Source. Wouter van Doorn and GPT-6 Astra Pro (the author line as
printed), Practical numbers and Egyptian fractions, Lemma 3.2 with its
proof and the definition of Md(X), physical p. 3 of the seven-page PDF
held by
van Doorn (2026).
The displays were read on the page image, since the text extraction garbles
them. Consumed by
the Lemma 3.3 reconstruction.
The note describes itself as a simplified and explicit version of the
bound of
the Price claim
(abstract and Section 1, physical p. 1; Section 2, p. 2) and does not
say which step of that argument this lemma replaces; the description of
the Price claim's analytic input as an exponential-sum theorem comes from
the site comment recorded on the Price card, not from the note.
Standing. Author-recorded reconstruction of a claimed result (see
the Lemma 3.1 page for
the note's standing); not an independent review; changes no status and
assigns no tier. The proof uses only Cauchy–Schwarz, Plancherel's identity
and character orthogonality on Z/dZ, all written out
below.
Definitions
D(n) is the set of positive divisors of n, and eq(z)=exp(2πiz/q).
For a finite nonempty set X of integers and an integer d≥1,
Fix a divisor d of A with d>1; d is odd. Three estimates are
established for fd, then combined.
Step 1: the pointwise bound at primitive frequencies. Since V1 and
V2 are coprime, every divisor of V factors uniquely as a divisor of
V1 times a divisor of V2, so multiplication is a bijection
X1×X2→X and ∣X∣=∣X1∣∣X2∣. Grouping the elements of X1
by residue class a modulo d, with N1(a)=∣{x∈X1:x≡a}∣,
since ∑aN1(a)2=∣X1∣2Md(X1). Expanding the square and summing over
a first, character orthogonality gives
∑amodded(ξa(y−y′))=d if d∣ξ(y−y′) and 0 otherwise.
When (ξ,d)=1 the condition is y≡y′(modd), so the inner sum is
d∣X2∣2Md(X2) and
∣fd(ξ)∣2≤dMd(X1)Md(X2)((ξ,d)=1).
Step 2: Plancherel. Expanding ∣fd(ξ)∣2 and summing over all
ξ modulo d, orthogonality gives
Step 3: the four-fold product. Since d is odd, multiplication by 2ℓ
permutes the residues modulo d and permutes the residues coprime to d.
Hence for (ξ,d)=1 both ξ and 2ξ are coprime to d, and Step 1
bounds ∣fd(ξ)∣∣fd(2ξ)∣≤dMd(X1)Md(X2); and for
ℓ∈{2,3}, ∑ξmodd∣fd(2ℓξ)∣2=dMd(X) by Step 2.
Bounding the first two factors pointwise, extending the sum to all ξ,
and applying Cauchy–Schwarz to the last two,
Step 4: orthogonality modulo A. Suppose that some residue c has no
representation (3.2). The number of quadruples
(z0,z1,z2,z3)∈X4 with z0+2z1+4z2+8z3≡c(modA) equals
and by assumption it is 0. The term h=0 equals 1. Every nonzero h
modulo A is uniquely h=(A/d)ξ with d=A/gcd(h,A)>1 a divisor of A
and 1≤ξ<d, (ξ,d)=1; then eA(hz)=ed(ξz) for every z, so
fA(2ℓh)=fd(2ℓξ). Moving the h=0 term to the left, taking
absolute values, and grouping by d,
by Step 3. Each term on the right is the cube of the corresponding
nonnegative term d2/3(Md(X1)Md(X2)Md(X))1/3 of S, and a sum of
cubes of nonnegative reals is at most the cube of their sum, so the right
side is at most S3<1. This contradiction proves the lemma.