Ursinus College

Digital Commons @ Ursinus College

Transforming Instruction in Undergraduate
Abstract Algebra Mathematics via Primary Historical Sources
(TRIUMPHS)

Fall 2017

The Roots of Early Group Theory in the Works of Lagrange

Janet Heine Barnett
Colorado State University-Pueblo, janet.barnett@csupueblo.edu

Follow this and additional works at: https://digitalcommons.ursinus.edu/triumphs_abstract

Part of the Algebra Commons, Curriculum and Instruction Commons, Educational Methods Commons,
Higher Education Commons, and the Science and Mathematics Education Commons
Click here to let us know how access to this document benefits you.

Recommended Citation
Barnett, Janet Heine, "The Roots of Early Group Theory in the Works of Lagrange" (2017). Abstract
Algebra. 2.
https://digitalcommons.ursinus.edu/triumphs_abstract/2

This Course Materials is brought to you for free and open access by the Transforming Instruction in Undergraduate
Mathematics via Primary Historical Sources (TRIUMPHS) at Digital Commons @ Ursinus College. It has been
accepted for inclusion in Abstract Algebra by an authorized administrator of Digital Commons @ Ursinus College.
For more information, please contact aprock@ursinus.edu.
 The Roots of Early Group Theory
in the Works of Lagrange

Janet Heine Barnett∗

February 2, 2019

1 Introduction
The problem of solving polynomial equations is nearly as old as mathematics itself. In addition
to methods for solving linear equations in ancient India, China, Egypt and Babylonia, solution
methods for quadratic equations were known in Babylonia as early as 1700 BCE. Written out
entirely in words as a set of directions for calculating a solution from the given numerical values,
the Babylonian procedure can easily be translated√ into a formula which is an early predecessor
of today’s well-known quadratic formula: x = −b± 2ab −4ac , given that ax2 +bx+c = 0. But what
2

about a ‘cubic formula’ for third degree equations, or a ‘quartic formula’ for equations of degree
four? More generally, is it always possible to compute the solutions of polynomial equations of
a given degree by means of a formula based on its coefficients, elementary arithmetic operations
and the extraction of roots alone?
As basic as this latter question may sound, its answer eluded mathematicians until the early
nineteenth century. In connection with its rather surprising resolution (revealed later in this
introduction), there emerged a new and abstract algebraic structure known as a group. Algebra,
understood prior to that time as the study of solution techniques for equations, was forever
changed as a result. In this project, we will explore an example of a particularly important
type of group, and will also meet ideas related to a second important type of group, by reading
historical excerpts from the earliest phase of the evolution of the group structure. In the
remainder of this introduction, we place this reading in context with a brief historical sketch of
efforts to find formulas for higher degree polynomial equations.
In a sense, the search for the solution to higher degree polynomials was also pursued by
ancient Babylonian mathematicians, whose repertoire included methods for approximating so-
lutions of certain types of cubic equations. The problem of finding exact (versus approximate)
solutions might be traced to the somewhat later geometric tradition of ancient Greece, in which
there arose problems such as the ‘duplication of a cube’ (i.e., constructing the side of a cube
with twice the volume of a given cube, using only straightedge and compass beginning with

∗
Department of Mathematics and Physics; Colorado State University-Pueblo; Pueblo, CO 81001-4901;
janet.barnett@csupueblo.edu.

1
the side of the given cube) that correspond to cubic equations when translated into today’s
algebraic symbolism (i.e., x3 = 2). In order to construct the line segments which served as
solutions of the geometrical problems in which they were interested, Greek mathematicians
developed various new curves, including the conic sections (i.e., parabolas, hyperbolas and
ellipses).1 Much later, the Islamic mathematician and poet Omar Khayyam (1048–1131) ex-
plicitly solved cubic equations by intersecting appropriate conic sections. For example, for a
cubic equation of the form x3 + d = bx2 , the points of intersection of the hyperbola xy = d
and the parabola y 2 + dx − db = 0 correspond to the real roots of the polynomial.2 Because
negative numbers were not allowed as coefficients (or roots) of equations at the time, however,
x3 + d = bx2 was only one of thirteen cubic forms which did not reduce to linear or quadratic
equations, each of which required different combinations of conic sections for their solution.
Khayyam gave the first systematic classification of all thirteen forms and demonstrated how to
solve each geometrically.
Unlike the Greeks, Khayyam and his fellow medieval Islamic mathematicians hoped to find
algebraic algorithms for cubic equations — similar to the quadratic formula for second degree
equations — in addition to geometric constructions based on curves. Although they were
unsuccessful in this regard, it was through Islamic texts that algebra became known in Western
Europe. As a result, the search for an algebraic method of solution for higher degree equations
was next taken up in Renaissance Italy beginning in the fourteenth century. Some equations
studied in this setting were solved through substitutions which reduced the given equation to
a quadratic or to a special form like xn = a. Most higher degree equations, however, cannot be
solved this way. Only in the sixteenth century, when methods applying to all cubic and quartic
polynomials were finally developed, did the Italian search for general solutions achieve some
success. The culmination of this search — described in footnote 3 below — is one of the great
stories in the history of mathematics.3

1
Using abstract algebra, it can now be shown that certain of these construction problems, including the
duplication of the cube, are impossible using only the Euclidean tools of a collapsible compass and an unmarked
straightedge. Likewise, these ideal Euclidean tools are not sufficient to construct a conic section.
2
To verify this, substitute y = xd from the equation of the hyperbola into the equation of the parabola, and
simplify. Of course, modern symbolism was not available to Khayyam, who instead wrote out his mathematics
entirely in words.
3
Set in the university world of sixteenth century Italy, where tenure did not exist and faculty appointments
were influenced by a scholar’s ability to win public challenges, the tale of the discovery of general formulas
for cubic and quartic equations begins with Scipione del Ferro (1465–1526), a professor at the University of
Bologna. Having discovered a solution method for equations of the form x3 + cx = d, del Ferro guarded his
method as a secret until just before his death, when he disclosed it to his student Antonio Maria Fiore (ca.
1506). Although this was only one of thirteen forms which cubic equations could assume, knowing its solution
was sufficient to encourage Fiore to challenge Niccolò Tartaglia of Brescia (1499–1557) to a public contest in
1535. Tartaglia, who had been boasting he could solve cubics of the form x3 + bx2 = d, accepted the challenge
and went to work on solving Fiore’s form x3 + cx = d. Finding a solution just days before the contest, Tartaglia
won the day, but declined the prize of 30 banquets to be prepared by the loser for the winner and his friends.
(Poisoning, it seems, was not an unknown occurrence.) Hearing of the victory, the mathematician, physician
and gambler Gerolamo Cardano (1501–1576) wrote to Tartaglia seeking permission to publish the method in an
arithmetic book. Cardano eventually convinced Tartaglia to share his method, which Tartaglia did in the form
of a poem, but only under the condition that Cardano would not publish the result. Although Cardano did not
publish Tartaglia’s solution in his arithmetic text, his celebrated 1545 algebra text Ars Magna included a cubic

2
 Although the solution methods discovered in the sixteenth century are interesting in and
of themselves, certain consequences of their discovery were of even greater importance to later
developments in mathematics. One such consequence was the discovery of complex numbers,
whose eventual acceptance was promoted by the fact that square roots of negative numbers of-
ten appear in the course of applying the cubic or quartic formulas to specific equations, only to
cancel out and leave only real roots in the end.4 The increased use of symbolism which charac-
terized Western European algebra in the Renaissance period also had important consequences,
including the relative ease with which this symbolism allowed theoretical questions to be asked
and answered by mathematicians of subsequent generations. Questions about the number and
kind of roots that a polynomial equation possesses, for example, led to the formulation of the
Fundamental Theorem of Algebra, the now well-known assertion that an nth degree equation
has exactly n roots, counting complex and multiple roots. Algebraic symbolism also allowed
questions about the relation between roots and factors to be carefully formulated, thereby lead-
ing to discoveries like the Factor Theorem, which states that r is a root of a polynomial if and
only if (x − r) is one of its factors.
Despite this progress in understanding the theory of equations, the problem of finding an
algebraic solution expressible only in terms of the coefficients, elementary arithmetic operations,
and extraction of roots for equations of degree higher than four resisted solution until the
Norwegian Niels Abel (1802–1829) settled it in a somewhat unexpected way. In a celebrated
1824 pamphlet, Abel proved that a ‘quintic formula’ for the general fifth degree polynomial
is impossible.5 The same is true for equations of higher degree, making the long search for
algebraic solutions to general polynomial equations perhaps seem fruitless. Then, as often
happens in mathematics, Abel’s ‘negative’ result produced fruit. Beginning with the central
idea of Abel’s proof — the concept of a ‘permutation’ — the French mathematician Évariste
Galois (1811–1832) used a concept which he called a ‘group of permutations’ as a means to
classify those equations which are solvable by radicals. Soon after publication of Galois’ work,
permutation groups in the sense that we know them today6 appeared as just one example of a
more general group concept in the paper On the theory of groups, as depending on the symbolic
equation θn = 1, written by British mathematician Arthur Cayley (1821–1895). Although it
drew little attention at the time of its publication, Cayley’s paper has since been recognized
as the inaugural paper in abstract group theory. By the end of the nineteenth century, group
theory was playing a central role in a number of mathematical sub-disciplines, as it continues
to do today.

equation solution method which Cardano claimed to have found in papers of del Ferro, then 20 years dead.
Not long after, a furious Tartaglia was defeated in a public contest by one of Cardano’s students, Lodovico
Ferrari (1522–1565), who had discovered a solution for equations of degree four. The cubic formula discovered
by Tartaglia now bears the name ‘Cardano’s formula.’
4
The first detailed discussion of complex numbers and rules for operating with them appeared in Rafael
Bombelli’s Algebra of 1560. Their full acceptance by mathematicians did not occur until much later, however,
in the nineteenth century.
5
The Italian mathematician Paolo Ruffini published this same result in 1799, but his proposed proof contained
a gap. Abel, who was not aware of Ruffini’s work until 1826, described it as “so complicated that it is very
difficult to decide the correctness of his reasoning” (as quoted in [Wussing, 1984, p. 97]).
6
Galois used the term ‘group’ in a slightly different way than we do today.

3
 In this project, we will study one of the early precursors of abstract group theory through
selections from the writings of the French mathematician J. L. Lagrange (1736–1813). An
important figure in the development of group theory, Lagrange made the first real advance in
the problem of solving polynomial equations by radicals since the work of Cardano and his
sixteenth century contemporaries. In particular, Lagrange was the first to suggest a relation
between permutations and the solution of equations by radicals that was later exploited by
the mathematicians Abel and Galois. We begin, in Section 2, with an overview of Lagrange’s
approach to the problem of solving general polynomial equations.

