RESEARCH / Focused target
Coefficientwise log-concavity of the two Karp--Zhang series
RELEASED FOUNDATIONS / PROPOSED CONNECTION
Research connections
All source anchors (2)
- D5/S3/Analytic/KarpZhangConjecture/Result.claimAbsent from this graph
- D5/S3/Analytic/KarpZhangConjecture/Result.resultAbsent from this graph
Problem
The question
Dmitrii Karp and Yi Zhang, Log-concavity and log-convexity of series containing multiple Pochhammer symbols, arXiv:2305.09029v3, page 12, Conjecture 5.2:
Suppose that ${y_k}_{k=0}^{\infty}$ is a log-concave sequence. Then the functions $\mu\to\widetilde y(\mu;x)$ and $\mu\to\widehat y(\mu;x)$ are coefficient-wise log-concave on $\mu>0$.
The two series are
$$ \widetilde y(\mu;x)=\sum_{k=0}^{\infty}y_k\frac{(2\mu)_k}{(\mu)k k!}x^k, \qquad \widehat y(\mu;x)=\sum{k=0}^{\infty}y_k\frac{\Gamma(\mu+k)}{\Gamma(2\mu+k)k!}x^k. $$
Here $(a)_k$ is the rising factorial. Definition 1.1, page 2:
Let ${f_k}{k=0}^{\infty}$ be a non-negative nontrivial real sequence. We call ${f_k}{k=0}^{\infty}$ log-concave (or Pólya frequency two – $PF_2$, or doubly positive) if $f_k^2\ge f_{k-1}f_{k+1}$ for each $k\in\mathbb N$, and ${f_k}{k=0}^{\infty}$ is supported on an interval of integers (i.e., $f_N=0$ implies either $f{N+i}=0$ for all $i\in\mathbb N_0$ or $f_{N-i}=0$ for $i=0,\ldots,N$).
The Lean predicate retains nonnegativity, nontriviality, adjacent inequalities for positive natural indices, and the printed zero-support disjunction. Natural subtraction is used only with the stated bounds.
For either series $f$, coefficientwise log-concavity means that every coefficient of
$$ f(\mu+\alpha;x)f(\mu+\beta;x)-f(\mu;x)f(\mu+\alpha+\beta;x) $$
is nonnegative for every $\mu>0$ and every real $\alpha,\beta\ge0$. This is an assertion about finite Cauchy sums in each natural degree, so it requires no convergence or growth assumption on $y$.
Formal correspondence. All Lean names below have prefix D5.S3.Analytic.
| Source object | Lean name | Correspondence |
| --- | --- | --- |
| Definition 1.1 | KarpZhangReciprocal.Result.IsLogConcave | All four sequence conditions, including nontriviality and the exact zero convention. |
| Reciprocal-series coefficient before multiplication by $y_k$ | KarpZhangReciprocal.Coeff.c | $(2\mu)_k/((\mu)_k k!)$. |
| Reciprocal Turanian coefficient | KarpZhangReciprocal.Result.delta | The degree-$m$ finite Cauchy coefficient with both factorials retained. |
| Gamma-series coefficient before multiplication by $y_k$ | KarpZhangGamma.Series.gammaCoeff | $\Gamma(\mu+k)/(\Gamma(2\mu+k)k!)$. |
| Gamma Turanian coefficient | KarpZhangGamma.Series.gammaDelta | The degree-$m$ finite Cauchy coefficient of the gamma-series Turanian. |
| Conjecture 5.2 | KarpZhangConjecture.Result.claim | Both coefficient inequalities for the same $y,\mu,\alpha,\beta,m$, universally quantified. |
| Settlement | KarpZhangConjecture.Result.result | A proof of exactly claim. |
Motivation
Our foothold
The conjecture asks for preservation of coefficientwise log-concavity by two different parameter-dependent transforms of an arbitrary log-concave sequence. Nonnegativity of a series at positive arguments would not suffice: the assertion must hold separately in every coefficient and for arbitrary nonnegative real shifts.
Gap
Missing bridges
Status: Proved (Lean) for both halves of Conjecture 5.2. The result has neither a degree cutoff nor an integrality restriction on either shift. Conjectures 3.8 and 4.2 of the same paper remain open in this development.
Route
Proposed approach
For the reciprocal series, put $c_k(t)=(2t)_k/((t)_k k!)$ and
$$ F_{m,L,U}(s,t)=\sum_{k=L}^{U}c_k(s)c_{m-k}(t). $$
On a line $s+t=S$, normalize the summands to a probability vector $\pi$. The second derivative of the logarithm of their sum is the score variance minus the expected negative logarithmic curvature. Reversible birth-and-death rates on the finite index interval, with rates zero at its endpoints, give a curvature-weighted variance inequality. The coefficient curvature estimate absorbs its edge terms, proving concavity of $\log F_{m,L,U}(s,S-s)$. Only this direction of concavity is needed. For a symmetric interval, symmetry under $s\leftrightarrow t$ then gives
$$ F_{m,q,m-q}(\mu,\mu+\alpha+\beta) \le F_{m,q,m-q}(\mu+\alpha,\mu+\beta). $$
Log-concavity and interval support of $y$ imply that
$W_q=y_qy_{m-q}$ increases towards the central index. Finite Abel summation
expresses the reciprocal Turanian coefficient as $W_0$ times the full tail
plus the increments $W_{q+1}-W_q$ times the successive symmetric tails.
