Keywords. Twisted quantum affine algebra; Lusztig’s restricted integral form; integral form in Drinfeld presentation; quantum Frobenius; cyclotomic localization.
2020 Mathematics Subject Classification. 17B37, 17B67, 20G42.
Introduction
Quantum affine algebras admit two complementary presentations. The Drinfeld–Jimbo definition is attached directly to an affine Cartan datum and is expressed in Chevalley–Serre generators [12 , 17 ] , whereas Drinfeld’s new realization is organized by loop modes and current generators [13 ] . The current realization is particularly well adapted to the study of finite-dimensional representations [6 ] and the construction of level-one vertex representations [15 ] , while the Drinfeld–Jimbo presentation is the natural setting for Lusztig’s divided powers, braid symmetries, and specialization at roots of unity. Explicit formulas for the comparison between the Drinfeld–Jimbo and Drinfeld current presentations were given by Jing [16 ] and, for twisted types including A 2 n ( 2 ) A_{2n}^{(2)} A 2 n ( 2 ) , were also recorded by Jing and Misra [18 ] . These works stated the comparison as an isomorphism; a complete proof over C ( q ) \mathbb{C}(q) C ( q ) was later obtained by Damiani [8 , 9 ] . A braid group construction of it appeared in [19 ] .
The integral case is more interesting. Although the comparison isomorphism identifies the two algebras over C ( q ) \mathbb{C}(q) C ( q ) , it need not identify their prescribed Z [ q , q − 1 ] \mathbb{Z}[q,q^{-1}] Z [ q , q − 1 ] -forms. A denominator invisible over the rational function field can change the specialized algebra, and a generating set adapted to current modes need not generate the Chevalley divided-power lattice. The comparison also depends on exactly which current generators are placed in the integral form; this point is part of the definition, not a harmless normalization choice.
There is a useful classical precedent. The works of Chevalley and Kostant initiated the divided-power integral theory for finite-dimensional semisimple Lie algebras [4 , 20 ] . Garland treated untwisted loop algebras [14 ] , and Mitzman constructed integral bases for affine Kac–Moody algebras in the Chevalley presentation [22 ] ; see also [Pr]. Akasaka [1 ] constructed an integral PBW basis and an integral crystal basis of U q + ( A 2 n ( 2 ) ) U_q^+(A_{2n}^{(2)}) U q + ( A 2 n ( 2 ) ) , but did not determine the least ordinary localization over which the integral forms in Drinfeld presentation and Lusztig’s restricted integral forms coincide. At the classical specialization q ↦ 1 q\mapsto1 q ↦ 1 , Damiani–Paolini proved in rank one that the integral algebra generated by the divided powers of the Drinfeld generators of A 2 ( 2 ) A_{2}^{(2)} A 2 ( 2 ) is properly contained in the corresponding Chevalley–Mitzman integral form [10 ] . Paolini subsequently extended this classical comparison to all affine types and identified A 2 n ( 2 ) A_{2n}^{(2)} A 2 n ( 2 ) as the exceptional family [23 ] . These classical results predict where an obstruction may occur, while they do not determine its cyclotomic support in the generic quantum algebra.
The comparison of the integral forms for quantum twisted affine algebras is substantially complex and new ideas are needed. We pursue a structural reduction for the comparison problem in this paper. We use local q q q -Weyl contractions and braid transport to establish sufficiency, while modified quantum Frobenius and parity separation detect necessity. This approach does not require a global computation of all commutation relations, which were extremely complicated in quantum situation.
The untwisted simply-laced case provides an important quantum precedent. In this setting, the real-root comparison and the recovery of the affine simple-root divided powers established by Beck–Chari–Pressley show that no additional coefficient obstruction occurs [3 ] . Chari–Jing subsequently recorded the resulting equality between the corresponding Drinfeld and Lusztig integral forms [CJ, Prop. 1.2].
The goal of this paper is to compare the ”bare real-current core” generated as an A \mathcal{A} A -algebra by the group-like toral elements K δ ± 1 K_{\delta}^{\pm1} K δ ± 1 , K i ± 1 K_{i}^{\pm1} K i ± 1 , and the divided powers of the Drinfeld current modes x i , r ± x_{i,r}^{\pm} x i , r ± with the full Lusztig’s divided-power form. We denote the bare real-current core by U A D r ( g ) U_{\mathcal{A}}^{\mathrm{Dr}}(\mathfrak{g}) U A Dr ( g ) . No imaginary generator or Heisenberg generators, no additional real-root vector, and no toral quantum binomial is inserted into this definition. This choice separates the genuine real-current obstruction from artifacts of an enlarged generating set.
Our first result is that no obstruction occurs outside twisted type A 2 n ( 2 ) A_{2n}^{(2)} A 2 n ( 2 ) . For
A 2 n − 1 ( 2 ) , D n + 1 ( 2 ) , E 6 ( 2 ) , D 4 ( 3 ) , A_{2n-1}^{(2)},\quad D_{n+1}^{(2)},\quad E_{6}^{(2)},\quad D_{4}^{(3)}, A 2 n − 1 ( 2 ) , D n + 1 ( 2 ) , E 6 ( 2 ) , D 4 ( 3 ) ,
the bare real-current core equals Lusztig’s integral form over A \mathcal{A} A . The proof does not use information of a global PBW basis. An ordinary A 1 ( 1 ) A_{1}^{(1)} A 1 ( 1 ) vertex supplies a direct integral starting point for each finite color. Current translations and finite braid operators preserve the core; an explicit extended Weyl element then carries one starting root α i ∗ \alpha_{i_*} α i ∗ to α 0 \alpha_0 α 0 . Lifting that extended Weyl group element by a braid operator transports E i ∗ E_{i_*} E i ∗ to E 0 E_0 E 0 . Finite braid stability then places the affine simple divided powers in the current integral form, which therefore contains the full Lusztig form.
Type A 2 n ( 2 ) A_{2n}^{(2)} A 2 n ( 2 ) stands out in the family and splits sharply according to rank. Let
B ∗ = A [ Φ 8 r ( q ) − 1 ∣ r ≥ 1 ] , C ∗ = A [ Φ 4 r ( q ) − 1 ∣ r ≥ 1 ] , B^*=\mathcal{A}[\Phi_{8r}(q)^{-1}\mid r\ge1],\qquad C^*=\mathcal{A}[\Phi_{4r}(q)^{-1}\mid r\ge1], B ∗ = A [ Φ 8 r ( q ) − 1 ∣ r ≥ 1 ] , C ∗ = A [ Φ 4 r ( q ) − 1 ∣ r ≥ 1 ] ,
where Φ k ( q ) ∈ Z [ q ] \Phi_k(q)\in\mathbb{Z}[q] Φ k ( q ) ∈ Z [ q ] are the cyclotomic polynomials of degree ϕ ( k ) \phi(k) ϕ ( k ) .
For A 2 ( 2 ) A_{2}^{(2)} A 2 ( 2 ) , a single q q q -Weyl cell produces explicit coefficients
d a = ∏ h = 1 a ( q 2 h + 1 ) ( q 4 h + 1 ) d_a=\prod_{h=1}^{a}(q^{2h}+1)(q^{4h}+1) d a = h = 1 ∏ a ( q 2 h + 1 ) ( q 4 h + 1 )
with d a E η ( a ) , d a F η ( a ) ∈ U A D r d_aE_{\eta}^{(a)},d_aF_{\eta}^{(a)}\in U_{\mathcal{A}}^{\mathrm{Dr}} d a E η ( a ) , d a F η ( a ) ∈ U A Dr . Their irreducible support is contained in the Φ 4 r \Phi_{4r} Φ 4 r ’s, so C ∗ C^* C ∗ is sufficient. For A 2 n ( 2 ) A_{2n}^{(2)} A 2 n ( 2 ) , n ≥ 2 n\ge2 n ≥ 2 , the relevant coefficients are instead
c a = ∏ h = 1 a ( q 4 h + 1 ) . c_a=\prod_{h=1}^{a}(q^{4h}+1). c a = h = 1 ∏ a ( q 4 h + 1 ) .
We first create c a E η n ( a ) c_aE_{\eta_n}^{(a)} c a E η n ( a ) inside a special three-root cell. Denominator-free braid transport carries this integrality along a precisely normalized chain from η n \eta_n η n to η 1 = α 0 \eta_1=\alpha_0 η 1 = α 0 . Since the nonunit factors of c a c_a c a are exactly supported on Φ 8 r \Phi_{8r} Φ 8 r , the ring B ∗ B^* B ∗ is sufficient.
At roots of unity we use Lusztig’s modified quantum Frobenius formalism [21 ] . Sufficiency alone does not identify a least ring. For this we introduce the coefficient ideals
c n , r : = ( U A D r , ( n ) : A E 0 ( r ) ) . \mathfrak{c}_{n,r}:=\left(U^{\mathrm{Dr},(n)}_{A}:_{A}E_{0}^{(r)}\right). c n , r := ( U A Dr , ( n ) : A E 0 ( r ) ) .
At each relevant cyclotomic prime ideal, a modified quantum Frobenius map preserves the root degree of every nonzero block. We equip the scaled root lattice with the parity given by the coefficient of the scaled affine simple root α 0 \alpha_0 α 0 modulo 2 2 2 . With respect to this parity, every nonzero image of a real-current block is even, whereas the image of E 0 ( r ) 1 0 E_{0}^{(r)}1_0 E 0 ( r ) 1 0 is nonzero and odd. This forces the relevant coefficient ideal to be contained in that cyclotomic prime ideal. The standard Frobenius theorem applies at ( Φ 8 r ( q ) ) (\Phi_{8r}(q)) ( Φ 8 r ( q )) for r ≥ 2 r\ge2 r ≥ 2 , while a second scaled datum treats ( Φ 4 r ( q ) ) (\Phi_{4r}(q)) ( Φ 4 r ( q )) in rank one for odd r ≥ 3 r\ge3 r ≥ 3 .
Two endpoint roots require special treatment: ( Φ 8 ) (\Phi_8) ( Φ 8 ) in every rank and ( Φ 4 ) (\Phi_4) ( Φ 4 ) in rank one. The standard Frobenius size condition fails there, so it would be circular to invoke the same theorem. We instead form a common triangular quotient of the specialized modified form. Root-of-unity Gaussian vanishing shows that triangular straightening preserves badness, where a word is bad if some associated weight λ j ( w ) \lambda_j(w) λ j ( w ) lies outside X ∘ X^\circ X ∘ , or if it contains a divided-power letter E i ( a ) E_i^{(a)} E i ( a ) or F i ( a ) F_i^{(a)} F i ( a ) with ℓ i ∤ a \ell_i\nmid a ℓ i ∤ a . Consequently, the span J J J of the evaluations of all bad words agrees with the span of the evaluations of the triangularly ordered bad words. The resulting subspace J J J is root graded, and concatenation of words shows that it is a two-sided ideal. The quotient kills the evaluations of all bad words, preserves E 0 1 0 E_0 1_0 E 0 1 0 , and carries the same parity separation. Thus the endpoint primes are forced as well. The detailed Gaussian, straightening, and half-ideal arguments are collected in Appendix § 6.8 , while § 6.5 retains the quotient construction and its role in the parity argument.
Combining the constructive and obstructive directions gives the following classification. Among all ordinary localizations B = S − 1 A ⊂ K B=S^{-1}A\subset K B = S − 1 A ⊂ K ,
B U A D r ( A 2 ( 2 ) ) = B U A L u s ( A 2 ( 2 ) ) ⟺ Φ 4 r ( q ) ∈ B × for every r ≥ 1 , BU^{\mathrm{Dr}}_{A}\left(A_{2}^{(2)}\right)=BU^{\mathrm{Lus}}_{A}\left(A_{2}^{(2)}\right)
\quad\Longleftrightarrow\quad
\Phi_{4r}(q)\in B^{\times}\text{ for every }r\ge1, B U A Dr ( A 2 ( 2 ) ) = B U A Lus ( A 2 ( 2 ) ) ⟺ Φ 4 r ( q ) ∈ B × for every r ≥ 1 ,
and, for n ≥ 2 n\ge2 n ≥ 2 ,
B U A D r ( A 2 n ( 2 ) ) = B U A L u s ( A 2 n ( 2 ) ) ⟺ Φ 8 r ( q ) ∈ B × for every r ≥ 1. BU^{\mathrm{Dr}}_{A}\left(A_{2n}^{(2)}\right)=BU^{\mathrm{Lus}}_{A}\left(A_{2n}^{(2)}\right)
\quad\Longleftrightarrow\quad
\Phi_{8r}(q)\in B^{\times}\text{ for every }r\ge1. B U A Dr ( A 2 n ( 2 ) ) = B U A Lus ( A 2 n ( 2 ) ) ⟺ Φ 8 r ( q ) ∈ B × for every r ≥ 1.
Consequently the least rings are C ∗ C^{*} C ∗ in rank one and B ∗ B^{*} B ∗ in higher rank, whereas A A A itself is least in every other twisted family. In particular, for every n ≥ 1 n\ge1 n ≥ 1 , no localization of A A A obtained by inverting only finitely many elements is sufficient to make the two integral forms of type A 2 n ( 2 ) A_{2n}^{(2)} A 2 n ( 2 ) coincide.
Both exceptional coefficient rings have the same classical shadow. For m ≥ 1 m\ge1 m ≥ 1 , one has Φ m ( 1 ) = p \Phi_m(1)=p Φ m ( 1 ) = p when m m m is a power of the prime p p p , and Φ m ( 1 ) = 1 \Phi_m(1)=1 Φ m ( 1 ) = 1 otherwise. Consequently, base change along q ↦ 1 q\mapsto1 q ↦ 1 sends both C ∗ C^{*} C ∗ and B ∗ B^{*} B ∗ to Z [ 1 / 2 ] \mathbb{Z}[1/2] Z [ 1/2 ] , which is consistent with the classical conclusions obtained by Damiani–Paolini [10 ] and Paolini [23 ] . The classical strictness results are therefore a consistency check, but they cannot imply the individual coefficient-ideal obstructions proved here.
The paper is organized as follows. § 2 fixes the comparison isomorphism between the Drinfeld and Drinfeld–Jimbo realizations and places the two integral forms in a common ambient algebra. The remaining argument is divided into two parts. Part I consists of § 3.1 to § 3.3 and proves equality in the four non-A 2 n ( 2 ) A_{2n}^{(2)} A 2 n ( 2 ) twisted families. Part II begins with the type A 2 n ( 2 ) A_{2n}^{(2)} A 2 n ( 2 ) data in § 5.1 . Sections 6 and § 6 prove sufficiency in rank one and higher rank, respectively; § 6.2 develops the coefficient-ideal and parity-separation machinery; and § 6.3 and § 6.5 prove cyclotomic necessity for r ≥ 2 r\ge2 r ≥ 2 and r = 1 r=1 r = 1 , respectively, including the two exceptional endpoints. Finally, § 6.7 assembles the implications, proves minimality, and shows that in type A 2 n ( 2 ) A^{(2)}_{2n} A 2 n ( 2 ) , n ≥ 1 n \ge1 n ≥ 1 , no localization obtained by inverting finitely many nonzero Laurent polynomials makes the two integral forms equal. Appendix § 6.8 contains the technical proofs for the endpoint quotient used in § 6.5 .
Conventions, two integral forms, and minimality
Types, parameters, and the fixed comparison isomorphism
Throughout the paper, set
A = Z [ q , q − 1 ] , K = Q ( q ) . \mathcal{A} = \mathbb{Z}[q,q^{-1}], \qquad\mathcal{K} = \mathbb{Q}(q). A = Z [ q , q − 1 ] , K = Q ( q ) .
Let g \mathfrak{g} g be an irreducible twisted affine Cartan type in
T : = { A 2 n ( 2 ) ( n ≥ 1 ) , A 2 n − 1 ( 2 ) ( n ≥ 3 ) , D n + 1 ( 2 ) ( n ≥ 2 ) , E 6 ( 2 ) , D 4 ( 3 ) } . \mathcal{T}:=\left\{ A^{(2)}_{2n}\ (n\ge1),\ A^{(2)}_{2n-1}\ (n\ge3),\ D^{(2)}_{n+1}\ (n\ge2),\ E^{(2)}_6,\ D^{(3)}_4 \right\}. T := { A 2 n ( 2 ) ( n ≥ 1 ) , A 2 n − 1 ( 2 ) ( n ≥ 3 ) , D n + 1 ( 2 ) ( n ≥ 2 ) , E 6 ( 2 ) , D 4 ( 3 ) } .
The low-rank isomorphism A 3 ( 2 ) ≅ D 3 ( 2 ) A^{(2)}_3 \cong D^{(2)}_3 A 3 ( 2 ) ≅ D 3 ( 2 ) is understood through the case n = 2 n=2 n = 2 of the family D n + 1 ( 2 ) D^{(2)}_{n+1} D n + 1 ( 2 ) . Write I = { 0 } ⊔ I 0 I=\{0\}\sqcup I_0 I = { 0 } ⊔ I 0 for the set of affine nodes. We work throughout with the derived quantum affine algebra: the central element is retained, whereas the degree element is omitted.
Let A = ( a i j ) i , j ∈ I A=(a_{ij})_{i,j\in I} A = ( a ij ) i , j ∈ I be the affine Cartan matrix, and choose relatively prime positive integer symmetrizers ( d i ) i ∈ I (d_i)_{i\in I} ( d i ) i ∈ I satisfying
d i a i j = d j a j i ( i , j ∈ I ) . d_i a_{ij}=d_j a_{ji}\qquad(i,j\in I). d i a ij = d j a j i ( i , j ∈ I ) .
Outside type A 2 n ( 2 ) A^{(2)}_{2n} A 2 n ( 2 ) , the d i d_i d i are those used in equation (2.1). In type A 2 n ( 2 ) A^{(2)}_{2n} A 2 n ( 2 ) , they are normalized by
( d 0 , d 1 , … , d n ) = ( 4 , 2 , … , 2 , 1 ) ( n ≥ 2 ) , ( d 0 , d 1 ) = ( 4 , 1 ) ( n = 1 ) . (d_0,d_1,\ldots,d_n)=(4,2,\ldots,2,1)\quad(n\ge2),\qquad(d_0,d_1)=(4,1)\quad(n=1). ( d 0 , d 1 , … , d n ) = ( 4 , 2 , … , 2 , 1 ) ( n ≥ 2 ) , ( d 0 , d 1 ) = ( 4 , 1 ) ( n = 1 ) .
Set
q i : = q d i ( i ∈ I ) . (1) q_i:=q^{d_i}\qquad(i\in I). \tag*{(1)} q i := q d i ( i ∈ I ) . (1)
We use the corresponding integral invariant form
( α i ∣ α j ) : = d i a i j ; (\alpha_i\mid\alpha_j):=d_i a_{ij}; ( α i ∣ α j ) := d i a ij ;
thus ( α i ∣ α i ) = 2 d i (\alpha_i\mid\alpha_i)=2d_i ( α i ∣ α i ) = 2 d i , and the Cartan action and all root parameters are expressed through this form. Thus the local parameters in type A 2 n ( 2 ) A^{(2)}_{2n} A 2 n ( 2 ) are ( q 0 , q 1 , … , q n ) = ( q 4 , q 2 , … , q 2 , q ) (q_0,q_1,\ldots,q_n)=(q^4,q^2,\ldots,q^2,q) ( q 0 , q 1 , … , q n ) = ( q 4 , q 2 , … , q 2 , q ) for n ≥ 2 n\ge2 n ≥ 2 , and ( q 0 , q 1 ) = ( q 4 , q ) (q_0,q_1)=(q^4,q) ( q 0 , q 1 ) = ( q 4 , q ) for n = 1 n=1 n = 1 .
Let Δ r e \Delta^{\mathrm{re}} Δ re be the set of affine real roots. For i ∈ I 0 i\in I_0 i ∈ I 0 , define
ν i : = min { m > 0 ∣ m δ − α i ∈ Δ r e } , I i : = ν i Z ⊆ Z . (2) \nu_i:=\min\{m>0\mid m\delta-\alpha_i\in\Delta^{\mathrm{re}}\},\qquad I_i:=\nu_i\mathbb{Z}\subseteq\mathbb{Z}. \tag*{(2)} ν i := min { m > 0 ∣ m δ − α i ∈ Δ re } , I i := ν i Z ⊆ Z . (2)
Thus I i I_i I i is the set of allowed current modes of color i i i . See Figure 1 and Figure 2 for numbering of Dynkin nodes.
Figure 1. The non-A 2 n ( 2 ) A_{2n}^{(2)} A 2 n ( 2 ) twisted affine Dynkin diagrams. The low-rank isomorphism A 3 ( 2 ) ≅ D 3 ( 2 ) A_{3}^{(2)}\cong D_{3}^{(2)} A 3 ( 2 ) ≅ D 3 ( 2 ) is represented by the D 3 ( 2 ) D_{3}^{(2)} D 3 ( 2 ) row.
Figure 2. The affine Dynkin diagrams of A 2 n ( 2 ) A_{2n}^{(2)} A 2 n ( 2 ) used in Part II.
In the four non-A 2 n ( 2 ) A^{(2)}_{2n} A 2 n ( 2 ) types, ν i = d i \nu_i=d_i ν i = d i ; in type A 2 n ( 2 ) A^{(2)}_{2n} A 2 n ( 2 ) , ν i = 1 \nu_i=1 ν i = 1 . For an indeterminate v v v , use the symmetric quantum integers, factorials, and binomial coefficients
[ r ] v : = v r − v − r v − v − 1 , [ a ] v ! : = ∏ r = 1 a [ r ] v , [ a b ] v : = [ a ] v ! [ b ] v ! [ a − b ] v ! , (3) [r]_v:=\frac{v^r-v^{-r}}{v-v^{-1}},\qquad[a]_v!:=\prod_{r=1}^{a}[r]_v,\qquad
\begin{bmatrix}a\\b\end{bmatrix}_v:=\frac{[a]_v!}{[b]_v![a-b]_v!}, \tag*{(3)} [ r ] v := v − v − 1 v r − v − r , [ a ] v ! := r = 1 ∏ a [ r ] v , [ a b ] v := [ b ] v ! [ a − b ] v ! [ a ] v ! , (3)
where r ∈ Z r\in\mathbb{Z} r ∈ Z , 0 ≤ b ≤ a 0\le b\le a 0 ≤ b ≤ a , and [ 0 ] v ! : = 1 [0]_v!:=1 [ 0 ] v ! := 1 .
The Drinfeld–Jimbo presentation. The algebra U q D J ( g ) U_q^{\mathrm{DJ}}(\mathfrak{g}) U q DJ ( g ) is the unital K \mathcal{K} K -algebra generated by { E i , F i , K i ± 1 ∣ i ∈ I } \{E_i,F_i,K_i^{\pm1}\mid i\in I\} { E i , F i , K i ± 1 ∣ i ∈ I } , subject to
K i K i − 1 = K i − 1 K i = 1 , K i K j = K j K i , (4) K_iK_i^{-1}=K_i^{-1}K_i=1,\qquad K_iK_j=K_jK_i, \tag*{(4)} K i K i − 1 = K i − 1 K i = 1 , K i K j = K j K i , (4)
K i E j K i − 1 = q i a i j E j , K i F j K i − 1 = q i − a i j F j , (5) K_iE_jK_i^{-1}=q_i^{a_{ij}}E_j,\qquad K_iF_jK_i^{-1}=q_i^{-a_{ij}}F_j, \tag*{(5)} K i E j K i − 1 = q i a ij E j , K i F j K i − 1 = q i − a ij F j , (5)
[ E i , F j ] = δ i j K i − K i − 1 q i − q i − 1 ( i , j ∈ I ) , (6) [E_i,F_j]=\delta_{ij}\frac{K_i-K_i^{-1}}{q_i-q_i^{-1}}\qquad(i,j\in I), \tag*{(6)} [ E i , F j ] = δ ij q i − q i − 1 K i − K i − 1 ( i , j ∈ I ) , (6)
∑ p = 0 1 − a i j ( − 1 ) p E i ( p ) E j E i ( 1 − a i j − p ) = 0 ( i ≠ j ) , (7) \sum_{p=0}^{1-a_{ij}}(-1)^pE_i^{(p)}E_jE_i^{(1-a_{ij}-p)}=0\qquad(i\ne j), \tag*{(7)} p = 0 ∑ 1 − a ij ( − 1 ) p E i ( p ) E j E i ( 1 − a ij − p ) = 0 ( i = j ) , (7)
∑ p = 0 1 − a i j ( − 1 ) p F i ( p ) F j F i ( 1 − a i j − p ) = 0 ( i ≠ j ) . (8) \sum_{p=0}^{1-a_{ij}}(-1)^pF_i^{(p)}F_jF_i^{(1-a_{ij}-p)}=0\qquad(i\ne j). \tag*{(8)} p = 0 ∑ 1 − a ij ( − 1 ) p F i ( p ) F j F i ( 1 − a ij − p ) = 0 ( i = j ) . (8)
Here the Chevalley divided powers are
E i ( a ) : = E i a [ a ] q i ! , F i ( a ) : = F i a [ a ] q i ! ( i ∈ I , a ≥ 0 ) . (9) E_i^{(a)}:=\frac{E_i^a}{[a]_{q_i}!},\qquad F_i^{(a)}:=\frac{F_i^a}{[a]_{q_i}!}\qquad(i\in I,\ a\ge0). \tag*{(9)} E i ( a ) := [ a ] q i ! E i a , F i ( a ) := [ a ] q i ! F i a ( i ∈ I , a ≥ 0 ) . (9)
Write Q : = ⨁ i ∈ I Z α i Q:=\bigoplus_{i\in I}\mathbb{Z}\alpha_i Q := ⨁ i ∈ I Z α i for the root lattice. For μ = ∑ i ∈ I m i α i ∈ Q \mu=\sum_{i\in I}m_i\alpha_i\in Q μ = ∑ i ∈ I m i α i ∈ Q , put K μ : = ∏ i ∈ I K i m i K_\mu:=\prod_{i\in I}K_i^{m_i} K μ := ∏ i ∈ I K i m i ; in particular, writing δ = α 0 + ∑ i ∈ I 0 r i α i \delta=\alpha_0+\sum_{i\in I_0}r_i\alpha_i δ = α 0 + ∑ i ∈ I 0 r i α i , we have K δ = K 0 ∏ i ∈ I 0 K i r i K_\delta=K_0\prod_{i\in I_0}K_i^{r_i} K δ = K 0 ∏ i ∈ I 0 K i r i . Put
Q + : = ∑ i ∈ I N α i . Q_+:=\sum_{i\in I}\mathbb{N}\alpha_i. Q + := i ∈ I ∑ N α i .
Let U q + U_q^+ U q + denote the K \mathcal{K} K -subalgebra of U q D J ( g ) U_q^{\mathrm{DJ}}(\mathfrak{g}) U q DJ ( g ) generated by the E i E_i E i , i ∈ I i\in I i ∈ I . Its root-lattice grading is written as
U q + = ⨁ γ ∈ Q + ( U q + ) γ , deg ( E i ) = α i . U_q^+=\bigoplus_{\gamma\in Q_+}(U_q^+)_\gamma,\qquad\deg(E_i)=\alpha_i. U q + = γ ∈ Q + ⨁ ( U q + ) γ , deg ( E i ) = α i .
The Drinfeld presentation. Set ν i j : = max { ν i , ν j } \nu_{ij}:=\max\{\nu_i,\nu_j\} ν ij := max { ν i , ν j } , and write [ u , v ] a : = u v − a v u [u,v]_a:=uv-avu [ u , v ] a := uv − a v u . For r ∈ Z ∖ { 0 } r\in\mathbb{Z}\setminus\{0\} r ∈ Z ∖ { 0 } , define
b i j ( r ) : = { 0 , ν i j ∤ r , [ 2 r ] q ( q 2 r + ( − 1 ) r − 1 + q − 2 r ) r , g = A 2 n ( 2 ) , i = j = n , [ ( r / ν i j ) a i j ] q i r / ν i j , otherwise . b_{ij}(r):=
\begin{cases}
0,&\nu_{ij}\nmid r,\\
\dfrac{[2r]_{q}\left(q^{2r}+(-1)^{r-1}+q^{-2r}\right)}{r},&\mathfrak{g}=A_{2n}^{(2)},\ i=j=n,\\
\dfrac{[(r/\nu_{ij})a_{ij}]_{q_i}}{r/\nu_{ij}},&\text{otherwise}.
\end{cases} b ij ( r ) := ⎩ ⎨ ⎧ 0 , r [ 2 r ] q ( q 2 r + ( − 1 ) r − 1 + q − 2 r ) , r / ν ij [( r / ν ij ) a ij ] q i , ν ij ∤ r , g = A 2 n ( 2 ) , i = j = n , otherwise .
The algebra U q D r ( g ) U_q^{\mathrm{Dr}}(\mathfrak{g}) U q Dr ( g ) is the unital K \mathcal{K} K -algebra generated by
{ K δ ± 1 , K i ± 1 , x i , r ± , h i , s ∣ i ∈ I 0 , r ∈ I i , s ∈ I i ∖ { 0 } } \{K_\delta^{\pm1},\ K_i^{\pm1},\ x_{i,r}^{\pm},\ h_{i,s}\mid i\in I_0,\ r\in I_i,\ s\in I_i\setminus\{0\}\} { K δ ± 1 , K i ± 1 , x i , r ± , h i , s ∣ i ∈ I 0 , r ∈ I i , s ∈ I i ∖ { 0 }}
with the relations listed below. Use the zero-mode convention
x i , r ± = h i , r = 0 ( r ∉ I i ) , (10) x_{i,r}^{\pm}=h_{i,r}=0\qquad(r\notin I_i), \tag*{(10)} x i , r ± = h i , r = 0 ( r ∈ / I i ) , (10)
and define
∑ a ≥ 0 ϕ i , ± a ± u a = K i ± 1 exp ( ± ( q i − q i − 1 ) ∑ a > 0 h i , ± a u a ) , ϕ i , t ± = 0 if ± t < 0. (11) \sum_{a\ge0}\phi_{i,\pm a}^{\pm}u^a=K_i^{\pm1}\exp\left(\pm(q_i-q_i^{-1})\sum_{a>0}h_{i,\pm a}u^a\right),\qquad\phi_{i,t}^{\pm}=0\ \text{if }\pm t<0. \tag*{(11)} a ≥ 0 ∑ ϕ i , ± a ± u a = K i ± 1 exp ( ± ( q i − q i − 1 ) a > 0 ∑ h i , ± a u a ) , ϕ i , t ± = 0 if ± t < 0. (11)
For ε ∈ { 1 , − 1 } \varepsilon\in\{1,-1\} ε ∈ { 1 , − 1 } , write x i , r 1 = x i , r + x_{i,r}^{1}=x_{i,r}^{+} x i , r 1 = x i , r + , x i , r − 1 = x i , r − x_{i,r}^{-1}=x_{i,r}^{-} x i , r − 1 = x i , r − . The notation Sym r 1 , … , r a \operatorname{Sym}_{r_1,\ldots,r_a} Sym r 1 , … , r a means the sum over all permutations of these mode indices. For m = − a i j ∈ { 2 , 3 } m=-a_{ij}\in\{2,3\} m = − a ij ∈ { 2 , 3 } , let
P i j ε ( u , v ; s ) : = x j , s ε x i , u ε x i , v ε − [ 2 ] q m x i , u ε x j , s ε x i , v ε + x i , u ε x i , v ε x j , s ε . P_{ij}^{\varepsilon}(u,v;s):=x_{j,s}^{\varepsilon}x_{i,u}^{\varepsilon}x_{i,v}^{\varepsilon}-[2]_{q_m}x_{i,u}^{\varepsilon}x_{j,s}^{\varepsilon}x_{i,v}^{\varepsilon}+x_{i,u}^{\varepsilon}x_{i,v}^{\varepsilon}x_{j,s}^{\varepsilon}. P ij ε ( u , v ; s ) := x j , s ε x i , u ε x i , v ε − [ 2 ] q m x i , u ε x j , s ε x i , v ε + x i , u ε x i , v ε x j , s ε .
In the following list, i , j ∈ I 0 i,j\in I_0 i , j ∈ I 0 , ε , η ∈ { 1 , − 1 } \varepsilon,\eta\in\{1,-1\} ε , η ∈ { 1 , − 1 } , and z z z ranges over the generators. In the first five relations, input modes belong to their respective I i I_i I i , and Heisenberg modes are nonzero; all mode indices in the remaining relations range over Z \mathbb{Z} Z , using the zero convention above. In the last two relations, M = 1 − a i j M=1-a_{ij} M = 1 − a ij and m = − a i j m=-a_{ij} m = − a ij , respectively. The defining relations are
[ K δ , z ] = 0 , K δ K δ − 1 = K δ − 1 K δ = 1 , [K_\delta,z]=0,\qquad K_\delta K_\delta^{-1}=K_\delta^{-1}K_\delta=1, [ K δ , z ] = 0 , K δ K δ − 1 = K δ − 1 K δ = 1 ,
K i K i − 1 = K i − 1 K i = 1 , K i K j = K j K i , (12) K_iK_i^{-1}=K_i^{-1}K_i=1,\qquad K_iK_j=K_jK_i, \tag*{(12)} K i K i − 1 = K i − 1 K i = 1 , K i K j = K j K i , (12)
K i x j , r ε K i − 1 = q i ε a i j x j , r ε , [ K i , h j , s ] = 0 , (13) K_i x_{j,r}^{\varepsilon}K_i^{-1}=q_i^{\varepsilon a_{ij}}x_{j,r}^{\varepsilon},\qquad[K_i,h_{j,s}]=0, \tag*{(13)} K i x j , r ε K i − 1 = q i ε a ij x j , r ε , [ K i , h j , s ] = 0 , (13)
[ x i , r + , x j , s − ] = δ i j K δ − s ϕ i , r + s + − K δ − r ϕ i , r + s − q i − q i − 1 , (14) [x_{i,r}^{+},x_{j,s}^{-}]=\delta_{ij}\frac{K_\delta^{-s}\phi_{i,r+s}^{+}-K_\delta^{-r}\phi_{i,r+s}^{-}}{q_i-q_i^{-1}}, \tag*{(14)} [ x i , r + , x j , s − ] = δ ij q i − q i − 1 K δ − s ϕ i , r + s + − K δ − r ϕ i , r + s − , (14)
[ h i , r , x j , s ε ] = ε b i j ( r ) K δ ( r − ε ∣ r ∣ ) / 2 x j , r + s ε , (15) [h_{i,r},x_{j,s}^{\varepsilon}]=\varepsilon b_{ij}(r)K_\delta^{(r-\varepsilon|r|)/2}x_{j,r+s}^{\varepsilon}, \tag*{(15)} [ h i , r , x j , s ε ] = ε b ij ( r ) K δ ( r − ε ∣ r ∣ ) /2 x j , r + s ε , (15)
[ h i , r , h j , s ] = δ r + s , 0 b i j ( r ) K δ r − K δ − r q j − q j − 1 , (16) [h_{i,r},h_{j,s}]=\delta_{r+s,0}b_{ij}(r)\frac{K_\delta^{r}-K_\delta^{-r}}{q_j-q_j^{-1}}, \tag*{(16)} [ h i , r , h j , s ] = δ r + s , 0 b ij ( r ) q j − q j − 1 K δ r − K δ − r , (16)
[ x i , r + ε ν i j ε , x j , s ε ] q i a i j + [ x j , s + ε ν i j ε , x i , r ε ] q i a i j = 0 ( i ≠ j , a i j < 0 ) , (17) [x_{i,r+\varepsilon\nu_{ij}}^{\varepsilon},x_{j,s}^{\varepsilon}]_{q_i^{a_{ij}}}+[x_{j,s+\varepsilon\nu_{ij}}^{\varepsilon},x_{i,r}^{\varepsilon}]_{q_i^{a_{ij}}}=0\qquad(i\ne j,\ a_{ij}<0), \tag*{(17)} [ x i , r + ε ν ij ε , x j , s ε ] q i a ij + [ x j , s + ε ν ij ε , x i , r ε ] q i a ij = 0 ( i = j , a ij < 0 ) , (17)
Sym r 1 , r 2 [ x i , r 1 + ε ν i ε , x i , r 2 ε ] q i 2 = 0 ( ( g , i ) ≠ ( A 2 n ( 2 ) , n ) ) , (18) \operatorname{Sym}_{r_1,r_2}[x_{i,r_1+\varepsilon\nu_i}^{\varepsilon},x_{i,r_2}^{\varepsilon}]_{q_i^2}=0\qquad((\mathfrak g,i)\ne(A_{2n}^{(2)},n)), \tag*{(18)} Sym r 1 , r 2 [ x i , r 1 + ε ν i ε , x i , r 2 ε ] q i 2 = 0 (( g , i ) = ( A 2 n ( 2 ) , n )) , (18)
Sym r 1 , r 2 ( [ x n , r 1 + 2 ε ε , x n , r 2 ε ] q 2 − q 4 [ x n , r 1 + ε ε , x n , r 2 + ε ε ] q − 6 ) = 0 \operatorname{Sym}_{r_1,r_2}\left([x_{n,r_1+2\varepsilon}^{\varepsilon},x_{n,r_2}^{\varepsilon}]_{q^2}-q^4[x_{n,r_1+\varepsilon}^{\varepsilon},x_{n,r_2+\varepsilon}^{\varepsilon}]_{q^{-6}}\right)=0 Sym r 1 , r 2 ( [ x n , r 1 + 2 ε ε , x n , r 2 ε ] q 2 − q 4 [ x n , r 1 + ε ε , x n , r 2 + ε ε ] q − 6 ) = 0
( g = A 2 n ( 2 ) ) , (19) (\mathfrak g=A_{2n}^{(2)}), \tag*{(19)} ( g = A 2 n ( 2 ) ) , (19)
Sym r 1 , r 2 , r 3 ( [ x n , r 1 + ε η ε , x n , r 2 ε ] q 2 η , x n , r 3 ε ) q 4 η = 0 ( g = A 2 n ( 2 ) ) , (20) \operatorname{Sym}_{r_1,r_2,r_3}\left([x_{n,r_1+\varepsilon\eta}^{\varepsilon},x_{n,r_2}^{\varepsilon}]_{q^{2\eta}},x_{n,r_3}^{\varepsilon}\right)_{q^{4\eta}}=0\qquad(\mathfrak g=A_{2n}^{(2)}), \tag*{(20)} Sym r 1 , r 2 , r 3 ( [ x n , r 1 + ε η ε , x n , r 2 ε ] q 2 η , x n , r 3 ε ) q 4 η = 0 ( g = A 2 n ( 2 ) ) , (20)
Sym r 1 , … , r M ∑ p = 0 M ( − 1 ) p [ M p ] q i x i , r 1 ε ⋯ x i , r p ε x j , s ε x i , r p + 1 ε ⋯ x i , r M ε = 0 \operatorname{Sym}_{r_1,\ldots,r_M}\sum_{p=0}^{M}(-1)^p\begin{bmatrix}M\\p\end{bmatrix}_{q_i}
x_{i,r_1}^{\varepsilon}\cdots x_{i,r_p}^{\varepsilon}x_{j,s}^{\varepsilon}x_{i,r_{p+1}}^{\varepsilon}\cdots x_{i,r_M}^{\varepsilon}=0 Sym r 1 , … , r M p = 0 ∑ M ( − 1 ) p [ M p ] q i x i , r 1 ε ⋯ x i , r p ε x j , s ε x i , r p + 1 ε ⋯ x i , r M ε = 0
( i ≠ j , a i j ∈ { 0 , − 1 } ) , (21) (i\ne j,\ a_{ij}\in\{0,-1\}), \tag*{(21)} ( i = j , a ij ∈ { 0 , − 1 }) , (21)
Sym r 1 , r 2 ∑ p = 0 m − 1 q m − 1 − 2 p P i j ε ( r 1 + ε ( m − 1 − p ) , r 2 + ε p ; s ) = 0 ( m ∈ { 2 , 3 } ) . (22) \operatorname{Sym}_{r_1,r_2}\sum_{p=0}^{m-1}q^{m-1-2p}P_{ij}^{\varepsilon}(r_1+\varepsilon(m-1-p),r_2+\varepsilon p;s)=0\qquad(m\in\{2,3\}). \tag*{(22)} Sym r 1 , r 2 p = 0 ∑ m − 1 q m − 1 − 2 p P ij ε ( r 1 + ε ( m − 1 − p ) , r 2 + εp ; s ) = 0 ( m ∈ { 2 , 3 }) . (22)
The comparison map and braid action. Over K \mathcal{K} K , we fix the comparison isomorphism
Ψ g : U q D r ( g ) → ∼ U q D J ( g ) , (23) \Psi_{\mathfrak g}:U_q^{\mathrm{Dr}}(\mathfrak g)\xrightarrow{\sim}U_q^{\mathrm{DJ}}(\mathfrak g), \tag*{(23)} Ψ g : U q Dr ( g ) ∼ U q DJ ( g ) , (23)
whose values on the generators are specified below. We use Beck’s braid operators T i T_i T i [2 Section 1.3] . The standard finite and extended braid lifts preserve the Lusztig integral form:
T w ± 1 ( U A L u s ( g ) ) = U A L u s ( g ) ; (24) T_w^{\pm1}\left(U_{\mathcal A}^{\mathrm{Lus}}(\mathfrak g)\right)=U_{\mathcal A}^{\mathrm{Lus}}(\mathfrak g); \tag*{(24)} T w ± 1 ( U A Lus ( g ) ) = U A Lus ( g ) ; (24)
see [21 Section 41.1.2] . For i ≠ j i\ne j i = j , let m = − a i j m=-a_{ij} m = − a ij . The braid operators T i T_i T i act on the positive Chevalley generators by
T i ( E i ) = − F i K i , T i − 1 ( E i ) = − K i − 1 F i , T_i(E_i)=-F_iK_i,\qquad T_i^{-1}(E_i)=-K_i^{-1}F_i, T i ( E i ) = − F i K i , T i − 1 ( E i ) = − K i − 1 F i ,
T i ( E j ) = ∑ r = 0 m ( − 1 ) m + r q i − r E i ( m − r ) E j E i ( r ) , T_i(E_j)=\sum_{r=0}^{m}(-1)^{m+r}q_i^{-r}E_i^{(m-r)}E_jE_i^{(r)}, T i ( E j ) = r = 0 ∑ m ( − 1 ) m + r q i − r E i ( m − r ) E j E i ( r ) ,
T i − 1 ( E j ) = ∑ r = 0 m ( − 1 ) m + r q i − r E i ( r ) E j E i ( m − r ) . (25) T_i^{-1}(E_j)=\sum_{r=0}^{m}(-1)^{m+r}q_i^{-r}E_i^{(r)}E_jE_i^{(m-r)}. \tag*{(25)} T i − 1 ( E j ) = r = 0 ∑ m ( − 1 ) m + r q i − r E i ( r ) E j E i ( m − r ) . (25)
For a reduced expression w = s i 1 ⋯ s i t τ w=s_{i_1}\cdots s_{i_t}\tau w = s i 1 ⋯ s i t τ , with τ \tau τ a length-zero diagram automorphism, set
T w : = T i 1 ⋯ T i t T τ , T τ ( E i ) = E τ ( i ) , T τ ( F i ) = F τ ( i ) , T τ ( K i ) = K τ ( i ) . (26) T_w:=T_{i_1}\cdots T_{i_t}T_\tau,\qquad T_\tau(E_i)=E_{\tau(i)},\qquad T_\tau(F_i)=F_{\tau(i)},\qquad T_\tau(K_i)=K_{\tau(i)}. \tag*{(26)} T w := T i 1 ⋯ T i t T τ , T τ ( E i ) = E τ ( i ) , T τ ( F i ) = F τ ( i ) , T τ ( K i ) = K τ ( i ) . (26)
For ordinary Weyl-group elements, τ = 1 \tau=1 τ = 1 . Weyl-group products act from right to left, and T i ( K μ ) = K s i ( μ ) T_i(K_\mu)=K_{s_i(\mu)} T i ( K μ ) = K s i ( μ ) . We use the semilinear antiautomorphism Ω \Omega Ω given by
Ω ( q ) = q − 1 , Ω ( E i ) = F i , Ω ( F i ) = E i , Ω ( K μ ) = K − μ . (27) \Omega(q)=q^{-1},\qquad\Omega(E_i)=F_i,\qquad\Omega(F_i)=E_i,\qquad\Omega(K_\mu)=K_{-\mu}. \tag*{(27)} Ω ( q ) = q − 1 , Ω ( E i ) = F i , Ω ( F i ) = E i , Ω ( K μ ) = K − μ . (27)
It is easy to check that
Ω T w = T w Ω . (28) \Omega T_w=T_w\Omega. \tag*{(28)} Ω T w = T w Ω. (28)
Thus T i ± 1 ( F j ) = Ω ( T i ± 1 ( E j ) ) T_i^{\pm1}(F_j)=\Omega(T_i^{\pm1}(E_j)) T i ± 1 ( F j ) = Ω ( T i ± 1 ( E j )) , which gives the formulas for the braid action on the generators.
Real root vectors. For i ∈ I 0 i\in I_0 i ∈ I 0 , let ω i ∨ \omega_i^\vee ω i ∨ be the fundamental coweight, and put
λ i : = ν i ω i ∨ , t λ i ( α ) : = α − ⟨ λ i , α ⟩ δ , T λ i : = T t λ i . (29) \lambda_i:=\nu_i\omega_i^\vee,\qquad t_{\lambda_i}(\alpha):=\alpha-\langle\lambda_i,\alpha\rangle\delta,\qquad T_{\lambda_i}:=T_{t_{\lambda_i}}. \tag*{(29)} λ i := ν i ω i ∨ , t λ i ( α ) := α − ⟨ λ i , α ⟩ δ , T λ i := T t λ i . (29)
Here ⟨ ω i ∨ , α j ⟩ = δ i j \langle\omega_i^\vee,\alpha_j\rangle=\delta_{ij} ⟨ ω i ∨ , α j ⟩ = δ ij for j ∈ I 0 j\in I_0 j ∈ I 0 , and ⟨ λ i , δ ⟩ = 0 \langle\lambda_i,\delta\rangle=0 ⟨ λ i , δ ⟩ = 0 . Order I 0 I_0 I 0 increasingly in the non-A 2 n ( 2 ) A_{2n}^{(2)} A 2 n ( 2 ) types and decreasingly in type A 2 n ( 2 ) A_{2n}^{(2)} A 2 n ( 2 ) , and denote this order by ( j 1 , … , j ∣ I 0 ∣ ) (j_1,\ldots,j_{|I_0|}) ( j 1 , … , j ∣ I 0 ∣ ) . For each j ∈ I 0 j\in I_0 j ∈ I 0 , choose a reduced expression for t λ j t_{\lambda_j} t λ j ending in s j s_j s j ; this is possible because t λ j ( α j ) = α j − ν j δ < 0 t_{\lambda_j}(\alpha_j)=\alpha_j-\nu_j\delta<0 t λ j ( α j ) = α j − ν j δ < 0 . Concatenate these expressions in the chosen order and move all length-zero diagram automorphisms to the right. This gives a reduced expression
t λ j 1 ⋯ t λ j ∣ I 0 ∣ = s i 1 ⋯ s i N τ , N = ∑ i ∈ I 0 ℓ ( t λ i ) . (30) t_{\lambda_{j_1}}\cdots t_{\lambda_{j_{|I_0|}}}=s_{i_1}\cdots s_{i_N}\tau,\qquad N=\sum_{i\in I_0}\ell(t_{\lambda_i}). \tag*{(30)} t λ j 1 ⋯ t λ j ∣ I 0 ∣ = s i 1 ⋯ s i N τ , N = i ∈ I 0 ∑ ℓ ( t λ i ) . (30)
In type A 2 n ( 2 ) A_{2n}^{(2)} A 2 n ( 2 ) , choose the initial block
( s 0 s 1 ⋯ s n ) n = t λ n . (31) (s_0s_1\cdots s_n)^n=t_{\lambda_n}. \tag*{(31)} ( s 0 s 1 ⋯ s n ) n = t λ n . (31)
Extend the indices to Z \mathbb{Z} Z by i r + N = τ ( i r ) i_{r+N}=\tau(i_r) i r + N = τ ( i r ) , and set
w r : = { s i 1 ⋯ s i r − 1 , r ≥ 1 , s i 0 ⋯ s i r + 1 , r ≤ 0 , ρ r : = w r ( α i r ) . (32) w_r:=\begin{cases}
s_{i_1}\cdots s_{i_{r-1}},&r\geq1,\\
s_{i_0}\cdots s_{i_{r+1}},&r\leq0,
\end{cases}
\qquad\rho_r:=w_r(\alpha_{i_r}). \tag*{(32)} w r := { s i 1 ⋯ s i r − 1 , s i 0 ⋯ s i r + 1 , r ≥ 1 , r ≤ 0 , ρ r := w r ( α i r ) . (32)
The roots ρ r \rho_r ρ r , r ∈ Z r\in\mathbb{Z} r ∈ Z , enumerate Δ r e + \Delta_{\mathrm{re}}^+ Δ re + without repetition. We define
E ρ r : = { T i 1 ⋯ T i r − 1 ( E i r ) , r ≥ 1 , T i 0 − 1 ⋯ T i r + 1 − 1 ( E i r ) , r ≤ 0 , F ρ r : = Ω ( E ρ r ) . (33) E_{\rho_r}:=\begin{cases}
T_{i_1}\cdots T_{i_{r-1}}(E_{i_r}),&r\geq1,\\
T_{i_0}^{-1}\cdots T_{i_{r+1}}^{-1}(E_{i_r}),&r\leq0,
\end{cases}
\qquad F_{\rho_r}:=\Omega(E_{\rho_r}). \tag*{(33)} E ρ r := { T i 1 ⋯ T i r − 1 ( E i r ) , T i 0 − 1 ⋯ T i r + 1 − 1 ( E i r ) , r ≥ 1 , r ≤ 0 , F ρ r := Ω ( E ρ r ) . (33)
with empty products equal to the identity. These are straightforward generalization of untwisted cases (Damiani [7 ] ). Note that the initial block in type A 2 n ( 2 ) A_{2n}^{(2)} A 2 n ( 2 ) is the reduced block therein. In particular,
E m ν i δ + α i = T λ i − m ( E i ) ( m ≥ 0 ) , (34) E_{m\nu_i\delta+\alpha_i}=T_{\lambda_i}^{-m}(E_i)\qquad(m\geq0), \tag*{(34)} E m ν i δ + α i = T λ i − m ( E i ) ( m ≥ 0 ) , (34)
E m ν i δ − α i = T λ i m T i − 1 ( E i ) ( m > 0 ) . E_{m\nu_i\delta-\alpha_i}=T_{\lambda_i}^{m}T_i^{-1}(E_i)\qquad(m>0). E m ν i δ − α i = T λ i m T i − 1 ( E i ) ( m > 0 ) .
Choose the bipartite sign function o : I 0 → { ± 1 } o:I_0\to\{ \pm1\} o : I 0 → { ± 1 } satisfying
o ( i ) o ( j ) = − 1 ( i ≠ j , a i j ≠ 0 ) , o(i)o(j)=-1\qquad(i\ne j,\ a_{ij}\ne0), o ( i ) o ( j ) = − 1 ( i = j , a ij = 0 ) ,
o ( i ) = 1 ( g ≠ A 2 n ( 2 ) , a i j = − 2 ) , o ( i ) = ( − 1 ) n − i ( g = A 2 n ( 2 ) ) . (35) o(i)=1\qquad(\mathfrak{g}\ne A_{2n}^{(2)},\ a_{ij}=-2),\qquad o(i)=(-1)^{n-i}\qquad(\mathfrak{g}=A_{2n}^{(2)}). \tag*{(35)} o ( i ) = 1 ( g = A 2 n ( 2 ) , a ij = − 2 ) , o ( i ) = ( − 1 ) n − i ( g = A 2 n ( 2 ) ) . (35)
Then the comparison isomorphism can be stated as follows.
Ψ g ( K δ ) = K δ , Ψ g ( K i ) = K i , \Psi_{\mathfrak{g}}(K_\delta)=K_\delta,\qquad\Psi_{\mathfrak{g}}(K_i)=K_i, Ψ g ( K δ ) = K δ , Ψ g ( K i ) = K i ,
Ψ g ( x i , ν i r + ) = o ( i ) r T λ i − r ( E i ) , Ψ g ( x i , ν i r − ) = o ( i ) r T λ i r ( F i ) ( r ∈ Z ) . (36) \Psi_{\mathfrak{g}}(x^+_{i,\nu_i r})=o(i)^rT_{\lambda_i}^{-r}(E_i),\qquad\Psi_{\mathfrak{g}}(x^-_{i,\nu_i r})=o(i)^rT_{\lambda_i}^{r}(F_i)\qquad(r\in\mathbb{Z}). \tag*{(36)} Ψ g ( x i , ν i r + ) = o ( i ) r T λ i − r ( E i ) , Ψ g ( x i , ν i r − ) = o ( i ) r T λ i r ( F i ) ( r ∈ Z ) . (36)
For complete proofs, see [8 , 9 ] (also see [19 ] ). The coefficients in the defining relations and in the formulas for the comparison map belong to Q ( q ) \mathbb{Q}(q) Q ( q ) , so the comparison homomorphism is first defined over K \mathcal{K} K . Since C ( q ) / Q ( q ) \mathbb{C}(q)/\mathbb{Q}(q) C ( q ) / Q ( q ) is faithfully flat, its kernel and cokernel already vanish over Q ( q ) \mathbb{Q}(q) Q ( q ) . Thus (23) is indeed an isomorphism of K \mathcal{K} K -algebras.
