Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

On Gibbs-Heaviside “interaction at a distance”: the curl of a tensor and the generalization

of Divergence and Stokes theorems.

Slide 1

First of all I would like to thank the local committee for organizing the conference and also for
having accepted my contribution.

Slide 2

This is the outline of the talk

Slide 3: INTRODUCTION

Slide 4

According to the American historian Micheal Crowe, the roots of modern vector and tensor
calculus go back to the end of the nineteenth century, when Josiah Willard Gibbs and Oliver
Heaviside independently developed the modern system. The two scientists worked on opposite
sides of the Atlantic Ocean.

Slide 5

Gibbs was born in New Heaven (Connecticut), he journeyed to Europe for three years of study
divided between Paris, Berlin, and Heidelberg, then he returned to New Heaven where he was
made professor of mathematical physics at Yale, a position he held until his death.

Slide 6

Oliver Heaviside was born here in Candem Town, he worked as a telegrapher in Denmark then
came back to London. Plagued by poor health, Heaviside had to deal with deafness and other
illnesses throughout his life. Maybe this was why he remained a reclusive bachelor when he
moved to Torquay in Devonshire where he died. The blue plaque has been inaugurated very
recently!
Slide 7

The modern vector calculus emerged from Hamilton’s system of quaternions, but the
transition was neither smooth nor easy. Let me spend a couple of words on that. The Scottish
mathematicians Peter Tait and Cargill Knott, ardent quaternionists, took Gibbs to task because
of his misuse of Hamilton's mathematical invention. Their criticism was that Gibbs had merely
invented a new notation and had actually made mathematical matters worse by not using
Hamilton’s quaternions. The acrimonious confrontation started in the preface to the third edition
of Tait’s Treatise on Quaternions (1890). Tait expressed his disappointment at “how little
progress has recently been made with the development of Quaternions”. Tait went on to
remark: “Even Prof. Willard Gibbs must be ranked as one of the retarders of Quaternion
progress, in virtue of his pamphlet on Vector Analysis; a sort of hermaphrodite monster,
compounded of the notations of Hamilton and of Grassmann”. Tait was referring to Gibbs’
Elements of Vector Analysis, which marks the birth of modern vector calculus. In the April 1891
issue of Nature, next year, there appeared an article by Gibbs written in response to Tait's
references to Gibbs and Tait's reply was published within the month. Gibbs’ second article was
published in Nature four weeks after Tait's article.

Slide 8

The debate continued also in Philosophical Transaction where Heaviside defended Gibbs’s
position: Heaviside had great respect for Gibbs’s work, although he did not like, and sometimes
criticized, Gibbs’s choices for vector notation. Heaviside expressed a very different opinion. In
1892, he added some footnotes to his Electromagnetic Theory (Vol.2), which republished his
paper appeared in another journal, The Electrician, where he clearly stated “This rather one-
sided discussion arouse out of Professor Tait stigmatizing Pr. Gibbs […] Prof. Gibbs is anything
but a retarder of progress in vector analysis and its application in physics” (p. 528). And
regarding Gibbs booklet: “Gibbs’ work is scarcely known, and deserves to be known […] His
treatment of linear vector-operator is specially deserving of notice” (p. 529). Regarding the
importance of quaternions, Heaviside stated clearly that “it rarely occurs that any advantage is
gained by the use of the quaternion”. Also in (later) review of the book Vector Analysis, based
on Gibbs’ pamphlet and written by Gibbs disciple Edwin Wilson, Heaviside recollected his
personal difficulties in introducing himself to quaternionic system, after having known it through
Maxwell’s Treatise, and emphasized that he realized that “quaternionic can be employed
consistently in Vectorial work” (EMT2; p. 557). Indeed, both authors abandoned the quaternionic
system to develop the vector calculus. Heaviside remarked: “Up to 1888 I imagined that I was the
only one doing Vectorial work on positive physical principles; but then I received a copy of Prof.
Gibbs’ Vector Analysis (unpublished, 1881-4). This was a sort of condensed synopsis of a treatise.
Though different in appearance, it was essentially the same Vectorial algebra and analysis to
which I had been led.” This debate informed us that Heaviside studied Gibbs’ work, hence, we
addressed the question: what was the impact that Gibbs had on Heaviside’s work after he
received Gibbs’ pamphlet? Here, we shall focus on some applications of differential and integral
operators to vectors and dyadics. Our analysis is based on the paper:

