Available online at www.sciencedirect.

com

ScienceDirect
Historia Mathematica 44 (2017) 150–169
www.elsevier.com/locate/yhmat

Arthur Cayley, Robert Harley and the quintic equation: newly

discovered letters 1859–1863
Tony Crilly a , Steven H. Weintraub b , Paul R. Wolfson c,∗
a Middlesex University, The Burroughs, London, NW4 4BT, UK
b Department of Mathematics, Lehigh University, Bethlehem, PA 18015, USA
c Department of Mathematics, West Chester University, West Chester, PA 19383, USA

Available online 24 October 2016

Abstract
Beginning in the 1840’s, Arthur Cayley (1821–1895) led a vast invariant theory programme in algebra. After learning of results
of James Cockle (1819–1895) and Robert Harley (1828–1910), he applied the techniques of invariant theory to the calculation of
resolvents of quintic equations. Letters recently discovered reveal the priorities of Cayley and Harley with respect to the quintic,
an approach which was at variance with that via the theory of groups. As another recently discovered manuscript reveals, Cayley
returned to this subject in his final days.
© 2016 Elsevier Inc. All rights reserved.

Résumé
À partir de la période de 1840, Arthur Cayley (1821–1895) a mené un vaste programme de la théorie des invariants en algèbre.
Après avoir appris les résultats de James Cockle (1819–1895) et Robert Harley (1828–1910), il a appliqué les techniques de la
théorie des invariants pour le calcul de résolvantes d’équations du cinquième degré. Lettres récemment découvertes révèlent les
priorités de Cayley et Harley par rapport aux équations du cinquième degré, une approche differente que par la théorie des groupes.
Comme un autre manuscrit récemment découvert révèle, Cayley est revenu à ce sujet dans ses derniers jours.
© 2016 Elsevier Inc. All rights reserved.

MSC: 01A55
Keywords: Arthur Cayley; Robert Harley; James Cockle; Quintic equation; Resolvent equation; Invariant theory

* Corresponding author.
E-mail addresses: t.crilly@btinternet.com (T. Crilly), shw2@lehigh.edu (S.H. Weintraub), pwolfson@wcupa.edu
(P.R. Wolfson).

http://dx.doi.org/10.1016/j.hm.2016.09.003
0315-0860/© 2016 Elsevier Inc. All rights reserved.
 T. Crilly et al. / Historia Mathematica 44 (2017) 150–169 151

1. Introduction

When Arthur Cayley’s wife died in 1923 his correspondence, assiduously collected for over half a cen-
tury, was mostly destroyed. Aside from some letters written to and from James Joseph Sylvester, held in
an archive at St. John’s College Cambridge, few written to Cayley survive, and those written by Cayley are
scattered. This background makes the coming to light of a batch of letters written by Cayley to the little
known mathematician Robert A. Harley especially valuable, even though the period of this batch is brief
– a mere four years. Robert Broadhead Honeyman, a well-known manuscript collector, donated a large
collection of material to the Lehigh University library in 1958. Much of this collection lay unexamined
by scholars for decades, but in the past few years it has been digitized and made available on the library’s
website [Lehigh]. This collection includes a set of 40 letters from Arthur Cayley to Robert Harley, written
between 1859 and 1863. It also includes a manuscript, drafts of a Memoir on the Quintic Equation, on
which Cayley was working at the time of his death in 1895 which, as will appear, is related in subject mat-
ter to some of the letters. We report on this material, which sheds light on Cayley’s working relationship
with Harley, and by extension on his relationship with coworkers, and also on Cayley’s lifelong fascination
with the quintic equation. We note that the letters are half of a correspondence, the other half of which has
not been preserved. (Actually, it is slightly less than half, as some of the letters refer to material sent from
Cayley to Harley that is not part of this collection and apparently has not been preserved.) The letters are all
simply addressed “Dear Sir”, but we have been able to identify the recipient by internal evidence, as will
be clear below. Also, the draft Memoir is undated, but again, internal evidence allows us to date it, as will
also be clear below.
Cayley’s papers sometimes present the results of extensive calculations without providing much context.
These letters make clear that many of the papers from the period belong to a large research programme
concerning the quintic equation which encompassed other English mathematicians with whom Cayley
interacted. We begin this paper by giving the personal background to the correspondence, then give an
introduction to the mathematics discussed in the correspondence and in the draft Memoir. In the heart of
the paper, we consider the Cayley–Harley correspondence, followed by a brief summary of the contents of
the Memoir and a conclusion.

2. The correspondents and the occasion of their correspondence

The letters in the Lehigh collection begin when Cayley was 38 years and Harley 31 years old. During
the 1850s Cayley was firmly established as England’s leading pure mathematician, but he was unsuccessful
in finding an academic position until June 1863, (just after the end of the correspondence being discussed
here) when he became the first Sadleirian professor of Pure Mathematics at Cambridge. Instead, he was
practicing law, based at Stone Buildings Lincoln’s Inn in London. His professional activity lay in drafting
legal documents, an activity which hardly demanded his appearance in court and to a large extent allowed
him to regulate his own time. (For more biographical details, see [Forsyth, 1895], [Crilly, 2006].) During
the same period, Harley was a Congregationalist minister at Brighouse in Yorkshire. In 1844, at the age of
16, he had become interested in problems in “The Lady’s and Gentleman’s Diary” (the magazine in which
the famous Kirkman’s schoolgirls problem appeared in 1850), which brought him into contact with James
Cockle, then well-known to English mathematicians. Harley was very much interested in, and influenced
by, Cockle’s work, and they became life-long friends. Indeed, upon Cockle’s death in 1895 Harley wrote his
obituary for the Proceedings of the Royal Society [Corley and Crilly, 2004]. In the 1850s Cockle and Harley
joined forces on the quintic problem, and between 1858 and 1862 they were in almost daily communication.
Harley also became a good friend, and follower, of George Boole. Harley’s early mathematical work was
concentrated in algebra, particularly in the theory of equations and invariant theory, which were closely
linked, but later on he became interested in differential equations, particularly in their connection with
152 T. Crilly et al. / Historia Mathematica 44 (2017) 150–169

algebraic equations. (Further biographical information on Harley may be found in [MacMahon, 1914–1915]
and [Bell and Panteki, 2004].)
During the period of the correspondence, Cayley was writing a number of papers on the theory of equa-
tions. His investigations in the theory of equations had begun with his work on symmetric functions and his
revision of the tables of Meyer–Hirsch [Cayley, 1857], and his early work included an unconvincing proof
of the classical theorem that every algebraic equation has a root [Cayley, 1859], but he really concentrated
on the quintic equation in 1860–1862, the period of his correspondence with Harley, evidently stimulated
by the work of Harley and Cockle. A flurry of papers dealt with the equation of differences, Tschirnhausen
transformations, and resolvents [Cayley, 1860a, 1860b, 1861a, 1861b, 1861c, 1861d, 1861e, 1861f, 1861g,
1862a, 1862b, 1862c]. Thereafter, he worked sporadically on the quintic equation throughout his life [Cay-
ley, 1868, 1880, 1882, 1890]. His interest was revived a last time as he was editing his Collected Works for
publication. (He edited the first seven volumes; the remainder appeared posthumously and were edited by
Forsyth, his successor as Sadleirian Professor.) He wrote a seven-page commentary [Cayley, 1891] on his
paper [Cayley, 1861d] discussing work on the quintic by Emory McClintock [McClintock, 1884, 1886];
under other circumstances Cayley would have undoubtedly published this as a separate paper. He wrote an-
other paper on the quintic equation later [Cayley, 1894]. Around that time, near the end of his life, Cayley
started on a memoir – the draft Memoir in this collection – but did not get to complete it.

3. Mathematical background to the correspondence and the Memoir

While still a student, Cayley had become aware of Abel’s proof that it is impossible to solve the general
quintic equation by radicals. That is, there is no expression in the coefficients, consisting only of the four
basic arithmetic operations and the extraction of roots, which yields the solutions of every quintic complex
polynomial equation. However, Cayley, along with Harley, Cockle, and other mathematicians sought to
develop a theory of the solution of such quintic equations as are solvable by radicals. The research sought
to extend, for those quintics, the methods of analyzing and obtaining solutions that are applicable to lower
degree equations. Most of the mathematical content of both the correspondence and the draft Memoir deals
with calculations with precisely this aim.

