Worked solution: Hilbert's existence proof for Waring's problem (1909)

Step 2 of 7: Reduction: rational combinations of

k

-th powers already suffice

Previous step Next step

In plain words

Hilbert's strategy does not attack integers directly. Instead, it first shows something a little different: that a large class of integers can be written as a positive rational-number combination of $k$ -th powers of integers (like $\tfrac{2}{3}y_1^k+\tfrac{5}{3}y_2^k+\cdots$ ), not necessarily whole-number multiples.

A simple but crucial observation bridges the gap: once you know such a rational-combination representation exists for every sufficiently large integer, clearing denominators (multiplying everything by a fixed common denominator $q$ ) converts it into an honest sum of $k$ -th powers for the multiple $qN$ of every large $N$ — and a short argument then patches up all the small leftover cases. So it is enough to find the rational-combination representation.

\text{if } y=\sum_{i=1}^{r} y_i^k \text{ has a rational-combination form, then } g(k)<\infty

Detailed analysis

The key reduction lemma (Nathanson, Ch. 3; as recounted in course notes following his text, e.g. Baxter, 'Waring's Problem in General') states: if there exist positive rational numbers $a_1,\dots,a_M$ such that every sufficiently large integer $n\ge n_0$ can be written as $n=\sum_{i=1}^M a_iy_i^k$ for non-negative integers $y_1,\dots,y_M$ , then $g(k)$ is finite. The proof is short: let $q$ be the least common denominator of $a_1,\dots,a_M$ ; then $qa_i$ is a positive integer for each $i$ , so $qn=\sum_i(qa_i)y_i^k$ is a sum of $\sum_i qa_i$ non-negative $k$ -th powers, for every $n\ge n_0$ .

Every integer $N\ge qn_0$ can then be written $N=qn+s$ with $n\ge n_0$ and $0\le s\le q-1$ , so $N$ is a sum of $\sum_i qa_i+(q-1)$ $k$ -th powers (the extra $s\le q-1$ contributed by at most $q-1$ copies of $1^k=1$ ). Finitely many remaining small integers $N<qn_0$ are each individually a sum of a bounded number of $k$ -th powers (at worst, $N$ ones), so an overall finite $g(k)$ exists.

This reduces Waring's problem to a purely algebraic task: exhibit, for each $k$ , positive rationals $a_i$ and a way of writing every large integer as $\sum a_iy_i^k$ . The next step supplies exactly this, via Hilbert's central identity.

Terms in this step

Rational-combination representation: Writing an integer $n$ as $\sum_i a_iy_i^k$ where the $a_i$ are fixed positive rational numbers (not necessarily $1$ ) and $y_i$ range over non-negative integers — a weaker, more flexible relative of an honest sum of $k$ -th powers, but one that can be converted into the latter by clearing denominators.