Arch. Hist. Exact Sci. DOI 10.1007/s00407-020-00264-x

Slide 9: Interaction at a distance

Slide 10

Elements on Vector Analysis was intended as a quick guide. The historian Elizabeth Gerber
specifies that “in his work Gibbs went beyond Tait and others in his ability to treat physical
problems. He demonstrated how vector analysis could be used in astronomy and gave courses in
vector analysis with applications to electricity and magnetism”. The pamphlet was printed for
use only in Gibbs's classes at Yale, hence the phrase "not published" on the cover sheet. Heaviside
introduced his elements of vector analysis in the context of electromagnetic theory, but he also
tried to give a more pedagogical and complete treatment of the subject. Heaviside’s published
work is contained in the two-volumes Electrical Papers (EP) and in the three volumes collection
known as Electromagnetic Theory (EMT).

In his lectures on vector calculus, Heaviside introduced the nabla operator and popularized
the term “divergence”. Crowe stated that Heaviside used the term “divergence” for the first time
in 1883 and the historian Charles Stolze emphasized that Heaviside introduced it for the first time
in connection with the famous Theorem in 1893. By analyzing Heaviside’s work, we clarified that
Heaviside introduced it on December 2, in 1882, when his paper by Heaviside was published in
The Electrician. In this paper, Heaviside explicitly considered the ‘Theorem of Divergence’
(Heaviside 1892a, p. 209). Heaviside presented proof of the theorem emphasizing that he had
read about it ‘in a German work’ (Heaviside 1892a, p. 208), without quoting it.

Heaviside introduced some innovations in vector analysis until 1885, but it was after receiving
a copy of Gibbs' work, in 1888, that he improved his use of linear operators praising Gibbs’
developments in this context. Heaviside and Gibbs never met. Hence, we can speak of a sort of
“action at a distance”. But this action at a distance was supported by a mediator and we had the
opportunity of studying it. Indeed, a rare copy of Gibbs’ work is kept in custody at the Dibner
Library of the History of Science in Washington, DC, which contains a number of manuscript notes
in the text. Its description on Dibner’s website of the collection Heralds of Science declared that
there is an inscription on the cover “From the author, June 1888” and that “there are copious
manuscript notes, apparently in the same hand” written by Heaviside. This suggested that the
booklet held in the Dibner Library is the copy received and by Heaviside. Firstly, we verified the
authenticity of the handwriting. For this purpose, we used manuscripts signed by Heaviside
himself (MSS 677 A) also held at the Dibner Library.

Slide 11

Secondly, I addressed the question: when did Heaviside write his annotations? I did not have
access to any of Heaviside’s personal diaries, if they exist. Therefore, the holographs cannot be
dated exactly, but by analyzing Heaviside’s comments in his published work, I argued that the
shortest lapse of time can start with June 1888 (of course the year Heaviside received the
booklet) and April 1892. Indeed, I recognized Heaviside’s use of specific terms, namely dyad and
dyadic, used by Gibbs for linear operator and I identified the published version of some
annotations. Let me show you the structure of the pamphlet.

Slide 12

Ch.1 vector algebra. In chapter two Gibbs introduced the nabla operator, using the usual
Assyrian harp (EMT1; p178) symbol ∇, and its applications (the mathematics of potential theory).
After having introduced line, surface and volume integrals, Gibbs presented Stokes’s theorem
and the Divergence theorem for vector fields. The chapter ends with the introduction of
irrotational and solenoidal vector fields. In chapter three Gibbs introduced the reader to the
concept of linear operators, by defining the terms dyad and dyadics. Gibbs considers the
divergence and the curl of a linear operator and presents a generalization of the Divergence and
Stokes’ theorem for linear operators. I’ll skip the other chapters.

Also Heaviside published in March 1892 both an “Extension of the Theorem of


Divergence” (EMT2; p. 190) and Stokes theorem, which he called “The Theorem of Version”
(EMT2; p. 191). Hence, Heaviside conceived his extensions after having received Gibbs’ booklet.
Furthermore, there are also some annotations where Gibbs introduces the scalar and the cross
product between the nabla operator and dyadics. Let us see how the influence that Gibbs had
emerges from the annotations.

Slide 13: The curl of a tensor

I will start with the latter by showing you how Heaviside reacted when he met Gibbs’ application
of the differential calculus to dyadics.