3.1. Tschirnhausen transformations

During the early 1860’s, Cayley wrote several papers on Tschirnhausen transformations, and his late
Memoir begins with the reduction of a quintic equation to simpler form via a Tschirnhausen transformation.
With modern eyes, we can see the first glimmering of such an idea in the reduction of a quadratic equation
x 2 + bx = c to y 2 = d via the substitution y = x + 12 b, which was done geometrically by the ancient
Babylonians; and in the Renaissance reduction [Cardano, 1545/1993] (again in modern notation) of the
cubic equation

x 3 + bx 2 + cx + d = 0 (1)

to

y 3 + Hy + G = 0 (2)

by letting y = x + 13 b. For the cubic equation, however, it is possible further to choose the coefficients u, v
of the substitution y = x 2 + ux + v so that the transformed equation is of the form

y 3 + D = 0. (3)
 T. Crilly et al. / Historia Mathematica 44 (2017) 150–169 153

In this way, the solutions of the original cubic equation can be computed. To find the transformation, one
uses the formulas relating roots to coefficients that was used by François Viète as early as 1591 [Viète,
1591/2006]. Specifically, if we call the roots of (1) and (3) x1 , x2 , x3 and y1 , y2 , y3 , respectively, then

−b = x1 + x2 + x3 (4)
c = x1 x2 + x2 x3 + x3 x1
−d = x1 x2 x3

and

0 = y1 + y2 + y3 (5)
0 = y1 y2 + y2 y3 + y3 y1
−D = y1 y2 y3 .

Now, letting yi = xi2 + uxi + v for i = 1, 2, 3, substituting into the first two of the conditions (5) yields

3v = −b2 + 2c + bu (6)
 
0 = cu2 − 2buv + (3d − bc) u + 3v 2 + 2 b2 − 2c v + c2 − 2bd

which lead to a quadratic in either u or v. Thus the coefficients of the transformation y = x 2 + ux + v

can be determined, although only after eliminating some solutions for u (or v). The coefficient D of the
equation (3) may then be computed directly from the equation −D = y1 y2 y3 and the formulas (4).
In 1683, Ehrenfried Walther von Tschirnhausen extended this idea of transforming an equation, assert-
ing that by introducing a suitable polynomial y of degree less than 5 in x, he could eliminate all of the
intermediate terms of a general quintic, thereby reducing it to an equation of the form y 5 + D = 0, which is
evidently solvable by radicals [Tschirnhausen, 1683]. This claim was false. What is possible is to eliminate
all but one of the intermediate terms, thereby reducing it to an equation of the form

y 5 + y + F = 0. (7)

An explicit reduction to this form was found by the English mathematician George Jerrard [Jerrard, 1834],
who believed that by its aid all quintic equations could be solved by radicals. It was later realized that the
Swedish mathematician Erland Samuel Bring had earlier discovered the same transformation, although,
unlike Jerrard, he had never asserted that thereby all quintic equations were solvable [Bring, 1786]. Cayley
referred to this reduction of a quintic as the Bring–Jerrard form. Of course, the computations for the quintic
equation which are analogous to those shown above for the cubic are much more extensive.
For this and similar purposes, Cayley made use of a network of human calculators in his numerical work.
Three of these were William Barrett Davis (1806–1879), Charles Creedy (1815–1872), and William Curtis
Otter (1826–1878). Besides offering their services as calculators, both Creedy and Davis contributed pa-
pers to the literature in pure mathematics, while Otter was involved with actuarial mathematics. Creedy had
worked in Charles Babbage’s offices dedicated to the analytical and difference engines. Of the other two,
Davis was most often employed by Cayley. In his youth, Davis had operated in Calcutta as an East India
merchant, where he had been in the tailoring business, supplying uniforms to the British Army in colonial
India. His children were born in India, but he evidently returned to England and obtained a London Uni-
versity degree in 1855. In mathematics, he contributed to the calculation of prime numbers between eight
and nine million. Besides his work on Cayley’s paper [Cayley, 1861e] on Tschirnhausen transformations,
154 T. Crilly et al. / Historia Mathematica 44 (2017) 150–169

he helped Cayley in the completion of the extensive calculations involved with the final four covariants of
the list of 23 irreducible covariants of the quintic form in his Ninth Memoir of Quantics [Cayley, 1871]. He
also aided Cayley with the labyrinthine calculations in the theory of elliptic motion and on the subject of
modular equations. Davis was also connected with the astronomer John Couch Adams, who had launched
himself into the calculation of Bernoulli numbers in the 1870s. Cayley was not alone in employing human
calculators. Harley employed Samuel Bills (1807–1876) of Hawton, near Newark-upon-Trent in Notting-
hamshire. Bills was a warehouseman, but was an able mathematician who contributed to the solutions of
Thomas Kirkman’s famous “schoolgirls problem”.

3.2. Symmetric polynomials

The computations above involving Viète’s formulas give one example of the occurrence of symmetric
polynomials in Cayley and Harley’s calculations. This was background knowledge for Cayley and his
contemporaries. Indeed, a few years before his correspondence with Harley, Cayley had written a paper
on the subject [Cayley, 1857]. We summarize in modern terminology some basic facts about symmetric
polynomials. Consider an arbitrary monic polynomial

f (x) = x n + an−1 x n−1 + an−2 x n−2 + . . . + a2 x 2 + a1 x + a0

with roots x1 , . . . , xn . A polynomial ϕ(x1 , . . . , xn ) is said to be a symmetric polynomial in the roots

x1 , . . . , xn if it is fixed under every permutation of the roots, i.e., if for every permutation σ of {x1 , . . . , xn },

ϕ(σ (x1 ), . . . , σ (xn )) = ϕ(x1 , . . . , xn ).

(In modern parlance, it is often said that ϕ(x1 , . . . , xn ) is invariant under permutations, but we will avoid
that language so as not to conflict with the language of Cayley and Harley.) Viète’s formulas show that the
coefficients of f (x) are symmetric polynomials in the roots x1 , . . . , xn , with (−1)i an−i being equal to the
ith elementary symmetric polynomial si (x1 , . . . , xn ), where si (x1 , . . . , xn ) is the sum of the products of the
roots taken i at a time, i.e.,

s0 (x1 , . . . , xn ) = 1
s1 (x1 , . . . , xn ) = x1 + . . . + xn
s2 (x1 , . . . , xn ) = x1 x2 + . . . + xn−1 xn
...
sn (x1 , . . . , xn ) = x1 · · · xn .

The Fundamental Theorem on Symmetric Polynomials states that any symmetric polynomial ϕ(x1 , . . . , xn )
can itself be expressed as a polynomial in the elementary symmetric functions {si (x1 , . . . , xn )}, and hence
in terms of the coefficients of f (x). The computation in section 3.1 uses the fact that the coefficients of
the equation in y are symmetric polynomials in x1 , x2 , x3 (since they are symmetric polynomials in the
roots y1 , y2 , y3 ). To perform the computation leading to (6), one rewrites these coefficients in terms of the
coefficients of the original equation in x.
In his analysis of the successful solution of equations of degrees 2, 3, and 4, Joseph-Louis Lagrange
[Lagrange, 1771/1867] had shown that these solutions depended upon the use of polynomials in the roots
x1 , . . . , xn that are not symmetric but rather remain fixed for only some among all the possible permu-
tations – in modern language, fixed under a subgroup of the full group of permutations of {x1 , . . . , xn }.
These polynomials are resolvents of the original equation. Since the letters between Cayley and Harley
 T. Crilly et al. / Historia Mathematica 44 (2017) 150–169 155

largely concern the calculations of resolvents of the quintic equation, as does Cayley’s Memoir, we discuss
resolvents further in section 3.4.
One symmetric polynomial which is of particular interest is the discriminant of f (x),

(f (x)) = (xi − xj )2 . (8)
1≤i<j ≤n

For example, let f (x) = ax 2 + bx + c be a quadratic

√ with roots r1 and r2 , so that f (x) = a(x − r1 )(x − r2 ).
−b ± b − 4ac
2
From the quadratic formula r1 , r2 = and Viète’s formulas we can calculate directly that
2a
 2  
b c 1 
(f (x)) = −4 = 2 b2 − 4ac .
a a a

With more effort we can obtain the discriminant of a cubic f (x) = ax 3 + bx 2 + cx + d. It is

 
1  
(f (x)) = −4b 3
d + b 2 2
c + 18abcd − 4ac 3
− 27a 2 2
d .
a4

It is important to note that we can find the discriminant of f (x) without first having to find the roots of
f (x). We can see directly from the definition that (f (x)) = 0 if and only if f (x) has distinct roots.

