The two congruences for DSOME(n)
For a nonnegative integer n, let DSOME(n) be the sum of the odd parts minus the sum of the even parts, taken over all partitions of n into distinct parts. We prove Baruah and Gogoi's Conjecture 4.1: for every n≥0,
DSOME(50n+21)≡0(mod8),DSOME(100n+71)≡0(mod16).
The first congruence was proved by Bardhan and Saikia in Theorem 4.11 and Remark 4.12. We prove the second. We will also use their quadratic-representation involution, explicitly given below. The additional step is a coefficientwise theta-series calculation modulo 16; two parity decompositions then finish the argument.
1. A weighted pentagonal identity
All series in the proof are formal power series with integer coefficients. Put
f=r≥1∏(1−qr),X=m≥1∑(−1)mqm2,Xd=X(qd),D=qdqd.
Thus X1=X. Write ϕ=1+2X. The generating-function identity in Baruah and Gogoi, Theorem 1.1, together with ϕ=∏r≥1(1−qr)2/∏r≥1(1−q2r), gives
F(q):=n≥0∑DSOME(n)qn=8f(ϕ−1−ϕ3).(1)
Set Pj=j(3j−1)/2 for j∈Z. We use the classical identities recorded as (2.2) and (4.54) in Bardhan and Saikia:
f=j∈Z∑(−1)jqPj,fϕ2=j∈Z∑(1−6j)qPj.(2)
Define U=X(1+X) and Y=fU. Since ϕ2=1+4U, (2) implies
Y=j∈Z∑(−1)jvjqPj,v2k=−3k,v2k+1=3k+1.(3)
A direct calculation, valid for every integer j, gives
3Pj=2vj2+vj.
Consequently vj≡Pj(mod4). Substituting this into
3(Pj−vj)=2vj(vj−1) modulo 8 gives
vj≡2Pj2−Pj(mod8). Therefore
Y≡(2D2−D)f(mod8),Y≡Df(mod4).(4)
In particular, both f and Y have zero coefficient at every index not of the form Pj.
2. The coefficient calculation modulo 16
Fix an integer N≥0 such that N≡3(mod4) and N=Pj for every j∈Z. For a series H, write ⟨H⟩=[qN]H, and put
ad=⟨fXd⟩(d=1,2,4,8,16).
We claim that
DSOME(N)≡−2Na1−12a4−8(a2+a8+a16)(mod16).(5)
Here are the details. Factoring the numerator of (1) gives the exact identity
F=−Y(1+1+2X2X2).
Since X has zero constant term, expansion of the inverse is valid formally. With
Z=X2−2X3+4X4,
we obtain F≡−Y−2YZ(mod16). Our assumption on N gives ⟨Y⟩=0, so
DSOME(N)≡−2⟨YZ⟩(mod16).(6)
For any series H,
⟨(Df)H⟩⟨(D2f)H⟩=N⟨fH⟩−⟨fDH⟩,=N2⟨fH⟩−2N⟨fDH⟩+⟨fD2H⟩.
Using (4), then N≡3(mod4), yields
⟨YZ⟩≡(2−N)⟨fZ⟩+5⟨fDZ⟩+2⟨fD2Z⟩(mod8).(7)
The following congruences follow coefficientwise by separating odd and even m in the definition of X:
DXD2X(DX)2≡X+X4−2X16(mod4),≡X+X4(mod2),≡X2+X8(mod2).(8)
For the last line, square the preceding parity expression and use Xd2≡X2d(mod2).
Differentiating Z gives
DZ2D2Z≡2UDX(mod8),≡4(DX)2+4UD2X(mod8).
Substitution into (7), followed by (8), gives
⟨YZ⟩≡⟨f(−NX2+4X4+6UX4−4UX16+4X2+4X8)⟩(mod8).(9)
Indeed, before using N≡3(mod4), the coefficient of X3 in this simplification is 2N+2, which vanishes modulo 8.
To simplify (9), use ⟨fX2⟩=−a1, which follows from ⟨Y⟩=0, and X4≡X4(mod2). Also (4) and DXd≡0(modd) give
⟨YX4⟩≡Na4(mod4),⟨YX16⟩≡Na16(mod2).
It follows that
⟨YZ⟩≡Na1+(4+6N)a4−4Na16+4a2+4a8≡Na1+6a4+4(a2+a8+a16)(mod8).
Together with (6), this proves (5).
3. The progression 100n+71
Now let N=100n+71 with n≥0, and put
M=24N+1=5(480n+341).
Thus the exponent of 5 in M is exactly one. In particular M is not a square, so N is not pentagonal, since 24Pj+1=(6j−1)2. Also N≡3(mod4), and (5) applies.
By (2), the sums ad have the explicit form
ad=j∈Z, m≥1Pj+dm2=N∑(−1)j+m.(10)
For d=2, a summand would give M=(6j−1)2+48m2. Reduction modulo 5 forces both 6j−1 and m to be divisible by 5, because 2 is a quadratic nonresidue modulo 5. This contradicts the exponent of 5 in M. The same argument for d=8 uses the nonresidue 3. Hence
a2=a8=0.(11)
Consider the finite set
R={(j,m)∈Z×Z>0:Pj+m2=N}.
We use the involution from the proof of Bardhan and Saikia's Theorem 4.11. For (j,m)∈R, put u=6j−1. Then u2+24m2=M. Neither u nor m is divisible by 5, and there is a unique s∈{1,−1} such that 5∣u+sm. Define
u′=524sm−u,m′=5∣su+m∣,j′=6u′+1.(12)
Both u′ and m′ are integers, u′≡−1(mod6), and m′>0: equality su+m=0 would make M=25m2. Moreover,
(u′)2+24(m′)2=M,
so (j′,m′)∈R. To check the inverse explicitly, let ε be the sign of su+m and put s′=sε. Then
s′m′=5u+sm,u′+s′m′=5sm.
Thus s′ is the unique sign selected for (u′,m′), and applying (12) again returns u and m.
Furthermore,
m′≡m+1(mod2),j′=54sm−j+1≡j+1(mod2).
The involution therefore exchanges even and odd m and preserves the sign (−1)j+m.
For d=2,4,8, set
Ed=(j,m)∈Rd∣m∑(−1)j.
Pairing by (12) gives a1=2E2. In addition, simply separating even and odd values of m/2 and m/4 gives
E2+a4=2E4,E4+a16=2E8.(13)
Using N≡3(mod4) and (13), we obtain
NE2+3a4+2a16≡3(E2+a4)+2a16=6E4+2a16≡2(E4+a16)=4E8≡0(mod4).
Finally, (5), (11) and a1=2E2 imply
DSOME(N)≡−4(NE2+3a4+2a16)≡0(mod16).
This proves the second congruence for every n≥0. Together with the previously proved modulo-8 congruence, it establishes all of Conjecture 4.1.