Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.
Updated
Arithmetic characterizations of Sidon sets
Gilles Pisier, “Arithmetic characterizations of Sidon sets,” Bulletin of the American Mathematical Society (New Series) 8 (1983), no. 1, 87–89.
Scope and notation
Let be a compact abelian group and let be its discrete dual. For , Pisier writes for the finitely supported relations
The larger set consists of all finitely supported -families, whether or not they give a zero relation. For , counts the representations coming from , and counts those with .
A set is quasi-independent when : its only relation is the zero relation. It is a Rider set when, for some , . A Sidon set is a set for which there is such that every trigonometric polynomial with Fourier support in satisfies
The least such is denoted . These definitions are on pp. 87–88 (the article's first two pages).
For , quasi-independence is exactly dissociation in the sense of E0774. Indeed, a nonzero signed relation partitions its support into two distinct finite subsets with equal sums. Conversely, equality of two distinct subset sums, after cancelling their intersection, gives a nonzero signed relation.
Main arithmetic characterizations
Theorem 1 (p. 88; the article's second page). Suppose . The following are equivalent:
- (i) is Sidon.
- (ii) For some , every finite satisfies
- (iii) For some , every finite satisfies
- (iv) For some , every finite satisfies
The accompanying proposition (p. 88; the article's second page) says these conditions are also equivalent to two analytic/metric formulations. First, there are and such that, for every finite ,
Second, there is such that for every such there are points with
The paper says the equivalence of these last two formulations is formal. It also notes that the equivalence of the two representation-count bounds follows easily from .
Theorem 2 (p. 89; the article's third page). “A subset of is a Sidon set iff (vii) there is an integer such that any finite subset of contains a quasi-independent subset with .”
Thus, for , “proportionately dissociated” is equivalent to “Sidon.” Pisier says that Theorem 2, in some sense, reduces the outstanding finite-union problem to “a purely combinatorial question: Is every set satisfying (vii) a finite union of quasi-independent sets?” (p. 89). Under the integer translation above, that is precisely E0774.
Proof ideas useful for E0774
- A union of quasi-independent sets has the proportional extraction property immediately: one color class contains at least elements of every finite . E0774 asks for the converse.
- The identity displayed as (1) on p. 88 expands in terms of the weighted counts . Pisier says the Sidon-to-relation-count implication uses this identity at together with integrability properties of .
Exact limits of this source
- This three-page article is an announcement. It sends the details of Theorem 1, the proposition, and the difficult direction of Theorem 2 to Pisier’s reference [5], then listed as forthcoming. It does not contain those proofs.
- The converse direction in Theorem 2 is attributed to Theorem 2.3 of Pisier’s 1981 paper, using the fact that every quasi-independent set has Sidon constant bounded by an absolute constant.
- The paper states that every Rider set is a finite union of quasi-independent sets, again referring to [5] rather than proving it here.
- No explicit dependence of , , , or on the Sidon constant is given. The equivalences are qualitative and uniform over finite subsets.
- Pisier does not answer the finite-union question and provides no integer counterexample, block construction, or encoding scheme. The positive results it mentions, for with prime (its reference [3]) or a product of distinct primes (Bourgain, private communication), do not cover .