3.3. Transcendental solutions

Cayley’s work on Tschirnhausen transformations [Cayley, 1861b, 1861c, and 1861e], done during the
period of his correspondence with Harley, may be viewed in the context of transcendental solutions of the
quintic equation, since the first step in computing such solutions is to bring the equation into Bring–Jerrard
form. From that form x 5 + x + F = 0, it follows that the roots of the equation are values of a function
of F . This function is transcendental, analogous to the trigonometric solution of the cubic equation. The
trigonometric solution is obtained by starting with the reduced cubic (2)

y 3 + Hy + G = 0

and rewriting it in the form

4z3 − 3z − a = 0

by an invertible algebraic change of variable. If this equation in z can solved, then inverting the change of
variable yields a solution to the equation in y. But substituting a = cos(3θ ) into the trigonometric identity

cos(3θ ) = 4 cos3 (θ ) − 3 cos(θ )

reveals the roots of 4z3 − 3z − a = 0 to be

   
2π 4π
z = cos(θ ), z = cos θ + , z = cos θ + .
3 3
156 T. Crilly et al. / Historia Mathematica 44 (2017) 150–169

Thus, we find the roots of 4z3 − 3z − a = 0 by first finding 3θ = cos−1 (a); i.e., we use a transcendental
function, the inverse cosine. Charles Hermite showed that the quintic is solvable by analogous transcen-
dental methods, using elliptic functions [Hermite, 1858] (see also [Goldstein, 2011]) and independently,
Leopold Kronecker developed his own form of transcendental solution [Kronecker, 1858]. Cayley gave
Harley his opinion of the desirable form of a transcendental solution in his letter of 29 April 1862, men-
tioned below, and mentioned the transcendental solutions again in his Memoir, but he did not discuss them
in detail.

3.4. Resolvents

Despite some references to transcendental solutions of the quintic equation, Cayley’s main interest was
algebraic and especially concerned with their resolvents. The term “resolvent” had been introduced by Euler
around 1732. As Cayley used it, a “resolvent” (or “auxiliary equation”) of an equation x n + an−1 x n−1 +
an−2 x n−2 + . . . + a2 x 2 + a1 x + a0 = 0 is a polynomial p (θ) whose roots help one to solve the original
equation. In this sense, reduced forms of the equation, such as the Bring–Jerrard form, are resolvents, and
the equation of differences, discussed below, is also a resolvent. Cayley’s “new auxiliary equation in the
theory of equations of the fifth order” [Cayley, 1861d], however, is a resolvent in the eighteenth-century
tradition of Euler, Lagrange, Alexandre-Théophile Vandermonde, and Gian Francesco Malfatti. Cayley’s
computation of this resolvent is his most important work related to the quintic equation.
In modern notation, the construction of such a resolvent is as follows. Let the solutions of x n +
an−1 x n−1 + an−2 x n−2 + . . . + a2 x 2 + a1 x + a0 = 0 be x1 , x2 , . . . , xn . The group Sn of all permutations of n
letters acts on the set of polynomials C [x1 , x2 , . . . , xn ] as follows. If f (x1 , x2 , . . . , xn ) ∈ C [x1 , x2 , . . . , xn ]
and σ ∈ Sn , then define σf (x1 , x2 , . . . , xn ) ≡ f (xσ (1) , xσ (2) , . . . , xσ (n) ). Suppose that the distinct images

of f under Sn are f1 , f2 , . . . , fk . Then the polynomial p(θ ) = ki=1 (θ − fi ) is symmetric in x1 , x2 , . . . , xn
and so its coefficients are polynomials in an−1 , . . . , a0 . The equation p(θ ) = 0 is a resolvent equation of
the original equation. Sometimes, an expression used to create p(θ ) is also referred to as a resolvent.
For example, if the solutions of a cubic equation are x1 , x2 , x3 , then under the action of S3 on x1 , x2 , x3
the expression (x1 + ωx2 + ω2 x3 )3 takes on two values, (x1 + ωx2 + ω2 x3 )3 and (x1 + ω2 x2 + ωx3 )3 , where
ω is a primitive cube root of unity. The resolvent equation is then the quadratic equation in θ ,

  3   3 
θ − x1 + ωx2 + ω2 x3 θ − x1 + ω2 x2 + ωx3 = 0. (9)

To write this quadratic in terms of the coefficients of the original cubic equation is somewhat tedious, but
it yields θ 2 + 27Gθ − 27H 3 = 0 for the reduced cubic (2) and θ 2 + 27( 27 b − 13 cb + d)θ − 27(c − 13 b2 )3
2 3

for the general cubic (1).

The cubic resolvent x1 + ωx2 + ω2 x3 was introduced by Lagrange, who also gave resolvents for quartic
and quintic equations [Lagrange, 1771/1867], but whereas the resolvents for the cubic and quartic were in
each case a polynomial of one degree less than the degree of the equation, the resolvent for the quintic was
a polynomial of one degree more, a sextic. As is well known, this asymmetry and Lagrange’s analysis of it
inspired the work of Niels Henrik Abel and Évariste Galois. Despite Abel’s impossibility proof, however,
in the nineteenth century several nonprofessional mathematicians thought that the quintic equation could be
solved by an algebraic formula, and this view took hold in Britain. In addition to Jerrard and Harley, Charles
Hargreave and Thomas P. Kirkman, at various stages of their lives, believed that a solution in radicals was
possible. This was also true of James Cockle.
James Cockle, who graduated from Cambridge a year earlier than Cayley, became a London legal bar-
rister, and like Cayley, kept up his passion for mathematics. Just before Harley started his correspondence
with Cayley on the theory of equations, Cockle had written to him on 23 April 1859 about Cayley:
 T. Crilly et al. / Historia Mathematica 44 (2017) 150–169 157

[He] is a man of eminent (or rather super-eminent) abilities. Literary courtesies have passed between us, but
I have not the honour and pleasure of his personal friendship. He was at Cambridge at the same time that I
was – though, I think, rather junior to me – and at the same College many years ago. Of his high qualities
not a moment’s doubt can be entertained. [Bennett, 2003, p. 6]

Cockle initially did not accept the conclusion of Abel’s impossibility theorem, and in his own investigation
of solutions of the quintic equation, discovered a sextic “auxiliary equation” (resolvent) [Cockle, 1860].
Cockle’s sextic resolvent, expressed in terms of the roots x1 , x2 , x3 , x4, and x5 , is

x1 x2 + x2 x3 + x3 x4 + x4 x5 + x5 x1 − x1 x3 − x3 x5 − x5 x2 − x2 x4 − x4 x1 , (10)

which is simpler to work with than the one that had been given earlier by Lagrange, namely

x1 + ωx2 + ω2 x3 + ω3 x4 + ω4 x5 ,

where ω is a primitive fifth root of unity, so Cockle’s work attracted attention in England, first by Harley
[Harley, 1860a], then by Cayley [Cayley, 1861a]. The spur provided by Cockle proved to be a known path,
however: Carl Gustav Jacob Jacobi had discovered the same sextic resolvent [Jacobi, 1835], as Cayley
acknowledged in a note published in November 1862 [Cayley, 1862a]. Yet there was more to investigate.
It was naturally desirable to express the coefficients of the resolvent equation in terms of the coefficients
of the original equation. Harley had computed the resolvent equation for the Bring–Jerrard form. Cayley
wanted the work done for the full form of the quintic, in order to see the results in a global setting and thus
to link it to the developing theory of invariants [Cayley, 1861d], discussed below in section 3.7.

3.5. Symmetric products

Cockle had developed his sextic resolvent by the use of what he called symmetric products [Cockle,
1852, 1853, 1854, 1860]. His work attracted Harley and then Cayley. In their correspondence, Cayley
and Harley discussed symmetric products as part of their analysis of the quintic equation, and [Cayley,
1861d] reflects this interest. In modern notation, these symmetric products are constructed as follows. From
the roots x1 , x2 , . . . , xn of an equation x n + an−1 x n−1 + an−2 x n−2 + . . . + a2 x 2 + a1 x + a0 = 0, Cockle
considered linear combinations (with as yet unspecified coefficients vr2 , . . . , vrn )

Vr = x1 + vr2 x2 + · · · + vrn xn

for 1 ≤ r ≤ n and from these created products

πm = V1 V2 · · · Vm

for m ≥ 1. Adding the requirement that a product πm be symmetric in the roots x1 , x2 , . . . , xn imposes con-
ditions on the coefficients vri . Cockle showed how to use these symmetric products to derive the solutions of
equations of degree 3 and 4 [Cockle, 1852]. For example, if the roots of a cubic equation are x1 , x2 , x3 , then
V1 = x1 + v12 x2 + v13 x3 , V2 = x1 + v22 x2 + v23 x3 , and π2 = (x1 + v12 x2 + v13 x3 ) (x1 + v22 x2 + v23 x3 ). If
one demands that v12 = v23 = k1 , that v13 = v22 = k2 , and that π2 be symmetric but that V1 and V2 not be
symmetric (else they equal −b), then k1 and k2 are each nonreal cube roots of unity. Thus the expressions
V1 and V2 are x1 + ωx2 + ω2 x3 and x1 + ω2 x2 + ωx3 , the previously seen cubic resolvent. Cockle noted
that
158 T. Crilly et al. / Historia Mathematica 44 (2017) 150–169

[F]or cubics, we employ two [values] only in place of the six which, when approached from first principles,
Lagrange’s process involves. . . . In the theory of quintics we do not take one hundred and twenty values as
our point of departure, but, starting from four, we are conducted to twenty-four as involved in the discussion.
[Cockle, 1854, p. 133; italics in the original]

He originally hoped that by judicious choice of symmetric product, he could solve all quintic equations.
His paper [Cockle, 1852, p. 492] begins with the statement

The conclusions of Abel and Sir W. R. Hamilton respecting the impossibility of solving equations of the
fifth degree are rendered doubtful by recent investigations of Mr. G. B. Jerrard. New fields of research thus
seem to open upon us. My present object is to point out the general scope of the method of symmetric
products, and to offer remarks which may assist in the inquiry as to how far that method is calculated to
throw light upon the theory of equations of higher degree.

Nevertheless, Cockle did not assert absolutely that the quintic is solvable in general, and later wrote that
“[i]n the present state of the subject of equations of the fifth degree, I do not desire to have any remark
of mine placed in a higher category than that of conjecture” [Cockle, 1854, p. 135]. By 1860, Cockle had
conceded: “[A] finite algebraic solution ... indeed appears to be absolutely unattainable” [Cockle, 1861,
p. 5].

3.6. The equation of differences

Some of Cayley’s correspondence with Harley dealt with yet another way of investigating the properties
of an equation, what he called the equation of differences. Papers from this period [Cayley, 1860a, 1860b,
1861a] reveal his interest. He defined the equation of differences as that monic polynomial whose roots
are the squares of the differences of every two of the roots of a given equation. For a polynomial f (x) of
degree n with roots x1 , . . . , xn , the equation of differences is the polynomial pf (θ) in θ defined as

pf (θ ) = (θ − (xi − xj )2 ). (11)
1≤i<j ≤n

Thus the discriminant (8) is, up to sign, just the constant term of the equation of differences. For example,
the equation of differences of a cubic equation (1)

x 3 + bx 2 + cx + d = 0

with roots x1 , x2 , x3 is
   
θ − (x1 − x2 )2 θ − (x2 − x3 )2 θ − (x3 − x1 )2 = 0,

which gives information about the roots of the original equation. For example, the cubic equation in θ has
a negative root if and only if the original real cubic (1), and hence the reduced cubic (2), has a pair of
imaginary roots. Since the equation of differences is a symmetric polynomial in x1 , x2 , x3 , its coefficients
are polynomials in the coefficients b, c, d. In terms of the coefficients of (2), the equation of differences is

θ 3 + 6H θ 2 + 9H 2 θ + (27G2 + 4H 3 ) = 0.

The computation of these coefficients is tedious; Cayley discussed the computation in letters to Harley. In
[Cayley, 1860b], Cayley calculated these coefficients for the cubic, the quartic, and the quintic equations
by noting that they are semi-invariants, that is, by using invariant theory.
 T. Crilly et al. / Historia Mathematica 44 (2017) 150–169 159

3.7. Invariant theory

Many of the computations discussed in the letters between Cayley and Harley, published in Cayley’s pa-
pers, and presented in Cayley’s Memoir employ methods of invariant theory, a subject which Cayley with
J.J. Sylvester virtually invented in the 1840s and for which, perhaps more than any other single mathemat-
ical contribution, Cayley is remembered. (See [Crilly, 1986 and 1988] and [Parshall, 1989 and 1998].) The
calculations of coefficients of the symmetric product, the equation of differences, and the resolvent were
simplified by noting that they were semi-invariants. To define semi-invariant, begin with a homogeneous
form of degree n in two variables

(x, y) = an x n + an−1 x n−1 y + an−2 x n−2 y 2 + . . . + a2 x 2 y n−2 + a1 xy n−1 + a0 y n ,

where x and y are variables and an , . . . , a0 are coefficients. An invertible linear change of variables T :
(x, y) → (x
, y
) gives a change of coefficients (an , . . . , a0 ) → (an
, . . . , a0
). A function f (an , . . . , a0 ) is
said to be an invariant if for some integer w,

f (an
, . . . , a0
) = (det(T ))w f (an , . . . , a0 ),

where det(T ) is the determinant of the linear transformation T ; if (x

, y
) = T (x, y) = (px + qy, rx + sy),
then det(T ) = ps − qr. More generally, a function f (an , . . . , a0 , x, y) is said to be a covariant if for some
integer w,

f (an
, . . . , a0
, x
, y
) = (det(T ))w f (an , . . . , a0 , x, y).

The simplest and most classical invariant is that of a quadratic form. The quadratic form (x, y) =
2 2  2
ax 2+ bxy + cy 2 has an invariant ac − b4 = − a4 a12 (b2 − 4ac) , which is, apart from the factor − a4 , the
discriminant of the polynomial f (x) = (x, 1) as we have defined it above.
Less simple but also classical is the Hessian, a covariant defined by Otto Hesse [Hesse, 1844] and later
named by Sylvester. Given the general form in two variables as above, its Hessian is defined to be
 2 2
∂ 2 ∂ 2 ∂ 
2 2
− .
∂x ∂y ∂x∂y

For example, the Hessian of the cubic form (x, y) = ax 3 + bx 2 y + cxy 2 + dy 3 is 4[(3ac − b2 )x 2 +
(9ad − bc) xy + (3bd − c2 )y 2 ].
The example of the cubic equation illustrates an occurrence of invariant theory in the solution of equa-
tions. If x1 , x2 , x3 are the roots of the original equation (1),

x 3 + bx 2 + cx + d = 0,

then the roots of the reduced equation y 3 + Hy + G = 0 are y1 = x1 + 13 b, y2 = x2 + 13 b, y3 = x3 + 13 b,

respectively. It follows that any symmetric function of the roots x1 , x2 , x3 which is a function only of their
differences can be expressed as a symmetric function of the y1 , y2 , y3 and thus in terms of the coefficients
H and G, which are themselves symmetric functions of the differences of the roots x1 , x2 , x3 . For example,
3H = (x1 − x2 )(x2 − x3 ) + (x2 − x3 )(x3 − x1 ) + (x3 − x1 )(x1 − x2 ). Now associating to the polynomial
x 3 + bx 2 + cx + d the homogeneous polynomial in two indeterminates
(x, y) = x 3 + bx 2 y + cxy 2 + dy 3 ,
H and G can be described and computed by invariant theory. For example, the coefficient H of the reduced
1
cubic is just the numerical multiple 12 of the coefficient of the leading term of the Hessian of
(x, y).
160 T. Crilly et al. / Historia Mathematica 44 (2017) 150–169

Cayley introduced the name semi-invariant (sometimes written seminvariant) for the coefficient of the
leading term of a covariant. In his letter to Harley of 22 March 1860, Cayley gave three characterizations
of semi-invariants:

• the leading coefficient of a covariant;

• a polynomial f (an , . . . , a0 ) annihilated (Sylvester’s word) by one of the differential operators which
defines an invariant; and
• a polynomial f (an , . . . , a0 ) which is invariant under linear transformations of the form Tq (x, y) =
(x + qy, y).

The last of these criteria was especially relevant to the calculations that Cayley and Harley discussed.
Since Tq (x, y) = (x + qy, y), when applied to the polynomial an x n + an−1 x n−1 + an−2 x n−2 + . . . + a1 x +
a0 =
(x, 1), sends each root xi to xi + q, any polynomial which is symmetric in the roots and is a function
only of their differences must be a semi-invariant. Thus, for example, the coefficients of the quadratic
resolvent equation (9) of a cubic equation can also be expressed in terms of the differences of the roots. For
convenience, set r = x1 − x2 , s = x2 − x3 , t = x3 − x1 . Then the coefficient of the linear term is

− [(x1 + ωx2 + ω2 x3 )3 + (x1 + ω2 x2 + ωx3 )3 ]

= (r − s)(s − t)(t − r),

and the constant term is

(x1 + ωx2 + ω2 x3 )3 (x1 + ω2 x2 + ωx3 )3

= [13(r 2 + s 2 + t 2 )(r 2 s 2 + s 2 t 2 + r 2 t 2 ) − 3(r 2 + s 2 + t 2 )3 ]/2.