2 Lagrange on the solution of algebraic equations

Born on January 25, 1736 in Turin, Italy to parents of French ancestry, Joseph Louis Lagrange
was appointed as a professor of mathematics at the Turin Royal Artillery School by the age of
19. He spent the first eleven years of his professional life teaching there, the next twenty as a
mathematical researcher at the Berlin Academy of Sciences, and the final twenty-seven as both
a teacher and researcher in Paris, where he died on April 10, 1813. Largely self-taught, he is
remembered today for contributions to every branch of eighteenth century mathematics, and
especially for his work on the foundations of calculus. His work in algebra is also recognized as
sowing one of the seeds that led to the development of group theory in the nineteenth century.
Lagrange’s work on the algebraic solution of polynomial equations — that is, a solution
expressible only in terms of the coefficients, elementary arithmetic operations and extraction of
roots — was first published as a lengthy article entitled ‘Réflexions sur la résolution algébrique
des équations’ in the Mémoire de l’Académie de Berlin of 1770 and 1771. As with all his
research, generality was Lagrange’s primary goal in his works on equations. In seeking a general
method of algebraically solving all polynomial equations, he began by looking for the common
features of the solutions methods for quadratics, cubics, and quartics. Following a detailed
analysis of the known methods of solution, he concluded that one thing they had in common
was the existence of an auxiliary (or resolvent) equation whose roots (if they could be found)
would allow one to easily find the roots of the originally given equation. Furthermore, Lagrange
discovered that it was always possible, for any given equation, to find a resolvent equation with
roots which are related to the roots of the original equation in a very special way. This special
relationship, and a summary of its importance to his work, are nicely described in Lagrange’s
own words in the following excerpt from his Traité de la résolution des équations numériques7
[Lagrange, 1808].8

7
As suggested by its title, Traité de la résolution des équations numériques primarily addressed the problem
of numerically approximating solutions to equations; it also included a summary of Lagrange’s 1770/1771 work
on finding exact solutions to polynomial equations as an appendix entitled ‘Note XIII: Sur le résolution des
équations algébriques’ (see [Lagrange, 1808, pp. 295-327]).
8
The English translation of this and all other Lagrange excerpts in this project are due to the author.

4
 ∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞
On the solution of algebraic equations 9

The solution of second degree equations is found in Diophantus and can also be
deduced from several propositions in Euclid’s Data; but it seems that the first Italian
algebraists learned of it from the Arabs. They then solved third and fourth degree
equations; but all efforts made since then to push the solution of equations further
has accomplished nothing more than finding new methods for the third and the fourth
degree, without being able to make a real start on higher degrees, other than for certain
particular classes of equations, such as the reciprocal equations, which can always be
reduced to a degree less than half [the original degree] …
In Mémoire de l’Académie de Berlin (1770, 1771), I examined and compared the
principal methods known for solving algebraic equations, and I found that the methods
all reduced, in the final analysis, to the use of a secondary equation called the resolvent,
for which the roots are of the form

x′ + αx′′ + α2 x′′′ + α3 x(iv) + . . .

where x′ , x′′ , x′′′ , . . . designate the roots of the proposed equation, and α designates one
of the roots of unity, of the same degree as that of the equation.
I next started from this general form of roots, and sought a priori for the degree of
the resolvent equation and the divisors which it could have, and I gave reasons why this
[resolvent] equation, which is always of a degree greater than that of the given equation,
can be reduced in the case of equations of the third and the fourth degree and thereby
can serve to resolve them.
∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞

The rest of this section includes some comments and an example that elaborate on the ideas
that Lagrange proposed in this excerpt. Before continuing with your reading, check your
understanding of some basic vocabulary by completing the following task.

Task 1 This task reviews some basic vocabulary and mathematical developments dis-
cussed in the introduction and the preceding Lagrange excerpt.
(a) What do we mean when we say a polynomial equation is ‘algebraically solv-
able’? Before Lagrange, for which polynomial degrees were algebraic solu-
tions known?
(b) What is a ‘resolvent equation’? How does the degree of the resolvent equa-
tion compare to the degree of the given equation?

9
To set them apart from the project narrative, all original source excerpts are set
in sans serif font and bracketed by the following symbol at their beginning and end:
∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞

5
 Lagrange’s emphasis on the form of the roots of the resolvent equation is a mark of the more
abstract approach he adopted in all his work. Although we will not go through the analysis
which led him to identify this form, we will later (in section 4) read through several of the
arguments that Lagrange deduced from its existence.10 Notice for now that the expression for
the roots t of the resolvent equation for a polynomial of degree m is actually the sum of only
finitely many terms:

t = x′ + αx′′ + α2 x′′′ + α3 x(iv) + . . . + αm−1 x(m) .

Lagrange used prime notation (x′ , x′′ , x′′′ , . . .) here instead of subscripts (x1 , x2 , x3 , . . .) to denote
the m roots of the original equation (and not to indicate derivatives). Similarly, x(m) denotes
one of these m roots (and neither a derivative nor a power of x). The symbols α2 , α3 , . . . here
do denote powers of α, however, where α itself denotes an mth root of unity: that is, a number
for which αm = 1. Of course, α = 1 is always a solution of the equation αm = 1. By the
Fundamental Theorem of Algebra, however, we know that α = 1 is only one of m (possibly
√
distinct) solutions of the equation αm = 1. Letting i = −1, for example, the four fourth roots
of unity are {1, i, −1, −i}.11 Although Lagrange specified no restrictions on the value which α
may assume in the formula for the resolvent’s roots in the preceding excerpt, his 1808 summary
later made it clear that for m > 2, this formula requires α to be a complex-valued root of unity.
Because complex roots of unity played a major role in his analysis, Lagrange included a
detailed discussion of their properties in his works on the theory of equations. We will consider
excerpts of this discussion — especially those parts which relate to group theory concepts —
in Section 3 below. We will also look at methods for computing the values of complex roots of
unity in that section. In Section 4, we will then return to the formula Lagrange gave for the
roots of the resolvent equation (t = x′ + αx′′ + α2 x′′′ + α3 x(iv) + . . . + αm−1 x(m) ), and examine
how rearranging (or permuting) the roots x′ , x′′ , x′′′ , . . . within this formula provides us with
information about the degree of the resolvent equation.
To set the stage for these discussions, let’s first look at a specific example to see how
an auxiliary equation, even one of higher degree than the original, can be used to solve a
given equation. We note that our primary purpose with this example and its continuation (in
several parts) later in this project is to provide a context for the group-theoretic connections
which emerge from Lagrange’s theoretical work on resolvent equations (e.g., roots of unity and
permutations), rather than to provide a complete development of that work.

10
As indicated in the excerpt, Lagrange arrived at this expression for the roots of the resolvent equation
by examining several specific methods for solving cubics and quartics; that is, his knowledge of this general
principle arose from an a posteriori analysis of particular instances. An a priori analysis, in contrast, is one
that starts from a general law and moves to particular instances. The literal meanings of these Latin terms are
‘from what comes after’ (a posteriori) and ‘from what comes first’ (a priori). The arguments we will see below
illustrate how Lagrange argued from general algebraic properties in an a priori fashion to deduce the degree of
the resolvent equation.
11
This can be verified by solving (α2 − 1)(α2 + 1) = 0, or by raising each number to the fourth power. As a
preview of ideas to come later in this project, note that powers of i give all four roots: i1 = i , i2 = −1 , i3 = −i
, i4 = 1.

6
A Specific Resolvent Equation Example: Part I
This example is based on Lagrange’s analysis [Lagrange, 1770-1771] of the solution given by Gerolamo
Cardano (see footnote 3) to the following cubic equation, where n, p ∈ R+ :
x3 + nx + p = 0, (1)
Using basic algebra, it is straightforward to check (do this!) that the substitution x = t
3
− n
t
, where
t ̸= 0, transforms the cubic equation (1) into the following sixth degree equation:
t6 + 27pt3 − 27n3 = 0 (2)
In Lagrange’s terminology, equation (2) is a resolvent equation for the given cubic equation (1).
This means we can find the three roots of the given cubic equation (1) using only elementary arithmetic
operations and extraction of roots involving the given coefficients (1, n, p), by first finding the six roots
of the resolvent equation (2). Here’s how:
2
(i) Notice that the resolvent equation (2) is quadratic in t3 : [t3 ] + 27p [t3 ] − 27n3 = 0.

The quadratic formula gives us the roots of θ2 + 27pθ − 27n3 in the required algebraic form.
Denote these two roots as θ1 and θ2 .
Because n and p are positive reals, it’s easy to check that θ1 , θ2 are distinct real numbers.a
√ √
It follows that t1 = 3 θ1 and t2 = 3 θ2 are distinct real roots of the resolvent equation,
and also that t1 , t2 are expressed in the required form. (Do you see why?)
(ii) Lagrange then showed that the remaining four (complex) roots of the resolvent equation (2)
can be written in the required algebraic form, by first writing each of these four (complex) roots
as products involving t1 , t2 and the two complex-valued cubic roots of unity.b
For instance, letting α denote a complex number with α3 = 1,
the following computation shows that t3 = αt1 is also root of the resolvent equation:

(αt1 )6 + 27p(αt1 )3 − 27n3 = α6 t61 + 27pα3 t31 − 27n3 by basic algebra

= t61 + 27pt31 − 27n3 since α3 = 1, α6 = (α3 )2 = 1
= 0 since t1 is a root of the resolvent

Letting α, β denote the two complex cubic roots of unity, the remaining roots of the resolvent
equation can then be similarly obtained by setting t4 = βt1 , t5 = αt2 and t6 = βt2 .
(iii) Now that every solution t of the resolvent equation is expressed in the required algebraic form,
the substitution x = 3t − nt gives the three solutions of the original cubic equation in that same
required form; that is, using only elementary arithmetic operations and the extraction of roots
beginning from the coefficients (1, n, p) of the given cubic.

a
One way to check that θ1 , θ2 are distinct real numbers is to show that the discriminant b2 − 4ac = 272 p2 + 4(27n3 )
is strictly positive. Can you think of other ways to do this?
b
We could also express the four complex roots of the resolvent directly in terms of i and the roots θ1 , θ2 of the
quadratic θ2 + 27pθ − 27n3 by using the substitution θ = t3 to first rewrite the resolvent equation as follows:
√ √
t6 + 27pt3 − 27n3 = (t3 − θ1 )(t3 − θ2 ) = (t3 − ( 3 θ1 )3 )(t3 − ( 3 θ2 )3 )
√ √ √ 2 √ √ √ 2
= (t − 3 θ1 )(t2 + 3 θ1 t + 3 θ1 )(t − 3 θ2 )(t2 + 3 θ2 t + 3 θ2 )

Applying the quadratic formula to the quadratic factors then gives the resolvent roots in the required algebraic form.

7
 As you read the example on the preceding page, did you wonder how the six roots of the
resolvent equation give us only three roots for the original cubic? Here’s how this works for the
real-valued roots.
√
• Begin by setting x1 = t31 − tn1 and x2 = t32 − tn2 , and remember that t1 = 3 θ1 and
√
t2 = 3 θ2 , where θ1 and θ2 denote the two real-valued roots of the quadratic function
f (θ) = θ2 + 27pθ − 27n3 .

• By the Factor Theorem, we know that f (θ) = (θ − θ1 )(θ − θ2 ). Expand this to get
f (θ) = θ2 − (θ1 + θ2 )θ + θ1 θ2 .

• Since the two boxed expressions above represent the same quadratic function, the coeffi-
cients must be the same. Equating the constant coefficient gives us θ1 θ2 = (−3n)3 .

• Since θ1 = t31 and θ2 = t32 , it follows that t1 t2 = −3n, or 3n = −t1 t2 .

• Substituting this last fact in the defining expression for x1 , we have:

t1 n t1 3n t1 (−t1 t2 ) 1
x1 = − = − = − = (t1 + t2 ).
3 t1 3 3t1 3 3t1 3

• Proceeding similarly (do this!), we can show that x2 = 31 (t1 + t2 ) , so that x1 = x2 !

Analogous computations with the 2-to-1 function x(t) = 3t − nt will show that x3 = x6 and
x4 = x5 . We will look at the complex computations associated with these two equalities in
Section 4, after first briefly studying the properties of complex roots of unity.

3 Roots of unity in Lagrange’s analysis

As suggested above, roots of unity played an important role in Lagrange’s formula for the roots
of the resolvent equation, and more generally in the theory of equations. In this section, we
consider some properties of these special roots as they were described by Lagrange.
Our first excerpt on roots of unity also touches on an important point about ‘solvability’
which we have already raised; namely, the notion of ‘solvability’ can be defined in a variety of
ways. We know from the introduction, for example, that Omar Khayyam continued to seek an
algebraic algorithm for the roots of cubic equations, even after he showed these roots existed
by intersecting conic sections in a way that would have satisfied Greek mathematicians as a
legitimate solution. In our context, ‘algebraic solvability’ also specifically requires that the roots
of a given equation can be determined from its coefficients by way of a finite number of steps
involving only elementary arithmetic operations (+, −, ×, ÷) and the extraction of roots. The
solution involving trigonometric functions described below by Lagrange would thus not count
as an algebraic solution for today’s algebraist (nor would it for Lagrange) unless the specific
trigonometric values involved could be expressed in the permitted form. We examine this issue
further in the following excerpt, taken from Note XIV of Lagrange’s Traité de la résolution des
équations numériques [Lagrange, 1808].

8
 ∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞
Although equations with two terms such as

xm − A = 0, or more simply xm − 1 = 0
√
(since the former is always reducible to the latter, by putting x m A in for x), are always
solvable by trigonometric tables in a manner that allows one to approximate the roots
as closely as desired, by employing the known formula12

k k √
x = cos 360◦ + sin 360◦ −1
m m
and letting k = 1, 2, 3, . . . , m successively, their algebraic solution is no less interesting
for Analysis, and mathematicians have greatly occupied themselves with it.

∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞

Today, this trigonometric formula for the mth roots of unity is written in terms of the radian
√
measure of a circle, 2π, and the now-standard symbol i = −1:
   
2πk 2πk
x − 1 = 0 ⇔ x = cos
m
+ i sin , where k = 1, 2, 3, . . . , m.
m m
For example, the solutions of x3 − 1 = 0 are readily obtained from this formula as follows:




√
k = 1 ⇒ x = cos 2π·1
3

+ i sin 2π·1
3

= cos 2π3

+ i sin 2π
3

= − 12 + √2
3
i
k = 2 ⇒ x = cos 2π·2
3

+ i sin 2π·2
3

= cos 4π3
+ i sin 4π
3
= − 12 − 2
3
i
k = 3 ⇒ x = cos 2π·3
3
+ i sin 2π·3
3
= cos (2π) + i sin (2π) = 1

Notice that these roots include one real root and two complex roots which are conjugates of
each other. Also note that the particular trigonometric values involved here can be expressed
in terms of only finitely many basic arithmetic operations on rational numbers and their roots
√
(including i = −1).13 In fact, we could instead solve x3 − 1 = 0 by first factoring [x3 − 1 =
(x − 1)(x2 + x + 1)] and then using the quadratic formula. Thus, the cubic roots of unity are
bona fide algebraic solutions of this equation.
In his celebrated doctoral dissertation Disquisitiones Arithmeticae of 1801, the great German
mathematician Friederich Gauss (1777–1855) proved that xm − 1 = 0 can always be solved
algebraically. This fact in turn implied Lagrange could legitimately use roots of unity in his
formula for the roots of the resolvent and still obtain an algebraic solution. Lagrange commented
on this technical point in his 1808 summary, saying of Gauss’ work that it was ‘as original as
it was ingenious’ [Lagrange, 1808, p. 329].

12
A proof of this formula lies outside the scope of this project.
13
Had the coefficients of the original equation included irrational or imaginary numbers, then any number
obtained in a finite number of steps from those coefficients with basic arithmetic operations or extraction of
roots would also be allowed. Since the coefficients of x3 − 1 = 0 are integers, these operations allow only rational
numbers and their roots.

9


Task 2 Without employing a calculator or computer, use the formula x = cos 2πk +
2πk

m
i sin m (with m = 6) to express all sixth roots of unity in terms of the elemen-
tary arithmetic operations on rational numbers and their roots only.
What difficulties do we encounter when we try to do this for the fifth roots of
unity?14

As Lagrange commented, the formula stated in the preceding excerpt for expressing the
th
m roots of unity in terms of the trigonometric functions was well-known by his time. It is
straightforward to check the correctness of this formula (see Appendix II, Task I) using another
formula that was then well-known:

de Moivre’s Formula: (cos θ + i sin θ)n = cos(nθ) + i sin(nθ)

The mathematician for which this formula is named, Abraham de Moivre15 (1667–1754), ap-
pears to have had some understanding of it as early as 1707, but never published a proof. In the
first-ever published precalculus text, Introductio in analysin infinitorum (1748), the celebrated
mathematician Leonhard Euler (1707–1783) used standard trigonometric identities to prove de
Moivre’s formula in the case where n is a natural number (see Appendix II, Task II). As was
typical of Euler, he then took matters further and used power series (see Appendix II, Task
III) to establish another amazing formula:

Euler’s formula: eiθ = cos θ + i sin θ

As a corollary to Euler’s formula, de Moivre’s formula is easy to prove, although the almost
magical ease of this proof makes it seem no less mysterious:16

(cos θ + i sin θ)n = (eiθ )n = ei(nθ) = cos(nθ) + i sin(nθ)

Additionally, Euler’s formula gives us another (more concise) way to represent roots of unity:
   
2πk 2πk 2πk
x − 1 = 0 ⇔ x = cos
m
+ i sin ⇔ x=e m
i
, where k = 1, 2, 3, . . . , m.
m m

Although Lagrange himself did not make use of this exponential representation in his own
writing (despite his familiarity with Euler’s identity), we will use it as a convenient notational

√ √ √ √ √ √
10+2 5 10−2 5
14
Lagrange found the values of the fifth roots of unity to be 1, 5−1
4 ± 4 i, and − 5+1
4 ± 4 i
— but not by using trigonometric functions! See Appendix I for a sketch of one of the approaches that he used
for deriving these values.
15
Although born in France, de Moivre lived as an exile in England for most of his life as the result of severe
religious persecution of Protestants in France following the issue of the Edict of Fontainebleau in 1685.
16
Another easy but mysterious consequence of Euler’s formula, obtained by letting θ = π, is Euler’s Identity
relating the five most important mathematical constants in a single equation: eiπ + 1 = 0. After proving
Euler’s identity to an undergraduate class at Harvard, the American algebraist Benjamin Peirce (1809–1880)
is reported in [Archibald, 1925] to have said: “Gentlemen, that is surely true, it is absolutely paradoxical; we
cannot understand it, and we don’t know what it means. But we have proved it, and therefore we know it must
be the truth.”

10
abbreviation and to simplify computations. For example, the three cubic roots of unity can be
written as follows:

2π·1 2πi
k=1 ⇒ x=e 3 i=e 3
2π·2 4πi
k=2 ⇒ x=e 3 i=e 3
2π·3
k = 3 ⇒ x = e 3 i = e2πi

We also make use of the geometric representation of roots of unity as points on the unit circle
2πk


(see Figure 1) which is naturally suggested by the formula e m i = cos 2πk
m
+ i sin 2πk
m
.

2πi
e 3

2π
3
e2πi
1

4πi
e 3

   
2πk
i 2πk 2πk
Figure 1: Cubic roots of unity: e 3 = cos + i sin , k = 1, 2, 3.
3 3

Task 3 (a) Represent the sixth roots of unity on a unit circle. (See also Task 2.)
2πk πk
i i
Label these in exponential form: e 6 =e 3 , where k = 1, 2, 3, 4, 5, 6.
Also indicate clearly which of these six roots are real and which are complex.
(b) On the unit circle from part (a), identify which of the sixth roots of unity
are also square roots of unity, and which are also cubic roots of unity.
(c) Use a separate unit circle to represent the fifth roots of unity; again label with
exponential notation and indicate which are real and which are complex.

Although the geometric representation of the set of all complex numbers as points in the
two-dimensional plane shown in Figure 2 (below) was first published after Lagrange’s work on
algebraic solvability,17 mathematicians of his era were well aware of the connection of roots
of unity to the unit circle.18 In fact, Lagrange explicitly used the geometric representation of

17
The name of Parisian bookkeeper Jean-Robert Argand (1768–1822) is frequently cited in connection with
the development of the complex plane, in recognition of an 1806 pamphlet which he produced on the topic.
Although there are indications that Gauss was in possession of the geometric representation of complex numbers
as early as 1796, he did not publish on the subject until 1831. Credit for the first publication on the subject
instead belongs to the Norwegian Caspar Wessel (1745–1818); unfortunately, Wessel’s 1797 paper was written
in Danish and went unnoticed until 1897.
18
It was this relation of the mth roots of unity to points on the unit circle that led to the use of the term
‘cyclotomic polynomial’ to refer to expressions of the form ‘xm−1 + xm−2 + . . . + 1’ which arise as a factor of
xm − 1 = (x − 1)(xm−1 + xm−2 + . . . + 1).

11
roots of unity as points on the unit circle to determine the total number and type (real versus
complex) of mth roots of unity, as we read in the following excerpt from his 1770 Mémoire.19 It
may be helpful to refer back to the unit circles in Figure 1 and Task 3 as you read this excerpt.

Imaginary

z = a + bi
b

r
θ

Real
a

√
Figure 2: Geometric representation of z = a + bi = reiθ , where r = a2 + b 2 .

∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞

We first remark concerning this solution that each of the roots of the equation
x − 1 = 0 should be different from each other, since on the circumference there are
m

not two different arcs which simultaneously have the same sine and the same cosine.
It is further easy to see that all the roots will be imaginary, with the exception of the
last which corresponds to k = m and which will always equal 1, and of that which
corresponds to k = m2 , when m is even, which will equal −1; since in order for the
imaginary part of the expression of x to vanish, it is necessary to have
 
k
sin 360◦ = 0,
m
which never occurs unless the arc is equal to 360◦ or to 180◦ ; in which case we will
k
have either m = 1 or = 21 , and consequently either k = m or k = m2 ; in the first case,


k
the real part cos m 360◦ will become cos 360◦ = 1; and in the second it will become
cos 180◦ = −1.

∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞

Task 4 Compare Lagrange’s claim concerning the number and type of mth roots of unity
in the preceding excerpt to what you found for m = 6 in parts (a) and (b) of
Task 3. How convincing do you find Lagrange’s argument in general?

19
Because the following excerpt comes from a different source than that which we have quoted thus far, we
have altered the notation in it slightly (replacing, for example, n by m) to be consistent with the notation used
in our other excerpts.

12
 We now read the continuation of Lagrange’s comments on roots of unity, in which ideas
related to what later came to be known as a ‘cyclic group’ arise for the first (but not last!) time
in this project.

∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞

Now, if we let
   
360◦ 360◦ √
α = cos + sin −1,
m m
we will have …
   
k k √
k
α = cos 360◦ + sin 360 ◦
−1;
m m
so that the different roots of xm − 1 = 0 will all be expressed by the powers of this
quantity α; and thus these roots will be α, α2 , α3 , . . . αm , of which the last αm will
always be equal to 1 ….

∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞

We pause at this point in our reading of Lagrange to illustrate his central idea for the
2πi
specific case of m = 3. Using exponential notation, we set α = e 3 . Then, as noted by
Lagrange, the
h remaining
i cubic roots hof unity
i can be obtained simply by taking powers of α,
2πi 2 4πi 2πi 3
since α2 = e 3 = e 3 and α3 = e 3 = e2πi = 1. Because it is possible to generate all
the cubic roots of unity from α in this way, α is called a primitive cubic root of unity.

2π

2π

πi
Task 5 Let m = 6 and set α = cos 6
+ i sin 6
=e3.
Verify that α is a primitive sixth root of unity by computing αk for k = 2, 3, 4, 5, 6.
Comment on how these results compare to your answer to Task 3(a).

We return now to Lagrange’s comments on primitive roots of unity.

∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞

It is good to observe here that if m is a prime number, one can always represent
all the roots of xm − 1 = 0 by the successive powers of any one of these same roots,
excepting only the last; let, for example, m = 3, the roots will be α, α2 , α3 : if we take
the next root α2 in place of α, we have the three roots α2 , α4 , α6 ; but, since α3 = 1, it
is clear that α4 = α and that α6 = α3 ; so that these roots will be α2 , α, α3 , the same
as before.

∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞

13
 Let us consider what Lagrange is claiming here in more detail. In light of his comments
on primitive roots of unity in the first excerpt on the preceding page, we know the expression
‘the last root’ is a reference to the real number ‘1’, obtained by taking the ‘last’ power (αm )
2πi 2πi
of α = e m . We also know from what Lagrange has already said that α = e m is a primitive
mth root of unity, since taking powers of α has the effect of cycling through all the mth roots
of unity. In the excerpt we have just read, Lagrange has gone beyond this to claim that every
mth root of unity — other than the last root 1 — behaves in exactly this same way, provided
m is prime. In other words, Lagrange has asserted the following theorem:

Theorem
If m ∈ Z+ is prime and β is an mth root of unity with β ̸= 1,
then β is a primitive mth root of unity.

Lagrange’s illustration of this theorem for the prime m = 3 in the preceding excerpt used the
2πi
fact that α = e 3 is already known to be a primitive cubic root of unity; thus, the only other
complex root of unity β can be written as a power of α; namely, β = α2 . Using this notation,
we can re-write Lagrange’s power computations as follows: β 1 = α2 , β 2 = α4 = α3 α = α and
β 3 = α6 = [α3 ]2 = 1. This shows that β = α2 is also a primitive root of unity, and the theorem
holds in this case.

Task 6 Following his discussion of cubic roots of unity in the preceding excerpt, Lagrange


2πi
next considered the case m = 5, where α = cos 2π 5
+ i sin 2π
5
= e 5 and the
five fifth roots of unity are α, α2 , α3 , α4 , α5 = 1.
Complete the following to prove that α2 , α3 , and α4 are also primitive fifth root
of unity.
(a) Find the first five powers of α2 , and show that these are the same as the five
original roots rearranged in the following order: α2 , α4 , α, α3 , α5
(b) Find the first five powers of α3 , and show that these generate the five original
roots rearranged in the following order: α3 , α, α4 , α2 , α5 .
(c) Determine the order in which the original roots are generated using powers
of α4 .

In the next task, the specific case of m = 6 is used to show that the restriction to prime numbers
in the preceding theorem is necessary; that is, when m is composite, it is no longer the case
that every complex root of unity is also a primitive root of unity. A proof of the theorem for
the prime case is then outlined in Task 8, followed by further explorations of the composite
case in Task 9.

14
 Before turning to these tasks, we introduce some current notation and terminology that will
be convenient to use in the rest of this section. Given m ∈ Z+ , let’s denote the set of all mth
roots of unity as:

2kπ
Um = {x ∈ C|xm = 1} = {e m
i
|k = 1, 2, . . . , m}

Given any root of unity β, we will also write

⟨β⟩ = {β n |n ∈ Z},

where the set ⟨β⟩ is called either the set generated by β, or the group generated by β.
Using this terminology and notation for the results from Task 6c, for instance, we can write
2π
the group generated by α4 (where α = e 5 i ) as:

⟨α4 ⟩ = { α4 , α3 , α2 , α , α5 } = U5

Notice that, since ⟨α4 ⟩ = U5 , we can conclude that α4 is a primitive fifth root of unity. Similarly,
in the general case, β is a primitive root of unity iff ⟨β⟩ = Um . When this occurs, we will also
say that β generates Um , or that β is generator of Um .

Task 7 This task uses the case m = 6 to show that the restriction to prime numbers in the
preceding theorem is necessary; that is, if m is composite, it is no longer the case
that every root of unity can be used to generate all m roots of unity via powers.


π
Set α = cos 2π6
+ i sin 2π
6
= e 3 i.

(a) Explain why all the sixth roots of unity are obtained by taking powers of α.
That is, show that ⟨α⟩ = U6 .
(It may be useful to review the unit circle diagram from Task 3, parts (a) and (b).)
(b) Now find ⟨α2 ⟩ by taking powers of β = α2 .
Which of the sixth roots of unity do you obtain in this case?
(c) Now find ⟨α3 ⟩ by taking powers of γ = α3 .
Which of the sixth roots of unity do you obtain in this case?
(d) Which of the powers of α, other than α1 = α, are also primitive roots of
unity? That is, which powers of α generate all six of the sixth roots of unity?
Justify your response.

15
Task 8 In this task, we return to the case where m is prime, and sketch a general proof
for Lagrange’s claim that every complex mth root of unity β except β = 1 is
primitive.

Begin by assuming that m is prime and that β ̸= 1 is a complex mth root of unity.
2πi
Also let α = e m , and choose n ∈ Z+ such that β = αn with 1 ≤ n < m, where
Z+ denotes the set of positive integers. (How do we know that such a value of
n exists?)

Our goal is to prove that the powers of β generate all possible mth root of unity
by proving that ⟨β⟩ = Um . In other words, we wish to show that the list
β , β 2 , . . . , β m consisting of the first m positive integer powers of β corresponds
to some arrangement of the list α , α2 , . . . , αm of all mth roots of unity.

(a) Begin by explaining why β s is an mth root of unity for all s ∈ Z+ with
1 ≤ s ≤ m.
Note:
Since different powers of β could produce the same complex number, this
proves only that the list β , β 2 , . . . , β m contains at most m distinct mth
roots of unity.

(b) Use the fact that m is prime to prove the following:

2πi
Lemma Assume m, n ∈ Z+ with m prime, α = e m , and β = αn .
For all s ∈ Z+ , β s = 1 if and only if m divides s.
2πi


Hint? Remember that α = e m = cos 2π m
+ i sin 2π
m
.
Note:
This shows that m is the first positive power of β that generates ‘the last
root’ 1; that is, the real root 1 = β m appears only once within the list
β , β 2 , . . . , β m.
It remains to show that no other mth roots of unity are repeated in this list.

(c) Suppose that the list β , β 2 , . . . , β m contains fewer than m distinct values.
That is, suppose that β k = β l for some integers k, l with 1 ≤ k < l ≤ m.
Use the lemma proven in part (b) to derive a contradiction.
Hint? Notice that 1 ≤ l − k < m.

(d) Optional: Re-write the proof that there are no repeated elements in the list
β , β 2 , . . . , β m from part (c) without using proof by contradiction.

16
 Task 9 We now return to the case where m is a composite number and consider the
2π
number of primitive mth roots of unity in this case. To this end, let α = e m i .
(a) Recall that for m = 4, the primitive fourth roots of unity are α = i and
α3 = −i. Also review the results which you obtained in Task 7 for the case
m = 6. Use this data to develop a general conjecture concerning exactly
which powers of α are primitive roots and which are not in the case where
m is composite.
(b) Test your conjecture from part (a) in the cases of m = 8 and m = 9. Clearly
record your evidence that αs is or is not a primitive root for each value of s.
Refine your conjecture as needed before testing it for the case of m = 12.
Continue to refine and re-test it further as needed.
Once you are satisfied with your conjecture, write a general proof for it. Dis-
cuss proof strategies as needed with other students or your course instructor.
(c) Optional: Modify your conjecture for the case where m is composite so that
it applies to all values of m, both prime and composite. Also modify your
proof as needed to apply to this more general conjecture.

We close this section with a continuation of the specific resolvent equation example from the
end of Section 2. In Section 4, we will return to this example once more, and use the expressions
obtained for x′ , x′′ , x′′′ below to establish that the special relationship t = x′ + αx′′ + α2 x′′′
which Lagrange claimed will hold between the roots of an equation (x, x′ , x′′ ) and the roots of
its resolvent (t) does indeed hold in this particular example.

A Specific Resolvent Equation Example: Part II

Recall that the roots of the cubic equation x3 + nx + p = 0, where n, p ∈ R+ , are related to the roots
of the sixth degree resolvent equation t6 + 27pt3 − 27n3 = 0 by the expression x = 3t − nt .
Also recall that the six (distinct) roots t1 , t2 , t3 , t4 , t5 , t6 of this resolvent equation can be expressed as
products of its two real roots (t1 , t2 ) and the cubic roots of unity. For example, denoting the cubic
roots of unity by 1, α, α2 , where α is a primitive cubic root of unity, and letting θ1 , θ2 denote the two
(distinct) real roots of θ2 + 27pθ − 27n3 , we obtain:
√
t 1 = 3 θ1 t3 = αt1 t5 = α 2 t1
√
t 2 = 3 θ2 t4 = αt2 t6 = α 2 t2 ,

Further recall that we can use the fact that t1 t2 = −3n to show that t31 − tn1 = 13 (t1 + t2 ) = t32 − tn2 ,
concluding that x′ = 13 (t1 + t2 ) is the only real root of the given cubic equation.
Using this same fact [t1 t2 = −3n], we see that t4 t5 = (α2 t1 )(αt2 ) = t1 t2 = −3n, which in turn implies:

t4 n t4 3n t4 (−t4 t5 ) 1 t5 (−t4 t5 ) t5 3n t5 n
x4 = − = − = − = (t4 + t5 ) = − = − = − = x5
3 t4 3 3t4 3 3t4 3 3 3t6 3 3t5 3 t5

This shows that x′′ = 31 (t4 + t5 ) is one of the two complex roots of the given cubic.
Denoting the second complex root of the given cubic by x′′′ , a similar computation will show that
x′′′ = 13 (t3 + t6 ). (Be sure to check this!)