Slide 14

To appreciate Heaviside’s annotations, let me briefly show you the different notations involved.
Heaviside adopted Clarendon’s notation with Latin letters, while Gibbs used Greek letters only.
(a funny thing: I recently discover, thank to prof. Duncan Agnew (Univ San Diego) that Clarendon
was the name of the font produced at the time!) Heaviside used the period as a mere separator,
while Gibbs introduced it as a symbol for the scalar product. The period became a dot like the
one used today in Wilson’s book (Wilson-Gibbs 1901, p. 55). The use of different notations
reflected the different philosophical approach the author had towards mathematics. I have no
time to discuss it: let me just say that Gibbs’s notation and Gibbs’s philosophy survived, as far as
I know for reasons not fully explored yet.

In Gibbs’s booklet, the tensor product of two vectors is called a dyad and it represents a linear
operator dyads can be identified with special tensors because a generic element of the vector
space V ⊗ V cannot always be written as the tensor product of two vectors. Using modern
language, they correspond to simple, or elementary, tensors (I-Shih 2002). In Gibbs’s language, a
linear combination of dyads is a dyadic. In the last line you have a dyadic constructed using six
vectors a, b, c, l, m and n: it represents the more general form of a linear operator. In the
application considered by Heaviside and Gibbs, dyadics can be identified with generic tensors of
rank two.

Slide 15

Heaviside was particularly interested in Gibbs’ application of nabla operator to the dyadics (p. 66
of the booklet), because since 1886 (on the left) Heaviside had discussed mechanical forces and
stresses by using the divergence of a tensor without using dyads. ‘the translational force due to
stress’ (Heaviside Electrical Papers 1892a, p. 543). The force is defined using three vector stresses
P and hence using a second rank tensor. Heaviside repeated quite a similar presentation in 1891,
i.e. after having studied Gibbs’s booklet (Heaviside 1892b, p. 533), using the nabla symbol. On
the right you have his annotations: you can see in Gibbs’ formulas the curl and the divergence of
the tensor φ, written in different forms.

Slide 16

Here you have the transcription of Heaviside’s annotations: Heaviside recognized the
connection between Gibbs’s abstract definition and the theory of continuum mechanics. He used
the same letters, for stresses, force and the use of nabla, that we can find in his second
presentation dated 1891. Heaviside recognized the equivalence between the tensor he used for
defining his stress vectors and the divergence of a dyadic, rewrote it using his notation as a sum
of dyads like Gibbs. And then comes the curl of a tensor.

Slide 17

On the back cover, he tried to write down its components starting from his tensor for stresses
and struggled with the question: What is [the physical] meaning of curl φ ? In red, you have the
components using index notation (Einstein convention is implied)
As far as we know, neither the Electrical Papers nor the Electromagnetic Theory contains an
explicit discussion of this mathematical tool, i.e. the curl of a tensor, and Heaviside never
addressed explicitly the question of its physical meaning. Despite this, first, some of the formulas
used by Heaviside seem to involve this concept implicitly in the electromagnetic context, as I will
tell you in a moment. Second, he expressed explicit appreciation for this concept in a paper
published by Nature at the beginning of 1894. Heaviside considered Gibbs’ advances in the theory
of dyadics and emphasized: ‘see Gibbs’s “Elements of Vector Analysis” (1881-4a) for the direct
[cfr. Scalar] product of ∇ and φ. (Also for the skew product, a more advanced idea; it, too, is a
physically useful result.)’ [emphasis added]. This statement joint with the annotation shows the
impact that had Gibbs’ booklet on Heaviside and that Heaviside met this concept by studying
Gibbs’ work. But it is not clear to me where he found it useful. Let us see why.

Slide 18

The curl of a tensor is a mathematical object which has different applications in physics, but
during the 1890s the applicability of this concept had not yet emerged explicitly. In the paper I
briefly discuss the “emergence” of this concept in the three frameworks connected with Gibbs-
Heaviside interaction: the context of electromagnetic theory in Heaviside’s published work, the
context of the theory of elasticity, where Heaviside asked himself the meaning for the first time,
and the context of abstract vector calculus in Gibbs’s pamphlet. Here, I briefly mention the first
frameworks.