Since these terms are polynomials which are symmetric in the roots and are functions only of their dif-
ferences, they are semi-invariants and can be calculated by the techniques of invariant theory. Cayley and
Harley’s calculations of the coefficients of the equation of differences and of the resolvent of a quintic equa-
tion was greatly facilitated by the recognition that they are semi-invariants. Nevertheless, the calculations
were onerous. Cayley, as well as Harley, launched himself into these, and, on occasions (as noted above),
would also draw on the services of human computers.

4. The correspondence

With these mathematical preliminaries, we are now in a position to examine the contents of the extant
correspondence between Cayley and Harley, a dialog which begins in medias res. In the first letter, dated
24 September 1859, Cayley wrote

I ought to have written sooner to thank you for the interesting paper on symmetric products &c. which I
found here on my return from Scotland. I received from Mr. Cockle your transformed expression (with R
instead of S) for the symmetric product, and have been transforming it into my own notation where the
quintic (if the second term vanishes) is ax 5 + 5.0bx 4 + 10cx 3 + 10dx 2 + 5ex + f = 0. This is of course
easy enough but I should like to see what the expression would be with the general value b, instead of the
value zero, and I am afraid that would be a very tedious calculation.

Here, Cayley was using the binary form of a polynomial, which he often favored in his papers. In it, the
appropriate binomial coefficient is factored out of each coefficient. For example, the polynomial
 T. Crilly et al. / Historia Mathematica 44 (2017) 150–169 161

f (x) = 2x 3 − 21x 2 + 3x + 1

is of the form

ax 3 + bx 2 + cx + d

with a = 2, b = −21, c = 3, and d = 1. Cayley called this the denumerate form of the polynomial. In the
binary form,

ax 3 + 3bx 2 y + 3cxy 2 + dy 3 ,

each literal coefficient is multiplied by the appropriate binomial coefficient. Thus, in the same numerical
example, a = 2, b = −7, c = 1, and d = 1. Including the binomial coefficients as factors gives greater
symmetry of expression, which is useful in invariant theory.
As the letter of 24 September 1859 makes clear, Harley had already been doing work on the quintic
equation that was of interest to Cayley. This letter, however, marks a new phase, in which they were begin-
ning to collaborate more closely and directly, with Cayley the master and Harley the assistant. In [Harley,
1860a], to which Cayley referred, Harley was following up on work of Cockle [Cockle, 1860] that Harley
had communicated to the Manchester Literary and Philosophical Society. Cockle had computed a resolvent
for the quintic and of course obtained a sextic. Harley showed that if his resolvent vanishes, then the orig-
inal quintic is solvable (by radicals). At this point Harley had apparently not given up on the possibility of
solving the general quintic by radicals. He wrote,

I may however remark that even if the method fail to achieve the solution of the general problem, it will
probably help to settle a controversy in which mathematicians of the greatest eminence have taken opposite
sides, and to throw light on the question, respecting which so much has been written, of the possibility
or impossibility of expressing a root of the general equation of the fifth degree by a finite combination of
radicals and rational functions. [Harley, 1860a, p. 216]

The quotation comes from the first of a pair of papers by Harley on this topic, and in letters of 18 November
and 12 December 1859 Cayley thanked Harley for advance copies of some of them. Although they were
intended to be published together, or one right after the other, Part II only appeared after some delay. In his
letter of 19 February 1861, Cayley wrote, referring to Part I [Harley, 1860b],

It appears to me that the conclusion on p. 346 of your paper on quintics [Harley, 1860b] is, that β1 , etc.
can be expressed as rational functions of t and the coefficients is erroneous: for t is a 12-valued function
whereas β1 is a 120-valued function and even β15 a 24-valued function. [underlining in the original]

Indeed, Harley, owing to Cayley’s critique, began part II of his paper [Harley, 1862, p. 248]

From the last three equations in Art. 5, I inferred that β can be expressed as a rational function of t , and
the coefficients of the trinomial equation. The inference is erroneous; for t is a 12-valued function, whereas
β is a 120-valued function and even β 5 a 24-valued function. It hence appears that β cannot be expressed
as a rational function of t , and that, consequently, the sextic in t is not, as I was tempted to suppose, an
Abelian. Misgivings on this point occurred to me in writing my paper, as will be seen on referring to the last
paragraph of Art. 8.
Mr. Cockle had previously committed a similar oversight . . . Subsequently, however, . . . he corrected the
oversight, and abandoned the idea of the algebraic solvability of the quintic.
162 T. Crilly et al. / Historia Mathematica 44 (2017) 150–169

By characterizing the sextic resolvent equation as Abelian, Harley meant that if r1 , r2, r3 , r4 , r5 , r6 are the
roots of the sextic, then there are rational functions θi , 2 ≤ i ≤ 6, such that ri = θi (r1 ) with the further
condition that θi θj (r1 ) = θj θi (r1 ) for 2 ≤ i, j ≤ 6. Having realized that the sextic resolvent equation was
not Abelian, Harley too abandoned his belief in the solvability of the quintic equation by radicals.
The letters and papers reveal growing awareness of the connection between invariant theory and the
quintic equation. Cockle himself had seen a connection between the theory of equations and invariant
theory as early 1854, for in a footnote to a paper, he wrote:

Mr. Cayley (Camb. and Dub. Math. Journ for May 1851) [Cayley, 1851] has given a coefficient to x 3 , gen-
eralizing my process.... The intrinsic interest of Mr. Cayley’s result is enhanced by its connexion with
Dr. Boole’s functions θ and θ
, and with his own researches on hyperdeterminants [invariant theory].
[Cockle, 1854, p. 132]

Although Cayley had given invariant theoretic solutions to lower degree equations in [Cayley, 1858], the
relevance of invariant theory for the quintic appears in a letter of 28 September 1859, in which he wrote to
Harley that

Mr Cockle suggested to me some time ago that the last coefficient of the resolvent equation of the quintic
might be the leading term of a covariant [i.e. a semi-invariant] but I did not until just now perceive that this
was the case as well with respect to the last as to every other coefficient of the resolvent equation.

Two weeks earlier, in his letter of 15 September 1860 Cayley had acknowledged receipt of Harley’s
manuscript intended for publication in the Philosophical Transactions of the Royal Society [Harley, 1861].
Cayley’s support of Harley’s mathematical career was extremely beneficial. He congratulated Harley on
completing the work and undertook to present his paper to the Royal Society. It was formally received on
18 November and read to a meeting on 13 December 1860.
This paper contains staggeringly complex computations of the sort that was typical for papers on invari-
ant theory by Cayley, J.J. Sylvester, and George Salmon. It is no wonder that Royal Society referees James
Booth and Philip Kelland hesitated. They were mathematicians, but the technical detail in this paper was
not in their current line of country, and, perhaps unaware of Cayley’s involvement with the project, they fell
back on his expertise for a definitive recommendation. Kelland struggled to gain an insight into Harley’s
“laborious calculations”, writing, “I would advise the Council of the Royal Society to give more weight
to Mr. Cayley’s opinion on such a matter without examination than to mine with it” [23 Mar. 1861, Roy.
Soc. Rr.4.122]. Booth admitted that it was unfamiliar territory, but after studying the material over a period
suggested it should be published, venturing, “but I should much wish to have my view confirmed by the
opinion of Mr Cayley or Mr Sylvester or Mr Spottiswoode, gentlemen who have made this branch of pure
mathematics their Special Study, and who may to a great extent by considered its Inventors” [11 Mar. 1861,
Roy. Soc. RR.4.121]. The printed abstract [Harley, 1860c, p. 44] indicated the intricacies ahead:

Employing the property of seminvariancy pointed out by Mr. Cayley, the author succeeds in effecting the
calculation of the symmetric product for the complete quintic. This product is composed of three hundred
and twenty-five functions of twenty-four dimensions. [italics in the original]

In the paper itself Harley quoted Cayley’s letter of 28 September almost verbatim, a practice not uncommon
at the time:

In a letter under date September 28, 1859, Mr. Cayley called my attention to the circumstance that the several
coefficients of the resolvent equation of the quintic
 are leading coefficients of a covariant. Mr. Cockle had
previously suggested that the symmetric product , or the last coefficient, was such a term.... [In fact] each
 T. Crilly et al. / Historia Mathematica 44 (2017) 150–169 163

coefficient of the sextic in θ is the leading coefficient of a covariant. At present, however, we have only to
deal with the last coefficient, that is, the symmetric product. [Harley, 1861, p. 348]