Every factor is nonnegative. This is
KarpZhangReciprocal.Result.reciprocal_half.
For the gamma series, introduce the formal exponential smoothings
$$ P_p(z)=\sum_{r\ge0}y_{p+r}\frac{z^r}{r!}, \qquad H(a;z,w)=\sum_{j\ge0}\frac{P_{2j}(z)}{j!\Gamma(a+j)}w^j. $$
Simultaneous induction in coefficient degree, using Pascal's recurrence, shows that $P_{p+1}P_{q+1}-P_pP_{q+2}$ has nonnegative coefficients when $p\le q+1$. Consequently the coefficient weights in $P_{2k}P_{2(m-k)}$ increase towards the centre. Pair the reciprocal-gamma kernel at $k$ and $m-k$, counting the middle term once. A finite Vandermonde identity and log-convexity of Gamma show that the kernel's total is nonnegative. The Gamma recurrence propagates nonpositivity of the four-term bracket backwards, establishing at most one change from negative to positive. The finite weighting principle therefore proves nonnegativity of every bivariate Turanian coefficient of $H$.
Finally Gamma duplication and a simultaneous recurrence in degree give
$$ \frac{\Gamma(t+n)}{\Gamma(2t+n)n!} =\sqrt\pi,2^{1-2t} \sum_{j=0}^{\lfloor n/2\rfloor} \frac{1}{(n-2j)!j!\Gamma(t+1/2+j)2^{n-2j}16^j}. $$
Thus the gamma series is $\sqrt\pi,2^{1-2t}H(t+1/2;x/2,x^2/16)$.
The scale products at $t+\alpha,t+\beta$ and $t,t+\alpha+\beta$ agree and
are positive. Each substituted coefficient is a finite nonnegative weighted
sum of the bivariate coefficients. This proves
KarpZhangGamma.Series.gamma_half. The settling result combines the two
halves without changing any hypothesis.
Falsifier
What would falsify this route
The correspondence would fail if either coefficient omitted a factorial, reversed the Turanian sign, restricted a shift to integers, or replaced the source's interval-support condition by adjacent inequalities alone. The paired kernel must count its middle index once. The positive diagonal substitution must retain every index satisfying $r+2j=n$. The exact universal claim and both imported halves must compile without proof placeholders or additional axioms.
Evidence
Evidence to collect
The source is arXiv:2305.09029v3, pages 2 and 12, with DOI
10.1007/s13540-023-00238-0; the quotation and sequence convention are recorded
in Library/Asymptotics/karpzhang2023pochhammer.md. The reciprocal proof is in
D5/S3/Analytic/KarpZhangReciprocal/{BirthDeath,IntervalConvolution,Result}.lean;
the gamma proof is in
D5/S3/Analytic/KarpZhangGamma/{Weighting,Smoothing,Result,Series}.lean.
D5/S3/Analytic/KarpZhangConjecture/Result.lean states claim and proves it as
result by the two imported half theorems.
The paper proofs were written with the assistance of codex-cli research seats and checked by independent codex-cli referee seats, the Lean proofs were drafted with the assistance of codex-cli seats, and the work was orchestrated with the assistance of Claude Opus 5.5.
The settlement is associated with preregistration
https://github.com/the-omega-institute/trureturing/issues/20997.
The scoped Lean build of D5.S3.Analytic.KarpZhangConjecture.Result verifies
the settling theorem and its import closure. Its axiom closure is contained
in ${\texttt{propext},\texttt{Classical.choice},\texttt{Quot.sound}}$.
This local verification is distinct from publication, freezing and required CI.
Triage
Scope assessment
What the settlement shows
The reciprocal mechanism is variance absorption for interval convolutions, followed by symmetric-tail balancing and Abel summation. The gamma mechanism is preservation of balanced minors under exponential smoothing, a single negative-to-positive sign change in the reciprocal-gamma kernel, and a positive diagonal transformation. Both mechanisms are uniform in coefficient degree and allow arbitrary nonnegative real shifts, including zero. The conclusions concern formal coefficients; analytic evaluation additionally requires the relevant convergence hypotheses. Conjecture 5.2 can therefore be used as a proved coefficient inequality under precisely its printed sequence conditions. This supplies neither a new hypothesis nor a proof of the paper's other conjectures.
- PROVED (Lean,
KarpZhangReciprocal.Result.reciprocal_half): Every reciprocal-series Turanian coefficient is nonnegative for every admissible sequence, positive base parameter and nonnegative real shifts. - PROVED (Lean,
KarpZhangGamma.Series.gamma_half): Every gamma-series Turanian coefficient is nonnegative under the same conditions. - PROVED (Lean,
KarpZhangConjecture.Result.result): The conjunction of both universal coefficient assertions, exactlyclaim. - OPEN: Conjectures 3.8 and 4.2 of the same paper. Numerical support is reported by the research line; neither conjecture is proved here. The numerical scope has not been independently revalidated in this delivery.
ASSUMED-UNVERIFIED
Unverified assumptions
Exhaustive literature coverage, citation-index completeness and priority are
unverified. Numerical support for Conjectures 3.8 and 4.2 is supplied by the
research line; no numerical experiment or uniform extension is certified by
this delivery. The producer attribution above is supplied by the coordinating
research line. Four-slot escape registrations for the public theorems of the
seven supporting modules and for the settling result are not completed in
this delivery; their mathematical proofs do not constitute those registrations.