Lagrange's four-square theorem


Lagrange's four-square theorem, also known as Bachet's conjecture, states that every natural number can be represented as the sum of four integer squares. That is, the squares form an additive basis of order four.
where the four numbers are integers. For illustration, 3, 31 and 310 can be represented as the sum of four squares as follows:
This theorem was proven by Joseph Louis Lagrange in 1770. It is a special case of the Fermat polygonal number theorem.

Historical development

From examples given in the Arithmetica it is clear that Diophantus was aware of the theorem. This book was translated in 1621 into Latin by Bachet, who stated the theorem in the notes of his translation. But the theorem was not proved until 1770 by Lagrange.
Adrien-Marie Legendre extended the theorem in 1797–8 with his three-square theorem, by proving that a positive integer can be expressed as the sum of three squares if and only if it is not of the form for integers and. Later, in 1834, Carl Gustav Jakob Jacobi discovered a simple formula for the number of representations of an integer as the sum of four squares with his own four-square theorem.
The formula is also linked to Descartes' theorem of four "kissing circles", which involves the sum of the squares of the curvatures of four circles. This is also linked to Apollonian gaskets, which were more recently related to the Ramanujan–Petersson conjecture.

The classical proof

Several very similar modern versions of Lagrange's proof exist. The proof below is a slightly simplified version, in which the cases for which m is even or odd do not require separate arguments.
It is sufficient to prove the theorem for every odd prime number p. This immediately follows from Euler's four-square identity.
The residues of a2 modulo p are distinct for every a between 0 and /2.
To see this, take some a and define
c as a2 mod p.
a is a root of the polynomial
over the field
Finite field|.
So is .
In a field K, any polynomial of degree n has at most n distinct roots,
so there are no other a with this property, in particular not among 0 to.
Similarly, for b taking integral values between 0 and , the are distinct.
By the pigeonhole principle, there are a and b in this range, for which a2 and are congruent modulo p, that is for which
Now let m be the smallest positive integer such that mp is the sum of four squares, . We show by contradiction that m equals 1: supposing it is not the case, we prove the existence of a positive integer r less than m, for which rp is also the sum of four squares.
For this purpose, we consider for each xi the yi which is in the same residue class modulo m and between and m/2. It follows that, for some strictly positive integer r less than m.
Finally, another appeal to Euler's four-square identity shows that. But the fact that each xi is congruent to its corresponding yi implies that all of the zi are divisible by m. Indeed,
It follows that, for,, and this is in contradiction with the minimality of m.
In the descent above, we must rule out both the case y1 = y2 = y3 = y4 = m/2, and also the case y1 = y2 = y3 = y4 = 0. For both of those cases, one can check that would be a multiple of m2, contradicting the fact that p is a prime greater than m.

Proof using the Hurwitz integers

One of the ways to prove the theorem relies on Hurwitz quaternions, which are the analog of integers for quaternions. The Hurwitz quaternions consist of all quaternions with integer components and all quaternions with half-integer components. These two sets can be combined into a single formula
where are integers. Thus, the quaternion components are either all integers or all half-integers, depending on whether is even or odd, respectively. The set of Hurwitz quaternions forms a ring; that is to say, the sum or product of any two Hurwitz quaternions is likewise a Hurwitz quaternion.
The norm of a rational quaternion is the nonnegative rational number
where is the conjugate of. Note that the norm of a Hurwitz quaternion is always an integer.
Since quaternion multiplication is associative, and real numbers commute with other quaternions, the norm of a product of quaternions equals the product of the norms:
For any,. It follows easily that is a unit in the ring of Hurwitz quaternions if and only if.
The proof of the main theorem begins by reduction to the case of prime numbers. Euler's four-square identity implies that if Langrange's four-square theorem holds for two numbers, it holds for the product of the two numbers. Since any natural number can be factored into powers of primes, it suffices to prove the theorem for prime numbers. It is true for. To show this for an odd prime integer, represent it as a quaternion and assume for now that it is not a Hurwitz irreducible; that is, it can be factored into two non-unit Hurwitz quaternions
The norms of are integers such that
and. This shows that both and are equal to , and is the sum of four squares
If it happens that the chosen has half-integer coefficients, it can be replaced by another Hurwitz quaternion. Choose in such a way that has even integer coefficients. Then
Since has even integer coefficients, will have integer coefficients and can be used instead of the original to give a representation of as the sum of four squares.
As for showing that is not a Hurwitz irreducible, Lagrange proved that any odd prime divides at least one number of the form, where and are integers. This can be seen as follows: since is prime, can hold for integers, only when. Thus, the set of squares contains distinct residues modulo. Likewise, contains residues. Since there are only residues in total, and, the sets and must intersect.
The number can be factored in Hurwitz quaternions:
The norm on Hurwitz quaternions satisfies a form of the Euclidean property: for any quaternion with rational coefficients we can choose a Hurwitz quaternion so that by first choosing so that and then so that for. Then we obtain
It follows that for any Hurwitz quaternions with, there exists a Hurwitz quaternion such that
The ring of Hurwitz quaternions is not commutative, hence it is not an actual Euclidean domain, and it does not have unique factorization in the usual sense. Nevertheless, the property above implies that every right ideal is principal. Thus, there is a Hurwitz quaternion such that
In particular, for some Hurwitz quaternion. If were a unit, would be a multiple of, however this is impossible as is not a Hurwitz quaternion for. Similarly, if were a unit, we would have
so divides, which again contradicts the fact that is not a Hurwitz quaternion. Thus, is not Hurwitz irreducible, as claimed.

Generalizations

Lagrange's four-square theorem is a special case of the Fermat polygonal number theorem and Waring's problem. Another possible generalization is the following problem: Given natural numbers, can we solve
for all positive integers in integers ? The case is answered in the positive by Lagrange's four-square theorem. The general solution was given by Ramanujan. He proved that if we assume, without loss of generality, that then there are exactly 54 possible choices for such that the problem is solvable in integers for all.

Algorithms

and Jeffrey Shallit have found randomized polynomial-time algorithms for computing a single representation for a given integer, in expected running time.

Number of representations

The number of representations of a natural number n as the sum of four squares is denoted by r4. Jacobi's four-square theorem states that this is eight times the sum of the divisors of n if n is odd and 24 times the sum of the odd divisors of n if n is even, i.e.
Equivalently, it is eight times the sum of all its divisors which are not divisible by 4, i.e.
We may also write this as
where the second term is to be taken as zero if n is not divisible by 4. In particular, for a prime number p we have the explicit formula r4 = 8.
Some values of r4 occur infinitely often as r4 = r4 whenever n is even. The values of r4/n can be arbitrarily large: indeed, r4/n is infinitely often larger than 8.

Uniqueness

The sequence of positive integers which have only one representation as a sum of four squares is:
These integers consist of the seven odd numbers 1, 3, 5, 7, 11, 15, 23 and all numbers of the form or.
The sequence of positive integers which cannot be represented as a sum of four non-zero squares is:
These integers consist of the eight odd numbers 1, 3, 5, 9, 11, 17, 29, 41 and all numbers of the form or.

Further refinements

Lagrange's four-square theorem can be refined in various ways. For example, Zhi-Wei Sun proved that each natural number can be written as the sum of a sixth power and three squares.
One may also wonder whether it is necessary to use the entire set of square integers to write each natural as the sum of four squares. Wirsing proved that there exists a set of squares with such that every positive integer smaller than or equal can be written as a sum of at most 4 elements of .