17
4 Permutations of roots in Lagrange’s analysis
We now return to Lagrange’s treatment of general polynomial equations in his 1808 note on this
topic. The first excerpt we consider states a relationship between the roots and the coefficients
of an equation which was well known to algebraists of his time; following the excerpt, we will
see how this relationship leads to the idea of permuting roots.
∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞
We represent the proposed equation by the general formula

xm − Axm−1 + Bxm−2 − Cxm−3 + . . . = 0,

and we designate its m roots by x′ , x′′ , x′′′ , . . . , x(m) ; we will then have, by the known
properties of equations,

A = x′ + x′′ + x′′′ + . . . + x(m) ,

B = x′ x′′ + x′ x′′′ + . . . + x′′ x′′′ + . . . ,
C = x′ x′′ x′′′ + . . .
∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞

Task 10 (a) For m = 2, note that the general equation becomes x2 − Ax + B = 0, where
Lagrange claimed that A = x′ +x′′ and B = x′ x′′ . Verify that these formulas
for A and B are correct by expanding the polynomial: (x − x′ )(x − x′′ ).
(b) Now write down the formulas for the coefficients A, B,C of the cubic poly-
nomial x3 − Ax2 + Bx − C in terms of its roots x′ , x′′ , x′′′ , and again verify
that these are correct by expanding the factored form of the polynomial.
(c) Use the formulas found in part (b) for the coefficients A, B, C of the cubic
polynomial x3 − Ax2 + Bx − C to determine the expanded form of the
following polynomials without multiplying out the given factors.
(i) (x − 2)(x − 3)(x − 5) (ii) (x − 1)(x − (1 + 2i))(x − (1 − 2i))

In Lagrange’s expressions for the coefficients A, B, C, . . ., note that the roots x′ , x′′ , x′′′ ,
. . . , x(m) can be permuted in any way we wish without changing the (formal) value of the
expression. For example, if we exchange x′ for x′′ (and vice-versa) in the case where m = 2,
we get A = x′′ + x′ and B = x′′ x′ , both clearly equal to the original expressions (A = x′ + x′′
and B = x′ x′′ ). For m = 3, more complicated permutations of the roots arise. For example,
we could simultaneously replace each occurrence of x′ by x′′′ , each occurrence of x′′ by x′ and
each occurrence of x′′′ by x′′ in the original expressions for A, B, C to obtain:

A = x′ + x′′ + x′′′ ⇒ A = x′′′ + x′ + x′′

B = x′ x′′ + x′ x′′′ + x′′ x′′′ ⇒ B = x′′′ x′ + x′′′ x′′ + x′ x′′
C = x′ x′′ x′′′ ⇒ C = x′′′ x′ x′′

Again we see that the expressions resulting from this particular permutation of the given roots
are formally equivalent to the original expressions. It is similarly straightforward to check that
this occurs with every possible permutation of the three roots. (Try it!)

18
 Expressions with the property that every permutation of the variables results in the same
formal value are said to be symmetric functions.20 In contrast, the expression x1 x2 + x3
is not a symmetric function since, for example, exchanging x1 and x3 results in a different
formal value (x3 x2 + x1 ), even though exchanging x1 and x2 results in an expression (x2 x1 + x3 )
equivalent to the original (x1 x2 + x3 ).

Task 11 Determine which of the following are symmetric expressions in x1 , x2 , x3 .

For any that is not, describe a permutation of x1 , x2 , x3 that changes the formal
expression, and (if possible) another permutation that does not change the formal
expression.
(a) (x1 +x2 +x3 )2 (b) x21 +(x2 +x3 )2 (c) (x1 +x2 )(x2 +x3 )

Returning now to Lagrange, we read two suggestions concerning how one might proceed to
find the resolvent equation whose solution would allow us to find an algebraic solution of the
original equation.
∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞
To obtain the [resolvent] equation …, it will be necessary to eliminate the m unknowns
x ,x′′ , x′′′ , . . . , x(m) by means of the preceding equations, which are also m in number;
′

but this process requires long calculations, and it will have, moreover, the inconvenience
of arriving at a final equation of degree higher than it needs to be.
One can obtain the equation in question directly and in a simpler fashion, by employ-
ing a method which we have made frequent use of here, which consists in first finding
the form of all the roots of the equation sought, and then composing this equation by
means of its roots.
∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞

In other words, we can try to write a resolvent equation either by working with m (non-
linear) equations in m unknowns through a series of long calculations involving symmetric
functions . . . or we can simplify this process by using the form of the roots to obtain the
resolvent equation by way the Factor Theorem, with each root contributing a factor towards
building up the resolvent equation.
In our remaining excerpts from Lagrange’s work, we will see how he set out to implement this
second plan. The tasks interspersed between these excerpts examine his argument in the specific
case m = 3. We begin with an excerpt in which Lagrange first reminded his readers about the
way in which the roots t of the resolvent appear as a function of the roots x′ , x′′ , . . . x(m) of
the original equation and the powers of a primitive mth root of unity α. His main goal in this
excerpt was to deduce the degree of the resolvent equation, based on the total number of roots
which can be formed in this way.

20
Qm
The symmetric functions given by the coefficients of the polynomial k=1 (x−xk ) are called the elementary
symmetric polynomials. An example of a non-elementary symmetric function on three variables is given by
x21 + x22 + x23 .

19
 ∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞
Let t be the unknown of the resolvent equation; in keeping with what was just said,
we set
t = x′ + αx′′ + α2 x′′′ + α3 x(iv) + . . . + αm−1 x(m) ,
the quantity α being one of the mth roots of unity, that is to say, one of the roots of
the binomial equation y m − 1 = 0. …………
It is first of all clear that, in the expression t, one can interchange the roots x′ ,
x′′ , x′′′ , . . . , x(m) at will since there is nothing to distinguish them here from one and
another; from this it follows that one obtains all the different values of t by making all
possible permutations of the roots x′ , x′′ , x′′′ , . . . , x(m) and these values will necessarily
be the roots of the resolvent in t which we wish to construct.
Now one knows, by the theory of combinations, that the number of permutations
which can be obtained from m things is expressed in general by the product 1.2.3 . . . m;
but we are going to see that this equation is capable of being reduced by the very form
of its roots.
∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞

Task 12 Consider the case m = 3 and let x′ , x′′ , x′′′ denote the three roots of an arbitrary
cubic equation and α denote a primitive cubic root of unity. According to La-
grange’s analysis in the preceding excerpt, the resolvent for the given cubic will
have a total of 3! = 6 roots, arising from the 3! = 6 possible permutations of
x′ , x′′ , x′′′ in the given formula.
Complete the list of these six roots below.
t = x′ + αx′′ + α2 x′′′ t = t =
t = x′ + αx′′′ + α2 x′′ t = t =

The previous task illustrates how permuting the roots of the original equation in the formula
for the resolvent’s roots produces m! resolvent roots in the specific case of m = 3. Of course,
since m! > m for m > 2, Lagrange’s conclusion that an equation of degree m has a resolvent
equation of degree m! hardly seems like progress. But remember what we’ve seen in our
example of the cubic equation x3 + nx + p = 0: even though the original degree 3 equation
has a resolvent equation of degree 6, that resolvent is quadratic in form, allowing us to use
substitution to reduce the resolvent degree to just 2. Lagrange ended the preceding excerpt
with the claim that a similar reduction in the degree of the resolvent is always possible, for
any polynomial of any degree.21 As Lagrange emphasized throughout his work, the key to this
reduction will be to consider the form of these roots, t = x′ + αx′′ + α2 x′′′ + . . . + αm−1 x(m) ,
and the effect of permutations on this form. To set the stage for our reading of Lagrange’s
analysis of the general case, we return to the specific cubic equation example x3 + nx + p = 0
and complete the proof that the roots of its resolvent equation do indeed assume this form.

21
Once this reduction is achieved, the next question will be whether the reduced degree is sufficiently small
that one could proceed to find an algebraic solution with known methods; if not, then some further reduction
in the resolvent’s degree would be required to complete the process.

20
A Specific Resolvent Example: Part III
In this example, we return to our exploration of the cubic equation x3 + nx + p = 0, for which n, p ∈ R+ and
its sixth degree resolvent equation is t6 + 27pt3 − 27n3 = 0.
Denoting the real roots of the resolvent as t1 , t2 and letting α be a primitive cubic root of unity, recall that
the six roots of the resolvent are:
t1 t3 = αt1 t 5 = α 2 t1 ; t2 t4 = αt2 t 6 = α 2 t2 ,
Further recall that the three roots of the given cubic can be written as follows:

x′ = 13 (t1 + t2 ) x′′ = 13 (t4 + t5 ) x′′′ = 13 (t3 + t6 )

Let’s now look at how the six roots of the resolvent can be obtained via permutations of x′ , x′′ , x′′′ in the
expression t = x′ + αx′′ + α2 x′′′ .

(i) We first note the following useful facta about sums of powers of primitive cubic roots of unity:

1 + α + α2 = 0

(ii) Substituting the above values for x′ , x′′ , x′′′ and t3 , t4 , t5 , t6 into the expression x′ + αx′′ + α2 x′′′ , then
simplifying using the facts that α3 = 1 and 1 + α + α2 = 0, we obtain:

x′ + αx′′ + α2 x′′′ = 1
3 (t1 + t2 ) + α 1
3 (t4 + t5 ) + α 2 1
3 (t3 + t6 )

1



= 3 (t1 + t2 ) + α αt2 + α2 t1 + α2 αt1 + α2 t2

1


= 3 t1 + α3 t1 + α3 t1 ) + 13 (t2 + α2 t2 + α4 t2

= 1
3 (3t1 ) + 13 t2 (1 + α2 + α) = t1
| {z }
0

A similar computation shows that x′ + αx′′′ + α2 x′′ = t2 .

(iii) Using the fact that α(x′ + αx′′ + α2 x′′′ ) = x′′′ + αx′ + α2 x′′ , along with the result of part (ii), we obtain:

x′′′ + αx′ + α2 x′′ = α(x′ + αx′′ + α2 x′′′ ) = αt1 = t3

Then using the fact that α2 (x′ + αx′′′ + α2 x′′ ) = x′′′ + αx′′ + α2 x′ along with the result of part (ii)
similarly allows us to conclude that x′′′ + αx′′ + α2 x′ = t6 .

(iv) Similar computations can then be done (do these!) to determine which of the remaining two permuta-
tions of x′ , x′′ , x′′′ in the general formula t = x′ + αx′′ + α2 x′′′ correspond to the remaining two resolvent
roots, t4 and t5 .

a
One way to verify this useful fact (1 + α + α2 = 0) is by direct computation, remembering that since α√ is not
specified to be any particular primitive cubic root of unity, there are technically two cases to check: α = − 12 + 23 i and
√
α = − 21 − 23 i. (A geometric diagram is suggestive of what is happening in this sum, but does not constitute a proof!)
Alternatively, we can verify that 1 + α + α2 = 0 by factoring α3 − 1 = (α − 1)(α2 + α + 1), and then applying the
given assumptions α3 = 1, α ̸= 1 to conclude that the second factor must equal 0.

21
 We now return to Lagrange’s argument that, given any mth degree polynomial, it is always
possible to reduce the degree of the resolvent equation to a number less than m!. As you
read through the first of these two excerpts, remember that Lagrange has already established
that every permutation of x′ , x′′ , x′′′ , . . . in the expression t will result in a root of the resolvent
equation.

∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞
One first sees that this expression is an unvariable function of the quantities α0 x′ ,
αx′′ , α2 x′′′ , . . ., and also that the result of permuting the roots x′ , x′′ , x′′′ , . . . among
themselves will be the same as that of [permuting] the powers of α among themselves.
It follows from this that αt will be the result of the simultaneous permutations of
[substituting] x′ in for x′′ , x′′ in for x′′′ , …x(m) in for x′ , since αm = 1. Similarly, α2 t
will be the result of the simultaneous permutations of [substituting] x′ in for x′′′ , x′′ in
for xiv , …x(m−1) in for x′ and x(m) in for x′′ , since αm = 1, αm+1 = α , and so on.
∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞

Task 13 Consider the case m = 3, so that t = x′ + αx′′ + α2 x′′′ and α3 = 1.

