not attract sufficient attention after Hamilton,
despite Klein’s use of this optical analogy in the development of the Hamilton–Jacobi theory
as a kind of optics in a multidimensional configuration space (Klein 1979, pages 186–7,
Schrödinger 1982, page 13 and footnote 3 thereto).
(ii) Even if equation (8) is assumed so that (11) follows, we need an extra hypothesis for (11)
to be the SE, namely that σ is a universal constant. Taking account of (9) in (8) for p, E,
this extra condition is provided by de Broglie’s insight to extend the Planck–Einstein relations
to all ‘material’ particles. This was the crucial physical idea which led from Hamilton’s
mathematically unified treatment of the conceptually so different theories of geometrical optics
and analytical mechanics, to a deep, fruitful physical theory of atomic phenomena (Kragh 1982,
page 168).
74 C Tzanakis

3. Concluding remarks
In this paper, inspired by the historical development of wave mechanics, we have obtained
Schrödinger’s equation, as an important example illustrating the role which analogy arguments
have played in the formulation of new physical concepts or theories, thus emphasizing the
importance of such arguments in the teaching or learning process.
In this connection, the paper stresses the role which a basic knowledge of the historical
development of a subject can play in its presentation: first the key steps of this development are
appreciated (here, the analogy between equations (1), (2) and Hamilton’s elaboration of it) and
then they are reconstructed in a modern context, so that they become didactically appropriate.
In the present case, for instance, our reconstruction is closer to Sommerfeld’s derivation of
the SE in 1929 (Mehra and Rechenberg 1987, pages 520–2, see Tzanakis and Coutsomitros
1988, page 280), than to Schrödinger’s approach, in which the optical analogy is used in a
somewhat different way (Schrödinger 1982, paper II, particularly section 2, see Kragh 1982,
pages 166–8, Yourgrau and Mandelstam 1968, section 11); moreover, many details can be
left as carefully designed exercises, enabling the student to become better acquainted with the
subject (for example, parts of the derivation of equations (3) from (1), (2), or of the eikonal
equation from the wave equation, etc, can be left as exercises). For more details on this general
scheme, see Tzanakis (1993, 1995, 1996, 1997b).
For the SE, the advantages of this approach are evident (for details see Tzanakis and
Coutsomitros 1988):
• the deep relationship between classical and quantum mechanics is clearly illustrated;
• the SE is derived independently of the number of degrees of freedom of the system under
consideration, a basic difficulty in conventional approaches;
• semiclassical approximations (in particular the WKB method), or the idea of the path
integral approach, are easily introduced on the basis of equations (4), (8);
• no physical interpretation of ψ is needed in order to derive the SE. Therefore the student
is free to think about the physical content of the formalism, already having the SE at
his disposal. Thus he has a better opportunity of gaining a deeper understanding of the
conceptual foundations of the theory.

References
Beem J K and Ehrlich P E 1981 Global Lorentzian Geometry (New York: Dekker) chs 10, 11
Bekenstein J 1973 Phys. Rev. D 7 2333 (particularly section II)
Bekenstein J 1974 Phys. Rev. D 12 3992
Brousseau G 1983 Recherche en Didactique des Mathématiques 4 (165) 193
Brush S G 1983 Statistical Physics and the Atomic Theory of Matter (Princeton, NJ: Princeton University Press)
pp 131–2
de Broglie L 1927 Ann. Phys., Paris 3 22
Choquet-Bruhat Y, de Witt-Morette C and Dillard-Bleick M 1982 Analysis, Manifolds and Physics (Amsterdam:
North-Holland) ch II.A
Courant R and Hilbert D 1962 Methods of Mathematical Physics vol II (New York: Wiley) section II.9.2
Dugas R 1988 A history of mechanics (New York: Dover) part IV, ch VI
Edwards H M 1977 Fermat’s Last Theorem: A Genetic Introduction to Algebraic Number Theory (Berlin: Springer)
Goldstein G 1980 Classical Mechanics (Reading, MA: Addison-Wesley)
Hairer E and Wanner G 1996 Analysis by its History (Berlin: Springer)
Halmos P R 1950 Measure Theory (New York: Van Nostrand) ch V
Hawking S W 1975 Commun. Math. Phys. 43 199
Hawking S W and Ellis G F R 1973 The Large-scale Structure of Spacetime (Cambridge: Cambridge University
Press) ch 8
Hawking S W and Penrose R 1996 The Nature of Space and Time (Princeton, NJ: Princeton University Press) pp 29-30
Heisenberg W 1949 The Physical Principles of the Quantum Theory (New York: Dover) appendix
Jammer M 1965 The Conceptual Development of Quantum Mechanics (London: McGraw-Hill) section 5.3
(particularly pp 257–8)
Discovering by analogy 75