Since 1883, by generalizing Ohm’s law, Heaviside had been considering the existence of
electrical ‘eolotropy’ (Heaviside 1892a, p. 286), i.e. the phenomenon whereby the electric
conductivity of a body depends on the direction in which it is measured. To describe the
phenomenon, Heaviside introduced the linear operators. In his investigation on the origin of
double refraction in the context of an elastic ether, in 1893, in the Proceedings of the Royal
Society, Heaviside not only quoted Gibbs’ term dyadic, but he also adopted explicitly Gibbs’
notation, like in his annotations on the booklet. because there are formulas where the vector
product between nabla and an operator is present, but it is always applied to a vector as you can
see in the slide.
Slide 19: The extensions of Divergence and Stokes Theorems.

Let us move to the last part of the talk. The historian Ido Yavetz described in 1995 the
generalizations of the Divergence and Stokes theorem presented by Heaviside. Heaviside gave a
proof of the Divergence Theorem in 1886 for elementary surfaces and volumes (unit cubes).

Ten years later, he had studied Gibbs work and here are the annotations:

Slide 20

Heaviside translated in his notation equations 1,3,5. On the right, there is also an additional
formula at the bottom. Let us consider the generalization of the Divergence Theorem.

Slide 21

I’m using modern notation for Gibbs formulas on the left and the original Heaviside notation on
the right. Equation 3 is the Divergence Theorem, eq. 1 is a sort of extension for the scalar
functions and eq. 5 involves the cross product. N is the unit normal to the surface. We notice that
Heaviside seemed to be interested in scalar and vector functions only, because he did not
translated the other formulas involving the dyadic Phi. First, let us notice that in Heaviside’s
notation the difference between surface and volume integrals is implied and the same was for
Stokes-like formulas. Second, the left-hand parts of the translated formulas can be interpreted
as odd functions of the unit normal N. Indeed, in enunciating his extension of Divergence
Theorem, Heaviside noticed that a key ingredient in the proof is the fact that, I’m quoting
Heaviside, “the validity of the process whereby we pass from a surface- to a volume-summation,
depends solely upon the quantity summed up, [the scalar product between the normal and the
vector field], changing its sign with N”. Hence, he inferred (actually without proof), that using this
hypothesis the theorem can be extended.

Slide 22

Follow the right column for Heaviside’s notation or the left one for Gibbs/modern notation.
Heaviside stated that, second row, if the capital F is an odd function of the unit normal N, hence
the surface volume can be transformed the a volume integral of a function f. He specified that in
its Cartesian form the function, when the surface is a cubical element, should have the form in
the third row. Hence, from my point of view, it is highly probable that Heaviside started to
elaborate his generalization of the theorem after reading Gibbs’s booklet. But another question
arises. Indeed, one of the non-translated Gibbs formulas seems to give the same result. Are the
two result connected? Was Heaviside cheating? (I’m kidding! He didn’t…). In none of his
annotations, Heaviside translated an integral formula involving the dyadics, maybe because he
did not find a physical application. Furthermore, Gibbs’ formula works if all the functions are
supposed to be differentiable in the sense of functions. Heaviside did not discuss the properties
of the unit vector field N(x). If the surface is a smooth orientable manifold, the Gauss map
provides a continuous differentiable normalized vector field over the whole volume enclosed by
the surface (except for a set of null-measure); Heaviside’s result is a special case of Gibbs’
formula. But reading Heaviside’s Electromagnetic Theory, the vector field N seems to be intended
as living on the surface only. This means that, from the three-dimensional point of view, it does
not correspond to a continuously differentiable vector field. And indeed, in his proof, Yavetz
(even if he did not underlined this fact), the additional Heaviside’s additional hypothesis is
needed to reconstruct by-hand the derivative of capital F function (giving the nablas). Nowadays,
we know that it can be generalized for tempered distributions, by introducing the correct
measure in order to carefully give sense to the integrals involved (Šilhavý. 2005; van Dijk 2013)
and the introduction of Heaviside’s hypothesis can be avoided.

I have no time to discuss the Stokes part. Let me only say that there is an “additional
formula”, that Heaviside published, but in its published version the order of the various part of
the formula is different. From my point of view, this is another hint of the fact that Heaviside
conceived his extensions by studying Gibbs booklet.

Slide 23

A final remark. At present, I am curing the archive of the Solvay Institutes and the Solvay
conferences, in Brussels. I would like to attract your attention to the fact the proceedings of the
conference could be interesting for exploring the application of mathematics to modern physics.
Hence, you can find them on the website of the Solvay Science Project (see at the bottom of the
Institutes’ webpage).

Thank you for your attention.

You might also like