He continued that “ is a seminvariant . . .” and at this point referred in a footnote to Cayley’s letter
of 22 March 1860 in which the term “seminvariant” (or “semi-invariant”) is defined. Harley specifically
mentioned the other two characterizations listed above.
Here, then, was the link between invariant theory and the quintic equation: every coefficient of the
resolvent equation was a semi-invariant. Cockle had suggested that the last coefficient (the symmetric
product) was a semi-invariant, justifying this in [Cockle, 1852] by the third of the characterizations of a
semi-invariant. Cayley, because of his familiarity with invariant theory, saw that every other coefficient of
the resolvent equation is also a semi-invariant. (It is easy to check that the resolvent (10) is invariant under
the translation x → x + b and so the coefficients of the resolvent equation (being symmetric in the roots)
are semi-invariants.) This observation seems to have galvanized Cayley. It may be significant that whereas
Cayley, 1858 paper, “A Fifth Memoir Upon Quantics” [Cayley, 1858], discussed the solutions only of the
quadratic, cubic, and quartic equations in invariant theoretic terms, in the following years Cayley turned
much of his attention to the quintic equation.
By 19 January 1860, Cayley had enlisted Harley’s help in a project he was working on. For a polynomial
f (x) of degree n with roots {r1 , . . . , rn }, Cayley wished to calculate the equation of differences (11)

pf (θ ) = (θ − (ri − rj )2 )
1≤i<j ≤n

in the case where f (x) is a quintic. Cayley’s method was a quite clever inductive procedure, to which he
referred in a letter of 31 December 1859, first mentioning it in connection with the problem of computing
the resolvent. Cayley introduced the polynomial

qf (θ ) = (θ − ri2 )
1≤i<j ≤n

Now suppose that rn is a root of f (x) and make the substitution x = y + rn to obtain a new polynomial
g(y). Then g(y) has 0 as a root, and so has 0 as its constant term, i.e., g(y) = yh(y) where h(y) is a
polynomial of degree n − 1. Call the roots of g(y) r1
, . . . , rn−1

with rn
= 0, and consider the corresponding
polynomial pg (θ ). Note that ri − rj = ri − rj if 1 ≤ i < j ≤ n − 1, while ri
− rn
= ri
. Thus

     
  
pg (θ ) = θ − (ri
− rj
)2
2
θ − (ri ) .
1≤i<j ≤n−1 1≤i≤n−1

The first factor is ph (θ ), while the second factor is qh (θ ). But by induction on the degree, Cayley could
compute both of these polynomials. This was the easy part of the work. The answer so obtained depends
on the new coefficients, and these not only involve the coefficients of f (x) but also the root rn . Thus
the expression so obtained needs to be “resymmetrized”, and this is where the hard work comes in. The
answer is to be found in Cayley’s paper [Cayley, 1860b]. The equations of differences for a quadratic,
cubic, quartic, and quintic have 3, 11, 61, and 336 terms respectively, with the constant term, which is,
up to sign, the discriminant (as noted in section 3.6), having 2, 5, 16, and 59 terms respectively. (Cayley
cleared denominators. For example, in the case of the quadratic, the equation of differences is simply
θ − (f (x)) = 0, and Cayley simply wrote a 2 θ − (b2 − 4ac) = 0.) These are massive computations for
which Cayley solicited Harley’s help. He described what he was doing in the letters of 31 December 1859
and 19 January 1860 and asked for Harley’s help in finishing the work. He discussed this further in his
164 T. Crilly et al. / Historia Mathematica 44 (2017) 150–169

letter of 25 January 1860, and in a postscript to this letter wrote “I have just received your letter & will
write again when I have examined it.” The very next day, in his letter of 26 January, Cayley corrected
an error that Harley had made. Several other letters deal with these computations, and Cayley explicitly
acknowledged Harley’s contribution in [Cayley, 1860b, p. 98].
Cayley had another project under way at this time. He had his own approach to the quintic through
calculating a resolvent that involved massive computations as well, and he enlisted Harley’s help with
these, too. On 5 February 1860 Cayley presaged that

There will be something very interesting in the forms of the combinations ϕ1 ϕ4 + ϕ2 ϕ5 + ϕ3 ϕ6 which give
the five roots of the quintic. There are in all six [sic] groupings such as

12. 34. 56
13. 25. 46
14. 26. 35
15. 24. 36
16. 23. 45

which contain all the duads each once and once only.

The background to this reference to groupings of pairs or “duads” is a paper by J.J. Sylvester [Sylvester,
1844] whose concerns were principally combinatorial. In a letter of 16 August 1860 to Sylvester [Parshall,
1998, pp. 97–101], Cayley had observed that the permutations of the duads constructed out of six indices
form a group that is (in our words) isomorphic to the group of permutations of five things. (Sylvester
promptly wrote back to assure Cayley that he was aware of this fact, and on 18 August Cayley acknowl-
edged Sylvester’s priority.) Now in a footnote to his paper, Sylvester had remarked on a connection between
his arrangements of duads and the solution of equations [Sylvester, 1844, p. 92]. There are three groupings
of pairs of four objects, giving a reduction of the general quartic equation to a cubic. The three independent
partitions of four items, x1 , x2 , x3 , x4 are

x1 x2 x3 x4
x1 x3 x2 x4
x1 x4 x2 x3

Thus, if x1 , x2 , x3 , x4 are solutions of a quartic equation, then under all the permutations of S4 , the quantity
x1 x2 + x3 x4 takes on just three values, namely x1 x2 + x3 x4 , x1 x3 + x2 x4 , and x1 x4 + x2 x3 . These values
are the solutions of a cubic equation – a resolvent cubic for the original quartic. In the letter to Harley of
5 February 1860 just quoted, Cayley was suggesting the reversal of Sylvester’s procedure: one identifies
each of the five (not six!) independent groupings of six quantities with one of the five roots of a quintic.
Two weeks later, however, on 16 February, Cayley decided that

It is consequently nearly as interesting, instead of the sextic for θ [i.e., instead of considering the original
resolvent] to investigate that for


a 4 ϑ = a 2 bd − a 2 c2 + 5a 4 x12 (x2 x5 + x3 x4 ).
 T. Crilly et al. / Historia Mathematica 44 (2017) 150–169 165

Here x1 , . . . , x5 are the roots of the quintic, and Cayley was adopting Harley’s notation, with
being the
sum over the cyclic permutations of the roots, i.e., the sum of the five terms obtained by the powers of the
permutation (x1 , x2 , x3 , x4 , x5 ) −→ (x2 , x3 , x4 , x5 , x1 ). Cayley then proposed the same trick he had used
previously, to translate the polynomial so as to eliminate the constant term f , whence this last summation
becomes a variable he denoted by λ. He then remarked on properties of λ, and noted that

it is practicable [Cayley originally wrote “easy” and then crossed it out] to find the sextic in λ, and thence
the sextic in ϑ ; the coefficients of the last mentioned sextic having ultimately to be completed by the intro-
duction of the terms in f .
The sextic in λ will be worth having, and I would suggest our each calculating it, for comparison, and
when it is done it will be possible to judge of the amount of the ulterior work.

Cayley and Harley continued to work on this project and discuss it in their letters, culminating in [Cayley,
1861d]. In this paper Cayley constructed what he thought was a new resolvent, later realizing that he
had been anticipated by Jacobi. The paper begins by discussing work of Cockle and Harley, and refers to
[Cayley, 1860b] and two papers by Harley, [Harley, 1860a] and [Harley, 1860b]. It points out the value of
invariant theory, stating that “The direct calculation of the auxiliary [resolvent] equation by the method of
symmetric functions would, I imagine, be very laborious. But the coefficients are seminvariants, and the
process explained in my memoir on the Equation of Differences [Cayley, 1860b] was therefore applica-
ble, and by means of it, the equation, it will be seen, is readily obtained” [Cayley, 1861d, pp. 264–265].
How readily it is obtained is seen in the next fourteen pages. “No mathematician of Cayley’s time pos-
sessed comparable powers over long expressions and tiresome calculations,” wrote Percy A. MacMahon
[MacMahon, 1896, p. 7]. The fact that others did some calculations for him does not materially detract
from this assessment.
During the period of this collaboration Cayley lived in London and Harley in Yorkshire, and it is unclear
whether they met personally at this time. On 26 April 1860 Cayley wrote, “I look forward with great
pleasure to seeing you on your visit here. I am almost always in [my rooms in Lincoln’s Inn] during the
mornings.” That meeting evidently did not occur, for Cayley wrote in the following week, on 4 May 1860,
“I rather expected you this week.. . . I will certainly be in my rooms the time you mention”. But there is no
evidence in later letters that this meeting took place, either. Cayley and Harley certainly met later in life,
however, as both were active in the affairs of the British Association for the Advancement of Science and
the London Mathematical Society – the latter formed in 1865, after this work was done.
Although the quintic equation was a lifelong interest of Cayley’s, during the 1870s he published nothing
about it. The reason for this is probably to be found in two of Cayley’s letters. On 19 February 1861, he
wrote of the computations of the sextic resolvent, culminating in [Cayley, 1861d] that

This seems to carry the theory of the solution of a quintic, as far as it can be carried.

A year later, on 25 February 1862 he expressed a similar sense of completion.

I think now, when my paper [Cayley, 1861e] on the Tschirnhausen transformation is printed we have really
all that is wanted in the theory of quintic equations.

The computations which were the principal subject of the (extant) correspondence were algebraic, but
Cayley also referred to transcendental solutions in a letter of 29 April 1862. There, he wrote:

The series for the solution of your resolvent equation should I think be assimilated to the form of the hyper-
geometric series. . . . I attach no value whatever to the transformation of the series into definite integrals.
166 T. Crilly et al. / Historia Mathematica 44 (2017) 150–169

As to your own remark “what is wanted is a solution without the aid of definite integrals” – it seems to
me that the series is the solution and that what is wanted is a study of these series, so as that they may be
considered as known transcendants; which considering the great extent of the theory of the simplest cases,
that of the hypergeometric series, is likely to remain a desideratum. [underlining in the original]

Cayley and Harley did not pursue this together. However, Cayley had developed so much esteem for
Harley during the course of their correspondence that he put forth Harley for election to the Royal Society
in 1862. However, he had disappointing news to relate to him on 14 April 1862:

I am sorry that I have not the contrary news to send you – but the only mathematical candidates nominated
by the Council of the R.S. are Capt. Clarke and Mr. Todhunter – and your election therefore cannot take
place until next year. I was very much in hopes you would have succeed [sic] this time.

Cayley tried again the next year, and this time, in his letter of 2 May 1863, the last letter in this collection,
he had happy news to report:

I ought to have written yesterday to tell you of your being one of the selected candidates for the R.S. I hope
that I am still the first to give you the news of it. Your election is practically quite certain.

Once on the selected list of twelve candidates, election was most often a formality; Harley was elected to
the Royal Society on 4 June 1863.
After his work with Cayley, Harley maintained his interest in mathematics. In addition to his pastoral
duties at Brighouse, Harley became Professor of Mathematics and Logic in nearby Airedale College (in
Bradford) for the years 1864–1868, the college which became the United College (circa 1914). He wrote
further papers, but failed to complete a projected work entitled a “Treatise on Quintics.” Meanwhile, after
fourteen years as a barrister in London, Cayley was appointed in June 1863 as the first Sadleirian professor
of Pure Mathematics in the University of Cambridge, a position he held with distinction for the remainder
of his life.

5. The Memoir

The quintic equation was one of Cayley’s central mathematical concerns throughout his life, and as that
life was drawing to a close he started on a memoir – apparently intended to be a complete history of the
various approaches, the algebraic and possibly the transcendental as well, to the solution of the quintic –
which we may presume he intended to be part of his legacy. A draft Memoir on the Quintic Equation begins
on page 3 of the collected manuscript and runs to some 25 pages. Cayley did not complete this draft, but
rather started to rewrite it. He did not get very far; pages 1–2 are the start of a rewritten version. Most of
the pages in this manuscript are of the written memoir, but some are pages of scratch work (calculations).
One of these is on the back of an advertisement for a collection of cartoons from “Punch” magazine that
had appeared between 1871 and 1891. The first draft ends with a list of references, the first of which is
to a paper of Tschirnhausen in 1683. On the first page of this draft (page 3 of the manuscript) we find a
reference to [Cayley, 1894]. Since Cayley died in 1895, we see that he was working on it at the end.
The revised manuscript begins with a list of several forms of resolvent for the quintic equation and
associates two of them with transcendental solutions, just as the Bring–Jerrard form is associated with a
solutions via elliptic functions. “But,” wrote Cayley on page 1, “it is proper in regard to all such transcen-
dental solutions to separate them sharply off from the algebraical theory.” Cayley then proceeded to survey
the various sextic resolvent equations. He discussed in most detail the comultiple resolvent – that is, the
line that goes through Malfatti and Lagrange to Jacobi (recall that Jacobi’s resolvent is the one that Cayley
independently rediscovered) – but also discussed another resolvent due to Kronecker, building on other
 T. Crilly et al. / Historia Mathematica 44 (2017) 150–169 167

work of Jacobi [Kronecker, 1858]. This resolvent involves multipliers for elliptic functions, and Cayley
cited Hermite’s solution of the quintic in terms of these functions [Hermite, 1858]. He referred to [Cayley,
1894] and promised to describe that further, and then mentioned the icosahedral resolvent due to Felix
Klein [Klein, 1884/1956].
In his detailed analysis Cayley began with the algebraic theory. Tschirnhausen had believed that his
transformation (a change of variables) could be used to transform an arbitrary quintic to the form y 5 +
F = 0, which could evidently be solved. This was too optimistic. Bring and Jerrard had showed that the
quintic could be reduced to the form (7), but in general, no further. Cayley went through this reduction,
explaining why it works in terms of algebraic geometry. He then introduced what we would now consider
as group theoretical/combinatorial considerations to derive the comultiple resolvent, also discussing the
work of Brioschi on it. Turning to Klein’s work, Cayley reconsidered it from his own viewpoint. This
discussion is lengthy but incomplete, and indeed Cayley presented two versions of it. In all this work
Cayley dealt only with the derivation of the resolvents, and not with the use of transcendental functions to
obtain solutions of the quintic. He never managed to cast his algebraic discussion in final form, much less
to begin a discussion of transcendental solutions. Cayley was certainly familiar with this work, being an
expert on elliptic functions, the subject of the only book he wrote, but time caught up with him.

6. Conclusion

The Cayley letters that we have examined allow us to see a great mathematician at his workbench creating
the mathematics that would be published in several of his papers. Perhaps most significantly, we see him
getting the hint to connect the quintic equation to invariant theory and then exploiting that link. The letters
confirm how much he was influenced by mathematicians who are now almost forgotten, although his work
belongs to a long tradition that includes contributions by Lagrange, Malfatti, Jacobi, Kronecker, Klein,
and later, Leonard Eugene Dickson [Dickson, 1925]. That tradition did not stop in the early twentieth
century. Although the long computations may have seemed almost pointless to some of the more formal
and abstract mathematicians of the later twentieth century, the tradition never fully disappeared, and it
gained new vitality in the age of computers [Berndt et al., 2002]. Perhaps Cayley wrote his Memoir as a
valediction to a certain mathematical tradition, but its revival shows this work as an important part of his
legacy.

Acknowledgments

We would like to thank the referees for detailed and helpful comments. We would also like to thank the Royal
Society of London for permission to quote from the Referee’s Reports held in their archives.

References

Bell, H.T.M., Panteki, M., 2004. Robert Harley (1828–1910). Oxford Dictionary of National Biography.
Bennett, J.M., 2003. Sir James Cockle: First Chief Justice of Queensland 1863–1879. Federation Press.
Berndt, B.C., Spearman, B.K., Williams, K.S., 2002. Commentary on an unpublished lecture by G.N. Watson on
solving the quintic. Math. Intell. 24 (4), 15–33.
Bring, E.S., 1786. Meletemata quaedam mathematiica circa transformationem aequationum algebraicarum. Lund.
Cardano, G., 1545/1993. The Rules of Algebra. Dover.
Cayley, A., 1851. Note on Mr. Cockle’s solution of a cubic equation. Camb. Dub. Math. J. 6, 186.
Cayley, A., 1857. A memoir on the symmetric functions of the roots of an equation. Philos. Trans. Roy. Soc. 147,
489–496 (rec. 18 Dec. 1856, read 8 Jan. 1857) = Coll. Math. Papers [147], 2, 417–430.
Cayley, A., 1858. A fifth memoir upon quantics. Philos. Trans. Roy. Soc. 148, 429–460 (rec. 11 Feb. 1858, read 18
Mar. 1858) = Coll. Math. Papers [156] 2, 527–557.
168 T. Crilly et al. / Historia Mathematica 44 (2017) 150–169

