Alphabeta Math
LemmaStatement: Literature-sourcedProof: AI-adaptedPipeline-generatedprecheck passjudge pass (gpt-6.1-sol)
How statement and proof provenance work

The first chip identifies the source of the statement or construction; the second identifies the source of its local proof or verification.

  • Literature-sourced: the exact statement appears in a cited source; only wording and notation differ.
  • AI-adapted: a semantically identical restatement of literature-sourced material, modulo indexing, notation, and boundary cases adopted by the library.
  • AI-generated: a genuinely novel statement formulated by AI, with no source for the claim itself.

These labels describe origin, not correctness: citations and verification chips remain separate evidence.

The unmodified integral torus has a quantum-binomial basis

Statement

Let (I,A,D,P,P∨,q) be a finite symmetrizable Cartan datum (Symmetrizable Cartan data for quantum groups), put A=Z[q,q−1] (The Laurent polynomial ring as the principal localisation of Z[t] at t), let L:=∑i∈IZ(dihi)⊆P∨, and let UA0 be the A-subalgebra of Uq0 generated by the Kh±1 and the toral quantum binomials Ti,c,t (The Lusztig divided-power integral form UA).

(a) One simple torus. Fix i∈I and write z:=qi, K:=Ki, Tt:=Ti,0,t for t≥0. Then for all r,s≥0, c∈Z and t≥1, TrTs=∑j=0min⁡(r,s)zrsK−j(r+s−jr)z(rj)zTr+s−j,Tc+1,t=ztTc,t+K−1z−c+t−1Tc,t−1, where (mr)z is the Gaussian binomial of Quantum integers, factorials, Gaussian binomials and divided powers at qi; all coefficients lie in A. Moreover the A-subalgebra of Uq0 generated by K±1 and all Tc,t equals the A-span of {KδTt:δ∈{0,1}, t≥0}, and that set is an A-basis.

(b) Full torus. Fix a Z-basis of P∨, let J be the finite set of coordinates occurring in the finitely many vectors dihi, and let D:=⨁j∈JZej be the corresponding coordinate direct summand. For every coset of L in P∨ choose the representative c whose D-component has minimal sum of coordinate absolute values, ties broken lexicographically, and whose complementary component is the coset's common complementary component. Then UA0=⨁cKc⋅A-span{∏i∈IKiδiTi,0,ti:δi∈{0,1}, ti≥0}, the sum over the chosen representatives, so the displayed elements form an A-basis of UA0.

Both statements are proved directly on the unmodified generators over the non-PID ring A: no integral PBW or canonical-basis input, no inference from torsion-freeness to freeness, and no modified-AU˙ result is used.

Facts & Assumptions

Given: The Cartan datum with finitely many indices, the Laurent ring A, the subgroup L=∑iZ(dihi)⊆P∨, and the toral binomials Ti,c,t=∏s=1t(Kiqic−s+1−Ki−1qi−c+s−1)/(qis−qi−s) for i∈I, c∈Z, t≥0.

[F1]

A=Z[q,q−1] is a subring of Q(q) containing z±m and (mr)z for all m,r≥0: the Gaussian binomials are Laurent polynomials in qi with integer coefficients (The Laurent polynomial ring as the principal localisation of Z[t] at t, The quantum Pascal recurrences, the Gauss product formula and Gaussian integrality (iv)), and qi±1∈A (Quantum integers, factorials, Gaussian binomials and divided powers at qi).

[F2]

Each Ti,c,t is a Laurent polynomial in Ki±1 with coefficients in Q(q), all Ti,c,t commute with one another and with every Kh, Ti,c,0=1, Ti,c,1=(Kiqic−Ki−1qi−c)/(qi−qi−1), and Ti,c,t=Ti,c,t−1(Kiqic−t+1−Ki−1qi−c+t−1)/(qit−qi−t) (The Lusztig divided-power integral form UA, Divided-power multiplication, toral binomials and the Ei(a)Fi(b) reordering identity (ii)).

[F3]

Uq0≅Q(q)[P∨] as algebras, with KhKh′=Kh+h′ and K0=1; the Kh for h in a Z-basis of P∨ are algebraically independent, and Uq0 is commutative (Triangular decomposition of a quantized enveloping algebra (ii)).

[F4]

The simple coroots hi are linearly independent over Q in P∨⊗ZQ, and all di>0; hence L=⨁iZ(dihi)≅Zn and each dihi has finite support in the fixed basis of P∨, so the set J of coordinates occurring in the dihi is finite and L⊆D (Symmetrizable Cartan data for quantum groups, Free abelian group on a set).

[F5]

For every nonzero integer t, qt≠1: after replacing t by ∣t∣, equality would make q a root of T∣t∣−1. Thus zs−z−s≠0 for s≥1 (Quantum integers, factorials, Gaussian binomials and divided powers at qi).