(a) Write the expression that results from t under the permutation of roots22
that simultaneously substitutes x′ in for x′′ ; x′′ in for x′′′ ; and x′′′ in for x′ .
(b) Write the expression that results from t under the permutation of powers23
of α that simultaneously substitutes α in for α0 ; α2 in for α; and α0 in for α2 .
(c) Compare the results of (a) and (b), and comment on how this illustrates
that ‘the result of permuting the roots x′ , x′′ , x′′′ , . . . among themselves will be
the same as that of [permuting] the powers of α among themselves.’
(d) Now determine the product αt and compare it to the results of parts (a)
and (b). Explain why this proves that αt is also a root of the resolvent.
(e) Determine the permutation of the powers of α that corresponds to the per-
mutation of roots that simultaneously substitutes x′ in for x′′′ , x′′ in for x′
and x′′′ in for x′′ . How could we obtain this same expression as a product
of t by a power of α?

22
One way to represent this permutation is with a function table such as the following, where the top row
represents the inputs and the bottom row represents the output value to be substituted in place of each input:
 ′′ 
x x′′′ x′
x′ x′′ x′′′
Although Lagrange himself did not use this particular representation, it was introduced soon afterwards by the
French mathematician Cauchy.
23
Representing this permutation with Cauchy’s notation (described in the previous footnote) gives:
 0 
α α α2
α α 2 α0

22
 We now continue with the remainder of Lagrange’s argument that the degree of the resolvent
can always be reduced to something less than m!.

∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞

Thus, t being one of the roots of the resolvent equation in t, then αt, α2 t, α3 t,
…αm−1 t will also be roots of this same equation; consequently, the [resolvent] equation
…will be such that it does not change when t is replaced there by αt, by α2 t, by α3 t,
…, by αm−1 t, from which it is easy to conclude first that this equation can only contain
powers of t for which the exponent will be a multiple of m.
If therefore one substitutes θ = tm , one will have an equation in θ which will be of
degree only 1.2.3 . . . [m − 1].

∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞

Let us pause here to consider exactly what Lagrange has just claimed, and why he believed
his claims to be true. Based on Task 13, the first claim in the excerpt should seem quite
believable; namely,

Thus, t being one of the roots of the resolvent equation in t, then αt, α2 t, α3 t, …αm−1 t
will also be roots of this same equation.

Note that no claim has been made that these m resolvent roots (t , αt , α2 t , α3 t, . . . , αm−1 t)
are distinct, but only that this list accounts for m of the m! roots of the resolvent. Lagrange
continued by stating, without proof, the following consequence of this fact:

…consequently, the [resolvent] equation …will be such that it does not change when t is
replaced there by αt, by α2 t, by α3 t, …, by αm−1 t, from which it is easy to conclude
first that this equation can only contain powers of t for which the exponent
will be a multiple of m.

To get an idea of what Lagrange meant here, remember that his overall plan is to build up the
resolvent equation ‘by means of its roots’ — in other words, by using what we would now call
the Factor Theorem. Let’s see how this plan plays out for the case of a general cubic equation.

23
A General Resolvent Equation Example
Start with an arbitrary cubic equation (so that m = 3), and denote its roots by x′ , x′′ , and x′′′ .
Take α to be either primitive cubic root of unity.a
Let t1 , t2 denote the following two roots of the resolvent equation:b
t1 = x′ + αx′′ + α2 x′′′
t2 = x′ + αx′′′ + α2 x′′
We then knowc that the full set of all six roots of the resolvent is as follows:

t1 , αt1 , α2 t1 ; t2 , αt2 , α2 t2 .

By the Factor Theorem, we can thus write the resolvent equation in the form

(t − t1 )(t − αt1 )(t − α2 t1 )(t − t2 )(t − αt2 )(t − α2 t2 ) = 0.

Notice that these six factors can be grouped to give two separate cubic functions:

g1 (t) = (t − t1 )(t − αt1 )(t − α2 t1 ) ; g2 (t) = (t − t2 )(t − αt2 )(t − α2 t2 ).

Expanding each separately, it is straightforwardd to show that

g1 (t) = t3 − t31 and g2 (t) = t3 − t32 .

Piecing these back together, the resolvent equation is thus given by the following function of t3 :

(t3 − t31 )(t3 − t32 ) = 0,

for which the expanded form clearly contains only powers of t for which the exponent is a multiple of m = 3:

t6 − (t31 + t32 )t3 + (t1 t2 )3 = 0.

√ √
a
Although there are two primitive cubic roots of unity, − 12 + 3
2 i and − 21 − 3
2 i, the argument outlined below does
not depend on which of these is used.
b
Note that we are not assuming that t1 and t2 are real-valued here! Nor are we assuming that t1 ̸= t2 . Rather, we are
only assuming that t1 , t2 are the roots of the sixth degree resolvent equation given by these particular arrangements of
x′ , x′′ , x′′′ in the formula for the resolvent’s roots. Our choice of which of the six possible arrangements to label as t1 was
completely arbitrary (other than a desire to maintain consistency with the notation used in the example of the specific
cubic equation x3 + nx + p = 0). Once t1 was selected, however, our choice for t2 had to differ from the arrangements
given by t1 , αt1 and α2 t1 (for reasons which should become clear later in this task).
c
In our analysis from Part I of the specific cubic equation example x3 +nx+p = 0, we arrived at these same conclusions
by substituting values into the resolvent equation which we already knew to be quadratic in form. We can not do that
in this general case, since we are now trying to prove the resolvent is quadratic in form.
d
To check this — and you should check this! — it may be useful to review Task 10 to recall how the elementary
symmetric functions can be used to avoid literally multiplying out this expression.

Notice that we can take this example one step further, since the final form of the resolvent
allows us to make the substitution θ = t3 , which reduces the resolvent equation to a quadratic:

(θ − t31 )(θ − t32 ) = 0.

24
In other words, starting from a polynomial of degree m = 3, the resolvent equation of degree
m! = 6 will always reduce (after a substitution) to an equation of degree (m − 1)! = 2! = 2 —
just as Lagrange claimed is the case in the final sentence of the last excerpt:

∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞

If therefore one substitutes θ = tm , one will have an equation in θ which will be of

degree only 1.2.3 . . . [m − 1].

∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞∞

In Task 14, you will examine this claim for the case of m = 4. Unfortunately, despite the
fact that the resolvent ‘can only contain powers of t for which the exponent will be a multiple of
m’ reduces the degree of the resolvent from m! to (m − 1)!, some difficulty remains even with
relatively small values of m. Granted, a resolvent for a quintic equation undergoes a reduction
from degree 5! = 120 down to degree 4! = 24 — but 24 is still considerably larger than the
original equation’s degree of 5.
Although a similar problem would seem to arise for quartics, where the initial resolvent
degree of 4! = 24 is reduced only to 3! = 6, Lagrange used other arguments to show that the
resolvent in this case (m = 4) could be further reduced to just a cubic equation.24 He also
explained how this reduction relates to the effect of permuting x′ , x′′ , x′′′ , x(iv) in the expression
for the resolvent’s roots t, again focusing on the form of the expression in question. In essence,
he showed that the resolvent root t can be written in a sufficiently symmetric way that only 3
‘values’ arise when x′ , x′′ , x′′′ , x(iv) are permuted in all possible ways. More specifically, Lagrange
′ ′′ ′′′ (iv)
showed that t = x x +x2 x , and that this (nearly symmetric) expression assumes only three
distinct forms when x′ , x′′ , x′′′ , x(iv) are permuted in all 24 possible ways — check this if you
like!
Despite his success with polynomials of degree 3 and 4, Lagrange suspected (and Abel and
Galois later confirmed) that it is not always possible to express the resolvent roots of equations
in a form that is sufficiently symmetric to achieve a similar result for polynomials of degree five
and higher. Nevertheless, Lagrange’s introduction of permutations into the picture was the first
significant step forward in centuries in the study of algebraic solvability. It also paved the way
for Cauchy’s development of a more general theory of permutations and a second important
type of group called a permutation group. The similarities between the algebraic structure of
the set of roots of unity and the algebraic structure of a permutation group in turn promoted
further Cayley’s ability to eventually define the notion of a completely abstract group that is
now the object of study in every abstract algebra course. But that is the subject of the next
phase of the evolution of the abstract group structure — with Task 14 on the following page,
we bring the particular story of Lagrange’s contributions to that evolution to a close.

24
The strategy of reducing a quartic to a cubic was known to Renaissance algebraists; see footnote 3.

25
 Task 14 This task outlines a rigorous proof of Lagrange’s claim that the resolvent ‘can
only contain powers of t for which the exponent will be a multiple of m’ in the
case of m = 4.
Let x′ , x′′ , x′′′ , x(iv) denote the four roots of an arbitrary quartic equation.
Let α be a primitive fourth root of unity.25

(a) Show that the 4!=24 roots of the resolvent for the given quartic can be
partitioned into six disjoint sets of 4 roots, each of which has the form
Si = {ti , αti , α2 ti , α3 ti }, where i ∈ {1, 2, 3, 4, 5, 6} and ti is a particular
root of the resolvent.
You might start, for example, by setting

t1 = x′ + αx′′ + α2 x′′′ + α3 x(iv) .

To abbreviate writing, denote the order in which the four roots appear in
this expression by ‘1, 2, 3, 4’ and note that the set S1 = {t1 , αt1 , α2 t1 , α3 t1 }
contains the roots corresponding to the following formal expressions:

t1 = x′ + αx′′ + α2 x′′′ + α3 x(iv) (1, 2, 3, 4)

αt1 = x(iv) + αx′ + α2 x′′ + α3 x′′′ (4, 1, 2, 3)
α 2 t1 = x′′′ + αx(iv) + α2 x′ + α3 x′′ (3, 4, 1, 2)
α 3 t1 = x′′ + αx′′′ + α2 x(iv) + α3 x′ (2, 3, 4, 1)
You can then choose t2 to be any of the remaining 20 expressions obtained
from another permutation of x′ , x′′ , x′′′ , x(iv) in the resolvent root formula.
Explain how you can now be sure that the formal expressions for the ele-
ments in the set S1 = {t1 , αt1 , α2 t1 , α3 t1 } are distinct from those in the
set S2 = {t2 , αt2 , α2 t2 , α3 t2 }.
Then determine suitable values of t3 , t4 , t5 , t6 (without necessarily writing out
the full formal expression for each) and explain how you are sure the sets S1 ,
S2 , S3 , S4 , S5 , S6 are mutually disjoint with respect to formal expressions.
(b) For i ∈ {1, 2, 3, 4, 5, 6}, let gi (t) = (t − ti )(t − αti )(t − α2 ti )(t − α3 ti ).
Show that gi (t) = t4 − t4i .
Hint? To avoid literally multiplying out the four factors in gi , it may be help-
ful to review Task 10 and instead write out the elementary symmetric func-
tions that define the coefficients of a quartic t4 − At3 + Bt2 − Ct + D in terms
of its roots. Also remember that α is a primitive fourth root of unity (either i
or −i), and think about the values of α2 +1 and 1+α+α2 +α3 in either case.
(c) Conclude that the resolvent is a function of t4 , and explain why it can there-
fore be treated as a polynomial of degree 6 only. Why is this not a sufficient
reduction to complete the algebraic solution of the original quartic?

Although there are also two primitive fourth roots of unity, i and −i, the argument outlined below does not
25

depend on which of these is used. Be careful not to implicitly assume that α = i in your proof!