Cayley, A., 1859. Sketch of a proof of the theorem that every algebraic equation has a root. Philos. Mag. 18, 436–439
= Coll. Math. Papers [248] 4, 116–119.
Cayley, A., 1860a. Note on the equation for the squared differences of the roots of a cubic equation. Q. J. Math. 3,
307–309 (dated 3 Nov. 1859) = Coll. Math. Papers [287] 4, 463–465.
Cayley, A., 1860b. On the equation of differences for an equation of any order and in particular for the equations of
the orders, two, three, four, and five. Philos. Trans. Roy. Soc. 150, 93–112 (rec. 2 Mar. 1860) = Coll. Math. Papers,
[262] 4, 240–261.
Cayley, A., 1861a. On the equation for the product of the differences of all but one of the roots of a given equation.
Philos. Trans. Roy. Soc. 151, 45–59 (rec. 30 Nov. 1860, read 10 Jan. 1861) = Coll. Math. Papers [266] 4, 276–291.
Cayley, A., 1861b. Note sur la transformation de Tschirnhausen. J. Reine Angew. Math. 58, 259–262 (dated 18 Apr.
1860) = Coll. Math. Papers [273] 4, 364–367.
Cayley, A., 1861c. Deuxième Note sur la transformation de Tschirnhausen. J. Reine Angew. Math. 58, 263–269 (dated
11 May 1860) = Coll. Math. Papers [274] 4, 368–374.
Cayley, A., 1861d. On a new auxiliary equation in the theory of equations of the fifth order. Philos. Trans. Roy.
Soc. 151, 263–276 (rec. 20 Feb. 1861, read 7 Mar. 1861) = Coll. Math. Papers [268] 4, 309–324.
Cayley, A., 1861e. On Tschirnhausen’s transformation. Philos. Trans. Roy. Soc. 151, 561–578 (rec. 7 Nov. 1861, read
5 Dec. 1861) = Coll. Math. Papers [275] 4, 375–394.
Cayley, A., 1861f. Note on Mr Jerrard’s researches on the equation of the fifth order. Philos. Mag. 21, 210–214 = Coll.
Math. Papers [310] 5, 50–54.
Cayley, A., 1861g. On a theorem of Abel’s relating to equations of the fifth order. Philos. Mag. 21, 257–263 = Coll.
Math. Papers [311] 5, 55–61.
Cayley, A., 1862a. Addition to [on a new auxiliary equation in the theory of equations of the fifth order]. Philos. Trans.
Roy. Soc. 151, 561–578 (dated 10 Nov. 1862) = Coll. Math. Papers [268] 4, 324.
Cayley, A., 1862b. Note on the [Jerrard’s] solution of an equation of the fifth order. Philos. Mag. 23, 195–196 = Coll.
Math. Papers [316] 5, 77.
Cayley, A., 1862c. Final remarks on Mr Jerrard’s theory of equations of the fifth order. Philos. Mag. 24, 290 = Coll.
Math. Papers [321] 5, 89.
Cayley, A., 1868. Note on the solvability of equations by means of radicals. Philos. Mag. 36, 386–387 = Coll. Math.
Papers [421] 7, 13–14.
Cayley, A., 1871. A ninth memoir on quantics. Philos. Trans. Roy. Soc. 161, 17–50 = Coll. Math. Papers [462] 7,
334–353.
Cayley, A., 1880. On a theorem of Abel’s relating to a quintic equation. Proc. Camb. Philos. Soc. 3, 155–159 = Coll.
Math. Papers [741] 11, 132–135.
Cayley, A., 1882. A solvable case of the quintic equation. Q. J. Math. 18, 154–157 = Coll. Math. Papers [777] 11,
402–404.
Cayley, A., 1890. On a soluble quintic equation. Am. J. Math. 13, 53–58 = Coll. Math. Papers [914] 13, 88–92.
Cayley, A. 1891. [Note on McClintock, 1884, 1886; Cayley, 1861c] = Coll. Math. Papers 4, 609–616.
Cayley, A., 1894. On the sextic resolvent equations of Jacobi and Kronecker. J. Reine Angew. Math. 113, 42–49 =
Coll. Math. Papers [950] 13, 473–479.
Cockle, James, 1852. On the method of symmetric products. Philos. Mag. 4, 492–496.
Cockle, James, 1853. On the method of symmetric products. Philos. Mag. 5, 170–174.
Cockle, James, 1854. On the method of symmetric products. Philos. Mag. 7, 130–138.
Cockle, James, 1860. Researches in the Higher Algebra. Mem. Lit. Philos. Soc. Manchester, 2nd Ser. 15, 131–142
(sub. 1858).
Cockle, James, 1861. On the resolution of quintics. Q. J. Pure Appl. Math. 4, 5–7 (sub. 1860).
Corley, T.A.B., Crilly, A.J., 2004. Sir James Cockle (1819–1895). Oxford Dictionary of National Biography.
Crilly, A., 1986. The rise of Cayley’s invariant theory (1841–1862). Hist. Math. 13, 241–254.
Crilly, A., 1988. The decline of Cayley’s invariant theory (1863–1895). Hist. Math. 15, 332–347.
Crilly, A., 2006. Arthur Cayley: Mathematician Laureate of the Victorian Age. Johns Hopkins Univ. Press.
Dickson, L.E., 1925. Resolvent sextics of quintic equations. Bull. Am. Math. Soc. 31, 515–523.
Forsyth, A.R., 1895. Arthur Cayley. Philos. Trans. Roy. Soc. 58, ix–xliv.
Goldstein, C., 2011. Charles Hermite’s stroll through the Galois fields. Rev. Hist. Math. 17, 211–270.
 T. Crilly et al. / Historia Mathematica 44 (2017) 150–169 169

Harley, R.A., 1860a. On the method of symmetric products and its application to the finite algebraic solution of
equations. Mem. Liter. Philos. Soc. Manchester, 2nd Ser. 15, 172–219.
Harley, R.A., 1860b. On the theory of quintics. Q. J. Pure Appl. Math. 3, 343–359.
Harley, R.A., 1860c. On the method of symmetric products, and certain circular functions connected with that method
(abstract). Proc. Roy. Soc. 11, 43–44.
Harley, R.A., 1861. On the method of symmetric products, and certain circular functions connected with that method.
Philos. Trans. Roy. Soc. 151, 327–356.
Harley, R.A., 1862. On the theory of quintics II. Q. J. Pure Appl. Math. 5, 248–259.
Hermite, C., 1858. Sur la résolution de l’équation du cinquième degré. C. R. Acad. Sci. 46, 508–512.
Hesse, O., 1844. Über die Wendepunkte der Curven dritter Ordnung. J. Reine Angew. Math. 28, 68–96.
Jacobi, C.G.J., 1835. Observatinculae ad theoriam aequationum pertinentes. J. Reine Angew. Math. 13, 340–352.
Jerrard, G.B., 1834. Mathematical Researches. Part II. Bristol and London.
Klein, F., 1884/1956. Lectures on the Icosahedron and the Solution of Equations of the Fifth Degree (G.G. Morrice,
Trans.). Dover.
Kronecker, L., 1858. Sur la résolution de l’équation du cinquième degré, extrait d’une lettre addressée à M. Hermite.
C. R. Acad. Sci. 46, 1150–1152.
Lagrange, J.L., 1771/1867. Réflexions sur la résolution algébrique des équations. In: Oeuvres de Lagrange, vol. III,
Serret, J.-A. (Ed.). Gauthier-Villars. Paris, pp. 265–421.
Lehigh. Cayley Materials & Research, Lehigh University Library, n.d. http://digital.lib.lehigh.edu/remain/con/
cayley.html, 29 September 2015.
McClintock, Emory, 1884. On the resolution of equations of the fifth degree. Am. J. Math. 6, 301–314.
McClintock, Emory, 1886. Analysis of quintic equations. Am. J. Math. 8, 45–84.
MacMahon, P.A., 1896. Combinatory analysis: a review of the present state of knowledge. Proc. Lond. Math. Soc. 28,
5–32.
MacMahon, P.A., 1914–1915. Robert Harley, 1828–1910. Proc. Roy. Soc. 91, i–v.
Parshall, K.H., 1989. Toward a history of nineteenth-century invariant theory. In: Rowe, D., McCleary, J. (Eds.), The
History of Modern Mathematics, vol. 1. Academic Press, pp. 157–206.
Parshall, K.H., 1998. James Joseph Sylvester: Life and Work in Letters. Oxford Univ. Press.
Sylvester, J.J., 1844. Elementary researches in the analysis of combinatorial aggregation. Philos. Mag. 24, 285–296 =
Coll. Math. Papers I, 91–102.
von Tschirnhausen, E.W., 1683. Methodus auferendi omnes terminos intermedios ex data equatione. Acta Erudito-
rium 2, 204–207.
Viète, F., 1591/2006. The Analytic Art (T.R. Whitmer, Trans.). Dover.