Verification

technique · evaluate both sides of the product identity at the infinitely many toral points $K=z^n$, using the Gaussian Vandermonde identity proved from the weighted-subset model of the ordinary Gaussian binomial; deduce the shift and stability identities by direct factorization; then span all shifts by the stable pair $\{K^{\delta}T_t\}$ and read off independence from the extremal exponents
1.1F2F5algebra

The simple torus and its evaluation. Fix i and write z=qi, K=Ki, Tt=Ti,0,t. By [F2], Tt=∏s=1t(Kz1−s−K−1zs−1)/(zs−z−s), a Laurent polynomial in K whose support is contained in [−t,t]: indeed each factor equals Kz1−s(1−K−2z2s−2), so Tt=Kt∏s=1tz1−s(1−K−2z2s−2)/∏s=1t(zs−z−s), a Q(q)-combination of Kt−2l for 0≤l≤t, and the two extreme coefficients (of Kt and of K−t) are nonzero by [F5]. Evaluating at K=zn for an integer n≥0 gives Tt(zn)=∏s=1t[n−s+1]z/[s]z with [m]z=(zm−z−m)/(z−z−1), which equals the Gaussian binomial (nt)z when t≤n, and equals 0 when t>n (the factor s=n+1 has numerator z0−z0=0); accordingly we use the convention (nt)z=0 for t>n or t<0. Finally, if a Laurent polynomial in K with coefficients in Q(q) vanishes at K=zn for infinitely many n, it is zero: after multiplying by a power of K it becomes a polynomial whose infinitely many pairwise distinct roots zn force all coefficients to vanish (the zn are distinct because q is transcendental by [F5]).

1.2F1algebra

The Gaussian Vandermonde. For an indeterminate Q and m,r≥0 put [m]Q:=1+Q+⋯+Qm−1 and B(m,r):=[m]Q!/([r]Q![m−r]Q!), so that with Q=z2 one has B(m,r)=(mr)z2 in the notation of [F1]. We first record the weighted-subset model: if inv(S):=#{(a,b):a<b, a∉S, b∈S} for S⊆{1,…,m}, then B(m,r)=∑S⊆{1,…,m}, ∣S∣=rQinv(S). Indeed both sides equal 1 for m=r=0 and 0 for r<0 or r>m; splitting the subsets according to whether m lies in S gives the recurrence B(m,r)=B(m−1,r)+Qm−rB(m−1,r−1) for the right-hand side, and the same recurrence holds for the left-hand side because B(m−1,r)+Qm−rB(m−1,r−1)=[m−1]Q![r]Q![m−r]Q!([m−r]Q+Qm−r[r]Q)=[m−1]Q![m]Q[r]Q![m−r]Q!=B(m,r), using [m−r]Q+Qm−r[r]Q=[m]Q. Splitting a subset S⊆{1,…,n} with ∣S∣=s into A:=S∩{1,…,r}, ∣A∣=j, and B′:=S∩{r+1,…,n}, the inversion count splits as inv(S)=inv(A)+inv(B′)+(r−j)(s−j), because a crossing inversion a<b with a∈{1,…,r}∖A, b∈B′ can always be realized, while pairs (b,a) with b∈{r+1,…,n}, a≤r are never inversions; hence for 0≤r≤n, B(n,s)=∑j=0min⁡(r,s)Q(r−j)(s−j)B(r,j)B(n−r,s−j).

1.3F1F2algebra

The shift and stability identities. Put Ns(c)=Kzc−s+1−K−1z−c+s−1. Since Ns(c+1)=Ns−1(c) for s≥2, factoring the first t−1 common factors reduces the shift to N1(c+1)−ztNt(c)=K−1z−c+t−1(zt−z−t), proving Tc+1,t=ztTc,t+K−1z−c+t−1Tc,t−1. The product definition also gives (zt+1−z−t−1)Tt+1=Tt(Kz−t−K−1zt). Rearranging yields K−1Tt=z−2tKTt−z−t(zt+1−z−t−1)Tt+1; multiplying by Kz2t yields K2Tt=z2tTt+zt(zt+1−z−t−1)KTt+1. Every coefficient is in A.

2.1step 1.2F1algebra

Normalization to the symmetric Gaussian binomials. By the identification of Quantum integers, factorials, Gaussian binomials and divided powers at qi, (mr)z=z−r(m−r)(mr)z2=z−r(m−r)B(m,r) for 0≤r≤m. First multiply step 1.2 by B(n,r) and use the factorial identity B(n,r)B(n−r,s−j)=B(n,r+s−j)B(r+s−j,r). Then multiply by z−r(n−r)z−s(n−s) and substitute B(m,r)=zr(m−r)(mr)z in each term, the powers of z on the right-hand side combine to 2(r−j)(s−j)+(r+s−j)(n−r−s+j)+r(s−j)+j(r−j)−r(n−r)−s(n−s)=rs−jn, as a direct expansion shows; in the resulting identity (nr)z(ns)z=∑j=0min⁡(r,s)zrs−jn(nr+s−j)z(r+s−jr)z(rj)z both sides are zero when n<r or n<s (then r+s−j≥max⁡(r,s)>n for every j, so the factor (nr+s−j)z vanishes), and the identity therefore holds for all n,r,s≥0 under the convention of step 1.1.

3.1step 1.1step 2.1F1F2algebra

The product identity. With notation as in step 1.1, we claim TrTs=∑j=0min⁡(r,s)zrsK−j(r+s−jr)z(rj)zTr+s−j for all r,s≥0. Both sides lie in Q(q)[K±1]: the left side by [F2], the right side because the coefficients are in A⊆Q(q) and Tr+s−j is a Laurent polynomial in K [F2]. Evaluating at K=zn and using step 1.1, the left side becomes (nr)z(ns)z and the right side becomes ∑jzrs−jn(r+s−jr)z(rj)z(nr+s−j)z, and these agree for every n≥0 by step 2.1. By the evaluation principle of step 1.1 the two Laurent polynomials coincide, proving the identity; its coefficients lie in A by [F1].

4.1step 3.1step 1.3F1F2algebra

Spanning and the simple-torus basis. Let M:=A-span{KδTt:δ∈{0,1}, t≥0}. First, KmTt∈M for all m∈Z, t≥0: for m≥2, multiply the second identity of step 1.3(b) by Km−2 to get KmTt=z2tKm−2Tt+zt(zt+1−z−t−1)Km−1Tt+1, which reduces ∣m∣ and lies in M by induction on ∣m∣ from the base m=0,1; for m=−1 use the first identity of step 1.3(b) directly; for m≤−2, multiply it by Km+1 to get KmTt=z−2tKm+2Tt−z−t(zt+1−z−t−1)Km+1Tt+1, again reducing ∣m∣. Second, all Tc,t∈M by induction on t (the case t=0 gives Tc,0=1): for c≥0 the shift identity Tc+1,t=ztTc,t+K−1z−c+t−1Tc,t−1 of step 1.3 (with unit coefficient zt∈A×) iterates upward from T0,t=Tt∈M, and for c≤−1 its other solution Tc,t=z−tTc+1,t−K−1z−c−1Tc,t−1 iterates downward from c=0. Third, M is an A-subalgebra: (KmTr)(KnTs)=Km+nTrTs expands by step 3.1 into the terms zrs(r+s−jr)z(rj)zKm+n−jTr+s−j, which lie in M by the first part. Since M contains K±1 and every Tc,t, it contains the generated subalgebra; conversely KδTt is a product of the generators K and T0,t, so M equals the A-subalgebra generated by K±1 and all Tc,t, which is the simple-torus part of UA0. Fourth, {KδTt} is A-independent: if a finite combination with coefficients (aδ,t) is zero and not all coefficients vanish, let N:=max⁡{t:a0,t≠0 or a1,t≠0}; the exponent N+1 of K occurs only in KTN (each Tt has support in [−t,t] by step 1.1, and KδTt with t≤N can reach exponent N+1 only for (δ,t)=(1,N)), and its coefficient of KN+1 is the nonzero extreme coefficient of TN, so a1,N=0; then the exponent −N occurs only in TN (K1Tt has support in [1−t,1+t] with 1−t>−N for t≤N), forcing a0,N=0; descending on N gives all coefficients zero. Hence {KδTt} is an A-basis of M, as claimed in (a).

5.1step 4.1F3F4algebra∎

The full torus. By [F3], Uq0=Q(q)[P∨] with basis-like multiplication KhKh′=Kh+h′; by [F4], L=⨁iZ(dihi) is free of rank ∣I∣ and the Kdihi=Ki are algebraically independent. Let B be the A-subalgebra of Q(q)[L] generated by the Ki±1 and all Ti,c,t; since Q(q)[L]=⨂iQ(q)[Ki±1] and the generators with distinct i lie in distinct tensor factors, step 4.1 applied in each factor gives B=A-span{∏iKiδiTi,0,ti:δi∈{0,1},ti≥0} with the displayed products forming an A-basis. In Q(q)[P∨]=⨁cKc Q(q)[L], the direct sum running over the chosen representatives (whose existence and explicitness: the D-components of a fixed coset form a coset of L in the finitely generated free group D, and the elements of minimal coordinate absolute-value sum are finitely many by boundedness of the coordinates, so the lexicographic tie-break is a finite comparison; the complementary component is determined by the coset because L⊆D by [F4]), every product of generators of UA0 lies in ⨁cKcB because Kh1⋯Khm=K∑hl and the T-products lie in B; conversely Kc∈UA0 and B⊆UA0 (each KiδiTi,0,ti is a product of generators). Hence UA0=⨁cKcB, and the union of Kc times the A-basis of B over the representatives is an A-basis of UA0, proving (b). The construction uses only finitely many explicit coordinate comparisons per coset and no choice function on an infinite family; no integral PBW basis, canonical basis or modified-algebra input is used, and no freeness is inferred from torsion-freeness over the non-PID ring A.

Depends on

Used by

Dependency tree · two levels

32 results within two dependency steps of this one, each drawn at its shortest distance from it. An arrow runs from a result to what uses it, so the chart reads left to right and ends at this result, which carries a heavier outline. Every node is a link to that result. Click elsewhere on the chart to enlarge it.

Sources