Conjecture on the integer part of non-perfect-power roots

From papers

Let a,nZ+a,n\in\mathbb{Z}^+ satisfy a>2a>2, log2(a)+1n>1\left\lfloor\log_2(a)\right\rfloor+1\geq n>1, and suppose there is no kZ+k\in\mathbb{Z}^+ such that kn=ak^n=a. Integer-part root conjecture.

an=(a2an+1)2an+1mod(a2an2a)(a2an+1)2anmod(a2an2a)1.\left\lfloor\sqrt[n]{a}\right\rfloor=\left\lfloor\frac{(a^{2an}+1)^{2an+1}\bmod(a^{2an^2}-a)}{(a^{2an}+1)^{2an}\bmod(a^{2an^2}-a)}-1\right\rfloor.

This gives a proposed elementary formula for the integer part of an nn-th root using polynomial quotient-ring reductions. The surrounding discussion indicates that the formula is motivated by convergence observations, while a deeper analysis of the convergence properties is left for future work; no resolution is supplied here.

Progress summary

Open

No public discussion or published progress on this conjecture was found.

No public discussion or published progress was found for the proposed formula.

Current status (as of August 2026): The conjecture appears open, with no recorded public activity.

Sources & referencesView supporting material

Primary source

Joseph M. Shunia, “Polynomial quotient rings and Kronecker substitution for deriving combinatorial identities”, arXiv:2404.00332 (2024).

Solutions 1

Proof

Let a>2 and 1<n≤⌊log₂a⌋+1, with a not an nth power. Put α=a^{1/n}, k=2an, b=a^k, M=b^n−a, q=1+α. We prove ⌊α⌋=⌊(((b+1)^{k+1} mod M)/((b+1)^k mod M))−1⌋.

Define P_j(X)=(1+X)^j mod (X^n−a)=Σ_{r=0}^{n−1}c_{j,r}X^r, c_{j,r}=Σ_{h≥0}C(j,r+hn)a^h. These coefficients are nonnegative and Σ_r c_{j,r}≤P_j(α)=q^j. In fact q^{k+1}<b: if a≥4, then q≤1+√a≤3a/4, k≥4a, and (4/3)^{4a}≥1+4a/3>a; the sole remaining case a=3,n=2 follows from 11^13<3^12·4^13 and √3<7/4. Hence for j=k,k+1, 0<P_j(b)≤b^{n−1}(b−1)<b^n−a=M. Since b^n≡a mod M, both modular residues are therefore exactly P_j(b).

Let ζ=e^{2πi/n}. Roots-of-unity filtering gives c_{k,r}=q^k/(nα^r)·(1+ε_r), ε_r=Σ_{s=1}^{n−1}ζ^{−rs}((1+αζ^s)/(1+α))^k. Set ρ²=1−[4α/(1+α)²]sin²(π/n), E=(n−1)ρ^k. Then |ε_r|≤E. Multiplication by 1+X modulo X^n−a gives c_{k+1,0}=c_{k,0}+ac_{k,n−1} and c_{k+1,r}=c_{k,r}+c_{k,r−1} for r≥1. With w_r=(b/α)^r, P_{k+1}(b)/P_k(b)−1 =α[Σ_rw_r(1+ε_{r−1 mod n})]/[Σ_rw_r(1+ε_r)]. Consequently, whenever E<1, |P_{k+1}(b)/P_k(b)−1−α|≤2αE/(1−E).

We now establish uniformly E<1/[4αn(1+α)^{n−1}]. (*) For n=2, ρ=(α−1)/(α+1), k=4α², and log((α+1)/(α−1))≥2/α, so E≤e^{−8α}. Since α≥√3>3/2, monotonicity gives 8α(1+α)E<30e^{−12}<1.

For n≥3, a≥2^{n−1}, hence α≥2^{(n−1)/n}. Using sin(π/n)≥3√3/(2n), (1+α)²<3α², and log(1−t)≤−t gives ρ^k<exp(−9α^{n−1}/n). It suffices to show H_n(x)=4n(n−1)x(1+x)^{n−1}exp(−9x^{n−1}/n)<1 for x≥2^{(n−1)/n}. Differentiation shows H_n is decreasing there. For n=3 its endpoint is bounded by 432e^{−7}<1. For n≥4 the endpoint is bounded by K_n=8n(n−1)3^{n−1}exp(−9·2^{n−2}/n). Here K_4=2592e^{−9}<1 and K_{n+1}/K_n=[3(n+1)/(n−1)]exp(−9·2^{n−2}(n−1)/(n(n+1)))<5e^{−5}<1. This proves (*).

Finally let t=⌊α⌋. Since a is not an nth power, both a−t^n and (t+1)^n−a are positive integers. Factoring differences of powers gives min(α−t,t+1−α)≥δ=1/[n(1+α)^{n−1}]. By (*), E<δ/(4α)<1/4, so |P_{k+1}(b)/P_k(b)−1−α|<2δ/3<δ. Thus P_{k+1}(b)/P_k(b)−1 lies in the same open unit interval (t,t+1) as α. Taking floors and substituting the exact modular residues proves the formula throughout the full conjectured parameter range.

0 endorsements
Shivam Patel ·