Klein F 1979 Development of Mathematics in the 19th Century (Brookline, MA: Mathematics and Science Press).
Originally published as: 1928 Vorlesungen uber die Entwicklung der Mathematik im 19 Jahrhundert (Berlin:
Springer)
Kragh H 1982 Centaurus 26 154
—— 1990 Dirac: A Scientific Biography (Cambridge: Cambridge University Press) pp 49–51
Landau L and Lifchitz E 1969 Électrodynamique des milieux continus (Moscow: MIR) ch X
Mehra J and Rechenberg H 1987 The Historical Development of Quantum Theory vol 5 (Berlin: Springer)
Naber G L 1987 Spacetime and Singularities (Cambridge: Cambridge University Press) ch 4, p VIII
O’ Neill B 1983 Semi-Riemannian Geometry (New York: Academic Press) ch 14
Pais A 1991 Niels Bohr’s Times (Oxford: Oxford University Press) section 14(b)
Pauli W 1973 Optics and the Theory of Electrons (Cambridge MA: MIT Press) ch I
Polya G 1954 Induction and Analogy in Mathematics (Princeton, NJ: Princeton University Press) preface
—— 1968 Patterns of Plausible Inference (Princeton, NJ: Princeton University Press) ch XV
Rüdinger E 1976 Am. J. Phys. 44 144
Schrödinger E 1982 Wave Mechanics (New York: Chelsea) paper II, lecture I. Paper II originally published as: 1926
Ann. Phys. (4) 79; lecture I originally published as: 1928 Four Lectures on Wave Mechanics (Glasgow)
Schumbring G 1977 Zentralblatt für Didaktik der Mathematik 9 (209) 211
Sierpinska A 1991 Cahiers de didactique des Mathématiques 7 (11) 13
Sommerfeld A 1964 Optics (New York: Academic) section 35.A, footnote to p 207
Toeplitz O 1963 Calculus: A Genetic Approach (Chicago, IL: Chicago University Press)
Tomonaga S I 1962 Quantum Mechanics vol I (Amsterdam: North-Holland)
—— 1966 Quantum Mechanics vol II (Amsterdam: North-Holland)
Tzanakis C 1993 Proc. 1st European Summer University on History and Epistemology in Mathematics Education ed
F Lalande, F Jaboeuf and Y Nouazé (Montpellier: IREM de Montpellier) p 271
—— 1995 Int. J. Math. Education Sci. Techn. 26 45
—— 1996 Proc. 2nd European Summer University on History and Epistemology in Mathematics Education vol II ed
M J Lagarto, A Vieira and E Veloso (Braga, Portugal: Universidade do Minho) p 96
—— 1997a Proc. 2nd World Congress of Nonlinear Analysts ed V Lakshmikantham (Amsterdam: Elsevier) to be
published
—— 1997b Presenting the relation between Mathematics and Physics on the basis of their history: a genetic approach
Using History of Mathematics in Teaching Mathematics ed V Katz (Washington, DC: The Mathematical
Association of America) to be published
Tzanakis C and Coutsomitros C 1988 Eur. J. Phys. 9 276
van der Waerden 1967 Sources of Quantum Mechanics (New York: Dover) ch 12 (particularly section 1)
Wald R 1984 General Relativity (Chicago, IL: Chicago University Press)
Yourgrau W and Mandelstam S 1968 Variational Principles in Dynamics and Quantum Theory (Philadelphia, PA:
Saunders)

You might also like