26
APPENDIX I: Optional exercises on algebraic solvability of x5 − 1 = 0
From our reading of Lagrange, we know that the five fifth roots of unity are given by the following
trigonometric expressions:
   
2πk 2πk
x = cos + i sin , for k = 1, 2, 3, 4, 5
5 5

Note that it is not be immediately obvious that these (trigonometric) values can be expressed in the
proper (algebraic) form. Indeed, Lagrange commented (page 9 of project) that the general problem of
proving this for roots of unity is one that “mathematicians have greatly occupied themselves with …”, and
commended Gauss’ eventual (1801) proof of this fact for being “as original as it was ingenious.” For the
specific case of the fifth roots of unity, however, Lagrange also noted (in footnote 14) that these can be
(algebraically) expressed as follow:
√ p √ √ p √
5−1 10 + 2 5 5+1 10 − 2 5
1 , ± i , − ± i
4 4 4 4

Task I
Verify that Lagrange’s values are correct by (algebraically) solving the equation x5 − 1 = 0
as follows.

• Begin by factoring the equation x5 −1 = 0 to obtain: (x−1)(x4 +x3 +x2 +x+1) =

0
The first factor gives the obvious fifth root, x = 1.
Thus, we need only (algebraically) solve the
equation associated with the second factor: x4 + x3 + x2 + x + 1 = 0
1 1
• Divide this (fourth-degree) equation by x2 to obtain: x2 + x + 1 + x
+ x2
=0
(How do we know that it is okay to do this?)
• Now use the (clever) substitution y = x + x1 to obtain a quadratic equation in y.
Use the quadratic formula to (algebraically) solve this equation.
• After (algebraically) solving the quadratic equation from the previous step for y, ‘undo’
the substitution y = x + x1 to obtain the (four) complex-valued solutions of x5 − 1 = 0
which were stated by Lagrange. (The quadratic formula will be needed again here.)

Task II
Use Lagrange’s values for the fifth roots of unity to algebraically express (e.g., using only
elementary arithmetic operations and extraction of roots) the following trigonometric values,
and explain how you know these are correct:

2π

4π

6π

8π


cos 5
cos 5
cos 5
cos 5

2π

4π

6π

8π


sin 5
sin 5
sin 5
sin 5

27
APPENDIX II: Optional exercises on de Moivre’s and Euler’s Formulas

The following tasks explore proofs of the following famous formulas, discussed in Subsection 1.1:

de Moivre’s Formula: (cos θ + i sin θ)n = cos(nθ) + i sin(nθ)

Euler’s formula: eiθ = cos θ + i sin θ

Task I



Let x = cos 2πk
m
+ i sin 2πk
m
where k ∈ {1, 2, 3, . . . , m}.
Use de Moivre’s formula to show that xm = 1.

Task II
This exercise outlines a proof of de Moivre’s formula for natural numbers n, based only on trigono-
metric sum identities and induction.
Recall the following trigonometric identities:

cos(x + y) = cos x cos y − sin x sin y ; sin(x + y) = cos x sin y + cos y sin x

(a) Prove that de Moivre’s formula holds for n = 2 by expanding (cos θ + i sin θ)2 and applying
the sum identities with x = y = θ.

(b) Use the result of part (a) to expand (cos θ + i sin θ)3 ; then use the sum identities with x = θ
and y = 2θ to prove de Moivre’s formula holds for n = 3.

(c) Use the sum identities and induction on n to prove de Moivre’s formula in the case where n
is a natural number.

Task III
Use the power series for ex , sin x and cos x given below to prove Euler’s formula.26
Recall that i2 = −1, i3 = −i and i4 = 1.

X∞
1 k X∞
(−1)k 2k+1 X
∞
(−1)k
x
e = x sin x = x cos x = x2k
k=0
k! k=0
(2k + 1)! k=0
(2k)!

Note: Recall that all three series are absolutely convergent on R; thus, we can legitimately
rearrange their terms, without in so doing changing the values to which they converge.

26
Euler’s derivation of this formula used a somewhat different approach in which infinitesimals appeared.

28
References
Raymond Clare Archibald. Benjamin Peirce, 1809–1880: Biographical Sketch and Bibliography.
The Mathematical Association of America, Oberlin, OH, 1925.
Arthur Cayley. On the theory of groups, as depending on the symbolic equation θn = 1 - Part I.
Philosophical Magazine, 7:151–242, 1854. Also in The Collected Mathematical Papers of Arthur
Cayley, Cambridge: Cambridge University Press, Volume 2 (1889), 123–130.
J. L. Lagrange. Réflexions sur la résolution algébrique des équations. Mémoire de l’Académie de
Berlin, 1770-1771. Also in Œuvres complètes de Lagrange, Tome 3, 205-421.
J. L. Lagrange. Traité de la résolution des èquations numériques de tous les degrés, avec des notes
sur plusieurs points de la théorie des équations algébriques. Courcier, Paris, 1808.
Hans Wussing. The Genesis of the Abstract Group Concept: A Contribution to the History of the
Origins of Abstract Group Theory. MIT Press, Cambridge and London, 1984.

29
Notes to Instructors
This Primary Source Project (PSP) draws on works by one of the early precursors of abstract group,
French mathematician J. L. Lagrange (1736-1813). An important figure in the development of group
theory, Lagrange made the first real advance in the problem of solving polynomial equations by
radicals since the work of Cardano and his sixteenth century contemporaries. In particular, Lagrange
was the first to suggest the existence of a relation between permutations and the solution of equations
by radicals, a suggestion later exploited by Abel and Galois. In addition to the important group-
theoretic concept of a permutation, the project employs excerpts from Lagrange’s study of roots
of unity to develop the concept of a finite cyclic group. Lagrange’s description of his quest for a
general method of algebraically solving all polynomial equations is also a model of mathematical
research that make him a master well worth reading by today’s students of mathematics. Through
their guided reading of excerpts from Lagrange, students thus encounter his original motivations
while developing their own understanding of these important group-theoretic concepts via the very
familiar and concrete context of solving polynomial equations.
Absolutely no familiarity with group theory is assumed in this PSP! Instead, it was explicitly
designed to serve as students’ very first encounter with group-related ideas. This approach has been
tested with good effect by instructors at a variety of institutions who have taught with its parent
PSP, an extended primary source project entitled Abstract Awakenings in Group Theory that was
developed with funding from a prior NSF grant.27
In fact, the current PSP is based on the first section of that earlier PSP, which then goes on
to develop a significant portion of the core topics in elementary group theory from the standard
curriculum of a one semester junior-level abstract algebra course.28 Instructors who begin their
study of group theory with the PSP The Roots of Early Group Theory in the Works of Lagrange
and then wish to continue with the pedagogy of primary source projects throughout their students’
study of group theory could thus easily shift over to the PSP Abstract Awakenings of Algebra.29

27
The full title of this earlier PSP (developed under the NSF grant DUE-0715392) is Abstract Awakenings in Group
Theory: Early group theory in the works of Lagrange, Cauchy, and Cayley. Its centerpiece is the 1854 inaugural paper
on abstract group, Arthur Cayley’s On the theory of groups, as depending on the symbolic equation θn = 1 [Cayley,
1854]. In keeping with the historical record, and to provide concrete examples on which to base their abstraction of
the group concept, Section 1 of that project begins with the material from Lagrange in the PSP that you are currently
reading. Section 2 then employs selections from writings by Augustin Cauchy in which a more general theory of
permutations and symmetric groups was developed independently of the theory of equations, and today’s current
notation for permutations was first introduced. Section 2 also includes Cauchy’s statement and proof of Lagrange’s
Theorem for Symmetric Groups, both of which are easily adapted to the more general case of any finite group. The
project then turns to a detailed reading of Cayley’s complete paper in Sections 3 and 4, paying careful attention to
the similarities between the theory of permutation groups as it was developed by Cauchy and the modern notion of
an abstract group as it was unveiled by Cayley.
28
Topics developed in the PSP Abstract Awakenings in Group Theory include roots of unity, permutations, definition
and elementary properties of group (including results related to the order of group elements), abelian groups, cyclic
groups, symmetric groups, alternating groups, Cayley tables, Lagrange’s Theorem, group isomorphisms, classification
of groups of small order, and direct products. The concept of cosets are also introduced in the main body of the
project, and further developed in an appendix which also states the definitions of normal subgroup and factor group.
Completion of the entire project takes approximately 10 weeks, but (un)covers the vast majority of the elementary
group theory typically studied in a junior level abstract algebra course.
29
To obtain the most recent version of Abstract Awakenings in Group Theory, contact the author at

30
For those who prefer a less extended use of this instructional practice, the PSP The Roots of Early
Group Theory in the Works of Lagrange could also be used in conjunction with a more traditional
textbook. In either case, this PSP will be more effective as an exploratory introduction to the group
concept if it is used before students have studied the concepts of cyclic groups and permutations /
permutations groups in much, if any, detail.
Classroom implementation of this and other PSPs may be accomplished through individually
assigned work, small group work and/or whole class discussion. A combination of these instructional
strategies is recommended in order to take advantage of the variety of questions included in the
project. To reap the full pedagogical and mathematical benefits offered by the PSP approach,
students should be required to read assigned sections and complete advance work on tasks related
to that reading prior to in-class discussions. The author’s method of ensuring that advance reading
takes place is to require student completion of “Reading Guides” (or “Entrance Tickets”); see pages
36–37 below for a sample guide based on this particular PSP.30 The author’s students do receive
credit for completion of each Reading Guide, but with no penalty for errors in solutions.
A sample implementation schedule is recommended on pages 37–38 below. The following de-
scription of the content of each section of the PSP should assist instructors in determining how best
to adapt that recommended schedule to their own course goals and students’ needs.

• Section 1: Introduction
This section includes a broad overview of the early history of the theory of equations, as it
relates to the eventual development of group theory. This material is intended to provide
students with a context for their study of Lagrange’s work in this PSP, and for their later
study of the more abstract concept of a group.
• Section 2: Lagrange on the algebraic solution of equations
The primary objective of this section is to introduce the language and notation used by
Lagrange in his analysis of the problem of solving equations by radical. The concept of a
‘resolvent equation’ is discussed in the context of Lagrange’s motivation for studying such
equations. A very concrete example, based on a problem initially solved by Cardano, is given
to illustrate how a resolvent equation can be used to solve a polynomial equation. Aspects of
this example are also considered in later parts of the project, in connection with Lagrange’s
work on roots of unity and permutations. Examining this example in this way31 connects well
with students’ prior, largely procedural, experiences with algebra, thereby providing a nice
bridge between their former and future algebraic studies.

janet.barnett@csupueblo.edu, or visit www.cs.nmsu.edu/historical-projects/projects.php for an earlier

version. Within that earlier version, all resolvent equation examples are instead presented as tasks for students
to complete themselves. An alternative version of the PSP The Roots of Early Group Theory in the Works of La-
grange which adopts that more open-ended/inquiry-based approach is also available upon request from the author.
30
The author’s Reading Guides typically include “Classroom Preparation” exercises (drawn from the PSP Tasks)
for students to complete prior to arriving in class; they may also include “Discussion Questions” that ask students
only to read a given task and jot down some notes in preparation for class work. On occasion, tasks are also assigned
as follow-up to a prior class discussion. In addition to supporting students’ advance preparation efforts, these guides
provide helpful feedback to the instructor about individual and whole class understanding of the material.
31
The author’s decision to present this and other resolvent equations in the PSP as examples (rather than tasks
for students to work through themselves) is due to the fact that resolvent equations themselves are no longer a focal
point in the study of group theory. As noted in a previous footnote, however, an alternate version of the PSP which
presents these examples as student tasks is available upon request.