We henceforth identify the Drinfeld generators with their images under Ψ g \Psi_{\mathfrak{g}} Ψ g and write U q ( g ) U_q(\mathfrak{g}) U q ( g ) for the common algebra. All root–coroot pairings are denoted by ⟨ h i , γ ⟩ \langle h_i,\gamma\rangle ⟨ h i , γ ⟩ . For m ≥ 1 m\geq1 m ≥ 1 , let Φ m ( q ) \Phi_m(q) Φ m ( q ) denote the m m m -th cyclotomic polynomial.
Whenever a modified form is used, we fix a regular realization ( X , Y , I , ⟨ , ⟩ ) (X,Y,I,\langle\, ,\,\rangle) ( X , Y , I , ⟨ , ⟩) of the same Cartan datum, solely to define the weight idempotents, modified blocks, and modified braid action. The ordinary quantum affine algebra remains the derived subalgebra, so no degree element is added to either the ordinary integral form.
For i ∈ I 0 i\in I_0 i ∈ I 0 , r ∈ I i r\in I_i r ∈ I i , and a ≥ 0 a\geq0 a ≥ 0 , define
( x i , r ± ) ( a ) : = ( x i , r ± ) a [ a ] q i ! . (37) \left(x^\pm_{i,r}\right)^{(a)}:=\frac{\left(x^\pm_{i,r}\right)^a}{[a]_{q_i}!}. \tag*{(37)} ( x i , r ± ) ( a ) := [ a ] q i ! ( x i , r ± ) a . (37)
with the factorials of (3) .
Writing
δ = α 0 + ∑ i ∈ I 0 r i α i , \delta=\alpha_0+\sum_{i\in I_0}r_i\alpha_i, δ = α 0 + i ∈ I 0 ∑ r i α i ,
the affine toral generator is recovered by
K 0 = K δ ∏ i ∈ I 0 K i − r i . (38) K_0=K_\delta\prod_{i\in I_0}K_i^{-r_i}. \tag*{(38)} K 0 = K δ i ∈ I 0 ∏ K i − r i . (38)
Lemma 1.1. [2 Lemma 3.2] Let w w w lie in the (extended) affine Weyl group and let i , j i,j i , j be nodes with w ( α i ) = α j w(\alpha_i)=\alpha_j w ( α i ) = α j . Then w s i ws_i w s i and s j w s_jw s j w are length additive and
T w T i = T j T w , T w ( E i ) = E j , T w ( F i ) = F j . T_wT_i=T_jT_w,\qquad T_w(E_i)=E_j,\qquad T_w(F_i)=F_j. T w T i = T j T w , T w ( E i ) = E j , T w ( F i ) = F j .
The two integral forms
Throughout this paper, an integral form of a K \mathcal{K} K -algebra U U U means a unital A \mathcal{A} A -subalgebra U A ⊆ U U_{\mathcal{A}}\subseteq U U A ⊆ U for which the natural map K ⊗ A U A → U \mathcal{K}\otimes_{\mathcal{A}}U_{\mathcal{A}}\to U K ⊗ A U A → U is an isomorphism. Freeness as an A \mathcal{A} A -module is not required in this definition. We use the divided powers defined in (9) .
Definition 1.2 (Lusztig restricted integral form). Inside U q D J ( g ) U_q^{\mathrm{DJ}}(\mathfrak{g}) U q DJ ( g ) , define
U A L u s ( g ) : = A ⟨ E i ( a ) , F i ( a ) , K i ± 1 ∣ i ∈ I , a ≥ 0 ⟩ . U_{\mathcal{A}}^{\mathrm{Lus}}(\mathfrak{g}):=\mathcal{A}\left\langle E_i^{(a)},F_i^{(a)},K_i^{\pm1}\mid i\in I,\ a\geq0\right\rangle. U A Lus ( g ) := A ⟨ E i ( a ) , F i ( a ) , K i ± 1 ∣ i ∈ I , a ≥ 0 ⟩ .
Definition 1.3 (Bare real-current core or integral form in Drinfeld presentation). Inside the fixed ambient algebra U q D r ( g ) U_q^{\mathrm{Dr}}(\mathfrak{g}) U q Dr ( g ) , define, uniformly in all types,
U A D r ( g ) : = A ⟨ K δ ± 1 , K i ± 1 , ( x i , r ± ) ( a ) ∣ i ∈ I 0 , r ∈ I i , a ≥ 0 ⟩ . (39) U_{\mathcal{A}}^{\mathrm{Dr}}(\mathfrak{g}):=\mathcal{A}\left\langle K_\delta^{\pm1},K_i^{\pm1},(x^\pm_{i,r})^{(a)}\mid i\in I_0,\ r\in I_i,\ a\geq0\right\rangle. \tag*{(39)} U A Dr ( g ) := A ⟨ K δ ± 1 , K i ± 1 , ( x i , r ± ) ( a ) ∣ i ∈ I 0 , r ∈ I i , a ≥ 0 ⟩ . (39)
Lemma 1.4. For every g ∈ T \mathfrak{g}\in\mathcal{T} g ∈ T ,
U A D r ( g ) ⊆ U A L u s ( g ) . U_{\mathcal{A}}^{\mathrm{Dr}}(\mathfrak{g})\subseteq U_{\mathcal{A}}^{\mathrm{Lus}}(\mathfrak{g}). U A Dr ( g ) ⊆ U A Lus ( g ) .
Proof. The toral generators in (39) already belong to U A L u s ( g ) U^{\mathrm{Lus}}_{\mathcal{A}}(\mathfrak{g}) U A Lus ( g ) . For i ∈ I 0 i \in I_0 i ∈ I 0 and r ∈ I i r \in I_i r ∈ I i , write r = ν i k r=\nu_i k r = ν i k , where k ∈ Z k \in\mathbb{Z} k ∈ Z . Equation (36) gives, for every a ≥ 0 a \ge0 a ≥ 0 ,
( x i , r + ) ( a ) = o ( i ) k a T λ i − k ( E i ( a ) ) , ( x i , r − ) ( a ) = o ( i ) k a T λ i k ( F i ( a ) ) . (40) \begin{aligned}
(x^+_{i,r})^{(a)} &= o(i)^{ka}T_{\lambda_i}^{-k}(E_i^{(a)}),\\
(x^-_{i,r})^{(a)} &= o(i)^{ka}T_{\lambda_i}^{k}(F_i^{(a)}). \tag*{(40)}
\end{aligned} ( x i , r + ) ( a ) ( x i , r − ) ( a ) = o ( i ) k a T λ i − k ( E i ( a ) ) , = o ( i ) k a T λ i k ( F i ( a ) ) . (40)
Since o ( i ) ∈ { ± 1 } o(i) \in\{\pm1\} o ( i ) ∈ { ± 1 } and the extended braid operators preserve U A L u s ( g ) U^{\mathrm{Lus}}_{\mathcal{A}}(\mathfrak{g}) U A Lus ( g ) , both divided powers belong to U A L u s ( g ) U^{\mathrm{Lus}}_{\mathcal{A}}(\mathfrak{g}) U A Lus ( g ) . This proves the inclusion.
Lemma 1.5 (q-Weyl normal ordering and the highest contraction). Let D D D be an A \mathcal{A} A -subalgebra of a K \mathcal{K} K -algebra, let Q ∈ A × Q \in\mathcal{A}^{\times} Q ∈ A × , and suppose that A A A , B B B , Z Z Z satisfy
Z = A B − B A , A Z = Q − 2 Z A , Z B = Q − 2 B Z . Z=AB-BA,\qquad AZ=Q^{-2}ZA,\qquad ZB=Q^{-2}BZ. Z = A B − B A , A Z = Q − 2 Z A , Z B = Q − 2 B Z . For m ≥ 0 m \ge0 m ≥ 0 , set
A Q ( m ) = A m [ m ] Q ! , B Q ( m ) = B m [ m ] Q ! . A_Q^{(m)}=\frac{A^m}{[m]_Q!},\qquad B_Q^{(m)}=\frac{B^m}{[m]_Q!}. A Q ( m ) = [ m ] Q ! A m , B Q ( m ) = [ m ] Q ! B m . Then, for all M , N ≥ 0 M,N \ge0 M , N ≥ 0 ,
A Q ( M ) B Q ( N ) = ∑ t = 0 min ( M , N ) Q − t ( M + N − 2 t ) − ( t 2 ) B Q ( N − t ) Z t [ t ] Q ! A Q ( M − t ) . (41) A_Q^{(M)}B_Q^{(N)}
=
\sum_{t=0}^{\min(M,N)}
Q^{-t(M+N-2t)-\binom{t}{2}}
B_Q^{(N-t)}
\frac{Z^t}{[t]_Q!}
A_Q^{(M-t)}. \tag*{(41)} A Q ( M ) B Q ( N ) = t = 0 ∑ m i n ( M , N ) Q − t ( M + N − 2 t ) − ( 2 t ) B Q ( N − t ) [ t ] Q ! Z t A Q ( M − t ) . (41) If A Q ( m ) , B Q ( m ) ∈ D A_Q^{(m)},B_Q^{(m)} \in D A Q ( m ) , B Q ( m ) ∈ D for every m ≥ 0 m \ge0 m ≥ 0 , then
Z a [ a ] Q ! ∈ D ( a ≥ 0 ) . \frac{Z^a}{[a]_Q!}\in D\qquad(a\ge0). [ a ] Q ! Z a ∈ D ( a ≥ 0 ) .
Proof. The three defining relations first imply, by induction on N N N , that
A B N = B N A + Q − ( N − 1 ) [ N ] Q B N − 1 Z ( N ≥ 1 ) . (42) AB^N=B^N A+Q^{-(N-1)}[N]_Q B^{N-1}Z\qquad(N\ge1). \tag*{(42)} A B N = B N A + Q − ( N − 1 ) [ N ] Q B N − 1 Z ( N ≥ 1 ) . (42)
The case N = 1 N=1 N = 1 is precisely Z = A B − B A Z=AB-BA Z = A B − B A . Assume that (42) holds for N N N . Multiplication on the right by B B B gives
A B N + 1 = ( B N A + Q − ( N − 1 ) [ N ] Q B N − 1 Z ) B = B N + 1 A + B N Z + Q − ( N − 1 ) [ N ] Q B N − 1 Z B = B N + 1 A + ( 1 + Q − ( N + 1 ) [ N ] Q ) B N Z = B N + 1 A + Q − N [ N + 1 ] Q B N Z . \begin{aligned}
AB^{N+1}
&=(B^N A+Q^{-(N-1)}[N]_Q B^{N-1}Z)B\\
&=B^{N+1}A+B^N Z+Q^{-(N-1)}[N]_Q B^{N-1}ZB\\
&=B^{N+1}A+(1+Q^{-(N+1)}[N]_Q)B^N Z\\
&=B^{N+1}A+Q^{-N}[N+1]_Q B^N Z.
\end{aligned} A B N + 1 = ( B N A + Q − ( N − 1 ) [ N ] Q B N − 1 Z ) B = B N + 1 A + B N Z + Q − ( N − 1 ) [ N ] Q B N − 1 Z B = B N + 1 A + ( 1 + Q − ( N + 1 ) [ N ] Q ) B N Z = B N + 1 A + Q − N [ N + 1 ] Q B N Z .
Here the third equality uses Z B = Q − 2 B Z ZB=Q^{-2}BZ Z B = Q − 2 B Z , and the last one uses
1 + Q − ( N + 1 ) [ N ] Q = Q − N [ N + 1 ] Q . 1+Q^{-(N+1)}[N]_Q=Q^{-N}[N+1]_Q. 1 + Q − ( N + 1 ) [ N ] Q = Q − N [ N + 1 ] Q .
This completes the induction. Dividing (42) by [ N ] Q ! [N]_Q! [ N ] Q ! yields
A B Q ( N ) = B Q ( N ) A + Q − ( N − 1 ) B Q ( N − 1 ) Z . (43) AB_Q^{(N)}=B_Q^{(N)}A+Q^{-(N-1)}B_Q^{(N-1)}Z. \tag*{(43)} A B Q ( N ) = B Q ( N ) A + Q − ( N − 1 ) B Q ( N − 1 ) Z . (43)
We next prove (41) by induction on M M M . The cases M = 0 M=0 M = 0 and M = 1 M=1 M = 1 follow from the definition and (43) , respectively. Put
c M , N , t : = Q − t ( M + N − 2 t ) − ( t 2 ) . c_{M,N,t}:=Q^{-t(M+N-2t)-\binom{t}{2}}. c M , N , t := Q − t ( M + N − 2 t ) − ( 2 t ) .
Assume that
A Q ( M ) B Q ( N ) = ∑ s = 0 min ( M , N ) c M , N , s B Q ( N − s ) Z s [ s ] Q ! A Q ( M − s ) . (44) A_Q^{(M)}B_Q^{(N)}=\sum_{s=0}^{\min(M,N)}c_{M,N,s}B_Q^{(N-s)}\frac{Z^s}{[s]_Q!}A_Q^{(M-s)}. \tag*{(44)} A Q ( M ) B Q ( N ) = s = 0 ∑ m i n ( M , N ) c M , N , s B Q ( N − s ) [ s ] Q ! Z s A Q ( M − s ) . (44)
Since
A Q ( M + 1 ) B Q ( N ) = 1 [ M + 1 ] Q A ( A Q ( M ) B Q ( N ) ) , A_Q^{(M+1)}B_Q^{(N)}=\frac{1}{[M+1]_Q}A\left(A_Q^{(M)}B_Q^{(N)}\right), A Q ( M + 1 ) B Q ( N ) = [ M + 1 ] Q 1 A ( A Q ( M ) B Q ( N ) ) ,
fix an index s s s in (44) . Using (1.29) together with
A Z s = Q − 2 s Z s A , A A Q ( M − s ) = [ M − s + 1 ] Q A Q ( M − s + 1 ) , Z s + 1 [ s ] Q ! = [ s + 1 ] Q Z s + 1 [ s + 1 ] Q ! , AZ^s=Q^{-2s}Z^sA,\qquad AA_Q^{(M-s)}=[M-s+1]_QA_Q^{(M-s+1)},\qquad\frac{Z^{s+1}}{[s]_Q!}=[s+1]_Q\frac{Z^{s+1}}{[s+1]_Q!}, A Z s = Q − 2 s Z s A , A A Q ( M − s ) = [ M − s + 1 ] Q A Q ( M − s + 1 ) , [ s ] Q ! Z s + 1 = [ s + 1 ] Q [ s + 1 ] Q ! Z s + 1 ,
we obtain
A ( B Q ( N − s ) Z s [ s ] Q ! A Q ( M − s ) ) = Q − 2 s [ M − s + 1 ] Q B Q ( N − s ) Z s [ s ] Q ! A Q ( M − s + 1 ) + Q − ( N − s − 1 ) [ s + 1 ] Q B Q ( N − s − 1 ) Z s + 1 [ s + 1 ] Q ! A Q ( M − s ) . A\left(B_Q^{(N-s)}\frac{Z^s}{[s]_Q!}A_Q^{(M-s)}\right)=Q^{-2s}[M-s+1]_QB_Q^{(N-s)}\frac{Z^s}{[s]_Q!}A_Q^{(M-s+1)}
+Q^{-(N-s-1)}[s+1]_QB_Q^{(N-s-1)}\frac{Z^{s+1}}{[s+1]_Q!}A_Q^{(M-s)}. A ( B Q ( N − s ) [ s ] Q ! Z s A Q ( M − s ) ) = Q − 2 s [ M − s + 1 ] Q B Q ( N − s ) [ s ] Q ! Z s A Q ( M − s + 1 ) + Q − ( N − s − 1 ) [ s + 1 ] Q B Q ( N − s − 1 ) [ s + 1 ] Q ! Z s + 1 A Q ( M − s ) .
The second summand is omitted when s = N s=N s = N . Fix the standard term
B Q ( N − t ) Z t [ t ] Q ! A Q ( M + 1 − t ) B_Q^{(N-t)}\frac{Z^t}{[t]_Q!}A_Q^{(M+1-t)} B Q ( N − t ) [ t ] Q ! Z t A Q ( M + 1 − t )
in the desired formula for M + 1 M+1 M + 1 . By the preceding identity, it can arise only from the first summand with s = t s=t s = t , or from the second summand with s = t − 1 s=t-1 s = t − 1 . Its coefficient is therefore
Q − 2 t [ M − t + 1 ] Q c M , N , t + Q − ( N − t ) [ t ] Q c M , N , t − 1 [ M + 1 ] Q , (45) \frac{Q^{-2t}[M-t+1]_Qc_{M,N,t}+Q^{-(N-t)}[t]_Qc_{M,N,t-1}}{[M+1]_Q}, \tag*{(45)} [ M + 1 ] Q Q − 2 t [ M − t + 1 ] Q c M , N , t + Q − ( N − t ) [ t ] Q c M , N , t − 1 , (45)
where an out-of-range term is understood to be zero. Directly from the definition of c M , N , t c_{M,N,t} c M , N , t ,
Q − 2 t c M , N , t = Q − t c M + 1 , N , t , Q − ( N − t ) c M , N , t − 1 = Q M − t + 1 c M + 1 , N , t . Q^{-2t}c_{M,N,t}=Q^{-t}c_{M+1,N,t},\qquad Q^{-(N-t)}c_{M,N,t-1}=Q^{M-t+1}c_{M+1,N,t}. Q − 2 t c M , N , t = Q − t c M + 1 , N , t , Q − ( N − t ) c M , N , t − 1 = Q M − t + 1 c M + 1 , N , t .
Consequently (45) equals
c M + 1 , N , t Q − t [ M − t + 1 ] Q + Q M − t + 1 [ t ] Q [ M + 1 ] Q = c M + 1 , N , t . c_{M+1,N,t}\frac{Q^{-t}[M-t+1]_Q+Q^{M-t+1}[t]_Q}{[M+1]_Q}=c_{M+1,N,t}. c M + 1 , N , t [ M + 1 ] Q Q − t [ M − t + 1 ] Q + Q M − t + 1 [ t ] Q = c M + 1 , N , t .
The last equality is the symmetric quantum-integer identity
[ a + b ] Q = Q − b [ a ] Q + Q a [ b ] Q [a+b]_Q=Q^{-b}[a]_Q+Q^a[b]_Q [ a + b ] Q = Q − b [ a ] Q + Q a [ b ] Q
with a = M − t + 1 a=M-t+1 a = M − t + 1 and b = t b=t b = t . If M < N M<N M < N , the new endpoint t = M + 1 t=M+1 t = M + 1 arises only from the second summand with s = M s=M s = M , and its coefficient is
Q − ( N − M − 1 ) c M , N , M = c M + 1 , N , M + 1 . Q^{-(N-M-1)}c_{M,N,M}=c_{M+1,N,M+1}. Q − ( N − M − 1 ) c M , N , M = c M + 1 , N , M + 1 .
If N ≤ M N\le M N ≤ M , the upper endpoint remains N N N . Thus every boundary term is accounted for, and the induction proves (41) .
It remains to prove the integrality assertion. We use strong induction on a a a . The case a = 0 a=0 a = 0 is immediate. Let a > 0 a>0 a > 0 , and assume that
Z t [ t ] Q ! ∈ D ( 0 ≤ t < a ) . \frac{Z^t}{[t]_Q!}\in D\qquad(0\le t<a). [ t ] Q ! Z t ∈ D ( 0 ≤ t < a ) .
Setting M = N = a M=N=a M = N = a in (41) gives
A Q ( a ) B Q ( a ) = ∑ t = 0 a Q − t ( 2 a − 2 t ) − ( t 2 ) B Q ( a − t ) Z t [ t ] Q ! A Q ( a − t ) . A_Q^{(a)}B_Q^{(a)}=\sum_{t=0}^{a}Q^{-t(2a-2t)-\binom{t}{2}}B_Q^{(a-t)}\frac{Z^t}{[t]_Q!}A_Q^{(a-t)}. A Q ( a ) B Q ( a ) = t = 0 ∑ a Q − t ( 2 a − 2 t ) − ( 2 t ) B Q ( a − t ) [ t ] Q ! Z t A Q ( a − t ) .
For every t < a t<a t < a , the corresponding term belongs to D D D : the middle factor belongs to D D D by the induction hypothesis, the two divided-power factors belong to D D D by assumption, and the displayed power of Q Q Q is a unit of A \mathcal{A} A . The left-hand side also lies in D D D . After subtracting the terms indexed by t < a t<a t < a , the remaining term is
Q − ( a 2 ) Z a [ a ] Q ! ∈ D , Q^{-\binom{a}{2}}\frac{Z^a}{[a]_Q!}\in D, Q − ( 2 a ) [ a ] Q ! Z a ∈ D ,
because A Q ( 0 ) = B Q ( 0 ) = 1 A_Q^{(0)}=B_Q^{(0)}=1 A Q ( 0 ) = B Q ( 0 ) = 1 . Finally, Q ( a 2 ) ∈ A × Q^{\binom{a}{2}}\in\mathcal{A}^{\times} Q ( 2 a ) ∈ A × , and multiplication by this unit gives Z a / [ a ] Q ! ∈ D Z^a/[a]_Q!\in D Z a / [ a ] Q ! ∈ D . This completes the strong induction.
A similar result without the integrality assertion is given in [11 Theorem 3.3(1)] , with different notation and normalization. The lemma isolates the algebraic operation used in both exceptional sufficiency arguments. In Sections 6 and 7, two integral real-root families form a q q q -Weyl cell, and the highest contraction produces the missing affine direction.
Ordinary localizations and the main theorem
If M M M is an A \mathcal{A} A -submodule of the common ambient K \mathcal{K} K -algebra and B = S − 1 A ⊆ K B=S^{-1}\mathcal{A}\subseteq\mathcal{K} B = S − 1 A ⊆ K , write B M BM B M for the image of
B ⊗ A M ⟶ U q D J ( g ) , b ⊗ m ⟼ b m . B\otimes_{\mathcal{A}}M\longrightarrow U_q^{\mathrm{DJ}}(\mathfrak{g}),\qquad b\otimes m\longmapsto bm. B ⊗ A M ⟶ U q DJ ( g ) , b ⊗ m ⟼ bm .
Localizing the inclusion M ↪ U q D J ( g ) M\hookrightarrow U_q^{\mathrm{DJ}}(\mathfrak{g}) M ↪ U q DJ ( g ) gives
S − 1 M ↪ S − 1 U q D J ( g ) = U q D J ( g ) , S^{-1}M\hookrightarrow S^{-1}U_q^{\mathrm{DJ}}(\mathfrak{g})=U_q^{\mathrm{DJ}}(\mathfrak{g}), S − 1 M ↪ S − 1 U q DJ ( g ) = U q DJ ( g ) ,
because localization is flat and every element of S S S acts invertibly on the ambient K \mathcal{K} K -algebra. Thus the displayed multiplication map is injective, and we identify B M BM B M with B ⊗ A M B\otimes_{\mathcal{A}}M B ⊗ A M . The following proposition shows that if the two localized integral forms coincide, then their coefficient rings must also coincide.
Proposition 1.6. Let B 1 , B 2 ⊆ K B_1,B_2\subseteq\mathcal{K} B 1 , B 2 ⊆ K be ordinary localizations of A \mathcal{A} A . If
B 1 U A D r ( g ) = B 2 U A L u s ( g ) B_1U_{\mathcal{A}}^{\mathrm{Dr}}(\mathfrak{g})=B_2U_{\mathcal{A}}^{\mathrm{Lus}}(\mathfrak{g}) B 1 U A Dr ( g ) = B 2 U A Lus ( g ) inside U q D J ( g ) U_q^{\mathrm{DJ}}(\mathfrak{g}) U q DJ ( g ) , then
Proof. Let
ε : U q D J ( g ) ⟶ K \varepsilon:U_q^{\mathrm{DJ}}(\mathfrak{g})\longrightarrow\mathcal{K} ε : U q DJ ( g ) ⟶ K be the counit, and let 1 U 1_U 1 U denote the identity element of U q D J ( g ) U_q^{\mathrm{DJ}}(\mathfrak{g}) U q DJ ( g ) . Since both integral forms contain A 1 U \mathcal{A}1_U A 1 U and the counit takes each of their defining generators into A \mathcal{A} A , we have
ε ( U A D r ( g ) ) = A = ε ( U A L u s ( g ) ) . \varepsilon\bigl(U_{\mathcal{A}}^{\mathrm{Dr}}(\mathfrak{g})\bigr)=\mathcal{A}=\varepsilon\bigl(U_{\mathcal{A}}^{\mathrm{Lus}}(\mathfrak{g})\bigr). ε ( U A Dr ( g ) ) = A = ε ( U A Lus ( g ) ) . Since ε \varepsilon ε is K \mathcal{K} K -linear, it follows that
ε ( B 1 U A D r ( g ) ) = B 1 , ε ( B 2 U A L u s ( g ) ) = B 2 . \varepsilon\bigl(B_1U_{\mathcal{A}}^{\mathrm{Dr}}(\mathfrak{g})\bigr)=B_1,\qquad
\varepsilon\bigl(B_2U_{\mathcal{A}}^{\mathrm{Lus}}(\mathfrak{g})\bigr)=B_2. ε ( B 1 U A Dr ( g ) ) = B 1 , ε ( B 2 U A Lus ( g ) ) = B 2 . Applying ε \varepsilon ε to the assumed equality therefore gives
B 1 = ε ( B 1 U A D r ( g ) ) = ε ( B 2 U A L u s ( g ) ) = B 2 . B_1=\varepsilon\bigl(B_1U_{\mathcal{A}}^{\mathrm{Dr}}(\mathfrak{g})\bigr)=\varepsilon\bigl(B_2U_{\mathcal{A}}^{\mathrm{Lus}}(\mathfrak{g})\bigr)=B_2. B 1 = ε ( B 1 U A Dr ( g ) ) = ε ( B 2 U A Lus ( g ) ) = B 2 .
Thus allowing the two integral forms to be localized over different coefficient rings does not yield any additional comparisons.
Definition 1.7 (Sufficient ordinary localization and least sufficient ring). Let S ⊂ A ∖ { 0 } S \subset A \setminus\{0\} S ⊂ A ∖ { 0 } be a multiplicatively closed set containing 1 1 1 , and put B = S − 1 A ⊂ K B=S^{-1}A\subset\mathcal{K} B = S − 1 A ⊂ K . We call B B B a sufficient ordinary localization for g \mathfrak{g} g if
B U A D r ( g ) = B U A L u s ( g ) . BU^{\mathrm{Dr}}_{A}(\mathfrak{g})=BU^{\mathrm{Lus}}_{A}(\mathfrak{g}). B U A Dr ( g ) = B U A Lus ( g ) . If the poset of sufficient ordinary localizations, ordered by inclusion as subrings of K \mathcal{K} K , has a least element, we denote it by B min ( g ) B_{\min}(\mathfrak{g}) B m i n ( g ) .
Set
B ∗ : = A [ Φ 8 r ( q ) − 1 ∣ r ≥ 1 ] , C ∗ : = A [ Φ 4 r ( q ) − 1 ∣ r ≥ 1 ] . (46) B^{*}:=A[\Phi_{8r}(q)^{-1}\mid r\geq1],\qquad C^{*}:=A[\Phi_{4r}(q)^{-1}\mid r\geq1]. \tag*{(46)} B ∗ := A [ Φ 8 r ( q ) − 1 ∣ r ≥ 1 ] , C ∗ := A [ Φ 4 r ( q ) − 1 ∣ r ≥ 1 ] . (46)
Theorem 1.8 (Uniform least-ring theorem). For every g ∈ T \mathfrak{g}\in\mathcal{T} g ∈ T , the least element in Definition 1.7 exists, and
B min ( g ) = { A , g ∉ { A 2 m ( 2 ) ∣ m ≥ 1 } , C ∗ , g = A 2 ( 2 ) , B ∗ , g = A 2 n ( 2 ) , n ≥ 2. B_{\min}(\mathfrak{g})=
\begin{cases}
\mathcal{A}, & \mathfrak{g}\notin\{A_{2m}^{(2)}\mid m\geq1\},\\
C^{*}, & \mathfrak{g}=A_{2}^{(2)},\\
B^{*}, & \mathfrak{g}=A_{2n}^{(2)},\ n\geq2.
\end{cases} B m i n ( g ) = ⎩ ⎨ ⎧ A , C ∗ , B ∗ , g ∈ / { A 2 m ( 2 ) ∣ m ≥ 1 } , g = A 2 ( 2 ) , g = A 2 n ( 2 ) , n ≥ 2. More precisely , for every ordinary localization B = S − 1 A ⊂ K B=S^{-1}A\subset\mathcal{K} B = S − 1 A ⊂ K ,
B U A D r ( A 2 ( 2 ) ) = B U A L u s ( A 2 ( 2 ) ) ⟺ Φ 4 r ( q ) ∈ B × ( r ≥ 1 ) , BU^{\mathrm{Dr}}_{A}(A_{2}^{(2)})=BU^{\mathrm{Lus}}_{A}(A_{2}^{(2)})
\quad\Longleftrightarrow\quad
\Phi_{4r}(q)\in B^{\times}\quad(r\geq1), B U A Dr ( A 2 ( 2 ) ) = B U A Lus ( A 2 ( 2 ) ) ⟺ Φ 4 r ( q ) ∈ B × ( r ≥ 1 ) , whereas, for n ≥ 2 n\geq2 n ≥ 2 ,
B U A D r ( A 2 n ( 2 ) ) = B U A L u s ( A 2 n ( 2 ) ) ⟺ Φ 8 r ( q ) ∈ B × ( r ≥ 1 ) . BU^{\mathrm{Dr}}_{A}(A_{2n}^{(2)})=BU^{\mathrm{Lus}}_{A}(A_{2n}^{(2)})
\quad\Longleftrightarrow\quad
\Phi_{8r}(q)\in B^{\times}\quad(r\geq1). B U A Dr ( A 2 n ( 2 ) ) = B U A Lus ( A 2 n ( 2 ) ) ⟺ Φ 8 r ( q ) ∈ B × ( r ≥ 1 ) .
The following theorem shows that adjoining D ± 1 D^{\pm1} D ± 1 does not change the ordinary localizations over which the two integral forms coincide.
Theorem 1.9. Let U ^ q D J ( g ) \widehat{U}^{\mathrm{DJ}}_{q}(\mathfrak{g}) U q DJ ( g ) be obtained from U q D J ( g ) U^{\mathrm{DJ}}_{q}(\mathfrak{g}) U q DJ ( g ) by adjoining the degree element D ± 1 = q ± d D^{\pm1}=q^{\pm d} D ± 1 = q ± d , where
D E i D − 1 = q δ i 0 E i , D F i D − 1 = q − δ i 0 F i , DE_iD^{-1}=q^{\delta_{i0}}E_i,\qquad DF_iD^{-1}=q^{-\delta_{i0}}F_i, D E i D − 1 = q δ i 0 E i , D F i D − 1 = q − δ i 0 F i , D x i , r ± D − 1 = q ± r x i , r ± , D K i D − 1 = K i , D K δ D − 1 = K δ . Dx_{i,r}^{\pm}D^{-1}=q^{\pm r}x_{i,r}^{\pm},\qquad D K_iD^{-1}=K_i,\qquad D K_{\delta}D^{-1}=K_{\delta}. D x i , r ± D − 1 = q ± r x i , r ± , D K i D − 1 = K i , D K δ D − 1 = K δ . For ∙ ∈ { D r , L u s } \bullet\in\{\mathrm{Dr},\mathrm{Lus}\} ∙ ∈ { Dr , Lus } , let
U ^ A ∙ ( g ) : = ⟨ U A ∙ ( g ) , D ± 1 ⟩ A . \widehat{U}^{\bullet}_{A}(\mathfrak{g}):=\langle U^{\bullet}_{A}(\mathfrak{g}),D^{\pm1}\rangle_{A}. U A ∙ ( g ) := ⟨ U A ∙ ( g ) , D ± 1 ⟩ A . Then , for every ordinary localization B = S − 1 A ⊂ K B=S^{-1}A\subset\mathcal{K} B = S − 1 A ⊂ K ,
B U ^ A D r ( g ) = B U ^ A L u s ( g ) ⟺ B U A D r ( g ) = B U A L u s ( g ) . B\widehat{U}^{\mathrm{Dr}}_{A}(\mathfrak{g})=B\widehat{U}^{\mathrm{Lus}}_{A}(\mathfrak{g})
\quad\Longleftrightarrow\quad
BU^{\mathrm{Dr}}_{A}(\mathfrak{g})=BU^{\mathrm{Lus}}_{A}(\mathfrak{g}). B U A Dr ( g ) = B U A Lus ( g ) ⟺ B U A Dr ( g ) = B U A Lus ( g ) . Consequently, adjoining D ± 1 D^{\pm1} D ± 1 does not change any conclusion of theorem 1.8; in particular, the least sufficient rings remain A A A , C ∗ C^{*} C ∗ , and B ∗ B^{*} B ∗ in the respective cases listed there.
Proof. Let τ = Ad ( D ) \tau=\operatorname{Ad}(D) τ = Ad ( D ) . The defining relations give
τ ( E i ( a ) ) = q a δ i 0 E i ( a ) , τ ( F i ( a ) ) = q − a δ i 0 F i ( a ) , τ ( ( x i , r ± ) ( a ) ) = q a r ( x i , r ± ) ( a ) , \tau(E_i^{(a)})=q^{a\delta_{i0}}E_i^{(a)},\qquad
\tau(F_i^{(a)})=q^{-a\delta_{i0}}F_i^{(a)},\qquad
\tau((x_{i,r}^{\pm})^{(a)})=q^{ar}(x_{i,r}^{\pm})^{(a)}, τ ( E i ( a ) ) = q a δ i 0 E i ( a ) , τ ( F i ( a ) ) = q − a δ i 0 F i ( a ) , τ (( x i , r ± ) ( a ) ) = q a r ( x i , r ± ) ( a ) , and τ \tau τ fixes all toral generators. Since the displayed powers of q q q belong to A × A^{\times} A × , both integral forms are preserved by τ \tau τ and τ − 1 \tau^{-1} τ − 1 . Hence
U ^ A ∙ ( g ) = ⨁ m ∈ Z U A ∙ ( g ) D m , ∙ ∈ { D r , L u s } , \widehat{U}^{\bullet}_{A}(\mathfrak{g})=\bigoplus_{m\in\mathbb{Z}}U^{\bullet}_{A}(\mathfrak{g})D^{m},\qquad\bullet\in\{\mathrm{Dr},\mathrm{Lus}\}, U A ∙ ( g ) = m ∈ Z ⨁ U A ∙ ( g ) D m , ∙ ∈ { Dr , Lus } ,
and therefore
B ^ A ( g ) = ⨁ m ∈ Z B A U ( g ) D m . \widehat{B}_{A}(\mathfrak{g})=\bigoplus_{m\in\mathbb{Z}}B^{U}_{A}(\mathfrak{g})^{D^{m}}. B A ( g ) = m ∈ Z ⨁ B A U ( g ) D m .
Equality of the degree-zero components proves the forward implication; the reverse implication follows by extending an equality in degree zero to every D D D -degree.
The remainder of the paper is devoted entirely to the proof of Theorem 1.8 .
I The non-A 2 n ( 2 ) A_{2n}^{(2)} A 2 n ( 2 ) twisted types
Root data and the equality theorem in the nonexceptional types
Throughout Part I, fix
g ∈ { A 2 n − 1 ( 2 ) ( n ≥ 3 ) , D n + 1 ( 2 ) ( n ≥ 2 ) , E 6 ( 2 ) , D 4 ( 3 ) } . \mathfrak{g}\in\left\{A_{2n-1}^{(2)}\ (n\geq3),\ D_{n+1}^{(2)}\ (n\geq2),\ E_{6}^{(2)},\ D_{4}^{(3)}\right\}. g ∈ { A 2 n − 1 ( 2 ) ( n ≥ 3 ) , D n + 1 ( 2 ) ( n ≥ 2 ) , E 6 ( 2 ) , D 4 ( 3 ) } .
We use the following numbering of the nodes and the parameters (1) . Write δ = α 0 + θ \delta=\alpha_{0}+\theta δ = α 0 + θ . The following table records the finite root data and fixes the finite node i ∗ i^{*} i ∗ and the Weyl-group element w w w used in § 3.3 .
Table 2.1.
The corresponding affine Dynkin diagrams are displayed in Figure 1 .
Theorem 2.1 (equality in the nonexceptional types). For each of the four families above, after fixing the comparison isomorphism (23) ,
U A D r ( g ) = U A L u s ( g ) inside U q D J ( g ) . U_A^{\mathrm{Dr}}(\mathfrak{g})=U_A^{\mathrm{Lus}}(\mathfrak{g})\quad\text{inside }U_q^{\mathrm{DJ}}(\mathfrak{g}). U A Dr ( g ) = U A Lus ( g ) inside U q DJ ( g ) . Sections 3 and 4 are devoted to the proof of Theorem 2.1 .
Braid stability and the β 0 \beta_0 β 0 -root divided powers
Use the translations in (29) ; for these four types, ν i = d i \nu_i=d_i ν i = d i . Writing ( λ ∣ α ) : = ⟨ λ , α ⟩ (\lambda\mid\alpha):=\langle\lambda,\alpha\rangle ( λ ∣ α ) := ⟨ λ , α ⟩ for the canonical pairing with the finite root lattice, we have
( λ i ∣ α j ) = d i δ i j . (47) (\lambda_i\mid\alpha_j)=d_i\delta_{ij}. \tag*{(47)} ( λ i ∣ α j ) = d i δ ij . (47)
The sign and zero-mode conventions are (35) and (10) .
The current formulas in the proof of Lemma 1.4, together with (28) , give
Ω ( ( x i , d i r ± ) ( a ) ) = ( x i , − d i r ∓ ) ( a ) , Ω ( U A D r ( g ) ) = U A D r ( g ) . (48) \Omega\left((x_{i,d_i r}^{\pm})^{(a)}\right)=(x_{i,-d_i r}^{\mp})^{(a)},\qquad\Omega\left(U_A^{\mathrm{Dr}}(\mathfrak{g})\right)=U_A^{\mathrm{Dr}}(\mathfrak{g}). \tag*{(48)} Ω ( ( x i , d i r ± ) ( a ) ) = ( x i , − d i r ∓ ) ( a ) , Ω ( U A Dr ( g ) ) = U A Dr ( g ) . (48)
Lemma 3.1. For every h ∈ I 0 h\in I_0 h ∈ I 0 ,
T λ h ( U A D r ( g ) ) = U A D r ( g ) . T_{\lambda_h}\left(U_A^{\mathrm{Dr}}(\mathfrak{g})\right)=U_A^{\mathrm{Dr}}(\mathfrak{g}). T λ h ( U A Dr ( g ) ) = U A Dr ( g ) .
Proof. Under the comparison isomorphism Ψ g \Psi_{\mathfrak g} Ψ g , for ϵ ∈ { ± 1 } \epsilon\in\{\pm1\} ϵ ∈ { ± 1 } ,
T λ h ϵ ( K δ ) = K δ , T_{\lambda_h}^{\epsilon}(K_\delta)=K_\delta, T λ h ϵ ( K δ ) = K δ ,
T λ h ϵ ( K j ) = K j K δ − ϵ δ h j d h , T_{\lambda_h}^{\epsilon}(K_j)=K_jK_\delta^{-\epsilon\delta_{hj}d_h}, T λ h ϵ ( K j ) = K j K δ − ϵ δ hj d h ,
T λ h ϵ ( ( x j , s + ) ( a ) ) = o ( h ) a δ h j ( x j , s − ϵ δ h j d h + ) ( a ) , T_{\lambda_h}^{\epsilon}\left((x_{j,s}^{+})^{(a)}\right)=o(h)^{a\delta_{hj}}\left(x_{j,s-\epsilon\delta_{hj}d_h}^{+}\right)^{(a)}, T λ h ϵ ( ( x j , s + ) ( a ) ) = o ( h ) a δ hj ( x j , s − ϵ δ hj d h + ) ( a ) ,
T λ h ϵ ( ( x j , s − ) ( a ) ) = o ( h ) a δ h j ( x j , s + ϵ δ h j d h − ) ( a ) . (49) T_{\lambda_h}^{\epsilon}\left((x_{j,s}^{-})^{(a)}\right)=o(h)^{a\delta_{hj}}\left(x_{j,s+\epsilon\delta_{hj}d_h}^{-}\right)^{(a)}. \tag*{(49)} T λ h ϵ ( ( x j , s − ) ( a ) ) = o ( h ) a δ hj ( x j , s + ϵ δ hj d h − ) ( a ) . (49)
If s ∈ d j Z s\in d_j\mathbb{Z} s ∈ d j Z , then every mode occurring on the right-hand side is again in d j Z d_j\mathbb{Z} d j Z . Hence both T λ h T_{\lambda_h} T λ h and T λ h − 1 T_{\lambda_h}^{-1} T λ h − 1 preserve the generators in (39) , proving the equality.
For j ∈ I 0 j\in I_0 j ∈ I 0 , set
A j : = Z [ q j , q j − 1 ] , β j : = d j δ − α j , K β j : = K δ d j K j − 1 . \mathcal{A}_j:=\mathbb{Z}[q_j,q_j^{-1}],\qquad\beta_j:=d_j\delta-\alpha_j,\qquad K_{\beta_j}:=K_\delta^{d_j}K_j^{-1}. A j := Z [ q j , q j − 1 ] , β j := d j δ − α j , K β j := K δ d j K j − 1 .
By (34) and (33) ,
E β j = T λ j T j − 1 ( E j ) , F β j = Ω ( E β j ) . (50) E_{\beta_j}=T_{\lambda_j}T_j^{-1}(E_j),\qquad F_{\beta_j}=\Omega(E_{\beta_j}). \tag*{(50)} E β j = T λ j T j − 1 ( E j ) , F β j = Ω ( E β j ) . (50)
There is an A 1 ( 1 ) A_1^{(1)} A 1 ( 1 ) -vertex homomorphism [D1][JZ]
ϕ j D J : U q D J ( A 1 ( 1 ) ) ⟶ U q D J ( g ) \phi_j^{\mathrm{DJ}}:U_q^{\mathrm{DJ}}(A_1^{(1)})\longrightarrow U_q^{\mathrm{DJ}}(\mathfrak{g}) ϕ j DJ : U q DJ ( A 1 ( 1 ) ) ⟶ U q DJ ( g )
sending the two source simple pairs ( E 1 , F 1 ) (E_1,F_1) ( E 1 , F 1 ) and ( E 0 , F 0 ) (E_0,F_0) ( E 0 , F 0 ) to ( E j , F j ) (E_j,F_j) ( E j , F j ) and ( E β j , F β j ) (E_{\beta_j},F_{\beta_j}) ( E β j , F β j ) respectively, and the corresponding toral elements to K j K_j K j and K β j K_{\beta_j} K β j .
Choose the sign normalization on the source ordinary vertex so that its finite current has sign o ( j ) o(j) o ( j ) . Write δ ( 1 ) \delta^{(1)} δ ( 1 ) for the null root of the source A 1 ( 1 ) A_1^{(1)} A 1 ( 1 ) . The homomorphism ϕ j D J \phi_j^{\mathrm{DJ}} ϕ j DJ satisfies
ϕ j D J ( K δ ( 1 ) ) = K δ d j , ϕ j D J ( K 1 ) = K j , \phi_j^{\mathrm{DJ}}(K_{\delta^{(1)}})=K_\delta^{d_j},\qquad\phi_j^{\mathrm{DJ}}(K_1)=K_j, ϕ j DJ ( K δ ( 1 ) ) = K δ d j , ϕ j DJ ( K 1 ) = K j ,
ϕ j D J ( x 1 , r ± ) = x j , d j r ± ( r ∈ Z ) , T j ϕ j D J = ϕ j D J T 1 . (51) \phi_j^{\mathrm{DJ}}(x_{1,r}^{\pm})=x_{j,d_jr}^{\pm}\quad(r\in\mathbb{Z}),\qquad T_j\phi_j^{\mathrm{DJ}}=\phi_j^{\mathrm{DJ}}T_1. \tag*{(51)} ϕ j DJ ( x 1 , r ± ) = x j , d j r ± ( r ∈ Z ) , T j ϕ j DJ = ϕ j DJ T 1 . (51)
Here x 1 , r ± x_{1,r}^{\pm} x 1 , r ± and T 1 T_1 T 1 denote, respectively, the real currents and the Beck braid operator in the source U q j J D ( A 1 ( 1 ) ) U_{q_j}^{\mathrm{JD}}(A_1^{(1)}) U q j JD ( A 1 ( 1 ) ) .
Let V j , A V_{j,A} V j , A denote the image of
A ⊗ A j U A j L u s ( A 1 ( 1 ) ) → 1 ⊗ ϕ j D J U q D J ( g ) . A\otimes_{A_j}U_{A_j}^{\mathrm{Lus}}(A_1^{(1)}) \xrightarrow{1\otimes\phi_j^{\mathrm{DJ}}} U_q^{\mathrm{DJ}}(\mathfrak{g}). A ⊗ A j U A j Lus ( A 1 ( 1 ) ) 1 ⊗ ϕ j DJ U q DJ ( g ) .
Lemma 3.2. For every j ∈ I 0 j\in I_0 j ∈ I 0 ,
V j , A ⊆ U A D r ( g ) , V_{j,A}\subseteq U_A^{\mathrm{Dr}}(\mathfrak{g}), V j , A ⊆ U A Dr ( g ) , and V j , A V_{j,A} V j , A contains
( x j , d j r ± ) ( a ) ( r ∈ Z , a ≥ 0 ) . \left(x_{j,d_jr}^{\pm}\right)^{(a)}\qquad(r\in\mathbb{Z},\ a\geq0). ( x j , d j r ± ) ( a ) ( r ∈ Z , a ≥ 0 ) .
Proof. For r ∈ Z r\in\mathbb{Z} r ∈ Z and a ≥ 0 a\geq0 a ≥ 0 , the source A 1 ( 1 ) A_1^{(1)} A 1 ( 1 ) current formulas give
( x 1 , r + ) ( a ) = o ( j ) r a T λ 1 − r ( E 1 ( a ) ) , ( x 1 , r − ) ( a ) = o ( j ) r a T λ 1 − r ( F 1 ( a ) ) . \left(x_{1,r}^{+}\right)^{(a)}=o(j)^{ra}T_{\lambda_1}^{-r}\left(E_1^{(a)}\right),\qquad
\left(x_{1,r}^{-}\right)^{(a)}=o(j)^{ra}T_{\lambda_1}^{-r}\left(F_1^{(a)}\right). ( x 1 , r + ) ( a ) = o ( j ) r a T λ 1 − r ( E 1 ( a ) ) , ( x 1 , r − ) ( a ) = o ( j ) r a T λ 1 − r ( F 1 ( a ) ) .
The braid-integrality statement (24) , applied over A j A_j A j , shows that Beck’s braid operators and their inverses preserve U A j L u s ( A 1 ( 1 ) ) U_{A_j}^{\mathrm{Lus}}(A_1^{(1)}) U A j Lus ( A 1 ( 1 ) ) . Hence
( x 1 , r ± ) ( a ) ∈ U A j L u s ( A 1 ( 1 ) ) . \left(x_{1,r}^{\pm}\right)^{(a)}\in U_{A_j}^{\mathrm{Lus}}(A_1^{(1)}). ( x 1 , r ± ) ( a ) ∈ U A j Lus ( A 1 ( 1 ) ) .
Equation (51) now gives
ϕ j D J ( ( x 1 , r ± ) ( a ) ) = ( x j , d j r ± ) ( a ) ∈ V j , A . \phi_j^{\mathrm{DJ}}\left(\left(x_{1,r}^{\pm}\right)^{(a)}\right)=\left(x_{j,d_jr}^{\pm}\right)^{(a)}\in V_{j,A}. ϕ j DJ ( ( x 1 , r ± ) ( a ) ) = ( x j , d j r ± ) ( a ) ∈ V j , A .
It remains to prove V j , A ⊆ U A D r ( g ) V_{j,A}\subseteq U_A^{\mathrm{Dr}}(\mathfrak{g}) V j , A ⊆ U A Dr ( g ) . For the source finite simple generators, one has
ϕ j D J ( E 1 ( a ) ) = E j ( a ) = ( x j , 0 + ) ( a ) , ϕ j D J ( F 1 ( a ) ) = F j ( a ) = ( x j , 0 − ) ( a ) . \phi_j^{\mathrm{DJ}}\left(E_1^{(a)}\right)=E_j^{(a)}=\left(x_{j,0}^{+}\right)^{(a)},\qquad
\phi_j^{\mathrm{DJ}}\left(F_1^{(a)}\right)=F_j^{(a)}=\left(x_{j,0}^{-}\right)^{(a)}. ϕ j DJ ( E 1 ( a ) ) = E j ( a ) = ( x j , 0 + ) ( a ) , ϕ j DJ ( F 1 ( a ) ) = F j ( a ) = ( x j , 0 − ) ( a ) .
Hence their images belong to U A D r ( g ) U_A^{\mathrm{Dr}}(\mathfrak{g}) U A Dr ( g ) . For the other positive simple generator, (49) , (50) , and T j − 1 ( E j ) = − K j − 1 F j T_j^{-1}(E_j)=-K_j^{-1}F_j T j − 1 ( E j ) = − K j − 1 F j give
E β j = T λ j T j − 1 ( E j ) = − T λ j ( K j − 1 ) T λ j ( F j ) = − o ( j ) K β j x j , d j − . \begin{aligned}
E_{\beta_j}&=T_{\lambda_j}T_j^{-1}(E_j)\\
&=-T_{\lambda_j}(K_j^{-1})T_{\lambda_j}(F_j)\\
&=-o(j)K_{\beta_j}x_{j,d_j}^{-}.
\end{aligned} E β j = T λ j T j − 1 ( E j ) = − T λ j ( K j − 1 ) T λ j ( F j ) = − o ( j ) K β j x j , d j − .
Moreover,
K β j x j , d j − = q j 2 x j , d j − K β j . K_{\beta_j}x_{j,d_j}^{-}=q_j^2x_{j,d_j}^{-}K_{\beta_j}. K β j x j , d j − = q j 2 x j , d j − K β j .
Consequently,
E β j ( a ) = ( − o ( j ) ) a q j − a ( a − 1 ) K β j a ( x j , d j − ) ( a ) ∈ U A D r ( g ) . E_{\beta_j}^{(a)}=(-o(j))^a q_j^{-a(a-1)}K_{\beta_j}^a\left(x_{j,d_j}^{-}\right)^{(a)}\in U_A^{\mathrm{Dr}}(\mathfrak{g}). E β j ( a ) = ( − o ( j ) ) a q j − a ( a − 1 ) K β j a ( x j , d j − ) ( a ) ∈ U A Dr ( g ) .
Equation (48) gives
F β j ( a ) = Ω ( E β j ( a ) ) ∈ U A D r ( g ) . F_{\beta_j}^{(a)}=\Omega\left(E_{\beta_j}^{(a)}\right)\in U_A^{\mathrm{Dr}}(\mathfrak{g}). F β j ( a ) = Ω ( E β j ( a ) ) ∈ U A Dr ( g ) .
Finally, the source toral generators are sent to K j ± 1 K_j^{\pm1} K j ± 1 and K β j ± 1 K_{\beta_j}^{\pm1} K β j ± 1 , which already belong to U A D r ( g ) U_A^{\mathrm{Dr}}(\mathfrak{g}) U A Dr ( g ) . Thus the images of all defining generators of the source Lusztig integral form lie in U A D r ( g ) U_A^{\mathrm{Dr}}(\mathfrak{g}) U A Dr ( g ) , proving the inclusion. ∎
Let
U A , 0 L u s : = A ⟨ E j ( a ) , F j ( a ) , K j ± 1 ∣ j ∈ I 0 , a ≥ 0 ⟩ U^{\mathrm{Lus}}_{A,0}:=A\left\langle E_j^{(a)},F_j^{(a)},K_j^{\pm1}\mid j\in I_0,\ a\geq0\right\rangle U A , 0 Lus := A ⟨ E j ( a ) , F j ( a ) , K j ± 1 ∣ j ∈ I 0 , a ≥ 0 ⟩
be the finite Lusztig integral form. By (40) with k = 0 k=0 k = 0 ,
E j ( a ) = ( x j , 0 + ) ( a ) , F j ( a ) = ( x j , 0 − ) ( a ) ( j ∈ I 0 , a ≥ 0 ) . E_j^{(a)}=(x^+_{j,0})^{(a)},\qquad F_j^{(a)}=(x^-_{j,0})^{(a)}\qquad(j\in I_0,\ a\geq0). E j ( a ) = ( x j , 0 + ) ( a ) , F j ( a ) = ( x j , 0 − ) ( a ) ( j ∈ I 0 , a ≥ 0 ) .
Together with the defining toral generators K j ± 1 K_j^{\pm1} K j ± 1 , this gives
U A , 0 L u s ⊆ U A D r ( g ) . U^{\mathrm{Lus}}_{A,0}\subseteq U^{\mathrm{Dr}}_A(\mathfrak{g}). U A , 0 Lus ⊆ U A Dr ( g ) .
