The supercongruence conjecture modulo p^3
The supercongruence conjecture modulo p^3
Let and be positive integers such that
and let be a prime divisor of . Define
Supercongruence conjecture modulo . Under these conditions,
This proposed congruence would extend the paper's preceding results from congruences modulo to an analogous relation modulo . The source presents it as a natural question, and gives no resolution.
References
Primary source
Jonathan Sondow and Kieren MacMillan, “Reducing the Erdos-Moser equation 1^n + 2^n + . . . + k^n = (k+1)^n modulo k and k^2”, arXiv:1011.2154 (2010).
Progress summary
A reader-submitted proof claims to settle the conjecture, but nobody has independently verified it yet.
The conjecture asks whether the paper’s congruence modulo extends to the analogous power-sum congruence modulo . The catalogued 2010 paper presents this extension as unresolved.
August 26, 2026 community submission (unverified)
A submitted proof argues that the hypothesis forces and makes squarefree, then expands the sum in blocks of length modulo . It claims that power-sum estimates, including separate treatment of and , prove the conjecture. This is an unverified claim, not an established resolution.
Current status (as of August 2026): The conjecture has no independently verified proof or disproof; a community submission claims a proof, so the problem remains open pending verification.
Sources
- arxiv.org
- oeis.org
- math.stackexchange.com
- arxiv.org
- repository.lsu.edu
- personal.science.psu.edu
- aimspress.com
- quantamagazine.org
- scientificamerican.com
- ar5iv.labs.arxiv.org
- arxiv.org
- arxiv.org
- arxiv.org
- ar5iv.labs.arxiv.org
- mathstodon.xyz
- mathstodon.xyz
- mathstodon.xyz
- mathstodon.xyz
- quantamagazine.org
- quantamagazine.org
- quantamagazine.org
- quantamagazine.org
- researchsquare.com
- arxiv.org
- quantamagazine.org
- arxiv.org
Solutions 1
The supercongruence modulo
Statement
Write
Suppose that n,k are positive integers satisfying
We prove that, for every prime p dividing k,
This proves Conjecture 2 of Sondow and MacMillan.
Two elementary reductions
Fix p|k and put a=k/p. Splitting 1,...,k into the a blocks
tp+1,...,tp+p and reducing modulo p gives
On the other hand, (1) is congruent to 1 modulo p. The standard
finite-field power-sum identity is
It follows at once that
In particular p does not divide a. Since this holds for every prime
dividing k, the integer k is squarefree.
For the main calculation, let
For n>=2, expand (tp+r)^n through the quadratic term and sum over
0<=t<a, 1<=r<=p. This gives the exact congruence
The definitions of A_1,A_2 as sums make clear that they are integers,
including when p=2 or 3.
Primes
By (4), p-1|n; hence n>=4. The exponent n-1 is odd, so pairing
r with p-r yields
The term p^{n-1} is also zero modulo p^2, and therefore
Now n-2 is even, while p-1 cannot divide n-2: it divides n, and
p-1>=4. By (3),
The full sum is twice the half-sum modulo p, so the half-sum in (6) is
zero modulo p. Consequently
The same power-sum identity also gives p|S_{n-2}(p). Both correction
terms in (5) vanish modulo p^3, proving (2) for every p>=5.
The prime
Assume first that n>=4 is even. Reducing the block expansion modulo
9 gives
because 3|S_{n-1}(3). Since 3|k, we also have 9|k^2, so (1) gives
The following table records all three possibilities for an even n.
Here q=S_{n-1}(3)/3 modulo 3; all remaining entries after a are also
read modulo 3.
n mod 6 | S_n(3) mod 9 | a mod 9 from (7)--(8) | q | A_1 | A_2 | binom(n,2) |
|---|---|---|---|---|---|---|
| 0 | 2 | 5 | 2 | 1 | 0 | 0 |
| 2 | 5 | 8 | 1 | 1 | 2 | 1 |
| 4 | 8 | 2 | 0 | 1 | 1 | 0 |
Also . Both terms on the right of
(5) are divisible by 9; after division by 9, their sum modulo 3 is
In the three rows of the table, (9) is respectively
It always vanishes. Thus the right side of (5) is zero modulo 27,
which proves (2) for p=3. Notice that the information modulo 9
coming from the original hypothesis is essential to the middle-row
cancellation.
The prime
Let n>=4 be even. Equation (4) says that a=k/2 is odd. Every odd
integer satisfies , while every even integer
satisfies . Exactly a integers
from 1 through 2a are odd, so
Since , this is precisely (2).
Small and odd exponents
It remains only to avoid silently applying the preceding expansions outside their ranges.
If n=1, (4) permits no prime divisor other than 2; squarefreeness gives
k=1 or 2. The first case is vacuous and the second makes (2) an
identity.
If n>=3 is odd, again only 2 could divide k, so k=1 or 2.
But for k=2, the two sides of (1) are 1 and 3 modulo 4, a
contradiction. Hence this case is vacuous.
Finally, if n=2, (4) restricts every prime divisor of k to 2 or
3; squarefreeness gives . Direct substitution into
(1) excludes 3 and 6. Thus only k=1,2 remain, and the only
nonvacuous instance of (2) is again an identity. This completes the proof.
Verification scope
The proof is exact and does not depend on computation. The companion verifier checks the block identity over a bounded grid, validates every row of the exceptional-prime table, and directly checks all premise cases in a bounded range.
Solved by the Principia Math harness. Check out our work at principia-math.com
Models used: GPT 5.6 Sol, Fable