31
 • Section 3: Roots of unity in Lagrange’s analysis
This section introduces students to the concept of finite cyclic groups via the context of
Lagrange’s writing on the solution of equations. Current notation and terminology for cyclic
groups is included. Two particular tasks in this section are especially critical. Task 8 leads
students through a particular proof of Lagrange’s claim that every non-unity complex mth
root of unity β is a generator of the (cyclic) group formed by the mth roots of unity in the
case where m is a prime. The proof strategy in question takes advantage of the fact that
Um is finite by having students prove that < β >⊆ Um with | < β > | = m, in order to
conclude that < β >= Um . Task 9 then asks students to explore and prove a conjecture
about which complex mth root of unity is a generator of the (cyclic) group formed by the mth
roots of unity in the case where m is a composite. Both these exercises allow students to draw
on the very concrete context of complex numbers, but without assuming much in terms of
background knowledge about the complex numbers. The proofs generated by students within
this familiar context also provide a good basis for proving these same results in the more
general case of an arbitrary cyclic group. The scaffolding offered by the proof outline given
in Task 8 is intended to offer students further support in their early efforts at formal proof
writing in abstract algebra.
• Section 4: Permutations of roots in Lagrange’s analysis
This section returns to Lagrange’s analysis of the degree of the resolvent equation, in which
the concept of a permutation naturally arises within the context of his writing on the solution
of equations. The elementary symmetric functions are introduced as formulas that give the
coefficients of a polynomial in terms of its roots; students find this a surprising and fascinating
method of expanding the factored form of polynomial. The focus of the excerpts and tasks
in this section, however, is on the concept of a permutations. The culminating resolvent
equation example and the related Task 14 also offer a natural context for the concept of a
n-cycle. Although operations on permutations are not introduced in this section, the modern
notation for permutations and for n-cycles (both originally due to Cauchy) is introduced.
• Appendix I: Optional exercises on fifth roots of unity
This optional appendix includes very straightforward exercises (involving little more than the
quadratic formula) that are related to Lagrange’s discussion of the algebraic solvability of
equations of the form xm − 1 = 0. Together with those in Appendix II, these exercises can be
used to offer students an opportunity to begin practicing more formal mathematical writing,
but within the familiar context of algebraic equations / manipulations (rather than proofs).
• Appendix II: Optional exercises on de Moivre’s and Euler’s Formulas
This optional appendix includes exercises related to the material on roots of unity from Section
2, including a proof by induction of deMoivre’s Theorem.

LATEXcode of the entire PSP is available from the author by request to facilitate preparation of
reading guides or ‘in-class task sheets’ based on tasks included in the project. The PSP itself can
also be modified by instructors as desired to better suit their goals for the course.

32
Acknowledgments
The development of this student project has been partially supported by the Learning Discrete
Mathematics and Computer Science via Primary Historical Sources project with funding from the
National Science Foundation’s Course, Curriculum & Laboratory Improvement Program (CCLI)
Program under grant number 0715392, and by the Transforming Instruction in Undergraduate
Mathematics via Primary Historical Sources (TRIUMPHS) project with funding from the National
Science Foundation’s Improving Undergraduate STEM Education (IUSE) Program under grant
number 1523494. Any opinions, findings, and conclusions or recommendations expressed in this
project are those of the author and do not necessarily reflect the views of the National Science
Foundation.

With the exception of excerpts taken from published transla-

tions of the primary sources used in this project and any direct
quotes from published secondary sources, this work is licensed under
a Creative Commons Attribution-ShareAlike 4.0 International License
(https://creativecommons.org/licenses/by-sa/4.0/legalcode).
It allows re-distribution and re-use of a licensed work on the con-
ditions that the creator is appropriately credited and that any
derivative work is made available under “the same, similar or a
compatible license.”

For more information about Learning Discrete Mathematics and Computer Science via Primary
Historical Sources, visit https://www.cs.nmsu.edu/historical-projects/.

For more information about Transforming Instruction in Undergraduate Mathematics via Primary
Historical Sources (TRIUMPHS), visit http://webpages.ursinus.edu/nscoville/TRIUMPHS.html.

33
Sample PSP Implementation Schedule (based on a 50-minute class period)

The following 7-day implementation schedule includes 2.5 days for small group work on
Tasks 8 and 9. These tasks represent the core material of Section 3, and the PSP itself!
Depending on the student group, however, one fewer day of in-class work may suffice for
small group discussion of these two tasks.
Some instructors may also choose to spend one additional day on the permutation material
in Section 4. The sample schedule below provides for only 2 days on that section, since its
purpose is primarily to bring closure to the historical context provided by this PSP with
regard to why permutations were initially of interest to algebraists. A suggestion for how a
third day could be profitably spent is included with the Day 7 Homework suggestion below.

• Advance Preparation Work for Day 1 (to be completed before class)

Read pages 1–9 in Sections 1 and 2 and complete Tasks 1–2 for class discussion, per the sample
Reading Guide on pages 37–38 below.
• Day 1 of Class Work
– Whole class and/or small group discussion of the following:
∗ (Optional) Historical and mathematical ideas from Section 1 (Introduction), if desired
∗ Mathematical concepts in Section 2, including answers to Task 1.
∗ Resolvent Equation Example - Part I (page 7).
∗ Assigned reading in Section 3, including answers to Task 2; this discussion could also be
postponed for Day 2 of Class Work, after instructor’s review students’ advance work.
– Time permitting, the instructor could preview the formulas on page 10, or have students
continue reading (individually or in small group) Section 3, pages 10–11 (through Task 3).
– Homework (optional): A complete formal write-up of some or all of the Tasks in Appendix
I could be assigned, to be due at a later date (e.g., one week after completion of the in-class
work).

34
• Advance Preparation Work for Day 2
In Section 3, read pages 10–14, completing Tasks 3, 4, 5, and 6(a) for class discussion along the
way. (Depending on the general background of the students, the advance reading for Day 2 could
also include page 15 and preparation of Task 7(a), but this is ambitious.)

• Day 2 of Class Work

– Brief whole class discussion of terminology and notation introduced in advanced reading:
de Moivre’s Formula, Euler’s Formula, primitive root of unity, etc.
This could include review of the answers to some or all of the following: Tasks 3, 4, 5, 6(a).
– Small group work on remaining parts of Task 6.
– Time permitting, the instructor could preview the notation and terminology discussed on
page 15, or have students continue reading (individually or in small group) Section 3, page
15 (through Task 7).
– Homework (optional): A complete formal write-up of some or all of the Tasks in Appendix
II could be assigned, to be due at a later date (e.g., one week after completion of the in-class
work).

• Advance Preparation Work for Day 3 (to be completed before class)

Read page 15 and complete Task 7; also read through all parts of Task 8 (page 16), and prepare
notes for class discussion.

• Day 3 of Class Work

– Brief whole class discussion of terminology and notation introduced in advanced reading:
Um , ⟨β⟩, primitive root of unity, generator, etc.
This could include answers to Task 7 by way of illustration.
– Begin small group work on Task 8.
– Homework: A complete formal write-up of student work on Task 8 should be assigned, to
be due at a later date (e.g., one week after completion of the in-class work).

• Advance Preparation Work for Day 4 (to be completed before class)

Review Task 8 in-class work from Day 3, and prepare notes for its continuation.
Complete Task 9, part (a), and perhaps also the first portion of part (b).

• Day 4 of Class Work

– Small group discussion of Task 8 and Task 9.
– Summarizing whole class discussion of cyclic group ideas introduced in Section 3.
– Time permitting, whole class or small group review and discussion of Resolvent Equation
Example - Part II.

35
• Advance Preparation Work for (optional) Day 5 (to be completed before class)
Review Task 9 in-class work from Day 4, and prepare notes for its continuation.
Also finish reading Section 3 (last half of page 17).
Note: Tasks 8 and 9(b) are the core material of Section 3, and the PSP itself! De-
pending on the student group, however, one day of in-class work may suffice for small
group discussion of these two tasks.

• (Optional) Day 5 of Class Work

– Continued small group discussion of Task 9 and Task 10, supplemented by whole class dis-
cussion as needed
– Whole class discussion of the Resolvent Equation Example - Part II (page 17).
– Time permitting, a summarizing whole class discussion of cyclic group ideas introduced in
Section 3.
– Homework: A complete formal write-up of student work on 10(b) should be assigned,
to be due at a later date (e.g., one week after completion of the in-class work).

• Advance Preparation Work for Day 6 (to be completed before class)

In Section 4, read pages 18–20 (stopping above Task 12), completing work on the following for
class discussion along the way: some/all of Task 10, some/all of Task 11, and all of Task 12.

• Day 6 of Class Work

– As needed, a summarizing whole class discussion of cyclic group ideas introduced in Section
3, including whole class or small group review and discussion of Resolvent Equation Example:
Part II as desired.
– Whole and/or small group discussion of the ideas in advance reading, to include answers to
some/all of Tasks 10–12.
– Time permitting, the instructor could preview the ideas discussed at the bottom of page 20.

• Advance Preparation Work for Day 7 (to be completed before class)

Complete reading of Section 4, pages 20–24, completing work on some/all of Task 13 for class
discussion along the way.

• Day 7 of Class Work

– Whole group discussion of ideas from assigned reading, including the General Resolvent
Equation Example on page 24.
– Summarizing whole group discussion of ideas related to permutations from Section 4.
– Time permitting, begin small group work on Task 14.
– Homework: A complete formal write-up of student work on Task 14 should also be assigned,
to be due at a later date (e.g., one week after completion of the in-class work).
Note: As desired an additional day of class work can be scheduled to allow for
students to work together in small groups on Task 14.

36
SAMPLE READING GUIDE

Background Information: The primary goal of this two-page reading and tasks assigned in this guide is
tto familiarize students with Lagrange’s notation and terminology related to resolvent equations in order
to prepare them for in-class small group work on Task 1 and Task 2 .

**********************************************************************************************
Reading Assignment - The Roots of Early Group Theory in the Works of Lagrange, pages 1–9.

1. Read the Introduction, pages 1–4. Jot down any comments or questions you have here.

2. From Section 2, read pages 4–5.

Them complete Task 1 (page 5) in preparation for class discussion here:

Task 1 This task reviews some basic vocabulary and mathematical developments discussed in
the introduction and the preceding Lagrange excerpt.
(a) What do we mean when we say a polynomial equation is ‘algebraically solvable’? For what
degree polynomials were algebraic solutions known when Lagrange was writing?

(b) What is a ‘resolvent equation’? How does the degree of the resolvent equation compare to
the degree of the given equation?

37
3. Now complete your reading of Section 2, by finishing pages 6–8.

• What specific equation is being solved in the Example on page 7?

• Write at least one comment and at least one question about the ideas presented in this
project thus far.

4. In Section 3, read pages 8–9. Jot down any comments or questions you have here.

5. DISCUSSION Jot down your notes about Task 2 from page 10 here.



Task 2 Without employing a calculator or computer, use the formula x = cos 2πk m + i sin m
2πk

(with m = 6) to express all sixth roots of unity in terms of the elementary arithmetic operations
on rational numbers and their roots only.

What difficulties do we encounter when you try to do this for the fifth roots of unity?
(See footnote 14 on page 10 for Lagrange’s list of the five fifth roots of unity.)

38