Proposition 3.3 (Stability under finite braid operators). For every i ∈ I 0 i\in I_0 i ∈ I 0 ,
T i ε ( U A D r ( g ) ) = U A D r ( g ) . T_i^\varepsilon\left(U^{\mathrm{Dr}}_A(\mathfrak{g})\right)=U^{\mathrm{Dr}}_A(\mathfrak{g}). T i ε ( U A Dr ( g ) ) = U A Dr ( g ) . Proof. We prove that T i ε T_i^\varepsilon T i ε preserves U A D r ( g ) U^{\mathrm{Dr}}_A(\mathfrak{g}) U A Dr ( g ) for ε ∈ { ± 1 } \varepsilon\in\{\pm1\} ε ∈ { ± 1 } . First consider the currents of color i i i . Equation (51) and its inverse give
T i ε Φ i D J = Φ i D J T 1 ε ( ε ∈ { ± 1 } ) , T_i^\varepsilon\Phi_i^{\mathrm{DJ}}=\Phi_i^{\mathrm{DJ}}T_1^\varepsilon\qquad(\varepsilon\in\{\pm1\}), T i ε Φ i DJ = Φ i DJ T 1 ε ( ε ∈ { ± 1 }) , where T 1 T_1 T 1 denotes the corresponding braid operator at the finite node of the source U q i D J ( A 1 ( 1 ) ) U_{q_i}^{\mathrm{DJ}}(A_1^{(1)}) U q i DJ ( A 1 ( 1 ) ) . Since
T 1 ε ( U A i L u s ( A 1 ( 1 ) ) ) = U A i L u s ( A 1 ( 1 ) ) , T_1^\varepsilon\left(U_{A_i}^{\mathrm{Lus}}(A_1^{(1)})\right)=U_{A_i}^{\mathrm{Lus}}(A_1^{(1)}), T 1 ε ( U A i Lus ( A 1 ( 1 ) ) ) = U A i Lus ( A 1 ( 1 ) ) , it follows that
T i ε ( V i , A ) = V i , A . T_i^\varepsilon(V_{i,A})=V_{i,A}. T i ε ( V i , A ) = V i , A . Lemma 3.2 therefore handles all divided real currents of color i i i .
Now let j ≠ i j\ne i j = i . By (47) and (29) ,
t λ j ( α i ) = α i , s i t λ j = t λ j s i . t_{\lambda_j}(\alpha_i)=\alpha_i,\qquad s_i t_{\lambda_j}=t_{\lambda_j}s_i. t λ j ( α i ) = α i , s i t λ j = t λ j s i . Since
t λ j ± 1 ( α i ) = α i > 0 , t_{\lambda_j}^{\pm1}(\alpha_i)=\alpha_i>0, t λ j ± 1 ( α i ) = α i > 0 , the corresponding factorizations are length additive. Hence the standard braid lifts commute, and therefore
T i ε T λ j r = T λ j r T i ε ( ε ∈ { ± 1 } , r ∈ Z ) . (52) T_i^\varepsilon T_{\lambda_j}^r=T_{\lambda_j}^rT_i^\varepsilon\qquad(\varepsilon\in\{\pm1\},\ r\in\mathbb{Z}). \tag*{(52)} T i ε T λ j r = T λ j r T i ε ( ε ∈ { ± 1 } , r ∈ Z ) . (52) Applying [21 Section 41.1.2] to the finite Cartan datum ( a j k ) j , k ∈ I 0 (a_{jk})_{j,k\in I_0} ( a j k ) j , k ∈ I 0 , we obtain
T i ε ( U A , 0 L u s ) = U A , 0 L u s ( ε ∈ { ± 1 } ) . T_i^\varepsilon\left(U_{A,0}^{\mathrm{Lus}}\right)=U_{A,0}^{\mathrm{Lus}}\qquad(\varepsilon\in\{\pm1\}). T i ε ( U A , 0 Lus ) = U A , 0 Lus ( ε ∈ { ± 1 }) . Consequently,
T i ε ( E j ( a ) ) ∈ U A , 0 L u s ⊆ U A D r ( g ) . T_i^\varepsilon(E_j^{(a)})\in U_{A,0}^{\mathrm{Lus}}\subseteq U_A^{\mathrm{Dr}}(\mathfrak{g}). T i ε ( E j ( a ) ) ∈ U A , 0 Lus ⊆ U A Dr ( g ) . Using (40) , (52) , and Lemma 3.1, we obtain
T i ε ( ( x j , d j r + ) ( a ) ) = o ( j ) r T i ε T λ j − r ( E j ( a ) ) = o ( j ) r T λ j − r ( T i ε ( E j ( a ) ) ) ∈ U A D r ( g ) . \begin{aligned}
T_i^\varepsilon\left((x^+_{j,d_jr})^{(a)}\right)
&=o(j)^rT_i^\varepsilon T_{\lambda_j}^{-r}(E_j^{(a)})\\
&=o(j)^rT_{\lambda_j}^{-r}\left(T_i^\varepsilon(E_j^{(a)})\right)\\
&\in U_A^{\mathrm{Dr}}(\mathfrak{g}).
\end{aligned} T i ε ( ( x j , d j r + ) ( a ) ) = o ( j ) r T i ε T λ j − r ( E j ( a ) ) = o ( j ) r T λ j − r ( T i ε ( E j ( a ) ) ) ∈ U A Dr ( g ) . The corresponding assertion for the negative currents follows from
Ω T i ε = T i ε Ω \Omega T_i^\varepsilon=T_i^\varepsilon\Omega Ω T i ε = T i ε Ω and (48) .
Finally, T i ε T_i^\varepsilon T i ε fixes K δ K_\delta K δ and sends every K j ± 1 K_j^{\pm1} K j ± 1 to a Laurent monomial in the finite toral generators. Thus both T i T_i T i and T i − 1 T_i^{-1} T i − 1 preserve U A D r ( g ) U_A^{\mathrm{Dr}}(\mathfrak{g}) U A Dr ( g ) , proving the equality.
Fix the node i ∗ i^* i ∗ and the element w w w in (2.1). The table gives
d i ∗ = d 0 = 1. (53) d_{i^*}=d_0=1. \tag*{(53)} d i ∗ = d 0 = 1. (53)
Set
β 0 : = δ − α i ∗ = β i ∗ . \beta_0:=\delta-\alpha_{i^*}=\beta_{i^*}. β 0 := δ − α i ∗ = β i ∗ .
The vertex homomorphism ϕ i ∗ D J \phi^{{\rm DJ}}_{i^*} ϕ i ∗ DJ sends the affine simple positive generator E 0 E_0 E 0 of U q i ∗ D J ( A 1 ( 1 ) ) U^{{\rm DJ}}_{q_{i^*}}(A_1^{(1)}) U q i ∗ DJ ( A 1 ( 1 ) ) to E β i ∗ = E β 0 E_{\beta_{i^*}}=E_{\beta_0} E β i ∗ = E β 0 . Consequently,
( 1 ⊗ ϕ i ∗ D J ) ( 1 ⊗ E 0 ( a ) ) = E β 0 ( a ) . (1\otimes\phi^{{\rm DJ}}_{i^*})(1\otimes E_0^{(a)})=E_{\beta_0}^{(a)}. ( 1 ⊗ ϕ i ∗ DJ ) ( 1 ⊗ E 0 ( a ) ) = E β 0 ( a ) .
Lemma 3.2 therefore gives
E β 0 ( a ) ∈ V i ∗ , A ⊆ U A Dr ( g ) ( a ≥ 0 ) . (54) E_{\beta_0}^{(a)}\in V_{i^*,A}\subseteq U_A^{\operatorname{Dr}}(\mathfrak{g})\qquad(a\geq0). \tag*{(54)} E β 0 ( a ) ∈ V i ∗ , A ⊆ U A Dr ( g ) ( a ≥ 0 ) . (54)
Recovery of the affine node
Lemma 4.1. For each row of (2.1), the displayed Weyl-group element satisfies
w ( α i ∗ ) = θ . w(\alpha_{i^*})=\theta. w ( α i ∗ ) = θ .
Proof. We verify the four types separately. In type A 2 n − 1 ( 2 ) A_{2n-1}^{(2)} A 2 n − 1 ( 2 ) , use the coordinate realization
α 1 = 2 e 1 , α r = e r − e r − 1 ( 2 ≤ r ≤ n ) . \alpha_1=2e_1,\qquad\alpha_r=e_r-e_{r-1}\qquad(2\leq r\leq n). α 1 = 2 e 1 , α r = e r − e r − 1 ( 2 ≤ r ≤ n ) .
Here θ = e n − 1 + e n \theta=e_{n-1}+e_n θ = e n − 1 + e n . Applying the word in (2.1) from right to left to α n = e n − e n − 1 \alpha_n=e_n-e_{n-1} α n = e n − e n − 1 gives
e n − e n − 1 → s n − 1 e n − e n − 2 → ⋯ → s 2 e n − e 1 → s 1 e n + e 1 → s 2 ⋯ → s n − 1 e n + e n − 1 . e_n-e_{n-1}\xrightarrow{s_{n-1}}e_n-e_{n-2}\to\cdots\xrightarrow{s_2}e_n-e_1\xrightarrow{s_1}e_n+e_1\xrightarrow{s_2}\cdots\xrightarrow{s_{n-1}}e_n+e_{n-1}. e n − e n − 1 s n − 1 e n − e n − 2 → ⋯ s 2 e n − e 1 s 1 e n + e 1 s 2 ⋯ s n − 1 e n + e n − 1 .
Thus w ( α n ) = θ w(\alpha_n)=\theta w ( α n ) = θ .
In type D n + 1 ( 2 ) D_{n+1}^{(2)} D n + 1 ( 2 ) , take
α 1 = e 1 , α r = e r − e r − 1 ( 2 ≤ r ≤ n ) . \alpha_1=e_1,\qquad\alpha_r=e_r-e_{r-1}\qquad(2\leq r\leq n). α 1 = e 1 , α r = e r − e r − 1 ( 2 ≤ r ≤ n ) .
Now θ = e n \theta=e_n θ = e n , and
e 1 → s 2 e 2 → s 3 ⋯ → s n e n . e_1\xrightarrow{s_2}e_2\xrightarrow{s_3}\cdots\xrightarrow{s_n}e_n. e 1 s 2 e 2 s 3 ⋯ s n e n .
Therefore w ( α 1 ) = θ w(\alpha_1)=\theta w ( α 1 ) = θ .
For type E 6 ( 2 ) E_6^{(2)} E 6 ( 2 ) , the finite Cartan matrix and an F 4 F_4 F 4 coordinate realization are
( 2 − 1 0 0 − 1 2 − 2 0 0 − 1 2 − 1 0 0 − 1 2 ) , α 1 = 1 2 ( e 1 − e 2 − e 3 − e 4 ) , α 2 = e 4 , α 3 = e 3 − e 4 , α 4 = e 2 − e 3 . \begin{pmatrix}
2 & -1 & 0 & 0\\
-1 & 2 & -2 & 0\\
0 & -1 & 2 & -1\\
0 & 0 & -1 & 2
\end{pmatrix},
\qquad
\alpha_1=\frac{1}{2}(e_1-e_2-e_3-e_4),\quad
\alpha_2=e_4,\quad
\alpha_3=e_3-e_4,\quad
\alpha_4=e_2-e_3. 2 − 1 0 0 − 1 2 − 1 0 0 − 2 2 − 1 0 0 − 1 2 , α 1 = 2 1 ( e 1 − e 2 − e 3 − e 4 ) , α 2 = e 4 , α 3 = e 3 − e 4 , α 4 = e 2 − e 3 .
Here θ = e 1 = ( 2 , 3 , 2 , 1 ) \theta=e_1=(2,3,2,1) θ = e 1 = ( 2 , 3 , 2 , 1 ) in simple-root coordinates. Beginning with α 1 \alpha_1 α 1 and applying, from right to left, s 2 , s 3 , s 2 , s 4 , s 3 , s 2 , s 1 s_2,s_3,s_2,s_4,s_3,s_2,s_1 s 2 , s 3 , s 2 , s 4 , s 3 , s 2 , s 1 , the successive simple-root coordinates are written in the compact form a 1 a 2 a 3 a 4 = a 1 α 1 + ⋯ + a 4 α 4 a_1a_2a_3a_4=a_1\alpha_1+\cdots+a_4\alpha_4 a 1 a 2 a 3 a 4 = a 1 α 1 + ⋯ + a 4 α 4 :
Table 1.
Thus w ( α 1 ) = θ w(\alpha_1)=\theta w ( α 1 ) = θ .
Finally, in type D 4 ( 3 ) D^{(3)}_4 D 4 ( 3 ) , the finite G 2 G_2 G 2 Cartan matrix is
( 2 − 3 − 1 2 ) , \begin{pmatrix} 2 & -3 \\ -1 & 2 \end{pmatrix}, ( 2 − 1 − 3 2 ) ,
with α 1 \alpha_1 α 1 short and α 2 \alpha_2 α 2 long. Direct calculation gives
α 1 → s 2 α 1 + α 2 → s 1 2 α 1 + α 2 = θ . \alpha_1 \xrightarrow{s_2} \alpha_1+\alpha_2 \xrightarrow{s_1} 2\alpha_1+\alpha_2=\theta. α 1 s 2 α 1 + α 2 s 1 2 α 1 + α 2 = θ .
Again w ( α 1 ) = θ w(\alpha_1)=\theta w ( α 1 ) = θ .
Lemma 4.2. Fix the node i = i ∗ i=i^* i = i ∗ and the Weyl-group element w w w from (2.1), and set
y = w t λ i , w ^ = y s i , w ^ 0 = t λ i s i . y=wt_{\lambda_i},\qquad\widehat{w}=ys_i,\qquad\widehat{w}_0=t_{\lambda_i}s_i. y = w t λ i , w = y s i , w 0 = t λ i s i . Then
ℓ ( w t λ i ) = ℓ ( w ) + ℓ ( t λ i ) . (55) \ell(wt_{\lambda_i})=\ell(w)+\ell(t_{\lambda_i}). \tag*{(55)} ℓ ( w t λ i ) = ℓ ( w ) + ℓ ( t λ i ) . (55) ℓ ( w ^ ) = ℓ ( w ) + ℓ ( t λ i ) − 1. (56) \ell(\widehat{w})=\ell(w)+\ell(t_{\lambda_i})-1. \tag*{(56)} ℓ ( w ) = ℓ ( w ) + ℓ ( t λ i ) − 1. (56) T w ^ = T w T λ i T i − 1 , w ^ ( α i ) = α 0 . (57) \begin{aligned}
T_{\widehat{w}}&=T_wT_{\lambda_i}T_i^{-1},&
\widehat{w}(\alpha_i)&=\alpha_0. \tag*{(57)}
\end{aligned} T w = T w T λ i T i − 1 , w ( α i ) = α 0 . (57) T w ^ 0 = T λ i T i − 1 . (58) T_{\widehat{w}_0}=T_{\lambda_i}T_i^{-1}. \tag*{(58)} T w 0 = T λ i T i − 1 . (58) In particular,
T w ( E β 0 ) = T w ^ ( E i ) ∈ ( U q + ) α 0 . (59) T_w(E_{\beta_0})=T_{\widehat{w}}(E_i)\in(U_q^+)_{\alpha_0}. \tag*{(59)} T w ( E β 0 ) = T w ( E i ) ∈ ( U q + ) α 0 . (59)
Proof. By Lemma 4.1, w ( α i ) = θ w(\alpha_i)=\theta w ( α i ) = θ . Choose a reduced expression w = s j L ⋯ s j 1 w=s_{j_L}\cdots s_{j_1} w = s j L ⋯ s j 1 , and put w r = s j r ⋯ s j 1 w_r=s_{j_r}\cdots s_{j_1} w r = s j r ⋯ s j 1 , with w 0 = 1 w_0=1 w 0 = 1 . For r ≥ 1 r\ge1 r ≥ 1 ,
γ r = w r − 1 − 1 ( α j r ) \gamma_r=w_{r-1}^{-1}(\alpha_{j_r}) γ r = w r − 1 − 1 ( α j r )
is a positive finite root. If γ r = ∑ h c h α h \gamma_r=\sum_h c_h\alpha_h γ r = ∑ h c h α h , then ( λ i ∣ γ r ) = d i c i ≥ 0 (\lambda_i\mid\gamma_r)=d_ic_i\ge0 ( λ i ∣ γ r ) = d i c i ≥ 0 , and hence
( w r − 1 t λ i ) − 1 ( α j r ) = t − λ i ( γ r ) = γ r + ( λ i ∣ γ r ) δ > 0. (w_{r-1}t_{\lambda_i})^{-1}(\alpha_{j_r})=t_{-\lambda_i}(\gamma_r)=\gamma_r+(\lambda_i\mid\gamma_r)\delta>0. ( w r − 1 t λ i ) − 1 ( α j r ) = t − λ i ( γ r ) = γ r + ( λ i ∣ γ r ) δ > 0.
Consequently,
ℓ ( w r t λ i ) = ℓ ( w r − 1 t λ i ) + 1 ( 1 ≤ r ≤ L ) . \ell(w_rt_{\lambda_i})=\ell(w_{r-1}t_{\lambda_i})+1\qquad(1\le r\le L). ℓ ( w r t λ i ) = ℓ ( w r − 1 t λ i ) + 1 ( 1 ≤ r ≤ L ) .
Induction from w 0 = 1 w_0=1 w 0 = 1 gives (55) . Since d i = 1 d_i=1 d i = 1 and w ( α i ) = θ w(\alpha_i)=\theta w ( α i ) = θ ,
y ( α i ) = w ( α i − δ ) = θ − δ = − α 0 < 0. y(\alpha_i)=w(\alpha_i-\delta)=\theta-\delta=-\alpha_0<0. y ( α i ) = w ( α i − δ ) = θ − δ = − α 0 < 0.
Thus
ℓ ( y s i ) = ℓ ( y ) − 1. \ell(ys_i)=\ell(y)-1. ℓ ( y s i ) = ℓ ( y ) − 1.
So y = w ^ s i y=\widehat{w}s_i y = w s i and ℓ ( y ) = ℓ ( w ^ ) + 1 \ell(y)=\ell(\widehat{w})+1 ℓ ( y ) = ℓ ( w ) + 1 . This proves (56) . The multiplication rule for standard braid lifts gives
T w T λ i = T y = T w ^ T i , T_wT_{\lambda_i}=T_y=T_{\widehat{w}}T_i, T w T λ i = T y = T w T i ,
and hence the braid identity in (57) . Moreover,
w ^ ( α i ) = w t λ i ( − α i ) = w ( δ − α i ) = δ − θ = α 0 . \widehat{w}(\alpha_i)=wt_{\lambda_i}(-\alpha_i)=w(\delta-\alpha_i)=\delta-\theta=\alpha_0. w ( α i ) = w t λ i ( − α i ) = w ( δ − α i ) = δ − θ = α 0 .
Similarly,
t λ i ( α i ) = α i − δ = − β 0 < 0. t_{\lambda_i}(\alpha_i)=\alpha_i-\delta=-\beta_0<0. t λ i ( α i ) = α i − δ = − β 0 < 0.
Hence t λ i = w ^ s i t_{\lambda_i} = \widehat{w}s_i t λ i = w s i is a length-additive factorization, which proves (58) .
Finally, w ^ ( α i ) = α 0 \widehat{w}(\alpha_i) = \alpha_0 w ( α i ) = α 0 , so Lemma 1.1 gives
T w ^ ( E i ) = E 0 ∈ ( U q + ) α 0 . T_{\widehat{w}}(E_i) = E_0 \in(U_q^+)_{\alpha_0}. T w ( E i ) = E 0 ∈ ( U q + ) α 0 .
By (58) , E β 0 = T w ^ 0 ( E i ) E_{\beta_0} = T_{\widehat{w}_0}(E_i) E β 0 = T w 0 ( E i ) , and hence
T w ( E β 0 ) = T w T λ i T λ i − 1 ( E i ) = T w ^ ( E i ) . T_w(E_{\beta_0}) = T_wT_{\lambda_i}T_{\lambda_i}^{-1}(E_i) = T_{\widehat{w}}(E_i). T w ( E β 0 ) = T w T λ i T λ i − 1 ( E i ) = T w ( E i ) .
This proves (59) .
Proposition 4.3. For every a ≥ 0 a \ge0 a ≥ 0 , one has
T w ( E β 0 ( a ) ) = E 0 ( a ) ∈ U A D r . (60) T_w(E_{\beta_0}^{(a)}) = E_0^{(a)} \in U_A^{\mathrm{Dr}}. \tag*{(60)} T w ( E β 0 ( a ) ) = E 0 ( a ) ∈ U A Dr . (60)
Proof. By Lemma 4.2,
T w ( E β 0 ) = T w ^ ( E i ∗ ) , w ^ ( α i ∗ ) = α 0 . T_w(E_{\beta_0}) = T_{\widehat{w}}(E_{i_*}), \qquad\widehat{w}(\alpha_{i_*}) = \alpha_0. T w ( E β 0 ) = T w ( E i ∗ ) , w ( α i ∗ ) = α 0 .
Lemma 1.1 therefore gives
T w ( E β 0 ) = T w ^ ( E i ∗ ) = E 0 . (61) T_w(E_{\beta_0}) = T_{\widehat{w}}(E_{i_*}) = E_0. \tag*{(61)} T w ( E β 0 ) = T w ( E i ∗ ) = E 0 . (61)
By (54) , E β 0 ( a ) ∈ U A D r E_{\beta_0}^{(a)} \in U_A^{\mathrm{Dr}} E β 0 ( a ) ∈ U A Dr for every a ≥ 0 a \ge0 a ≥ 0 . Applying Proposition 3.3 along a reduced expression for w w w , we obtain
T w ( E β 0 ( a ) ) ∈ U A D r ( a ≥ 0 ) . (62) T_w(E_{\beta_0}^{(a)}) \in U_A^{\mathrm{Dr}} \qquad(a \ge0). \tag*{(62)} T w ( E β 0 ( a ) ) ∈ U A Dr ( a ≥ 0 ) . (62)
Since (53) gives q i ∗ = q 0 = q q_{i_*} = q_0 = q q i ∗ = q 0 = q , it follows that
T w ( E β 0 ( a ) ) = T w ( E β 0 ) a [ a ] q 0 ! = E 0 ( a ) ∈ U A D r ( a ≥ 0 ) . T_w(E_{\beta_0}^{(a)}) = \frac{T_w(E_{\beta_0})^a}{[a]_{q_0}!} = E_0^{(a)} \in U_A^{\mathrm{Dr}} \qquad(a \ge0). T w ( E β 0 ( a ) ) = [ a ] q 0 ! T w ( E β 0 ) a = E 0 ( a ) ∈ U A Dr ( a ≥ 0 ) .
Proof of Theorem 2.1. Proposition 4.3 gives E 0 ( a ) ∈ U A D r E_0^{(a)} \in U_A^{\mathrm{Dr}} E 0 ( a ) ∈ U A Dr for every a ≥ 0. A p p l y i n g a \ge0. Applying a ≥ 0. A ppl y in g \Omega$ gives F 0 ( a ) ∈ U A D r F_0^{(a)} \in U_A^{\mathrm{Dr}} F 0 ( a ) ∈ U A Dr . The divided powers at every finite simple node, as well as the elements K j ± 1 K_j^{\pm1} K j ± 1 , already belong to U A D r U_A^{\mathrm{Dr}} U A Dr by the zero-current modes and its definition. The identity (38) and its inverse expression
K 0 − 1 = K δ − 1 ∏ i ∈ I 0 K i r i K_0^{-1} = K_\delta^{-1}\prod_{i\in I_0} K_i^{r_i} K 0 − 1 = K δ − 1 i ∈ I 0 ∏ K i r i
show that K 0 ± 1 ∈ U A D r K_0^{\pm1} \in U_A^{\mathrm{Dr}} K 0 ± 1 ∈ U A Dr . Hence
U A L u s ( g ) ⊆ U A D r ( g ) . U_A^{\mathrm{Lus}}(\mathfrak{g}) \subseteq U_A^{\mathrm{Dr}}(\mathfrak{g}). U A Lus ( g ) ⊆ U A Dr ( g ) .
The reverse inclusion is by Lemma 1.4. Therefore the two integral forms are equal.
Part II
The twisted type A 2 n ( 2 ) A_{2n}^{(2)} A 2 n ( 2 ) and cyclotomic minimality
In type A 2 n ( 2 ) A_{2n}^{(2)} A 2 n ( 2 ) we need to use a different method to reach the affine simple root α 0 \alpha_0 α 0 because α 0 \alpha_0 α 0 is strictly longer than every finite simple root. As Weyl-group transport preserves root length, the method of Part I does not work.
Root data and real currents in type A 2 n ( 2 ) A_{2n}^{(2)} A 2 n ( 2 )
From this section onward, fix n ≥ 1 n \ge1 n ≥ 1 . When n ≥ 2 n \ge2 n ≥ 2 , the finite part is of type B n B_n B n , and
( d 0 , d 1 , … , d n ) = ( 4 , 2 , … , 2 , 1 ) , δ = α 0 + 2 α 1 + ⋯ + 2 α n , (63) (d_0,d_1,\ldots,d_n)=(4,2,\ldots,2,1),\qquad\delta=\alpha_0+2\alpha_1+\cdots+2\alpha_n, \tag*{(63)} ( d 0 , d 1 , … , d n ) = ( 4 , 2 , … , 2 , 1 ) , δ = α 0 + 2 α 1 + ⋯ + 2 α n , (63)
and the nonzero off-diagonal Cartan entries are
( a 01 , a 10 ) = ( − 1 , − 2 ) , ( a i , i + 1 , a i + 1 , i ) = ( − 1 , − 1 ) ( 1 ≤ i ≤ n − 2 ) , (a_{01},a_{10})=(-1,-2),\qquad(a_{i,i+1},a_{i+1,i})=(-1,-1)\qquad(1\le i\le n-2), ( a 01 , a 10 ) = ( − 1 , − 2 ) , ( a i , i + 1 , a i + 1 , i ) = ( − 1 , − 1 ) ( 1 ≤ i ≤ n − 2 ) ,
( a n − 1 , n , a n , n − 1 ) = ( − 1 , − 2 ) . (64) (a_{n-1,n},a_{n,n-1})=(-1,-2). \tag*{(64)} ( a n − 1 , n , a n , n − 1 ) = ( − 1 , − 2 ) . (64)
For n = 1 n=1 n = 1 , we instead prescribe
( a i j ) = ( 2 − 1 − 4 2 ) , ( d 0 , d 1 ) = ( 4 , 1 ) , δ = α 0 + 2 α 1 . (65) (a_{ij})=\begin{pmatrix}2&-1\\-4&2\end{pmatrix},\qquad(d_0,d_1)=(4,1),\qquad\delta=\alpha_0+2\alpha_1. \tag*{(65)} ( a ij ) = ( 2 − 4 − 1 2 ) , ( d 0 , d 1 ) = ( 4 , 1 ) , δ = α 0 + 2 α 1 . (65)
These conventions are summarized by Figure 2 . As above, arrows point toward the shorter simple root. The rank-one quadruple bond is displayed separately because (65) is not obtained by substituting n = 1 n=1 n = 1 into (64) .
Retain and specialize the regular root datum fixed in § 2 to this Cartan matrix:
( X , Y , I , ⟨ , ⟩ , { α i } i ∈ I , { h i } i ∈ I ) , I = { 0 , 1 , … , n } , (66) (X,Y,I,\langle\ ,\ \rangle,\{\alpha_i\}_{i\in I},\{h_i\}_{i\in I}),\qquad I=\{0,1,\ldots,n\}, \tag*{(66)} ( X , Y , I , ⟨ , ⟩ , { α i } i ∈ I , { h i } i ∈ I ) , I = { 0 , 1 , … , n } , (66)
where X X X and Y Y Y are free abelian groups in perfect duality, ⟨ h i , α j ⟩ = a i j \langle h_i,\alpha_j\rangle=a_{ij} ⟨ h i , α j ⟩ = a ij , and ⨁ i ∈ I Z α i ⊂ X \bigoplus_{i\in I}\mathbb{Z}\alpha_i\subset X ⨁ i ∈ I Z α i ⊂ X .
We abbreviate
U A L u s , ( n ) : = U A L u s ( A 2 n ( 2 ) ) . U_{\mathcal A}^{\mathrm{Lus},(n)}:=U_{\mathcal A}^{\mathrm{Lus}}(A_{2n}^{(2)}). U A Lus , ( n ) := U A Lus ( A 2 n ( 2 ) ) .
Specializing (29) to ν i = 1 \nu_i=1 ν i = 1 , write ω ~ i : = t λ i = t ω i ∨ \widetilde{\omega}_i:=t_{\lambda_i}=t_{\omega_i^\vee} ω i := t λ i = t ω i ∨ and T ω ~ i : = T λ i T_{\widetilde{\omega}_i}:=T_{\lambda_i} T ω i := T λ i . Then (40) with a = 1 a=1 a = 1 gives
x i , k + = o ( i ) k ( T ω ~ i ) − k ( E i ) , x i , k − = o ( i ) k ( T ω ~ i ) k ( F i ) ( k ∈ Z ) . (67) x_{i,k}^{+}=o(i)^k(T_{\widetilde{\omega}_i})^{-k}(E_i),\qquad x_{i,k}^{-}=o(i)^k(T_{\widetilde{\omega}_i})^k(F_i)\qquad(k\in\mathbb{Z}). \tag*{(67)} x i , k + = o ( i ) k ( T ω i ) − k ( E i ) , x i , k − = o ( i ) k ( T ω i ) k ( F i ) ( k ∈ Z ) . (67)
Equation (34) gives
E k δ + α i = ( T ω ~ i ) − k ( E i ) = o ( i ) k x i , k + ( k ≥ 0 ) , E_{k\delta+\alpha_i}=(T_{\widetilde{\omega}_i})^{-k}(E_i)=o(i)^k x_{i,k}^{+}\qquad(k\ge0), E k δ + α i = ( T ω i ) − k ( E i ) = o ( i ) k x i , k + ( k ≥ 0 ) ,
E k δ − α i = ( T ω ~ i ) k T i − 1 ( E i ) = − o ( i ) k K k δ − α i x i , k − ( k > 0 ) . E_{k\delta-\alpha_i}=(T_{\widetilde{\omega}_i})^kT_i^{-1}(E_i)=-o(i)^kK_{k\delta-\alpha_i}x_{i,k}^{-}\qquad(k>0). E k δ − α i = ( T ω i ) k T i − 1 ( E i ) = − o ( i ) k K k δ − α i x i , k − ( k > 0 ) .
In particular,
deg x i , k + = k δ + α i , deg x i , k − = k δ − α i . (68) \deg x_{i,k}^{+}=k\delta+\alpha_i,\qquad\deg x_{i,k}^{-}=k\delta-\alpha_i. \tag*{(68)} deg x i , k + = k δ + α i , deg x i , k − = k δ − α i . (68)
Write U A D r , ( n ) : = U A D r ( A 2 n ( 2 ) ) . U_{\mathcal A}^{\mathrm{Dr},(n)}:=U_{\mathcal A}^{\mathrm{Dr}}(A_{2n}^{(2)}). U A Dr , ( n ) := U A Dr ( A 2 n ( 2 ) ) . Since I i = Z I_i=\mathbb{Z} I i = Z , (39) becomes
U A D r , ( n ) = A ⟨ K δ ± 1 , K i ± 1 , ( x i , k ± ) ( a ) ∣ i ∈ I 0 , k ∈ Z , a ≥ 0 ⟩ . (69) U_{\mathcal A}^{\mathrm{Dr},(n)}=\mathcal{A}\left\langle K_\delta^{\pm1},K_i^{\pm1},(x_{i,k}^{\pm})^{(a)}\mathrel{\big|}i\in I_0,\ k\in\mathbb{Z},\ a\ge0\right\rangle. \tag*{(69)} U A Dr , ( n ) = A ⟨ K δ ± 1 , K i ± 1 , ( x i , k ± ) ( a ) i ∈ I 0 , k ∈ Z , a ≥ 0 ⟩ . (69)
Finally, (28) and (40) give
Ω ( ( x i , k ± ) ( a ) ) = ( x i , − k ∓ ) ( a ) , Ω ( U A D r , ( n ) ) = U A D r , ( n ) . (70) \Omega\left((x_{i,k}^{\pm})^{(a)}\right)=(x_{i,-k}^{\mp})^{(a)},\qquad\Omega\left(U_{\mathcal A}^{\mathrm{Dr},(n)}\right)=U_{\mathcal A}^{\mathrm{Dr},(n)}. \tag*{(70)} Ω ( ( x i , k ± ) ( a ) ) = ( x i , − k ∓ ) ( a ) , Ω ( U A Dr , ( n ) ) = U A Dr , ( n ) . (70)
Lemma 5.1. Let X \mathcal{X} X be an A \mathcal{A} A -subalgebra satisfying
U A D r , ( n ) ⊆ X ⊆ U A L u s , ( n ) , U_{\mathcal{A}}^{\mathrm{Dr},(n)} \subseteq\mathcal{X} \subseteq U_{\mathcal{A}}^{\mathrm{Lus},(n)}, U A Dr , ( n ) ⊆ X ⊆ U A Lus , ( n ) , and let B = S − 1 A ⊂ K B=S^{-1}\mathcal{A}\subset\mathcal{K} B = S − 1 A ⊂ K . If, for every a ≥ 0 a\geq0 a ≥ 0 ,
E 0 ( a ) , F 0 ( a ) ∈ B X , (71) E_0^{(a)},F_0^{(a)}\in B\mathcal{X}, \tag*{(71)} E 0 ( a ) , F 0 ( a ) ∈ B X , (71) then
B X = B U A L u s , ( n ) . B\mathcal{X}=BU_{\mathcal{A}}^{\mathrm{Lus},(n)}. B X = B U A Lus , ( n ) . Proof. Setting k = 0 k=0 k = 0 in (67) gives
E i ( a ) = ( x i , 0 + ) ( a ) , F i ( a ) = ( x i , 0 − ) ( a ) ∈ U A D r , ( n ) ( i ∈ I 0 , a ≥ 0 ) . E_i^{(a)}=(x_{i,0}^{+})^{(a)},\qquad F_i^{(a)}=(x_{i,0}^{-})^{(a)}\in U_{\mathcal{A}}^{\mathrm{Dr},(n)}\qquad(i\in I_0,\ a\geq0). E i ( a ) = ( x i , 0 + ) ( a ) , F i ( a ) = ( x i , 0 − ) ( a ) ∈ U A Dr , ( n ) ( i ∈ I 0 , a ≥ 0 ) . The elements K δ ± 1 , K i ± 1 K_\delta^{\pm1},K_i^{\pm1} K δ ± 1 , K i ± 1 ( i ∈ I 0 ) (i\in I_0) ( i ∈ I 0 ) also belong to U A D r , ( n ) ⊆ X U_{\mathcal{A}}^{\mathrm{Dr},(n)}\subseteq\mathcal{X} U A Dr , ( n ) ⊆ X , while (71) supplies the affine simple divided powers. Moreover,
K 0 ± 1 = ( K δ ∏ i ∈ I 0 K i − r i ) ± 1 ∈ B X . K_0^{\pm1}=\left(K_\delta\prod_{i\in I_0}K_i^{-r_i}\right)^{\pm1}\in B\mathcal{X}. K 0 ± 1 = ( K δ i ∈ I 0 ∏ K i − r i ) ± 1 ∈ B X . Thus B X B\mathcal{X} B X contains B B B and all defining generators of U A L u s , ( n ) U_{\mathcal{A}}^{\mathrm{Lus},(n)} U A Lus , ( n ) , so
B U A L u s , ( n ) ⊆ B X . BU_{\mathcal{A}}^{\mathrm{Lus},(n)}\subseteq B\mathcal{X}. B U A Lus , ( n ) ⊆ B X . The reverse inclusion follows from X ⊆ U A L u s , ( n ) \mathcal{X}\subseteq U_{\mathcal{A}}^{\mathrm{Lus},(n)} X ⊆ U A Lus , ( n ) .
Quantum scaled roots when n = 1 n=1 n = 1 and C*-recovery
The rank-one q q q -Weyl cell
By (65) ,
q 0 = q 4 , q 1 = q , δ − α 1 = α 0 + α 1 . q_0=q^4,\qquad q_1=q,\qquad\delta-\alpha_1=\alpha_0+\alpha_1. q 0 = q 4 , q 1 = q , δ − α 1 = α 0 + α 1 .
The root δ − α 1 \delta-\alpha_1 δ − α 1 has the same root parameter q q q as α 1 \alpha_1 α 1 . Thus the divided powers of E δ − α 1 E_{\delta-\alpha_1} E δ − α 1 and F 1 F_1 F 1 are normalized by [ s ] q ! [s]_q! [ s ] q ! , whereas those of E 0 E_0 E 0 and F 0 F_0 F 0 are normalized by [ a ] q 4 ! [a]_{q^4}! [ a ] q 4 ! .
Let
W : = K 1 E 0 . W:=K_1E_0. W := K 1 E 0 .
Using (31) , (33) , and (25) , we get that E δ − α 1 = T 0 ( E 1 ) = − E 0 E 1 + q − 4 E 1 E 0 E_{\delta-\alpha_1}=T_0(E_1)=-E_0E_1+q^{-4}E_1E_0 E δ − α 1 = T 0 ( E 1 ) = − E 0 E 1 + q − 4 E 1 E 0 . By the q q q -commutator techniques of [16 ] , we have that
E δ − α 1 F 1 − F 1 E δ − α 1 = − [ 4 ] q W . E_{\delta-\alpha_1}F_1-F_1E_{\delta-\alpha_1}=-[4]_qW. E δ − α 1 F 1 − F 1 E δ − α 1 = − [ 4 ] q W .
Since E 0 F 1 = F 1 E 0 E_0F_1=F_1E_0 E 0 F 1 = F 1 E 0 and
K 1 F 1 K 1 − 1 = q − 2 F 1 , K_1F_1K_1^{-1}=q^{-2}F_1, K 1 F 1 K 1 − 1 = q − 2 F 1 ,
we have
W F 1 = K 1 E 0 F 1 = K 1 F 1 E 0 = q − 2 F 1 K 1 E 0 = q − 2 F 1 W . WF_1=K_1E_0F_1=K_1F_1E_0=q^{-2}F_1K_1E_0=q^{-2}F_1W. W F 1 = K 1 E 0 F 1 = K 1 F 1 E 0 = q − 2 F 1 K 1 E 0 = q − 2 F 1 W .
Note that our real root vectors differ from Akasaka’s only by signs. It follows from [1 Lemma 3.9(2)] that
E 0 E δ − α 1 = q 4 E δ − α 1 E 0 . (72) E_0E_{\delta-\alpha_1}=q^4E_{\delta-\alpha_1}E_0. \tag*{(72)} E 0 E δ − α 1 = q 4 E δ − α 1 E 0 . (72)
while the weight of E δ − α 1 E_{\delta-\alpha_1} E δ − α 1 gives
K 1 E δ − α 1 K 1 − 1 = q − 2 E δ − α 1 , equivalently E δ − α 1 K 1 = q 2 K 1 E δ − α 1 . K_1E_{\delta-\alpha_1}K_1^{-1}=q^{-2}E_{\delta-\alpha_1}, \qquad\text{equivalently} \qquad E_{\delta-\alpha_1}K_1=q^2K_1E_{\delta-\alpha_1}. K 1 E δ − α 1 K 1 − 1 = q − 2 E δ − α 1 , equivalently E δ − α 1 K 1 = q 2 K 1 E δ − α 1 .
Consequently,
E δ − α 1 W = E δ − α 1 K 1 E 0 = q 2 K 1 E δ − α 1 E 0 = q − 2 K 1 E 0 E δ − α 1 = q − 2 W E δ − α 1 . E_{\delta-\alpha_1}W=E_{\delta-\alpha_1}K_1E_0=q^2K_1E_{\delta-\alpha_1}E_0=q^{-2}K_1E_0E_{\delta-\alpha_1}=q^{-2}WE_{\delta-\alpha_1}. E δ − α 1 W = E δ − α 1 K 1 E 0 = q 2 K 1 E δ − α 1 E 0 = q − 2 K 1 E 0 E δ − α 1 = q − 2 W E δ − α 1 .
We have therefore obtained
E δ − α 1 F 1 − F 1 E δ − α 1 = − [ 4 ] q W , W F 1 = q − 2 F 1 W , E δ − α 1 W = q − 2 W E δ − α 1 . (73) \begin{aligned}
E_{\delta-\alpha_1}F_1-F_1E_{\delta-\alpha_1}&=-[4]_qW,\\
WF_1&=q^{-2}F_1W,\\
E_{\delta-\alpha_1}W&=q^{-2}WE_{\delta-\alpha_1}. \tag*{(73)}
\end{aligned} E δ − α 1 F 1 − F 1 E δ − α 1 W F 1 E δ − α 1 W = − [ 4 ] q W , = q − 2 F 1 W , = q − 2 W E δ − α 1 . (73)
Formula (67) gives
x 1 , 1 − = − K δ − 1 K 1 E δ − α 1 , K 1 E δ − α 1 = q − 2 E δ − α 1 K 1 . (74) x_{1,1}^{-}=-K_{\delta}^{-1}K_1E_{\delta-\alpha_1}, \qquad K_1E_{\delta-\alpha_1}=q^{-2}E_{\delta-\alpha_1}K_1. \tag*{(74)} x 1 , 1 − = − K δ − 1 K 1 E δ − α 1 , K 1 E δ − α 1 = q − 2 E δ − α 1 K 1 . (74)
The first identity follows from (67) and (34) ; the second is the Cartan relation above. We now express the divided powers of E δ − α 1 E_{\delta-\alpha_1} E δ − α 1 in terms of the current generator x 1 , 1 − x_{1,1}^{-} x 1 , 1 − . From
E δ − α 1 K 1 = q 2 K 1 E δ − α 1 , E_{\delta-\alpha_1}K_1=q^2K_1E_{\delta-\alpha_1}, E δ − α 1 K 1 = q 2 K 1 E δ − α 1 ,
induction on s s s gives
( K 1 E δ − α 1 ) s = q 2 ( s ) K 1 s E δ − α 1 s . (K_1E_{\delta-\alpha_1})^s=q^{2^{(s)}}K_1^sE_{\delta-\alpha_1}^s. ( K 1 E δ − α 1 ) s = q 2 ( s ) K 1 s E δ − α 1 s .
Indeed, assuming the formula for s s s , one has
( K 1 E δ − α 1 ) s + 1 = ( K 1 E δ − α 1 ) s K 1 E δ − α 1 = q 2 ( s ) K 1 s E δ − α 1 s K 1 E δ − α 1 = q 2 ( s ) + 2 s K 1 s + 1 E δ − α 1 s + 1 = q 2 ( s + 1 ) K 1 s + 1 E δ − α 1 s + 1 . \begin{aligned}
(K_1E_{\delta-\alpha_1})^{s+1}
&=(K_1E_{\delta-\alpha_1})^sK_1E_{\delta-\alpha_1}\\
&=q^{2^{(s)}}K_1^sE_{\delta-\alpha_1}^sK_1E_{\delta-\alpha_1}\\
&=q^{2^{(s)}+2s}K_1^{s+1}E_{\delta-\alpha_1}^{s+1}\\
&=q^{2^{(s+1)}}K_1^{s+1}E_{\delta-\alpha_1}^{s+1}.
\end{aligned} ( K 1 E δ − α 1 ) s + 1 = ( K 1 E δ − α 1 ) s K 1 E δ − α 1 = q 2 ( s ) K 1 s E δ − α 1 s K 1 E δ − α 1 = q 2 ( s ) + 2 s K 1 s + 1 E δ − α 1 s + 1 = q 2 ( s + 1 ) K 1 s + 1 E δ − α 1 s + 1 .
Since K δ K_\delta K δ is central, (74) therefore gives
( x 1 , 1 − ) s = ( − 1 ) s K δ − s ( K 1 E δ − α 1 ) s = ( − 1 ) s q 2 ( s ) K δ − s K 1 s E δ − α 1 s . (x_{1,1}^{-})^s=(-1)^sK_\delta^{-s}(K_1E_{\delta-\alpha_1})^s=(-1)^sq^{2^{(s)}}K_\delta^{-s}K_1^sE_{\delta-\alpha_1}^s. ( x 1 , 1 − ) s = ( − 1 ) s K δ − s ( K 1 E δ − α 1 ) s = ( − 1 ) s q 2 ( s ) K δ − s K 1 s E δ − α 1 s .
Left multiplication by K δ s K 1 − s K_\delta^sK_1^{-s} K δ s K 1 − s and division by [ s ] q ! [s]_q! [ s ] q ! yield
E δ − α 1 ( s ) = ( − 1 ) s q − 2 ( s ) K δ s K 1 − s ( x 1 , 1 − ) ( s ) . E_{\delta-\alpha_1}^{(s)}=(-1)^sq^{-2^{(s)}}K_\delta^sK_1^{-s}(x_{1,1}^{-})^{(s)}. E δ − α 1 ( s ) = ( − 1 ) s q − 2 ( s ) K δ s K 1 − s ( x 1 , 1 − ) ( s ) .
All factors on the right belong to U A D r , ( 1 ) U_A^{\mathrm{Dr},(1)} U A Dr , ( 1 ) . Furthermore, (67) gives x 1 , 0 − = F 1 x_{1,0}^{-}=F_1 x 1 , 0 − = F 1 . Hence, for every s ≥ 0 s\geq0 s ≥ 0 ,
E δ − α 1 ( s ) = ( − 1 ) s q − 2 ( s ) K δ s K 1 − s ( x 1 , 1 − ) ( s ) ∈ U A D r , ( 1 ) , F 1 ( s ) = ( x 1 , 0 − ) ( s ) ∈ U A D r , ( 1 ) . (75) E_{\delta-\alpha_1}^{(s)}=(-1)^sq^{-2^{(s)}}K_\delta^sK_1^{-s}(x_{1,1}^{-})^{(s)}\in U_A^{\mathrm{Dr},(1)}, \qquad F_1^{(s)}=(x_{1,0}^{-})^{(s)}\in U_A^{\mathrm{Dr},(1)}. \tag*{(75)} E δ − α 1 ( s ) = ( − 1 ) s q − 2 ( s ) K δ s K 1 − s ( x 1 , 1 − ) ( s ) ∈ U A Dr , ( 1 ) , F 1 ( s ) = ( x 1 , 0 − ) ( s ) ∈ U A Dr , ( 1 ) . (75)
A direct multiplier for the affine simple root
Proposition 6.1 (Quantum scaled roots when n = 1 n=1 n = 1 ). For a ≥ 0 a\geq0 a ≥ 0 , define
d a : = ∏ h = 1 a ( q 2 h + 1 ) ( q 4 h + 1 ) , d 0 = 1. (76) d_a:=\prod_{h=1}^{a}(q^{2h}+1)(q^{4h}+1), \qquad d_0=1. \tag*{(76)} d a := h = 1 ∏ a ( q 2 h + 1 ) ( q 4 h + 1 ) , d 0 = 1. (76) Then d a ∈ A d_a\in\mathcal{A} d a ∈ A and
d a E 0 ( a ) , d a F 0 ( a ) ∈ U A D r , ( 1 ) . (77) d_aE_0^{(a)}, \qquad d_aF_0^{(a)}\in U_A^{\mathrm{Dr},(1)}. \tag*{(77)} d a E 0 ( a ) , d a F 0 ( a ) ∈ U A Dr , ( 1 ) . (77)
Proof. Write Z W : = E δ − α 1 F 1 − F 1 E δ − α 1 Z_W := E_{\delta-\alpha_1}F_1-F_1E_{\delta-\alpha_1} Z W := E δ − α 1 F 1 − F 1 E δ − α 1 . Then it follows from (6.1) that
E δ − α 1 Z W = q − 2 Z W E δ − α 1 , Z W F 1 = q − 2 F 1 Z W . E_{\delta-\alpha_1}Z_W=q^{-2}Z_WE_{\delta-\alpha_1},\qquad Z_WF_1=q^{-2}F_1Z_W. E δ − α 1 Z W = q − 2 Z W E δ − α 1 , Z W F 1 = q − 2 F 1 Z W .
Apply Lemma 1.5 with
D = U A D r , ( 1 ) , Q = q , A = E δ − α 1 , B = F 1 , Z = Z W . D=U_A^{\mathrm{Dr},(1)},\qquad Q=q,\qquad A=E_{\delta-\alpha_1},\qquad B=F_1,\qquad Z=Z_W. D = U A Dr , ( 1 ) , Q = q , A = E δ − α 1 , B = F 1 , Z = Z W .
The required divided-power hypotheses follow from (75) . Hence
Z W a [ a ] q ! = ( − 1 ) a [ 4 ] q a W a [ a ] q ! ∈ U A D r , ( 1 ) . (78) \frac{Z_W^a}{[a]_q!}=(-1)^a\frac{[4]_q^aW^a}{[a]_q!}\in U_A^{\mathrm{Dr},(1)}. \tag*{(78)} [ a ] q ! Z W a = ( − 1 ) a [ a ] q ! [ 4 ] q a W a ∈ U A Dr , ( 1 ) . (78)
We next rewrite W a W^a W a in terms of the affine simple divided power. The Cartan relation
K 1 E 0 K 1 − 1 = q − 4 E 0 K_1E_0K_1^{-1}=q^{-4}E_0 K 1 E 0 K 1 − 1 = q − 4 E 0
is equivalent to
E 0 K 1 = q 4 K 1 E 0 . E_0K_1=q^4K_1E_0. E 0 K 1 = q 4 K 1 E 0 .
It follows by induction that
W a = ( K 1 E 0 ) a = q 4 ( a 2 ) K 1 a E 0 a . W^a=(K_1E_0)^a=q^{4\binom{a}{2}}K_1^aE_0^a. W a = ( K 1 E 0 ) a = q 4 ( 2 a ) K 1 a E 0 a .
Indeed, if the identity holds for a a a , then
( K 1 E 0 ) a + 1 = q 4 ( a 2 ) K 1 a E 0 a K 1 E 0 = q 4 ( a 2 ) + 4 a K 1 a + 1 E 0 a + 1 = q 4 ( a + 1 2 ) K 1 a + 1 E 0 a + 1 . \begin{aligned}
(K_1E_0)^{a+1}
&=q^{4\binom{a}{2}}K_1^aE_0^aK_1E_0\\
&=q^{4\binom{a}{2}+4a}K_1^{a+1}E_0^{a+1}\\
&=q^{4\binom{a+1}{2}}K_1^{a+1}E_0^{a+1}.
\end{aligned} ( K 1 E 0 ) a + 1 = q 4 ( 2 a ) K 1 a E 0 a K 1 E 0 = q 4 ( 2 a ) + 4 a K 1 a + 1 E 0 a + 1 = q 4 ( 2 a + 1 ) K 1 a + 1 E 0 a + 1 .
Since q 0 = q 4 q_0=q^4 q 0 = q 4 , one has
E 0 a = [ a ] q 4 ! E 0 ( a ) . E_0^a=[a]_{q_4}!E_0^{(a)}. E 0 a = [ a ] q 4 ! E 0 ( a ) .
Therefore
W a = ( K 1 E 0 ) a = q 4 ( a 2 ) K 1 a E 0 a = q 4 ( a 2 ) [ a ] q 4 ! K 1 a E 0 ( a ) . (79) W^a=(K_1E_0)^a=q^{4\binom{a}{2}}K_1^aE_0^a=q^{4\binom{a}{2}}[a]_{q_4}!K_1^aE_0^{(a)}. \tag*{(79)} W a = ( K 1 E 0 ) a = q 4 ( 2 a ) K 1 a E 0 a = q 4 ( 2 a ) [ a ] q 4 ! K 1 a E 0 ( a ) . (79)
Substituting (79) into (78) gives
( − 1 ) a q 4 ( a 2 ) K 1 a [ 4 ] q a [ a ] q 4 ! [ a ] q ! E 0 ( a ) ∈ U A D r , ( 1 ) . (-1)^aq^{4\binom{a}{2}}K_1^a\frac{[4]_q^a[a]_{q_4}!}{[a]_q!}E_0^{(a)}\in U_A^{\mathrm{Dr},(1)}. ( − 1 ) a q 4 ( 2 a ) K 1 a [ a ] q ! [ 4 ] q a [ a ] q 4 ! E 0 ( a ) ∈ U A Dr , ( 1 ) .
Put
d ~ a : = [ 4 ] q a [ a ] q 4 ! [ a ] q ! . \widetilde d_a:=\frac{[4]_q^a[a]_{q_4}!}{[a]_q!}. d a := [ a ] q ! [ 4 ] q a [ a ] q 4 ! .
The scalar coefficient can be computed factor by factor:
d ~ a = ∏ h = 1 a [ 4 ] q [ h ] q 4 [ h ] q = ∏ h = 1 a [ 4 h ] q [ h ] q , \widetilde d_a=\prod_{h=1}^a\frac{[4]_q[h]_{q_4}}{[h]_q}=\prod_{h=1}^a\frac{[4h]_q}{[h]_q}, d a = h = 1 ∏ a [ h ] q [ 4 ] q [ h ] q 4 = h = 1 ∏ a [ h ] q [ 4 h ] q ,
where
[ 4 ] q [ h ] q 4 = q 4 − q − 4 q − q − 1 q 4 h − q − 4 h q 4 − q − 4 = q 4 h − q − 4 h q − q − 1 = [ 4 h ] q . [4]_q[h]_{q_4}=\frac{q^4-q^{-4}}{q-q^{-1}}\frac{q^{4h}-q^{-4h}}{q^4-q^{-4}}=\frac{q^{4h}-q^{-4h}}{q-q^{-1}}=[4h]_q. [ 4 ] q [ h ] q 4 = q − q − 1 q 4 − q − 4 q 4 − q − 4 q 4 h − q − 4 h = q − q − 1 q 4 h − q − 4 h = [ 4 h ] q .
For every h ≥ 1 h \ge1 h ≥ 1 ,
[ 4 h ] q [ h ] q = q 4 h − q − 4 h q h − q − h = q − 3 h q 8 h − 1 q 2 h − 1 = q − 3 h ( q 6 h + q 4 h + q 2 h + 1 ) = q − 3 h ( q 2 h + 1 ) ( q 4 h + 1 ) . (80) \begin{aligned}
\frac{[4h]_q}{[h]_q}
&=\frac{q^{4h}-q^{-4h}}{q^h-q^{-h}}\\
&=q^{-3h}\frac{q^{8h}-1}{q^{2h}-1}\\
&=q^{-3h}(q^{6h}+q^{4h}+q^{2h}+1)\\
&=q^{-3h}(q^{2h}+1)(q^{4h}+1). \tag*{(80)}
\end{aligned} [ h ] q [ 4 h ] q = q h − q − h q 4 h − q − 4 h = q − 3 h q 2 h − 1 q 8 h − 1 = q − 3 h ( q 6 h + q 4 h + q 2 h + 1 ) = q − 3 h ( q 2 h + 1 ) ( q 4 h + 1 ) . (80)
Equations (76) and (80) give
d ~ a = q − 3 a ( a + 1 ) 2 d a . \widetilde{d}_a=q^{-\frac{3a(a+1)}{2}}d_a. d a = q − 2 3 a ( a + 1 ) d a .
Thus
( − 1 ) a q a ( a − 7 ) 2 d a K 1 a E 0 ( a ) ∈ U A D r , ( 1 ) . (-1)^a q^{\frac{a(a-7)}{2}}d_aK_1^aE_0^{(a)}\in U_{\mathcal{A}}^{\mathrm{Dr},(1)}. ( − 1 ) a q 2 a ( a − 7 ) d a K 1 a E 0 ( a ) ∈ U A Dr , ( 1 ) .
Multiplication by the unit
( − 1 ) a q a ( 7 − a ) 2 K 1 − a (-1)^a q^{\frac{a(7-a)}{2}}K_1^{-a} ( − 1 ) a q 2 a ( 7 − a ) K 1 − a
gives
d a E 0 ( a ) ∈ U A D r , ( 1 ) . d_aE_0^{(a)}\in U_{\mathcal{A}}^{\mathrm{Dr},(1)}. d a E 0 ( a ) ∈ U A Dr , ( 1 ) .
The definition (76) makes d a ∈ Z [ q ] ⊆ A d_a\in\mathbb{Z}[q]\subseteq\mathcal{A} d a ∈ Z [ q ] ⊆ A immediate.
Finally, (70) gives
Ω ( U A D r , ( 1 ) ) = U A D r , ( 1 ) . \Omega\left(U_{\mathcal{A}}^{\mathrm{Dr},(1)}\right)=U_{\mathcal{A}}^{\mathrm{Dr},(1)}. Ω ( U A Dr , ( 1 ) ) = U A Dr , ( 1 ) .
The polynomial d a d_a d a satisfies
Ω ( d a ) = ∏ h = 1 a ( q − 2 h + 1 ) ( q − 4 h + 1 ) = q − 3 a ( a + 1 ) d a . \Omega(d_a)=\prod_{h=1}^{a}(q^{-2h}+1)(q^{-4h}+1)=q^{-3a(a+1)}d_a. Ω ( d a ) = h = 1 ∏ a ( q − 2 h + 1 ) ( q − 4 h + 1 ) = q − 3 a ( a + 1 ) d a .
Moreover,
Ω ( E 0 ( a ) ) = F 0 a [ a ] q − 4 ! = F 0 a [ a ] q 4 ! = F 0 ( a ) . \Omega\left(E_0^{(a)}\right)=\frac{F_0^a}{[a]_{q^{-4}}!}=\frac{F_0^a}{[a]_{q^4}!}=F_0^{(a)}. Ω ( E 0 ( a ) ) = [ a ] q − 4 ! F 0 a = [ a ] q 4 ! F 0 a = F 0 ( a ) .
Applying Ω \Omega Ω to d a E 0 ( a ) ∈ U A D r , ( 1 ) d_aE_0^{(a)}\in U_{\mathcal{A}}^{\mathrm{Dr},(1)} d a E 0 ( a ) ∈ U A Dr , ( 1 ) therefore gives
q − 3 a ( a + 1 ) d a F 0 ( a ) = Ω ( d a E 0 ( a ) ) ∈ U A D r , ( 1 ) . q^{-3a(a+1)}d_aF_0^{(a)}=\Omega\left(d_aE_0^{(a)}\right)\in U_{\mathcal{A}}^{\mathrm{Dr},(1)}. q − 3 a ( a + 1 ) d a F 0 ( a ) = Ω ( d a E 0 ( a ) ) ∈ U A Dr , ( 1 ) .
Multiplication by the unit q 3 a ( a + 1 ) q^{3a(a+1)} q 3 a ( a + 1 ) gives d a F 0 ( a ) ∈ U A D r , ( 1 ) d_aF_0^{(a)}\in U_{\mathcal{A}}^{\mathrm{Dr},(1)} d a F 0 ( a ) ∈ U A Dr , ( 1 ) . This proves (77) . □ \square □
Theorem 6.2 (Rank-one C ∗ C^* C ∗ -sufficiency). Recall C ∗ = A [ Φ 4 r ( q ) − 1 ∣ r ≥ 1 ] C^*=\mathcal{A}[\Phi_{4r}(q)^{-1}\mid r\ge1] C ∗ = A [ Φ 4 r ( q ) − 1 ∣ r ≥ 1 ] . One has
C ∗ U A D r , ( 1 ) = C ∗ U A L u s , ( 1 ) . (81) C^*U_{\mathcal{A}}^{\mathrm{Dr},(1)}=C^*U_{\mathcal{A}}^{\mathrm{Lus},(1)}. \tag*{(81)} C ∗ U A Dr , ( 1 ) = C ∗ U A Lus , ( 1 ) . (81) Proof. The forward inclusion
C ∗ U A D r , ( 1 ) ⊆ C ∗ U A L u s , ( 1 ) C^*U_{\mathcal{A}}^{\mathrm{Dr},(1)}\subseteq C^*U_{\mathcal{A}}^{\mathrm{Lus},(1)} C ∗ U A Dr , ( 1 ) ⊆ C ∗ U A Lus , ( 1 ) follows from U A D r , ( 1 ) ⊆ U A L u s , ( 1 ) U_{\mathcal{A}}^{\mathrm{Dr},(1)}\subseteq U_{\mathcal{A}}^{\mathrm{Lus},(1)} U A Dr , ( 1 ) ⊆ U A Lus , ( 1 ) . It remains to recover the affine simple divided powers after extending scalars to C ∗ C^* C ∗ . By (80) , d a = ∏ h = 1 a ( q 2 h + 1 ) ( q 4 h + 1 ) d_a=\prod_{h=1}^{a}(q^{2h}+1)(q^{4h}+1) d a = ∏ h = 1 a ( q 2 h + 1 ) ( q 4 h + 1 ) . We determine the cyclotomic factors of the two polynomials occurring here. The standard factorization
q N − 1 = ∏ d ∣ N Φ d ( q ) q^N-1=\prod_{d\mid N}\Phi_d(q) q N − 1 = d ∣ N ∏ Φ d ( q )
gives
q 2 h + 1 = q 4 h − 1 q 2 h − 1 = ∏ d ∣ 4 h d ∤ 2 h Φ d ( q ) . q^{2h}+1=\frac{q^{4h}-1}{q^{2h}-1}=\prod_{\substack{d\mid4h\\d\nmid2h}}\Phi_d(q). q 2 h + 1 = q 2 h − 1 q 4 h − 1 = d ∣ 4 h d ∤ 2 h ∏ Φ d ( q ) .
Write h = 2 ν m h=2^\nu m h = 2 ν m , where m m m is odd. An integer d d d divides 4 h 4h 4 h but not 2 h 2h 2 h precisely when
d = 2 ν + 2 t for some t ∣ m . d=2^{\nu+2}t\qquad\text{for some }t\mid m. d = 2 ν + 2 t for some t ∣ m .
Writing s = 2 ν t s=2^\nu t s = 2 ν t , this is equivalent to
d = 4 s , s ∣ h , h / s odd . d=4s,\qquad s\mid h,\qquad h/s\text{ odd}. d = 4 s , s ∣ h , h / s odd .
Therefore
q 2 h + 1 = ∏ s ∣ h h / s odd Φ 4 s ( q ) . q^{2h}+1=\prod_{\substack{s\mid h\\h/s\text{ odd}}}\Phi_{4s}(q). q 2 h + 1 = s ∣ h h / s odd ∏ Φ 4 s ( q ) .
Similarly,
q 4 h + 1 = q 8 h − 1 q 4 h − 1 = ∏ d ∣ 8 h d ∤ 4 h Φ d ( q ) . q^{4h}+1=\frac{q^{8h}-1}{q^{4h}-1}=\prod_{\substack{d\mid8h\\d\nmid4h}}\Phi_d(q). q 4 h + 1 = q 4 h − 1 q 8 h − 1 = d ∣ 8 h d ∤ 4 h ∏ Φ d ( q ) .
The same 2-adic argument shows that the surviving indices are d = 8 s d=8s d = 8 s , where s ∣ h s\mid h s ∣ h and h / s h/s h / s is odd. Hence
q 4 h + 1 = ∏ s ∣ h h / s odd Φ 8 s ( q ) = ∏ s ∣ h h / s odd Φ 4 ( 2 s ) ( q ) . q^{4h}+1=\prod_{\substack{s\mid h\\h/s\text{ odd}}}\Phi_{8s}(q)
=\prod_{\substack{s\mid h\\h/s\text{ odd}}}\Phi_{4(2s)}(q). q 4 h + 1 = s ∣ h h / s odd ∏ Φ 8 s ( q ) = s ∣ h h / s odd ∏ Φ 4 ( 2 s ) ( q ) .
It follows that
d a = ∏ h = 1 a ∏ s ∣ h h / s odd Φ 4 s ( q ) Φ 8 s ( q ) . d_a=\prod_{h=1}^{a}\prod_{\substack{s\mid h\\h/s\text{ odd}}}\Phi_{4s}(q)\Phi_{8s}(q). d a = h = 1 ∏ a s ∣ h h / s odd ∏ Φ 4 s ( q ) Φ 8 s ( q ) .
Every factor Φ 4 s ( q ) \Phi_{4s}(q) Φ 4 s ( q ) is invertible in C ∗ \mathbb{C}^* C ∗ , and
Φ 8 s ( q ) = Φ 4 ( 2 s ) ( q ) \Phi_{8s}(q)=\Phi_{4(2s)}(q) Φ 8 s ( q ) = Φ 4 ( 2 s ) ( q )
is invertible there as well. Consequently,
d a ∈ ( C ∗ ) × ( a ≥ 0 ) . d_a\in(\mathbb{C}^*)^\times\qquad(a\geq0). d a ∈ ( C ∗ ) × ( a ≥ 0 ) .
Proposition 6.1 gives
d a E 0 ( a ) , d a F 0 ( a ) ∈ U A D r , ( 1 ) ⊆ C ∗ U A D r , ( 1 ) . d_aE_0^{(a)},\qquad d_aF_0^{(a)}\in U_A^{\mathrm{Dr},(1)}\subseteq\mathbb{C}^*U_A^{\mathrm{Dr},(1)}. d a E 0 ( a ) , d a F 0 ( a ) ∈ U A Dr , ( 1 ) ⊆ C ∗ U A Dr , ( 1 ) .
Multiplication by d a − 1 ∈ C ∗ d_a^{-1}\in\mathbb{C}^* d a − 1 ∈ C ∗ therefore yields
E 0 ( a ) , F 0 ( a ) ∈ C ∗ U A D r , ( 1 ) ( a ≥ 0 ) . E_0^{(a)},\qquad F_0^{(a)}\in\mathbb{C}^*U_A^{\mathrm{Dr},(1)}\qquad(a\geq0). E 0 ( a ) , F 0 ( a ) ∈ C ∗ U A Dr , ( 1 ) ( a ≥ 0 ) .
Thus the hypothesis {(xref)}(5.8) of Lemma 5.1 holds with
X = U A D r , ( 1 ) , B = C ∗ . \mathcal{X}=U_A^{\mathrm{Dr},(1)},\qquad B=\mathbb{C}^*. X = U A Dr , ( 1 ) , B = C ∗ .
Applying that lemma gives
C ∗ U A D r , ( 1 ) = C ∗ U A L u s , ( 1 ) . \mathbb{C}^*U_A^{\mathrm{Dr},(1)}=\mathbb{C}^*U_A^{\mathrm{Lus},(1)}. C ∗ U A Dr , ( 1 ) = C ∗ U A Lus , ( 1 ) .
The sufficiency in Rank-one is now complete. Under the classical specialization q ↦ 1 q\mapsto1 q ↦ 1 , E 0 ↦ e 0 E_0\mapsto e_0 E 0 ↦ e 0 , we have d a ( 1 ) = 4 a d_a(1)=4^a d a ( 1 ) = 4 a . Thus d a E 0 ( a ) d_aE_0^{(a)} d a E 0 ( a ) specializes to ( 4 e 0 ) ( a ) (4e_0)^{(a)} ( 4 e 0 ) ( a ) , recovering the divided powers of classical scaled-roots in [10 Remark C.3] .
Quantum scaled roots when n ≥ 2 n \ge2 n ≥ 2 and higher-rank B ∗ B^{*} B ∗ -recovery
We next establish the corresponding higher-rank statement. The mechanism is genuinely different: the required multiplier is produced by a three-root cell, and its nonunit factors have only Φ 8 r \Phi_{8r} Φ 8 r -support.
Throughout this section, fix n ≥ 2 n \ge2 n ≥ 2 , retain the integral form U A D r , ( n ) U_{A}^{\mathrm{Dr},(n)} U A Dr , ( n ) of (69) , and put Q = q 2 Q=q^{2} Q = q 2 .
Braid group stability of U A D r , ( n ) U_{A}^{\mathrm{Dr},(n)} U A Dr , ( n )
Lemma 7.1. For 1 ≤ h ≤ n 1 \le h \le n 1 ≤ h ≤ n and ε ∈ { ± 1 } \varepsilon\in\{\pm1\} ε ∈ { ± 1 } ,
( T ω ~ h ) ε ( U A D r , ( n ) ) = U A D r , ( n ) , T i ( U A D r , ( n ) ) ⊆ U A D r , ( n ) ( 1 ≤ i < n ) . (82) \left(T_{\widetilde{\omega}_{h}}\right)^{\varepsilon}\left(U_{A}^{\mathrm{Dr},(n)}\right)=U_{A}^{\mathrm{Dr},(n)}, \qquad T_i\left(U_{A}^{\mathrm{Dr},(n)}\right)\subseteq U_{A}^{\mathrm{Dr},(n)} \qquad(1 \le i<n). \tag*{(82)} ( T ω h ) ε ( U A Dr , ( n ) ) = U A Dr , ( n ) , T i ( U A Dr , ( n ) ) ⊆ U A Dr , ( n ) ( 1 ≤ i < n ) . (82) Moreover, if 1 ≤ j < n 1 \le j<n 1 ≤ j < n , then
T n ( ( x j , r ± ) ( a ) ) ∈ U A D r , ( n ) ( r ∈ Z , a ≥ 0 ) . T_n\left(\left(x_{j,r}^{\pm}\right)^{(a)}\right)\in U_{A}^{\mathrm{Dr},(n)} \qquad(r\in\mathbb{Z},\ a\ge0). T n ( ( x j , r ± ) ( a ) ) ∈ U A Dr , ( n ) ( r ∈ Z , a ≥ 0 ) . Proof. The translation assertion follows from the same generator-by-generator argument as in Lemma 3.1, now using (67) . Indeed, ( T ω ~ h ) ± 1 \left(T_{\widetilde{\omega}_{h}}\right)^{\pm1} ( T ω h ) ± 1 shifts only the mode of a color-h h h current, up to a sign and central or toral Laurent monomials. Hence both operators preserve the generators of (69) , and therefore
( T ω ~ h ) ε ( U A D r , ( n ) ) = U A D r , ( n ) ( ε = ± 1 ) . \left(T_{\widetilde{\omega}_{h}}\right)^{\varepsilon}\left(U_{A}^{\mathrm{Dr},(n)}\right)=U_{A}^{\mathrm{Dr},(n)} \qquad(\varepsilon=\pm1). ( T ω h ) ε ( U A Dr , ( n ) ) = U A Dr , ( n ) ( ε = ± 1 ) . Now let 1 ≤ i < n 1 \le i<n 1 ≤ i < n . After the finite-node relabelling, the color-i i i currents lie in an ordinary A 1 ( 1 ) A_1^{(1)} A 1 ( 1 ) vertex. If p p p and q q q are its two nodes, then
a p q = a q p = − 2. a_{pq}=a_{qp}=-2. a pq = a q p = − 2. The argument is therefore similar to the proof of Proposition 3.3 : the embedded A 1 ( 1 ) A_1^{(1)} A 1 ( 1 ) vertex treats the color-i i i currents, while the different-color argument applies with t λ j t_{\lambda_j} t λ j replaced by ω ~ j = t ω j ∨ \widetilde{\omega}_j=t_{\omega_j}^{\vee} ω j = t ω j ∨ and uses (67) . The negative-current case follows from
Ω T i = T i Ω , Ω ( U A D r , ( n ) ) = U A D r , ( n ) , \Omega T_i=T_i\Omega,\qquad\Omega\left(U_{A}^{\mathrm{Dr},(n)}\right)=U_{A}^{\mathrm{Dr},(n)}, Ω T i = T i Ω , Ω ( U A Dr , ( n ) ) = U A Dr , ( n ) , and the toral generators are treated exactly as in that proof. Hence
T i ( U A D r , ( n ) ) ⊆ U A D r , ( n ) ( 1 ≤ i < n ) . T_i\left(U_{A}^{\mathrm{Dr},(n)}\right)\subseteq U_{A}^{\mathrm{Dr},(n)} \qquad(1\le i<n). T i ( U A Dr , ( n ) ) ⊆ U A Dr , ( n ) ( 1 ≤ i < n ) . It remains to prove the last assertion. Let 1 ≤ j < n 1\le j<n 1 ≤ j < n . Since
⟨ ω ~ j ∨ , α n ⟩ = 0 , s n ω ~ j = ω ~ j s n , \langle\widetilde{\omega}_j^{\vee},\alpha_n\rangle=0,\qquad s_n\widetilde{\omega}_j=\widetilde{\omega}_j s_n, ⟨ ω j ∨ , α n ⟩ = 0 , s n ω j = ω j s n , the different-color part of the proof of Proposition 3.3 applies with i = n i=n i = n . Together with (67) and (70) , it gives
T n ( ( x j , r ± ) ( a ) ) ∈ U A D r , ( n ) ( 1 ≤ j < n , r ∈ Z , a ≥ 0 ) . T_n\left(\left(x_{j,r}^{\pm}\right)^{(a)}\right)\in U_{A}^{\mathrm{Dr},(n)} \qquad(1\le j<n,\ r\in\mathbb{Z},\ a\ge0). T n ( ( x j , r ± ) ( a ) ) ∈ U A Dr , ( n ) ( 1 ≤ j < n , r ∈ Z , a ≥ 0 ) .
No assertion is made here for the color-n n n currents: the embedded A 1 ( 1 ) A_1^{(1)} A 1 ( 1 ) argument is unavailable at the node n n n , and only the different-color case is needed in the following.
Transport of the non-extra-long roots
Remark. We explain why the higher rank case needs only the Φ 8 r \Phi_{8r} Φ 8 r , not the Φ 4 r \Phi_{4r} Φ 4 r . In our numbering, there is an embedding
\iota_n: U_q(A_2^{(2)}) \lhook\joinrel\longrightarrow U_q(A_{2n}^{(2)})
that satisfies
ι n ( E 1 ) = E n , ι n ( E 0 ) = E δ − 2 α n , T n ∘ ι n = ι n ∘ T 1 . \iota_n(E_1)=E_n,\qquad\iota_n(E_0)=E_{\delta-2\alpha_n},\qquad T_n\circ\iota_n=\iota_n\circ T_1. ι n ( E 1 ) = E n , ι n ( E 0 ) = E δ − 2 α n , T n ∘ ι n = ι n ∘ T 1 .
(cf. [7 Defn. 4.1.1, Props. 4.3.11, 4.4.9–4.4.10] ). This embedding does not reduce theorem 7.9 to theorem 6.2. Indeed, the image of ι n \iota_n ι n is graded by Z α n + Z δ \mathbb{Z}\alpha_n+\mathbb{Z}\delta Z α n + Z δ , in which every element has even α n − 1 \alpha_{n-1} α n − 1 -coefficient, whereas that of β ′ = δ − ( α n − 1 + 2 α n ) \beta'=\delta-(\alpha_{n-1}+2\alpha_n) β ′ = δ − ( α n − 1 + 2 α n ) equals one; neither β = α n − 1 \beta=\alpha_{n-1} β = α n − 1 nor β ′ \beta' β ′ lies in that image. The three-root cell { β , β ′ , η n } \{\beta,\beta',\eta_n\} { β , β ′ , η n } of lemma 7.5 thus has at its disposal generators that are unavailable in rank one, and this is precisely what sharpens the multiplier: the cell produces the coefficient c a c_a c a of (88) , whose nonunit factors are exactly the Φ 8 d ( q ) \Phi_{8d}(q) Φ 8 d ( q ) , whereas the rank-one multiplier d a d_a d a of (76) carries in addition the factors Φ 4 s ( q ) \Phi_{4s}(q) Φ 4 s ( q ) with s s s odd.
Let
μ = δ − α n − 1 , β ′ = δ − ( α n − 1 + 2 α n ) , K μ : = K δ K n − 1 − 1 . \mu=\delta-\alpha_{n-1},\qquad\beta'=\delta-(\alpha_{n-1}+2\alpha_n),\qquad K_\mu:=K_\delta K_{n-1}^{-1}. μ = δ − α n − 1 , β ′ = δ − ( α n − 1 + 2 α n ) , K μ := K δ K n − 1 − 1 .
Proposition 7.2. One has ( μ ∣ μ ) = 4 (\mu\mid\mu)=4 ( μ ∣ μ ) = 4 ,
K μ x n − 1 , 1 − K μ − 1 = Q 2 x n − 1 , 1 − , K_\mu x_{n-1,1}^{-}K_\mu^{-1}=Q^2x_{n-1,1}^{-}, K μ x n − 1 , 1 − K μ − 1 = Q 2 x n − 1 , 1 − , and, for every a ≥ 0 a\geq0 a ≥ 0 ,
E μ = K μ x n − 1 , 1 − , E μ ( a ) = Q − 2 ( a ) K μ a ( x n − 1 , 1 − ) ( a ) ∈ U A D r , ( n ) . (83) E_\mu=K_\mu x_{n-1,1}^{-},\qquad E_\mu^{(a)}=Q^{-2(a)}K_\mu^a(x_{n-1,1}^{-})^{(a)}\in U_A^{\mathrm{Dr},(n)}. \tag*{(83)} E μ = K μ x n − 1 , 1 − , E μ ( a ) = Q − 2 ( a ) K μ a ( x n − 1 , 1 − ) ( a ) ∈ U A Dr , ( n ) . (83)
Proof. Since ⟨ ω n − 1 ∨ , α n − 1 ⟩ = 1 \langle\omega_{n-1}^{\vee},\alpha_{n-1}\rangle=1 ⟨ ω n − 1 ∨ , α n − 1 ⟩ = 1 and ω ~ n − 1 = t ω n − 1 ∨ \widetilde{\omega}_{n-1}=t_{\omega_{n-1}^{\vee}} ω n − 1 = t ω n − 1 ∨ ,
ω ~ n − 1 ( α n − 1 ) = α n − 1 − δ , ω ~ n − 1 ( δ ) = δ . (84) \widetilde{\omega}_{n-1}(\alpha_{n-1})=\alpha_{n-1}-\delta,\qquad\widetilde{\omega}_{n-1}(\delta)=\delta. \tag*{(84)} ω n − 1 ( α n − 1 ) = α n − 1 − δ , ω n − 1 ( δ ) = δ . (84)
Equations (34) and (25) give
E μ = E δ − α n − 1 = T ω ~ n − 1 T n − 1 − 1 ( E n − 1 ) , T n − 1 − 1 ( E n − 1 ) = − K n − 1 − 1 F n − 1 . E_\mu=E_{\delta-\alpha_{n-1}}=T_{\widetilde{\omega}_{n-1}}T_{n-1}^{-1}(E_{n-1}),\qquad T_{n-1}^{-1}(E_{n-1})=-K_{n-1}^{-1}F_{n-1}. E μ = E δ − α n − 1 = T ω n − 1 T n − 1 − 1 ( E n − 1 ) , T n − 1 − 1 ( E n − 1 ) = − K n − 1 − 1 F n − 1 .
By [21 Section 37.1.3] the braid group operators act on the torus by T w ( K η ) = K w ( η ) T_w(K_\eta)=K_{w(\eta)} T w ( K η ) = K w ( η ) , so (84) gives
T ω ~ n − 1 ( K n − 1 − 1 ) = K δ − α n − 1 = K δ K n − 1 − 1 = K μ . T_{\widetilde{\omega}_{n-1}}(K_{n-1}^{-1})=K_{\delta-\alpha_{n-1}}=K_\delta K_{n-1}^{-1}=K_\mu. T ω n − 1 ( K n − 1 − 1 ) = K δ − α n − 1 = K δ K n − 1 − 1 = K μ .
The transported sign function satisfies o ( n − 1 ) = − 1 o(n-1)=-1 o ( n − 1 ) = − 1 , so (67) at k = 1 k=1 k = 1 gives T ω ~ n − 1 ( F n − 1 ) = − x n − 1 , 1 − T_{\widetilde{\omega}_{n-1}}(F_{n-1})=-x_{n-1,1}^{-} T ω n − 1 ( F n − 1 ) = − x n − 1 , 1 − . Multiplying the two images yields
E μ = K μ x n − 1 , 1 − . E_\mu=K_\mu x_{n-1,1}^{-}. E μ = K μ x n − 1 , 1 − .
By (68) , deg x n − 1 , 1 − = δ − α n − 1 = μ \deg x_{n-1,1}^{-}=\delta-\alpha_{n-1}=\mu deg x n − 1 , 1 − = δ − α n − 1 = μ . Since ( δ ∣ γ ) = 0 (\delta\mid\gamma)=0 ( δ ∣ γ ) = 0 for every γ \gamma γ in the affine root lattice,
( μ ∣ μ ) = ( α n − 1 ∣ α n − 1 ) = 2 d n − 1 = 4 , (\mu\mid\mu)=(\alpha_{n-1}\mid\alpha_{n-1})=2d_{n-1}=4, ( μ ∣ μ ) = ( α n − 1 ∣ α n − 1 ) = 2 d n − 1 = 4 ,
and therefore
K μ x n − 1 , 1 − K μ − 1 = q ( μ ∣ μ ) x n − 1 , 1 − = Q 2 x n − 1 , 1 − , equivalently x n − 1 , 1 − K μ = Q − 2 K μ x n − 1 , 1 − . (85) K_\mu x_{n-1,1}^{-}K_\mu^{-1}=q^{(\mu\mid\mu)}x_{n-1,1}^{-}=Q^2x_{n-1,1}^{-},\qquad\text{equivalently}\qquad x_{n-1,1}^{-}K_\mu=Q^{-2}K_\mu x_{n-1,1}^{-}. \tag*{(85)} K μ x n − 1 , 1 − K μ − 1 = q ( μ ∣ μ ) x n − 1 , 1 − = Q 2 x n − 1 , 1 − , equivalently x n − 1 , 1 − K μ = Q − 2 K μ x n − 1 , 1 − . (85)
Writing x = x n − 1 , 1 − x=x_{n-1,1}^{-} x = x n − 1 , 1 − , induction on a a a gives
( K μ x ) a = Q − 2 ( a 2 ) K μ a x a . (K_\mu x)^a=Q^{-2\binom{a}{2}}K_\mu^a x^a. ( K μ x ) a = Q − 2 ( 2 a ) K μ a x a .
the case a = 0 a=0 a = 0 is trivial, and if the formula holds for a a a , then (85) gives x a K μ = Q − 2 a K μ x a x^aK_\mu=Q^{-2a}K_\mu x^a x a K μ = Q − 2 a K μ x a , whence
( K μ x ) a + 1 = Q − 2 ( a 2 ) K μ a x a K μ x = Q − 2 ( a 2 ) − 2 a K μ a + 1 x a + 1 = Q − 2 ( a + 1 2 ) K μ a + 1 x a + 1 . (K_\mu x)^{a+1}=Q^{-2\binom{a}{2}}K_\mu^a x^aK_\mu x=Q^{-2\binom{a}{2}-2a}K_\mu^{a+1}x^{a+1}=Q^{-2\binom{a+1}{2}}K_\mu^{a+1}x^{a+1}. ( K μ x ) a + 1 = Q − 2 ( 2 a ) K μ a x a K μ x = Q − 2 ( 2 a ) − 2 a K μ a + 1 x a + 1 = Q − 2 ( 2 a + 1 ) K μ a + 1 x a + 1 .
Finally, μ \mu μ has root parameter q ( μ ∣ μ ) / 2 = q 2 = Q = q n − 1 q^{(\mu\mid\mu)/2}=q^2=Q=q_{n-1} q ( μ ∣ μ ) /2 = q 2 = Q = q n − 1 , so E μ ( a ) E_\mu^{(a)} E μ ( a ) and x ( a ) x^{(a)} x ( a ) are normalized by the same factorial [ a ] Q ! [a]_Q! [ a ] Q ! . Dividing E μ a = ( K μ x ) a E_\mu^a=(K_\mu x)^a E μ a = ( K μ x ) a by [ a ] Q ! [a]_Q! [ a ] Q ! therefore gives the displayed formula for E μ ( a ) E_\mu^{(a)} E μ ( a ) . Its right-hand side lies in U A Dr , ( n ) U_A^{\operatorname{Dr},(n)} U A Dr , ( n ) , because K μ a = ( K δ K n − 1 − 1 ) a K_\mu^a=(K_\delta K_{n-1}^{-1})^a K μ a = ( K δ K n − 1 − 1 ) a is a monomial in the group-like toral generators, x ( a ) x^{(a)} x ( a ) is a defining generator of (69) , and Q − 2 ( a 2 ) ∈ A × Q^{-2\binom{a}{2}}\in A^\times Q − 2 ( 2 a ) ∈ A × .
Proposition 7.3 (Transport to the root β ′ \beta' β ′ ). Put
w ^ β : = s n w ~ n − 1 s n − 1 , T β : = T n T w ~ n − 1 ( T n − 1 ) − 1 , B ♯ : = T β ( E n − 1 ) . \widehat{w}_{\beta}:=s_n\widetilde{w}_{n-1}s_{n-1},\qquad\mathcal{T}_{\beta}:=T_n\mathcal{T}_{\widetilde{w}_{n-1}}(T_{n-1})^{-1},\qquad B^\sharp:=\mathcal{T}_{\beta}(E_{n-1}). w β := s n w n − 1 s n − 1 , T β := T n T w n − 1 ( T n − 1 ) − 1 , B ♯ := T β ( E n − 1 ) . Then w ^ β ( α n − 1 ) = β ′ \widehat{w}_{\beta}(\alpha_{n-1})=\beta' w β ( α n − 1 ) = β ′ , the operator T β \mathcal{T}_{\beta} T β is the standard lift T w ^ β T_{\widehat{w}_{\beta}} T w β , and
B ♯ = T n ( E μ ) , ( B ♯ ) ( a ) = T n ( E μ ( a ) ) ∈ U A Dr , ( n ) ( a ≥ 0 ) . (86) B^\sharp=T_n(E_\mu),\qquad(B^\sharp)^{(a)}=T_n(E_\mu^{(a)})\in U_A^{\operatorname{Dr},(n)}\qquad(a\geq0). \tag*{(86)} B ♯ = T n ( E μ ) , ( B ♯ ) ( a ) = T n ( E μ ( a ) ) ∈ U A Dr , ( n ) ( a ≥ 0 ) . (86) Proof. By (64) , a n , n − 1 = − 2 a_{n,n-1}=-2 a n , n − 1 = − 2 and a n − 1 , n = − 1 a_{n-1,n}=-1 a n − 1 , n = − 1 , so
s n ( α n − 1 ) = α n − 1 + 2 α n , s n − 1 ( α n ) = α n + α n − 1 . (87) s_n(\alpha_{n-1})=\alpha_{n-1}+2\alpha_n,\qquad s_{n-1}(\alpha_n)=\alpha_n+\alpha_{n-1}. \tag*{(87)} s n ( α n − 1 ) = α n − 1 + 2 α n , s n − 1 ( α n ) = α n + α n − 1 . (87) Using (84) , s n ( δ ) = δ s_n(\delta)=\delta s n ( δ ) = δ , and (87) , we compute
α n − 1 → s n − 1 − α n − 1 → w ~ n − 1 δ − α n − 1 = μ → s n δ − ( α n − 1 + 2 α n ) = β ′ , \alpha_{n-1}\xrightarrow{s_{n-1}}-\alpha_{n-1}\xrightarrow{\widetilde{w}_{n-1}}\delta-\alpha_{n-1}=\mu\xrightarrow{s_n}\delta-(\alpha_{n-1}+2\alpha_n)=\beta', α n − 1 s n − 1 − α n − 1 w n − 1 δ − α n − 1 = μ s n δ − ( α n − 1 + 2 α n ) = β ′ , which proves w ^ β ( α n − 1 ) = β ′ \widehat{w}_{\beta}(\alpha_{n-1})=\beta' w β ( α n − 1 ) = β ′ .
Next we identify the standard lift. Let y : = w ~ n − 1 s n − 1 y:=\widetilde{w}_{n-1}s_{n-1} y := w n − 1 s n − 1 . By (84) , w ~ n − 1 ( α n − 1 ) = α n − 1 − δ < 0 \widetilde{w}_{n-1}(\alpha_{n-1})=\alpha_{n-1}-\delta<0 w n − 1 ( α n − 1 ) = α n − 1 − δ < 0 , so the right-descent criterion gives ℓ ( y ) = ℓ ( w ~ n − 1 ) − 1 \ell(y)=\ell(\widetilde{w}_{n-1})-1 ℓ ( y ) = ℓ ( w n − 1 ) − 1 and hence
T y = T w ~ n − 1 ( T n − 1 ) − 1 . T_y=T_{\widetilde{w}_{n-1}}(T_{n-1})^{-1}. T y = T w n − 1 ( T n − 1 ) − 1 . Moreover ⟨ w ~ n − 1 , α n ⟩ = 0 \langle\widetilde{w}_{n-1},\alpha_n\rangle=0 ⟨ w n − 1 , α n ⟩ = 0 gives w ~ n − 1 − 1 ( α n ) = α n \widetilde{w}_{n-1}^{-1}(\alpha_n)=\alpha_n w n − 1 − 1 ( α n ) = α n , so by (87)
y − 1 ( α n ) = s n − 1 w ~ n − 1 − 1 ( α n ) = s n − 1 ( α n ) = α n + α n − 1 > 0 , y^{-1}(\alpha_n)=s_{n-1}\widetilde{w}_{n-1}^{-1}(\alpha_n)=s_{n-1}(\alpha_n)=\alpha_n+\alpha_{n-1}>0, y − 1 ( α n ) = s n − 1 w n − 1 − 1 ( α n ) = s n − 1 ( α n ) = α n + α n − 1 > 0 , whence ℓ ( s n y ) = ℓ ( y ) + 1 \ell(s_ny)=\ell(y)+1 ℓ ( s n y ) = ℓ ( y ) + 1 and
T w ^ β = T n T y = T n T w ~ n − 1 ( T n − 1 ) − 1 = T β . T_{\widehat{w}_{\beta}}=T_nT_y=T_nT_{\widetilde{w}_{n-1}}(T_{n-1})^{-1}=\mathcal{T}_{\beta}. T w β = T n T y = T n T w n − 1 ( T n − 1 ) − 1 = T β . Combining this with the factorization of E μ E_\mu E μ in the proof of proposition 7.2 gives
B ♯ = T n ( T w ~ n − 1 ( T n − 1 ) − 1 ( E n − 1 ) ) = T n ( E μ ) . B^\sharp=T_n\bigl(T_{\widetilde{w}_{n-1}}(T_{n-1})^{-1}(E_{n-1})\bigr)=T_n(E_\mu). B ♯ = T n ( T w n − 1 ( T n − 1 ) − 1 ( E n − 1 ) ) = T n ( E μ ) . Since s n s_n s n preserves the invariant form,
( β ′ ∣ β ′ ) = ( s n μ ∣ s n μ ) = ( μ ∣ μ ) = 4 , (\beta'\mid\beta')=(s_n\mu\mid s_n\mu)=(\mu\mid\mu)=4, ( β ′ ∣ β ′ ) = ( s n μ ∣ s n μ ) = ( μ ∣ μ ) = 4 ,
so β ′ \beta' β ′ and μ \mu μ have the same root parameter Q Q Q , and both E μ ( a ) E_\mu^{(a)} E μ ( a ) and ( B ♯ ) ( a ) (B^\sharp)^{(a)} ( B ♯ ) ( a ) are normalized by [ a ] Q ! [a]_Q! [ a ] Q ! . As T n T_n T n is a K \mathcal{K} K -algebra automorphism,
( B ♯ ) ( a ) = T n ( E μ ) a [ a ] Q ! = T n ( E μ ( a ) ) . (B^\sharp)^{(a)}=\frac{T_n(E_\mu)^a}{[a]_Q!}=T_n(E_\mu^{(a)}). ( B ♯ ) ( a ) = [ a ] Q ! T n ( E μ ) a = T n ( E μ ( a ) ) .
Applying T n T_n T n to (83) and using T n ( K μ ) = K s n ( μ ) = K β ′ T_n(K_\mu)=K_{s_n(\mu)}=K_{\beta'} T n ( K μ ) = K s n ( μ ) = K β ′ gives the explicit expression
( B ♯ ) ( a ) = Q − 2 ( a ) K β ′ a T n ( ( x n − 1 , 1 − ) ( a ) ) , K β ′ = K δ K n − 1 − 1 K n − 2 . (B^\sharp)^{(a)}=Q^{-2(a)}K_{\beta'}^aT_n\left((x_{n-1,1}^{-})^{(a)}\right),\qquad K_{\beta'}=K_\delta K_{n-1}^{-1}K_n^{-2}. ( B ♯ ) ( a ) = Q − 2 ( a ) K β ′ a T n ( ( x n − 1 , 1 − ) ( a ) ) , K β ′ = K δ K n − 1 − 1 K n − 2 .
Here K β ′ ± 1 ∈ U A Dr , ( n ) K_{\beta'}^{\pm1}\in U_A^{\operatorname{Dr},(n)} K β ′ ± 1 ∈ U A Dr , ( n ) , and T n ( ( x n − 1 , 1 − ) ( a ) ) ∈ U A Dr , ( n ) T_n\left((x_{n-1,1}^{-})^{(a)}\right)\in U_A^{\operatorname{Dr},(n)} T n ( ( x n − 1 , 1 − ) ( a ) ) ∈ U A Dr , ( n ) by the last assertion of lemma 7.1, applied with the color n − 1 ≠ n n-1\ne n n − 1 = n . This proves (86) .
Proposition 7.3 completes the integral transport for this non-extra-long root. It remains to identify B ♯ B^\sharp B ♯ with the root vector E β ′ E_{\beta'} E β ′ fixed by the convex order.
Let
β = α n − 1 , η j = δ − 2 ( α j + ⋯ + α n ) ( 1 ≤ j ≤ n ) , c a = ∏ h = 1 a ( q 4 h + 1 ) , c 0 = 1. (88) \beta=\alpha_{n-1},\qquad\eta_j=\delta-2(\alpha_j+\cdots+\alpha_n)\quad(1\le j\le n),\qquad c_a=\prod_{h=1}^{a}(q^{4h}+1),\qquad c_0=1. \tag*{(88)} β = α n − 1 , η j = δ − 2 ( α j + ⋯ + α n ) ( 1 ≤ j ≤ n ) , c a = h = 1 ∏ a ( q 4 h + 1 ) , c 0 = 1. (88)
The roots η j \eta_j η j are extra-long real roots with root parameter Q 2 Q^2 Q 2 , whereas β \beta β and β ′ \beta' β ′ have root parameter Q Q Q .
Lemma 7.4. With μ = δ − α n − 1 \mu=\delta-\alpha_{n-1} μ = δ − α n − 1 , β ′ = δ − ( α n − 1 + 2 α n ) \beta'=\delta-(\alpha_{n-1}+2\alpha_n) β ′ = δ − ( α n − 1 + 2 α n ) , and B ♯ = T n ( E μ ) B^\sharp=T_n(E_\mu) B ♯ = T n ( E μ ) as above, one has
B ♯ = E β ′ . B^\sharp=E_{\beta'}. B ♯ = E β ′ . Consequently ( E β ′ ) ( a ) ∈ U A Dr , ( n ) (E_{\beta'})^{(a)}\in U_A^{\operatorname{Dr},(n)} ( E β ′ ) ( a ) ∈ U A Dr , ( n ) for every a ≥ 0 a\ge0 a ≥ 0 .
Proof. Let e i = α i + ⋯ + α n e_i=\alpha_i+\cdots+\alpha_n e i = α i + ⋯ + α n and w = s 0 s 1 ⋯ s n w=s_0s_1\cdots s_n w = s 0 s 1 ⋯ s n . After transporting the finite-node numbering, [7 Lemma 4.2.3] gives
w m ( e i ) = { e i + m , i + m ≤ n , e i + m − n − δ , i + m > n , 0 ≤ m ≤ n . (89) w^m(e_i)=
\begin{cases}
e_{i+m}, & i+m\le n,\\
e_{i+m-n}-\delta, & i+m>n,
\end{cases}
\qquad0\le m\le n. \tag*{(89)} w m ( e i ) = { e i + m , e i + m − n − δ , i + m ≤ n , i + m > n , 0 ≤ m ≤ n . (89)
Set u = ( s 2 ⋯ s n ) ( s 1 ⋯ s n ) u=(s_2\cdots s_n)(s_1\cdots s_n) u = ( s 2 ⋯ s n ) ( s 1 ⋯ s n ) . Direct calculation gives
u ( e i ) = { e i + 2 , 1 ≤ i ≤ n − 2 , − e 2 , i = n − 1 , − e 1 , i = n . u(e_i)=
\begin{cases}
e_{i+2}, & 1\le i\le n-2,\\
-e_2, & i=n-1,\\
-e_1, & i=n.
\end{cases} u ( e i ) = ⎩ ⎨ ⎧ e i + 2 , − e 2 , − e 1 , 1 ≤ i ≤ n − 2 , i = n − 1 , i = n .
Since w ^ β = s n w ~ n − 1 s n − 1 \widehat{w}_\beta=s_n\widetilde{w}_{n-1}s_{n-1} w β = s n w n − 1 s n − 1 , the translation action gives
w ^ β ( e i ) = s n ( e i − δ ) = e i − δ ( 1 ≤ i ≤ n − 2 ) , \widehat{w}_\beta(e_i)=s_n(e_i-\delta)=e_i-\delta\qquad(1\le i\le n-2), w β ( e i ) = s n ( e i − δ ) = e i − δ ( 1 ≤ i ≤ n − 2 ) ,
w ^ β ( e n − 1 ) = s n ( e n ) = − e n , \widehat{w}_\beta(e_{n-1})=s_n(e_n)=-e_n, w β ( e n − 1 ) = s n ( e n ) = − e n ,
w ^ β ( e n ) = s n ( e n − 1 − δ ) = e n − 1 − δ . \widehat{w}_\beta(e_n)=s_n(e_{n-1}-\delta)=e_{n-1}-\delta. w β ( e n ) = s n ( e n − 1 − δ ) = e n − 1 − δ .
On the other hand, s 0 ( e 1 ) = δ − e 1 s_0(e_1)=\delta-e_1 s 0 ( e 1 ) = δ − e 1 and s 0 ( e j ) = e j s_0(e_j)=e_j s 0 ( e j ) = e j for j ≥ 2 j\ge2 j ≥ 2 , so (89) gives
w n − 2 s 0 u ( e i ) = w n − 2 ( e i + 2 ) = e i − δ ( 1 ≤ i ≤ n − 2 ) , w^{n-2}s_0u(e_i)=w^{n-2}(e_{i+2})=e_i-\delta\qquad(1\le i\le n-2), w n − 2 s 0 u ( e i ) = w n − 2 ( e i + 2 ) = e i − δ ( 1 ≤ i ≤ n − 2 ) ,
w n − 2 s 0 u ( e n − 1 ) = − w n − 2 ( e 2 ) = − e n , w^{n-2}s_0u(e_{n-1})=-w^{n-2}(e_2)=-e_n, w n − 2 s 0 u ( e n − 1 ) = − w n − 2 ( e 2 ) = − e n ,
w n − 2 s 0 u ( e n ) = w n − 2 ( e 1 − δ ) = e n − 1 − δ . w^{n-2}s_0u(e_n)=w^{n-2}(e_1-\delta)=e_{n-1}-\delta. w n − 2 s 0 u ( e n ) = w n − 2 ( e 1 − δ ) = e n − 1 − δ .
Both elements also fix δ \delta δ . Since δ , e 1 , … , e n \delta,e_1,\ldots,e_n δ , e 1 , … , e n form a basis of the affine root lattice, their actions coincide. The Weyl-group action is faithful, so
w ~ β = w n − 2 s 0 u . (90) \widetilde{w}_{\beta}=w^{n-2}s_0u. \tag*{(90)} w β = w n − 2 s 0 u . (90)
For w ~ n − 1 \widetilde{w}_{n-1} w n − 1 , the short and extra-long root strings each contribute n − 1 n-1 n − 1 inversions, while the long strings contribute 2 n − 1 2 + 2 ( n − 1 ) = n ( n − 1 ) 2^{\frac{n-1}{2}}+2(n-1)=n(n-1) 2 2 n − 1 + 2 ( n − 1 ) = n ( n − 1 ) . Hence
ℓ ( w ~ n − 1 ) = ( n − 1 ) + n ( n − 1 ) + ( n − 1 ) = ( n − 1 ) ( n + 2 ) . \ell(\widetilde{w}_{n-1})=(n-1)+n(n-1)+(n-1)=(n-1)(n+2). ℓ ( w n − 1 ) = ( n − 1 ) + n ( n − 1 ) + ( n − 1 ) = ( n − 1 ) ( n + 2 ) .
The descent and ascent calculations in the proof of Proposition Proposition 7.3 give
ℓ ( w ~ β ) = ℓ ( w ~ n − 1 s n − 1 ) + 1 = ℓ ( w ~ n − 1 ) . \ell(\widetilde{w}_{\beta})=\ell(\widetilde{w}_{n-1}s_{n-1})+1=\ell(\widetilde{w}_{n-1}). ℓ ( w β ) = ℓ ( w n − 1 s n − 1 ) + 1 = ℓ ( w n − 1 ) .
The word w n − 2 s 0 ( s 2 ⋯ s n ) ( s 1 ⋯ s n ) w^{n-2}s_0(s_2\cdots s_n)(s_1\cdots s_n) w n − 2 s 0 ( s 2 ⋯ s n ) ( s 1 ⋯ s n ) has exactly
( n − 2 ) ( n + 1 ) + 1 + ( n − 1 ) + n = ( n − 1 ) ( n + 2 ) (n-2)(n+1)+1+(n-1)+n=(n-1)(n+2) ( n − 2 ) ( n + 1 ) + 1 + ( n − 1 ) + n = ( n − 1 ) ( n + 2 )
letters, so (90) is a reduced factorization. Proposition Proposition 7.3 therefore gives
T n T w ~ n − 1 T n − 1 − 1 = T w ~ β = T w n − 2 T 0 T u . (91) T_nT_{\widetilde{w}_{n-1}}T_{n-1}^{-1}=T_{\widetilde{w}_{\beta}}=T_{w^{n-2}}T_0T_u. \tag*{(91)} T n T w n − 1 T n − 1 − 1 = T w β = T w n − 2 T 0 T u . (91)
Since u ( α n − 1 ) = α 1 u(\alpha_{n-1})=\alpha_1 u ( α n − 1 ) = α 1 , Lemma Lemma 1.1 gives T u ( E n − 1 ) = E 1 T_u(E_{n-1})=E_1 T u ( E n − 1 ) = E 1 . And since s 0 ( α 1 ) = α 0 + α 1 = δ − e 1 − e 2 s_0(\alpha_1)=\alpha_0+\alpha_1=\delta-e_1-e_2 s 0 ( α 1 ) = α 0 + α 1 = δ − e 1 − e 2 , (89) gives
w n − 2 s 0 ( α 1 ) = δ − e n − 1 − e n = δ − ( α n − 1 + 2 α n ) = β ′ . w^{n-2}s_0(\alpha_1)=\delta-e_{n-1}-e_n=\delta-(\alpha_{n-1}+2\alpha_n)=\beta'. w n − 2 s 0 ( α 1 ) = δ − e n − 1 − e n = δ − ( α n − 1 + 2 α n ) = β ′ .
By (31) and (33) , w n − 2 s 0 w^{n-2}s_0 w n − 2 s 0 is a prefix followed by s 1 s_1 s 1 , so
E β ′ = T w n − 2 T 0 ( E 1 ) . E_{\beta'}=T_{w^{n-2}}T_0(E_1). E β ′ = T w n − 2 T 0 ( E 1 ) .
Applying (91) to E n − 1 E_{n-1} E n − 1 yields
B ♯ = T n ( E μ ) = T w n − 2 T 0 T u ( E n − 1 ) = T w n − 2 T 0 ( E 1 ) = E β ′ . (92) B^{\sharp}=T_n(E_\mu)=T_{w^{n-2}}T_0T_u(E_{n-1})=T_{w^{n-2}}T_0(E_1)=E_{\beta'}. \tag*{(92)} B ♯ = T n ( E μ ) = T w n − 2 T 0 T u ( E n − 1 ) = T w n − 2 T 0 ( E 1 ) = E β ′ . (92)
The divided-power assertion follows from (86) .
The special three-root cell
Lemma 7.5 (The special three-root cell). Put
A = E β , B = E β ′ , G = E η n . A=E_\beta,\qquad B=E_{\beta'},\qquad G=E_{\eta_n}. A = E β , B = E β ′ , G = E η n . Then A ( a ) , B ( a ) ∈ U A D r , ( n ) A^{(a)},B^{(a)}\in U_A^{\mathrm{Dr},(n)} A ( a ) , B ( a ) ∈ U A Dr , ( n ) for every a ≥ 0 a\geq0 a ≥ 0 , and
A B − B A = [ 2 ] Q G , A G = Q − 2 G A , G B = Q − 2 B G . (93) AB-BA=[2]_QG,\qquad AG=Q^{-2}GA,\qquad GB=Q^{-2}BG. \tag*{(93)} A B − B A = [ 2 ] Q G , A G = Q − 2 G A , GB = Q − 2 B G . (93) Proof. The divided powers of A = E n − 1 = x n − 1 , 0 + A=E_{n-1}=x^+_{n-1,0} A = E n − 1 = x n − 1 , 0 + belong to U A D r , ( n ) U_A^{\mathrm{Dr},(n)} U A Dr , ( n ) by (69) ; those of B B B belong to U A D r , ( n ) U_A^{\mathrm{Dr},(n)} U A Dr , ( n ) by Lemma Lemma 7.4 .
Let W = T w n − 2 W=T_{w^{n-2}} W = T w n − 2 and
a = E 1 , b = T 0 ( E 1 ) , g = T 0 T 1 ( E 0 ) . (94) a=E_1,\qquad b=T_0(E_1),\qquad g=T_0T_1(E_0). \tag*{(94)} a = E 1 , b = T 0 ( E 1 ) , g = T 0 T 1 ( E 0 ) . (94) Equation (89) and Lemma Lemma 1.1 give W ( a ) = A W(a)=A W ( a ) = A . The proof of Lemma Lemma 7.4 gives W ( b ) = B W(b)=B W ( b ) = B . By (89) ,
w n − 1 ( α 0 ) = w n − 1 ( δ − 2 e 1 ) = δ − 2 e n = η n . w^{n-1}(\alpha_0)=w^{n-1}(\delta-2e_1)=\delta-2e_n=\eta_n. w n − 1 ( α 0 ) = w n − 1 ( δ − 2 e 1 ) = δ − 2 e n = η n .
By (31) , w n − 1 w^{n-1} w n − 1 is a prefix followed by s 0 s_0 s 0 ; (33) therefore gives
G = T w n − 1 ( E 0 ) = W T 0 T 1 ⋯ T n ( E 0 ) = W T 0 T 1 ( E 0 ) = W ( g ) , G=T_{w^{n-1}}(E_0)
=WT_0T_1\cdots T_n(E_0)
=WT_0T_1(E_0)=W(g), G = T w n − 1 ( E 0 ) = W T 0 T 1 ⋯ T n ( E 0 ) = W T 0 T 1 ( E 0 ) = W ( g ) ,
where T j ( E 0 ) = E 0 T_j(E_0)=E_0 T j ( E 0 ) = E 0 for j ≥ 2 j\ge2 j ≥ 2 , since a j 0 = 0 a_{j0}=0 a j 0 = 0 .
Moreover, s 1 s 0 s 1 s_1s_0s_1 s 1 s 0 s 1 is reduced in the finite Weyl subgroup of type B 2 B_2 B 2 generated by s 0 , s 1 s_0,s_1 s 0 , s 1 , and
s 1 s 0 s 1 ( α 0 ) = s 1 s 0 ( α 0 + 2 α 1 ) = s 1 ( α 0 + 2 α 1 ) = α 0 . s_1s_0s_1(\alpha_0)=s_1s_0(\alpha_0+2\alpha_1)=s_1(\alpha_0+2\alpha_1)=\alpha_0. s 1 s 0 s 1 ( α 0 ) = s 1 s 0 ( α 0 + 2 α 1 ) = s 1 ( α 0 + 2 α 1 ) = α 0 .
Lemma 1.1 therefore gives
T 1 T 0 T 1 ( E 0 ) = E 0 . T_1T_0T_1(E_0)=E_0. T 1 T 0 T 1 ( E 0 ) = E 0 .
Applying T 1 − 1 T_1^{-1} T 1 − 1 , we obtain
g = T 0 T 1 ( E 0 ) = T 1 − 1 ( E 0 ) . g=T_0T_1(E_0)=T_1^{-1}(E_0). g = T 0 T 1 ( E 0 ) = T 1 − 1 ( E 0 ) .
It therefore suffices to prove
a b − b a = [ 2 ] Q g , a g = Q − 2 g a , g b = Q − 2 b g . (95) ab-ba=[2]_Qg,\qquad ag=Q^{-2}ga,\qquad gb=Q^{-2}bg. \tag*{(95)} ab − ba = [ 2 ] Q g , a g = Q − 2 g a , g b = Q − 2 b g . (95)
Since a 01 = − 1 a_{01}=-1 a 01 = − 1 , a 10 = − 2 a_{10}=-2 a 10 = − 2 , q 0 = Q 2 q_0=Q^2 q 0 = Q 2 , and q 1 = Q q_1=Q q 1 = Q , equation 1.11 gives
b = T 0 ( E 1 ) = − E 0 E 1 + q 0 − 1 E 1 E 0 = − E 0 E 1 + Q − 2 E 1 E 0 , b=T_0(E_1)=-E_0E_1+q_0^{-1}E_1E_0=-E_0E_1+Q^{-2}E_1E_0, b = T 0 ( E 1 ) = − E 0 E 1 + q 0 − 1 E 1 E 0 = − E 0 E 1 + Q − 2 E 1 E 0 ,
g = T 1 − 1 ( E 0 ) = ∑ r = 0 2 ( − 1 ) r q 1 − r E 1 ( r ) E 0 E 1 ( 2 − r ) = E 0 E 1 ( 2 ) − Q − 1 E 1 E 0 E 1 + Q − 2 E 1 ( 2 ) E 0 . g=T_1^{-1}(E_0)=\sum_{r=0}^{2}(-1)^r q_1^{-r}E_1^{(r)}E_0E_1^{(2-r)}
=E_0E_1^{(2)}-Q^{-1}E_1E_0E_1+Q^{-2}E_1^{(2)}E_0. g = T 1 − 1 ( E 0 ) = r = 0 ∑ 2 ( − 1 ) r q 1 − r E 1 ( r ) E 0 E 1 ( 2 − r ) = E 0 E 1 ( 2 ) − Q − 1 E 1 E 0 E 1 + Q − 2 E 1 ( 2 ) E 0 .
Consequently,
a b − b a = E 0 E 1 2 − ( 1 + Q − 2 ) E 1 E 0 E 1 + Q − 2 E 1 2 E 0 = [ 2 ] Q g . ab-ba=E_0E_1^2-(1+Q^{-2})E_1E_0E_1+Q^{-2}E_1^2E_0=[2]_Qg. ab − ba = E 0 E 1 2 − ( 1 + Q − 2 ) E 1 E 0 E 1 + Q − 2 E 1 2 E 0 = [ 2 ] Q g .
For the remaining relations, the Chevalley relations give
[ E 0 , F 1 ] = 0 , K 1 E 0 K 1 − 1 = Q − 2 E 0 . [E_0,F_1]=0,\qquad K_1E_0K_1^{-1}=Q^{-2}E_0. [ E 0 , F 1 ] = 0 , K 1 E 0 K 1 − 1 = Q − 2 E 0 .
Since T 1 ( a ) = − F 1 K 1 T_1(a)=-F_1K_1 T 1 ( a ) = − F 1 K 1 and T 1 − 1 ( a ) = − K 1 − 1 F 1 T_1^{-1}(a)=-K_1^{-1}F_1 T 1 − 1 ( a ) = − K 1 − 1 F 1 , we obtain
T 1 ( a ) E 0 = Q − 2 E 0 T 1 ( a ) , E 0 T 1 − 1 ( a ) = Q − 2 T 1 − 1 ( a ) E 0 . T_1(a)E_0=Q^{-2}E_0T_1(a),\qquad E_0T_1^{-1}(a)=Q^{-2}T_1^{-1}(a)E_0. T 1 ( a ) E 0 = Q − 2 E 0 T 1 ( a ) , E 0 T 1 − 1 ( a ) = Q − 2 T 1 − 1 ( a ) E 0 .
Applying T 1 − 1 T_1^{-1} T 1 − 1 to the first equality and T 0 T 1 T_0T_1 T 0 T 1 to the second gives a g = Q − 2 g a ag=Q^{-2}ga a g = Q − 2 g a and g b = Q − 2 b g gb=Q^{-2}bg g b = Q − 2 b g , respectively. Applying the algebra automorphism W W W to (95) proves (93) .
Proposition 7.6 (Quantum scaled roots when n ≥ 2 n\ge2 n ≥ 2 ). For every a ≥ 0 a\ge0 a ≥ 0 ,
c a E η n ( a ) ∈ U A D r , ( n ) . (96) c_aE_{\eta_n}^{(a)}\in U_A^{\mathrm{Dr},(n)}. \tag*{(96)} c a E η n ( a ) ∈ U A Dr , ( n ) . (96) Proof. Set Z = A B − B A = [ 2 ] Q G Z=AB-BA=[2]_QG Z = A B − B A = [ 2 ] Q G . By Lemma 7.5,
A Z = Q − 2 Z A , Z B = Q − 2 B Z , A ( m ) , B ( m ) ∈ U A D r , ( n ) ( m ≥ 0 ) . AZ=Q^{-2}ZA,\qquad ZB=Q^{-2}BZ,\qquad A^{(m)},B^{(m)}\in U_A^{\mathrm{Dr},(n)}\quad(m\ge0). A Z = Q − 2 Z A , Z B = Q − 2 B Z , A ( m ) , B ( m ) ∈ U A Dr , ( n ) ( m ≥ 0 ) . Apply Lemma 1.5 with D = U A D r , ( n ) D=U_A^{\mathrm{Dr},(n)} D = U A Dr , ( n ) . It gives
Z a [ a ] Q ! ∈ U A D r , ( n ) . (97) \frac{Z^a}{[a]_Q!}\in U_A^{\mathrm{Dr},(n)}. \tag*{(97)} [ a ] Q ! Z a ∈ U A Dr , ( n ) . (97)
For every h ≥ 1 h \ge1 h ≥ 1 , the definitions of the symmetric quantum integers give
[ h ] Q 2 = [ h ] Q Q h + Q − h [ 2 ] Q . [h]_{Q^2} = [h]_Q \frac{Q^h+Q^{-h}}{[2]_Q}. [ h ] Q 2 = [ h ] Q [ 2 ] Q Q h + Q − h .
Multiplying this identity for h = 1 , … , a h=1,\ldots,a h = 1 , … , a gives
[ a ] Q 2 ! = [ a ] Q ! [ 2 ] Q a c ~ a , c ~ a : = ∏ h = 1 a ( Q h + Q − h ) . (98) [a]_{Q^2}! = \frac{[a]_Q!}{[2]_Q^a}\widetilde{c}_a,\qquad\widetilde{c}_a := \prod_{h=1}^{a}(Q^h+Q^{-h}). \tag*{(98)} [ a ] Q 2 ! = [ 2 ] Q a [ a ] Q ! c a , c a := h = 1 ∏ a ( Q h + Q − h ) . (98)
Since Q = q 2 Q=q^2 Q = q 2 , the definition (88) gives
c ~ a = ∏ h = 1 a q − 2 h ( q 4 h + 1 ) = q − a ( a + 1 ) c a . \widetilde{c}_a = \prod_{h=1}^{a}q^{-2h}(q^{4h}+1)=q^{-a(a+1)}c_a. c a = h = 1 ∏ a q − 2 h ( q 4 h + 1 ) = q − a ( a + 1 ) c a .
Since G = E η n G=E_{\eta_n} G = E η n has root parameter Q 2 Q^2 Q 2 , equations (97) –(98) give
Z a [ a ] Q ! = c ~ a G a [ a ] Q 2 ! = q − a ( a + 1 ) c a E η n ( a ) ∈ U A D r , ( n ) . \frac{Z^a}{[a]_Q!}=\widetilde{c}_a\frac{G^a}{[a]_{Q^2}!}=q^{-a(a+1)}c_aE_{\eta_n}^{(a)}\in U_A^{\mathrm{Dr},(n)}. [ a ] Q ! Z a = c a [ a ] Q 2 ! G a = q − a ( a + 1 ) c a E η n ( a ) ∈ U A Dr , ( n ) .
Multiplication by the unit q a ( a + 1 ) q^{a(a+1)} q a ( a + 1 ) proves (96) .
Therefore, the coefficient c a c_a c a is created entirely inside the special three-root cell.
Weyl transport of the scaled roots
Lemma 7.7. Let
w η = s 1 s 2 ⋯ s n − 1 , w_\eta=s_1s_2\cdots s_{n-1}, w η = s 1 s 2 ⋯ s n − 1 , Then
w η ( η n ) = η 1 = α 0 , T w η ( E η n ) = E 0 . (99) w_\eta(\eta_n)=\eta_1=\alpha_0,\qquad T_{w_\eta}(E_{\eta_n})=E_0. \tag*{(99)} w η ( η n ) = η 1 = α 0 , T w η ( E η n ) = E 0 . (99)
Proof. In the finite coordinates α i = e i − e i + 1 \alpha_i=e_i-e_{i+1} α i = e i − e i + 1 for i < n i<n i < n and α n = e n \alpha_n=e_n α n = e n , one has η j = δ − 2 e j \eta_j=\delta-2e_j η j = δ − 2 e j . Hence s j − 1 ( η j ) = η j − 1 s_{j-1}(\eta_j)=\eta_{j-1} s j − 1 ( η j ) = η j − 1 for 2 ≤ j ≤ n 2\le j\le n 2 ≤ j ≤ n , and the first identity follows.
After transporting the finite-node numbering, [7 Cor. 4.2.8] gives
E η n = T w η − 1 ( E 0 ) , E_{\eta_n}=T_{w_\eta}^{-1}(E_0), E η n = T w η − 1 ( E 0 ) ,
since s 0 w s n = s 1 ⋯ s n − 1 = w η s_0ws_n=s_1\cdots s_{n-1}=w_\eta s 0 w s n = s 1 ⋯ s n − 1 = w η . Applying T w η T_{w_\eta} T w η yields
T w η ( E η n ) = E 0 , T_{w_\eta}(E_{\eta_n})=E_0, T w η ( E η n ) = E 0 ,
which proves the second identity in (99) .
Proposition 7.8. For every a ≥ 0 a\ge0 a ≥ 0 ,
c a E 0 ( a ) , c a F 0 ( a ) ∈ U A D r , ( n ) . (100) c_aE_0^{(a)},\qquad c_aF_0^{(a)}\in U_A^{\mathrm{Dr},(n)}. \tag*{(100)} c a E 0 ( a ) , c a F 0 ( a ) ∈ U A Dr , ( n ) . (100)
Proof. Proposition 7.6 gives c a E η n ( a ) ∈ U A D r , ( n ) c_aE_{\eta_n}^{(a)}\in U_A^{\mathrm{Dr},(n)} c a E η n ( a ) ∈ U A Dr , ( n ) . Lemmas 7.1 and 7.7 yield
c a E 0 ( a ) = T w η ( c a E η n ( a ) ) ∈ U A D r , ( n ) . c_aE_0^{(a)}=T_{w_\eta}\left(c_aE_{\eta_n}^{(a)}\right)\in U_A^{\mathrm{Dr},(n)}. c a E 0 ( a ) = T w η ( c a E η n ( a ) ) ∈ U A Dr , ( n ) .
Finally,
c a ( q − 1 ) = q − 2 a ( a + 1 ) c a ( q ) . c_a(q^{-1})=q^{-2a(a+1)}c_a(q). c a ( q − 1 ) = q − 2 a ( a + 1 ) c a ( q ) .
Applying (70) to the positive-root assertion gives
q − 2 a ( a + 1 ) c a F 0 ( a ) ∈ U A D r , ( n ) . q^{-2a(a+1)}c_aF_0^{(a)}\in U_A^{\mathrm{Dr},(n)}. q − 2 a ( a + 1 ) c a F 0 ( a ) ∈ U A Dr , ( n ) .
Multiplication by the unit q 2 a ( a + 1 ) q^{2a(a+1)} q 2 a ( a + 1 ) yields the negative-root assertion.
Cyclotomic support and higher-rank recovery
Theorem 7.9 (Higher-rank B ∗ B^* B ∗ -sufficiency). Recall B ∗ = A [ Φ 8 r ( q ) − 1 ∣ r ≥ 1 ] B^*=\mathcal{A}[\Phi_{8r}(q)^{-1}\mid r\geq1] B ∗ = A [ Φ 8 r ( q ) − 1 ∣ r ≥ 1 ] . For every n ≥ 2 n\geq2 n ≥ 2 ,
B ∗ U A D r , ( n ) = B ∗ U A L u s , ( n ) . (101) B^*U_{\mathcal{A}}^{\mathrm{Dr},(n)}=B^*U_{\mathcal{A}}^{\mathrm{Lus},(n)}. \tag*{(101)} B ∗ U A Dr , ( n ) = B ∗ U A Lus , ( n ) . (101) Proof. By (88) and the cyclotomic factorization used in the proof of Theorem 6.2 , one has
c a = ∏ h = 1 a ( q 4 h + 1 ) , q 4 h + 1 = ∏ d ∣ h h / d o d d Φ 8 d ( q ) . c_a=\prod_{h=1}^{a}(q^{4h}+1),\qquad q^{4h}+1=\prod_{\substack{d\mid h\\h/d\ \mathrm{odd}}}\Phi_{8d}(q). c a = h = 1 ∏ a ( q 4 h + 1 ) , q 4 h + 1 = d ∣ h h / d odd ∏ Φ 8 d ( q ) . Thus every nonunit irreducible factor of c a c_a c a is some Φ 8 d ( q ) \Phi_{8d}(q) Φ 8 d ( q ) , so c a ∈ ( B ∗ ) × c_a\in(B^*)^\times c a ∈ ( B ∗ ) × . Proposition 7.8 therefore gives
E 0 ( a ) , F 0 ( a ) ∈ B ∗ U A D r , ( n ) ( a ≥ 0 ) . E_0^{(a)},F_0^{(a)}\in B^*U_{\mathcal{A}}^{\mathrm{Dr},(n)}\qquad(a\geq0). E 0 ( a ) , F 0 ( a ) ∈ B ∗ U A Dr , ( n ) ( a ≥ 0 ) . Applying Lemma 5.1 with X = U A D r , ( n ) \mathcal{X}=U_{\mathcal{A}}^{\mathrm{Dr},(n)} X = U A Dr , ( n ) and B = B ∗ B=B^* B = B ∗ proves (101) .
This completes the sufficiency direction in twisted type A 2 n ( 2 ) A_{2n}^{(2)} A 2 n ( 2 ) . Under the classical specialization q ↦ 1 q\mapsto1 q ↦ 1 , E 0 ↦ e 0 E_0\mapsto e_0 E 0 ↦ e 0 , we have c a ( 1 ) = 2 a c_a(1)=2^a c a ( 1 ) = 2 a . Thus c a E 0 ( a ) c_aE_0^{(a)} c a E 0 ( a ) specializes to ( 2 e 0 ) ( a ) (2e_0)^{(a)} ( 2 e 0 ) ( a ) , recovering the divided powers of the classical scaled roots in [23 Remark 28] .
Coefficient ideals, root-degree detection, and parity separation in modified form
We now turn to necessity. The present section isolates the common mechanism used in Sections 9 and 10. The successive implications are supplied by lemmas 8.1 to 8.4.
For n , r ≥ 1 n,r\geq1 n , r ≥ 1 , define
c n , r : = ( U A D r , ( n ) : A A E 0 ( r ) ) = { a ∈ A ∣ a E 0 ( r ) ∈ U A D r , ( n ) } . (102) c_{n,r}:=(U_{\mathcal{A}}^{\mathrm{Dr},(n)}:_{\mathcal{A}}AE_0^{(r)})=\{a\in\mathcal{A}\mid aE_0^{(r)}\in U_{\mathcal{A}}^{\mathrm{Dr},(n)}\}. \tag*{(102)} c n , r := ( U A Dr , ( n ) : A A E 0 ( r ) ) = { a ∈ A ∣ a E 0 ( r ) ∈ U A Dr , ( n ) } . (102)
Since U A D r , ( n ) U_{\mathcal{A}}^{\mathrm{Dr},(n)} U A Dr , ( n ) is an A \mathcal{A} A -submodule of U A L u s , ( n ) U_{\mathcal{A}}^{\mathrm{Lus},(n)} U A Lus , ( n ) , the quotient
U A L u s , ( n ) / U A D r , ( n ) U_{\mathcal{A}}^{\mathrm{Lus},(n)}/U_{\mathcal{A}}^{\mathrm{Dr},(n)} U A Lus , ( n ) / U A Dr , ( n )
is an A \mathcal{A} A -module. Writing [ E 0 ( r ) ] [E_0^{(r)}] [ E 0 ( r ) ] for the class of E 0 ( r ) E_0^{(r)} E 0 ( r ) in this quotient, we have
c n , r = Ann A ( [ E 0 ( r ) ] ) . c_{n,r}=\operatorname{Ann}_{\mathcal{A}}([E_0^{(r)}]). c n , r = Ann A ([ E 0 ( r ) ]) .
Hence c n , r c_{n,r} c n , r is an A \mathcal{A} A -ideal.
Unlike the explicit multipliers d r d_r d r and c r c_r c r , the ideal c n , r c_{n,r} c n , r records every coefficient that recovers E 0 ( r ) E_0^{(r)} E 0 ( r ) from the bare real-current core.
Lemma 8.1. For every ordinary localization B = S − 1 A ⊆ K B=S^{-1}\mathcal{A}\subseteq\mathcal{K} B = S − 1 A ⊆ K and every ( m , n , r ) ∈ Z ≥ 1 3 (m,n,r)\in\mathbb{Z}_{\geq1}^3 ( m , n , r ) ∈ Z ≥ 1 3 , if c n , r ⊆ ( Φ m ( q ) ) c_{n,r}\subseteq(\Phi_m(q)) c n , r ⊆ ( Φ m ( q )) and the integral forms U A D r , ( n ) U_{\mathcal{A}}^{\mathrm{Dr},(n)} U A Dr , ( n ) and U A L u s , ( n ) U_{\mathcal{A}}^{\mathrm{Lus},(n)} U A Lus , ( n ) coincide over B B B , then Φ m ( q ) \Phi_m(q) Φ m ( q ) is invertible in B B B .
Proof. Since
E 0 ( r ) ∈ B U A L u s , ( n ) = B U A D r , ( n ) = S − 1 U A D r , ( n ) , E_0^{(r)}\in BU_{\mathcal{A}}^{\mathrm{Lus},(n)}=BU_{\mathcal{A}}^{\mathrm{Dr},(n)}=S^{-1}U_{\mathcal{A}}^{\mathrm{Dr},(n)}, E 0 ( r ) ∈ B U A Lus , ( n ) = B U A Dr , ( n ) = S − 1 U A Dr , ( n ) , denominator clearing gives
∃ s ∈ S : s E 0 ( r ) ∈ U A D r , ( n ) . \exists\,s\in S:\qquad sE_0^{(r)}\in U_{\mathcal{A}}^{\mathrm{Dr},(n)}. ∃ s ∈ S : s E 0 ( r ) ∈ U A Dr , ( n ) .
Hence
s ∈ S ∩ c n , r ⊆ ( Φ m ( q ) ) , ∃ b ∈ A : s = Φ m ( q ) b . s \in S \cap c_{n,r} \subseteq(\Phi_m(q)), \qquad\exists b \in\mathcal{A}: \ s=\Phi_m(q)b. s ∈ S ∩ c n , r ⊆ ( Φ m ( q )) , ∃ b ∈ A : s = Φ m ( q ) b .
Since s − 1 ∈ B s^{-1}\in B s − 1 ∈ B ,
Φ m ( q ) − 1 = b s − 1 ∈ B , Φ m ( q ) ∈ B × . \Phi_m(q)^{-1}=bs^{-1}\in B,\qquad\Phi_m(q)\in B^{\times}. Φ m ( q ) − 1 = b s − 1 ∈ B , Φ m ( q ) ∈ B × .
We next fix the grading conventions for modified quantum groups. Let U ˙ A \dot{U}_{A} U ˙ A be the Lusztig modified integral form for the original regular A 2 n ( 2 ) A_{2n}^{(2)} A 2 n ( 2 ) root datum fixed in (66) . If z ∈ U A L u s , ( n ) z\in U_{A}^{\mathrm{Lus},(n)} z ∈ U A Lus , ( n ) is homogeneous of root degree ν \nu ν , first interpret in U ˙ A \dot{U}_{A} U ˙ A the identity
z 1 λ = 1 λ + ν z 1 λ . (103) z1_{\lambda}=1_{\lambda+\nu}z1_{\lambda}. \tag*{(103)} z 1 λ = 1 λ + ν z 1 λ . (103)
and only then specialize at a root of unity. We do not require the ordinary form to embed globally into the modified form after specialization.
Let ℓ i ∈ Z > 0 \ell_i\in\mathbb{Z}_{>0} ℓ i ∈ Z > 0 . Denote the weight lattice, root lattice, and simple roots of a scaled root datum by
X ♯ , Q ♯ , α i ♯ = ℓ i α i , X^{\sharp},\qquad Q^{\sharp},\qquad\alpha_i^{\sharp}=\ell_i\alpha_i, X ♯ , Q ♯ , α i ♯ = ℓ i α i ,
respectively.
For ϵ ∈ Hom Z ( Q ♯ , Z / 2 Z ) \epsilon\in\operatorname{Hom}_{\mathbb{Z}}(Q^{\sharp},\mathbb{Z}/2\mathbb{Z}) ϵ ∈ Hom Z ( Q ♯ , Z /2 Z ) , e ˉ ∈ Z / 2 Z \bar e\in\mathbb{Z}/2\mathbb{Z} e ˉ ∈ Z /2 Z , put
H e ˉ : = ⨁ λ , μ ∈ X ♯ μ − λ ∈ Q ♯ , ϵ ( μ − λ ) = e ˉ 1 μ U ˙ ♯ 1 λ . (104) \mathcal{H}_{\bar e}:=\bigoplus_{\substack{\lambda,\mu\in X^{\sharp}\\ \mu-\lambda\in Q^{\sharp},\ \epsilon(\mu-\lambda)=\bar e}}1_{\mu}\dot{U}^{\sharp}1_{\lambda}. \tag*{(104)} H e ˉ := λ , μ ∈ X ♯ μ − λ ∈ Q ♯ , ϵ ( μ − λ ) = e ˉ ⨁ 1 μ U ˙ ♯ 1 λ . (104)
The modified block decomposition [21 Sect. 23.1.2 and Lemma 23.2.2(b)] gives
U ˙ ♯ = ⨁ λ , μ ∈ X ♯ μ − λ ∈ Q ♯ 1 μ U ˙ ♯ 1 λ = H 0 ˉ ⊕ H 1 ˉ , H e ˉ H f ˉ ⊆ H e ˉ + f ˉ . (105) \dot{U}^{\sharp}=\bigoplus_{\substack{\lambda,\mu\in X^{\sharp}\\ \mu-\lambda\in Q^{\sharp}}}1_{\mu}\dot{U}^{\sharp}1_{\lambda}=\mathcal{H}_{\bar0}\oplus\mathcal{H}_{\bar1},\qquad\mathcal{H}_{\bar e}\mathcal{H}_{\bar f}\subseteq\mathcal{H}_{\bar e+\bar f}. \tag*{(105)} U ˙ ♯ = λ , μ ∈ X ♯ μ − λ ∈ Q ♯ ⨁ 1 μ U ˙ ♯ 1 λ = H 0 ˉ ⊕ H 1 ˉ , H e ˉ H f ˉ ⊆ H e ˉ + f ˉ . (105)
Indeed,
( 1 μ U ˙ ♯ 1 λ ) ( 1 ρ U ˙ ♯ 1 σ ) = 0 ( λ ≠ ρ ) , (1_{\mu}\dot{U}^{\sharp}1_{\lambda})(1_{\rho}\dot{U}^{\sharp}1_{\sigma})=0\qquad(\lambda\ne\rho), ( 1 μ U ˙ ♯ 1 λ ) ( 1 ρ U ˙ ♯ 1 σ ) = 0 ( λ = ρ ) ,
whereas, for λ = ρ \lambda=\rho λ = ρ ,
( 1 μ U ˙ ♯ 1 λ ) ( 1 λ U ˙ ♯ 1 σ ) ⊆ 1 μ U ˙ ♯ 1 σ , ϵ ( μ − σ ) = ϵ ( μ − λ ) + ϵ ( λ − σ ) . (1_{\mu}\dot{U}^{\sharp}1_{\lambda})(1_{\lambda}\dot{U}^{\sharp}1_{\sigma})\subseteq1_{\mu}\dot{U}^{\sharp}1_{\sigma},\qquad\epsilon(\mu-\sigma)=\epsilon(\mu-\lambda)+\epsilon(\lambda-\sigma). ( 1 μ U ˙ ♯ 1 λ ) ( 1 λ U ˙ ♯ 1 σ ) ⊆ 1 μ U ˙ ♯ 1 σ , ϵ ( μ − σ ) = ϵ ( μ − λ ) + ϵ ( λ − σ ) .
For every scalar extension k ⊗ A M k\otimes_A M k ⊗ A M used below, put
y ˉ : = 1 ⊗ y ∈ k ⊗ A M ( y ∈ M ) . \bar y:=1\otimes y\in k\otimes_A M\qquad(y\in M). y ˉ := 1 ⊗ y ∈ k ⊗ A M ( y ∈ M ) .
This convention is distinct from an overline on an integer, which denotes its class in Z / 2 Z \mathbb{Z}/2\mathbb{Z} Z /2 Z .
Lemma 8.2. For π ∈ { Φ h ( q ) ∣ h ∈ Z ≥ 1 } \pi\in\{\Phi_h(q)\mid h\in\mathbb{Z}_{\geq1}\} π ∈ { Φ h ( q ) ∣ h ∈ Z ≥ 1 } , let k : = Frac ( A / ( π ) ) k:=\operatorname{Frac}(A/(\pi)) k := Frac ( A / ( π )) , U ˙ k : = k ⊗ A U ˙ A \dot{U}_k:=k\otimes_A\dot{U}_A U ˙ k := k ⊗ A U ˙ A , and let U ˙ ♯ \dot{U}^{\sharp} U ˙ ♯ be the modified k k k -form attached to the scaled root datum, and let F ∈ Hom k -alg ( U ˙ k , U ˙ ♯ ) \mathcal{F}\in\operatorname{Hom}_{k\text{-alg}}(\dot{U}_k,\dot{U}^{\sharp}) F ∈ Hom k -alg ( U ˙ k , U ˙ ♯ ) .
Assume
∀ μ ∈ X , F ( 1 μ ) = { 1 μ ♯ , μ ∈ X ♯ , 0 , μ ∉ X ♯ . (106) \forall\mu\in X,\qquad\mathcal{F}(1_{\mu})=
\begin{cases}
1_{\mu}^{\sharp},&\mu\in X^{\sharp},\\
0,&\mu\notin X^{\sharp}.
\end{cases} \tag*{(106)} ∀ μ ∈ X , F ( 1 μ ) = { 1 μ ♯ , 0 , μ ∈ X ♯ , μ ∈ / X ♯ . (106) Then let ν ∈ Q \nu\in Q ν ∈ Q , z ∈ ( U A L u s , ( n ) ) ν z\in(U_A^{\mathrm{Lus},(n)})_{\nu} z ∈ ( U A Lus , ( n ) ) ν , and λ ∈ X \lambda\in X λ ∈ X . If F ( z 1 λ ) ≠ 0 \mathcal{F}(z1_{\lambda})\ne0 F ( z 1 λ ) = 0 , then ν ∈ Q ♯ \nu\in Q^{\sharp} ν ∈ Q ♯ , and F ( z 1 λ ) \mathcal{F}(z1_{\lambda}) F ( z 1 λ ) is a nonzero homogeneous element of U ˙ ♯ \dot{U}^{\sharp} U ˙ ♯ of degree ν \nu ν .
Proof. Let
y : = F ( z 1 λ ‾ ) . y := \mathcal{F}(\overline{z_1\lambda}). y := F ( z 1 λ ) .
By (103) ,
z 1 λ ‾ = 1 λ + ν z 1 1 λ ‾ , y = F ( 1 λ + ν ‾ ) y F ( 1 λ ‾ ) . \overline{z_1\lambda}=\overline{1_{\lambda+\nu}z_1 1_\lambda},\qquad y=\mathcal{F}(\overline{1_{\lambda+\nu}})y\mathcal{F}(\overline{1_\lambda}). z 1 λ = 1 λ + ν z 1 1 λ , y = F ( 1 λ + ν ) y F ( 1 λ ) .
Hence
y ≠ 0 ⟹ λ , λ + ν ∈ X ♯ , 0 ≠ y ∈ 1 λ + ν ♯ U ˙ ♯ 1 λ ♯ . y\ne0 \quad\Longrightarrow\quad\lambda,\lambda+\nu\in X^\sharp,\qquad0\ne y\in1^\sharp_{\lambda+\nu}\dot{U}^\sharp1^\sharp_\lambda. y = 0 ⟹ λ , λ + ν ∈ X ♯ , 0 = y ∈ 1 λ + ν ♯ U ˙ ♯ 1 λ ♯ .
The modified PBW block decomposition gives
1 μ ♯ U ˙ ♯ 1 λ ♯ ≠ 0 ⟹ μ − λ ∈ Q ♯ , 1 μ ♯ U ˙ ♯ 1 λ ♯ ⊆ ( U ˙ ♯ ) μ − λ . 1^\sharp_\mu\dot{U}^\sharp1^\sharp_\lambda\ne0 \quad\Longrightarrow\quad\mu-\lambda\in Q^\sharp,\qquad1^\sharp_\mu\dot{U}^\sharp1^\sharp_\lambda\subseteq(\dot{U}^\sharp)_{\mu-\lambda}. 1 μ ♯ U ˙ ♯ 1 λ ♯ = 0 ⟹ μ − λ ∈ Q ♯ , 1 μ ♯ U ˙ ♯ 1 λ ♯ ⊆ ( U ˙ ♯ ) μ − λ .
Taking μ = λ + ν \mu=\lambda+\nu μ = λ + ν yields
ν ∈ Q ♯ , 0 ≠ y ∈ ( U ˙ ♯ ) ν . \nu\in Q^\sharp,\qquad0\ne y\in(\dot{U}^\sharp)_\nu. ν ∈ Q ♯ , 0 = y ∈ ( U ˙ ♯ ) ν .
For n ∈ Z ≥ 1 n\in\mathbb{Z}_{\ge1} n ∈ Z ≥ 1 , let
C n : = { ( x i , k ε ) ( s ) ∣ 1 ≤ i ≤ n , k ∈ Z , s ∈ Z ≥ 0 , ε ∈ { 1 , − 1 } } ∪ { K δ ± 1 } ∪ { K i ± 1 ∣ 1 ≤ i ≤ n } . \mathcal{C}_n:=\left\{(x^\varepsilon_{i,k})^{(s)}\mid1\le i\le n,\ k\in\mathbb{Z},\ s\in\mathbb{Z}_{\ge0},\ \varepsilon\in\{1,-1\}\right\}\cup\{K_\delta^{\pm1}\}\cup\{K_i^{\pm1}\mid1\le i\le n\}. C n := { ( x i , k ε ) ( s ) ∣ 1 ≤ i ≤ n , k ∈ Z , s ∈ Z ≥ 0 , ε ∈ { 1 , − 1 } } ∪ { K δ ± 1 } ∪ { K i ± 1 ∣ 1 ≤ i ≤ n } .
be the set of generators of U A D r , ( n ) U_A^{\mathrm{Dr},(n)} U A Dr , ( n ) .
Lemma 8.3. Let k \mathbb{k} k be an A \mathcal{A} A -algebra, let
D = D ‾ 0 ⊕ D ‾ 1 , D e ‾ D f ‾ ⊆ D e + f ‾ , D=\overline{D}_0\oplus\overline{D}_1,\qquad D_{\overline e}D_{\overline f}\subseteq D_{\overline{e+f}}, D = D 0 ⊕ D 1 , D e D f ⊆ D e + f , be a Z / 2 Z \mathbb{Z}/2\mathbb{Z} Z /2 Z -graded k \mathbb{k} k -algebra, and let
φ ∈ Hom k -alg ( k ⊗ A U ˙ A , D ) . \varphi\in\operatorname{Hom}_{\mathbb{k}\text{-alg}}\left(\mathbb{k}\otimes_{\mathcal{A}}\dot{U}_A,D\right). φ ∈ Hom k -alg ( k ⊗ A U ˙ A , D ) . If any nonzero image φ ( u 1 λ ‾ ) \varphi(\overline{u1_\lambda}) φ ( u 1 λ ) , with λ ∈ X \lambda\in X λ ∈ X and u ∈ C n u\in\mathcal{C}_n u ∈ C n , belongs to D ‾ 0 \overline{D}_0 D 0 , then
\varphi\left(\overline{U_A^{\mathrm{Dr},(n)}_0}\right)\subseteq\overline{D}_0.
Proof. For
M = u 1 ⋯ u t , u j ∈ C n , γ j : = deg u j , λ j : = ∑ p = j + 1 t γ p , M=u_1\cdots u_t,\qquad u_j\in\mathcal{C}_n,\qquad\gamma_j:=\deg u_j,\qquad\lambda_j:=\sum_{p=j+1}^{t}\gamma_p, M = u 1 ⋯ u t , u j ∈ C n , γ j := deg u j , λ j := p = j + 1 ∑ t γ p ,
one has
M 1 0 = ( u 1 1 λ 1 ) ⋯ ( u t 1 0 ) . M1_0=(u_1 1_{\lambda_1})\cdots(u_t1_0). M 1 0 = ( u 1 1 λ 1 ) ⋯ ( u t 1 0 ) .
Thus
φ ( M 1 0 ‾ ) = 0 or φ ( M 1 0 ‾ ) ∈ ( D ‾ 0 ) t ⊆ D ‾ 0 . \varphi(\overline{M1_0})=0\quad\text{or}\quad\varphi(\overline{M1_0})\in(\overline{D}_0)^t\subseteq\overline{D}_0. φ ( M 1 0 ) = 0 or φ ( M 1 0 ) ∈ ( D 0 ) t ⊆ D 0 .
For t = 0 t=0 t = 0 , use 1 = ( x 1 , 0 + ) ( 0 ) ∈ C n 1=(x^+_{1,0})^{(0)}\in\mathcal{C}_n 1 = ( x 1 , 0 + ) ( 0 ) ∈ C n . Since
U A D r , ( n ) = span A { u 1 ⋯ u t ∣ t ≥ 0 , u j ∈ C n } , U_A^{\mathrm{Dr},(n)}=\operatorname{span}_{\mathcal{A}}\{u_1\cdots u_t\mid t\ge0,\ u_j\in\mathcal{C}_n\}, U A Dr , ( n ) = span A { u 1 ⋯ u t ∣ t ≥ 0 , u j ∈ C n } ,
k \mathbb{k} k -linearity yields the assertion.
Lemma 8.4. Let n , r ∈ Z ≥ 1 n,r\in\mathbb{Z}_{\geq1} n , r ∈ Z ≥ 1 , π ∈ { Φ h ( q ) ∣ h ∈ Z ≥ 1 } \pi\in\{\Phi_h(q)\mid h\in\mathbb{Z}_{\geq1}\} π ∈ { Φ h ( q ) ∣ h ∈ Z ≥ 1 } , k : = Frac ( A / ( π ) ) \mathsf{k}:=\operatorname{Frac}(\mathcal{A}/(\pi)) k := Frac ( A / ( π )) , and let
χ : k ⊗ A ( U A L u s , ( n ) ) 10 ⟶ D ‾ 0 ⊕ D ‾ 1 \chi:\mathsf{k}\otimes_{\mathcal{A}}(U^{\mathrm{Lus},(n)}_{\mathcal{A}})_{10}\longrightarrow\overline{D}_0\oplus\overline{D}_1 χ : k ⊗ A ( U A Lus , ( n ) ) 10 ⟶ D 0 ⊕ D 1 be k \mathsf{k} k -linear, where the sum is internal. Then
χ ( U A D r , ( n ) 10 ‾ ) ⊆ D ‾ 0 , 0 ≠ χ ( E 0 ( r ) 10 ‾ ) ∈ D ‾ 1 } ⟹ c n , r ⊆ ( π ) . (107) \left.
\begin{aligned}
\chi\left(\overline{U^{\mathrm{Dr},(n)}_{\mathcal{A}}{}_{10}}\right)&\subseteq\overline{D}_0,\\
0\ne\chi\left(\overline{E^{(r)}_0{}_{10}}\right)&\in\overline{D}_1
\end{aligned}
\right\}
\Longrightarrow\mathfrak{c}_{n,r}\subseteq(\pi). \tag*{(107)} χ ( U A Dr , ( n ) 10 ) 0 = χ ( E 0 ( r ) 10 ) ⊆ D 0 , ∈ D 1 ⎭ ⎬ ⎫ ⟹ c n , r ⊆ ( π ) . (107)
Proof. For a ∈ c n , r a\in\mathfrak{c}_{n,r} a ∈ c n , r , put
c : = a E 0 ( r ) ∈ U A D r , ( n ) . c:=aE^{(r)}_0\in U^{\mathrm{Dr},(n)}_{\mathcal{A}}. c := a E 0 ( r ) ∈ U A Dr , ( n ) .
Then
χ ( c 10 ‾ ) = a ‾ χ ( E 0 ( r ) 10 ‾ ) ∈ D ‾ 0 . \chi(\overline{c_{10}})=\overline{a}\,\chi\left(\overline{E^{(r)}_0{}_{10}}\right)\in\overline{D}_0. χ ( c 10 ) = a χ ( E 0 ( r ) 10 ) ∈ D 0 .
If a ∉ ( π ) a\notin(\pi) a ∈ / ( π ) , then a ‾ ∈ k × \overline{a}\in\mathsf{k}^{\times} a ∈ k × , whence
0 ≠ χ ( c 10 ‾ ) ∈ D ‾ 0 ∩ D ‾ 1 = { 0 } , 0\ne\chi(\overline{c_{10}})\in\overline{D}_0\cap\overline{D}_1=\{0\}, 0 = χ ( c 10 ) ∈ D 0 ∩ D 1 = { 0 } ,
a contradiction. Therefore
a ∈ ( π ) , c n , r ⊆ ( π ) . a\in(\pi),\qquad\mathfrak{c}_{n,r}\subseteq(\pi). a ∈ ( π ) , c n , r ⊆ ( π ) .
The scaled root datum and cyclotomic obstructions for r ≥ 2 r\geq2 r ≥ 2
The preceding lemmas are quite general, now we apply them to our case of the nonendpoint cyclotomic primes by means of Lusztig’s modified quantum Frobenius. Recall the coefficient ideals c n , a \mathfrak{c}_{n,a} c n , a from (102) . In this section, we use scaled root data and modified quantum Frobenius to prove
c n , r ⊆ ( Φ 8 r ( q ) ) ( n ≥ 1 , r ≥ 2 ) , c 1 , r ⊆ ( Φ 4 r ( q ) ) ( r ≥ 3 odd ) . \mathfrak{c}_{n,r}\subseteq(\Phi_{8r}(q))\quad(n\geq1,\ r\geq2),\qquad\mathfrak{c}_{1,r}\subseteq(\Phi_{4r}(q))\quad(r\geq3\text{ odd}). c n , r ⊆ ( Φ 8 r ( q )) ( n ≥ 1 , r ≥ 2 ) , c 1 , r ⊆ ( Φ 4 r ( q )) ( r ≥ 3 odd ) .
The scaled-data formulas below will also be used at r = 1 r=1 r = 1 in § 6.5 . We therefore retain that value in the definitions and lattice calculations; the Frobenius applications in the present section are restricted to the higher indices stated above.
The scaled datum at Φ 8 r ( q ) \Phi_{8r}(q) Φ 8 r ( q )
Fix r ≥ 1 r\geq1 r ≥ 1 . Consider k r ( 8 ) : = Frac ( A / ( Φ 8 r ( q ) ) ) \mathsf{k}^{(8)}_r:=\operatorname{Frac}(\mathcal{A}/(\Phi_{8r}(q))) k r ( 8 ) := Frac ( A / ( Φ 8 r ( q ))) and fix
ζ r : = q m o d Φ 8 r ( q ) . (108) \zeta_r:=q\bmod\Phi_{8r}(q). \tag*{(108)} ζ r := q mod Φ 8 r ( q ) . (108)
Then
ord ( ζ r ) = 8 r , ord ( ζ r 2 ) = 4 r . \operatorname{ord}(\zeta_r)=8r,\qquad\operatorname{ord}(\zeta_r^2)=4r. ord ( ζ r ) = 8 r , ord ( ζ r 2 ) = 4 r .
Write
( d 0 , d 1 , … , d n ) = ( 4 , 2 , … , 2 , 1 ) . (d_0,d_1,\ldots,d_n)=(4,2,\ldots,2,1). ( d 0 , d 1 , … , d n ) = ( 4 , 2 , … , 2 , 1 ) .
The local quantum parameter is q i = q d i q_i=q^{d_i} q i = q d i . Hence Lusztig’s even-order Frobenius integer is
ℓ i : = ord ( ( q i m o d Φ 8 r ( q ) ) 2 ) = ord ( ζ r 2 d i ) = 8 r gcd ( 8 r , 2 d i ) = 4 r gcd ( 4 r , d i ) . (109) \ell_i:=\operatorname{ord}\left((q_i\bmod\Phi_{8r}(q))^2\right)=\operatorname{ord}(\zeta_r^{2d_i})=\frac{8r}{\gcd(8r,2d_i)}=\frac{4r}{\gcd(4r,d_i)}. \tag*{(109)} ℓ i := ord ( ( q i mod Φ 8 r ( q ) ) 2 ) = ord ( ζ r 2 d i ) = g cd( 8 r , 2 d i ) 8 r = g cd( 4 r , d i ) 4 r . (109)
Therefore
ℓ 0 = r , ℓ i = 2 r ( 1 ≤ i < n ) , ℓ n = 4 r . (110) \ell_0=r,\qquad\ell_i=2r\quad(1\le i<n),\qquad\ell_n=4r. \tag*{(110)} ℓ 0 = r , ℓ i = 2 r ( 1 ≤ i < n ) , ℓ n = 4 r . (110)
The middle family is empty when n = 1 n=1 n = 1 .
Retain the regular A 2 n ( 2 ) A_{2n}^{(2)} A 2 n ( 2 ) root datum ( X , Y , I , ⟨ , ⟩ ) (X,Y,I,\langle\ ,\ \rangle) ( X , Y , I , ⟨ , ⟩) fixed in (66) as the source root datum.
Define
X ∙ : = { λ ∈ X ∣ ⟨ h i , λ ⟩ ∈ ℓ i Z for every i ∈ I } , (111) X^\bullet:=\{\lambda\in X\mid\langle h_i,\lambda\rangle\in\ell_i\mathbb{Z}\text{ for every }i\in I\}, \tag*{(111)} X ∙ := { λ ∈ X ∣ ⟨ h i , λ ⟩ ∈ ℓ i Z for every i ∈ I } , (111)
Y ∙ : = Hom Z ( X ∙ , Z ) , (112) Y^\bullet:=\operatorname{Hom}_{\mathbb{Z}}(X^\bullet,\mathbb{Z}), \tag*{(112)} Y ∙ := Hom Z ( X ∙ , Z ) , (112)
α i ∙ : = ℓ i α i , ⟨ h i ∙ , λ ⟩ : = ℓ i − 1 ⟨ h i , λ ⟩ ( λ ∈ X ∙ ) . (113) \alpha_i^\bullet:=\ell_i\alpha_i,\qquad\langle h_i^\bullet,\lambda\rangle:=\ell_i^{-1}\langle h_i,\lambda\rangle\qquad(\lambda\in X^\bullet). \tag*{(113)} α i ∙ := ℓ i α i , ⟨ h i ∙ , λ ⟩ := ℓ i − 1 ⟨ h i , λ ⟩ ( λ ∈ X ∙ ) . (113)
If
i ⋅ j : = d i a i j = d j a j i i\mathbin{\cdot}j:=d_i a_{ij}=d_j a_{ji} i ⋅ j := d i a ij = d j a j i
is the symmetric form on Z [ I ] \mathbb{Z}[I] Z [ I ] , put
i ⋅ ∙ j : = ℓ i ℓ j ( i ⋅ j ) , d i ∙ : = d i ℓ i 2 . (114) i\mathbin{\cdot^\bullet}j:=\ell_i\ell_j(i\mathbin{\cdot}j),\qquad d_i^\bullet:=d_i\ell_i^2. \tag*{(114)} i ⋅ ∙ j := ℓ i ℓ j ( i ⋅ j ) , d i ∙ := d i ℓ i 2 . (114)
The coefficients d i ∙ d_i^\bullet d i ∙ are the integral symmetrizers of the scaled Cartan datum.
Lemma 9.1. With the data of (111) –(114) , ( X ∙ , Y ∙ , I , ⟨ , ⟩ , ( α i ∙ ) i ∈ I , ( h i ∙ ) i ∈ I ) (X^\bullet,Y^\bullet,I,\langle\ ,\ \rangle,(\alpha_i^\bullet)_{i\in I},(h_i^\bullet)_{i\in I}) ( X ∙ , Y ∙ , I , ⟨ , ⟩ , ( α i ∙ ) i ∈ I , ( h i ∙ ) i ∈ I ) is a regular root datum.
For every ( i , j ) ∈ I 2 (i,j)\in I^2 ( i , j ) ∈ I 2 ,
a i j ∙ : = ⟨ h i ∙ , α j ∙ ⟩ = ℓ i − 1 ℓ j a i j = a j i . (115) a_{ij}^\bullet:=\langle h_i^\bullet,\alpha_j^\bullet\rangle=\ell_i^{-1}\ell_j a_{ij}=a_{ji}. \tag*{(115)} a ij ∙ := ⟨ h i ∙ , α j ∙ ⟩ = ℓ i − 1 ℓ j a ij = a j i . (115) Moreover,
L r ∙ : = ⨁ i ∈ I Z α i ∙ = r Z α 0 ⊕ ⨁ i = 1 n − 1 2 r Z α i ⊕ 4 r Z α n . (116) L_r^\bullet:=\bigoplus_{i\in I}\mathbb{Z}\alpha_i^\bullet=r\mathbb{Z}\alpha_0\oplus\bigoplus_{i=1}^{n-1}2r\mathbb{Z}\alpha_i\oplus4r\mathbb{Z}\alpha_n. \tag*{(116)} L r ∙ := i ∈ I ⨁ Z α i ∙ = r Z α 0 ⊕ i = 1 ⨁ n − 1 2 r Z α i ⊕ 4 r Z α n . (116)
Proof. By (109) ,
ℓ i = min { m ∈ Z ≥ 1 ∣ m d i ∈ 4 r Z } . \ell_i=\min\{m\in\mathbb{Z}_{\geq1}\mid md_i\in4r\mathbb{Z}\}. ℓ i = min { m ∈ Z ≥ 1 ∣ m d i ∈ 4 r Z } .
Thus (111) –(114) are Lusztig’s 4 r 4r 4 r -modified root datum [21 §§2.2.4–2.2.5] . Since the source datum is regular, so is the modified datum. Moreover,
d i ℓ i = 4 r ( i ∈ I ) , d_i\ell_i=4r\qquad(i\in I), d i ℓ i = 4 r ( i ∈ I ) ,
and hence
a i j ∙ = ℓ i − 1 ℓ j a i j = d i d j a i j = a j i . a_{ij}^\bullet=\ell_i^{-1}\ell_j a_{ij}=\frac{d_i}{d_j}a_{ij}=a_{ji}. a ij ∙ = ℓ i − 1 ℓ j a ij = d j d i a ij = a j i .
Finally, (110) gives
L r ∙ = r Z α 0 ⊕ ⨁ i = 1 n − 1 2 r Z α i ⊕ 4 r Z α n , L_r^\bullet=r\mathbb{Z}\alpha_0\oplus\bigoplus_{i=1}^{n-1}2r\mathbb{Z}\alpha_i\oplus4r\mathbb{Z}\alpha_n, L r ∙ = r Z α 0 ⊕ i = 1 ⨁ n − 1 2 r Z α i ⊕ 4 r Z α n ,
Define
ϵ r : L r ∙ ⟶ Z / 2 Z , ϵ r ( ∑ i = 0 n b i α i ∙ ) : = b ‾ 0 . (117) \epsilon_r:L_r^\bullet\longrightarrow\mathbb{Z}/2\mathbb{Z},\qquad
\epsilon_r\left(\sum_{i=0}^{n}b_i\alpha_i^\bullet\right):=\overline{b}_0. \tag*{(117)} ϵ r : L r ∙ ⟶ Z /2 Z , ϵ r ( i = 0 ∑ n b i α i ∙ ) := b 0 . (117)
Lemma 9.2 (Uniform evenness of real-current degrees). Let s ∈ Z ≥ 0 s \in\mathbb{Z}_{\ge0} s ∈ Z ≥ 0 , k ∈ Z k \in\mathbb{Z} k ∈ Z , ε ∈ { 1 , − 1 } \varepsilon\in\{1,-1\} ε ∈ { 1 , − 1 } , and 1 ≤ i ≤ n 1 \le i \le n 1 ≤ i ≤ n . Then deg ( ( x i , k ε ) ( s ) ) = s ( k δ + ε α i ) \deg\left((x_{i,k}^{\varepsilon})^{(s)}\right)=s(k\delta+\varepsilon\alpha_i) deg ( ( x i , k ε ) ( s ) ) = s ( k δ + ε α i ) , and if s ( k δ + ε α i ) ∈ L r ∙ s(k\delta+\varepsilon\alpha_i)\in L_r^{\bullet} s ( k δ + ε α i ) ∈ L r ∙ , then ε r ( s ( k δ + ε α i ) ) = 0 ˉ \varepsilon_r(s(k\delta+\varepsilon\alpha_i))=\bar{0} ε r ( s ( k δ + ε α i )) = 0 ˉ .
Proof. Let
ν : = s ( k δ + ε α i ) . \nu:= s(k\delta+\varepsilon\alpha_i). ν := s ( k δ + ε α i ) .
Since
δ = α 0 + 2 α 1 + ⋯ + 2 α n , \delta=\alpha_0+2\alpha_1+\cdots+2\alpha_n, δ = α 0 + 2 α 1 + ⋯ + 2 α n ,
we have
ν = s k α 0 + ∑ 1 ≤ j ≤ n j ≠ i 2 s k α j + s ( 2 k + ε ) α i . \nu=sk\alpha_0+\sum_{\substack{1\le j\le n\\j\ne i}}2sk\alpha_j+s(2k+\varepsilon)\alpha_i. ν = s k α 0 + 1 ≤ j ≤ n j = i ∑ 2 s k α j + s ( 2 k + ε ) α i .
If ν ∈ L r ∙ \nu\in L_r^{\bullet} ν ∈ L r ∙ , its α 0 = r α 0 \alpha_0=r\alpha_0 α 0 = r α 0 coordinate is
b 0 = s k r ∈ Z . b_0=\frac{sk}{r}\in\mathbb{Z}. b 0 = r s k ∈ Z .
It remains to prove b 0 ∈ 2 Z b_0\in2\mathbb{Z} b 0 ∈ 2 Z .
If i < n i<n i < n , the α n \alpha_n α n -coordinate of ν \nu ν is 2 s k 2sk 2 s k . Hence
ν ∈ L r ∙ ⟹ 4 r ∣ 2 s k ⟹ 2 r ∣ s k ⟹ b 0 = s k r ∈ 2 Z . \nu\in L_r^{\bullet}\ \Longrightarrow\ 4r\mid2sk\ \Longrightarrow\ 2r\mid sk\ \Longrightarrow\ b_0=\frac{sk}{r}\in2\mathbb{Z}. ν ∈ L r ∙ ⟹ 4 r ∣ 2 s k ⟹ 2 r ∣ s k ⟹ b 0 = r s k ∈ 2 Z .
Suppose i = n i=n i = n . The α 0 \alpha_0 α 0 - and α n \alpha_n α n -coordinates give
ν ∈ L r ∙ ⟹ r ∣ s k , 4 r ∣ s ( 2 k + ε ) . \nu\in L_r^{\bullet}\ \Longrightarrow\ r\mid sk,\qquad4r\mid s(2k+\varepsilon). ν ∈ L r ∙ ⟹ r ∣ s k , 4 r ∣ s ( 2 k + ε ) .
Since
gcd ( k , 2 k + ε ) = gcd ( k , ε ) = 1 , \gcd(k,2k+\varepsilon)=\gcd(k,\varepsilon)=1, g cd( k , 2 k + ε ) = g cd( k , ε ) = 1 ,
r ∣ s k , r ∣ s ( 2 k + ε ) ⟹ r ∣ s . r\mid sk,\quad r\mid s(2k+\varepsilon)\ \Longrightarrow\ r\mid s. r ∣ s k , r ∣ s ( 2 k + ε ) ⟹ r ∣ s .
Write s = r t s=rt s = r t . Then
4 r ∣ r t ( 2 k + ε ) ⟹ 4 ∣ t ( 2 k + ε ) . 4r\mid rt(2k+\varepsilon)\ \Longrightarrow\ 4\mid t(2k+\varepsilon). 4 r ∣ r t ( 2 k + ε ) ⟹ 4 ∣ t ( 2 k + ε ) .
Because 2 k + ε 2k+\varepsilon 2 k + ε is odd,
gcd ( 4 , 2 k + ε ) = 1 ⟹ 4 ∣ t . \gcd(4,2k+\varepsilon)=1\ \Longrightarrow\ 4\mid t. g cd( 4 , 2 k + ε ) = 1 ⟹ 4 ∣ t .
Consequently
b 0 = s k r = t k ∈ 4 Z ⊆ 2 Z . b_0=\frac{sk}{r}=tk\in4\mathbb{Z}\subseteq2\mathbb{Z}. b 0 = r s k = t k ∈ 4 Z ⊆ 2 Z .
Thus ε r ( ν ) = 0 ˉ \varepsilon_r(\nu)=\bar{0} ε r ( ν ) = 0 ˉ in both cases. The same chains include s = 0 s=0 s = 0 and k = 0 k=0 k = 0 . □
Let’s explain why the affine-coordinate parity is chosen. The parity ε r \varepsilon_r ε r detects the affine coordinate: the Frobenius image of E 0 ( r ) 1 E_0^{(r)}{}_1 E 0 ( r ) 1 has α 0 \alpha_0 α 0 -coordinate one, whereas the lattice divisibilities force every surviving real-current degree to have even affine coordinate. Lemma 9.2 remains valid for r = 1 r=1 r = 1 . The restriction r ≥ 2 r\ge2 r ≥ 2 below comes solely from the numerical inequalities ℓ i ≥ 1 − a i j \ell_i\ge1-a_{ij} ℓ i ≥ 1 − a ij required in Lusztig’s Frobenius theorem for i ≠ j i\ne j i = j with ℓ j ≥ 2 \ell_j\ge2 ℓ j ≥ 2 . This explains why Φ 8 \Phi_8 Φ 8 requires a different detector rather than a different parity.
We now assume r ≥ 2 r\ge2 r ≥ 2 . Let U ˙ A L u s , ( n ) \dot{U}_{\mathrm{A}}^{\mathrm{Lus},(n)} U ˙ A Lus , ( n ) be the source modified integral form, and let U ˙ A , r ∙ \dot{U}_{\mathrm{A},r}^{\bullet} U ˙ A , r ∙ be the modified integral form attached to the scaled datum. Let
U ˙ r L u s , ( n ) : = k r ( 8 ) ⊗ A , q ↦ ζ r U ˙ A L u s , ( n ) , U ˙ r ∙ : = k r ( 8 ) ⊗ A , q ↦ ζ r U ˙ A , r ∙ . (118) \dot{U}_{r}^{\mathrm{Lus},(n)}:=\mathbb{k}_{r}^{(8)}\otimes_{\mathbb{A},\,q\mapsto\zeta_r}\dot{U}_{\mathrm{A}}^{\mathrm{Lus},(n)},\qquad
\dot{U}_{r}^{\bullet}:=\mathbb{k}_{r}^{(8)}\otimes_{\mathbb{A},\,q\mapsto\zeta_r}\dot{U}_{\mathrm{A},r}^{\bullet}. \tag*{(118)} U ˙ r Lus , ( n ) := k r ( 8 ) ⊗ A , q ↦ ζ r U ˙ A Lus , ( n ) , U ˙ r ∙ := k r ( 8 ) ⊗ A , q ↦ ζ r U ˙ A , r ∙ . (118)
For e ˉ ∈ Z / 2 Z \bar e \in\mathbb{Z}/2\mathbb{Z} e ˉ ∈ Z /2 Z , let
H e ˉ ( r ) : = ⨁ λ , μ ∈ X ∙ μ − λ ∈ L r ∙ , ϵ r ( μ − λ ) = e ˉ 1 μ ∙ U ˙ r ∙ 1 λ . (119) \mathcal{H}_{\bar e}^{(r)} := \bigoplus_{\substack{\lambda,\mu\in X^\bullet\\ \mu-\lambda\in L_r^\bullet,\ \epsilon_r(\mu-\lambda)=\bar e}} 1_\mu^\bullet\dot{U}_r^\bullet1_\lambda. \tag*{(119)} H e ˉ ( r ) := λ , μ ∈ X ∙ μ − λ ∈ L r ∙ , ϵ r ( μ − λ ) = e ˉ ⨁ 1 μ ∙ U ˙ r ∙ 1 λ . (119)
Then (105) gives
U ˙ r ∙ = H 0 ˉ ( r ) ⊕ H 1 ˉ ( r ) , H e ˉ ( r ) H f ˉ ( r ) ⊆ H e ˉ + f ˉ ( r ) . (120) \dot{U}_r^\bullet=\mathcal{H}_{\bar0}^{(r)}\oplus\mathcal{H}_{\bar1}^{(r)},\qquad\mathcal{H}_{\bar e}^{(r)}\mathcal{H}_{\bar f}^{(r)}\subseteq\mathcal{H}_{\bar e+\bar f}^{(r)}. \tag*{(120)} U ˙ r ∙ = H 0 ˉ ( r ) ⊕ H 1 ˉ ( r ) , H e ˉ ( r ) H f ˉ ( r ) ⊆ H e ˉ + f ˉ ( r ) . (120)
The following is [21 §§35.1.1–35.1.3, Thm. 35.1.9] , written in the parameter ( q ) (q) ( q ) .
Lemma 9.3 (Lusztig’s modified quantum Frobenius criterion in the q q q -parameter). Let k k k be a field and q ˉ ∈ k × \bar q\in k^\times q ˉ ∈ k × . Assume and set
char k = 0 , N : = ord ( q ˉ 2 ) ∈ 2 Z > 0 , A N ′ : = A / ( Φ 2 N ( q ) ) , \operatorname{char} k=0,\qquad N:=\operatorname{ord}(\bar q^2)\in2\mathbb{Z}_{>0},\qquad\mathcal{A}'_N:=\mathcal{A}/(\Phi_{2N}(q)), char k = 0 , N := ord ( q ˉ 2 ) ∈ 2 Z > 0 , A N ′ := A / ( Φ 2 N ( q )) , and assume
( A → q ↦ q ˉ k ) = ( A ↠ A N ′ ⟶ k ) . \left(\mathcal{A}\xrightarrow{q\mapsto\bar q} k\right)=\left(\mathcal{A}\twoheadrightarrow\mathcal{A}'_N\longrightarrow k\right). ( A q ↦ q ˉ k ) = ( A ↠ A N ′ ⟶ k ) . Let
( d i ) i ∈ I ∈ Z > 0 I (the source symmetrizers) , d i a i j = d j a j i ( ( i , j ) ∈ I 2 ) , (d_i)_{i\in I}\in\mathbb{Z}_{>0}^{I}\quad\text{(the source symmetrizers)},\qquad d_i a_{ij}=d_j a_{ji}\quad((i,j)\in I^2), ( d i ) i ∈ I ∈ Z > 0 I (the source symmetrizers) , d i a ij = d j a j i (( i , j ) ∈ I 2 ) , and, for every i ∈ I i\in I i ∈ I , define
ℓ i : = min { m ∈ Z ≥ 1 ∣ m d i ∈ N Z } = N gcd ( N , d i ) = ord ( q ˉ 2 d i ) . \ell_i:=\min\{m\in\mathbb{Z}_{\geq1}\mid md_i\in N\mathbb{Z}\}=\frac{N}{\gcd(N,d_i)}=\operatorname{ord}(\bar q^{2d_i}). ℓ i := min { m ∈ Z ≥ 1 ∣ m d i ∈ N Z } = g cd( N , d i ) N = ord ( q ˉ 2 d i ) . Let U ˙ k \dot{U}_k U ˙ k be the specialized source modified form and U ˙ k ∙ \dot{U}_k^\bullet U ˙ k ∙ the specialized modified form for the associated ( ℓ i ) i ∈ I (\ell_i)_{i\in I} ( ℓ i ) i ∈ I -scaled root datum. Assume
(i) There exists a function o : I ⟶ { 1 , − 1 } o:I\longrightarrow\{1,-1\} o : I ⟶ { 1 , − 1 } such that o ( i ) = − o ( j ) o(i)=-o(j) o ( i ) = − o ( j ) whenever i ≠ j i\ne j i = j and a i j ≠ 0 a_{ij}\ne0 a ij = 0 .
(ii) For all i , j ∈ I i,j\in I i , j ∈ I with i ≠ j i\ne j i = j and ℓ j ≥ 2 \ell_j\geq2 ℓ j ≥ 2 , one has
ℓ i ≥ 1 − a i j . (121) \ell_i\geq1-a_{ij}. \tag*{(121)} ℓ i ≥ 1 − a ij . (121) Then there exists a k k k -algebra homomorphism Fr : U ˙ k ⟶ U ˙ k ∙ \operatorname{Fr}:\dot{U}_k\longrightarrow\dot{U}_k^\bullet Fr : U ˙ k ⟶ U ˙ k ∙ . For every λ ∈ X \lambda\in X λ ∈ X , it sends 1 λ 1_\lambda 1 λ to 1 λ ∙ 1_\lambda^\bullet 1 λ ∙ if λ ∈ X ∙ \lambda\in X^\bullet λ ∈ X ∙ , and to zero otherwise. Moreover, for all i ∈ I i\in I i ∈ I , s ≥ 0 s\geq0 s ≥ 0 , and λ ∈ X \lambda\in X λ ∈ X , with s s s an integer, it satisfies
Fr ( E i ( s ) 1 λ ) = { E i ∙ , ( s / ℓ i ) 1 λ ∙ , λ ∈ X ∙ , ℓ i ∣ s , 0 , otherwise , \operatorname{Fr}(E_i^{(s)}1_\lambda)=
\begin{cases}
E_i^{\bullet,(s/\ell_i)}1_\lambda^\bullet,& \lambda\in X^\bullet,\ \ell_i\mid s,\\
0,& \text{otherwise},
\end{cases} Fr ( E i ( s ) 1 λ ) = { E i ∙ , ( s / ℓ i ) 1 λ ∙ , 0 , λ ∈ X ∙ , ℓ i ∣ s , otherwise , Fr ( F i ( s ) 1 λ ) = { F i ∙ , ( s / ℓ i ) 1 λ ∙ , λ ∈ X ∙ , ℓ i ∣ s , 0 , otherwise . \operatorname{Fr}(F_i^{(s)}1_\lambda)=
\begin{cases}
F_i^{\bullet,(s/\ell_i)}1_\lambda^\bullet,& \lambda\in X^\bullet,\ \ell_i\mid s,\\
0,& \text{otherwise}.
\end{cases} Fr ( F i ( s ) 1 λ ) = { F i ∙ , ( s / ℓ i ) 1 λ ∙ , 0 , λ ∈ X ∙ , ℓ i ∣ s , otherwise .
Application to the Φ 8 r ( q ) \Phi_{8r}(q) Φ 8 r ( q ) specialization.
∀ ( n , r ) ∈ Z ≥ 1 × Z ≥ 2 , ∃ Fr n , r ∈ Hom Z k ( 8 ) -alg ( U ˙ r , U ˙ r ∙ ) : (122) \forall(n,r)\in\mathbb{Z}_{\geq1}\times\mathbb{Z}_{\geq2},\qquad\exists\ \operatorname{Fr}_{n,r}\in\operatorname{Hom}_{\mathbb{Z}_k^{(8)}}\text{-alg}(\dot{U}_r,\dot{U}_r^\bullet): \tag*{(122)} ∀ ( n , r ) ∈ Z ≥ 1 × Z ≥ 2 , ∃ Fr n , r ∈ Hom Z k ( 8 ) -alg ( U ˙ r , U ˙ r ∙ ) : (122)
∀ λ ∈ X , Fr n , r ( 1 λ ) = { 1 λ ∙ , λ ∈ X ∙ , 0 , λ ∉ X ∙ , (123) \forall\lambda\in X,\qquad\operatorname{Fr}_{n,r}(1_\lambda)=
\begin{cases}
1_\lambda^\bullet,& \lambda\in X^\bullet,\\
0,& \lambda\notin X^\bullet,
\end{cases} \tag*{(123)} ∀ λ ∈ X , Fr n , r ( 1 λ ) = { 1 λ ∙ , 0 , λ ∈ X ∙ , λ ∈ / X ∙ , (123)
Fr n , r ( E 0 ( r ) 1 0 ) = E 0 ∙ 1 0 ∙ ≠ 0. (124) \operatorname{Fr}_{n,r}(E_0^{(r)}1_0)=E_0^\bullet1_0^\bullet\ne0. \tag*{(124)} Fr n , r ( E 0 ( r ) 1 0 ) = E 0 ∙ 1 0 ∙ = 0. (124)
Proof. Here
N = ord ( ζ r 2 ) = 4 r , 2 N = 8 r N=\operatorname{ord}(\zeta_r^2)=4r,\qquad2N=8r N = ord ( ζ r 2 ) = 4 r , 2 N = 8 r
Since ξ 8 r ∈ C \xi_{8r}\in\mathbb{C} ξ 8 r ∈ C is a primitive 8 r 8r 8 r -th root of unity,
A / ( Φ 8 r ( q ) ) ≅ Z [ ξ 8 r , ξ 8 r − 1 ] = Z [ ξ 8 r ] . \mathcal{A}/(\Phi_{8r}(q))\cong\mathbb{Z}[\xi_{8r},\xi_{8r}^{-1}]=\mathbb{Z}[\xi_{8r}]. A / ( Φ 8 r ( q )) ≅ Z [ ξ 8 r , ξ 8 r − 1 ] = Z [ ξ 8 r ] .
This is an integral domain of characteristic zero. Consequently,
k r ( 8 ) = Frac ( A / ( Φ 8 r ( q ) ) ) ≅ Q ( ξ 8 r ) \mathbb{k}_r^{(8)}=\operatorname{Frac}(\mathcal{A}/(\Phi_{8r}(q)))\cong\mathbb{Q}(\xi_{8r}) k r ( 8 ) = Frac ( A / ( Φ 8 r ( q ))) ≅ Q ( ξ 8 r )
is a cyclotomic field of characteristic zero. Thus the field hypothesis in the criterion holds. Moreover, N N N is even and
A ⟶ k r ( 8 ) \mathcal{A}\longrightarrow\mathbb{k}_r^{(8)} A ⟶ k r ( 8 )
factors through A N ′ = A / ( Φ 8 r ( q ) ) \mathcal{A}'_N=\mathcal{A}/(\Phi_{8r}(q)) A N ′ = A / ( Φ 8 r ( q )) by (108) . Set
oFr ( i ) : = ( − 1 ) n − i ( i ∈ I ) . \operatorname{oFr}(i):=(-1)^{n-i}\qquad(i\in I). oFr ( i ) := ( − 1 ) n − i ( i ∈ I ) .
Then
i ≠ j , a i j ≠ 0 ⟹ ∣ i − j ∣ = 1 ⟹ oFr ( i ) = − oFr ( j ) . i\ne j,\quad a_{ij}\ne0\quad\Longrightarrow\quad|i-j|=1\quad\Longrightarrow\quad\operatorname{oFr}(i)=-\operatorname{oFr}(j). i = j , a ij = 0 ⟹ ∣ i − j ∣ = 1 ⟹ oFr ( i ) = − oFr ( j ) .
Since r ≥ 2 r\ge2 r ≥ 2 , (110) and the Cartan matrices fixed in § 5.1 give, for i ≠ j i\ne j i = j ,
1 − a i j ≤ { 2 , i = 0 , 3 , 1 ≤ i < n , 5 , i = n ≤ { r , i = 0 , 2 r , 1 ≤ i < n , 4 r , i = n = : ℓ i . 1-a_{ij}\le
\begin{cases}
2,&i=0,\\
3,&1\le i<n,\\
5,&i=n
\end{cases}
\le
\begin{cases}
r,&i=0,\\
2r,&1\le i<n,\\
4r,&i=n
\end{cases}
=:\ell_i. 1 − a ij ≤ ⎩ ⎨ ⎧ 2 , 3 , 5 , i = 0 , 1 ≤ i < n , i = n ≤ ⎩ ⎨ ⎧ r , 2 r , 4 r , i = 0 , 1 ≤ i < n , i = n =: ℓ i .
Thus conditions (i)–(ii) of the Frobenius criterion hold, and the criterion gives (122) –(123) . Since
0 ∈ X ∙ , ℓ 0 = r , 0\in X^\bullet,\qquad\ell_0=r, 0 ∈ X ∙ , ℓ 0 = r ,
its positive block formula gives
Fr n , r ( E 0 ( r ) 1 0 ) = E 0 ∙ , ( r / ℓ 0 ) 1 0 ∙ = E 0 ∙ 1 0 ∙ . \operatorname{Fr}_{n,r}(E_0^{(r)}1_0)=E_0^{\bullet,(r/\ell_0)}1_0^\bullet=E_0^\bullet1_0^\bullet. Fr n , r ( E 0 ( r ) 1 0 ) = E 0 ∙ , ( r / ℓ 0 ) 1 0 ∙ = E 0 ∙ 1 0 ∙ .
The final element is nonzero by the modified PBW basis [21 §§23.2] ; this proves (124) .
Theorem 9.4. For any integers n ≥ 1 n\ge1 n ≥ 1 and r ≥ 2 r\ge2 r ≥ 2 , one has
c n , r ⊆ ( Φ 8 r ( q ) ) . (125) c_{n,r}\subseteq(\Phi_{8r}(q)). \tag*{(125)} c n , r ⊆ ( Φ 8 r ( q )) . (125)
Proof. Fix
( n , r ) ∈ Z ≥ 1 × Z ≥ 2 , ( λ , u ) ∈ X × C n , ν : = deg u . (n,r)\in\mathbb{Z}_{\ge1}\times\mathbb{Z}_{\ge2},\qquad(\lambda,u)\in X\times\mathcal{C}_n,\qquad\nu:=\deg u. ( n , r ) ∈ Z ≥ 1 × Z ≥ 2 , ( λ , u ) ∈ X × C n , ν := deg u .
equation (122) supplies the algebra homomorphism required in lemma 8.2, while (123) is precisely its block-idempotent hypothesis (106) . If
Fr n , r ( u 1 λ ) ≠ 0 \operatorname{Fr}_{n,r}(u1_\lambda)\ne0 Fr n , r ( u 1 λ ) = 0
and u u u is a real-current generator, then
ν ∈ L r ∙ by xref‘lemma 8.2‘ , ϵ r ( ν ) = 0 by xref‘lemma 9.2‘ , \nu\in L_r^\bullet\quad\text{by {xref}`lemma 8.2`},\qquad\epsilon_r(\nu)=0\quad\text{by {xref}`lemma 9.2`}, ν ∈ L r ∙ by xref ‘lemma 8.2‘ , ϵ r ( ν ) = 0 by xref ‘lemma 9.2‘ ,
and hence
0 ≠ Fr n , r ( u 1 λ ) ∈ H 0 ˉ ( r ) . 0\ne\operatorname{Fr}_{n,r}(u1_\lambda)\in\mathcal{H}_{\bar{0}}^{(r)}. 0 = Fr n , r ( u 1 λ ) ∈ H 0 ˉ ( r ) .
For a toral generator,
ν = 0 , Fr n , r ( u 1 λ ‾ ) ∈ H 0 ( r ) . \nu= 0,\qquad\operatorname{Fr}_{n,r}(\overline{u_1\lambda}) \in\mathcal{H}^{(r)}_0. ν = 0 , Fr n , r ( u 1 λ ) ∈ H 0 ( r ) .
Thus lemma 8.3 gives
Fr n , r ( U A D r , ( n ) ‾ 1 0 ) ⊆ H 0 ( r ) . (126) \operatorname{Fr}_{n,r}\left(\overline{U^{\mathrm{Dr},(n)}_{\mathcal{A}}}_{1_0}\right) \subseteq\mathcal{H}^{(r)}_0. \tag*{(126)} Fr n , r ( U A Dr , ( n ) 1 0 ) ⊆ H 0 ( r ) . (126)
Moreover,
r α 0 = α 0 ⋆ ∈ L r ⋆ , ϵ r ( r α 0 ) = 1 , r\alpha_0=\alpha_0^\star\in L^\star_r,\qquad\epsilon_r(r\alpha_0)=1, r α 0 = α 0 ⋆ ∈ L r ⋆ , ϵ r ( r α 0 ) = 1 ,
and (124) gives
0 ≠ Fr n , r ( E 0 ( r ) 1 0 ‾ ) = E 0 ⋆ 1 0 ∈ H 1 ( r ) . 0\ne\operatorname{Fr}_{n,r}\left(\overline{E^{(r)}_0 1_0}\right)=E^\star_0 1_0\in\mathcal{H}^{(r)}_1. 0 = Fr n , r ( E 0 ( r ) 1 0 ) = E 0 ⋆ 1 0 ∈ H 1 ( r ) .
Hence lemma 8.4, applied to
χ n , r : = Fr n , r ∣ k r ( 8 ) ⊗ A ( U A L u s , ( n ) ) 1 0 , \chi_{n,r}:=\left.\operatorname{Fr}_{n,r}\right|_{\mathbb{k}^{(8)}_r\otimes_{\mathcal{A}}(U^{\mathrm{Lus},(n)}_{\mathcal{A}})_{1_0}}, χ n , r := Fr n , r ∣ k r ( 8 ) ⊗ A ( U A Lus , ( n ) ) 1 0 ,
gives
c n , r ⊆ ( Φ 8 r ( q ) ) , \mathfrak{c}_{n,r}\subseteq\left(\Phi_{8r}(q)\right), c n , r ⊆ ( Φ 8 r ( q ) ) ,
which is (125) .
The rank-one primes Φ 4 r ( q ) \Phi_{4r}(q) Φ 4 r ( q ) for odd r ≥ 3 r\ge3 r ≥ 3
We now take n = 1 n=1 n = 1 and let r ≥ 3 r\ge3 r ≥ 3 be odd. Let
k r ( 4 ) : = Frac ( A / ( Φ 4 r ( q ) ) ) , ξ r : = q m o d Φ 4 r ( q ) . (127) \mathbb{k}^{(4)}_r:=\operatorname{Frac}\left(\mathcal{A}/(\Phi_{4r}(q))\right),\qquad\xi_r:=q\bmod\Phi_{4r}(q). \tag*{(127)} k r ( 4 ) := Frac ( A / ( Φ 4 r ( q )) ) , ξ r := q mod Φ 4 r ( q ) . (127)
Then
ord ( ξ r ) = 4 r , ord ( ξ r 2 ) = 2 r . \operatorname{ord}(\xi_r)=4r,\qquad\operatorname{ord}(\xi_r^2)=2r. ord ( ξ r ) = 4 r , ord ( ξ r 2 ) = 2 r .
For ( d 0 , d 1 ) = ( 4 , 1 ) (d_0,d_1)=(4,1) ( d 0 , d 1 ) = ( 4 , 1 ) ,
ℓ i = ord ( ξ r 2 d i ) = 4 r gcd ( 4 r , 2 d i ) . \ell_i=\operatorname{ord}\left(\xi_r^{2d_i}\right)=\frac{4r}{\gcd(4r,2d_i)}. ℓ i = ord ( ξ r 2 d i ) = g cd( 4 r , 2 d i ) 4 r .
Since r r r is odd,
gcd ( 4 r , 8 ) = 4 , gcd ( 4 r , 2 ) = 2 , \gcd(4r,8)=4,\qquad\gcd(4r,2)=2, g cd( 4 r , 8 ) = 4 , g cd( 4 r , 2 ) = 2 ,
and hence
ℓ 0 = r , ℓ 1 = 2 r . (128) \ell_0=r,\qquad\ell_1=2r. \tag*{(128)} ℓ 0 = r , ℓ 1 = 2 r . (128)
Set
X r ⋆ : = { λ ∈ X ∣ ⟨ h i , λ ⟩ ∈ ℓ i Z for i = 0 , 1 } , α i ⋆ : = ℓ i α i . (129) X_r^\star:=\{\lambda\in X\mid\langle h_i,\lambda\rangle\in\ell_i\mathbb{Z}\text{ for }i=0,1\},\qquad\alpha_i^\star:=\ell_i\alpha_i. \tag*{(129)} X r ⋆ := { λ ∈ X ∣ ⟨ h i , λ ⟩ ∈ ℓ i Z for i = 0 , 1 } , α i ⋆ := ℓ i α i . (129)
Set also
Y r ⋆ : = Hom Z ( X r ⋆ , Z ) , ⟨ h i ⋆ , λ ⟩ : = ℓ i − 1 ⟨ h i , λ ⟩ , L r ⋆ : = r Z α 0 ⊕ 2 r Z α 1 . (130) Y_r^\star:=\operatorname{Hom}_{\mathbb{Z}}(X_r^\star,\mathbb{Z}),\qquad\langle h_i^\star,\lambda\rangle:=\ell_i^{-1}\langle h_i,\lambda\rangle,\qquad L^\star_r:=r\mathbb{Z}\alpha_0\oplus2r\mathbb{Z}\alpha_1. \tag*{(130)} Y r ⋆ := Hom Z ( X r ⋆ , Z ) , ⟨ h i ⋆ , λ ⟩ := ℓ i − 1 ⟨ h i , λ ⟩ , L r ⋆ := r Z α 0 ⊕ 2 r Z α 1 . (130)
If i ⋅ j = d i a i j = d j a j i i\cdot j=d_i a_{ij}=d_j a_{ji} i ⋅ j = d i a ij = d j a j i is the source symmetric form, put
i ⋅ ⋆ j : = ℓ i ℓ j ( i ⋅ j ) , d i ⋆ : = d i ℓ i 2 . (131) i\cdot_\star j:=\ell_i\ell_j(i\cdot j),\qquad d_i^\star:=d_i\ell_i^2. \tag*{(131)} i ⋅ ⋆ j := ℓ i ℓ j ( i ⋅ j ) , d i ⋆ := d i ℓ i 2 . (131)
The coefficients d i ⋆ d_i^\star d i ⋆ are the integral symmetrizers of the starred Cartan datum.
Lemma 9.5. With the data of (129) –(131) ,
( X r ⋆ , Y r ⋆ , { 0 , 1 } , ⟨ , ⟩ , ( α i ⋆ ) i = 0 , 1 , ( h i ⋆ ) i = 0 , 1 ) \left(X_r^\star,Y_r^\star,\{0,1\},\langle\ ,\ \rangle,(\alpha_i^\star)_{i=0,1},(h_i^\star)_{i=0,1}\right) ( X r ⋆ , Y r ⋆ , { 0 , 1 } , ⟨ , ⟩ , ( α i ⋆ ) i = 0 , 1 , ( h i ⋆ ) i = 0 , 1 ) is a regular root datum,
and
( a i j ⋆ ) i , j = 0 , 1 = ( 2 − 2 − 2 2 ) , ( d 0 ⋆ , d 1 ⋆ ) = ( 4 r 2 , 4 r 2 ) , ⨁ i = 0 1 Z α i ⋆ = L r ⋆ ⊆ X r ⋆ . (a_{ij}^\star)_{i,j=0,1}=\begin{pmatrix}2&-2\\-2&2\end{pmatrix},\qquad(d_0^\star,d_1^\star)=(4r^2,4r^2),\qquad\bigoplus_{i=0}^{1}\mathbb{Z}\alpha_i^\star=L_r^\star\subseteq X_r^\star. ( a ij ⋆ ) i , j = 0 , 1 = ( 2 − 2 − 2 2 ) , ( d 0 ⋆ , d 1 ⋆ ) = ( 4 r 2 , 4 r 2 ) , i = 0 ⨁ 1 Z α i ⋆ = L r ⋆ ⊆ X r ⋆ .
Proof. Since r r r is odd,
ℓ i = min { m ∈ Z ≥ 1 ∣ m d i ∈ 2 r Z } , ( ℓ 0 , ℓ 1 ) = ( r , 2 r ) , \ell_i=\min\{m\in\mathbb{Z}_{\geq1}\mid m d_i\in2r\mathbb{Z}\},\qquad(\ell_0,\ell_1)=(r,2r), ℓ i = min { m ∈ Z ≥ 1 ∣ m d i ∈ 2 r Z } , ( ℓ 0 , ℓ 1 ) = ( r , 2 r ) ,
Hence (129) –(131) are Lusztig’s 2 r 2r 2 r -modified root datum [21 §§2.2.4–2.2.5] , and regularity follows from that of the source datum. Directly,
( a i j ⋆ ) i , j = 0 , 1 = ( ℓ i − 1 ℓ j a i j ) i , j = 0 , 1 = ( 2 − 2 − 2 2 ) . (a_{ij}^\star)_{i,j=0,1}=\left(\ell_i^{-1}\ell_j a_{ij}\right)_{i,j=0,1}=\begin{pmatrix}2&-2\\-2&2\end{pmatrix}. ( a ij ⋆ ) i , j = 0 , 1 = ( ℓ i − 1 ℓ j a ij ) i , j = 0 , 1 = ( 2 − 2 − 2 2 ) .
( d 0 ⋆ , d 1 ⋆ ) = ( 4 r 2 , 4 r 2 ) , ⨁ i = 0 1 Z α i ⋆ = r Z α 0 ⊕ 2 r Z α 1 = L r ⋆ . (d_0^\star,d_1^\star)=(4r^2,4r^2),\qquad\bigoplus_{i=0}^{1}\mathbb{Z}\alpha_i^\star=r\mathbb{Z}\alpha_0\oplus2r\mathbb{Z}\alpha_1=L_r^\star. ( d 0 ⋆ , d 1 ⋆ ) = ( 4 r 2 , 4 r 2 ) , i = 0 ⨁ 1 Z α i ⋆ = r Z α 0 ⊕ 2 r Z α 1 = L r ⋆ .
□
Define
ϵ r ⋆ : L r ⋆ ⟶ Z / 2 Z , ϵ r ⋆ ( b 0 α 0 ⋆ + b 1 α 1 ⋆ ) = b 0 ‾ . (132) \epsilon_r^\star:L_r^\star\longrightarrow\mathbb{Z}/2\mathbb{Z},\qquad\epsilon_r^\star(b_0\alpha_0^\star+b_1\alpha_1^\star)=\overline{b_0}. \tag*{(132)} ϵ r ⋆ : L r ⋆ ⟶ Z /2 Z , ϵ r ⋆ ( b 0 α 0 ⋆ + b 1 α 1 ⋆ ) = b 0 . (132)
Let U ˙ A , r ⋆ \dot{U}_{A,r}^\star U ˙ A , r ⋆ be the modified integral form for the starred datum, and put
U ˙ r ⋆ : = k r ( 4 ) ⊗ A , q ↦ ξ r U ˙ A , r ⋆ . (133) \dot{U}_r^\star:=\mathbf{k}_r^{(4)}\otimes_{A,q\mapsto\xi_r}\dot{U}_{A,r}^\star. \tag*{(133)} U ˙ r ⋆ := k r ( 4 ) ⊗ A , q ↦ ξ r U ˙ A , r ⋆ . (133)
For e ˉ ∈ Z / 2 Z \bar e\in\mathbb{Z}/2\mathbb{Z} e ˉ ∈ Z /2 Z , put
G e ˉ ( r ) : = ⨁ λ , μ ∈ X ⋆ μ − λ ∈ L r ⋆ , ϵ r ⋆ ( μ − λ ) = e ˉ 1 μ U ˙ r ⋆ 1 λ . (134) \mathcal{G}_{\bar e}^{(r)}:=\bigoplus_{\substack{\lambda,\mu\in X^\star\\\mu-\lambda\in L_r^\star,\ \epsilon_r^\star(\mu-\lambda)=\bar e}}1_\mu\dot{U}_r^\star1_\lambda. \tag*{(134)} G e ˉ ( r ) := λ , μ ∈ X ⋆ μ − λ ∈ L r ⋆ , ϵ r ⋆ ( μ − λ ) = e ˉ ⨁ 1 μ U ˙ r ⋆ 1 λ . (134)
Then
U ˙ r ⋆ = G 0 ˉ ( r ) ⊕ G 1 ˉ ( r ) , G e ˉ ( r ) G f ˉ ( r ) ⊆ G e + f ‾ ( r ) . (135) \dot{U}_r^\star=\mathcal{G}_{\bar0}^{(r)}\oplus\mathcal{G}_{\bar1}^{(r)},\qquad\mathcal{G}_{\bar e}^{(r)}\mathcal{G}_{\bar f}^{(r)}\subseteq\mathcal{G}_{\overline{e+f}}^{(r)}. \tag*{(135)} U ˙ r ⋆ = G 0 ˉ ( r ) ⊕ G 1 ˉ ( r ) , G e ˉ ( r ) G f ˉ ( r ) ⊆ G e + f ( r ) . (135)
Application to the rank-one Φ 4 r ( q ) \Phi_{4r}(q) Φ 4 r ( q ) specialization.
∀ r ∈ ( 2 Z + 1 ) ∩ Z ≥ 3 , U ˙ ξ r ( 1 ) : = k r ( 4 ) ⊗ A , q ↦ ξ r U ˙ A L u s , ( 1 ) , \forall r\in(2\mathbb{Z}+1)\cap\mathbb{Z}_{\geq3},\qquad\dot{U}_{\xi_r}^{(1)}:=\mathbf{k}_r^{(4)}\otimes_{A,q\mapsto\xi_r}\dot{U}_{A}^{\mathrm{Lus},(1)}, ∀ r ∈ ( 2 Z + 1 ) ∩ Z ≥ 3 , U ˙ ξ r ( 1 ) := k r ( 4 ) ⊗ A , q ↦ ξ r U ˙ A Lus , ( 1 ) ,
∃ F r ( 4 ) ∈ Hom k r ( 4 ) -alg ( U ˙ ξ r ( 1 ) , U ˙ r ⋆ ) : (136) \exists F_r^{(4)}\in\operatorname{Hom}_{\mathbf{k}_r^{(4)}\text{-alg}}\left(\dot{U}_{\xi_r}^{(1)},\dot{U}_r^\star\right): \tag*{(136)} ∃ F r ( 4 ) ∈ Hom k r ( 4 ) -alg ( U ˙ ξ r ( 1 ) , U ˙ r ⋆ ) : (136)
∀ λ ∈ X , F r ( 4 ) ( 1 λ ) = { 1 λ ⋆ , λ ∈ X r ⋆ , 0 , λ ∉ X r ⋆ , \forall\lambda\in X,\qquad F_r^{(4)}(1_\lambda)=
\begin{cases}
1_\lambda^\star,&\lambda\in X_r^\star,\\
0,&\lambda\notin X_r^\star,
\end{cases} ∀ λ ∈ X , F r ( 4 ) ( 1 λ ) = { 1 λ ⋆ , 0 , λ ∈ X r ⋆ , λ ∈ / X r ⋆ ,
F r ( 4 ) ( E 0 ( r ) 1 0 ) = E 0 ⋆ 1 0 ⋆ ≠ 0. F_r^{(4)}(E_0^{(r)}1_0)=E_0^\star1_0^\star\ne0. F r ( 4 ) ( E 0 ( r ) 1 0 ) = E 0 ⋆ 1 0 ⋆ = 0.
Proof. Here
N = ord ( ξ r 2 ) = 2 r , 2 N = 4 r N=\operatorname{ord}(\xi_r^2)=2r,\qquad2N=4r N = ord ( ξ r 2 ) = 2 r , 2 N = 4 r
Similar to the proof above, k r ( 4 ) \mathbb{k}_r^{(4)} k r ( 4 ) has characteristic zero. Thus the field hypothesis in the criterion holds. Moreover, N N N is even and
A ⟶ k r ( 4 ) \mathcal{A}\longrightarrow\mathbb{k}_r^{(4)} A ⟶ k r ( 4 )
factors through A N ′ = A / ( Φ 4 r ( q ) ) \mathcal{A}'_N=\mathcal{A}/(\Phi_{4r}(q)) A N ′ = A / ( Φ 4 r ( q )) by (127) . The rank-one graph has no cycle. Since a 01 = − 1 a_{01}=-1 a 01 = − 1 and a 10 = − 4 a_{10}=-4 a 10 = − 4 , the two ordered size conditions are
ℓ 0 = r ≥ 3 ≥ 2 = 1 − a 01 , ℓ 1 = 2 r ≥ 6 ≥ 5 = 1 − a 10 . \ell_0=r\geq3\geq2=1-a_{01},\qquad\ell_1=2r\geq6\geq5=1-a_{10}. ℓ 0 = r ≥ 3 ≥ 2 = 1 − a 01 , ℓ 1 = 2 r ≥ 6 ≥ 5 = 1 − a 10 .
Thus Lusztig’s modified Frobenius criterion gives (136) and the displayed idempotent formula. Since
0 ∈ X r ⋆ , ℓ 0 = r , r ∣ r , 0\in X_r^\star,\qquad\ell_0=r,\qquad r\mid r, 0 ∈ X r ⋆ , ℓ 0 = r , r ∣ r ,
the positive simple-block formula gives
Fr r ( 4 ) ( E 0 ( r ) 1 0 ) = E 0 ⋆ , ( r / ℓ 0 ) 1 0 ⋆ = E 0 ⋆ 1 0 ⋆ . \operatorname{Fr}_r^{(4)}(E_0^{(r)}1_0)=E_0^{\star,(r/\ell_0)}1_0^\star=E_0^\star1_0^\star. Fr r ( 4 ) ( E 0 ( r ) 1 0 ) = E 0 ⋆ , ( r / ℓ 0 ) 1 0 ⋆ = E 0 ⋆ 1 0 ⋆ .
Its nonvanishing follows from the target modified PBW basis [21 §§23.2.1–23.2.2] .
Proposition 9.6. For every odd integer r ≥ 3 r\geq3 r ≥ 3 , one has
c 1 , r ⊆ ( Φ 4 r ( q ) ) . (137) \mathfrak{c}_{1,r}\subseteq(\Phi_{4r}(q)). \tag*{(137)} c 1 , r ⊆ ( Φ 4 r ( q )) . (137)
Proof. Fix
r ∈ ( 2 Z + 1 ) ∩ Z ≥ 3 , ( λ , u ) ∈ X × C 1 , ν : = deg u . r\in(2\mathbb{Z}+1)\cap\mathbb{Z}_{\geq3},\qquad(\lambda,u)\in X\times\mathcal{C}_1,\qquad\nu:=\deg u. r ∈ ( 2 Z + 1 ) ∩ Z ≥ 3 , ( λ , u ) ∈ X × C 1 , ν := deg u .
The idempotent formula in (136) verifies (106) . Suppose
Fr r ( 4 ) ( u 1 λ ) ≠ 0. \operatorname{Fr}_r^{(4)}(u1_\lambda)\ne0. Fr r ( 4 ) ( u 1 λ ) = 0.
If
u = ( x 1 , k ε ) ( s ) , ν = s k α 0 + s ( 2 k + ε ) α 1 , u=(x_{1,k}^{\varepsilon})^{(s)},\qquad\nu=sk\alpha_0+s(2k+\varepsilon)\alpha_1, u = ( x 1 , k ε ) ( s ) , ν = s k α 0 + s ( 2 k + ε ) α 1 ,
then lemma 8.2 gives
ν ∈ L r ⋆ , r ∣ s k , 2 r ∣ s ( 2 k + ε ) . \nu\in L_r^\star,\qquad r\mid sk,\qquad2r\mid s(2k+\varepsilon). ν ∈ L r ⋆ , r ∣ s k , 2 r ∣ s ( 2 k + ε ) .
Since
gcd ( k , 2 k + ε ) = 1 , \gcd(k,2k+\varepsilon)=1, g cd( k , 2 k + ε ) = 1 ,
one has r ∣ s r\mid s r ∣ s . Writing s = r t s=rt s = r t gives
2 ∣ t ( 2 k + ε ) ⟹ 2 ∣ t ⟹ ε r ⋆ ( ν ) = t k ‾ = 0 ‾ . 2\mid t(2k+\varepsilon)\Longrightarrow2\mid t\Longrightarrow\varepsilon_r^\star(\nu)=\overline{tk}=\overline{0}. 2 ∣ t ( 2 k + ε ) ⟹ 2 ∣ t ⟹ ε r ⋆ ( ν ) = t k = 0 .
Thus every nonzero real-current block image belongs to G 0 ( r ) \mathcal{G}_0^{(r)} G 0 ( r ) . For a toral generator,
ν = 0 , Fr r ( 4 ) ( u 1 λ ) ∈ G 0 ( r ) . \nu=0,\qquad\operatorname{Fr}_r^{(4)}(u1_\lambda)\in\mathcal{G}_0^{(r)}. ν = 0 , Fr r ( 4 ) ( u 1 λ ) ∈ G 0 ( r ) .
Hence lemma 8.3 gives
Fr r ( 4 ) ( U A D r , ( 1 ) ‾ A 1 0 ) ⊆ G 0 ( r ) . (138) \operatorname{Fr}_r^{(4)}\left(U_A^{\overline{\mathrm{Dr},(1)}_A}1_0\right)\subseteq\mathcal{G}_0^{(r)}. \tag*{(138)} Fr r ( 4 ) ( U A Dr , ( 1 ) A 1 0 ) ⊆ G 0 ( r ) . (138)
Moreover,
r α 0 = α 0 ⋆ ∈ L r ⋆ , ε r ⋆ ( r α 0 ) = 1 ‾ , r\alpha_0=\alpha_0^\star\in L_r^\star,\qquad\varepsilon_r^\star(r\alpha_0)=\overline{1}, r α 0 = α 0 ⋆ ∈ L r ⋆ , ε r ⋆ ( r α 0 ) = 1 ,
and
0 ≠ Fr r ( 4 ) ( E ‾ 0 ( r ) 1 0 ) = E 0 ⋆ 1 1 0 ⋆ ∈ G 1 ( r ) . 0 \ne\operatorname{Fr}^{(4)}_r\left(\overline{E}^{(r)}_0 1_0\right)=E^{\star1}_{0}1^{\star}_{0}\in\mathcal{G}^{(r)}_1. 0 = Fr r ( 4 ) ( E 0 ( r ) 1 0 ) = E 0 ⋆ 1 1 0 ⋆ ∈ G 1 ( r ) .
Thus lemma 8.4, applied to
χ r ( 4 ) : = Fr r ( 4 ) ∣ k r ( 4 ) ⊗ A ( U A L u s , ( 1 ) ) 1 0 , \chi^{(4)}_r:=\left.\operatorname{Fr}^{(4)}_r\right|_{\mathfrak{k}^{(4)}_r\otimes_A(U^{\mathrm{Lus},(1)}_A)_{1_0}}, χ r ( 4 ) := Fr r ( 4 ) k r ( 4 ) ⊗ A ( U A Lus , ( 1 ) ) 1 0 ,
gives
c 1 , r ⊆ ( Φ 4 r ( q ) ) , \mathfrak{c}_{1,r}\subseteq(\Phi_{4r}(q)), c 1 , r ⊆ ( Φ 4 r ( q )) ,
which is (137) .
The restriction to odd r r r is essential for the displayed starred datum. If r r r were even, then
ℓ 0 = ord ( ξ r 8 ) = 4 r gcd ( 4 r , 8 ) = r 2 , \ell_0=\operatorname{ord}(\xi^8_r)=\frac{4r}{\gcd(4r,8)}=\frac{r}{2}, ℓ 0 = ord ( ξ r 8 ) = g cd( 4 r , 8 ) 4 r = 2 r ,
so the scaled affine root would be α 0 ⋆ = ( r / 2 ) α 0 \alpha^\star_0=(r/2)\alpha_0 α 0 ⋆ = ( r /2 ) α 0 and, when the Frobenius criterion applies,
E 0 ( r ) 1 0 ⟼ E 0 ⋆ , ( r / ℓ 0 ) 1 0 ⋆ = E 0 ⋆ , ( 2 ) 1 0 ⋆ . E^{(r)}_0 1_0\longmapsto E^{\star,(r/\ell_0)}_0 1^\star_0=E^{\star,(2)}_0 1^\star_0. E 0 ( r ) 1 0 ⟼ E 0 ⋆ , ( r / ℓ 0 ) 1 0 ⋆ = E 0 ⋆ , ( 2 ) 1 0 ⋆ .
Its α 0 ⋆ \alpha^\star_0 α 0 ⋆ -coordinate is even, rather than the required odd coordinate. For an even index 2 s ≥ 4 2s\geq4 2 s ≥ 4 ,
Φ 4 ( 2 s ) ( q ) = Φ 8 s ( q ) , \Phi_{4(2s)}(q)=\Phi_{8s}(q), Φ 4 ( 2 s ) ( q ) = Φ 8 s ( q ) ,
and s ≥ 2 s\geq2 s ≥ 2 , so the corresponding prime is already covered by theorem 9.4. The remaining Φ 4 \Phi_4 Φ 4 and Φ 8 \Phi_8 Φ 8 endpoints, for which the affine scaling integer ℓ 0 \ell_0 ℓ 0 equals 1 1 1 , are treated in § 6.5 .
The scaled root datum and cyclotomic obstructions for r = 1 r=1 r = 1
Throughout Sections 10–11 and Appendix § 6.8 , we use N = Z ≥ 0 \mathbb{N}=\mathbb{Z}_{\geq0} N = Z ≥ 0 . At the two remaining primes, the scaled affine integer is ℓ 0 = 1 \ell_0=1 ℓ 0 = 1 , so the size condition (121) fails and the Frobenius criterion of § 6.3 does not apply. Using the root data of that section at r = 1 r=1 r = 1 , we replace the Frobenius target U ˙ k ∙ \dot{U}^{\bullet}_\mathsf{k} U ˙ k ∙ by a common triangular quotient of the specialized source form. Its technical construction is given in Appendix § 6.8 . Subsection 10.3 then applies Lemmas 8.3 and 8.4 to prove
c n , 1 ⊆ ( Φ 8 ( q ) ) ( n ≥ 1 ) , c 1 , 1 ⊆ ( Φ 4 ( q ) ) . \mathfrak{c}_{n,1}\subseteq(\Phi_8(q))\quad(n\geq1),\qquad\mathfrak{c}_{1,1}\subseteq(\Phi_4(q)). c n , 1 ⊆ ( Φ 8 ( q )) ( n ≥ 1 ) , c 1 , 1 ⊆ ( Φ 4 ( q )) .
The two endpoint instances
Convention 10.1 (Endpoint data). Fix n ∈ Z ≥ 1 n\in\mathbb{Z}_{\geq1} n ∈ Z ≥ 1 and retain the regular A 2 n ( 2 ) A^{(2)}_{2n} A 2 n ( 2 ) root datum ( X , Y , I , ⟨ , ⟩ ) (X,Y,I,\langle\ ,\ \rangle) ( X , Y , I , ⟨ , ⟩) of (66) as the source datum, with symmetrizers ( d i ) i ∈ I (d_i)_{i\in I} ( d i ) i ∈ I and q i = q d i q_i=q^{d_i} q i = q d i . Assume given a field
k , q ‾ ∈ k × , char k = 0 , N : = ord ( q ‾ 2 ) ∈ 2 Z > 0 , (139) \mathbb{k},\qquad\overline{q}\in\mathbb{k}^{\times},\qquad\operatorname{char}\mathbb{k}=0,\qquad N:=\operatorname{ord}(\overline{q}^{\,2})\in2\mathbb{Z}_{>0}, \tag*{(139)} k , q ∈ k × , char k = 0 , N := ord ( q 2 ) ∈ 2 Z > 0 , (139)
Here q ‾ \overline{q} q denotes the specialized value of q q q in k \mathbb{k} k . Let ( ℓ i ) i ∈ I (\ell_i)_{i\in I} ( ℓ i ) i ∈ I , X ∙ X^\bullet X ∙ , α i ∙ \alpha^\bullet_i α i ∙ , h i ∙ h^\bullet_i h i ∙ , L ∙ L^\bullet L ∙ be the associated scaled data of (111) –(113) and (116) , that is
ℓ i = ord ( q ‾ 2 d i ) = N gcd ( N , d i ) , α i ∙ = ℓ i α i , L ∙ = ⨁ i ∈ I Z α i ∙ . (140) \ell_i=\operatorname{ord}(\overline{q}^{\,2d_i})=\frac{N}{\gcd(N,d_i)},\qquad\alpha^\bullet_i=\ell_i\alpha_i,\qquad L^\bullet=\bigoplus_{i\in I}\mathbb{Z}\alpha^\bullet_i. \tag*{(140)} ℓ i = ord ( q 2 d i ) = g cd( N , d i ) N , α i ∙ = ℓ i α i , L ∙ = i ∈ I ⨁ Z α i ∙ . (140)
X ∙ = { λ ∈ X ∣ ⟨ h i , λ ⟩ ∈ ℓ i Z for every i ∈ I } . (141) X^\bullet=\{\lambda\in X\mid\langle h_i,\lambda\rangle\in\ell_i\mathbb{Z}\text{ for every }i\in I\}. \tag*{(141)} X ∙ = { λ ∈ X ∣ ⟨ h i , λ ⟩ ∈ ℓ i Z for every i ∈ I } . (141)
We assume throughout § 6.5 and Appendix § 6.8 that
( X ∙ , Y ∙ , I , ⟨ , ⟩ , ( α i ∙ ) i ∈ I , ( h i ∙ ) i ∈ I ) is a regular root datum , L ∙ ⊆ X ∙ ; (142) (X^\bullet,Y^\bullet,I,\langle\ ,\ \rangle,(\alpha_i^\bullet)_{i\in I},(h_i^\bullet)_{i\in I})\text{ is a regular root datum},\qquad L^\bullet\subseteq X^\bullet; \tag*{(142)} ( X ∙ , Y ∙ , I , ⟨ , ⟩ , ( α i ∙ ) i ∈ I , ( h i ∙ ) i ∈ I ) is a regular root datum , L ∙ ⊆ X ∙ ; (142)
ℓ 0 = 1. (143) \ell_0=1. \tag*{(143)} ℓ 0 = 1. (143)
Write q k , i : = q ˉ d i q_{k,i}:=\bar{q}^{d_i} q k , i := q ˉ d i for the specialized local parameter, so that ℓ i = ord ( q ˉ k , i d i ) \ell_i=\operatorname{ord}(\bar{q}^{d_i}_{k,i}) ℓ i = ord ( q ˉ k , i d i ) .
Lemma 10.2 (The two endpoint instances). Convention 10.1 is satisfied in the following two ways, both obtained from § 6.3 at r = 1 r=1 r = 1 .
(a) The Φ 8 ( q ) \Phi_8(q) Φ 8 ( q ) endpoint. Let n ∈ Z ≥ 1 n\in\mathbb{Z}_{\geq1} n ∈ Z ≥ 1 and put ( k , q ˉ ) : = ( k 1 ( 8 ) , ζ 1 ) (\mathbb{k},\bar{q}):=(\mathbb{k}_1^{(8)},\zeta_1) ( k , q ˉ ) := ( k 1 ( 8 ) , ζ 1 ) as in (108) with r = 1 r=1 r = 1 . Then N = 4 N=4 N = 4 ,
( ℓ 0 , ℓ 1 , … , ℓ n ) = ( 1 , 2 , … , 2 , 4 ) , L ∙ = L 1 ∙ = Z α 0 ⊕ ⨁ i = 1 n − 1 2 Z α i ⊕ 4 Z α n , (\ell_0,\ell_1,\ldots,\ell_n)=(1,2,\ldots,2,4),\qquad L^\bullet=L_1^\bullet=\mathbb{Z}\alpha_0\oplus\bigoplus_{i=1}^{n-1}2\mathbb{Z}\alpha_i\oplus4\mathbb{Z}\alpha_n, ( ℓ 0 , ℓ 1 , … , ℓ n ) = ( 1 , 2 , … , 2 , 4 ) , L ∙ = L 1 ∙ = Z α 0 ⊕ i = 1 ⨁ n − 1 2 Z α i ⊕ 4 Z α n , the middle sum being empty for n = 1 n=1 n = 1 .
(b) The Φ 4 ( q ) \Phi_4(q) Φ 4 ( q ) endpoint. Let n = 1 n=1 n = 1 and put ( k , q ˉ ) : = ( k 1 ( 4 ) , ξ 1 ) (\mathbb{k},\bar{q}):=(\mathbb{k}_1^{(4)},\xi_1) ( k , q ˉ ) := ( k 1 ( 4 ) , ξ 1 ) as in (127) with r = 1 r=1 r = 1 . Then N = 2 N=2 N = 2 ,
( ℓ 0 , ℓ 1 ) = ( 1 , 2 ) , X ∙ = X 1 ⋆ , L ∙ = L 1 ⋆ = Z α 0 ⊕ 2 Z α 1 . (\ell_0,\ell_1)=(1,2),\qquad X^\bullet=X_1^\star,\qquad L^\bullet=L_1^\star=\mathbb{Z}\alpha_0\oplus2\mathbb{Z}\alpha_1. ( ℓ 0 , ℓ 1 ) = ( 1 , 2 ) , X ∙ = X 1 ⋆ , L ∙ = L 1 ⋆ = Z α 0 ⊕ 2 Z α 1 .
Proof. We verify the two instances separately.
(a) In § 6.3.1 , (108) gives ord ( ζ 1 ) = 8 \operatorname{ord}(\zeta_1)=8 ord ( ζ 1 ) = 8 , hence N = 4 N=4 N = 4 ; (110) gives ( ℓ 0 , … , ℓ n ) = ( 1 , 2 , … , 2 , 4 ) (\ell_0,\ldots,\ell_n)=(1,2,\ldots,2,4) ( ℓ 0 , … , ℓ n ) = ( 1 , 2 , … , 2 , 4 ) ; and lemma 9.1, which is stated for every r ≥ 1 r\geq1 r ≥ 1 , gives (142) together with L ∙ = L 1 ∙ L^\bullet=L_1^\bullet L ∙ = L 1 ∙ .
(b) Here k = Frac ( A / ( Φ 4 ( q ) ) ) \mathbb{k}=\operatorname{Frac}(\mathcal{A}/(\Phi_4(q))) k = Frac ( A / ( Φ 4 ( q ))) has characteristic zero, and q ˉ = ξ 1 \bar{q}=\xi_1 q ˉ = ξ 1 has order 4 4 4 . Thus
N = ord ( q ˉ − 2 ) = 2 , ( ℓ 0 , ℓ 1 ) = ( 2 gcd ( 2 , 4 ) , 2 gcd ( 2 , 1 ) ) = ( 1 , 2 ) . N=\operatorname{ord}(\bar{q}^{-2})=2,\qquad(\ell_0,\ell_1)=\left(\frac{2}{\gcd(2,4)},\frac{2}{\gcd(2,1)}\right)=(1,2). N = ord ( q ˉ − 2 ) = 2 , ( ℓ 0 , ℓ 1 ) = ( g cd( 2 , 4 ) 2 , g cd( 2 , 1 ) 2 ) = ( 1 , 2 ) .
Evaluating the defining formulas of section 9.2 at r = 1 r=1 r = 1 , we obtain
X ∙ = X 1 ⋆ = { λ ∈ X ∣ ⟨ h 1 , λ ⟩ ∈ 2 Z } , L ∙ = L 1 ⋆ = Z α 0 ⊕ 2 Z α 1 . X^\bullet=X_1^\star=\{\lambda\in X\mid\langle h_1,\lambda\rangle\in2\mathbb{Z}\},\qquad L^\bullet=L_1^\star=\mathbb{Z}\alpha_0\oplus2\mathbb{Z}\alpha_1. X ∙ = X 1 ⋆ = { λ ∈ X ∣ ⟨ h 1 , λ ⟩ ∈ 2 Z } , L ∙ = L 1 ⋆ = Z α 0 ⊕ 2 Z α 1 .
We verify regularity directly. Since ⟨ h 1 , α 0 ⟩ = − 4 \langle h_1,\alpha_0\rangle=-4 ⟨ h 1 , α 0 ⟩ = − 4 and ⟨ h 1 , 2 α 1 ⟩ = 4 \langle h_1,2\alpha_1\rangle=4 ⟨ h 1 , 2 α 1 ⟩ = 4 , both scaled simple roots α 0 ∙ = α 0 \alpha_0^\bullet=\alpha_0 α 0 ∙ = α 0 and α 1 ∙ = 2 α 1 \alpha_1^\bullet=2\alpha_1 α 1 ∙ = 2 α 1 belong to X ∙ X^\bullet X ∙ . The scaled coroots
h 0 ∙ = h 0 ∣ X ∙ , h 1 ∙ = 1 2 h 1 ∣ X ∙ h_0^\bullet=h_0|_{X^\bullet},\qquad h_1^\bullet=\tfrac{1}{2}h_1|_{X^\bullet} h 0 ∙ = h 0 ∣ X ∙ , h 1 ∙ = 2 1 h 1 ∣ X ∙
belong to Y ∙ = Hom Z ( X ∙ , Z ) Y^\bullet=\operatorname{Hom}_{\mathbb{Z}}(X^\bullet,\mathbb{Z}) Y ∙ = Hom Z ( X ∙ , Z ) , and
( a i j ∙ ) i , j = 0 , 1 = ( 2 − 2 − 2 2 ) , ( d 0 ∙ , d 1 ∙ ) = ( 4 , 4 ) . (a_{ij}^\bullet)_{i,j=0,1}=\begin{pmatrix}2&-2\\-2&2\end{pmatrix},\qquad(d_0^\bullet,d_1^\bullet)=(4,4). ( a ij ∙ ) i , j = 0 , 1 = ( 2 − 2 − 2 2 ) , ( d 0 ∙ , d 1 ∙ ) = ( 4 , 4 ) .
Since 2 X ⊆ X ∙ ⊆ X 2X\subseteq X^\bullet\subseteq X 2 X ⊆ X ∙ ⊆ X , the lattice X ∙ X^\bullet X ∙ has finite index in X X X . The scaled simple roots and coroots therefore remain linearly independent, by regularity of the source datum. This proves (142) .
In both cases ℓ 0 = 1 \ell_0=1 ℓ 0 = 1 .
The endpoint quotient Q
We retain here the quotient and the properties needed for the parity argument. The detailed word calculus, Gaussian vanishing, and triangular straightening are collected in Appendix § 6.8 .
Let
V ˙ : = k ⊗ A , q ↦ q ˉ U ˙ A L u s , ( n ) . \dot{V}:=\mathbb{k}\otimes_{A,q\mapsto\bar{q}}\dot{U}_{A}^{\mathrm{Lus},(n)}. V ˙ := k ⊗ A , q ↦ q ˉ U ˙ A Lus , ( n ) .
For a labelled divided-power word w = ( u 1 , … , u m ; η ) \mathbf{w}=(u_1,\ldots,u_m;\eta) w = ( u 1 , … , u m ; η ) , where η ∈ X \eta\in X η ∈ X and u j = E i j ( a j ) u_j=E_{i_j}^{(a_j)} u j = E i j ( a j ) or F i j ( a j ) F_{i_j}^{(a_j)} F i j ( a j ) with a j ≥ 1 a_j\geq1 a j ≥ 1 , set
λ j ( w ) : = η + ∑ v = j + 1 m deg u v ( 0 ≤ j ≤ m ) , ev ( w ) : = u 1 ⋯ u m 1 η . \lambda_j(\mathbf{w}):=\eta+\sum_{v=j+1}^{m}\deg u_v\quad(0\leq j\leq m),\qquad\operatorname{ev}(\mathbf{w}):=u_1\cdots u_m1_\eta. λ j ( w ) := η + v = j + 1 ∑ m deg u v ( 0 ≤ j ≤ m ) , ev ( w ) := u 1 ⋯ u m 1 η .
For m = 0 m=0 m = 0 , the evaluation is 1 η 1_\eta 1 η .
The definition below encodes the two selection rules suggested by quantum Frobenius in Lemma 9.3 : only weights in X ∙ X^\bullet X ∙ and divided-power exponents divisible by the corresponding ℓ i \ell_i ℓ i are allowed to survive. The intermediate weights λ j ( w ) \lambda_j(\mathbf{w}) λ j ( w ) record the successive weight idempotents through which the word passes. We exclude any word that passes through an idempotent outside X ∙ X^\bullet X ∙ , or contains a divided-power letter whose exponent is not a multiple of its scaling integer. At the endpoints the Frobenius size condition fails, so we implement these selection rules directly in a quotient of V ˙ \dot{V} V ˙ .
Call w \mathbf{w} w bad if
( ∃ j ∈ { 0 , … , m } : λ j ( w ) ∉ X ∙ ) or ( ∃ j ∈ { 1 , … , m } : ℓ i j ∤ a j ) , \left(\exists j\in\{0,\ldots,m\}:\lambda_j(\mathbf{w})\notin X^\bullet\right)\quad\text{or}\quad\left(\exists j\in\{1,\ldots,m\}:\ell_{i_j}\nmid a_j\right), ( ∃ j ∈ { 0 , … , m } : λ j ( w ) ∈ / X ∙ ) or ( ∃ j ∈ { 1 , … , m } : ℓ i j ∤ a j ) ,
and define
J : = span k { ev ( w ) ∣ w is bad } . \mathcal{J}:=\operatorname{span}_{\mathbb{k}}\{\operatorname{ev}(\mathbf{w})\mid\mathbf{w}\text{ is bad}\}. J := span k { ev ( w ) ∣ w is bad } .
The formal definitions of these words and their evaluations are given in Appendix § 6.8 .§ 3.1 . To obtain the required obstruction, we must show that J \mathcal{J} J is a two-sided ideal and that E 010 E_{010} E 010 survives in the quotient. The following theorem establishes these facts; its proof uses the Gaussian and triangular arguments in the appendix.
Theorem 10.3. Under convention 10.1,
V ˙ J ⊆ J , J V ˙ ⊆ J . \dot{V}\mathcal{J}\subseteq\mathcal{J},\qquad\mathcal{J}\dot{V}\subseteq\mathcal{J}. V ˙ J ⊆ J , J V ˙ ⊆ J . Hence Q : = V ˙ / J Q:=\dot{V}/\mathcal{J} Q := V ˙ / J is a root-graded k \mathbb{k} k -algebra, and
{ ν ∣ ( Q ) ν ≠ 0 } ⊆ L ∙ , E 010 + J ∈ ( Q ) α 0 ∖ { 0 } . (144) \{\nu\mid(Q)_\nu\ne0\}\subseteq L^\bullet,\qquad E_{010}+\mathcal{J}\in(Q)_{\alpha_0}\setminus\{0\}. \tag*{(144)} { ν ∣ ( Q ) ν = 0 } ⊆ L ∙ , E 010 + J ∈ ( Q ) α 0 ∖ { 0 } . (144) Proof. A product of a divided-power word and a bad word, in either order, is zero or the evaluation of a bad word. Hence J \mathcal{J} J is a two-sided ideal. Since the evaluated bad words are root homogeneous, J \mathcal{J} J is root graded. The straightening arguments in Appendix § 6.8 .3 give its triangular decomposition. This decomposition yields the root support and the nonvanishing in (144) ; the full proof is given in Appendix § 6.8 .§ 3.3 .
The root support in (144) allows the parity on L ∙ L^\bullet L ∙ to grade Q Q Q . Subsection 10.3 verifies that every surviving current-generator block is even, whereas E 010 + J E_{010}+\mathcal{J} E 010 + J is nonzero and odd. Lemma 8.3 then makes the whole current core even, and Lemma 8.4 gives the endpoint coefficient-ideal obstructions.
Parity separation and coefficient ideals
Retain convention 10.1 and the notation of § 6.5.2 . Exactly as in (117) and (132) , define
ϵ : L ∙ ⟶ Z / 2 Z , ϵ ( ∑ i ∈ I b i α i ∨ ) : = b 0 ‾ , ϵ ∈ Hom Z ( L ∙ , Z / 2 Z ) . (145) \epsilon: L^\bullet\longrightarrow\mathbb{Z}/2\mathbb{Z}, \qquad\epsilon\left(\sum_{i\in I} b_i\alpha_i^\vee\right) := \overline{b_0}, \qquad\epsilon\in\operatorname{Hom}_{\mathbb{Z}}(L^\bullet,\mathbb{Z}/2\mathbb{Z}). \tag*{(145)} ϵ : L ∙ ⟶ Z /2 Z , ϵ ( i ∈ I ∑ b i α i ∨ ) := b 0 , ϵ ∈ Hom Z ( L ∙ , Z /2 Z ) . (145)
By (144) the root grading of Q Q Q is supported on L ∙ L^\bullet L ∙ , so
Q e ‾ : = ⨁ ν ∈ L ∙ ϵ ( ν ) = e ‾ ( Q ) ν ( e ‾ ∈ Z / 2 Z ) (146) Q_{\overline{e}} := \bigoplus_{\substack{\nu\in L^\bullet\\ \epsilon(\nu)=\overline{e}}} (Q)_\nu\qquad(\overline{e}\in\mathbb{Z}/2\mathbb{Z}) \tag*{(146)} Q e := ν ∈ L ∙ ϵ ( ν ) = e ⨁ ( Q ) ν ( e ∈ Z /2 Z ) (146)
satisfies
Q = Q 0 ‾ ⊕ Q 1 ‾ , Q e ‾ Q f ‾ ⊆ Q e + f ‾ , (147) Q=Q_{\overline{0}}\oplus Q_{\overline{1}}, \qquad Q_{\overline{e}}Q_{\overline{f}}\subseteq Q_{\overline{e+f}}, \tag*{(147)} Q = Q 0 ⊕ Q 1 , Q e Q f ⊆ Q e + f , (147)
the second inclusion because Q Q Q is root graded and ϵ \epsilon ϵ is additive. Let
φ : k ⊗ A U ˙ A L u s , ( n ) ⟶ q ↦ q ‾ V ˙ = Q (148) \varphi:k\otimes_A \dot{U}^{\mathrm{Lus},(n)}_A \mathrel{\mathop{\longrightarrow}^{q\mapsto\overline{q}}} \dot{V}=Q \tag*{(148)} φ : k ⊗ A U ˙ A Lus , ( n ) ⟶ q ↦ q V ˙ = Q (148)
be the canonical projection, a k k k -algebra homomorphism by theorem 10.3. Since J \mathcal{J} J is root graded,
φ ( ( k ⊗ A U ˙ A L u s , ( n ) ) ν ) ⊆ ( Q ) ν ( ν ∈ ⨁ i ∈ I Z α i ) . (149) \varphi\left(\left(k\otimes_A\dot{U}^{\mathrm{Lus},(n)}_A\right)_\nu\right)\subseteq(Q)_\nu\qquad\left(\nu\in\bigoplus_{i\in I}\mathbb{Z}\alpha_i\right). \tag*{(149)} φ ( ( k ⊗ A U ˙ A Lus , ( n ) ) ν ) ⊆ ( Q ) ν ( ν ∈ i ∈ I ⨁ Z α i ) . (149)
Thus φ \varphi φ plays here the role played by Fr n , r \operatorname{Fr}_{n,r} Fr n , r in (122) and by Fr r ( 4 ) \operatorname{Fr}^{(4)}_r Fr r ( 4 ) in (136) , and Q 0 ‾ ⊕ Q 1 ‾ Q_{\overline{0}}\oplus Q_{\overline{1}} Q 0 ⊕ Q 1 replaces (120) and (135) .
Lemma 10.4. Let ( s , k , ϵ , i ) ∈ N × Z × { 1 , − 1 } × { 1 , … , n } (s,k,\epsilon,i)\in\mathbb{N}\times\mathbb{Z}\times\{1,-1\}\times\{1,\ldots,n\} ( s , k , ϵ , i ) ∈ N × Z × { 1 , − 1 } × { 1 , … , n } . Then deg ( ( x i , k ϵ ) ( s ) ) = s ( k δ + ϵ α i ) \deg((x_{i,k}^{\epsilon})^{(s)})=s(k\delta+\epsilon\alpha_i) deg (( x i , k ϵ ) ( s ) ) = s ( k δ + ϵ α i ) . Let
γ : = s ( k δ + ϵ α i ) = s k α 0 + ∑ 1 ≤ j ≤ n j ≠ i 2 s k α j + s ( 2 k + ϵ ) α i . \gamma:=s(k\delta+\epsilon\alpha_i)=sk\alpha_0+\sum_{\substack{1\leq j\leq n\\j\neq i}}2sk\alpha_j+s(2k+\epsilon)\alpha_i. γ := s ( k δ + ϵ α i ) = s k α 0 + 1 ≤ j ≤ n j = i ∑ 2 s k α j + s ( 2 k + ϵ ) α i . Then, in either instance of lemma 10.2, if γ ∈ L ∙ \gamma\in L^\bullet γ ∈ L ∙ , then ϵ ( γ ) = s k ‾ = 0 ‾ \epsilon(\gamma)=\overline{sk}=\overline{0} ϵ ( γ ) = s k = 0 .
Proof. Since ℓ 0 = 1 \ell_0=1 ℓ 0 = 1 , the α 0 \alpha_0 α 0 -coordinate of γ \gamma γ is s k sk s k , so ϵ ( γ ) = s k ‾ \epsilon(\gamma)=\overline{sk} ϵ ( γ ) = s k ; it remains to show 2 ∣ s k 2\mid sk 2 ∣ s k . Note that 2 k + ϵ 2k+\epsilon 2 k + ϵ is odd.
In instance (a) of lemma 10.2, membership in L ∙ = L 1 ∙ L^\bullet=L_1^\bullet L ∙ = L 1 ∙ forces the α n \alpha_n α n -coordinate of γ \gamma γ to lie in 4 Z 4\mathbb{Z} 4 Z . If i < n i<n i < n that coordinate is 2 s k 2sk 2 s k , so 4 ∣ 2 s k 4\mid2sk 4 ∣ 2 s k and 2 ∣ s k 2\mid sk 2 ∣ s k . If i = n i=n i = n it is s ( 2 k + ϵ ) s(2k+\epsilon) s ( 2 k + ϵ ) , so 4 ∣ s ( 2 k + ϵ ) 4\mid s(2k+\epsilon) 4 ∣ s ( 2 k + ϵ ) ; since gcd ( 4 , 2 k + ϵ ) = 1 \gcd(4,2k+\epsilon)=1 g cd( 4 , 2 k + ϵ ) = 1 , this gives 4 ∣ s 4\mid s 4 ∣ s and hence 2 ∣ s k 2\mid sk 2 ∣ s k . This is lemma 9.2 at r = 1 r=1 r = 1 , where that lemma remains valid.
In instance (b) we have n = 1 n=1 n = 1 , hence i = 1 i=1 i = 1 , and membership in L ∙ = L 1 ⋆ L^\bullet=L_1^\star L ∙ = L 1 ⋆ forces the α 1 \alpha_1 α 1 -coordinate of γ \gamma γ to lie in 2 Z 2\mathbb{Z} 2 Z :
2 ∣ s ( 2 k + ϵ ) , gcd ( 2 , 2 k + ϵ ) = 1 ⟹ 2 ∣ s ⟹ 2 ∣ s k . 2\mid s(2k+\epsilon),\qquad\gcd(2,2k+\epsilon)=1 \quad\Longrightarrow\quad2\mid s \quad\Longrightarrow\quad2\mid sk. 2 ∣ s ( 2 k + ϵ ) , g cd( 2 , 2 k + ϵ ) = 1 ⟹ 2 ∣ s ⟹ 2 ∣ s k .
Theorem 10.5. For every integer n ≥ 1 n\geq1 n ≥ 1 , one has
c n , 1 ⊆ ( Φ 8 ( q ) ) , c 1 , 1 ⊆ ( Φ 4 ( q ) ) . (150) c_{n,1}\subseteq(\Phi_8(q)),\qquad c_{1,1}\subseteq(\Phi_4(q)). \tag*{(150)} c n , 1 ⊆ ( Φ 8 ( q )) , c 1 , 1 ⊆ ( Φ 4 ( q )) . (150)
Proof. Fix one of the two instances of lemma 10.2 and let
π : = { Φ 8 ( q ) , instance (a) , Φ 4 ( q ) , instance (b) , k = Frac ( A / ( π ) ) , \pi:= \begin{cases}
\Phi_{8}(q), & \text{instance (a)},\\
\Phi_{4}(q), & \text{instance (b)},
\end{cases}
\qquad\mathbb{k} = \operatorname{Frac}(\mathcal{A}/(\pi)), π := { Φ 8 ( q ) , Φ 4 ( q ) , instance (a) , instance (b) , k = Frac ( A / ( π )) ,
so that convention 10.1 holds and theorem 10.3 applies. Let φ \varphi φ be as in (148) .
Step 1: every nonzero generator block is even. Let ( λ , u ) ∈ X × C n (\lambda,u)\in X\times\mathcal{C}_{n} ( λ , u ) ∈ X × C n with φ ( u λ ‾ ) ≠ 0 \varphi(\overline{u\lambda})\ne0 φ ( u λ ) = 0 , and put γ : = deg u \gamma:=\deg u γ := deg u . If u ∈ { K δ ± 1 } ∪ { K i , δ ± 1 ∣ 1 ≤ i ≤ n } u\in\{K_{\delta}^{\pm1}\}\cup\{K_{i,\delta}^{\pm1}\mid1\le i\le n\} u ∈ { K δ ± 1 } ∪ { K i , δ ± 1 ∣ 1 ≤ i ≤ n } , then γ = 0 \gamma=0 γ = 0 , ϵ ( 0 ) = 0 ‾ \epsilon(0)=\overline{0} ϵ ( 0 ) = 0 , and (149) gives φ ( u λ ‾ ) ∈ ( Q ) 0 ⊆ Q ‾ 0 \varphi(\overline{u\lambda})\in(\mathcal{Q})_{0}\subseteq\overline{\mathcal{Q}}_{0} φ ( u λ ) ∈ ( Q ) 0 ⊆ Q 0 . If u = ( x i , k ϵ ) ( s ) u=(x_{i,k}^{\epsilon})^{(s)} u = ( x i , k ϵ ) ( s ) with 1 ≤ i ≤ n 1\le i\le n 1 ≤ i ≤ n , k ∈ Z k\in\mathbb{Z} k ∈ Z , s ∈ N s\in\mathbb{N} s ∈ N , ϵ ∈ { 1 , − 1 } \epsilon\in\{1,-1\} ϵ ∈ { 1 , − 1 } , then γ = s ( k δ + ϵ α i ) \gamma=s(k\delta+\epsilon\alpha_i) γ = s ( k δ + ϵ α i ) by (68) , and (149) together with (144) give
0 ≠ φ ( u λ ‾ ) ∈ ( Q ) γ ⟹ γ ∈ L ∙ . 0\ne\varphi(\overline{u\lambda})\in(\mathcal{Q})_{\gamma}
\quad\Longrightarrow\quad
\gamma\in L^{\bullet}. 0 = φ ( u λ ) ∈ ( Q ) γ ⟹ γ ∈ L ∙ .
Hence ϵ ( γ ) = 0 ‾ \epsilon(\gamma)=\overline{0} ϵ ( γ ) = 0 by lemma 10.4, and φ ( u λ ‾ ) ∈ Q ‾ 0 \varphi(\overline{u\lambda})\in\overline{\mathcal{Q}}_{0} φ ( u λ ) ∈ Q 0 . Therefore
∀ ( λ , u ) ∈ X × C n , φ ( u λ ‾ ) ≠ 0 ⟹ φ ( u λ ‾ ) ∈ Q ‾ 0 . \forall\,(\lambda,u)\in X\times\mathcal{C}_{n},\qquad
\varphi(\overline{u\lambda})\ne0\Longrightarrow
\varphi(\overline{u\lambda})\in\overline{\mathcal{Q}}_{0}. ∀ ( λ , u ) ∈ X × C n , φ ( u λ ) = 0 ⟹ φ ( u λ ) ∈ Q 0 .
Step 2: the whole real-current core is even. By (147) , Q = Q ‾ 0 ⊕ Q ‾ 1 \mathcal{Q}=\overline{\mathcal{Q}}_{0}\oplus\overline{\mathcal{Q}}_{1} Q = Q 0 ⊕ Q 1 is a Z / 2 Z \mathbb{Z}/2\mathbb{Z} Z /2 Z -graded k \mathbb{k} k -algebra, and φ \varphi φ is a k \mathbb{k} k -algebra homomorphism, so Step 1 and lemma 8.3 give
φ ( U A D r , ( n ) 1 0 ) ⊆ Q ‾ 0 . (151) \varphi\left(U_{\mathcal{A}}^{\mathrm{Dr},(n)}{}_{1_{0}}\right)\subseteq\overline{\mathcal{Q}}_{0}. \tag*{(151)} φ ( U A Dr , ( n ) 1 0 ) ⊆ Q 0 . (151)
Step 3: the affine divided power is odd. By lemma A.1(b), α 0 = α 0 ∙ ∈ L ∙ \alpha_{0}=\alpha_{0}^{\bullet}\in L^{\bullet} α 0 = α 0 ∙ ∈ L ∙ , whose α 0 ∙ \alpha_{0}^{\bullet} α 0 ∙ -coordinate is 1 1 1 ; hence ϵ ( α 0 ) = 1 ‾ \epsilon(\alpha_{0})=\overline{1} ϵ ( α 0 ) = 1 . Since ℓ 0 = 1 \ell_{0}=1 ℓ 0 = 1 we have E 0 ( 1 ) = E 0 E_{0}^{(1)}=E_{0} E 0 ( 1 ) = E 0 , and (144) gives
0 ≠ φ ( E 0 ( 1 ) 1 0 ) = E 0 1 0 + J ∈ ( Q ) α 0 ⊆ Q ‾ 1 . (152) 0\ne\varphi\left(E_{0}^{(1)}1_{0}\right)=E_{0}1_{0}+\mathfrak{J}\in(\mathcal{Q})_{\alpha_{0}}\subseteq\overline{\mathcal{Q}}_{1}. \tag*{(152)} 0 = φ ( E 0 ( 1 ) 1 0 ) = E 0 1 0 + J ∈ ( Q ) α 0 ⊆ Q 1 . (152)
Step 4: conclusion. Apply lemma 8.4 with r = 1 r=1 r = 1 , D e ‾ = Q e ‾ D_{\overline{e}}=\mathcal{Q}_{\overline{e}} D e = Q e , and
χ : = φ ∣ k ⊗ A ( U A L u s , ( n ) 1 0 ) . \chi:=\left.\varphi\right|_{\mathbb{k}\otimes_{\mathcal{A}}(U_{\mathcal{A}}^{\mathrm{Lus},(n)}{}_{1_{0}})}. χ := φ ∣ k ⊗ A ( U A Lus , ( n ) 1 0 ) .
Its two hypotheses are (151) and (152) , and the sum Q ‾ 0 ⊕ Q ‾ 1 \overline{\mathcal{Q}}_{0}\oplus\overline{\mathcal{Q}}_{1} Q 0 ⊕ Q 1 is internal by (147) . Hence
c n , 1 ⊆ ( π ) . \mathfrak{c}_{n,1}\subseteq(\pi). c n , 1 ⊆ ( π ) .
Instance (a) is available for every n ≥ 1 n\ge1 n ≥ 1 and gives c n , 1 ⊆ ( Φ 8 ( q ) ) \mathfrak{c}_{n,1}\subseteq(\Phi_{8}(q)) c n , 1 ⊆ ( Φ 8 ( q )) ; instance (b) has n = 1 n=1 n = 1 and gives c 1 , 1 ⊆ ( Φ 4 ( q ) ) \mathfrak{c}_{1,1}\subseteq(\Phi_{4}(q)) c 1 , 1 ⊆ ( Φ 4 ( q )) . This is (150) .
Corollary 10.6
Corollary 10.6. For every integer n ≥ 1 n\ge1 n ≥ 1 , one has
c n , 1 = { ( Φ 4 ( q ) Φ 8 ( q ) ) , n = 1 , ( Φ 8 ( q ) ) , n ≥ 2. (153) \mathfrak{c}_{n,1}=
\begin{cases}
(\Phi_{4}(q)\Phi_{8}(q)), & n=1,\\
(\Phi_{8}(q)), & n\ge2.
\end{cases} \tag*{(153)} c n , 1 = { ( Φ 4 ( q ) Φ 8 ( q )) , ( Φ 8 ( q )) , n = 1 , n ≥ 2. (153)
Proof. For n = 1 n=1 n = 1 , equation (77) with a = 1 a=1 a = 1 gives
Φ 4 ( q ) Φ 8 ( q ) = d 1 ∈ c 1 , 1 . \Phi_{4}(q)\Phi_{8}(q)=d_{1}\in\mathfrak{c}_{1,1}. Φ 4 ( q ) Φ 8 ( q ) = d 1 ∈ c 1 , 1 .
Since c 1 , 1 \mathfrak{c}_{1,1} c 1 , 1 is an A \mathcal{A} A -ideal, Theorem 10.5 yields
( Φ 4 ( q ) Φ 8 ( q ) ) ⊆ c 1 , 1 ⊆ ( Φ 4 ( q ) ) ∩ ( Φ 8 ( q ) ) = ( Φ 4 ( q ) Φ 8 ( q ) ) , (\Phi_{4}(q)\Phi_{8}(q))\subseteq\mathfrak{c}_{1,1}\subseteq
(\Phi_{4}(q))\cap(\Phi_{8}(q))=(\Phi_{4}(q)\Phi_{8}(q)), ( Φ 4 ( q ) Φ 8 ( q )) ⊆ c 1 , 1 ⊆ ( Φ 4 ( q )) ∩ ( Φ 8 ( q )) = ( Φ 4 ( q ) Φ 8 ( q )) ,
where the last equality follows because Φ 4 ( q ) \Phi_{4}(q) Φ 4 ( q ) and Φ 8 ( q ) \Phi_{8}(q) Φ 8 ( q ) are nonassociate prime elements of A \mathcal{A} A .
For n ≥ 2 n\ge2 n ≥ 2 , equation (100) with a = 1 a=1 a = 1 gives Φ 8 ( q ) = c 1 ∈ c n , 1 \Phi_{8}(q)=c_{1}\in\mathfrak{c}_{n,1} Φ 8 ( q ) = c 1 ∈ c n , 1 . Together with Theorem 10.5, this gives
( Φ 8 ( q ) ) ⊆ c n , 1 ⊆ ( Φ 8 ( q ) ) . (\Phi_{8}(q))\subseteq\mathfrak{c}_{n,1}\subseteq(\Phi_{8}(q)). ( Φ 8 ( q )) ⊆ c n , 1 ⊆ ( Φ 8 ( q )) .
The least ordinary localization
Retain the localization convention of definition 1.7 and the discussion preceding it; in particular, both sides of every equality below are extended to the same coefficient ring.
Proof of Theorem 1.8 . If
g ∉ { A 2 m ( 2 ) ∣ m ≥ 1 } , \mathfrak{g} \notin\left\{A_{2m}^{(2)} \mid m \ge1\right\}, g ∈ / { A 2 m ( 2 ) ∣ m ≥ 1 } ,
then theorem 2.1 gives
U A D r ( g ) = U A L u s ( g ) , B min ( g ) = A . U_A^{\mathrm{Dr}}(\mathfrak{g})=U_A^{\mathrm{Lus}}(\mathfrak{g}), \qquad B_{\min}(\mathfrak{g})=A. U A Dr ( g ) = U A Lus ( g ) , B m i n ( g ) = A .
Let
B = S − 1 A ⊆ K B=S^{-1}A\subseteq K B = S − 1 A ⊆ K
be an ordinary localization. We treat the two exceptional cases separately.
Suppose first that n = 1 n=1 n = 1 . By theorem 6.2,
C ∗ U A D r , ( 1 ) = C ∗ U A L u s , ( 1 ) . C^*U_A^{\mathrm{Dr},(1)}=C^*U_A^{\mathrm{Lus},(1)}. C ∗ U A Dr , ( 1 ) = C ∗ U A Lus , ( 1 ) .
Hence C ∗ ⊆ B C^*\subseteq B C ∗ ⊆ B implies, by extension of scalars,
B U A D r , ( 1 ) = B U A L u s , ( 1 ) . BU_A^{\mathrm{Dr},(1)}=BU_A^{\mathrm{Lus},(1)}. B U A Dr , ( 1 ) = B U A Lus , ( 1 ) .
Conversely, suppose that
B U A D r , ( 1 ) = B U A L u s , ( 1 ) . BU_A^{\mathrm{Dr},(1)}=BU_A^{\mathrm{Lus},(1)}. B U A Dr , ( 1 ) = B U A Lus , ( 1 ) .
The coefficient-ideal calculations give
c 1 , r ⊆ ( Φ 4 r ( q ) ) ( r ≥ 1 o d d ) \mathfrak{c}_{1,r}\subseteq(\Phi_{4r}(q)) \qquad(r\ge1\ \mathrm{odd}) c 1 , r ⊆ ( Φ 4 r ( q )) ( r ≥ 1 odd )
by theorem 10.5 and proposition 9.6, and
c 1 , r ⊆ ( Φ 8 r ( q ) ) ( r ≥ 1 ) \mathfrak{c}_{1,r}\subseteq(\Phi_{8r}(q)) \qquad(r\ge1) c 1 , r ⊆ ( Φ 8 r ( q )) ( r ≥ 1 )
by theorems 9.4 and 10.5. Therefore lemma 8.1 implies that
Φ 4 r ( q ) ∈ B × ( r ≥ 1 o d d ) , Φ 8 r ( q ) ∈ B × ( r ≥ 1 ) . \Phi_{4r}(q)\in B^\times\quad(r\ge1\ \mathrm{odd}), \qquad\Phi_{8r}(q)\in B^\times\quad(r\ge1). Φ 4 r ( q ) ∈ B × ( r ≥ 1 odd ) , Φ 8 r ( q ) ∈ B × ( r ≥ 1 ) .
Every positive integer is either odd or of the form 2 r 2r 2 r , and Φ 4 ( 2 r ) ( q ) = Φ 8 r ( q ) \Phi_{4(2r)}(q)=\Phi_{8r}(q) Φ 4 ( 2 r ) ( q ) = Φ 8 r ( q ) . Thus
Φ 4 r ( q ) ∈ B × ( r ≥ 1 ) , \Phi_{4r}(q)\in B^\times\qquad(r\ge1), Φ 4 r ( q ) ∈ B × ( r ≥ 1 ) ,
which is equivalent to C ∗ ⊆ B C^*\subseteq B C ∗ ⊆ B . Consequently,
B U A D r , ( 1 ) = B U A L u s , ( 1 ) ⟺ C ∗ ⊆ B . BU_A^{\mathrm{Dr},(1)}=BU_A^{\mathrm{Lus},(1)}
\quad\Longleftrightarrow\quad
C^*\subseteq B. B U A Dr , ( 1 ) = B U A Lus , ( 1 ) ⟺ C ∗ ⊆ B .
Now suppose that n ≥ 2 n\ge2 n ≥ 2 . By theorem 7.9,
B ∗ U A D r , ( n ) = B ∗ U A L u s , ( n ) , B^*U_A^{\mathrm{Dr},(n)}=B^*U_A^{\mathrm{Lus},(n)}, B ∗ U A Dr , ( n ) = B ∗ U A Lus , ( n ) ,
so B ∗ ⊆ B B^*\subseteq B B ∗ ⊆ B again implies equality after extension of scalars. Conversely, if
B U A D r , ( n ) = B U A L u s , ( n ) , BU_A^{\mathrm{Dr},(n)}=BU_A^{\mathrm{Lus},(n)}, B U A Dr , ( n ) = B U A Lus , ( n ) ,
then
c n , r ⊆ ( Φ 8 r ( q ) ) ( r ≥ 1 ) \mathfrak{c}_{n,r}\subseteq(\Phi_{8r}(q)) \qquad(r\ge1) c n , r ⊆ ( Φ 8 r ( q )) ( r ≥ 1 )
by theorems 9.4 and 10.5. Hence lemma 8.1 gives
Φ 8 r ( q ) ∈ B × ( r ≥ 1 ) , \Phi_{8r}(q) \in B^{\times}\qquad(r \ge1), Φ 8 r ( q ) ∈ B × ( r ≥ 1 ) ,
or equivalently B ∗ ⊂ B B^{*} \subset B B ∗ ⊂ B . Therefore
B U A D r , ( n ) = B U A L u s , ( n ) ⟺ B ∗ ⊂ B ( n ≥ 2 ) . BU_{\mathcal{A}}^{\mathrm{Dr},(n)} = BU_{\mathcal{A}}^{\mathrm{Lus},(n)}
\quad\Longleftrightarrow\quad B^{*} \subset B
\qquad(n \ge2). B U A Dr , ( n ) = B U A Lus , ( n ) ⟺ B ∗ ⊂ B ( n ≥ 2 ) .
It follows that
B min ( A 2 n ( 2 ) ) = { C ∗ , n = 1 , B ∗ , n ≥ 2. B_{\min}\left(A_{2n}^{(2)}\right)=
\begin{cases}
\mathcal{C}^{*}, & n=1,\\
B^{*}, & n\ge2.
\end{cases} B m i n ( A 2 n ( 2 ) ) = { C ∗ , B ∗ , n = 1 , n ≥ 2.
This proves all exceptional cases and completes the theorem.
Corollary 11.1 (No finite denominator set). For every
n ∈ Z ≥ 1 , t ∈ Z ≥ 0 , ( f 1 , … , f t ) ∈ ( A ∖ { 0 } ) t , n\in\mathbb{Z}_{\ge1},\qquad t\in\mathbb{Z}_{\ge0},\qquad(f_1,\ldots,f_t)\in(\mathcal{A}\setminus\{0\})^t, n ∈ Z ≥ 1 , t ∈ Z ≥ 0 , ( f 1 , … , f t ) ∈ ( A ∖ { 0 } ) t , the ordinary localization
B : = A [ f 1 − 1 , … , f t − 1 ] ⊂ K B:=\mathcal{A}[f_1^{-1},\ldots,f_t^{-1}]\subset\mathcal{K} B := A [ f 1 − 1 , … , f t − 1 ] ⊂ K satisfies
B U A D r , ( n ) ≠ B U A L u s , ( n ) . BU_{\mathcal{A}}^{\mathrm{Dr},(n)}\ne BU_{\mathcal{A}}^{\mathrm{Lus},(n)}. B U A Dr , ( n ) = B U A Lus , ( n ) .
Proof. By Gauss’s lemma, Z [ q ] \mathbb{Z}[q] Z [ q ] is a UFD, and every cyclotomic polynomial Φ m ( q ) \Phi_m(q) Φ m ( q ) is prime in Z [ q ] \mathbb{Z}[q] Z [ q ] . Localizing at the powers of q q q gives
A = Z [ q , q − 1 ] , \mathcal{A}=\mathbb{Z}[q,q^{-1}], A = Z [ q , q − 1 ] ,
which is again a UFD. Since Φ m ( 0 ) = ± 1 \Phi_m(0)=\pm1 Φ m ( 0 ) = ± 1 , the prime Φ m ( q ) \Phi_m(q) Φ m ( q ) is not associate to q q q and hence divides no power of q q q . Thus
( Φ m ( q ) ) ∩ { q k ∣ k ≥ 0 } = ∅ , (\Phi_m(q))\cap\{q^k\mid k\ge0\}=\varnothing, ( Φ m ( q )) ∩ { q k ∣ k ≥ 0 } = ∅ ,
so the prime ideal ( Φ m ( q ) ) (\Phi_m(q)) ( Φ m ( q )) remains prime after localization at the powers of q q q . Hence Φ m ( q ) \Phi_m(q) Φ m ( q ) is prime in A = Z [ q , q − 1 ] \mathcal{A}=\mathbb{Z}[q,q^{-1}] A = Z [ q , q − 1 ] .
The units of A \mathcal{A} A are precisely ± q k \pm q^k ± q k , k ∈ Z k\in\mathbb{Z} k ∈ Z . If two cyclotomic polynomials are associates in A \mathcal{A} A , then
Φ m ( q ) = ± q k Φ m ′ ( q ) \Phi_m(q)=\pm q^k\Phi_{m'}(q) Φ m ( q ) = ± q k Φ m ′ ( q )
for some k ∈ Z k\in\mathbb{Z} k ∈ Z . Comparing the lowest powers of q q q gives k = 0 k=0 k = 0 , because both cyclotomic polynomials have nonzero constant term. Their monicity then gives
Φ m ( q ) = Φ m ′ ( q ) , \Phi_m(q)=\Phi_{m'}(q), Φ m ( q ) = Φ m ′ ( q ) ,
and hence m = m ′ m=m' m = m ′ . Therefore both
{ Φ 4 r ( q ) ∣ r ≥ 1 } and { Φ 8 r ( q ) ∣ r ≥ 1 } \{\Phi_{4r}(q)\mid r\ge1\}\qquad\text{and}\qquad\{\Phi_{8r}(q)\mid r\ge1\} { Φ 4 r ( q ) ∣ r ≥ 1 } and { Φ 8 r ( q ) ∣ r ≥ 1 }
are infinite families of pairwise nonassociate primes in A \mathcal{A} A .
Now let
B = A [ f 1 − 1 , … , f t − 1 ] ⊂ K , B=\mathcal{A}[f_1^{-1},\ldots,f_t^{-1}]\subset\mathcal{K}, B = A [ f 1 − 1 , … , f t − 1 ] ⊂ K ,
where f 1 , … , f t ≠ 0 f_1,\ldots,f_t\ne0 f 1 , … , f t = 0 . If a nonunit prime p ∈ A p\in\mathcal{A} p ∈ A is invertible in B B B , then
p − 1 = a f 1 e 1 ⋯ f t e t (154) p^{-1}=\frac{a}{f_1^{e_1}\cdots f_t^{e_t}} \tag*{(154)} p − 1 = f 1 e 1 ⋯ f t e t a (154)
for some a ∈ A a \in\mathcal{A} a ∈ A and e 1 , … , e t ≥ 0 e_1,\ldots,e_t \ge0 e 1 , … , e t ≥ 0 . Thus
f 1 e 1 ⋯ f t e t = p a , f_1^{e_1}\cdots f_t^{e_t}=pa, f 1 e 1 ⋯ f t e t = p a ,
so p p p divides one of f 1 , … , f t f_1,\ldots,f_t f 1 , … , f t . Since each f j f_j f j has only finitely many prime divisors in the UFD A \mathcal{A} A , only finitely many pairwise nonassociate primes can become invertible in B B B .
On the other hand, theorem 1.8 shows that
B U A D r , ( n ) = B U A L u s , ( n ) BU_A^{\mathrm{Dr},(n)}=BU_A^{\mathrm{Lus},(n)} B U A Dr , ( n ) = B U A Lus , ( n )
would force every Φ 4 r ( q ) \Phi_{4r}(q) Φ 4 r ( q ) to be invertible in B B B when n = 1 n=1 n = 1 , and every Φ 8 r ( q ) \Phi_{8r}(q) Φ 8 r ( q ) to be invertible in B B B when n ≥ 2 n\ge2 n ≥ 2 . In either case this would make infinitely many pairwise nonassociate primes invertible in B B B , a contradiction.
Technical details in Section 10.2
We retain Convention 10.1 and the notation of Section 10, with N = Z ≥ 0 \mathbb{N}=\mathbb{Z}_{\ge0} N = Z ≥ 0 . This appendix supplies the endpoint verification, word calculus, and triangular decomposition used in the proof of theorem 10.3.
Endpoint data and Gaussian vanishing
Lemma A.1. (a) X ∘ X^\circ X ∘ is a subgroup of X X X , and λ + ν ∈ X ∘ \lambda+\nu\in X^\circ λ + ν ∈ X ∘ for all ( λ , ν ) ∈ X ∘ × L ∘ (\lambda,\nu)\in X^\circ\times L^\circ ( λ , ν ) ∈ X ∘ × L ∘ .
(b) α 0 = α 0 ∘ ∈ L ∘ \alpha_0=\alpha_0^\circ\in L^\circ α 0 = α 0 ∘ ∈ L ∘ , and
I ∘ : = { i ∈ I ∣ ℓ i > 1 } ⊆ I ∖ { 0 } . I^\circ:=\{i\in I\mid\ell_i>1\}\subseteq I\setminus\{0\}. I ∘ := { i ∈ I ∣ ℓ i > 1 } ⊆ I ∖ { 0 } . (c) For ( i , a ) ∈ I × Z ≥ 1 (i,a)\in I\times\mathbb{Z}_{\ge1} ( i , a ) ∈ I × Z ≥ 1 ,
ℓ i ∤ a ⟹ ℓ i > 1 , ℓ i ∣ a ⟹ a α i = a ℓ i α i ∘ ∈ L ∘ . \ell_i\nmid a\ \Longrightarrow\ \ell_i>1,\qquad
\ell_i\mid a\ \Longrightarrow\ a\alpha_i=\frac{a}{\ell_i}\alpha_i^\circ\in L^\circ. ℓ i ∤ a ⟹ ℓ i > 1 , ℓ i ∣ a ⟹ a α i = ℓ i a α i ∘ ∈ L ∘ .
Proof. (a) Each defining condition in (141) is a subgroup condition, so X ∘ ≤ X X^\circ\le X X ∘ ≤ X ; the second assertion then follows from L ∘ ⊆ X ∘ L^\circ\subseteq X^\circ L ∘ ⊆ X ∘ in (142) .
(b) α 0 ∘ = ℓ 0 α 0 = α 0 \alpha_0^\circ=\ell_0\alpha_0=\alpha_0 α 0 ∘ = ℓ 0 α 0 = α 0 by (143) , and the same identity gives 0 ∉ I ∘ 0\notin I^\circ 0 ∈ / I ∘ .
(c) If ℓ i = 1 \ell_i=1 ℓ i = 1 then ℓ i ∣ a \ell_i\mid a ℓ i ∣ a ; this is the first implication. The second is immediate.
For c ∈ Z c\in\mathbb{Z} c ∈ Z and t ∈ N t\in\mathbb{N} t ∈ N we use Lusztig’s symmetric Gaussian convention: for c ≥ 0 c\ge0 c ≥ 0 it is the usual Laurent polynomial, with value zero for t > c t>c t > c , and for c = − d < 0 c=-d<0 c = − d < 0 ,
\begin{bmatrix}
-d\\
t
\end{bmatrix}_{u}
:=(-1)^t
\begin{bmatrix}
d+t-1\\
t
\end{bmatrix}_{u}, \tag*{(155)}\end{bmatrix}
as in [21 §1.3.1(a)] .
Lemma A.2 (Endpoint Gaussian vanishing). Let k k k be a field with char k = 0 \operatorname{char} k=0 char k = 0 , let ξ ∈ k × \xi\in k^\times ξ ∈ k × , and put e : = ord ( ξ 2 ) < ∞ e:=\operatorname{ord}(\xi^2)<\infty e := ord ( ξ 2 ) < ∞ . Then
\begin{bmatrix}
c\\
t
\end{bmatrix}_{\xi}=0\qquad(c\in e\mathbb{Z},\ t\in\mathbb{N},\ e\nmid t), \tag*{(156)}\end{bmatrix}
with (A.1) understood for c < 0 c<0 c < 0 .
Proof. This is [21 Lemma 34.1.2(a)] with ℓ = e \ell=e ℓ = e , after specializing the quantum parameter to ξ \xi ξ . For e = 1 e=1 e = 1 the condition e ∤ t e\nmid t e ∤ t is empty.
Specialized halves, words, and half ideals
Specialized halves. Let
U A + : = A ⟨ E i ( a ) ∣ i ∈ I , a ∈ N ⟩ , U A − : = A ⟨ F i ( a ) ∣ i ∈ I , a ∈ N ⟩ , (157) U_{\mathcal A}^{+}:=\mathcal A\langle E_i^{(a)}\mid i\in I,\ a\in\mathbb N\rangle,\qquad U_{\mathcal A}^{-}:=\mathcal A\langle F_i^{(a)}\mid i\in I,\ a\in\mathbb N\rangle, \tag*{(157)} U A + := A ⟨ E i ( a ) ∣ i ∈ I , a ∈ N ⟩ , U A − := A ⟨ F i ( a ) ∣ i ∈ I , a ∈ N ⟩ , (157)
and let U ˙ A L u s , ( n ) \dot U_{\mathcal A}^{\mathrm{Lus},(n)} U ˙ A Lus , ( n ) be the source modified integral form of § 6.3 , that is the form denoted U ˙ A \dot U_{\mathcal A} U ˙ A in § 6.2 . Define
V ˙ : = k ⊗ A , q ↦ q ˉ U ˙ A L u s , ( n ) , V ± : = k ⊗ A , q ↦ q ˉ U A ± , (158) \dot V:=\mathbb{k}\otimes_{\mathcal A,q\mapsto\bar q}\dot U_{\mathcal A}^{\mathrm{Lus},(n)},\qquad V^{\pm}:=\mathbb{k}\otimes_{\mathcal A,q\mapsto\bar q}U_{\mathcal A}^{\pm}, \tag*{(158)} V ˙ := k ⊗ A , q ↦ q ˉ U ˙ A Lus , ( n ) , V ± := k ⊗ A , q ↦ q ˉ U A ± , (158)
so that V ± V^{\pm} V ± are unital k \mathbb{k} k -algebras graded by ⨁ i ∈ I Z α i \bigoplus_{i\in I}\mathbb{Z}\alpha_i ⨁ i ∈ I Z α i , and V ˙ \dot V V ˙ is a k \mathbb{k} k -algebra with orthogonal idempotents 1 λ 1_{\lambda} 1 λ (λ ∈ X \lambda\in X λ ∈ X ). We write E i ( a ) , F i ( a ) , 1 λ E_i^{(a)},F_i^{(a)},1_{\lambda} E i ( a ) , F i ( a ) , 1 λ again for the images of these elements. Lusztig’s integral triangular decomposition and its integral basis [21 §§23.2.1–23.2.2] are statements about free A \mathcal A A -modules and therefore remain direct after the base change (A.4):
V ˙ = ⨁ λ ∈ X V − 1 λ V + , V − ⊗ k k 1 λ ⊗ k V + → ∼ V − 1 λ V + ( λ ∈ X ) , (159) \dot V=\bigoplus_{\lambda\in X}V^{-}1_{\lambda}V^{+},\qquad V^{-}\otimes_{\mathbb{k}}\mathbb{k}1_{\lambda}\otimes_{\mathbb{k}}V^{+}\xrightarrow{\sim}V^{-}1_{\lambda}V^{+}\quad(\lambda\in X), \tag*{(159)} V ˙ = λ ∈ X ⨁ V − 1 λ V + , V − ⊗ k k 1 λ ⊗ k V + ∼ V − 1 λ V + ( λ ∈ X ) , (159)
the second map being multiplication. In particular the specialized basis gives
0 ≠ 1 ∈ ( V − ) 0 , 0 ≠ E 0 ∈ ( V + ) α 0 . (160) 0\ne1\in(V^{-})_{0},\qquad0\ne E_{0}\in(V^{+})_{\alpha_{0}}. \tag*{(160)} 0 = 1 ∈ ( V − ) 0 , 0 = E 0 ∈ ( V + ) α 0 . (160)
Letters and words. We use formal letters and words so that their labels remain part of the data. Put
S : = { 1 , − 1 } × I × Z ≥ 1 , S ε : = { ε } × I × Z ≥ 1 ( ε ∈ { 1 , − 1 } ) , (161) \mathcal S:=\{1,-1\}\times I\times\mathbb Z_{\geq1},\qquad\mathcal S^{\varepsilon}:=\{\varepsilon\}\times I\times\mathbb Z_{\geq1}\quad(\varepsilon\in\{1,-1\}), \tag*{(161)} S := { 1 , − 1 } × I × Z ≥ 1 , S ε := { ε } × I × Z ≥ 1 ( ε ∈ { 1 , − 1 }) , (161)
and call the elements of S \mathcal S S letters . Write V 1 : = V + V^{1}:=V^{+} V 1 := V + and V − 1 : = V − V^{-1}:=V^{-} V − 1 := V − . Define ev ε : S ε ⟶ V ε \operatorname{ev}_{\varepsilon}:\mathcal S^{\varepsilon}\longrightarrow V^{\varepsilon} ev ε : S ε ⟶ V ε by
ev ε ( ε , i , a ) : = { E i ( a ) , ε = 1 , F i ( a ) , ε = − 1 , deg ( ε , i , a ) : = ε a α i , S n b : = { ( ε , i , a ) ∈ S ∣ ℓ i ∤ a } . (162) \operatorname{ev}_{\varepsilon}(\varepsilon,i,a):=
\begin{cases}
E_i^{(a)},&\varepsilon=1,\\
F_i^{(a)},&\varepsilon=-1,
\end{cases}
\qquad
\deg(\varepsilon,i,a):=\varepsilon a\alpha_i,\qquad
\mathcal S^{\mathrm{nb}}:=\{(\varepsilon,i,a)\in\mathcal S\mid\ell_i\nmid a\}. \tag*{(162)} ev ε ( ε , i , a ) := { E i ( a ) , F i ( a ) , ε = 1 , ε = − 1 , deg ( ε , i , a ) := ε a α i , S nb := {( ε , i , a ) ∈ S ∣ ℓ i ∤ a } . (162)
A letter is nonblock if it belongs to S n b \mathcal S^{\mathrm{nb}} S nb , and block otherwise. By Lemma A.1 (b)–(c),
S n b ⊂ { 1 , − 1 } × I ∘ × Z ≥ 1 , deg ( S ∖ S n b ) ⊂ L ∙ . (163) \mathcal S^{\mathrm{nb}}\subset\{1,-1\}\times I^{\circ}\times\mathbb Z_{\geq1},\qquad
\deg(\mathcal S\setminus\mathcal S^{\mathrm{nb}})\subset L^{\bullet}. \tag*{(163)} S nb ⊂ { 1 , − 1 } × I ∘ × Z ≥ 1 , deg ( S ∖ S nb ) ⊂ L ∙ . (163)
For each ε ∈ { 1 , − 1 } \varepsilon\in\{1,-1\} ε ∈ { 1 , − 1 } , identify a letter with the corresponding one-letter word and extend ev ε \operatorname{ev}_{\varepsilon} ev ε to
ev ε : ∐ m ≥ 0 ( S ε ) m ⟶ V ε \operatorname{ev}_{\varepsilon}:\coprod_{m\geq0}(\mathcal S^{\varepsilon})^{m}\longrightarrow V^{\varepsilon} ev ε : m ≥ 0 ∐ ( S ε ) m ⟶ V ε
by
ev ε ( x 1 , … , x m ) : = ev ε ( x 1 ) ⋯ ev ε ( x m ) . \operatorname{ev}_{\varepsilon}(x_1,\ldots,x_m):=\operatorname{ev}_{\varepsilon}(x_1)\cdots\operatorname{ev}_{\varepsilon}(x_m). ev ε ( x 1 , … , x m ) := ev ε ( x 1 ) ⋯ ev ε ( x m ) .
The empty word has value 1 ∈ V ε 1\in V^{\varepsilon} 1 ∈ V ε . The set of words with a terminal weight is
M : = ∐ m ≥ 0 ( S m × X ) , w = ( x 1 , … , x m ; η ) . (164) \mathcal M:=\coprod_{m\geq0}(\mathcal S^{m}\times X),\qquad\mathbf w=(x_1,\ldots,x_m;\eta). \tag*{(164)} M := m ≥ 0 ∐ ( S m × X ) , w = ( x 1 , … , x m ; η ) . (164)
For x j = ( ε j , i j , a j ) x_j=(\varepsilon_j,i_j,a_j) x j = ( ε j , i j , a j ) , define the intermediate weights and the map ev V ˙ : M ⟶ V ˙ \operatorname{ev}_{\dot V}:\mathcal{M}\longrightarrow\dot V ev V ˙ : M ⟶ V ˙ by
λ j ( w ) : = η + ∑ u = j + 1 m deg ( x u ) ( 0 ≤ j ≤ m ) , \lambda_j(\mathbf{w}):=\eta+\sum_{u=j+1}^{m}\deg(x_u)\qquad(0\le j\le m), λ j ( w ) := η + u = j + 1 ∑ m deg ( x u ) ( 0 ≤ j ≤ m ) ,
ev V ˙ ( w ) : = { 1 η , m = 0 , ∏ j = 1 m ( ev ε j ( x j ) 1 λ j ( w ) ) , m ≥ 1. (165) \operatorname{ev}_{\dot V}(\mathbf{w}):=
\begin{cases}
1_\eta, & m=0,\\
\prod_{j=1}^{m}\left(\operatorname{ev}_{\varepsilon_j}(x_j)1_{\lambda_j(\mathbf{w})}\right), & m\ge1.
\end{cases} \tag*{(165)} ev V ˙ ( w ) := { 1 η , ∏ j = 1 m ( ev ε j ( x j ) 1 λ j ( w ) ) , m = 0 , m ≥ 1. (165)
The product is taken in increasing order of j j j . Each factor denotes the modified divided power E i j ( a j ) 1 λ j ( w ) E_{i_j}^{(a_j)}1_{\lambda_j(\mathbf{w})} E i j ( a j ) 1 λ j ( w ) or F i j ( a j ) 1 λ j ( w ) F_{i_j}^{(a_j)}1_{\lambda_j(\mathbf{w})} F i j ( a j ) 1 λ j ( w ) , according to the sign ε j \varepsilon_j ε j , and therefore belongs to V ˙ \dot V V ˙ . Since
λ m ( w ) = η , λ j − 1 ( w ) = λ j ( w ) + deg ( x j ) , \lambda_m(\mathbf{w})=\eta,\qquad\lambda_{j-1}(\mathbf{w})=\lambda_j(\mathbf{w})+\deg(x_j), λ m ( w ) = η , λ j − 1 ( w ) = λ j ( w ) + deg ( x j ) ,
the factors have matching intermediate weights, and
ev V ˙ ( w ) ∈ 1 λ 0 ( w ) V ˙ 1 η . \operatorname{ev}_{\dot V}(\mathbf{w})\in1_{\lambda_0(\mathbf{w})}\dot V1_\eta. ev V ˙ ( w ) ∈ 1 λ 0 ( w ) V ˙ 1 η .
We write both ev F \operatorname{ev}_{\mathcal{F}} ev F and ev V ˙ \operatorname{ev}_{\dot V} ev V ˙ as ev \operatorname{ev} ev for these evaluation maps when the domain is clear. With this convention, for m ≥ 1 m\ge1 m ≥ 1 ,
ev ( w ) = ev ( x 1 ) ⋯ ev ( x m ) 1 η = ( 1 λ 0 ( w ) ev ( x 1 ) 1 λ 1 ( w ) ) ⋯ ( 1 λ m − 1 ( w ) ev ( x m ) 1 λ m ( w ) ) . (166) \begin{aligned}
\operatorname{ev}(\mathbf{w})=\operatorname{ev}(x_1)\cdots\operatorname{ev}(x_m)1_\eta\\
&= \left(1_{\lambda_0(\mathbf{w})}\operatorname{ev}(x_1)1_{\lambda_1(\mathbf{w})}\right)\cdots\left(1_{\lambda_{m-1}(\mathbf{w})}\operatorname{ev}(x_m)1_{\lambda_m(\mathbf{w})}\right). \tag*{(166)}
\end{aligned} ev ( w ) = ev ( x 1 ) ⋯ ev ( x m ) 1 η = ( 1 λ 0 ( w ) ev ( x 1 ) 1 λ 1 ( w ) ) ⋯ ( 1 λ m − 1 ( w ) ev ( x m ) 1 λ m ( w ) ) . (166)
where the first line is shorthand for the product defined in (A.11). Since U A ± U_A^\pm U A ± is generated by the divided powers, the triangular decomposition (A.5) gives
V ˙ = span k ev ( M ) . (167) \dot V=\operatorname{span}_{\Bbbk}\operatorname{ev}(\mathcal{M}). \tag*{(167)} V ˙ = span k ev ( M ) . (167)
Define
B : = { w = ( x 1 , … , x m ; η ) ∈ M | ( ∃ i ∈ { 0 , … , m } : λ i ( w ) ∉ X ∙ ) ∨ ( ∃ j ∈ { 1 , … , m } : x j ∈ S n b ) } , \mathcal{B}:=\left\{\mathbf{w}=(x_1,\ldots,x_m;\eta)\in\mathcal{M}\ \middle|\ \left(\exists i\in\{0,\ldots,m\}:\lambda_i(\mathbf{w})\notin X^\bullet\right)\vee\left(\exists j\in\{1,\ldots,m\}:x_j\in\mathcal{S}^{\mathrm{nb}}\right)\right\}, B := { w = ( x 1 , … , x m ; η ) ∈ M ( ∃ i ∈ { 0 , … , m } : λ i ( w ) ∈ / X ∙ ) ∨ ( ∃ j ∈ { 1 , … , m } : x j ∈ S nb ) } ,
B △ : = { w = ( x 1 , … , x m ; η ) ∈ B | ∃ p ∈ { 0 , … , m } : x i ∈ S − 1 ( 1 ≤ i ≤ p ) , x j ∈ S 1 ( p < j ≤ m ) } , \mathcal{B}^{\triangle}:=\left\{\mathbf{w}=(x_1,\ldots,x_m;\eta)\in\mathcal{B}\ \middle|\ \exists p\in\{0,\ldots,m\}:x_i\in\mathcal{S}^{-1}\ (1\le i\le p),\ x_j\in\mathcal{S}^{1}\ (p<j\le m)\right\}, B △ := { w = ( x 1 , … , x m ; η ) ∈ B ∃ p ∈ { 0 , … , m } : x i ∈ S − 1 ( 1 ≤ i ≤ p ) , x j ∈ S 1 ( p < j ≤ m ) } ,
J : = span k ev ( B ) . (168) \mathcal{J}:=\operatorname{span}_{\Bbbk}\operatorname{ev}(\mathcal{B}). \tag*{(168)} J := span k ev ( B ) . (168)
Thus B △ \mathcal{B}^{\triangle} B △ is the subset of B \mathcal{B} B consisting of triangularly ordered bad words, namely those in which all negative letters precede all positive letters. Words in B \mathcal{B} B are called bad , the remaining words good .
The subspace J \mathcal{J} J imposes exactly the two selection rules that quantum Frobenius would impose: forbidden weight idempotents vanish, and only divided powers whose exponents are multiples of the scaling integers survive. At ℓ 0 = 1 \ell_0=1 ℓ 0 = 1 the criterion of section 9 is unavailable. The next three results prove that these rules are nevertheless compatible with triangular straightening and with multiplication.
The half ideals. Let
J + : = ∑ i ∈ I ∘ V + E i V + , J − : = ∑ i ∈ I ∘ V − F i V − . (169) \mathcal{J}^{+}:=\sum_{i\in I^\circ}V^{+}E_iV^{+},\qquad\mathcal{J}^{-}:=\sum_{i\in I^\circ}V^{-}F_iV^{-}. \tag*{(169)} J + := i ∈ I ∘ ∑ V + E i V + , J − := i ∈ I ∘ ∑ V − F i V − . (169)
with I ∘ = { i ∈ I ∣ ℓ i > 1 } ⊆ I ∖ { 0 } I^\circ=\{i\in I\mid\ell_i>1\}\subseteq I\setminus\{0\} I ∘ = { i ∈ I ∣ ℓ i > 1 } ⊆ I ∖ { 0 } as in lemma A.1(b). By construction J ± \mathcal{J}^{\pm} J ± is the two-sided ideal of the unital algebra V ± V^{\pm} V ± generated by { E i } i ∈ I ∘ \{E_i\}_{i\in I^\circ} { E i } i ∈ I ∘ , resp. { F i } i ∈ I ∘ \{F_i\}_{i\in I^\circ} { F i } i ∈ I ∘ :
V ± J ± ⊆ J ± , J ± V ± ⊆ J ± , J ± = ⨁ ν ( J ± ∩ ( V ± ) ν ) , (170) V^\pm\mathcal{J}^\pm\subseteq\mathcal{J}^\pm,\qquad\mathcal{J}^\pm V^\pm\subseteq\mathcal{J}^\pm,\qquad\mathcal{J}^\pm=\bigoplus_\nu\left(\mathcal{J}^\pm\cap(V^\pm)_\nu\right), \tag*{(170)} V ± J ± ⊆ J ± , J ± V ± ⊆ J ± , J ± = ν ⨁ ( J ± ∩ ( V ± ) ν ) , (170)
the last equality because the generators are root homogeneous. These are ideals of the half algebras only, not of V ˙ \dot V V ˙ .
Lemma A.3 (Endpoint half ideals). For ε ∈ { 1 , − 1 } \varepsilon\in\{1,-1\} ε ∈ { 1 , − 1 } put V 1 : = V + V^{1}:=V^{+} V 1 := V + , V − 1 : = V − V^{-1}:=V^{-} V − 1 := V − , J 1 : = J + \mathfrak{J}^{1}:=\mathfrak{J}^{+} J 1 := J + , J − 1 : = J − \mathfrak{J}^{-1}:=\mathfrak{J}^{-} J − 1 := J − ,
W ε : = { ( x 1 , … , x m ) | m ∈ Z ≥ 1 , x j ∈ S ε ( 1 ≤ j ≤ m ) , ∃ j ∈ { 1 , … , m } : x j ∈ S n b } . \mathcal{W}^{\varepsilon}:=\left\{(x_{1},\ldots,x_{m})\ \middle|\ m\in\mathbb{Z}_{\geq1},\quad x_{j}\in\mathcal{S}^{\varepsilon}\ (1\leq j\leq m),\quad\exists j\in\{1,\ldots,m\}:x_{j}\in\mathcal{S}^{\mathrm{nb}}\right\}. W ε := { ( x 1 , … , x m ) m ∈ Z ≥ 1 , x j ∈ S ε ( 1 ≤ j ≤ m ) , ∃ j ∈ { 1 , … , m } : x j ∈ S nb } . Then
J ε = span k ev ( W ε ) ( ε ∈ { 1 , − 1 } ) . \mathfrak{J}^{\varepsilon}=\operatorname{span}_{\mathbb{k}}\operatorname{ev}(\mathcal{W}^{\varepsilon})\qquad(\varepsilon\in\{1,-1\}). J ε = span k ev ( W ε ) ( ε ∈ { 1 , − 1 }) . Proof. Step 1: Let ( ε , i , a ) ∈ S n b (\varepsilon,i,a)\in\mathcal{S}^{\mathrm{nb}} ( ε , i , a ) ∈ S nb , so a ≥ 1 a\geq1 a ≥ 1 and ℓ i ∤ a \ell_i\nmid a ℓ i ∤ a ; by lemma A.1(c), ℓ i > 1 \ell_i>1 ℓ i > 1 and i ∈ I ∘ i\in I^{\circ} i ∈ I ∘ . Over A \mathcal{A} A one has [ a ] q i = ∑ j = 0 a − 1 q i a − 1 − 2 j [a]_{q_i}=\sum_{j=0}^{a-1}q_i^{a-1-2j} [ a ] q i = ∑ j = 0 a − 1 q i a − 1 − 2 j , and multiplication by q i 2 − 1 q_i^2-1 q i 2 − 1 telescopes to the identity of Laurent polynomials
( q i 2 − 1 ) [ a ] q i = q i 1 − a ( q i 2 a − 1 ) in A . (171) (q_i^2-1)[a]_{q_i}=q_i^{1-a}(q_i^{2a}-1)\qquad\text{in }\mathcal{A}. \tag*{(171)} ( q i 2 − 1 ) [ a ] q i = q i 1 − a ( q i 2 a − 1 ) in A . (171) Applying A → k \mathcal{A}\to\mathbb{k} A → k , q ↦ q ‾ q\mapsto\overline{q} q ↦ q , gives
( q k , i 2 − 1 ) [ a ] q k , i = q k , i 1 − a ( q k , i 2 a − 1 ) . (172) (q_{k,i}^2-1)[a]_{q_{k,i}}=q_{k,i}^{1-a}(q_{k,i}^{2a}-1). \tag*{(172)} ( q k , i 2 − 1 ) [ a ] q k , i = q k , i 1 − a ( q k , i 2 a − 1 ) . (172) Here q k , i ∈ k × q_{k,i}\in\mathbb{k}^{\times} q k , i ∈ k × , so q k , i 1 − a ∈ k × q_{k,i}^{1-a}\in\mathbb{k}^{\times} q k , i 1 − a ∈ k × ; and since ord ( q k , i 2 ) = ℓ i < ∞ \operatorname{ord}(q_{k,i}^2)=\ell_i<\infty ord ( q k , i 2 ) = ℓ i < ∞ ,
q k , i 2 a = ( q k , i 2 ) a = 1 ⟺ ℓ i ∣ a , q_{k,i}^{2a}=(q_{k,i}^2)^a=1\Longleftrightarrow\ell_i\mid a, q k , i 2 a = ( q k , i 2 ) a = 1 ⟺ ℓ i ∣ a , which fails by hypothesis. Hence the right-hand side of (A.18) is nonzero and, k \mathbb{k} k being a field, so are both factors on the left:
q k , i 2 ≠ 1 , [ a ] q k , i ∈ k × . (173) q_{k,i}^2\ne1,\qquad[a]_{q_{k,i}}\in\mathbb{k}^{\times}. \tag*{(173)} q k , i 2 = 1 , [ a ] q k , i ∈ k × . (173) The divided-power identities
E i ( a − 1 ) E i = [ a ] q i E i ( a ) , F i ( a − 1 ) F i = [ a ] q i F i ( a ) E_i^{(a-1)}E_i=[a]_{q_i}E_i^{(a)},\qquad F_i^{(a-1)}F_i=[a]_{q_i}F_i^{(a)} E i ( a − 1 ) E i = [ a ] q i E i ( a ) , F i ( a − 1 ) F i = [ a ] q i F i ( a ) hold in U A ± U_{\mathcal{A}}^{\pm} U A ± , hence in V ± V^{\pm} V ± after the base change (A.4); they cannot be obtained inside V ± V^{\pm} V ± by dividing by [ a ] q k , i [a]_{q_{k,i}} [ a ] q k , i , which may vanish. Multiplying by [ a ] q k , i − 1 [a]_{q_{k,i}}^{-1} [ a ] q k , i − 1 gives
ev ( ( 1 , i , a ) ) = E i ( a ) = [ a ] q k , i − 1 E i ( a − 1 ) E i ∈ J + , ev ( ( − 1 , i , a ) ) = F i ( a ) = [ a ] q k , i − 1 F i ( a − 1 ) F i ∈ J − . (174) \operatorname{ev}((1,i,a))=E_i^{(a)}=[a]_{q_{k,i}}^{-1}E_i^{(a-1)}E_i\in\mathfrak{J}^{+},\qquad\operatorname{ev}((-1,i,a))=F_i^{(a)}=[a]_{q_{k,i}}^{-1}F_i^{(a-1)}F_i\in\mathfrak{J}^{-}. \tag*{(174)} ev (( 1 , i , a )) = E i ( a ) = [ a ] q k , i − 1 E i ( a − 1 ) E i ∈ J + , ev (( − 1 , i , a )) = F i ( a ) = [ a ] q k , i − 1 F i ( a − 1 ) F i ∈ J − . (174) Step 2: span k ev ( W ε ) ⊆ J ε \operatorname{span}_{\mathbb{k}}\operatorname{ev}(\mathcal{W}^{\varepsilon})\subseteq\mathfrak{J}^{\varepsilon} span k ev ( W ε ) ⊆ J ε . Let ( x 1 , … , x m ) ∈ W ε (x_{1},\ldots,x_{m})\in\mathcal{W}^{\varepsilon} ( x 1 , … , x m ) ∈ W ε and choose j j j with x j ∈ S n b x_j\in\mathcal{S}^{\mathrm{nb}} x j ∈ S nb . Then
ev ( x 1 , … , x m ) = ( ev ( x 1 ) ⋯ ev ( x j − 1 ) ) ⏟ ∈ V ε ev ( x j ) ( ev ( x j + 1 ) ⋯ ev ( x m ) ) ⏟ ∈ V ε , \operatorname{ev}(x_1,\ldots,x_m)=\underbrace{\bigl(\operatorname{ev}(x_1)\cdots\operatorname{ev}(x_{j-1})\bigr)}_{\in V^{\varepsilon}}\operatorname{ev}(x_j)\underbrace{\bigl(\operatorname{ev}(x_{j+1})\cdots\operatorname{ev}(x_m)\bigr)}_{\in V^{\varepsilon}}, ev ( x 1 , … , x m ) = ∈ V ε ( ev ( x 1 ) ⋯ ev ( x j − 1 ) ) ev ( x j ) ∈ V ε ( ev ( x j + 1 ) ⋯ ev ( x m ) ) , empty products being 1 1 1 . By Step 1, ev ( x j ) ∈ J ε \operatorname{ev}(x_j)\in\mathfrak{J}^{\varepsilon} ev ( x j ) ∈ J ε , and (A.16) gives ev ( x 1 , … , x m ) ∈ J ε \operatorname{ev}(x_1,\ldots,x_m)\in\mathfrak{J}^{\varepsilon} ev ( x 1 , … , x m ) ∈ J ε . Taking k \mathbb{k} k -spans yields the inclusion.
Step 3: the reverse inclusion. Let i ∈ I ∘ i\in I^{\circ} i ∈ I ∘ and x , y ∈ V + x,y\in V^{+} x , y ∈ V + . Since V + V^{+} V + is spanned by the elements ev ( u 1 , … , u p ) \operatorname{ev}(u_1,\ldots,u_p) ev ( u 1 , … , u p ) with u 1 , … , u p ∈ S 1 u_1,\ldots,u_p\in\mathcal{S}^{1} u 1 , … , u p ∈ S 1 and p ∈ N p\in\mathbb{N} p ∈ N , we may expand x x x and y y y in such elements and obtain
x E i y ∈ span k { ev ( u 1 , … , u p , ( 1 , i , 1 ) , v 1 , … , v s ) ∣ u ∙ , v ∙ ∈ S 1 } . xE_i y\in\operatorname{span}_{\mathbb{k}}\left\{\operatorname{ev}(u_1,\ldots,u_p,(1,i,1),v_1,\ldots,v_s)\mid u_{\bullet},v_{\bullet}\in\mathcal{S}^{1}\right\}. x E i y ∈ span k { ev ( u 1 , … , u p , ( 1 , i , 1 ) , v 1 , … , v s ) ∣ u ∙ , v ∙ ∈ S 1 } . Every word occurring here contains the letter ( 1 , i , 1 ) (1,i,1) ( 1 , i , 1 ) , and ( 1 , i , 1 ) ∈ S n b (1,i,1)\in\mathcal{S}^{\mathrm{nb}} ( 1 , i , 1 ) ∈ S nb because ℓ i > 1 \ell_i>1 ℓ i > 1 does not divide 1 1 1 ; hence every such word lies in W 1 \mathcal{W}^{1} W 1 . Summing over i ∈ I ∘ i\in I^{\circ} i ∈ I ∘ gives J + ⊆ span k ev ( W 1 ) \mathfrak{J}^{+}\subseteq\operatorname{span}_{\mathbb{k}}\operatorname{ev}(\mathcal{W}^{1}) J + ⊆ span k ev ( W 1 ) . The case ε = − 1 \varepsilon=-1 ε = − 1 is identical in V − V^{-} V − .
Bad-word straightening and triangular decomposition
Lemma A.4.
span k ev ( B ) = span k ev ( B △ ) . (175) \operatorname{span}_{\mathbb{k}}\operatorname{ev}(\mathcal{B})=\operatorname{span}_{\mathbb{k}}\operatorname{ev}(\mathcal{B}^{\triangle}). \tag*{(175)} span k ev ( B ) = span k ev ( B △ ) . (175) Proof. Since B △ ⊆ B \mathcal{B}^{\triangle}\subseteq\mathcal{B} B △ ⊆ B , one inclusion is immediate. For the reverse inclusion we first analyse the two local replacements used in triangular straightening. We show that each local replacement sends a bad input word to a k \mathbb{k} k -linear combination of bad words. To prove that badness is preserved, we argue contrapositively: if an output word occurring with nonzero coefficient is good, then the input word is good.
Different colors. Let i ≠ j i\ne j i = j , a , b ≥ 1 a,b\ge1 a , b ≥ 1 , η ∈ X \eta\in X η ∈ X . Then
E i ( a ) F j ( b ) 1 η = F j ( b ) 1 η + a α i E i ( a ) , (176) E_i^{(a)}F_j^{(b)}1_\eta=F_j^{(b)}1_{\eta+a\alpha_i}E_i^{(a)}, \tag*{(176)} E i ( a ) F j ( b ) 1 η = F j ( b ) 1 η + a α i E i ( a ) , (176) so the input word ( ( 1 , i , a ) , ( − 1 , j , b ) ; η ) ((1,i,a),(-1,j,b);\eta) (( 1 , i , a ) , ( − 1 , j , b ) ; η ) , with intermediate weight η − b α j \eta-b\alpha_j η − b α j , is replaced by the single word ( ( − 1 , j , b ) , ( 1 , i , a ) ; η ) ((-1,j,b),(1,i,a);\eta) (( − 1 , j , b ) , ( 1 , i , a ) ; η ) , with intermediate weight η + a α i \eta+a\alpha_i η + a α i . Assume the output word is good. Then ℓ i ∣ a \ell_i\mid a ℓ i ∣ a , ℓ j ∣ b \ell_j\mid b ℓ j ∣ b and η ∈ X ∙ \eta\in X^\bullet η ∈ X ∙ , so lemma A.1(c) gives b α j ∈ L ∙ b\alpha_j\in L^\bullet b α j ∈ L ∙ and hence, by lemma A.1(a),
η − b α j ∈ η + L ∙ ⊆ X ∙ . \eta-b\alpha_j\in\eta+L^\bullet\subseteq X^\bullet. η − b α j ∈ η + L ∙ ⊆ X ∙ . Both input letters are block, its intermediate weight lies in X ∙ X^\bullet X ∙ , and the outer weights and outside data are unchanged; hence the input word is good.
Equal colors. Let i ∈ I i\in I i ∈ I , a , b ≥ 1 a,b\ge1 a , b ≥ 1 , η ∈ X \eta\in X η ∈ X . Specializing Lusztig’s modified divided-power identity [21 §23.1.3] gives
E i ( a ) F i ( b ) 1 η = ∑ t = 0 min ( a , b ) [ a − b + ⟨ h i , η ⟩ t ] q k , i F i ( b − t ) 1 η + ( a − t ) α i E i ( a − t ) 1 η . (177) E_i^{(a)}F_i^{(b)}1_\eta
=\sum_{t=0}^{\min(a,b)}
\begin{bmatrix}
a-b+\langle h_i,\eta\rangle\\
t
\end{bmatrix}_{q_{k,i}}
F_i^{(b-t)}1_{\eta+(a-t)\alpha_i}E_i^{(a-t)}1_\eta. \tag*{(177)} E i ( a ) F i ( b ) 1 η = t = 0 ∑ m i n ( a , b ) [ a − b + ⟨ h i , η ⟩ t ] q k , i F i ( b − t ) 1 η + ( a − t ) α i E i ( a − t ) 1 η . (177) In every summand − ( b − t ) α i + ( a − t ) α i = ( a − b ) α i -(b-t)\alpha_i+(a-t)\alpha_i=(a-b)\alpha_i − ( b − t ) α i + ( a − t ) α i = ( a − b ) α i , so the two outer weights are preserved. Set E i ( 0 ) = F i ( 0 ) = 1 E_i^{(0)}=F_i^{(0)}=1 E i ( 0 ) = F i ( 0 ) = 1 and delete every zero-exponent factor from the corresponding word; if t = a t=a t = a , resp. t = b t=b t = b , the deleted intermediate weight η + ( a − t ) α i \eta+(a-t)\alpha_i η + ( a − t ) α i equals the right, resp. the left, outer weight. Hence every summand is again the evaluation of a word in M \mathcal{M} M .
Fix a summand with nonzero coefficient and suppose its word is good. Write ℓ : = ℓ i \ell:=\ell_i ℓ := ℓ i . Goodness gives
ℓ ∣ a − t , ℓ ∣ b − t , η ∈ X ∙ , \ell\mid a-t,\qquad\ell\mid b-t,\qquad\eta\in X^\bullet, ℓ ∣ a − t , ℓ ∣ b − t , η ∈ X ∙ , the first two conditions holding trivially when the corresponding factor was deleted, since then the exponent is 0 0 0 . Therefore
a − b = ( a − t ) − ( b − t ) ∈ ℓ Z , ⟨ h i , η ⟩ ∈ ℓ Z , a − b + ⟨ h i , η ⟩ ∈ ℓ Z , a-b=(a-t)-(b-t)\in\ell\mathbb{Z},\qquad
\langle h_i,\eta\rangle\in\ell\mathbb{Z},\qquad
a-b+\langle h_i,\eta\rangle\in\ell\mathbb{Z}, a − b = ( a − t ) − ( b − t ) ∈ ℓ Z , ⟨ h i , η ⟩ ∈ ℓ Z , a − b + ⟨ h i , η ⟩ ∈ ℓ Z , and lemma A.2, applied with ξ = q k , i \xi=q_{k,i} ξ = q k , i and e = ℓ = ord ( q k , i 2 ) e=\ell=\operatorname{ord}(q_{k,i}^{2}) e = ℓ = ord ( q k , i 2 ) , forces ℓ ∣ t \ell\mid t ℓ ∣ t because the coefficient is nonzero. Consequently
ℓ ∣ a , ℓ ∣ b , η − b α i ∈ η + L ∙ ⊆ X ∙ , \ell\mid a,\qquad\ell\mid b,\qquad
\eta-b\alpha_i\in\eta+L^\bullet\subseteq X^\bullet, ℓ ∣ a , ℓ ∣ b , η − b α i ∈ η + L ∙ ⊆ X ∙ , so both input letters are block and the input intermediate weight lies in X ∙ X^\bullet X ∙ . As before the outer weights and all outside data agree with those of the good output; hence the input word is good.
By contraposition, in both cases every nonzero output word obtained from a bad input word is bad.
Global induction. For w = ( x 1 , … , x m ; η ) ∈ M \mathbf{w}=(x_1,\ldots,x_m;\eta)\in\mathcal{M} w = ( x 1 , … , x m ; η ) ∈ M , put
inv ( w ) : = # { ( r , s ) ∣ r < s , x r ∈ S 1 , x s ∈ S − 1 } . \operatorname{inv}(\mathbf{w}):=\#\{(r,s)\mid r<s,\ x_r\in\mathcal{S}^{1},\ x_s\in\mathcal{S}^{-1}\}. inv ( w ) := # {( r , s ) ∣ r < s , x r ∈ S 1 , x s ∈ S − 1 } .
We prove by induction on inv ( w ) \operatorname{inv}(\mathbf{w}) inv ( w ) that
w ∈ B ⟹ ev ( w ) ∈ span k ev ( B △ ) . \mathbf{w} \in\mathcal{B} \quad\Longrightarrow\quad\operatorname{ev}(\mathbf{w}) \in\operatorname{span}_{k} \operatorname{ev}(\mathcal{B}^{\triangle}). w ∈ B ⟹ ev ( w ) ∈ span k ev ( B △ ) .
If inv ( w ) = 0 \operatorname{inv}(\mathbf{w}) = 0 inv ( w ) = 0 , no positive letter precedes a negative letter. Hence all negative letters precede all positive letters, and therefore w ∈ B △ \mathbf{w} \in\mathcal{B}^{\triangle} w ∈ B △ .
Now suppose that
N : = inv ( w ) > 0 N := \operatorname{inv}(\mathbf{w}) > 0 N := inv ( w ) > 0
and that the assertion holds for all bad words of inversion number strictly smaller than N N N . The sign sequence of w \mathbf{w} w contains an adjacent pair
x u ∈ S 1 , x u + 1 ∈ S − 1 . x_u \in\mathcal{S}^{1}, \qquad x_{u+1} \in\mathcal{S}^{-1}. x u ∈ S 1 , x u + 1 ∈ S − 1 .
Apply the appropriate local replacement above. After omitting all terms with zero coefficient, this gives
ev ( w ) = ∑ w ′ c w ′ ev ( w ′ ) , c w ′ ∈ k × . \operatorname{ev}(\mathbf{w}) = \sum_{\mathbf{w}'} c_{\mathbf{w}'}\operatorname{ev}(\mathbf{w}'), \qquad c_{\mathbf{w}'} \in k^{\times}. ev ( w ) = w ′ ∑ c w ′ ev ( w ′ ) , c w ′ ∈ k × .
By the local analysis above, every output word w ′ \mathbf{w}' w ′ occurring in this sum is bad.
It remains to compare inversion numbers. If both letters of the chosen pair survive, their sign pattern changes from
( + , − ) to ( − , + ) , (+, -) \quad\text{to}\quad(-,+), ( + , − ) to ( − , + ) ,
so the inversion formed by this pair disappears. The remaining two possible outside signs contribute no change of the inversion. Consequently,
inv ( w ′ ) = inv ( w ) − 1 \operatorname{inv}(\mathbf{w}') = \operatorname{inv}(\mathbf{w}) - 1 inv ( w ′ ) = inv ( w ) − 1
whenever both letters survive.
If at least one letter of the pair is deleted, the chosen inversion again disappears. Moreover, deleting a letter can only remove inversions involving that letter, and the relative order of all remaining letters is unchanged. Hence in this case
inv ( w ′ ) ≤ inv ( w ) − 1. \operatorname{inv}(\mathbf{w}') \le\operatorname{inv}(\mathbf{w}) - 1. inv ( w ′ ) ≤ inv ( w ) − 1.
Thus every output word in the preceding sum satisfies
w ′ ∈ B , inv ( w ′ ) < N . \mathbf{w}' \in\mathcal{B}, \qquad\operatorname{inv}(\mathbf{w}') < N. w ′ ∈ B , inv ( w ′ ) < N .
The induction hypothesis therefore gives
ev ( w ′ ) ∈ span k ev ( B △ ) \operatorname{ev}(\mathbf{w}') \in\operatorname{span}_{k} \operatorname{ev}(\mathcal{B}^{\triangle}) ev ( w ′ ) ∈ span k ev ( B △ )
for every such w ′ \mathbf{w}' w ′ . Substituting these expressions into the straightening formula yields
ev ( w ) ∈ span k ev ( B △ ) . \operatorname{ev}(\mathbf{w}) \in\operatorname{span}_{k} \operatorname{ev}(\mathcal{B}^{\triangle}). ev ( w ) ∈ span k ev ( B △ ) .
This proves
span k ev ( B ) ⊆ span k ev ( B △ ) . \operatorname{span}_{k} \operatorname{ev}(\mathcal{B}) \subseteq\operatorname{span}_{k} \operatorname{ev}(\mathcal{B}^{\triangle}). span k ev ( B ) ⊆ span k ev ( B △ ) .
The reverse inclusion follows from B △ ⊆ B \mathcal{B}^{\triangle} \subseteq\mathcal{B} B △ ⊆ B , and the lemma follows.
Proposition A.5 (Half-ideal decomposition of J ~ \widetilde{\mathfrak{J}} J ).
J ~ = ⨁ λ ∈ X ∖ X ∙ V λ − 1 V + ⊕ ⨁ λ ∈ X ∙ ( J ~ λ − 1 V + + V λ − 1 J + ) . (178) \widetilde{\mathfrak{J}}=\bigoplus_{\lambda\in X\setminus X^\bullet}V^{-1}_{\lambda}V^+\mathbin{\oplus}\bigoplus_{\lambda\in X^\bullet}\left(\widetilde{\mathfrak{J}}^{-1}_{\lambda}V^++V^{-1}_{\lambda}\mathfrak{J}^+\right). \tag*{(178)} J = λ ∈ X ∖ X ∙ ⨁ V λ − 1 V + ⊕ λ ∈ X ∙ ⨁ ( J λ − 1 V + + V λ − 1 J + ) . (178) J ~ = ⨁ ν ∈ ⨁ i ∈ I Z α i ( J ~ ∩ ( V ˙ ) ν ) , (179) \widetilde{\mathfrak{J}}=\bigoplus_{\nu\in\bigoplus_{i\in I}\mathbb{Z}\alpha_i}\left(\widetilde{\mathfrak{J}}\cap(\dot{V})_\nu\right), \tag*{(179)} J = ν ∈ ⨁ i ∈ I Z α i ⨁ ( J ∩ ( V ˙ ) ν ) , (179) and, as root-graded k k k -vector spaces,
V ‾ / J ~ ≅ ⨁ λ ∈ X ∙ ( V − / J ~ − ) ⊗ k k 1 λ ⊗ k ( V + / J + ) . (180) \overline{V}/\widetilde{\mathfrak{J}}\cong\bigoplus_{\lambda\in X^\bullet}\left(V^-/\widetilde{\mathfrak{J}}^-\right)\otimes_k k1_\lambda\otimes_k\left(V^+/\mathfrak{J}^+\right). \tag*{(180)} V / J ≅ λ ∈ X ∙ ⨁ ( V − / J − ) ⊗ k k 1 λ ⊗ k ( V + / J + ) . (180)
Proof. By lemma A.4, J ~ = span k ev ( B △ ) \widetilde{\mathfrak{J}}=\operatorname{span}_k\operatorname{ev}(\mathcal{B}^{\triangle}) J = span k ev ( B △ ) .
Left side contained in right side. Let w = ( x 1 , … , x m ; η ) ∈ B △ \mathbf{w}=(x_1,\ldots,x_m;\eta)\in\mathcal{B}^{\triangle} w = ( x 1 , … , x m ; η ) ∈ B △ with splitting index p p p , and put λ : = λ p ( w ) \lambda:=\lambda_p(\mathbf{w}) λ := λ p ( w ) . By (A.12),
ev ( w ) = ( ev ( x 1 ) ⋯ ev ( x p ) ) 1 λ ( ev ( x p + 1 ) ⋯ ev ( x m ) 1 η ) ∈ V λ − 1 V + . \operatorname{ev}(\mathbf{w})=\left(\operatorname{ev}(x_1)\cdots\operatorname{ev}(x_p)\right)1_\lambda\left(\operatorname{ev}(x_{p+1})\cdots\operatorname{ev}(x_m)1_\eta\right)\in V^{-1}_{\lambda}V^+. ev ( w ) = ( ev ( x 1 ) ⋯ ev ( x p ) ) 1 λ ( ev ( x p + 1 ) ⋯ ev ( x m ) 1 η ) ∈ V λ − 1 V + .
If λ ∉ X ∙ \lambda\notin X^\bullet λ ∈ / X ∙ , this lies in the first sum of (A.23). Assume λ ∈ X ∙ \lambda\in X^\bullet λ ∈ X ∙ . If every letter of w \mathbf{w} w were block, then (A.9) would give deg ( x j ) ∈ L ∙ \deg(x_j)\in L^\bullet deg ( x j ) ∈ L ∙ for all j j j , hence
λ j ( w ) ∈ λ + L ∙ ⊆ X ∙ ( 0 ≤ j ≤ m ) \lambda_j(\mathbf{w})\in\lambda+L^\bullet\subseteq X^\bullet\qquad(0\leq j\leq m) λ j ( w ) ∈ λ + L ∙ ⊆ X ∙ ( 0 ≤ j ≤ m )
by lemma A.1(a), contradicting w ∈ B \mathbf{w}\in\mathcal{B} w ∈ B . So some x j x_j x j is nonblock. If j ≤ p j\leq p j ≤ p , then ( x 1 , … , x p ) ∈ W − 1 (x_1,\ldots,x_p)\in\mathcal{W}^{-1} ( x 1 , … , x p ) ∈ W − 1 and lemma A.3 places ev ( w ) \operatorname{ev}(\mathbf{w}) ev ( w ) in J ~ λ − 1 V + \widetilde{\mathfrak{J}}^{-1}_{\lambda}V^+ J λ − 1 V + ; if j > p j>p j > p , then ( x p + 1 , … , x m ) ∈ W 1 (x_{p+1},\ldots,x_m)\in\mathcal{W}^{1} ( x p + 1 , … , x m ) ∈ W 1 and ev ( w ) ∈ V λ − 1 J + \operatorname{ev}(\mathbf{w})\in V^{-1}_{\lambda}\mathfrak{J}^+ ev ( w ) ∈ V λ − 1 J + .
Right side contained in left side. Let λ ∈ X ∖ X ∙ \lambda\in X\setminus X^\bullet λ ∈ X ∖ X ∙ . Expanding both halves of V λ − 1 V + V^{-1}_{\lambda}V^+ V λ − 1 V + in evaluated words gives k k k -linear combinations of ev ( w ) \operatorname{ev}(\mathbf{w}) ev ( w ) with w ∈ B △ \mathbf{w}\in\mathcal{B}^{\triangle} w ∈ B △ and λ p ( w ) = λ ∉ X ∙ \lambda_p(\mathbf{w})=\lambda\notin X^\bullet λ p ( w ) = λ ∈ / X ∙ ; hence V λ − 1 V + ⊆ J ~ V^{-1}_{\lambda}V^+\subseteq\widetilde{\mathfrak{J}} V λ − 1 V + ⊆ J . Let λ ∈ X ∙ \lambda\in X^\bullet λ ∈ X ∙ . By lemma A.3, J ~ − \widetilde{\mathfrak{J}}^- J − is spanned by ev ( W − 1 ) \operatorname{ev}(\mathcal{W}^{-1}) ev ( W − 1 ) and V + V^+ V + by evaluated positive words, so J ~ λ − 1 V + \widetilde{\mathfrak{J}}^{-1}_{\lambda}V^+ J λ − 1 V + is spanned by ev ( w ) \operatorname{ev}(\mathbf{w}) ev ( w ) with w ∈ B △ \mathbf{w}\in\mathcal{B}^{\triangle} w ∈ B △ containing a nonblock letter; similarly for V λ − 1 J + V^{-1}_{\lambda}\mathfrak{J}^+ V λ − 1 J + . Both are therefore contained in J ~ \widetilde{\mathfrak{J}} J .
Directness, grading, quotient. The summands of (A.23) lie in pairwise distinct blocks V λ − 1 V + V^{-1}_{\lambda}V^+ V λ − 1 V + of (A.5), so the sum is direct. Each summand is root graded, the blocks by (A.5) and J ± \mathfrak{J}^{\pm} J ± by (A.16); this gives (A.24). Finally, fix λ ∈ X ∙ \lambda\in X^\bullet λ ∈ X ∙ and let
p ± : V ± ↠ V ± / J ± p^\pm:V^\pm\twoheadrightarrow V^\pm/\mathfrak{J}^\pm p ± : V ± ↠ V ± / J ±
be the quotient maps. Since tensoring over the field k k k is exact, the standard quotient–tensor isomorphism gives
V − ⊗ k k 1 λ ⊗ k V + J ~ − ⊗ k k 1 λ ⊗ k V + + V − ⊗ k k 1 λ ⊗ k J + ≅ ( V − / J ~ − ) ⊗ k k 1 λ ⊗ k ( V + / J + ) . \frac{V^-\otimes_k k1_\lambda\otimes_k V^+}{\widetilde{\mathfrak{J}}^-\otimes_k k1_\lambda\otimes_k V^++V^-\otimes_k k1_\lambda\otimes_k\mathfrak{J}^+}\cong(V^-/\widetilde{\mathfrak{J}}^-)\otimes_k k1_\lambda\otimes_k(V^+/\mathfrak{J}^+). J − ⊗ k k 1 λ ⊗ k V + + V − ⊗ k k 1 λ ⊗ k J + V − ⊗ k k 1 λ ⊗ k V + ≅ ( V − / J − ) ⊗ k k 1 λ ⊗ k ( V + / J + ) .
Equivalently,
ker ( p − ⊗ id ⊗ p + ) = J ~ − ⊗ k k 1 λ ⊗ k V + + V − ⊗ k k 1 λ ⊗ k J + . \ker(p^-\otimes\operatorname{id}\otimes p^+)=\widetilde{\mathfrak{J}}^-\otimes_k k1_\lambda\otimes_k V^++V^-\otimes_k k1_\lambda\otimes_k\mathfrak{J}^+. ker ( p − ⊗ id ⊗ p + ) = J − ⊗ k k 1 λ ⊗ k V + + V − ⊗ k k 1 λ ⊗ k J + .
Transporting this kernel through the multiplication isomorphism in (A.5) yields a root-graded isomorphism
V λ − 1 V + J ~ λ − 1 V + + V λ − 1 J + ≅ ( V − / J ~ − ) ⊗ k k 1 λ ⊗ k ( V + / J + ) . \frac{V^{-1}_{\lambda}V^+}{\widetilde{\mathfrak{J}}^{-1}_{\lambda}V^++V^{-1}_{\lambda}\mathfrak{J}^+}\cong(V^-/\widetilde{\mathfrak{J}}^-)\otimes_k k1_\lambda\otimes_k(V^+/\mathfrak{J}^+). J λ − 1 V + + V λ − 1 J + V λ − 1 V + ≅ ( V − / J − ) ⊗ k k 1 λ ⊗ k ( V + / J + ) .
For λ ∉ X ∙ \lambda\notin X^\bullet λ ∈ / X ∙ , (A.23) gives
V λ − 1 V + ⊆ J ~ , (181) V^{-1}_{\lambda}V^+\subseteq\widetilde{\mathfrak{J}}, \tag*{(181)} V λ − 1 V + ⊆ J , (181)
so the entire λ \lambda λ -block vanishes in V ˙ / J \dot{V}/\mathfrak{J} V ˙ / J . Since the triangular decomposition in (A.5) is an internal direct sum, taking the direct sum of the surviving blocks over λ ∈ X ∘ \lambda\in X^\circ λ ∈ X ∘ gives
V ˙ / J ≅ ⨁ λ ∈ X ∘ ( V − / J − ) ⊗ k k 1 λ ⊗ k ( V + / J + ) . \dot{V}/\mathfrak{J} \cong\bigoplus_{\lambda\in X^\circ} (V^-/\mathfrak{J}^-) \otimes_k k1_\lambda\otimes_k (V^+/\mathfrak{J}^+). V ˙ / J ≅ λ ∈ X ∘ ⨁ ( V − / J − ) ⊗ k k 1 λ ⊗ k ( V + / J + ) .
which is (A.25).
Proof of the quotient theorem
Proof of theorem 10.3. Two-sidedness. Let
u = ( u 1 , … , u p ; θ ) ∈ M , w = ( x 1 , … , x m ; η ) ∈ B . \mathbf{u}=(u_1,\ldots,u_p;\theta)\in\mathcal{M}, \qquad\mathbf{w}=(x_1,\ldots,x_m;\eta)\in\mathcal{B}. u = ( u 1 , … , u p ; θ ) ∈ M , w = ( x 1 , … , x m ; η ) ∈ B .
Since the 1 λ 1_\lambda 1 λ are orthogonal idempotents,
ev ( u ) ev ( w ) = { ev ( u 1 , … , u p , x 1 , … , x m ; η ) , θ = λ 0 ( w ) , 0 , θ ≠ λ 0 ( w ) , \operatorname{ev}(\mathbf{u})\operatorname{ev}(\mathbf{w})=
\begin{cases}
\operatorname{ev}(u_1,\ldots,u_p,x_1,\ldots,x_m;\eta), & \theta=\lambda_0(\mathbf{w}),\\
0, & \theta\ne\lambda_0(\mathbf{w}),
\end{cases} ev ( u ) ev ( w ) = { ev ( u 1 , … , u p , x 1 , … , x m ; η ) , 0 , θ = λ 0 ( w ) , θ = λ 0 ( w ) ,
and
ev ( w ) ev ( u ) = { ev ( x 1 , … , x m , u 1 , … , u p ; θ ) , η = λ 0 ( u ) , 0 , η ≠ λ 0 ( u ) . \operatorname{ev}(\mathbf{w})\operatorname{ev}(\mathbf{u})=
\begin{cases}
\operatorname{ev}(x_1,\ldots,x_m,u_1,\ldots,u_p;\theta), & \eta=\lambda_0(\mathbf{u}),\\
0, & \eta\ne\lambda_0(\mathbf{u}).
\end{cases} ev ( w ) ev ( u ) = { ev ( x 1 , … , x m , u 1 , … , u p ; θ ) , 0 , η = λ 0 ( u ) , η = λ 0 ( u ) .
Suppose first that the product on the left is nonzero, and denote the corresponding concatenated word by v \mathbf{v} v . Then
λ p + j ( v ) = λ j ( w ) ( 0 ≤ j ≤ m ) . \lambda_{p+j}(\mathbf{v})=\lambda_j(\mathbf{w}) \qquad(0\le j\le m). λ p + j ( v ) = λ j ( w ) ( 0 ≤ j ≤ m ) .
Thus every letter x j x_j x j of w \mathbf{w} w , and every boundary weight λ j ( w ) \lambda_j(\mathbf{w}) λ j ( w ) , occurs unchanged in v \mathbf{v} v . Hence a nonblock letter or a forbidden boundary weight witnessing w ∈ B \mathbf{w}\in\mathcal{B} w ∈ B also witnesses v ∈ B \mathbf{v}\in\mathcal{B} v ∈ B . Likewise, if the product on the right is nonzero, then its concatenated word v \mathbf{v} v satisfies
λ j ( v ) = λ j ( w ) ( 0 ≤ j ≤ m ) , \lambda_j(\mathbf{v})=\lambda_j(\mathbf{w}) \qquad(0\le j\le m), λ j ( v ) = λ j ( w ) ( 0 ≤ j ≤ m ) ,
and again all letters and boundary weights of w \mathbf{w} w occur unchanged in v \mathbf{v} v . Therefore v ∈ B \mathbf{v}\in\mathcal{B} v ∈ B in this case as well. Consequently,
ev ( M ) ev ( B ) ⊆ { 0 } ∪ ev ( B ) , ev ( B ) ev ( M ) ⊆ { 0 } ∪ ev ( B ) . \operatorname{ev}(\mathcal{M})\operatorname{ev}(\mathcal{B})\subseteq\{0\}\cup\operatorname{ev}(\mathcal{B}), \qquad\operatorname{ev}(\mathcal{B})\operatorname{ev}(\mathcal{M})\subseteq\{0\}\cup\operatorname{ev}(\mathcal{B}). ev ( M ) ev ( B ) ⊆ { 0 } ∪ ev ( B ) , ev ( B ) ev ( M ) ⊆ { 0 } ∪ ev ( B ) .
Since
V ˙ = span k ev ( M ) , J = span k ev ( B ) , \dot{V}=\operatorname{span}_k\operatorname{ev}(\mathcal{M}), \qquad\mathfrak{J}=\operatorname{span}_k\operatorname{ev}(\mathcal{B}), V ˙ = span k ev ( M ) , J = span k ev ( B ) ,
by (A.13) and the definition of J \mathfrak{J} J , k k k -bilinearity gives
V ˙ J ⊆ J , J V ˙ ⊆ J . \dot{V}\mathfrak{J}\subseteq\mathfrak{J}, \qquad\mathfrak{J}\dot{V}\subseteq\mathfrak{J}. V ˙ J ⊆ J , J V ˙ ⊆ J .
Thus J \mathfrak{J} J is a two-sided ideal of V ˙ \dot{V} V ˙ . Its root-gradedness was established in proposition A.5; hence Q = V ˙ / J Q=\dot{V}/\mathfrak{J} Q = V ˙ / J is a root-graded k k k -algebra.
Support. Let
V ‾ ± : = V ± / J ± , L + ∙ : = ⨁ i ∈ I N α i . \overline{V}^{\pm}:=V^{\pm}/\mathfrak{J}^{\pm}, \qquad L_+^\bullet:=\bigoplus_{i\in I}\mathbb{N}\alpha_i. V ± := V ± / J ± , L + ∙ := i ∈ I ⨁ N α i .
By lemma A.3, V ‾ + \overline{V}^{+} V + is spanned by the images of evaluated positive words containing no nonblock letter. For a letter x = ( 1 , i , a ) ∈ S 1 ∖ S n b x=(1,i,a)\in\mathcal{S}^{1}\setminus\mathcal{S}^{\mathrm{nb}} x = ( 1 , i , a ) ∈ S 1 ∖ S nb , one has ℓ i ∣ a \ell_i\mid a ℓ i ∣ a , and therefore
deg ( x ) = a α i = a ℓ i α i ∈ N α i . \deg(x)=a\alpha_i=\frac{a}{\ell_i}\alpha_i\in\mathbb{N}\alpha_i. deg ( x ) = a α i = ℓ i a α i ∈ N α i .
It follows that
Supp ( V + ‾ ) ⊆ L + ∙ . \operatorname{Supp}(\overline{V^+}) \subseteq L_+^\bullet. Supp ( V + ) ⊆ L + ∙ .
The same argument for negative words gives
Supp ( V − ‾ ) ⊆ − L + ∙ . \operatorname{Supp}(\overline{V^-}) \subseteq-L_+^\bullet. Supp ( V − ) ⊆ − L + ∙ .
By (A.25),
Q ≅ ⨁ λ ∈ X ∙ V − ‾ ⊗ k k 1 λ ⊗ k V + ‾ Q \cong\bigoplus_{\lambda\in X^\bullet} \overline{V^-}\otimes_{\mathbb{k}} \mathbb{k}1_\lambda\otimes_{\mathbb{k}}\overline{V^+} Q ≅ λ ∈ X ∙ ⨁ V − ⊗ k k 1 λ ⊗ k V +
as root-graded k \mathbb{k} k -vector spaces and more explicitly,
Q ν ≅ ⨁ λ ∈ X ∙ ⨁ β , γ ∈ L + ∙ γ − β = ν ( V − ‾ ) − β ⊗ k k 1 λ ⊗ k ( V + ‾ ) γ . Q_\nu\cong\bigoplus_{\lambda\in X^\bullet}\bigoplus_{\substack{\beta,\gamma\in L_+^\bullet\\\gamma-\beta=\nu}}(\overline{V^-})_{-\beta}\otimes_{\mathbb{k}}\mathbb{k}1_\lambda\otimes_{\mathbb{k}}(\overline{V^+})_\gamma. Q ν ≅ λ ∈ X ∙ ⨁ β , γ ∈ L + ∙ γ − β = ν ⨁ ( V − ) − β ⊗ k k 1 λ ⊗ k ( V + ) γ .
Since every 1 λ 1_\lambda 1 λ has root degree 0 0 0 , a nonzero homogeneous component ( Q ) ν (Q)_\nu ( Q ) ν can occur only if
ν ∈ − L + ∙ + L + ∙ ⊆ L ∙ . \nu\in-L_+^\bullet+L_+^\bullet\subseteq L^\bullet. ν ∈ − L + ∙ + L + ∙ ⊆ L ∙ .
Hence
{ ν ∣ ( Q ) ν ≠ 0 } ⊆ L ∙ . \{\nu\mid(Q)_\nu\ne0\}\subseteq L^\bullet. { ν ∣ ( Q ) ν = 0 } ⊆ L ∙ .
Survival of E 010 E_{0 10} E 010 . By lemma (A.1)(b),
α 0 = α 0 ∙ ∈ L ∙ ⊆ X ∙ , \alpha_0=\alpha_0^\bullet\in L^\bullet\subseteq X^\bullet, α 0 = α 0 ∙ ∈ L ∙ ⊆ X ∙ ,
while 0 ∈ X ∙ 0\in X^\bullet 0 ∈ X ∙ because ⟨ h i , 0 ⟩ = 0 \langle h_i,0\rangle=0 ⟨ h i , 0 ⟩ = 0 for every i ∈ I i\in I i ∈ I .
By lemma A.3, J − \mathcal{J}^- J − is spanned by evaluations of words in W − 1 \mathcal{W}^{-1} W − 1 . Every such word is nonempty and has strictly negative root degree, so
( J − ) 0 = 0. (\mathcal{J}^-)_{0}=0. ( J − ) 0 = 0.
Similarly, every word in W 1 \mathcal{W}^{1} W 1 contains a nonblock letter ( 1 , i , a ) (1,i,a) ( 1 , i , a ) . By (A.9), its color satisfies i ∈ I ∘ i\in I^\circ i ∈ I ∘ , and hence i ≠ 0 i\ne0 i = 0 by lemma A.1(b). Since all letters of the word have positive degree, its degree has a strictly positive α i \alpha_i α i -coefficient and therefore cannot equal α 0 \alpha_0 α 0 . Consequently,
( J + ) α 0 = 0. (\mathcal{J}^+)_{\alpha_0}=0. ( J + ) α 0 = 0.
Together with (A.6), these equalities give
1 + J − ≠ 0 in V − ‾ , E 0 + J + ≠ 0 in V + ‾ . 1+\mathcal{J}^-\ne0\quad\text{in }\overline{V^-},\qquad E_0+\mathcal{J}^+\ne0\quad\text{in }\overline{V^+}. 1 + J − = 0 in V − , E 0 + J + = 0 in V + .
In the λ = α 0 \lambda=\alpha_0 λ = α 0 summand of (A.25), the multiplication isomorphism uses
1 α 0 E 0 = E 0 1 0 . 1_{\alpha_0}E_0=E_0 1_0. 1 α 0 E 0 = E 0 1 0 .
Therefore
E 0 1 0 + J ⟷ ( 1 + J − ) ⊗ 1 α 0 ⊗ ( E 0 + J + ) . E_0 1_0+\mathcal{J}\longleftrightarrow(1+\mathcal{J}^-)\otimes1_{\alpha_0}\otimes(E_0+\mathcal{J}^+). E 0 1 0 + J ⟷ ( 1 + J − ) ⊗ 1 α 0 ⊗ ( E 0 + J + ) .
All three factors on the right are nonzero, so their pure tensor is nonzero over the field k \mathbb{k} k . Hence
E 0 1 0 + J ∈ ( Q ) α 0 ∖ { 0 } , E_0 1_0+\mathcal{J}\in(Q)_{\alpha_0}\setminus\{0\}, E 0 1 0 + J ∈ ( Q ) α 0 ∖ { 0 } ,
as required. □
Acknowledgments
The project is partially supported by the Simons Foundation grant no. MP-TSM-00002518.
Data Availability Statement. All data of this work are included in the manuscript and its supplemental materials.
Ziyang Jiang, Department of Mathematics, North Carolina State University, Raleigh, NC 27695, USA Email address: zjiang29@ncsu.edu
Naihuan Jing, Department of Mathematics, North Carolina State University, Raleigh, NC 27695, USA Email address: jing@ncsu.edu