A history of the Burnside problem

Algebra Index History Topics Index
Version for printing
Some definitions
  1. A Group G is said to be periodic if for all gG there exists nN with gn = 1.
    (Note that the number n may depend on the element g.)

  2. A Group G is said to be periodic of bounded exponent if there exists nN with gn = 1 for all gG. The minimal such n is called the exponent of G.

It is clear that any finite group is periodic. In his 1902 paper, Burnside [1] introduced what he termed "a still undetermined point" in the theory of groups:

General Burnside Problem:
Is a finitely generated periodic group necessarily finite?

Burnside immediately suggested the "easier" question:

Burnside Problem:
Is a finitely generated periodic group of bounded exponent necessarily finite?

Let Fm denote the free group of rank m. For a fixed n let Fmn denote the subgroup of Fm generated by gn for gG.
Then Fmn is a normal subgroup of Fm(it is even an invariant subgroup), and we define the Burnside Group B(m, n) to be the factor group Fm/ Fmn .

Burnside showed a number of results in his 1902 paper;

  1. B(1, n) Cn
  2. B(m, 2) is an elementary abelian group of order 2n (a direct product of n copies of C2)
  3. B(m, 3) is finite of order ≤ 32m-1
  4. B(2, 4) is finite of order ≤ 212. (in fact Burnside claimed equality)

Burnside and Schur made early progress on the problems in two papers, which confirmed that the problem would certainly not be straightforward:

Theorem (Burnside, 1905 [2])
A finitely generated linear group which is finite dimensional and has finite exponent is finite i.e. any subgroup of GL(n,C) with bounded exponent is finite.

Theorem (Schur, 1911 [3])
Every finitely generated periodic subgroup of GL(n,C) is finite.

These results imply that any counterexample to the Burnside Problems will have to be difficult, i.e. not expressible in terms of the well-known linear groups. After this initial flurry of results, no more progress was made on the Problems until the early 1930's, when the topic was resurrected by the suggestion of a variant on the original problem:

Restricted Burnside Problem:
Are there only finitely many finite m-generator groups of exponent n?

If the Restricted Burnside Problem has a positive solution for some m, n then we may factor B(m, n) by the intersection of all subgroups of finite index to obtain B0(m,n), the universal finite m-generator group of exponent n having all other finite m-generator groups of exponent n as homomorphic images.

Note that if B(m,n) is finite then B0(m,n) = B(m,n).

Despite this formulation having been present on the seminar circuit in the 1930's, it was not until 1940 that the first paper, by Grün [6], appeared specifically addressing the RBP, and not until 1950 that the term "Restricted Burnside Problem" was coined by Magnus [7].

1933Levi, Van der Waerden [4] (independently) showed that B(m, 3) has order 3c, c = m + mC2 + mC3 and is a metabelian group of nilpotency class 3.
1940Sanov [5] proved that B(m, 4) is finite.

1954Tobin [8] showed that B(2, 4) has order 212, and gave a presentation.
1955Kostrikin[9] established that B0(2, 5) exists.

1956Higman[10] proved that B0(m, 5) exists.

P Hall and G Higman [11] showed that B0(m, 6) exists and has order 2a3b where a = 1 + (m - 1)3c , b = 1 + (m - 1)2m , c = m + mC2 + mC3 and is hence soluble of derived length 3.

1958Marshall Hall Jr. [12] proved that B(m, 6) is finite, a contribution which was described as a "heroic piece of calculation" by one reviewer.

Kostrikin[13] showed that B0(m, p) exists for all p prime.

The 1956 Hall-Higman paper contains a remarkable reduction theorem for the Restricted Burnside Problem:

Theorem (Hall-Higman, 1956 [11])
Suppose that n = p1k1. ... .prkr with p1, ... , pr distinct primes.

  1. The RBP holds for groups of exponent piki,
  2. There are finitely many finite simple groups of exponent n,
  3. The outer automorphism group Out(G) = Aut(G)/Inn(G) is soluble for any finite simple group of exponent n.
Then the RBP holds for groups of exponent n.

(Note that iii. above is the so-called Schreier Conjecture)

Now (moving ahead), the classification of finite simple groups in the 1980's shows that ii. and iii. hold. Even earlier it was known for n odd by Feit-Thompson (the "odd-order paper" of 1962), and at the time of publication must have been a reasonable conjecture.

Consequently, to prove that B0(m,n) exists for all m, n we need only (!) show that B0(m, pk) exists for all m and prime powers pk. Kostrikin had "shown" that B0(m, p) exists.

In 1989 Zelmanov announced his proof of a positive solution of the Restricted Burnside Problem and was awarded a Fields medal for this in 1994.

1959Turning back to the original Burnside Problems, Novikov announced that B(m, n) is infinite for n odd, n > 71. Novikov published a collection of ideas and theorems [14], but no definitive proof was forthcoming. John Britton suspected Novikov's proof was wrong and he began to work on the problem.

1964Golod and Shafarevich [15] provided a counter-example to the General Burnside Problem -- an infinite, finitely generated, periodic group.

1968S I Adian, P S Novikov [16] proved that B(m, n) is infinite for n odd, n ≥ 4381 with an epic combinatorial proof based upon Novikov's earlier efforts.

This saddened Britton since he was close to publishing himself, but he continued and finished in 1970. His paper was published in 1973, but Adian discovered that it was wrong. There was not a single error in any lemma. However in order to apply them simultaneously the inequalities needed to make their hypotheses valid were inconsistent. Britton never really recovered, and this was to be the last major research paper he published.

1975S I Adian [17] proved that B(m, n) is infinite if n odd, n ≥ 665, improving the Adian-Novikov result of 1968.

1982Ol'shanskii showed that given p a prime, p > 1075, then there is an infinite group, every proper subgroup of which is cyclic of order p. (This is called the Tarski Monster)
1994S V Ivanov published his proof that B(m, n) is infinite for m ≥ 2 and n ≥ 248.
1996I G Lysenok proved that B(m, n) is infinite for m ≥ 2 and n ≥ 8000.
It is still an open question whether B(2, 5) is finite or not.

References (17 books/articles)

Other Web sites:

  1. Mathworld
  2. S V Ivanov (pdf)

Article by: J J O'Connor and E F Robertson. Thanks to Andrew Isherwood.

History Topics Index
Algebra Index
Main index Biographies Index

JOC/EFR July 2002

The URL of this page is: