Introduction
The Korteweg–de Vries equation u t = u x x x + 6 u u x u_t=u_{xxx}+6uu_x u t = u xxx + 6 u u x admits an infinite family of conserved quantities that commute under the KdV Poisson bracket. Their Hamiltonian flows form the KdV hierarchy [11 , 12 ] . The first densities are
u , u 2 , u 3 − 1 2 u x 2 . u,\qquad u^2,\qquad u^3-\frac{1}{2}u_x^2. u , u 2 , u 3 − 2 1 u x 2 .
Their integrals are constant for smooth periodic or rapidly decaying solutions. The corresponding algebraic statement is a local conservation law: the time derivative of a density is a total spatial derivative. Thus the equation and its conservation laws can be formulated entirely in the algebra of differential polynomials P 0 : = Q [ u , u − 1 , u − 2 , … ] \mathcal{P}_0 := \mathbb{Q}[u,u_{-1},u_{-2},\ldots] P 0 := Q [ u , u − 1 , u − 2 , … ] , where u − j : = ∂ x j u u_{-j}:=\partial_x^j u u − j := ∂ x j u and the total derivative satisfies ∂ u − j = u − j − 1 \partial u_{-j}=u_{-j-1} ∂ u − j = u − j − 1 . The KdV evolution acts on differential polynomials by a derivation X 0 X_0 X 0 , determined by X 0 u = u − 3 + 6 u u − 1 X_0u=u_{-3}+6uu_{-1} X 0 u = u − 3 + 6 u u − 1 and X 0 ∂ = ∂ X 0 X_0\partial=\partial X_0 X 0 ∂ = ∂ X 0 . Explicitly,
X 0 : = ∑ j ≥ 0 ∂ j ( u − 3 + 6 u u − 1 ) ∂ u − j . X_0:=\sum_{j\geq0}\partial^j(u_{-3}+6uu_{-1})\partial_{u_{-j}}. X 0 := j ≥ 0 ∑ ∂ j ( u − 3 + 6 u u − 1 ) ∂ u − j .
The quotient P 0 / ∂ P 0 \mathcal{P}_0/\partial\mathcal{P}_0 P 0 / ∂ P 0 is the space of local functionals; it expresses integration by parts without choosing boundary conditions. For the second KdV Poisson bracket, X 0 X_0 X 0 is the Hamiltonian flow generated by [ u 2 ] [u^2] [ u 2 ] (§ 6.1 ). A class [ f ] [f] [ f ] is a conserved quantity for the flow X 0 X_0 X 0 precisely when
X 0 [ f ] = 0 , equivalently X 0 f ∈ ∂ P 0 . X_0[f]=0,\qquad\text{equivalently}\qquad X_0f\in\partial\mathcal{P}_0. X 0 [ f ] = 0 , equivalently X 0 f ∈ ∂ P 0 .
Pairwise commutativity is a further property of the chosen conserved quantities. The pseudodifferential formulation places KdV within the KP hierarchy [10 ] .
The Virasoro vertex operator algebra provides the quantum version of this problem [8 , 9 ] . Let V c V^c V c be the universal Virasoro VOA of central charge c c c , with conformal vector ω \omega ω and translation T T T . We use modes Y ( a , z ) = ∑ n ∈ Z a ( n ) z − n − 1 Y(a,z)=\sum_{n\in\mathbb{Z}}a(n)z^{-n-1} Y ( a , z ) = ∑ n ∈ Z a ( n ) z − n − 1 ; in particular,
Y ( ω , z ) = ∑ n ∈ Z L n z − n − 2 , ω ( n ) = L n − 1 . (1) Y(\omega,z)=\sum_{n\in\mathbb{Z}}L_nz^{-n-2},\qquad\omega(n)=L_{n-1}. \tag*{(1)} Y ( ω , z ) = n ∈ Z ∑ L n z − n − 2 , ω ( n ) = L n − 1 . (1)
The space V c / T V c V^c/TV^c V c / T V c is a Lie algebra with bracket
{ [ a ] , [ b ] } : = [ a ( 0 ) b ] . \{[a],[b]\}:=[a(0)b]. {[ a ] , [ b ]} := [ a ( 0 ) b ] .
Our result concerns this universal VOA at every central charge, including c = 0 c=0 c = 0 .
Main result (Theorem 5.3). There is an explicit formula for vectors H 2 k ( c ) ∈ V c H^{(c)}_{2k}\in V^c H 2 k ( c ) ∈ V c , of weight 2 k 2k 2 k and with highest PBW term ω ( − 1 ) k \omega(-1)^k ω ( − 1 ) k , such that
H 2 k ( c ) ( 0 ) H 2 l ( c ) ∈ T V c ( k , l ≥ 1 ) . H^{(c)}_{2k}(0)H^{(c)}_{2l}\in TV^c\qquad(k,l\geq1). H 2 k ( c ) ( 0 ) H 2 l ( c ) ∈ T V c ( k , l ≥ 1 ) . The kernel of ( ω ( − 1 ) ω ) ( 0 ) (\omega(-1)\omega)(0) ( ω ( − 1 ) ω ) ( 0 ) on V c / T V c V^c/TV^c V c / T V c is spanned by [ H 2 k ( c ) ] [H^{(c)}_{2k}] [ H 2 k ( c ) ] in each positive even weight and is zero in positive odd weights.
Thus the conserved classes for the weight-four action are unique up to scalar in each positive even weight; the displayed identity also asserts their pairwise commutativity. The formula is (55) , followed by the change of variables (87) ; the first vectors are displayed in Appendix § 8 .
Earlier explicit low-weight densities were obtained by Sasaki and Yamanaka [7 ] . Feigin and Frenkel establish existence and commutativity of quantum integrals of motion for generic parameter by a homological deformation argument [8 Section 4.7] . Bazhanov, Lukyanov, and Zamolodchikov construct quantum monodromy and commuting transfer operators, and relate their asymptotic expansions to local integrals [9 ] . They also display the first densities, of weights two, four, and six [9 equations (5)–(6)] . More recent explicit expressions, including weights eight and ten, appear in [5 Appendix D] . These approaches provide existence and spectral information; here the aim is a direct formula for the polynomial densities in fixed VOA coordinates. The proof works directly with identities in the universal Virasoro VOA, independently of transfer-operator asymptotics. Our local-functional bracket uses the zeroth products a ( 0 ) b a(0)b a ( 0 ) b in the mode convention above. Comparison with cylinder currents also requires the exponential change of coordinates. In Zhu’s square-bracket formalism [6 ] , this is expressed by
Y [ a , z ] : = Y ( e z L 0 a , e z − 1 ) , ω ~ : = ω − c 24 1. Y[a,z] := Y(e^{zL_0}a,e^z-1), \qquad\widetilde{\omega} := \omega-\frac{c}{24}\mathbf{1}. Y [ a , z ] := Y ( e z L 0 a , e z − 1 ) , ω := ω − 24 c 1 .
Thus comparison with the formulas of [9 , 5 ] requires converting both coordinates and normal-ordering conventions. We do not need this change of coordinates in the construction below.
For the construction, we introduce a parameter ℏ \hbar ℏ and pass to polynomial coordinates obtained by symmetrization. The quantum KdV action then becomes an operator X ℏ X_{\hbar} X ℏ deforming X 0 X_0 X 0 , and the conserved densities take the form
F k = ( 1 + M ) − 1 u k . (2) F_k=(1+M)^{-1}u^k. \tag*{(2)} F k = ( 1 + M ) − 1 u k . (2)
Here M M M acts nilpotently at each weight after reduction modulo total derivatives, so the inverse is a finite sum. Sections 2–5 give the precise setup; equation (87) returns to the original central-charge coordinates.
Four steps explain the construction and its proof.
(i) Symmetric PBW coordinates realize the Virasoro modes as possibly infinite-order differential operators, finite on each polynomial. Their coefficients have degree at most one in the coordinates.
(ii) The weight-four action has a part that can be inverted explicitly after reduction modulo total derivatives. The remaining correction strictly lowers a finite ordering on each weight space. This reduces the construction to inverting the identity plus a nilpotent operator, giving (2) .
(iii) The commuting classical third and fifth KdV flows define a Koszul-type complex. Its middle exactness, together with the explicit weight-six conservation identity, shows that the constructed F k F_k F k are conserved densities for X ℏ X_{\hbar} X ℏ .
(iv) The same inverse determines the kernel in each weight. Odd-weight vanishing and the Jacobi identity then imply pairwise commutativity.
The paper is organized as follows. § 2 develops the differential representation, and § 3 computes the weight-four action. § 4 constructs reduction and extraction; § 5 gives the formula and states the main theorem. § 6 proves classical exactness, and § 7 completes the conservation, uniqueness, and specialization arguments. The appendix records explicit elements and computational checks. Readers primarily interested in the formula may start with (55) , Theorem 5.3, and Appendix § 8 . The arguments for the differential realization, finite inverse, and quantum conservation and commutativity are given explicitly. We use standard PBW and VOA identities, with conventions as in [1 , 2 ] , and recall the classical Gelfand–Dickey construction where it enters the proof [10 ] .
Symmetric PBW coordinates and differential operators
Virasoro modes and symmetric PBW coordinates
Introduce a parameter ℏ \hbar ℏ and work over the polynomial ring
k : = Q [ ℏ ] . \mathsf{k} := \mathbb{Q}[\hbar]. k := Q [ ℏ ] .
The completion Q [ [ ℏ ] ] \mathbb{Q}[[\hbar]] Q [[ ℏ ]] is used only in the lifting argument of § 7 . After adjoining ℏ − 1 \hbar^{-1} ℏ − 1 , set U : = ℏ ω U := \hbar\omega U := ℏ ω and c : = 6 / ℏ c := 6/\hbar c := 6/ℏ ; this rescales the Virasoro relations so that all commutators vanish at ℏ = 0 \hbar= 0 ℏ = 0 while their first-order terms retain the classical bracket. Let V \mathcal{V} V be the universal vertex algebra over k \mathsf{k} k generated by U U U with singular products
U ( z ) U ( w ) ∼ ℏ ( 3 ( z − w ) 4 + 2 U ( w ) ( z − w ) 2 + ∂ w U ( w ) z − w ) . (3) U(z)U(w) \sim\hbar\left(\frac{3}{(z-w)^4}+\frac{2U(w)}{(z-w)^2}+\frac{\partial_w U(w)}{z-w}\right). \tag*{(3)} U ( z ) U ( w ) ∼ ℏ ( ( z − w ) 4 3 + ( z − w ) 2 2 U ( w ) + z − w ∂ w U ( w ) ) . (3)
Here U ( z ) : = Y ( U , z ) U(z) := Y(U,z) U ( z ) := Y ( U , z ) , and ∼ \sim ∼ denotes equality of singular parts. Equivalently, its modes satisfy the algebraic relation
[ U ( m ) , U ( n ) ] = ℏ ( m − n ) U ( m + n − 1 ) + ℏ 2 m ( m − 1 ) ( m − 2 ) δ m + n , 2 . (4) [U(m),U(n)] = \hbar(m-n)U(m+n-1) + \frac{\hbar}{2}m(m-1)(m-2)\delta_{m+n,2}. \tag*{(4)} [ U ( m ) , U ( n )] = ℏ ( m − n ) U ( m + n − 1 ) + 2 ℏ m ( m − 1 ) ( m − 2 ) δ m + n , 2 . (4)
The family V \mathcal{V} V is a graded vertex algebra over k \mathsf{k} k and has conformal vector U / ℏ U/\hbar U /ℏ after adjoining ℏ − 1 \hbar^{-1} ℏ − 1 . Its ℏ = 0 \hbar= 0 ℏ = 0 specialization is the classical differential algebra.
The mode convention (1) and the scaling U = ℏ ω U = \hbar\omega U = ℏ ω motivate the following indices: nonpositive indices will correspond directly to the differential polynomial generators u − j u_{-j} u − j . Define
U − j : = j ! U ( − j − 1 ) ( j ≥ 0 ) , U n : = U ( n − 1 ) ( n ≥ 1 ) , U n 1 = 0 ( n ≥ 1 ) . (5) U_{-j}:=j!U(-j-1)\quad(j\geq0),\qquad U_n:=U(n-1)\quad(n\geq1),\qquad U_n\mathbf{1}=0\quad(n\geq1). \tag*{(5)} U − j := j ! U ( − j − 1 ) ( j ≥ 0 ) , U n := U ( n − 1 ) ( n ≥ 1 ) , U n 1 = 0 ( n ≥ 1 ) . (5)
Thus U − j 1 = T j U U_{-j}\mathbf{1}=T^jU U − j 1 = T j U , matching u − j = ∂ j u u_{-j}=\partial^j u u − j = ∂ j u without additional factorials in the symmetrization map. Positive indices label annihilation modes. After localization,
U = ℏ ω , c = 6 / ℏ , U − j = j ! ℏ L − j − 2 ( j ≥ 0 ) , U n = ℏ L n − 2 ( n ≥ 1 ) , U=\hbar\omega,\qquad c=6/\hbar,\qquad U_{-j}=j!\hbar L_{-j-2}\quad(j\geq0),\qquad U_n=\hbar L_{n-2}\quad(n\geq1), U = ℏ ω , c = 6/ℏ , U − j = j ! ℏ L − j − 2 ( j ≥ 0 ) , U n = ℏ L n − 2 ( n ≥ 1 ) ,
[ L m , L n ] = ( m − n ) L m + n + c 12 ( m 3 − m ) δ m + n , 0 . (6) [L_m,L_n]=(m-n)L_{m+n}+\frac{c}{12}(m^3-m)\delta_{m+n,0}. \tag*{(6)} [ L m , L n ] = ( m − n ) L m + n + 12 c ( m 3 − m ) δ m + n , 0 . (6)
By (5) , U n = max ( 0 , − n ) ! U ( n − 1 ) U_n=\max(0,-n)!U(n-1) U n = max ( 0 , − n )! U ( n − 1 ) ; hence (4) , with indices m − 1 m-1 m − 1 , n − 1 n-1 n − 1 , gives for all m , n ∈ Z m,n\in\mathbb{Z} m , n ∈ Z
[ U m , U n ] = ℏ ( m − n ) max ( 0 , − m ) ! max ( 0 , − n ) ! max ( 0 , 2 − m − n ) ! U m + n − 2 + ℏ 2 max ( 0 , − m ) ! max ( 0 , − n ) ! ( ( m − 2 ) 3 − ( m − 2 ) ) δ m + n , 4 . [U_m,U_n]=\hbar(m-n)\frac{\max(0,-m)!\max(0,-n)!}{\max(0,2-m-n)!}U_{m+n-2}
+\frac{\hbar}{2}\max(0,-m)!\max(0,-n)!((m-2)^3-(m-2))\delta_{m+n,4}. [ U m , U n ] = ℏ ( m − n ) max ( 0 , 2 − m − n )! max ( 0 , − m )! max ( 0 , − n )! U m + n − 2 + 2 ℏ max ( 0 , − m )! max ( 0 , − n )! (( m − 2 ) 3 − ( m − 2 )) δ m + n , 4 .
Here max ( 0 , − n ) ! \max(0,-n)! max ( 0 , − n )! is ( − n ) ! (-n)! ( − n )! for n ≤ 0 n\leq0 n ≤ 0 and $1f o r for f or n\geq 1. The induced vacuum module carries the vertex algebra V \mathcal{V} V . Its translation satisfies T 1 = 0 T1=0 T 1 = 0 and
[ T , U − j ] = U − j − 1 ( j ≥ 0 ) , [ T , U n ] = − ( n − 1 ) U n − 1 ( n ≥ 1 ) . [T,U_{-j}]=U_{-j-1}\quad(j\geq0),\qquad[T,U_n]=-(n-1)U_{n-1}\quad(n\geq1). [ T , U − j ] = U − j − 1 ( j ≥ 0 ) , [ T , U n ] = − ( n − 1 ) U n − 1 ( n ≥ 1 ) .
PBW gives the basis U − r 1 ⋯ U − r m 1 U_{-r_1}\cdots U_{-r_m}1 U − r 1 ⋯ U − r m 1 with r 1 ≥ ⋯ ≥ r m ≥ 0 r_1\geq\cdots\geq r_m\geq0 r 1 ≥ ⋯ ≥ r m ≥ 0 . Standard field reconstruction gives the vertex-algebra structure [1 ] . At fixed c c c , V c V^c V c is the universal Virasoro VOA. Its vacuum module is the highest-weight-zero Verma module modulo the submodule generated by L − 1 1 L_{-1}1 L − 1 1 . We work with this universal algebra rather than its simple quotient.
The vectors T j U T^jU T j U suggest the differential polynomial coordinates
P : = k [ u 0 , u − 1 , u − 2 , … ] , u : = u 0 , ∂ : = ∑ j ≥ 0 u − j − 1 ∂ u − j . (7) \mathcal{P}:=\mathbb{k}[u_0,u_{-1},u_{-2},\ldots],\qquad u:=u_0,\qquad\partial:=\sum_{j\geq0}u_{-j-1}\partial_{u_{-j}}. \tag*{(7)} P := k [ u 0 , u − 1 , u − 2 , … ] , u := u 0 , ∂ := j ≥ 0 ∑ u − j − 1 ∂ u − j . (7)
Thus P = k ⊗ Q P 0 \mathcal{P}=\mathbb{k}\otimes_{\mathbb{Q}}\mathcal{P}_0 P = k ⊗ Q P 0 , and specializing ℏ = 0 \hbar=0 ℏ = 0 recovers the algebra P 0 \mathcal{P}_0 P 0 introduced above. We use capitals for the generating vectors and modes on the Virasoro side, and lowercase letters for their commuting polynomial coordinates. Jet order always means derivative order: u − j u_{-j} u − j has order j j j , with subscript − j -j − j . Two gradings will be used:
(i) the conformal weight, wt ( u − j ) : = j + 2 \operatorname{wt}(u_{-j}):=j+2 wt ( u − j ) := j + 2 ;
(ii) the factor number , deg f ( u − j ) : = 1 \deg_{\mathrm{f}}(u_{-j}):=1 deg f ( u − j ) := 1 , extended additively to products.
Thus u u − 2 uu_{-2} u u − 2 and u − 1 2 u_{-1}^2 u − 1 2 have factor number two and weight six, whereas u 3 u^3 u 3 has factor number three and the same weight. The parameter ℏ \hbar ℏ has both degrees zero. Let P [ m ] \mathcal{P}^{[m]} P [ m ] denote the factor-homogeneous subspace of degree m m m , and put P ≤ m : = ⨁ r = 0 m P [ r ] \mathcal{P}^{\leq m}:=\bigoplus_{r=0}^{m}\mathcal{P}^{[r]} P ≤ m := ⨁ r = 0 m P [ r ] .
Since U − j 1 = T j U U_{-j}1=T^jU U − j 1 = T j U for j ≥ 0 j\geq0 j ≥ 0 , define the symmetrization map
σ ( u − j 1 ⋯ u − j m ) : = 1 m ! ∑ τ ∈ S m U − j τ ( 1 ) ⋯ U − j τ ( m ) 1 , σ ( 1 ) : = 1. (8) \sigma(u_{-j_1}\cdots u_{-j_m}):=\frac{1}{m!}\sum_{\tau\in S_m}U_{-j_{\tau(1)}}\cdots U_{-j_{\tau(m)}}1,\qquad\sigma(1):=1. \tag*{(8)} σ ( u − j 1 ⋯ u − j m ) := m ! 1 τ ∈ S m ∑ U − j τ ( 1 ) ⋯ U − j τ ( m ) 1 , σ ( 1 ) := 1. (8)
The average includes repeated indices. This is the standard PBW symmetrization underlying the symbol calculus of Berezin [3 ] ; the induced-module version needed here is proved below.
Proposition 2.1. The map σ : P → V \sigma:\mathcal{P}\to\mathcal{V} σ : P → V is a weight-preserving filtered k \mathbb{k} k -linear isomorphism, and
σ ∂ = T σ . (9) \sigma\partial=T\sigma. \tag*{(9)} σ ∂ = T σ . (9)
Proof. Let τ : P → V \tau:\mathcal{P}\to\mathcal{V} τ : P → V send each monomial to its ordered word in the U − j U_{-j} U − j . PBW makes τ \tau τ an isomorphism. Filter by factor number. Exchanging adjacent creation modes introduces a commutator and hence a word with one fewer factor. Consequently,
( τ − 1 σ − 1 ) P ≤ m ⊆ P ≤ m − 1 . (\tau^{-1}\sigma-1)\mathcal{P}^{\leq m}\subseteq\mathcal{P}^{\leq m-1}. ( τ − 1 σ − 1 ) P ≤ m ⊆ P ≤ m − 1 .
For Q : = τ − 1 σ − 1 Q := \tau^{-1}\sigma- 1 Q := τ − 1 σ − 1 , this gives Q m + 1 = 0 Q^{m+1} = 0 Q m + 1 = 0 on P ≤ m \mathcal{P}^{\le m} P ≤ m and the inverse σ − 1 = ( ∑ r = 0 m ( − Q ) r ) τ − 1 \sigma^{-1} = \left(\sum_{r=0}^{m}(-Q)^r\right)\tau^{-1} σ − 1 = ( ∑ r = 0 m ( − Q ) r ) τ − 1 . This proves the filtered isomorphism even with infinitely many generators. Finally [ T , U − j ] = U − j − 1 [T,U_{-j}] = U_{-j-1} [ T , U − j ] = U − j − 1 ; translation acts on each factor of a symmetrized word in turn, giving (9) . □
The map σ \sigma σ is linear; it does not identify polynomial multiplication with a ( − 1 ) b a(-1)b a ( − 1 ) b . For example,
σ ( u u − 1 ) = ( U − 1 U 0 + ℏ 12 U − 3 ) 1 , σ ( u − 3 ) = U − 3 1. (10) \sigma(uu_{-1}) = \left(U_{-1}U_0 + \frac{\hbar}{12}U_{-3}\right)1,\qquad\sigma(u_{-3}) = U_{-3}1. \tag*{(10)} σ ( u u − 1 ) = ( U − 1 U 0 + 12 ℏ U − 3 ) 1 , σ ( u − 3 ) = U − 3 1. (10)
Symbols of operators on differential polynomials
The first step is to express the Virasoro action as operators on differential polynomials. Transport the mode action by symmetrization:
ρ ( U n ) : = σ − 1 U n σ ∈ End k ( P ) . (11) \rho(U_n) := \sigma^{-1}U_n\sigma\in\operatorname{End}_{k}(\mathcal{P}). \tag*{(11)} ρ ( U n ) := σ − 1 U n σ ∈ End k ( P ) . (11)
Every such endomorphism can be written uniquely as a possibly infinite sum of differential operators with polynomial coefficients. We now specify the formal series that encode this expression; no bound on the differential order is assumed.
The auxiliary commuting parameters t j t_j t j and multi-index entries are indexed by derivative order j ≥ 0 j \ge0 j ≥ 0 , so t j t_j t j pairs with u − j u_{-j} u − j and U − j U_{-j} U − j . A multi-index I = ( I i ) I = (I_i) I = ( I i ) has nonnegative integer entries, all but finitely many of which are zero. We use
u I : = ∏ i u − i I i , I ! : = ∏ i I i ! , ∣ I ∣ : = ∑ i I i , ∂ u I : = ∏ i ∂ u − i I i . u^I := \prod_i u_{-i}^{I_i},\qquad I! := \prod_i I_i!,\qquad|I| := \sum_i I_i,\qquad\partial_u^I := \prod_i \partial_{u_{-i}}^{I_i}. u I := i ∏ u − i I i , I ! := i ∏ I i ! , ∣ I ∣ := i ∑ I i , ∂ u I := i ∏ ∂ u − i I i .
For a k k k -module W W W , put
W [ [ t ] ] : = ∏ I W t I , [ t I ] ( ∑ J w J t J ) : = w I . W[[t]] := \prod_I Wt^I,\qquad[t^I]\left(\sum_J w_Jt^J\right) := w_I. W [[ t ]] := I ∏ W t I , [ t I ] ( J ∑ w J t J ) := w I .
Thus [ t I ] [t^I] [ t I ] extracts the coefficient of the monomial t I t^I t I . We give this space the product topology in which W W W is discrete. If W W W is an algebra, each product coefficient is a finite sum over J ≤ I J \le I J ≤ I . Power series in an element of positive total t t t -degree are therefore well defined. A k k k -linear map extends coefficientwise to a continuous k [ [ t ] ] k[[t]] k [[ t ]] -linear map. The formal derivatives ∂ t i \partial_{t_i} ∂ t i are k k k -linear, but not k [ [ t ] ] k[[t]] k [[ t ]] -linear; they commute with these coefficientwise extensions.
Set U ( t ) : = ∑ j ≥ 0 t j U − j U(t) := \sum_{j\ge0} t_jU_{-j} U ( t ) := ∑ j ≥ 0 t j U − j . Here U − j U_{-j} U − j is a creation operator, whereas U − j 1 = T j U U_{-j}1 = T^jU U − j 1 = T j U is a vector. In the centrally extended mode algebra g \mathfrak{g} g specified below,
U ( t ) ∈ g [ [ t ] ] , e U ( t ) ∈ U ( g ) [ [ t ] ] , e U ( t ) 1 ∈ V [ [ t ] ] . U(t) \in\mathfrak{g}[[t]],\qquad e^{U(t)} \in U(\mathfrak{g})[[t]],\qquad e^{U(t)}1 \in\mathcal{V}[[t]]. U ( t ) ∈ g [[ t ]] , e U ( t ) ∈ U ( g ) [[ t ]] , e U ( t ) 1 ∈ V [[ t ]] .
On the polynomial side, let u ( t ) : = ∑ j ≥ 0 u − j t j u(t) := \sum_{j\ge0}u_{-j}t_j u ( t ) := ∑ j ≥ 0 u − j t j and e u ( t ) : = ∑ I u I t I / I ! ∈ P [ [ t ] ] e^{u(t)} := \sum_I u^It^I/I! \in\mathcal{P}[[t]] e u ( t ) := ∑ I u I t I / I ! ∈ P [[ t ]] . The coefficientwise extension of σ \sigma σ satisfies
σ ( e u ( t ) ) = e U ( t ) 1. (12) \sigma\left(e^{u(t)}\right) = e^{U(t)}1. \tag*{(12)} σ ( e u ( t ) ) = e U ( t ) 1. (12)
Indeed, each coefficient on either side is the same average of creation words. Thus symmetric PBW corresponds to the exponential of their sum. No completion in the u u u variables is needed: every t t t -coefficient is a polynomial. The right side is module-valued; we invert exponentials only in the respective algebras.
For a k \mathbb{k} k -linear endomorphism D D D of P \mathcal{P} P , define
symb ( D ) ( u , t ) : = e − u ( t ) D ( e u ( t ) ) = ∑ I c I ( u ) t I ∈ P [ [ t ] ] . (13) \operatorname{symb}(D)(u,t) := e^{-u(t)}D\left(e^{u(t)}\right) = \sum_I c_I(u)t^I \in\mathcal{P}[[t]]. \tag*{(13)} symb ( D ) ( u , t ) := e − u ( t ) D ( e u ( t ) ) = I ∑ c I ( u ) t I ∈ P [[ t ]] . (13)
Comparing coefficients after multiplication by e u ( t ) e^{u(t)} e u ( t ) gives
D ( u J ) J ! = ∑ I ≤ J c I ( u ) u J − I ( J − I ) ! , D = ∑ I c I ( u ) ∂ u I . (14) \frac{D(u^J)}{J!} = \sum_{I\leq J} c_I(u)\frac{u^{J-I}}{(J-I)!}, \qquad D = \sum_I c_I(u)\partial_u^I. \tag*{(14)} J ! D ( u J ) = I ≤ J ∑ c I ( u ) ( J − I )! u J − I , D = I ∑ c I ( u ) ∂ u I . (14)
Equivalently,
c I ( u ) = ∑ J ≤ I ( − u ) I − J ( I − J ) ! J ! D ( u J ) . c_I(u) = \sum_{J\leq I}\frac{(-u)^{I-J}}{(I-J)!J!}D(u^J). c I ( u ) = J ≤ I ∑ ( I − J )! J ! ( − u ) I − J D ( u J ) .
All coefficient multiplications stand to the left of the derivatives. Only finitely many derivatives contribute on a polynomial, although their orders need not be uniformly bounded. This reconstruction holds for every linear operator; the coordinate-degree bound for a mode will follow from its explicit symbol. The sums in (14) form the completion of the polynomial Weyl algebra used here: they act on each polynomial by a finite sum.
Explicit mode symbols from the exponential map
Berezin’s symbol calculus gives differential formulas for the regular enveloping-algebra action [3 ] . The induced-module exponential formula used below is a specialization of [4 Theorem 23 and equations (4.4)–(4.5)] . We include a coefficientwise proof adapted to the present Virasoro module and its countably many modes. The preceding reconstruction applies to any endomorphism. For a Lie algebra action we need a more explicit symbol, in particular a bound on the coordinate degree of its coefficients. This bound will isolate the terms that preserve or raise factor number in § 3 . To retain the central term before evaluating it on the vacuum, let g \mathfrak{g} g have the mode relations above with their scalar central term multiplied by c / 6 c/6 c /6 . The central generator c c c acts as the scalar 6 6 6 on the vacuum, so the displayed scalar mode relations are recovered. It is distinct from the central charge c c c . We use the splitting into Lie subalgebras
g = g − ⊕ g + , g − : = span k { U − j : j ≥ 0 } , g + : = span k { U n : n > 0 } ⊕ k c , \mathfrak{g}=\mathfrak{g}_{-}\oplus\mathfrak{g}_{+},\qquad\mathfrak{g}_{-}:=\operatorname{span}_{\mathbb{k}}\{U_{-j}:j\geq0\},\qquad\mathfrak{g}_{+}:=\operatorname{span}_{\mathbb{k}}\{U_n:n>0\}\oplus\mathbb{k}c, g = g − ⊕ g + , g − := span k { U − j : j ≥ 0 } , g + := span k { U n : n > 0 } ⊕ k c ,
with projections denoted by − - − and + + + . More explicitly,
V = U ( g ) ⊗ U ( g + ) k 1 , \mathcal{V}=U(\mathfrak{g})\otimes_{U(\mathfrak{g}_{+})}\mathbb{k}1, V = U ( g ) ⊗ U ( g + ) k 1 ,
where U n U_n U n for n > 0 n>0 n > 0 annihilates 1 1 1 and c c c acts as 6 6 6 . PBW multiplication gives U ( g ) ≃ U ( g − ) ⊗ k U ( g + ) U(\mathfrak{g})\simeq U(\mathfrak{g}_{-})\otimes_{\mathbb{k}}U(\mathfrak{g}_{+}) U ( g ) ≃ U ( g − ) ⊗ k U ( g + ) ; after imposing the vacuum action, symmetrization identifies V ≃ U ( g − ) ≃ P \mathcal{V}\simeq U(\mathfrak{g}_{-})\simeq\mathcal{P} V ≃ U ( g − ) ≃ P as k \mathbb{k} k -modules.
Set
φ ( s ) : = 1 − e − s s , ψ ( s ) : = φ ( s ) − 1 = s 1 − e − s = 1 + s 2 + s 2 12 − s 4 720 + ⋯ . (15) \varphi(s) := \frac{1-e^{-s}}{s}, \qquad\psi(s) := \varphi(s)^{-1} = \frac{s}{1-e^{-s}} = 1+\frac{s}{2}+\frac{s^2}{12}-\frac{s^4}{720}+\cdots. \tag*{(15)} φ ( s ) := s 1 − e − s , ψ ( s ) := φ ( s ) − 1 = 1 − e − s s = 1 + 2 s + 12 s 2 − 720 s 4 + ⋯ . (15)
These quotients denote power series. For X ∈ g X \in\mathfrak{g} X ∈ g , define
v X ( U , t ) : = ψ ( ad U ( t ) ) ( e − ad U ( t ) X ) − = ∑ j ≥ 0 v X , j ( t ) U − j , ( e − ad U ( t ) X ) + 1 = 6 c X ( t ) 1. (16) \begin{aligned}
v_X(U,t) &:= \psi(\operatorname{ad} U(t))(e^{-\operatorname{ad} U(t)}X)_{-} = \sum_{j\geq0}v_{X,j}(t)U_{-j},\\
(e^{-\operatorname{ad} U(t)}X)_{+}1 &= 6c_X(t)1. \tag*{(16)}
\end{aligned} v X ( U , t ) ( e − ad U ( t ) X ) + 1 := ψ ( ad U ( t )) ( e − ad U ( t ) X ) − = j ≥ 0 ∑ v X , j ( t ) U − j , = 6 c X ( t ) 1. (16)
where c X ( t ) c_X(t) c X ( t ) is the coefficient of c c c in the conjugated series. The projection precedes ψ \psi ψ . All scalar coefficients lie in k [ [ t ] ] k[[t]] k [[ t ]] , so their dependence on ℏ \hbar ℏ is implicit. For each t I t^I t I , only its finitely many input indices and at most ∣ I ∣ |I| ∣ I ∣ brackets contribute. Hence every coefficient of v X ( U , t ) v_X(U,t) v X ( U , t ) is a finite linear combination of the U − j U_{-j} U − j , and the shorthand
v X ( u , t ) 1 : = σ − 1 ( v X ( U , t ) 1 ) = ∑ j ≥ 0 v X , j ( t ) u − j v_X(u,t)1 := \sigma^{-1}(v_X(U,t)1) = \sum_{j\geq0}v_{X,j}(t)u_{-j} v X ( u , t ) 1 := σ − 1 ( v X ( U , t ) 1 ) = j ≥ 0 ∑ v X , j ( t ) u − j
is well defined in P [ [ t ] ] \mathcal{P}[[t]] P [[ t ]] , since σ ( u − j ) = U − j 1 \sigma(u_{-j})=U_{-j}1 σ ( u − j ) = U − j 1 . The next identity is the formal version of the differential of the exponential map; see [14 Section 5.4, Theorem 5.4] .
Lemma 2.2 (Formal differentiation of the exponential). In U ( g ) [ [ t ] ] U(\mathfrak{g})[[t]] U ( g ) [[ t ]] , the following identity holds coefficientwise:
∂ t j e U ( t ) = ∫ 0 1 e ( 1 − s ) U ( t ) U − j e s U ( t ) d s = e U ( t ) φ ( ad U ( t ) ) U − j . (17) \partial_{t_j}e^{U(t)}=\int_0^1 e^{(1-s)U(t)}U_{-j}e^{sU(t)}\,ds=e^{U(t)}\varphi(\operatorname{ad} U(t))U_{-j}. \tag*{(17)} ∂ t j e U ( t ) = ∫ 0 1 e ( 1 − s ) U ( t ) U − j e s U ( t ) d s = e U ( t ) φ ( ad U ( t )) U − j . (17)
Proof. To verify this identity formally, expand the exponential and differentiate each power: for p , q ≥ 0 p,q\geq0 p , q ≥ 0 , the word U ( t ) p U − j U ( t ) q U(t)^pU_{-j}U(t)^q U ( t ) p U − j U ( t ) q has coefficient 1 / ( p + q + 1 ) ! 1/(p+q+1)! 1/ ( p + q + 1 )! . Expanding the integrand gives the same coefficient, since the integral of ( 1 − s ) p s q / ( p ! q ! ) (1-s)^ps^q/(p!q!) ( 1 − s ) p s q / ( p ! q !) over [ 0 , 1 ] [0,1] [ 0 , 1 ] is 1 / ( p + q + 1 ) ! 1/(p+q+1)! 1/ ( p + q + 1 )! . Integration here is performed coefficientwise on polynomials in s s s , so no analytic convergence is required. Factoring out e U ( t ) e^{U(t)} e U ( t ) and using conjugation gives the last equality. □
For arbitrary X ∈ g X\in\mathfrak{g} X ∈ g , extend (2.9) by ρ ( X ) : = σ − 1 X σ \rho(X):=\sigma^{-1}X\sigma ρ ( X ) := σ − 1 X σ .
Proposition 2.3 (Symbol of a Lie algebra action). For X ∈ g X\in\mathfrak{g} X ∈ g , the symbol and action of X X X are
ℓ X ( u , t ) : = symb ( ρ ( X ) ) ( u , t ) = v X ( u , t ) 1 + 6 c X ( t ) , ρ ( X ) = ∑ I [ t I ℓ X ( u , t ) ] ∂ u I . (18) \begin{aligned}
\ell_X(u,t) &:= \operatorname{symb}(\rho(X))(u,t)=v_X(u,t)1+6c_X(t),\\
\rho(X) &= \sum_I [t^I\ell_X(u,t)]\partial_u^I. \tag*{(18)}
\end{aligned} ℓ X ( u , t ) ρ ( X ) := symb ( ρ ( X )) ( u , t ) = v X ( u , t ) 1 + 6 c X ( t ) , = I ∑ [ t I ℓ X ( u , t )] ∂ u I . (18)
In particular, every coefficient has coordinate degree at most one.
Proof. Step 1. Move the action to the vacuum. Conjugation in the completed enveloping algebra gives
X e U ( t ) = e U ( t ) ( e − ad U ( t ) X ) . Xe^{U(t)}=e^{U(t)}(e^{-\operatorname{ad}U(t)}X). X e U ( t ) = e U ( t ) ( e − ad U ( t ) X ) .
Each iterated commutator lies in g [ [ t ] ] \mathfrak{g}[[t]] g [[ t ]] and raises total t t t -degree. Its coefficient at any fixed t I t^I t I is a finite linear combination of modes and the central generator. The same therefore holds for formal power series in ad U ( t ) \operatorname{ad} U(t) ad U ( t ) applied to X X X . Splitting into the two subalgebras and acting on the vacuum gives
X e U ( t ) 1 = e U ( t ) ( e − ad U ( t ) X ) − 1 + 6 c X ( t ) e U ( t ) 1. (19) X e^{U(t)}1=e^{U(t)}\left(e^{-\operatorname{ad}U(t)}X\right)_{-}1+6c_X(t)e^{U(t)}1. \tag*{(19)} X e U ( t ) 1 = e U ( t ) ( e − ad U ( t ) X ) − 1 + 6 c X ( t ) e U ( t ) 1. (19)
Since g − \mathfrak{g}_{-} g − is a subalgebra, ad U ( t ) \operatorname{ad}U(t) ad U ( t ) preserves it and raises total t t t -degree. Thus φ ( ad U ( t ) ) \varphi(\operatorname{ad}U(t)) φ ( ad U ( t )) is invertible there, and the definition of v X v_X v X says exactly that
φ ( ad U ( t ) ) v X ( U , t ) = ( e − ad U ( t ) X ) − . (20) \varphi(\operatorname{ad}U(t))v_X(U,t)=\left(e^{-\operatorname{ad}U(t)}X\right)_{-}. \tag*{(20)} φ ( ad U ( t )) v X ( U , t ) = ( e − ad U ( t ) X ) − . (20)
Step 2. Transfer by symmetrization. Apply Lemma 2.2 to express derivatives of the noncommutative exponential, then transfer them to polynomial coordinates. The extension of σ \sigma σ acts on polynomial coefficients only, so ∂ t j σ ( ∑ I f I t I ) = ∑ I σ ( f I ) ∂ t j t I = σ ( ∂ t j ∑ I f I t I ) \partial_{t_j}\sigma(\sum_I f_I t^I)=\sum_I\sigma(f_I)\partial_{t_j}t^I=\sigma(\partial_{t_j}\sum_I f_I t^I) ∂ t j σ ( ∑ I f I t I ) = ∑ I σ ( f I ) ∂ t j t I = σ ( ∂ t j ∑ I f I t I ) .
Differentiating the PBW exponential therefore gives
∂ t j ( e u ( t ) 1 ) = σ ( ∂ t j e u ( t ) ) = σ ( u − j e u ( t ) ) . (21) \partial_{t_j}(e^{u(t)}1)=\sigma(\partial_{t_j}e^{u(t)})=\sigma(u_{-j}e^{u(t)}). \tag*{(21)} ∂ t j ( e u ( t ) 1 ) = σ ( ∂ t j e u ( t ) ) = σ ( u − j e u ( t ) ) . (21)
Multiply these two derivative identities by v X , j ( t ) v_{X,j}(t) v X , j ( t ) and sum:
e U ( t ) φ ( ad U ( t ) ) v X ( U , t ) 1 = ∑ j ≥ 0 v X , j ( t ) ∂ t j ( e u ( t ) 1 ) = σ [ v X ( u , t ) 1 e u ( t ) ] . (22) \begin{aligned}
e^{U(t)}\varphi(\operatorname{ad}U(t))v_X(U,t)1
&=\sum_{j\geq0}v_{X,j}(t)\partial_{t_j}(e^{u(t)}1)\\
&=\sigma\left[v_X(u,t)1e^{u(t)}\right]. \tag*{(22)}
\end{aligned} e U ( t ) φ ( ad U ( t )) v X ( U , t ) 1 = j ≥ 0 ∑ v X , j ( t ) ∂ t j ( e u ( t ) 1 ) = σ [ v X ( u , t ) 1 e u ( t ) ] . (22)
The sum is coefficientwise finite. The scalar coefficients multiply after differentiation and pass through the k [ [ t ] ] \mathbf{k}[[t]] k [[ t ]] -linear map σ \sigma σ ; no derivative acts on them.
Step 3. Recover the symbol. Substitution into the conjugation identity now yields
X e U ( t ) 1 = σ [ ( v X ( u , t ) 1 + 6 c X ( t ) ) e u ( t ) ] . (23) Xe^{U(t)}1=\sigma\left[(v_X(u,t)1+6c_X(t))e^{u(t)}\right]. \tag*{(23)} X e U ( t ) 1 = σ [ ( v X ( u , t ) 1 + 6 c X ( t )) e u ( t ) ] . (23)
Apply σ − 1 \sigma^{-1} σ − 1 and cancel e u ( t ) e^{u(t)} e u ( t ) in the symbol definition. This proves (18) , including the degree bound. □
For the Virasoro modes needed below, take X = U n X=U_n X = U n and write ℓ n : = ℓ U n \ell_n:=\ell_{U_n} ℓ n := ℓ U n and c n : = c U n c_n:=c_{U_n} c n := c U n . Thus (18) gives
ℓ n ( u , t ) = v U n ( u , t ) 1 + 6 c n ( t ) , ρ ( U n ) = ∑ I ( [ t I ℓ n ( u , t ) ] ) ∂ u I . \ell_n(u,t)=v_{U_n}(u,t)1+6c_n(t),\qquad\rho(U_n)=\sum_I([t^I\ell_n(u,t)])\partial_u^I. ℓ n ( u , t ) = v U n ( u , t ) 1 + 6 c n ( t ) , ρ ( U n ) = I ∑ ([ t I ℓ n ( u , t )]) ∂ u I .
For creation modes n ≤ 0 n\leq0 n ≤ 0 , all brackets stay in g − \mathfrak{g}_{-} g − , so c n = 0 c_n=0 c n = 0 . To illustrate an annihilation mode, set t 0 = t t_0=t t 0 = t and t j = 0 t_j=0 t j = 0 for j ≥ 1 j\geq1 j ≥ 1 . Then U ( t ) = t U 0 U(t)=tU_0 U ( t ) = t U 0 and
e − t ad U 0 U 4 = U 4 + 4 ℏ t U 2 + ℏ t 2 c + 4 ℏ 2 t 2 U 0 . e^{-t\operatorname{ad}U_0}U_4=U_4+4\hbar tU_2+\frac{\hbar t}{2}c+4\hbar^2t^2U_0. e − t ad U 0 U 4 = U 4 + 4ℏ t U 2 + 2 ℏ t c + 4 ℏ 2 t 2 U 0 .
The creation component commutes with U ( t ) U(t) U ( t ) , so v U 4 ( U , t ) = 4 ℏ 2 t 2 U 0 v_{U_4}(U,t)=4\hbar^2t^2U_0 v U 4 ( U , t ) = 4 ℏ 2 t 2 U 0 and 6 c 4 ( t ) = 3 ℏ t 6c_4(t)=3\hbar t 6 c 4 ( t ) = 3ℏ t . Consequently,
ℓ 4 ( u ; t , 0 , … ) = 3 ℏ t + 4 ℏ 2 u t 2 , ρ ( U 4 ) ∣ k [ u ] = 3 ℏ ∂ u + 4 ℏ 2 u ∂ u 2 . (24) \ell_4(u;t,0,\ldots)=3\hbar t+4\hbar^2u t^2,\qquad\rho(U_4)|_{\mathbf{k}[u]}=3\hbar\partial_u+4\hbar^2u\partial_u^2. \tag*{(24)} ℓ 4 ( u ; t , 0 , … ) = 3ℏ t + 4 ℏ 2 u t 2 , ρ ( U 4 ) ∣ k [ u ] = 3ℏ ∂ u + 4 ℏ 2 u ∂ u 2 . (24)
The second-order term is invisible on 1 1 1 and u u u ; coordinate degree at most one does not mean differential order at most one.
Lemma 2.4 (Parameter degree of a mode). For a multi-index I I I , every coefficient of [ t I ] ℓ n [t^I]\ell_n [ t I ] ℓ n is homogeneous of degree ∣ I ∣ |I| ∣ I ∣ in ℏ \hbar ℏ .
Proof. Each commutator with U ( t ) U(t) U ( t ) contributes one parameter and one factor of ℏ \hbar ℏ . A term of degree ∣ I ∣ |I| ∣ I ∣ in (16) has ∣ I ∣ |I| ∣ I ∣ brackets in total, before and after projection; the central coefficient c n c_n c n obeys the same count. Projection, factorials in U − j = j ! U ( − j − 1 ) U_{-j}=j!U(-j-1) U − j = j ! U ( − j − 1 ) , and evaluation of c \mathbf{c} c introduce only rational factors. Equation (18) proves the assertion. □ \square □
Leading mode terms and the weight-four zero mode
The construction depends on the zero-mode action of H 4 = U ( − 1 ) U H_4=U(-1)U H 4 = U ( − 1 ) U . We first isolate the terms of the individual modes with the largest factor-number shifts, then combine them to determine this action. Here “leading” refers to factor number, not to differential order or degree in ℏ \hbar ℏ .
Mode terms that raise or preserve factor number
For a differential term p ( u ) ∂ u I p(u)\partial_u^I p ( u ) ∂ u I with factor-homogeneous coefficient p p p , the factor-number shift is deg f p − ∣ I ∣ \operatorname{deg}_{f}p-|I| deg f p − ∣ I ∣ . Write D [ s ] D^{[s]} D [ s ] for the component of an operator D D D that changes factor number by s s s , so D [ s ] P [ m ] ⊆ P [ m + s ] D^{[s]}\mathcal{P}^{[m]}\subseteq\mathcal{P}^{[m+s]} D [ s ] P [ m ] ⊆ P [ m + s ] . By (18) , every single-mode coefficient has degree at most one. Hence terms of differential order at least two lower factor number. The multiplication and first-order coefficients of any D D D are recovered by
c 0 = D ( 1 ) , c e j = D ( u − j ) − u − j D ( 1 ) . c_0=D(1),\qquad c_{e_j}=D(u_{-j})-u_{-j}D(1). c 0 = D ( 1 ) , c e j = D ( u − j ) − u − j D ( 1 ) .
Here 0 0 0 is the zero multi-index and e j e_j e j has its only nonzero entry, equal to one, at position j j j . This determines the following pieces exactly, without determining the higher-order terms.
For i ≥ 0 i\geq0 i ≥ 0 , U − i 1 = σ ( u − i ) U_{-i}1=\sigma(u_{-i}) U − i 1 = σ ( u − i ) . The multiplication coefficient of ρ ( U − i ) : = σ − 1 U − i σ \rho(U_{-i}):=\sigma^{-1}U_{-i}\sigma ρ ( U − i ) := σ − 1 U − i σ is u − i u_{-i} u − i . Symmetric ordering gives
U − i U − j 1 − σ ( u − i u − j ) = 1 2 [ U − i , U − j ] 1 , [ U − i , U − j ] = ℏ ( j − i ) i ! j ! ( i + j + 2 ) ! U − i − j − 2 . U_{-i}U_{-j}1-\sigma(u_{-i}u_{-j})=\frac{1}{2}[U_{-i},U_{-j}]1,\qquad[U_{-i},U_{-j}]=\hbar\frac{(j-i)i!j!}{(i+j+2)!}U_{-i-j-2}. U − i U − j 1 − σ ( u − i u − j ) = 2 1 [ U − i , U − j ] 1 , [ U − i , U − j ] = ℏ ( i + j + 2 )! ( j − i ) i ! j ! U − i − j − 2 .
Consequently,
ρ ( U − i ) = u − i + ℏ 2 ∑ j ≥ 0 ( j − i ) i ! j ! ( i + j + 2 ) ! u − i − j − 2 ∂ u − j \rho(U_{-i})=u_{-i}+\frac{\hbar}{2}\sum_{j\geq0}\frac{(j-i)i!j!}{(i+j+2)!}u_{-i-j-2}\partial_{u_{-j}} ρ ( U − i ) = u − i + 2 ℏ j ≥ 0 ∑ ( i + j + 2 )! ( j − i ) i ! j ! u − i − j − 2 ∂ u − j
ρ ( U − i ) = + terms of differential order at least two . (25) \phantom{\rho(U_{-i})={}}+\text{terms of differential order at least two}. \tag*{(25)} ρ ( U − i ) = + terms of differential order at least two . (25)
The two displayed pieces have factor shifts + 1 +1 + 1 and 0 0 0 ; all omitted terms lower factor number.
For n ≥ − 1 n\geq-1 n ≥ − 1 , ρ ( U n + 2 ) ( 1 ) = 0 \rho(U_{n+2})(1)=0 ρ ( U n + 2 ) ( 1 ) = 0 , and the first-order coefficient is σ − 1 ( [ U n + 2 , U − j ] 1 ) \sigma^{-1}([U_{n+2},U_{-j}]1) σ − 1 ([ U n + 2 , U − j ] 1 ) . On the vacuum, the commutator is
[ U n + 2 , U − j ] 1 = { ℏ ( n + j + 2 ) j ! ( j − n ) ! U − j + n 1 , j ≥ n , 0 , j < n + ℏ 2 ( n 3 − n ) j ! δ n , j + 2 1. [U_{n+2},U_{-j}]1=
\begin{cases}
\hbar\frac{(n+j+2)j!}{(j-n)!}U_{-j+n}1, & j\geq n,\\
0, & j<n
\end{cases}
+\frac{\hbar}{2}(n^3-n)j!\delta_{n,j+2}1. [ U n + 2 , U − j ] 1 = { ℏ ( j − n )! ( n + j + 2 ) j ! U − j + n 1 , 0 , j ≥ n , j < n + 2 ℏ ( n 3 − n ) j ! δ n , j + 2 1.
Thus the factor-preserving part is
ρ ( U n + 2 ) [ 0 ] = ℏ ∑ j ≥ max ( 0 , n ) ( n + j + 2 ) j ! ( j − n ) ! u − j + n ∂ u − j , n ≥ − 1. (26) \rho(U_{n+2})^{[0]}=\hbar\sum_{j\geq\max(0,n)}\frac{(n+j+2)j!}{(j-n)!}u^{-j+n}\partial_{u^{-j}},\qquad n\geq-1. \tag*{(26)} ρ ( U n + 2 ) [ 0 ] = ℏ j ≥ m a x ( 0 , n ) ∑ ( j − n )! ( n + j + 2 ) j ! u − j + n ∂ u − j , n ≥ − 1. (26)
For n ≥ 2 n\geq2 n ≥ 2 , the remaining first-order term is the scalar operator
ℏ 2 ( n 3 − n ) ( n − 2 ) ! ∂ u − n + 2 . (27) \frac{\hbar}{2}(n^3-n)(n-2)!\partial_{u^{-n+2}}. \tag*{(27)} 2 ℏ ( n 3 − n ) ( n − 2 )! ∂ u − n + 2 . (27)
It lowers factor number by one. Higher-order terms also lower factor number, so (27) is not the complete shift-(−1) component. In particular, the second-order term in (24) must not be discarded from the full action. For the first two modes the following formulas are exact:
ρ ( U 1 ) = ℏ ∂ , ρ ( U 2 ) = ℏ ∑ j ≥ 0 ( j + 2 ) u − j ∂ u − j . (28) \rho(U_1)=\hbar\partial,\qquad\rho(U_2)=\hbar\sum_{j\geq0}(j+2)u^{-j}\partial_{u^{-j}}. \tag*{(28)} ρ ( U 1 ) = ℏ ∂ , ρ ( U 2 ) = ℏ j ≥ 0 ∑ ( j + 2 ) u − j ∂ u − j . (28)
Indeed U 1 = ℏ T U_1=\hbar T U 1 = ℏ T on the vacuum module, and U 2 U_2 U 2 acts by ℏ \hbar ℏ times the weight. Their commutators with each creation factor preserve symmetric ordering, so no higher-order correction occurs.
The weight-four action
Define
H 4 : = U ( − 1 ) U , X ℏ : = σ − 1 ℏ − 1 H 4 ( 0 ) σ . (29) H_4:=U(-1)U,\qquad X_{\hbar}:=\sigma^{-1}\hbar^{-1}H_4(0)\sigma. \tag*{(29)} H 4 := U ( − 1 ) U , X ℏ := σ − 1 ℏ − 1 H 4 ( 0 ) σ . (29)
Because the (−1)-product is normal ordered, its two summands at mode zero coincide after reindexing. Thus
X ℏ = 2 ℏ ∑ r ≥ 2 1 ( r − 2 ) ! ρ ( U − r + 2 ) ρ ( U r − 1 ) . (30) X_{\hbar}=\frac{2}{\hbar}\sum_{r\geq2}\frac{1}{(r-2)!}\rho(U_{-r+2})\rho(U_{r-1}). \tag*{(30)} X ℏ = ℏ 2 r ≥ 2 ∑ ( r − 2 )! 1 ρ ( U − r + 2 ) ρ ( U r − 1 ) . (30)
On an input of weight w w w , terms with r > w + 3 r>w+3 r > w + 3 vanish because the right mode lowers weight by r − 3 r-3 r − 3 . Every such right mode has index at least 1 and is divisible by ℏ \hbar ℏ : it kills the vacuum, and every commutation through a creation word supplies a factor of ℏ \hbar ℏ . The displayed division therefore defines an operator over k \mathsf{k} k , not an operator with a pole at ℏ = 0 \hbar=0 ℏ = 0 .
For operators whose right symbol has coordinate degree at most one, the Leibniz rule gives
symb ( D 1 D 2 ) = symb ( D 1 ) symb ( D 2 ) + ∑ j ( ∂ t j symb ( D 1 ) ) ( ∂ u − j symb ( D 2 ) ) . \operatorname{symb}(D_1D_2)=\operatorname{symb}(D_1)\operatorname{symb}(D_2)+\sum_j(\partial_{t_j}\operatorname{symb}(D_1))(\partial_{u^{-j}}\operatorname{symb}(D_2)). symb ( D 1 D 2 ) = symb ( D 1 ) symb ( D 2 ) + j ∑ ( ∂ t j symb ( D 1 )) ( ∂ u − j symb ( D 2 )) .
Derivatives of order at least two kill the right coefficients. At a fixed coefficient in t t t , those coefficients involve finitely many coordinates. Applied to (30) , this yields
symb ( X ℏ ) = 2 ℏ ∑ r ≥ 2 1 ( r − 2 ) ! ( ℓ 2 − r ℓ r − 1 + ∑ j ≥ 0 ( ∂ t j ℓ 2 − r ) ( ∂ u − j ℓ r − 1 ) ) . (31) \operatorname{symb}(X_{\hbar})=\frac{2}{\hbar}\sum_{r\geq2}\frac{1}{(r-2)!}\left(\ell_{2-r}\ell_{r-1}+\sum_{j\geq0}(\partial_{t_j}\ell_{2-r})(\partial_{u^{-j}}\ell_{r-1})\right). \tag*{(31)} symb ( X ℏ ) = ℏ 2 r ≥ 2 ∑ ( r − 2 )! 1 ( ℓ 2 − r ℓ r − 1 + j ≥ 0 ∑ ( ∂ t j ℓ 2 − r ) ( ∂ u − j ℓ r − 1 ) ) . (31)
In particular, its coordinate degree is at most two.
Proposition 3.1. The operator X ℏ X_{\hbar} X ℏ commutes with ∂ \partial ∂ , raises conformal weight by three, and raises factor number by at most one. Its factor-increasing part is X H X_{\mathrm{H}} X H , where
X H : = 6 ∑ j ≥ 0 ∂ j ( u u − 1 ) ∂ u − j . (32) X_{\mathrm{H}} := 6 \sum_{j\geq0} \partial^{j}(u u_{-1})\partial_{u_{-j}}. \tag*{(32)} X H := 6 j ≥ 0 ∑ ∂ j ( u u − 1 ) ∂ u − j . (32) Consequently,
R : = X ℏ − X H (33) R := X_{\hbar}-X_{\mathrm{H}} \tag*{(33)} R := X ℏ − X H (33) preserves or lowers factor number, and all its coefficients lie in Q [ ℏ ] \mathbb{Q}[\hbar] Q [ ℏ ] .
Proof. A zero-mode operator commutes with translation. Together with σ ∂ = T σ \sigma\partial=T\sigma σ ∂ = T σ , this gives [ X ℏ , ∂ ] = 0 [X_{\hbar},\partial]=0 [ X ℏ , ∂ ] = 0 . Each product of modes in (30) raises weight by three. The creation factor raises factor number by at most one, and the annihilation factor does not raise it.
More precisely, the shift-+1 part can only arise from the multiplication term in (25) composed with the factor-preserving part (26) . All other combinations have shift at most zero. The resulting operator is first-order, and its coefficient at ∂ u − j \partial_{u_{-j}} ∂ u − j is
( X ℏ ) u − j [ + 1 ] = 2 j ! ∑ a = 0 j + 1 j + a + 1 a ! ( j + 1 − a ) ! u − a u − j − 1 + a = 3 ∂ j + 1 ( u 2 ) = 6 ∂ j ( u u − 1 ) . (34) \begin{aligned}
(X_{\hbar})^{[+1]}_{u_{-j}}=2j!\sum_{a=0}^{j+1}\frac{j+a+1}{a!(j+1-a)!}u_{-a}u_{-j-1+a} \\
&=3\partial^{j+1}(u^2)=6\partial^j(uu_{-1}). \tag*{(34)}
\end{aligned} ( X ℏ ) u − j [ + 1 ] = 2 j ! a = 0 ∑ j + 1 a ! ( j + 1 − a )! j + a + 1 u − a u − j − 1 + a = 3 ∂ j + 1 ( u 2 ) = 6 ∂ j ( u u − 1 ) . (34)
For the middle equality, pair the terms indexed by a a a and j + 1 − a j+1-a j + 1 − a ; the average of their numerator factors is 3 ( j + 1 ) / 2 3(j+1)/2 3 ( j + 1 ) /2 . This identifies the whole shift-+1 operator as (32) , not merely its value on u u u . Polynomial dependence on ℏ \hbar ℏ was established before (31) . □
The classical limit and the factor-preserving terms
By Lemma 2.4, a differential-order-b b b coefficient of a single mode carries ℏ b \hbar^b ℏ b . In (30) , after division by ℏ \hbar ℏ , the only terms surviving at ℏ = 0 \hbar=0 ℏ = 0 are creation multiplication times the complete first-order annihilation part. The differentiation terms in the composition formula require a positive-order contribution from each mode, hence vanish after division by ℏ \hbar ℏ and specialization. The coordinate-linear part gives X H X_{\mathrm{H}} X H . With n = r − 3 ≥ 2 n=r-3\geq2 n = r − 3 ≥ 2 , the scalar part (27) contributes
2 ℏ u − n − 1 ( n + 1 ) ! ℏ 2 ( n 3 − n ) ( n − 2 ) ! ∂ u − n + 2 = u − n − 1 ∂ u − n + 2 . \frac{2}{\hbar}\frac{u_{-n-1}}{(n+1)!}\frac{\hbar}{2}(n^3-n)(n-2)!\,\partial_{u_{-n+2}}=u_{-n-1}\partial_{u_{-n+2}}. ℏ 2 ( n + 1 )! u − n − 1 2 ℏ ( n 3 − n ) ( n − 2 )! ∂ u − n + 2 = u − n − 1 ∂ u − n + 2 .
Hence
X ℏ ∣ ℏ = 0 = X H + ∑ j ≥ 0 u − j − 3 ∂ u − j = X 0 . (35) \left.X_{\hbar}\right|_{\hbar=0}=X_{\mathrm{H}}+\sum_{j\geq0}u_{-j-3}\partial_{u_{-j}}=X_0. \tag*{(35)} X ℏ ∣ ℏ = 0 = X H + j ≥ 0 ∑ u − j − 3 ∂ u − j = X 0 . (35)
This explains directly why the dispersive term survives: it comes from a central term that lowers the factor number of a single mode. The PBW-leading part and the classical limit are different operations.
The action on the first two powers of u u u . On the one-generator vector only r = 2 , 3 , 5 r=2,3,5 r = 2 , 3 , 5 in (30) contribute. Before applying σ − 1 \sigma^{-1} σ − 1 the result is
( 2 U 0 U − 1 + 4 U − 1 U 0 + U − 3 ) 1. (2U_0U_{-1}+4U_{-1}U_0+U_{-3})1. ( 2 U 0 U − 1 + 4 U − 1 U 0 + U − 3 ) 1.
Using (10) gives
X ℏ u = 6 u u − 1 + ( 1 − ℏ 6 ) u − 3 . (36) X_{\hbar}u=6uu_{-1}+\left(1-\frac{\hbar}{6}\right)u_{-3}. \tag*{(36)} X ℏ u = 6 u u − 1 + ( 1 − 6 ℏ ) u − 3 . (36)
Applying the same mode sum to u 2 u^2 u 2 and reordering its terms gives the following identity. The ordered-mode calculation and the symmetrization conversions are recorded in Appendix § 8 .§ 4 :
X ℏ u 2 = 12 u 2 u − 1 + ( 2 + 5 ℏ 3 ) u u − 3 + 3 ℏ u − 1 u − 2 + ( 3 ℏ 20 − ℏ 2 5 ) u − 5 . (37) X_{\hbar}u^2=12u^2u_{-1}+\left(2+\frac{5\hbar}{3}\right)uu_{-3}+3\hbar u_{-1}u_{-2}+\left(\frac{3\hbar}{20}-\frac{\hbar^2}{5}\right)u_{-5}. \tag*{(37)} X ℏ u 2 = 12 u 2 u − 1 + ( 2 + 3 5ℏ ) u u − 3 + 3ℏ u − 1 u − 2 + ( 20 3ℏ − 5 ℏ 2 ) u − 5 . (37)
For a factor-homogeneous input, let R [ s ] R^{[s]} R [ s ] be the part changing factor number by s s s : R [ s ] P [ m ] ⊆ P [ m + s ] R^{[s]}\mathcal{P}^{[m]}\subseteq\mathcal{P}^{[m+s]} R [ s ] P [ m ] ⊆ P [ m + s ] . Thus R = ∑ s ≤ 0 R [ s ] R=\sum_{s\leq0}R^{[s]} R = ∑ s ≤ 0 R [ s ] . The coordinate-degree-two bound now makes the factor-preserving action on a polynomial in u u u alone completely explicit:
R [ 0 ] f ( u ) = ( 1 − ℏ 6 ) u − 3 f ′ ( u ) + ℏ ( u u − 3 + 3 2 u − 1 u − 2 ) f ′ ′ ( u ) . (38) R^{[0]}f(u)=\left(1-\frac{\hbar}{6}\right)u_{-3}f'(u)+\hbar\left(uu_{-3}+\frac{3}{2}u_{-1}u_{-2}\right)f''(u). \tag*{(38)} R [ 0 ] f ( u ) = ( 1 − 6 ℏ ) u − 3 f ′ ( u ) + ℏ ( u u − 3 + 2 3 u − 1 u − 2 ) f ′′ ( u ) . (38)
Indeed a factor-preserving term with coefficients on the left must have equal coordinate degree and differential order; the latter is therefore at most two. Equation (36) determines the first-order coefficient. Half the two-factor part of X ℏ ( u 2 ) − 2 u X ℏ ( u ) X_{\hbar}(u^2)-2uX_{\hbar}(u) X ℏ ( u 2 ) − 2 u X ℏ ( u ) is ℏ ( u u − 3 + 3 2 u − 1 u − 2 ) \hbar(uu_{-3}+\frac{3}{2}u_{-1}u_{-2}) ℏ ( u u − 3 + 2 3 u − 1 u − 2 ) , determining the second-order coefficient. The one-factor u − 5 u_{-5} u − 5 term of (37) cannot contribute here.
These truncations have distinct purposes. Creation multiplication and factor-preserving annihilation prove the Hopf-leading statement; the scalar first-order annihilation terms are also needed for the classical limit; (38) supplies the seed calculation. None of these truncated formulas is the full remainder R R R . Evaluating the higher-weight construction uses the complete symbols (16) and (18) , or equivalently finite PBW commutation, in (30) .
Local functionals, reduction, and extraction
Local functionals and reduced representatives
A local functional is a differential polynomial considered modulo total derivatives: it is the algebraic counterpart of the integral of a density. These classes form the k-module
F : = P / ∂ P . (39) \mathcal{F}:=\mathcal{P}/\partial\mathcal{P}. \tag*{(39)} F := P / ∂ P . (39)
The image ∂ P \partial\mathcal{P} ∂ P need not be an ideal, so this quotient is not an algebra for polynomial multiplication. By (9) , F \mathcal{F} F is identified with V / T V \mathcal{V}/T\mathcal{V} V / T V . Its Lie bracket is
{ [ a ] , [ b ] } : = [ a ( 0 ) b ] . (40) \{[a],[b]\}:=[a_{(0)}b]. \tag*{(40)} {[ a ] , [ b ]} := [ a ( 0 ) b ] . (40)
Indeed, ( T a ) ( 0 ) = 0 (Ta)(0)=0 ( T a ) ( 0 ) = 0 and [ a ( 0 ) , T ] = 0 [a(0),T]=0 [ a ( 0 ) , T ] = 0 give well-definedness; skew-symmetry and the mode commutator identity give antisymmetry and Jacobi modulo T V T\mathcal{V} T V . The bracket has weight wt ( a ) + wt ( b ) − 1 \operatorname{wt}(a)+\operatorname{wt}(b)-1 wt ( a ) + wt ( b ) − 1 .
The choice below follows the elimination of monomials linear in their highest derivative used for scalar polynomial densities in [15 Section 1.3 of the arXiv version, Step 1] . We specify the reduction map and prove uniqueness of the resulting representative in our polynomial setting.
Definition 4.1. A monomial is reduced if it is a power of u u u , including 1 1 1 , or its highest-order jet occurs at least twice.
For example, u 4 u^4 u 4 , u u 2 uu_2 u u 2 , u 2 2 u_2^2 u 2 2 are reduced, whereas u 2 u − 2 u^2u_{-2} u 2 u − 2 and u − 1 u − 3 u_{-1}u_{-3} u − 1 u − 3 are not. If the highest-order jet u − j u_{-j} u − j , j ≥ 2 j\ge2 j ≥ 2 , occurs once, write the monomial as u − j u − j + 1 a f u_{-j}u_{-j+1}^{a}f u − j u − j + 1 a f with f ∈ k [ u , … , u − j + 2 ] f\in\mathbb{k}[u,\ldots,u_{-j+2}] f ∈ k [ u , … , u − j + 2 ] . Formal integration by parts gives
u − j u − j + 1 a f = − u − j + 1 a + 1 a + 1 ∂ f ( m o d ∂ P ) . (41) u_{-j}u_{-j+1}^{a}f=-\frac{u_{-j+1}^{a+1}}{a+1}\partial f\pmod{\partial\mathcal{P}}. \tag*{(41)} u − j u − j + 1 a f = − a + 1 u − j + 1 a + 1 ∂ f ( mod ∂ P ) . (41)
Every resulting term has lower maximum derivative order. Define reduction by
Π ( m ) : = { m , m reduced , 0 , m = u a u − 1 , − 1 a + 1 Π ( u − j + 1 a + 1 ∂ f ) , m = u − j u − j + 1 a f , j ≥ 2 , f ∈ k [ u , … , u − j + 2 ] . (42) \Pi(\mathfrak{m}):=
\begin{cases}
\mathfrak{m}, & \mathfrak{m}\text{ reduced},\\
0, & \mathfrak{m}=u^a u_{-1},\\
-\dfrac{1}{a+1}\Pi\left(u_{-j+1}^{a+1}\partial f\right), & \mathfrak{m}=u_{-j}u_{-j+1}^{a}f,\ j\ge2,\ f\in\mathbb{k}[u,\ldots,u_{-j+2}].
\end{cases} \tag*{(42)} Π ( m ) := ⎩ ⎨ ⎧ m , 0 , − a + 1 1 Π ( u − j + 1 a + 1 ∂ f ) , m reduced , m = u a u − 1 , m = u − j u − j + 1 a f , j ≥ 2 , f ∈ k [ u , … , u − j + 2 ] . (42)
The last case assumes that u − j u_{-j} u − j occurs once. Recursion terminates because the highest derivative order decreases; the second case follows from u a u − 1 = ∂ ( u a + 1 ) / ( a + 1 ) u^a u_{-1}=\partial(u^{a+1})/(a+1) u a u − 1 = ∂ ( u a + 1 ) / ( a + 1 ) . Extend Π \Pi Π linearly.
Proposition 4.2 (Reduced representatives). The reduced monomials form a basis of F \mathcal{F} F . Reduction preserves weight and factor number, and
P = im Π ⊕ ∂ P , Π 2 = Π , Π ∂ = 0 , [ f ] = [ Π f ] . (43) \mathcal{P}=\operatorname{im}\Pi\oplus\partial\mathcal{P},\qquad\Pi^2=\Pi,\qquad\Pi\partial=0,\qquad[f]=[\Pi f]. \tag*{(43)} P = im Π ⊕ ∂ P , Π 2 = Π , Π ∂ = 0 , [ f ] = [ Π f ] . (43)
Proof. The terminating reduction proves spanning and preserves both gradings. For independence, let g g g be nonconstant and u − r u_{-r} u − r its highest-order jet. The part of ∂ g \partial g ∂ g containing the next jet is u − r − 1 ∂ g / ∂ u − r u_{-r-1}\partial g/\partial u_{-r} u − r − 1 ∂ g / ∂ u − r . It is nonzero and contains u − r − 1 u_{-r-1} u − r − 1 once; all remaining terms have lower maximum derivative order. Thus ∂ g \partial g ∂ g is not a linear combination of reduced monomials. For r = 0 r=0 r = 0 the same argument uses u − 1 g ′ ( u ) u_{-1}g'(u) u − 1 g ′ ( u ) .
We henceforth represent each class by its unique reduced polynomial. An operator D D D commuting with ∂ \partial ∂ acts on classes by [ f ] ↦ [ D f ] [f]\mapsto[Df] [ f ] ↦ [ D f ] ; its action in the reduced model is Π D \Pi D Π D . We use the same symbol for these induced actions when the domain is F \mathcal{F} F . Thus D f = 0 Df=0 D f = 0 in F \mathcal{F} F means Π D f = 0 \Pi Df=0 Π D f = 0 , not necessarily D f = 0 Df=0 D f = 0 in P \mathcal{P} P . Maps defined only on reduced representatives, such as the extraction map below, are not applied to arbitrary representatives of a class. Derivative-free evaluation is the separate map
π ( f ) : = f ( u , 0 , 0 , … ) . (44) \pi(f):=f(u,0,0,\ldots). \tag*{(44)} π ( f ) := f ( u , 0 , 0 , … ) . (44)
It descends to F \mathcal{F} F because π ∂ = 0 \pi\partial=0 π ∂ = 0 , and π Π = π \pi\Pi=\pi π Π = π . Thus Π \Pi Π selects a representative, while π \pi π extracts its component in k [ u ] \mathbb{k}[u] k [ u ] . Useful reductions are
Π ( u 2 u − 2 ) = − 2 u u − 1 2 , Π ( u u − 4 ) = u − 2 2 , Π ( u 2 u − 3 ) = u − 1 3 , Π ( u u − 1 u − 2 ) = − 1 2 u − 1 3 . (45) \begin{aligned}
\Pi(u^2u_{-2})&=-2uu_{-1}^2, & \Pi(uu_{-4})&=u_{-2}^2,\\
\Pi(u^2u_{-3})&=u_{-1}^3, & \Pi(uu_{-1}u_{-2})&=-\frac{1}{2}u_{-1}^3. \tag*{(45)}
\end{aligned} Π ( u 2 u − 2 ) Π ( u 2 u − 3 ) = − 2 u u − 1 2 , = u − 1 3 , Π ( u u − 4 ) Π ( u u − 1 u − 2 ) = u − 2 2 , = − 2 1 u − 1 3 . (45)
At odd weight, classes with one or two factors vanish. One-factor positive derivatives are exact; for two factors, [ u − i u − j ] = ( − 1 ) i [ u u − i − j ] [u_{-i}u_{-j}]=(-1)^i[uu_{-i-j}] [ u − i u − j ] = ( − 1 ) i [ u u − i − j ] , and reversing the factors makes this class zero when i + j i+j i + j is odd.
A partial inverse of the dispersionless action
The derivation X H X_{\mathrm{H}} X H corresponds to the dispersionless equation u t = 6 u u x u_t=6uu_x u t = 6 u u x , often called the Hopf equation. Its action modulo total derivatives contains a term that is diagonal after removing one factor u − 1 u_{-1} u − 1 . To describe this term, introduce the diagonal degree operator and its eigenvalue
E : = u − 1 ∂ u − 1 + ∑ j ≥ 2 ( j + 1 ) u − j ∂ u − j , E m = ϵ ( m ) m , ϵ ( m ) : = a 1 + ∑ j ≥ 2 ( j + 1 ) a j (46) E:=u_{-1}\partial_{u_{-1}}+\sum_{j\geq2}(j+1)u_{-j}\partial_{u_{-j}},\qquad E\mathfrak{m}=\epsilon(\mathfrak{m})\mathfrak{m},\qquad\epsilon(\mathfrak{m}):=a_1+\sum_{j\geq2}(j+1)a_j \tag*{(46)} E := u − 1 ∂ u − 1 + j ≥ 2 ∑ ( j + 1 ) u − j ∂ u − j , E m = ϵ ( m ) m , ϵ ( m ) := a 1 + j ≥ 2 ∑ ( j + 1 ) a j (46)
for m = ∏ j u − j a j \mathfrak{m}=\prod_j u_{-j}^{a_j} m = ∏ j u − j a j . This is a different grading from factor number: E E E assigns weights 0 , 1 , 3 , 4 , … 0,1,3,4,\ldots 0 , 1 , 3 , 4 , … to u , u − 1 , u − 2 , u − 3 , … u,u_{-1},u_{-2},u_{-3},\ldots u , u − 1 , u − 2 , u − 3 , … , whereas factor number assigns weight one to every jet. Define the splitting derivation
N : = 6 ∑ j ≥ 3 ∑ a = 2 j − 1 ( j a ) u − a u − j − 1 + a ∂ u − j . (47) N:=6\sum_{j\geq3}\sum_{a=2}^{j-1}\binom{j}{a}u_{-a}u_{-j-1+a}\partial_{u_{-j}}. \tag*{(47)} N := 6 j ≥ 3 ∑ a = 2 ∑ j − 1 ( a j ) u − a u − j − 1 + a ∂ u − j . (47)
Expansion of ∂ ( u u − 1 ) \partial(uu_{-1}) ∂ ( u u − 1 ) on the generators gives
X H = 6 u ∂ + 6 u − 1 E + N . (48) X_{\mathrm{H}}=6u\partial+6u_{-1}E+N. \tag*{(48)} X H = 6 u ∂ + 6 u − 1 E + N . (48)
For example, the terms outside u ∂ u\partial u ∂ in the unscaled expansion are u − 1 2 u_{-1}^2 u − 1 2 at u − 1 u_{-1} u − 1 , and ( j + 1 ) u − 1 u − j + ∑ a = 2 j − 1 ( j a ) u − a u − j − 1 + a (j+1)u_{-1}u_{-j}+\sum_{a=2}^{j-1}\binom{j}{a}u_{-a}u_{-j-1+a} ( j + 1 ) u − 1 u − j + ∑ a = 2 j − 1 ( a j ) u − a u − j − 1 + a at u − j u_{-j} u − j for j ≥ 2 j\geq2 j ≥ 2 . Since u ∂ m ≡ − u − 1 m u\partial\mathfrak{m}\equiv-u_{-1}\mathfrak{m} u ∂ m ≡ − u − 1 m modulo ∂ P \partial\mathcal{P} ∂ P , (48) implies
Π X H m = Π ( 6 ( ϵ ( m ) − 1 ) u − 1 m + N m ) . (49) \Pi X_{\mathrm{H}}\mathfrak{m}=\Pi\left(6(\epsilon(\mathfrak{m})-1)u_{-1}\mathfrak{m}+N\mathfrak{m}\right). \tag*{(49)} Π X H m = Π ( 6 ( ϵ ( m ) − 1 ) u − 1 m + N m ) . (49)
The following map inverts the contribution 6 ( ϵ ( m ) − 1 ) u − 1 m 6(\epsilon(\mathfrak{m})-1)u_{-1}\mathfrak{m} 6 ( ϵ ( m ) − 1 ) u − 1 m in (49) on reduced monomials containing derivatives.
Definition 4.3 (Partial inverse of the diagonal term). For a reduced monomial m \mathfrak{m} m , set
C ( m ) : = { m ′ 6 ( ϵ ( m ′ ) − 1 ) , m = u − 1 m ′ , m ′ reduced and containing a positive-order jet , 0 , otherwise . (50) \mathcal{C}(\mathfrak{m}):=
\begin{cases}
\dfrac{\mathfrak{m}'}{6(\epsilon(\mathfrak{m}')-1)}, & \mathfrak{m}=u_{-1}\mathfrak{m}',\ \mathfrak{m}'\text{ reduced and containing a positive-order jet},\\
0, & \text{otherwise}.
\end{cases} \tag*{(50)} C ( m ) := ⎩ ⎨ ⎧ 6 ( ϵ ( m ′ ) − 1 ) m ′ , 0 , m = u − 1 m ′ , m ′ reduced and containing a positive-order jet , otherwise . (50) Extend C \mathcal{C} C k \mathbb{k} k -linearly on reduced representatives.
In the nonzero case ϵ ( m ) > 1 \epsilon(\mathfrak{m})>1 ϵ ( m ) > 1 , so this is division by a positive integer. The map lowers weight by three and factor number by one; its image contains a positive-order jet and at least two factors. For example,
C ( u − 1 3 ) = 1 6 u − 1 2 , C ( u − 1 u − 2 2 ) = 1 3 u − 2 2 , C ( u a u − 1 2 ) = C ( u − 2 2 ) = C ( u a ) = 0. (51) \mathcal{C}(u_{-1}^{3})=\frac{1}{6}u_{-1}^{2},\qquad\mathcal{C}(u_{-1}u_{-2}^{2})=\frac{1}{3}u_{-2}^{2},\qquad\mathcal{C}(u^{a}u_{-1}^{2})=\mathcal{C}(u_{-2}^{2})=\mathcal{C}(u^{a})=0. \tag*{(51)} C ( u − 1 3 ) = 6 1 u − 1 2 , C ( u − 1 u − 2 2 ) = 3 1 u − 2 2 , C ( u a u − 1 2 ) = C ( u − 2 2 ) = C ( u a ) = 0. (51)
Lemma 4.4 (Extraction identity). On reduced representatives,
C Π X H = 1 − π + C Π N . (52) \mathcal{C}\Pi X_{\mathrm{H}}=1-\pi+\mathcal{C}\Pi N. \tag*{(52)} C Π X H = 1 − π + C Π N . (52)
Proof. By linearity it suffices to consider a reduced monomial m \mathfrak{m} m . Applying C \mathcal{C} C to (49) gives
C Π X H m = 6 ( ϵ ( m ) − 1 ) C Π ( u − 1 m ) + C Π N m . \mathcal{C}\Pi X_{\mathrm{H}}\mathfrak{m}=6(\epsilon(\mathfrak{m})-1)\mathcal{C}\Pi(u_{-1}\mathfrak{m})+\mathcal{C}\Pi N\mathfrak{m}. C Π X H m = 6 ( ϵ ( m ) − 1 ) C Π ( u − 1 m ) + C Π N m .
If m \mathfrak{m} m contains a positive-order jet, then u − 1 m u_{-1}\mathfrak{m} u − 1 m is reduced: its highest-order jet still occurs at least twice. Hence Π ( u − 1 m ) = u − 1 m \Pi(u_{-1}\mathfrak{m})=u_{-1}\mathfrak{m} Π ( u − 1 m ) = u − 1 m , and Definition 4.3 gives
6 ( ϵ ( m ) − 1 ) C Π ( u − 1 m ) = m . 6(\epsilon(\mathfrak{m})-1)\mathcal{C}\Pi(u_{-1}\mathfrak{m})=\mathfrak{m}. 6 ( ϵ ( m ) − 1 ) C Π ( u − 1 m ) = m .
Here ϵ ( m ) > 1 \epsilon(\mathfrak{m})>1 ϵ ( m ) > 1 , and π ( m ) = 0 \pi(\mathfrak{m})=0 π ( m ) = 0 , so this is exactly ( 1 − π ) m (1-\pi)\mathfrak{m} ( 1 − π ) m .
Otherwise m = u a \mathfrak{m}=u^{a} m = u a for some a ≥ 0 a\geq0 a ≥ 0 . In this case u − 1 m = ∂ ( u a + 1 ) / ( a + 1 ) u_{-1}\mathfrak{m}=\partial(u^{a+1})/(a+1) u − 1 m = ∂ ( u a + 1 ) / ( a + 1 ) , so Π ( u − 1 m ) = 0 \Pi(u_{-1}\mathfrak{m})=0 Π ( u − 1 m ) = 0 , while ( 1 − π ) m = 0 (1-\pi)\mathfrak{m}=0 ( 1 − π ) m = 0 as well. Thus in both cases the first term in the displayed identity equals ( 1 − π ) m (1-\pi)\mathfrak{m} ( 1 − π ) m , proving (52) . □ \square □
For example,
N ( u u − 2 3 ) = 36 u u − 3 u − 2 2 , Π N ( u u − 2 3 ) = − 12 u − 1 u − 2 3 , C Π N ( u u − 2 3 ) = − 1 4 u − 2 3 . N(uu_{-2}^{3})=36uu_{-3}u_{-2}^{2},\qquad\Pi N(uu_{-2}^{3})=-12u_{-1}u_{-2}^{3},\qquad\mathcal{C}\Pi N(uu_{-2}^{3})=-\frac{1}{4}u_{-2}^{3}. N ( u u − 2 3 ) = 36 u u − 3 u − 2 2 , Π N ( u u − 2 3 ) = − 12 u − 1 u − 2 3 , C Π N ( u u − 2 3 ) = − 4 1 u − 2 3 .
The map C Π N \mathcal{C}\Pi N C Π N preserves factor number but lowers the higher-jet exponents in a finite order. We prove this decrease, together with the contribution of R R R , in the next proposition.
The explicit construction and the main theorem
The inverse formula on each weight space
Write F w \mathcal{F}_{w} F w for the component of conformal weight w w w , represented by reduced monomials of that weight. It is a free k k k -module of finite rank. On this reduced model of F \mathcal{F} F , define
M : = C Π ( N + R ) . (53) M:=\mathcal{C}\Pi(N+R). \tag*{(53)} M := C Π ( N + R ) . (53)
The decomposition X ℏ = X H + R X_{\hbar}=X_{\mathrm{H}}+R X ℏ = X H + R and (52) give
C Π X ℏ = 1 + M − π . (54) \mathcal{C}\Pi X_{\hbar}=1+M-\pi. \tag*{(54)} C Π X ℏ = 1 + M − π . (54)
The parameter degree below is ordinary polynomial degree in ℏ \hbar ℏ . It measures the coefficients, whereas conformal weight and factor number measure the jet monomials; ℏ \hbar ℏ has degree zero for the latter two gradings.
Proposition 5.1 (Finite inverse on each weight space). On every weight space F w \mathcal{F}_w F w , the operator M M M is nilpotent. For k ≥ 1 k \ge1 k ≥ 1 , the reduced polynomial
F k : = ( 1 + M ) − 1 u k = ∑ r = 0 d 2 k − 1 ( − M ) r u k , d w : = rank K F w (55) F_k := (1+M)^{-1}u^k = \sum_{r=0}^{d_{2k-1}}(-M)^r u^k,\qquad d_w := \operatorname{rank}_{\mathbb{K}}\mathcal{F}_w \tag*{(55)} F k := ( 1 + M ) − 1 u k = r = 0 ∑ d 2 k − 1 ( − M ) r u k , d w := rank K F w (55) has π F k = u k \pi F_k=u^k π F k = u k . Every remaining term has between two and k − 1 k-1 k − 1 factors. A coefficient of a term with m m m factors has degree at most k − m k-m k − m in ℏ \hbar ℏ .
Proof. Step 1. A finite decreasing order. For a reduced monomial m = ∏ j = 0 w − 2 u − j a j \mathfrak{m}=\prod_{j=0}^{w-2}u_{-j}^{a_j} m = ∏ j = 0 w − 2 u − j a j of weight w w w , put
μ w ( m ) : = ( deg f m , a w − 2 , a w − 3 , … , a 2 ) , \mu_w(\mathfrak{m}):=(\deg_f\mathfrak{m},a_{w-2},a_{w-3},\ldots,a_2), μ w ( m ) := ( deg f m , a w − 2 , a w − 3 , … , a 2 ) ,
with the higher-jet list empty for w < 4 w<4 w < 4 , and compare the tuples lexicographically. At fixed weight and factor number, the exponents of u − j u_{-j} u − j for j ≥ 2 j\ge2 j ≥ 2 determine those of u , u − 1 u,u_{-1} u , u − 1 , so this orders the finite reduced basis without ties.
A term of N N N replaces one u − j u_{-j} u − j by two jets of strictly smaller derivative orders. It increases factor number by one and decreases the higher-jet tuple. Each integration-by-parts step in (41) decreases that tuple again, whereas extraction removes only u − 1 u_{-1} u − 1 . Thus C Π N \mathcal{C}\Pi N C Π N preserves factor number and strictly lowers the tuple. The part C Π R \mathcal{C}\Pi R C Π R decreases factor number: R R R does not increase it, reduction preserves it, and extraction removes one factor. Every term of M m M\mathfrak{m} M m therefore has a smaller tuple.
List the reduced basis as m 1 , … , m d w \mathfrak{m}_1,\ldots,\mathfrak{m}_{d_w} m 1 , … , m d w in increasing order. Then M m i ∈ span { m 1 , … , m i − 1 } M\mathfrak{m}_i\in\operatorname{span}\{\mathfrak{m}_1,\ldots,\mathfrak{m}_{i-1}\} M m i ∈ span { m 1 , … , m i − 1 } , so M d w = 0 M^{d_w}=0 M d w = 0 , proving (55) . Also π M = 0 \pi M=0 π M = 0 , and the image of extraction has a positive-order jet and at least two factors. Starting from u k u^k u k , the first nonzero step decreases factor number because N u k = 0 Nu^k=0 N u k = 0 . This proves the asserted range of factors.
Step 2. The parameter degree. Lemma 2.4 shows that a single-mode symbol coefficient of differential order b b b has ℏ \hbar ℏ -degree b b b . In (31) , a product term of order b b b has parameter degree b − 1 b-1 b − 1 and coordinate degree at most two; a contraction term has parameter degree b b b and coordinate degree at most one. Thus a term of coordinate degree a a a and differential order b b b has parameter degree at most 1 − a + b 1-a+b 1 − a + b .
For a term of R R R the factor shift is a − b ≤ 0 a-b\le0 a − b ≤ 0 . Extraction lowers it once more, so a term of C Π R \mathcal{C}\Pi R C Π R lowering factor number by r ≥ 1 r\ge1 r ≥ 1 has parameter degree at most r r r . The component C Π N \mathcal{C}\Pi N C Π N preserves factor number and is independent of ℏ \hbar ℏ . Along a chain from k k k factors to m m m factors, these degrees sum to at most k − m k-m k − m . □
Proposition 5.2. For every k ≥ 1 k\ge1 k ≥ 1 , the reduced polynomial F k F_k F k defined by (55) satisfies
C Π X ℏ F k = 0. (56) \mathcal{C}\Pi X_\hbar F_k=0. \tag*{(56)} C Π X ℏ F k = 0. (56)
Proof. By Proposition 5.1, M d 2 k = 0 M^{d_{2k}}=0 M d 2 k = 0 on F 2 k \mathcal{F}_{2k} F 2 k . Multiplying the finite sum in (55) by 1 + M 1+M 1 + M therefore gives
( 1 + M ) F k = ( 1 − ( − M ) d 2 k ) u k = u k . (57) (1+M)F_k=\left(1-(-M)^{d_{2k}}\right)u^k=u^k. \tag*{(57)} ( 1 + M ) F k = ( 1 − ( − M ) d 2 k ) u k = u k . (57)
Every term in the image of M M M contains a positive-order jet, since M = C Π ( N + R ) M=\mathcal{C}\Pi(N+R) M = C Π ( N + R ) . Thus π M = 0 \pi M=0 π M = 0 , and applying π \pi π to (55) gives π F k = u k \pi F_k=u^k π F k = u k . Finally, the operator identity (54) , applied to the reduced polynomial F k F_k F k , yields
C Π X ℏ F k = ( 1 + M ) F k − π F k = u k − u k = 0. \mathcal{C}\Pi X_{\hbar}F_k=(1+M)F_k-\pi F_k=u^k-u^k=0. C Π X ℏ F k = ( 1 + M ) F k − π F k = u k − u k = 0.
Proposition 5.2 gives only the extracted conservation equation. The main theorem requires Π X ℏ F k = 0 \Pi X_{\hbar}F_k=0 Π X ℏ F k = 0 , or equivalently X ℏ [ F k ] = 0 X_{\hbar}[F_k]=0 X ℏ [ F k ] = 0 in F \mathcal{F} F ; this does not require X ℏ F k = 0 X_{\hbar}F_k=0 X ℏ F k = 0 as a polynomial. Since C \mathcal{C} C has a kernel, it cannot simply be cancelled. § 6 proves an injective test for compatible classical residual pairs. § 7 lifts that test to the quantum flows and applies it to ( X ℏ F k , Y ℏ F k ) (X_{\hbar}F_k,Y_{\hbar}F_k) ( X ℏ F k , Y ℏ F k ) for k ≥ 2 k\geq2 k ≥ 2 , thereby removing extraction from the conservation condition. The case k = 1 k=1 k = 1 is checked directly below.
Theorem 5.3 (Main theorem). Let F k F_k F k be given by (55) and put H 2 k : = σ ( F k ) H_{2k}:=\sigma(F_k) H 2 k := σ ( F k ) for k ≥ 1 k\geq1 k ≥ 1 . Then:
(i) Conservation and commutativity. The polynomials F k F_k F k have weight 2 k 2k 2 k and satisfy
X ℏ [ F k ] = 0 in F , (58) X_{\hbar}[F_k]=0\quad\text{in }\mathcal{F}, \tag*{(58)} X ℏ [ F k ] = 0 in F , (58) and
H 2 k ( 0 ) H 2 l ∈ T V ( k , l ≥ 1 ) . (59) H_{2k}(0)H_{2l}\in T\mathcal{V}\quad(k,l\geq1). \tag*{(59)} H 2 k ( 0 ) H 2 l ∈ T V ( k , l ≥ 1 ) . (59) (ii) Uniqueness. At every positive weight w w w ,
ker ( X ℏ : F w → F w + 3 ) = { K ⋅ [ F k ] , w = 2 k , 0 , w odd . \operatorname{ker}(X_{\hbar}:\mathcal{F}_w\to\mathcal{F}_{w+3})=
\begin{cases}
\mathbb{K}\cdot[F_k], & w=2k,\\
0, & w\text{ odd}.
\end{cases} ker ( X ℏ : F w → F w + 3 ) = { K ⋅ [ F k ] , 0 , w = 2 k , w odd . Thus the normalized reduced conserved density with derivative-free part u k u^k u k is F k F_k F k .
(iii) Classical limit. With L : = ∂ 2 + u L:=\partial^2+u L := ∂ 2 + u ,
F k ∣ ℏ = 0 = I k , I k : = 4 k ( 2 k k ) Π Res ∂ L k − 1 2 . (60) F_k\big|_{\hbar=0}=I_k,\qquad I_k:=\frac{4^k}{\binom{2k}{k}}\Pi\operatorname{Res}_{\partial}L^{k-\frac{1}{2}}. \tag*{(60)} F k ℏ = 0 = I k , I k := ( k 2 k ) 4 k Π Res ∂ L k − 2 1 . (60) (iv) Every central charge. The inverse scaling (87) defines polynomials F k ( c ) F_k^{(c)} F k ( c ) over Q [ c ] \mathbb{Q}[c] Q [ c ] . The vectors H 2 k ( c ) : = σ ω ( F k ( c ) ) H_{2k}^{(c)}:=\sigma_\omega(F_k^{(c)}) H 2 k ( c ) := σ ω ( F k ( c ) ) , where σ ω \sigma_\omega σ ω uses jet coordinates v − j v_{-j} v − j for T ω T\omega T ω , have commuting classes in V c / T V c \mathcal{V}^c/T\mathcal{V}^c V c / T V c for every c ∈ C c\in\mathbb{C} c ∈ C . A coefficient of an m m m -factor term of F k ( c ) F_k^{(c)} F k ( c ) has degree at most k − m k-m k − m in c c c . The centralizer of [ ω ( − 1 ) ω ] [\omega(-1)\omega] [ ω ( − 1 ) ω ] is spanned by [ H 2 k ( c ) ] [H_{2k}^{(c)}] [ H 2 k ( c ) ] in weight 2 k 2k 2 k and is zero in positive odd weights, including at c = 0 c=0 c = 0 .
For k = 1 k=1 k = 1 , the formula gives F 1 = u F_1=u F 1 = u and H 2 = U H_2=U H 2 = U . Equation (36) shows X ℏ u ∈ ∂ P X_{\hbar}u\in\partial\mathcal{P} X ℏ u ∈ ∂ P . Since U ( 0 ) = ℏ T U(0)=\hbar T U ( 0 ) = ℏ T , its class is central in V / T V \mathcal{V}/T\mathcal{V} V / T V ; hence its conservation and commutativity are immediate. Also I 1 = u I_1=u I 1 = u by the residue formula. The residual argument below is needed only for k ≥ 2 k\geq2 k ≥ 2 .
Proposition 5.1 already proves the explicit formula, its normalization, and the parameter-degree bound. The weight-six calculation below gives a second commuting zero mode. § 6 proves the classical exactness needed to remove C \mathcal{C} C from (56) ; § 7 then proves conservation, uniqueness, commutativity, and the claimed specializations, completing Theorem 5.3.
The weight-six conserved quantity and its action
The weight-six calculation supplies more than the first nontrivial example: it produces a second operator Y ℏ Y_{\hbar} Y ℏ commuting with X ℏ X_{\hbar} X ℏ . This commutation relation will make the two conservation residuals compatible, as required by the test in § 6 . For weight four, M u 2 = 0 Mu^2=0 M u 2 = 0 and F 2 = u 2 F_2=u^2 F 2 = u 2 . To find the weight-six member, apply (38) to u 3 u^3 u 3 . Both X ℏ u 3 X_{\hbar}u^3 X ℏ u 3 and X ℏ u − 1 2 X_{\hbar}u^2_{-1} X ℏ u − 1 2 have weight nine. All their terms with at most two factors vanish after reduction, by the one- and two-factor observation following (45) . The factor-preserving part of R u 3 Ru^3 R u 3 is given by (38) , while R u − 1 2 Ru^2_{-1} R u − 1 2 has at most two factors. Hence
Π R u 3 = Π ( ( 3 + 11 2 ℏ ) u − 3 2 + 9 ℏ u − 1 u − 2 ) = ( 3 + ℏ ) u − 1 3 , Π R ( u − 1 2 ) = 0. \Pi Ru^3=\Pi\left((3+\frac{11}{2}\hbar)u^2_{-3}+9\hbar u_{-1}u_{-2}\right)=(3+\hbar)u^3_{-1},\qquad\Pi R(u^2_{-1})=0. Π R u 3 = Π ( ( 3 + 2 11 ℏ ) u − 3 2 + 9ℏ u − 1 u − 2 ) = ( 3 + ℏ ) u − 1 3 , Π R ( u − 1 2 ) = 0.
Since C ( u − 1 3 ) = u − 1 2 / 6 C(u^3_{-1})=u^2_{-1}/6 C ( u − 1 3 ) = u − 1 2 /6 and N N N annihilates u 3 , u − 1 2 u^3,u^2_{-1} u 3 , u − 1 2 , the inverse formula gives
F 3 = u 3 − ( 1 2 + ℏ 6 ) u − 1 2 . (61) F_3=u^3-\left(\frac{1}{2}+\frac{\hbar}{6}\right)u^2_{-1}. \tag*{(61)} F 3 = u 3 − ( 2 1 + 6 ℏ ) u − 1 2 . (61)
Also Π X H ( u 3 ) = 0 \Pi X_H(u^3)=0 Π X H ( u 3 ) = 0 and Π X H ( u − 1 2 ) = 6 u − 1 3 \Pi X_H(u^2_{-1})=6u^3_{-1} Π X H ( u − 1 2 ) = 6 u − 1 3 . Consequently,
Π X ℏ F 3 = ( 3 + ℏ ) u − 1 3 − 6 ( 1 2 + ℏ 6 ) u − 1 3 = 0. (62) \Pi X_{\hbar}F_3=(3+\hbar)u^3_{-1}-6\left(\frac{1}{2}+\frac{\hbar}{6}\right)u^3_{-1}=0. \tag*{(62)} Π X ℏ F 3 = ( 3 + ℏ ) u − 1 3 − 6 ( 2 1 + 6 ℏ ) u − 1 3 = 0. (62)
Choose the following ordered weight-six vector, which represents the same local functional as the symmetric vector σ ( F 3 ) \sigma(F_3) σ ( F 3 ) modulo translation:
H ~ 6 : = U ( − 1 ) ( U ( − 1 ) U ) + ( 1 2 + ℏ 6 ) ( T 2 U ) ( − 1 ) U . (63) \widetilde{H}_6:=U(-1)(U(-1)U)+\left(\frac{1}{2}+\frac{\hbar}{6}\right)(T^2U)(-1)U. \tag*{(63)} H 6 := U ( − 1 ) ( U ( − 1 ) U ) + ( 2 1 + 6 ℏ ) ( T 2 U ) ( − 1 ) U . (63)
Indeed, the two terms are U 0 3 1 U^3_{0}{}_{1} U 0 3 1 and U − 2 U 0 1 U_{-2}U_{0}{}_{1} U − 2 U 0 1 , respectively, and
σ ( F 3 ) − H ~ 6 = − ( 1 2 + ℏ 6 ) T ( U − 1 U 0 1 ) . (64) \sigma(F_3)-\widetilde{H}_6=-\left(\frac{1}{2}+\frac{\hbar}{6}\right)T(U_{-1}U_{0}{}_{1}). \tag*{(64)} σ ( F 3 ) − H 6 = − ( 2 1 + 6 ℏ ) T ( U − 1 U 0 1 ) . (64)
Both vectors therefore have the same zero-mode operator. By (62) , H 4 ( 0 ) H ~ 6 ∈ T V H_4(0)\widetilde{H}_6\in T\mathcal{V} H 4 ( 0 ) H 6 ∈ T V , so [ a ( 0 ) , b ( 0 ) ] = ( a ( 0 ) b ) ( 0 ) [a(0),b(0)]=(a(0)b)(0) [ a ( 0 ) , b ( 0 )] = ( a ( 0 ) b ) ( 0 ) gives [ H 4 ( 0 ) , H ~ 6 ( 0 ) ] = 0 [H_4(0),\widetilde{H}_6(0)]=0 [ H 4 ( 0 ) , H 6 ( 0 )] = 0 . The following operators consequently commute after localization and, by freeness, over k \mathbf{k} k :
Y ℏ : = σ − 1 1 ℏ H ~ 6 ( 0 ) σ , [ X ℏ , Y ℏ ] = 0. (65) Y_{\hbar}:=\sigma^{-1}\frac{1}{\hbar}\widetilde{H}_6(0)\sigma,\qquad[X_{\hbar},Y_{\hbar}]=0. \tag*{(65)} Y ℏ := σ − 1 ℏ 1 H 6 ( 0 ) σ , [ X ℏ , Y ℏ ] = 0. (65)
The factor 2 makes the coefficient of u − 5 u_{-5} u − 5 in the classical limit equal to one; see (68) . The operator has coefficients in k \mathbf{k} k : the specialization at ℏ = 0 \hbar=0 ℏ = 0 is a commutative vertex algebra, whose nonnegative products vanish. Hence H ~ 6 ( 0 ) V ⊆ ℏ V \widetilde{H}_6(0)\mathcal{V}\subseteq\hbar\mathcal{V} H 6 ( 0 ) V ⊆ ℏ V . Its weight shift is five, and it commutes with translation, so it also acts on F \mathcal{F} F .
Classical KdV flows and a compatibility criterion
The classical coefficient algebra is P 0 = P / ℏ P \mathcal{P}_0=\mathcal{P}/\hbar\mathcal{P} P 0 = P /ℏ P . Put F 0 : = P 0 / ∂ P 0 \mathcal{F}_0:=\mathcal{P}_0/\partial\mathcal{P}_0 F 0 := P 0 / ∂ P 0 . The reduced basis identifies F \mathcal{F} F with k ⊗ Q F 0 \mathbb{k}\otimes_{\mathbb{Q}}\mathcal{F}_0 k ⊗ Q F 0 ; write F 0 , w \mathcal{F}_{0,w} F 0 , w for its classical weight-w w w part. Define the evolutionary derivations on the generator u u u by
X 0 u : = u − 3 + 6 u u − 1 , Y 0 u : = u − 5 + 10 u u − 3 + 20 u − 1 u − 2 + 30 u 2 u − 1 . (66) X_0u:=u_{-3}+6uu_{-1},
\qquad
Y_0u:=u_{-5}+10uu_{-3}+20u_{-1}u_{-2}+30u^2u_{-1}. \tag*{(66)} X 0 u := u − 3 + 6 u u − 1 , Y 0 u := u − 5 + 10 u u − 3 + 20 u − 1 u − 2 + 30 u 2 u − 1 . (66)
and extend uniquely by [ X 0 , ∂ ] = [ Y 0 , ∂ ] = 0 [X_0,\partial]=[Y_0,\partial]=0 [ X 0 , ∂ ] = [ Y 0 , ∂ ] = 0 and the Leibniz rule. Thus X 0 u − j = ∂ j X 0 u X_0u_{-j}=\partial^jX_0u X 0 u − j = ∂ j X 0 u and Y 0 u − j = ∂ j Y 0 u Y_0u_{-j}=\partial^jY_0u Y 0 u − j = ∂ j Y 0 u . On F 0 \mathcal{F}_0 F 0 , we identify classes with their reduced representatives, so their actions include reduction by Π \Pi Π . We first identify these derivations as the classical limits of the weight-four and weight-six VOA actions and relate them to the KdV hierarchy. We then prove that, for a pair ( a , b ) (a,b) ( a , b ) in the required odd weights, the two conditions Y 0 a = X 0 b Y_0a=X_0b Y 0 a = X 0 b and C a = 0 \mathcal{C}a=0 C a = 0 force a = b = 0 a=b=0 a = b = 0 . This is the classical ingredient needed to remove C \mathcal{C} C from the conservation equation.
Classical limits and KdV conserved quantities
The specialization of V \mathcal{V} V at ℏ = 0 \hbar=0 ℏ = 0 is commutative. Dividing the singular products in (3) by ℏ \hbar ℏ and taking this specialization gives the Poisson vertex bracket
{ u λ u } = ( ∂ + 2 λ ) u + 1 2 λ 3 . (67) \{u_\lambda u\}=(\partial+2\lambda)u+\frac{1}{2}\lambda^3. \tag*{(67)} { u λ u } = ( ∂ + 2 λ ) u + 2 1 λ 3 . (67)
Thus the constant terms of X ℏ , Y ℏ X_\hbar,Y_\hbar X ℏ , Y ℏ are evolutionary derivations of the commutative differential algebra. Let J J J be the second KdV Hamiltonian operator,
J : = 1 2 ∂ 3 + 2 u ∂ + u − 1 . J:=\frac{1}{2}\partial^3+2u\partial+u_{-1}. J := 2 1 ∂ 3 + 2 u ∂ + u − 1 .
The classical symbols of H 4 , H ~ 6 H_4,\widetilde{H}_6 H 4 , H 6 are u 2 u^2 u 2 and u 3 + 1 2 u u − 2 u^3+\frac{1}{2}uu_{-2} u 3 + 2 1 u u − 2 respectively. For this classical identification only, write the variational derivative as
δ f : = ∑ j ≥ 0 ( − ∂ ) j ∂ u − j f . \delta f:=\sum_{j\geq0}(-\partial)^j\partial_{u_{-j}}f. δ f := j ≥ 0 ∑ ( − ∂ ) j ∂ u − j f .
It vanishes on total derivatives and therefore depends only on the class of f f f . The Hamiltonian formula u ↦ J δ f u\mapsto J\delta f u ↦ J δ f [10 Section 1.4] then gives
X ℏ = X 0 + O ( ℏ ) , Y ℏ = Y 0 + O ( ℏ ) . (68) X_\hbar=X_0+O(\hbar),
\qquad
Y_\hbar=Y_0+O(\hbar). \tag*{(68)} X ℏ = X 0 + O ( ℏ ) , Y ℏ = Y 0 + O ( ℏ ) . (68)
Indeed
J δ ( u 2 ) = u − 3 + 6 u u − 1 , 2 J δ ( u 3 + 1 2 u u − 2 ) = u − 5 + 10 u u − 3 + 20 u − 1 u − 2 + 30 u 2 u − 1 . J\delta(u^2)=u_{-3}+6uu_{-1},
\qquad
2J\delta\left(u^3+\frac{1}{2}uu_{-2}\right)=u_{-5}+10uu_{-3}+20u_{-1}u_{-2}+30u^2u_{-1}. J δ ( u 2 ) = u − 3 + 6 u u − 1 , 2 J δ ( u 3 + 2 1 u u − 2 ) = u − 5 + 10 u u − 3 + 20 u − 1 u − 2 + 30 u 2 u − 1 .
The second expression confirms the factor 2 in (65) . The limiting first-order property follows from the commutative specialization of the ( − 1 ) (-1) ( − 1 ) -product; it does not assert that the higher ℏ \hbar ℏ -coefficients obey polynomial Leibniz rules.
We use the classical Gelfand–Dickey construction [12 ] . A formulation in the present differential-algebra language, including both the KdV and KP hierarchies, is given in [10 Section 3] . Only the third and fifth KdV flows and their common conserved densities are needed here.
Lemma 6.1 (Classical conservation and normalization). For L : = ∂ 2 + u L := \partial^2+u L := ∂ 2 + u , the reduced polynomials I k I_k I k of (60) have weight 2 k 2k 2 k and satisfy
π I k = u k , X 0 I k = Y 0 I k = 0 in F 0 . (69) \pi I_k=u^k,\qquad X_0I_k=Y_0I_k=0\quad\text{in }\mathcal{F}_0. \tag*{(69)} π I k = u k , X 0 I k = Y 0 I k = 0 in F 0 . (69) Moreover [ X 0 , Y 0 ] = 0 [X_0,Y_0]=0 [ X 0 , Y 0 ] = 0 . In particular, I 1 = u I_1=u I 1 = u , I 2 = u 2 I_2=u^2 I 2 = u 2 , and I 3 = u 3 − 1 2 u 1 2 I_3=u^3-\frac{1}{2}u_1^2 I 3 = u 3 − 2 1 u 1 2 .
Proof. The classical result supplies the commuting Lax flows and their residue integrals. We record the normalization check to identify them with (66) . Formal multiplication uses
∂ a f = ∑ r ≥ 0 ( a r ) ( ∂ r f ) ∂ a − r , a ∈ Z . (70) \partial^a f=\sum_{r\geq0}\binom{a}{r}(\partial^r f)\partial^{a-r},\qquad a\in\mathbb{Z}. \tag*{(70)} ∂ a f = r ≥ 0 ∑ ( r a ) ( ∂ r f ) ∂ a − r , a ∈ Z . (70) The monic root Q : = L 1 / 2 = ∂ + ∑ r ≥ 1 q r ∂ − r Q:=L^{1/2}=\partial+\sum_{r\geq1}q_r\partial^{-r} Q := L 1/2 = ∂ + ∑ r ≥ 1 q r ∂ − r is uniquely determined by Q 2 = L Q^2=L Q 2 = L : each new coefficient occurs with coefficient two. Operators have order bounded above, so every coefficient in a product is a finite sum. Let ( ⋅ ) + (\cdot)_+ ( ⋅ ) + retain nonnegative powers of ∂ \partial ∂ , and write P 3 : = 4 ( Q 3 ) + P_3:=4(Q^3)_+ P 3 := 4 ( Q 3 ) + , P 5 : = 16 ( Q 5 ) + P_5:=16(Q^5)_+ P 5 := 16 ( Q 5 ) + . Explicitly,
P 3 = 4 ∂ 3 + 6 u ∂ + 3 u − 1 , P_3=4\partial^3+6u\partial+3u_{-1}, P 3 = 4 ∂ 3 + 6 u ∂ + 3 u − 1 , P 5 = 16 ∂ 5 + 40 u ∂ 3 + 60 u − 1 ∂ 2 + ( 50 u − 2 + 30 u 2 ) ∂ + 15 u − 3 + 30 u u − 1 . (71) P_5=16\partial^5+40u\partial^3+60u_{-1}\partial^2+(50u_{-2}+30u^2)\partial+15u_{-3}+30uu_{-1}. \tag*{(71)} P 5 = 16 ∂ 5 + 40 u ∂ 3 + 60 u − 1 ∂ 2 + ( 50 u − 2 + 30 u 2 ) ∂ + 15 u − 3 + 30 u u − 1 . (71) Their commutators with L L L have order zero, with coefficients exactly u − 3 + 6 u u − 1 u_{-3}+6uu_{-1} u − 3 + 6 u u − 1 and u − 5 + 10 u u − 3 + 20 u − 1 u − 2 + 30 u 2 u − 1 u_{-5}+10uu_{-3}+20u_{-1}u_{-2}+30u^2u_{-1} u − 5 + 10 u u − 3 + 20 u − 1 u − 2 + 30 u 2 u − 1 . Hence
X 0 L = [ P 3 , L ] , Y 0 L = [ P 5 , L ] . (72) X_0L=[P_3,L],\qquad Y_0L=[P_5,L]. \tag*{(72)} X 0 L = [ P 3 , L ] , Y 0 L = [ P 5 , L ] . (72) This identifies the normalizations of the commuting classical flows. The Lax identities extend to Q 2 k − 1 Q^{2k-1} Q 2 k − 1 . Taking residues gives conservation because the residue of a pseudodifferential commutator is a total derivative. This standard trace identity follows directly from (70) : for r = m + n + 1 ≥ 0 r=m+n+1\geq0 r = m + n + 1 ≥ 0 the residue of [ a ∂ m , b ∂ n ] [a\partial^m,b\partial^n] [ a ∂ m , b ∂ n ] is ( m r ) a ∂ r b − ( n r ) b ∂ r a \binom{m}{r}a\partial^r b-\binom{n}{r}b\partial^r a ( r m ) a ∂ r b − ( r n ) b ∂ r a , which is zero modulo total derivatives since ( n r ) = ( − 1 ) r ( r − n − 1 r ) \binom{n}{r}=(-1)^r\binom{r-n-1}{r} ( r n ) = ( − 1 ) r ( r r − n − 1 ) . Finally, setting all positive-order jets to zero gives
π Res ∂ L k − 1 / 2 = ( k − 1 / 2 k ) u k = ( 2 k k ) 4 k u k . \pi\operatorname{Res}_{\partial}L^{k-1/2}=\binom{k-1/2}{k}u^k=\frac{\binom{2k}{k}}{4^k}u^k. π Res ∂ L k − 1/2 = ( k k − 1/2 ) u k = 4 k ( k 2 k ) u k . Assigning weight one to ∂ \partial ∂ makes Q Q Q homogeneous of weight one; its ∂ − 1 \partial^{-1} ∂ − 1 coefficient in Q 2 k − 1 Q^{2k-1} Q 2 k − 1 therefore has weight 2 k 2k 2 k . Reduction preserves this weight and the derivative-free term, proving the assertions.
Kernel after reduction and extraction
Let D 3 : = ∑ j ≥ 0 u − j − 3 ∂ u − j D_3 := \sum_{j\geq0} u_{-j-3}\partial u_{-j} D 3 := ∑ j ≥ 0 u − j − 3 ∂ u − j , so X 0 = X H + D 3 X_0 = X_H + D_3 X 0 = X H + D 3 . Composing X 0 X_0 X 0 first with reduction Π \Pi Π and then with C \mathcal C C gives
M 0 : = C Π ( N + D 3 ) , C Π X 0 = 1 + M 0 − π . (73) M_0 := \mathcal C\Pi(\mathcal N + D_3), \qquad\mathcal C\Pi X_0 = 1 + M_0 - \pi. \tag*{(73)} M 0 := C Π ( N + D 3 ) , C Π X 0 = 1 + M 0 − π . (73)
The order in Proposition 5.1 makes M 0 M_0 M 0 nilpotent: D 3 D_3 D 3 preserves factor number, so C Π D 3 \mathcal C\Pi D_3 C Π D 3 lowers it. Equation (68) also identifies M 0 M_0 M 0 with M ∣ ℏ = 0 M|_{\hbar=0} M ∣ ℏ = 0 .
Lemma 6.2 (Kernel of the extracted classical action). For every positive weight w w w ,
ker ( C Π X 0 ∣ F 0 , w ) = { Q I k , w = 2 k , 0 , w odd . (74) \ker(\mathcal C\Pi X_0|_{\mathcal F_{0,w}}) =
\begin{cases}
\mathbb{Q}I_k, & w = 2k,\\
0, & w\ \text{odd}.
\end{cases} \tag*{(74)} ker ( C Π X 0 ∣ F 0 , w ) = { Q I k , 0 , w = 2 k , w odd . (74) The even-weight line is contained in ker X 0 ∩ ker Y 0 \ker X_0 \cap\ker Y_0 ker X 0 ∩ ker Y 0 . In particular, X 0 X_0 X 0 is injective on positive odd-weight spaces.
Proof. An element of this kernel satisfies g = ( 1 + M 0 ) − 1 π g g = (1 + M_0)^{-1}\pi g g = ( 1 + M 0 ) − 1 π g . If w w w is odd, π g = 0 \pi g = 0 π g = 0 because every power of u u u has even weight; hence g = 0 g = 0 g = 0 . If w = 2 k w = 2k w = 2 k , π g \pi g π g is a scalar multiple of u k u^k u k , so the kernel has dimension at most one. The normalized class I k I_k I k of Lemma 6.1 belongs to this kernel and is killed by both flows, proving equality. □
Middle exactness of the KdV Koszul complex
The commuting endomorphisms X 0 X_0 X 0 and Y 0 Y_0 Y 0 define a Koszul complex on the graded vector space F 0 \mathcal F_0 F 0 . Its middle exactness says that every compatible pair ( a , b ) (a,b) ( a , b ) comes from a common density. The proof reduces this assertion, one factor number at a time, to divisibility in a polynomial ring.
Lemma 6.3 (Middle exactness). For even d = 2 k ≥ 4 d = 2k \geq4 d = 2 k ≥ 4 , the sequence
F 0 , d → ( X 0 , Y 0 ) F 0 , d + 3 ⊕ F 0 , d + 5 → ( Y 0 , − X 0 ) F 0 , d + 8 (75) \mathcal F_{0,d} \xrightarrow{(X_0,Y_0)} \mathcal F_{0,d+3} \oplus\mathcal F_{0,d+5} \xrightarrow{(Y_0,-X_0)} \mathcal F_{0,d+8} \tag*{(75)} F 0 , d ( X 0 , Y 0 ) F 0 , d + 3 ⊕ F 0 , d + 5 ( Y 0 , − X 0 ) F 0 , d + 8 (75) is exact at the middle term. Equivalently, Y 0 a = X 0 b Y_0a = X_0b Y 0 a = X 0 b implies ( a , b ) = ( X 0 g , Y 0 g ) (a,b) = (X_0g,Y_0g) ( a , b ) = ( X 0 g , Y 0 g ) for some g ∈ F 0 , d g \in\mathcal F_{0,d} g ∈ F 0 , d .
Proof. Step 1. Replace derivatives by symmetric polynomials. Translation preserves factor number, so F 0 \mathcal F_0 F 0 is the direct sum of its factor-homogeneous parts. For a product of m m m jets, assign one auxiliary variable ξ i \xi_i ξ i to each factor and record its derivative order as an exponent. Averaging over the assignments makes the result independent of the order of the factors. This is the symbolic representation of differential polynomials [13 Section 2.3, Definition 6] :
u − j 1 ⋯ u − j m ⟼ 1 m ! ∑ σ ∈ S m ξ σ ( 1 ) j 1 ⋯ ξ σ ( m ) j m . (76) u_{-j_1}\cdots u_{-j_m} \longmapsto\frac{1}{m!}\sum_{\sigma\in S_m}\xi_{\sigma(1)}^{j_1}\cdots\xi_{\sigma(m)}^{j_m}. \tag*{(76)} u − j 1 ⋯ u − j m ⟼ m ! 1 σ ∈ S m ∑ ξ σ ( 1 ) j 1 ⋯ ξ σ ( m ) j m . (76)
Applying ∂ \partial ∂ differentiates each factor in turn, so it becomes multiplication by p 1 = ξ 1 + ⋯ + ξ m p_1=\xi_1+\cdots+\xi_m p 1 = ξ 1 + ⋯ + ξ m . Passing to densities modulo total derivatives therefore amounts to imposing p 1 = 0 p_1=0 p 1 = 0 . Consequently the factor-m m m part of F 0 \mathcal{F}_0 F 0 is identified with the graded polynomial ring
R m : = Q [ ξ 1 , … , ξ m ] S m / ( p 1 ) = Q [ e 2 , … , e m ] , (77) \mathcal{R}_m := \mathbb{Q}[\xi_1,\ldots,\xi_m]^{S_m}/(p_1)=\mathbb{Q}[e_2,\ldots,e_m], \tag*{(77)} R m := Q [ ξ 1 , … , ξ m ] S m / ( p 1 ) = Q [ e 2 , … , e m ] , (77)
where e j e_j e j is the elementary symmetric polynomial, of degree j j j . Weight w w w corresponds to symbol degree w − 2 m w-2m w − 2 m . The constant in R m \mathcal{R}_m R m represents u m u^m u m , not the vacuum. These rings are used one factor number at a time, not as a multiplication on F 0 \mathcal{F}_0 F 0 .
The factor-preserving parts of X 0 , Y 0 X_0,Y_0 X 0 , Y 0 are D 3 , D 5 D_3,D_5 D 3 , D 5 , where D s u − j = u − j − s D_su_{-j}=u_{-j-s} D s u − j = u − j − s . Each D s D_s D s raises the derivative order of one factor by s s s and sums over all factors. Under (76) it therefore becomes multiplication by the power sum p s = ∑ i = 1 m ξ i s p_s=\sum_{i=1}^m \xi_i^s p s = ∑ i = 1 m ξ i s :
p 3 = 3 e 3 , p 5 = 5 ( e 5 − e 2 e 3 ) . (78) p_3=3e_3,\qquad p_5=5(e_5-e_2e_3). \tag*{(78)} p 3 = 3 e 3 , p 5 = 5 ( e 5 − e 2 e 3 ) . (78)
These are Newton’s identities with e 1 = 0 e_1=0 e 1 = 0 ; set e j = 0 e_j=0 e j = 0 when j > m j>m j > m . All nonlinear terms in X 0 , Y 0 X_0,Y_0 X 0 , Y 0 raise factor number.
By the observation after (45) , both odd-weight residuals start at factor number at least three. In these coordinates, this is also immediate from R 1 = Q \mathcal{R}_1=\mathbb{Q} R 1 = Q and R 2 = Q [ e 2 ] \mathcal{R}_2=\mathbb{Q}[e_2] R 2 = Q [ e 2 ] .
Step 2. Divisibility. Suppose the residual pair has no components of factor number less than m m m . Let a m , b m a_m,b_m a m , b m denote its factor-m m m symbols, allowing either to be zero. Nonlinear terms cannot contribute at this lowest factor number, so compatibility reads
p 5 a m = p 3 b m . (79) p_5a_m=p_3b_m. \tag*{(79)} p 5 a m = p 3 b m . (79)
For m = 3 , 4 m=3,4 m = 3 , 4 , the symbol a m a_m a m has odd degree, and e 3 e_3 e 3 is the only odd-degree generator of R m \mathcal{R}_m R m . Every monomial in a m a_m a m is therefore divisible by e 3 e_3 e 3 . For m ≥ 5 m\geq5 m ≥ 5 , reduce (79) modulo e 3 e_3 e 3 . The result is e 5 ( a m m o d e 3 ) = 0 e_5(a_m\bmod e_3)=0 e 5 ( a m mod e 3 ) = 0 in the polynomial domain R m / ( e 3 ) \mathcal{R}_m/(e_3) R m / ( e 3 ) . Again e 3 e_3 e 3 divides a m a_m a m .
Step 3. Elimination. Write a m = p 3 q m a_m=p_3q_m a m = p 3 q m . Cancellation in the domain R m \mathcal{R}_m R m gives b m = p 5 q m b_m=p_5q_m b m = p 5 q m . Its degree is
deg q m = ( d + 3 − 2 m ) − 3 = d − 2 m . \deg q_m=(d+3-2m)-3=d-2m. deg q m = ( d + 3 − 2 m ) − 3 = d − 2 m .
Thus q m q_m q m represents an m m m -factor source g m g_m g m of weight d d d . Subtract the full pair ( X 0 g m , Y 0 g m ) (X_0g_m,Y_0g_m) ( X 0 g m , Y 0 g m ) . Its leading pieces remove ( a m , b m ) (a_m,b_m) ( a m , b m ) , its remaining pieces have higher factor number, and [ X 0 , Y 0 ] = 0 [X_0,Y_0]=0 [ X 0 , Y 0 ] = 0 preserves compatibility. No earlier component is recreated.
Step 4. Termination. Perform this elimination for m = 3 , … , k m=3,\ldots,k m = 3 , … , k . Afterward the first residual can have only factor number k + 1 k+1 k + 1 , with symbol degree one; there is no such element in (77) . Larger factor numbers have negative symbol degree. Hence the first residual is zero. Compatibility now says that X 0 X_0 X 0 kills the second residual, which has odd weight. Lemma 6.2 makes it zero as well. The sum of the g m g_m g m is the required g g g . For k = 2 k=2 k = 2 the elimination range is empty and the same final argument applies.
For m ≥ 5 m\geq5 m ≥ 5 , divisibility uses that e 5 e_5 e 5 is a non-zero-divisor modulo e 3 e_3 e 3 ; for m = 3 , 4 m=3,4 m = 3 , 4 , it uses parity instead. This is why the argument proves middle exactness in the stated weights without requiring a regular sequence at every factor number.
Compatibility and extraction determine a pair
We now combine middle exactness with the common conserved line. Compatibility represents a pair as ( X 0 g , Y 0 g ) (X_0g,Y_0g) ( X 0 g , Y 0 g ) ; the additional condition C Π X 0 g = 0 \mathcal{C}\Pi X_0g=0 C Π X 0 g = 0 then puts g g g on that line and makes both entries vanish.
Proposition 6.4. For every even d ≥ 4 d \ge4 d ≥ 4 , the following map is injective:
Θ 0 : F 0 , d + 3 ⊕ F 0 , d + 5 ⟶ F 0 , d + 8 ⊕ F 0 , d , ( a , b ) ⟼ ( Y 0 a − X 0 b , C a ) (80) \begin{aligned}
\Theta_0:\mathcal{F}_{0,d+3}\oplus\mathcal{F}_{0,d+5}&\longrightarrow\mathcal{F}_{0,d+8}\oplus\mathcal{F}_{0,d},\\
(a,b)&\longmapsto(Y_0a-X_0b,\mathcal{C}a) \tag*{(80)}
\end{aligned} Θ 0 : F 0 , d + 3 ⊕ F 0 , d + 5 ( a , b ) ⟶ F 0 , d + 8 ⊕ F 0 , d , ⟼ ( Y 0 a − X 0 b , C a ) (80)
Proof. For a pair in its kernel, Lemma 6.3 gives ( a , b ) = ( X 0 g , Y 0 g ) (a,b)=(X_0g,Y_0g) ( a , b ) = ( X 0 g , Y 0 g ) . The second equation is C Π X 0 g = 0 \mathcal{C}\Pi X_0g=0 C Π X 0 g = 0 . Lemma 6.2 makes g g g a multiple of I k I_k I k , and both X 0 X_0 X 0 and Y 0 Y_0 Y 0 annihilate that class. Thus a = b = 0 a=b=0 a = b = 0 .
Conservation, uniqueness, and commutativity
Proposition 5.2 gives C Π X ℏ F k = 0 \mathcal{C}\Pi X_{\hbar}F_k=0 C Π X ℏ F k = 0 . To obtain conservation, we must deduce Π X ℏ F k = 0 \Pi X_{\hbar}F_k=0 Π X ℏ F k = 0 . The classical limits are now identified by (68) ; we lift the injectivity of Proposition 6.4 to formal power series in ℏ \hbar ℏ and apply it to the compatible pair ( X ℏ F k , Y ℏ F k ) (X_{\hbar}F_k,Y_{\hbar}F_k) ( X ℏ F k , Y ℏ F k ) . The finite inverse then proves uniqueness, and the Jacobi identity gives commutativity.
Lifting the injectivity to formal power series
Extend scalars at each fixed weight:
F ^ w : = F w ⊗ Q Q [ [ ℏ ] ] ≃ F 0 , w [ [ ℏ ] ] . \widehat{\mathcal{F}}_w:=\mathcal{F}_w\otimes_{\mathbb{Q}}\mathbb{Q}[[\hbar]]\simeq\mathcal{F}_{0,w}[[\hbar]]. F w := F w ⊗ Q Q [[ ℏ ]] ≃ F 0 , w [[ ℏ ]] .
Write F ^ = ⨁ w F ^ w \widehat{\mathcal{F}}=\bigoplus_w\widehat{\mathcal{F}}_w F = ⨁ w F w . These are finite-rank free modules at each fixed weight. For even d ≥ 4 d \ge4 d ≥ 4 , on F ^ d + 3 ⊕ F ^ d + 5 \widehat{\mathcal{F}}_{d+3}\oplus\widehat{\mathcal{F}}_{d+5} F d + 3 ⊕ F d + 5 define
Θ ℏ ( a , b ) : = ( Y ℏ a − X ℏ b , C a ) . (81) \Theta_{\hbar}(a,b):=(Y_{\hbar}a-X_{\hbar}b,\mathcal{C}a). \tag*{(81)} Θ ℏ ( a , b ) := ( Y ℏ a − X ℏ b , C a ) . (81)
Its constant term is the map Θ 0 \Theta_0 Θ 0 of Proposition 6.4.
Lemma 7.1 (Injectivity over formal power series). The map Θ ℏ \Theta_{\hbar} Θ ℏ is injective. For every k ≥ 2 k \ge2 k ≥ 2 and every f ∈ F ^ 2 k f\in\widehat{\mathcal{F}}_{2k} f ∈ F 2 k ,
C Π X ℏ f = 0 ⟹ X ℏ f = Y ℏ f = 0. (82) \mathcal{C}\Pi X_{\hbar}f=0\quad\Longrightarrow\quad X_{\hbar}f=Y_{\hbar}f=0. \tag*{(82)} C Π X ℏ f = 0 ⟹ X ℏ f = Y ℏ f = 0. (82) Proof. If a nonzero formal pair lies in ker Θ ℏ \ker\Theta_{\hbar} ker Θ ℏ , its first nonzero coefficient in ℏ \hbar ℏ lies in ker Θ 0 \ker\Theta_0 ker Θ 0 , contrary to Proposition 6.4. Hence Θ ℏ \Theta_{\hbar} Θ ℏ is injective. Since [ X ℏ , Y ℏ ] = 0 [X_{\hbar},Y_{\hbar}]=0 [ X ℏ , Y ℏ ] = 0 by (65) , we have
Θ ℏ ( X ℏ f , Y ℏ f ) = ( 0 , C Π X ℏ f ) . \Theta_{\hbar}(X_{\hbar}f,Y_{\hbar}f)=(0,\mathcal{C}\Pi X_{\hbar}f). Θ ℏ ( X ℏ f , Y ℏ f ) = ( 0 , C Π X ℏ f ) . The stated implication follows.
Conservation and the classical limit of the constructed densities
For k ≥ 2 k \ge2 k ≥ 2 , the polynomial F k F_k F k has coefficients in Q [ ℏ ] \mathbb{Q}[\hbar] Q [ ℏ ] and satisfies C Π X ℏ F k = 0 \mathcal{C}\Pi X_{\hbar}F_k = 0 C Π X ℏ F k = 0 by (56) . Applying Lemma 7.1 over Q [ [ ℏ ] ] \mathbb{Q}[[\hbar]] Q [[ ℏ ]] gives
X ℏ F k = Y ℏ F k = 0 in F ^ . (83) X_{\hbar}F_k = Y_{\hbar}F_k = 0 \quad\text{in } \widehat{\mathcal{F}}. \tag*{(83)} X ℏ F k = Y ℏ F k = 0 in F . (83)
Because F \mathcal{F} F is a free Q [ ℏ ] \mathbb{Q}[\hbar] Q [ ℏ ] -module with a fixed reduced basis, the equality already holds over Q [ ℏ ] \mathbb{Q}[\hbar] Q [ ℏ ] . In VOA notation it is exactly ℏ − 1 H 4 ( 0 ) H 2 k ∈ T V \hbar^{-1}H_4(0)H_{2k} \in T\mathcal{V} ℏ − 1 H 4 ( 0 ) H 2 k ∈ T V , which is the required conservation statement.
Taking the constant term gives X 0 F k ( 0 ) = 0 X_0F_k(0) = 0 X 0 F k ( 0 ) = 0 , together with π F k ( 0 ) = u k \pi F_k(0) = u^k π F k ( 0 ) = u k . Lemma 6.2 , independently derived from the Lax residues, forces
F k ( 0 ) = I k . (84) F_k(0) = I_k. \tag*{(84)} F k ( 0 ) = I k . (84)
This proves the classical-limit part of the theorem.
Uniqueness and commutativity from the Jacobi identity
Let g ∈ F w g \in\mathcal{F}_w g ∈ F w be a homogeneous class annihilated by X ℏ X_{\hbar} X ℏ . The identity (54) and nilpotence of M M M give
( 1 + M ) g = π g , g = ( 1 + M ) − 1 π g . (85) (1+M)g = \pi g,\qquad g = (1+M)^{-1}\pi g. \tag*{(85)} ( 1 + M ) g = π g , g = ( 1 + M ) − 1 π g . (85)
Derivative-free monomials have even positive weight, so this equation forces g = 0 g=0 g = 0 when w w w is odd. At weight 2 k 2k 2 k it forces g g g to be a multiple of F k F_k F k ; conservation gives the required nonzero class. This includes F 1 = u F_1=u F 1 = u , whose centrality was checked above. These kernel conclusions hold over Q [ ℏ ] \mathbb{Q}[\hbar] Q [ ℏ ] and at every specialization ℏ ∈ C \hbar\in\mathbb{C} ℏ ∈ C . At ℏ = 0 \hbar=0 ℏ = 0 they concern X 0 X_0 X 0 , rather than the zero-mode bracket, which vanishes in that specialization.
Write h k : = [ H 2 k ] h_k := [H_{2k}] h k := [ H 2 k ] and h 2 : = [ H 4 ] h_2 := [H_4] h 2 := [ H 4 ] in the Lie algebra V / T V \mathcal{V}/T\mathcal{V} V / T V , with bracket { [ a ] , [ b ] } = [ a ( 0 ) b ] \{[a],[b]\}=[a(0)b] {[ a ] , [ b ]} = [ a ( 0 ) b ] . Conservation gives { h 2 , h k } = 0 \{h_2,h_k\}=0 { h 2 , h k } = 0 . By the Jacobi identity,
{ h 2 , { h k , h l } } = { { h 2 , h k } , h l } + { h k , { h 2 , h l } } = 0. \{h_2,\{h_k,h_l\}\}=\{\{h_2,h_k\},h_l\}+\{h_k,\{h_2,h_l\}\}=0. { h 2 , { h k , h l }} = {{ h 2 , h k } , h l } + { h k , { h 2 , h l }} = 0.
In the reduced model, { h 2 , a } = ℏ X ℏ a \{h_2,a\}=\hbar X_{\hbar}a { h 2 , a } = ℏ X ℏ a . Because the local-functional module is ℏ \hbar ℏ -torsion free, the Jacobi identity implies X ℏ { h k , h l } = 0 X_{\hbar}\{h_k,h_l\}=0 X ℏ { h k , h l } = 0 . This bracket has weight 2 k + 2 l − 1 2k+2l-1 2 k + 2 l − 1 , which is odd; it therefore vanishes by the preceding kernel theorem. Thus
H 2 k ( 0 ) H 2 l ∈ T V . (86) H_{2k}(0)H_{2l} \in T\mathcal{V}. \tag*{(86)} H 2 k ( 0 ) H 2 l ∈ T V . (86)
On every module the corresponding zero-mode operators commute, by [ a ( 0 ) , b ( 0 ) ] = ( a ( 0 ) b ) ( 0 ) [a(0),b(0)]=(a(0)b)(0) [ a ( 0 ) , b ( 0 )] = ( a ( 0 ) b ) ( 0 ) and ( T a ) ( 0 ) = 0 (Ta)(0)=0 ( T a ) ( 0 ) = 0 .
Specialization to every central charge
The preceding argument uses the rescaling c = 6 / ℏ c = 6/\hbar c = 6/ℏ . To include c = 0 c = 0 c = 0 , we return to the original Virasoro coordinates and show that both the constructed densities and the inverse identity are polynomial in c c c . This also proves the uniqueness statement at c = 0 c = 0 c = 0 . To distinguish the two PBW coordinate systems, write v − j v_{-j} v − j for the coordinate of T j ω T^{j}\omega T j ω and v : = v 0 v := v_{0} v := v 0 . The map σ ω \sigma_{\omega} σ ω is defined by (8) with u − j u_{-j} u − j replaced by v − j v_{-j} v − j and U − j U_{-j} U − j replaced by j ! L − j − 2 j!L_{-j-2} j ! L − j − 2 . Reduction, evaluation, and extraction use the same formulas in the v v v -variables; we retain the symbols Π \Pi Π , π \pi π , C \mathcal{C} C on this polynomial space.
After adjoining ℏ − 1 \hbar^{-1} ℏ − 1 , define S ℏ S_{\hbar} S ℏ by S ℏ f ( u 0 , u − 1 , … ) : = f ( ℏ v 0 , ℏ v − 1 , … ) S_{\hbar}f(u_{0},u_{-1},\ldots) := f(\hbar v_{0},\hbar v_{-1},\ldots) S ℏ f ( u 0 , u − 1 , … ) := f ( ℏ v 0 , ℏ v − 1 , … ) . Thus it multiplies an m m m -factor monomial by ℏ m \hbar^{m} ℏ m and changes its coordinate letters. Set
F k ( c ) : = ℏ − k S ℏ F k ∣ ℏ = 6 / c . (87) F_{k}^{(c)} := \left.\hbar^{-k}S_{\hbar}F_{k}\right|_{\hbar=6/c}. \tag*{(87)} F k ( c ) := ℏ − k S ℏ F k ℏ = 6/ c . (87)
For an m m m -factor monomial m \mathfrak{m} m , Proposition 5.1 allows only terms ℏ q m \hbar^{q}\mathfrak{m} ℏ q m with 0 ≤ q ≤ k − m 0 \le q \le k-m 0 ≤ q ≤ k − m . Under (87) , such a term becomes
( c / 6 ) k − m − q m ( v 0 , v − 1 , … ) . (c/6)^{k-m-q}\mathfrak{m}(v_{0},v_{-1},\ldots). ( c /6 ) k − m − q m ( v 0 , v − 1 , … ) .
Consequently F k ( c ) ∈ Q [ c ] [ v 0 , v − 1 , … ] F_{k}^{(c)} \in\mathbb{Q}[c][v_{0},v_{-1},\ldots] F k ( c ) ∈ Q [ c ] [ v 0 , v − 1 , … ] , its m m m -factor coefficients have c c c -degree at most k − m k-m k − m , and π F k ( c ) = v k \pi F_{k}^{(c)} = v^{k} π F k ( c ) = v k . Equation (87) therefore defines the density at c = 0 c = 0 c = 0 by polynomial evaluation.
Let the original weight-four action be
A c : = σ ω − 1 ( ω ( − 1 ) ω ) ( 0 ) σ ω . A_{c}:=\sigma_{\omega}^{-1}(\omega(-1)\omega)(0)\sigma_{\omega}. A c := σ ω − 1 ( ω ( − 1 ) ω ) ( 0 ) σ ω .
The PBW identifications satisfy σ = σ ω S ℏ \sigma=\sigma_{\omega}S_{\hbar} σ = σ ω S ℏ after c = 6 / ℏ c=6/\hbar c = 6/ℏ . Since H 4 = ℏ 2 ω ( − 1 ) ω H_{4}=\hbar^{2}\omega(-1)\omega H 4 = ℏ 2 ω ( − 1 ) ω , we have
X ℏ = ℏ S ℏ − 1 A 6 / ℏ S ℏ . (88) X_{\hbar}=\hbar S_{\hbar}^{-1}A_{6/\hbar}S_{\hbar}. \tag*{(88)} X ℏ = ℏ S ℏ − 1 A 6/ℏ S ℏ . (88)
Likewise σ ω ( F k ( c ) ) = ℏ − k H 2 k \sigma_{\omega}(F_{k}^{(c)})=\hbar^{-k}H_{2k} σ ω ( F k ( c ) ) = ℏ − k H 2 k after localization. The conservation and commutativity identities thus hold after the substitution c = 6 / ℏ c=6/\hbar c = 6/ℏ . Virasoro PBW reordering makes the original mode action polynomial in c c c , and the reduced basis is free over Q [ c ] \mathbb{Q}[c] Q [ c ] . Injectivity of Q [ c ] ↪ Q ( ( ℏ ) ) \mathbb{Q}[c]\hookrightarrow\mathbb{Q}((\hbar)) Q [ c ] ↪ Q (( ℏ )) then proves these identities over Q [ c ] \mathbb{Q}[c] Q [ c ] , including their evaluation at c = 0 c=0 c = 0 .
For the centralizer statement, put M c : = S ℏ M S ℏ − 1 ∣ ℏ = 6 / c M_{c}:=\left.S_{\hbar}MS_{\hbar}^{-1}\right|_{\hbar=6/c} M c := S ℏ M S ℏ − 1 ℏ = 6/ c . A term of M M M lowering factor number by r r r has parameter degree at most r r r . Conjugation multiplies it by ℏ − r \hbar^{-r} ℏ − r , so M c M_{c} M c also has polynomial coefficients in c c c . Both Π \Pi Π and π \pi π commute with S ℏ S_{\hbar} S ℏ , while C S ℏ = ℏ S ℏ C \mathcal{C}S_{\hbar}=\hbar S_{\hbar}\mathcal{C} C S ℏ = ℏ S ℏ C . Conjugating (54) and using (88) gives
C Π A c = 1 + M c − π . (89) \mathcal{C}\Pi A_{c}=1+M_{c}-\pi. \tag*{(89)} C Π A c = 1 + M c − π . (89)
The decreasing order of Proposition 5.1 is unchanged, so M c M_{c} M c is nilpotent after every specialization. If A c g = 0 A_{c}g=0 A c g = 0 in the local-functional space, then g = ( 1 + M c ) − 1 π g g=(1+M_{c})^{-1}\pi g g = ( 1 + M c ) − 1 π g . Odd weights therefore give zero, and an even-weight kernel has dimension at most one. The conserved class F k ( c ) F_{k}^{(c)} F k ( c ) , whose evaluation is v k v^{k} v k , is nonzero at every c c c and spans that line. This completes the proof of Theorem 5.3.
Low-weight formulas and exact checks
We first display the vectors H 2 k ( c ) H^{(c)}_{2k} H 2 k ( c ) in the original Virasoro generators L n L_n L n and conformal vector ω \omega ω , at central charge c c c . The later calculations retain the rescaled coordinates used in the construction. These exact examples illustrate the formulas; the proofs in the main text establish the results at all weights.
Conserved vectors in the original Virasoro notation
Recall H 2 ( c ) = σ ω ( F 1 ( c ) ) H^{(c)}_{2}=\sigma_{\omega}(F^{(c)}_{1}) H 2 ( c ) = σ ω ( F 1 ( c ) ) from Theorem 5.3, with H 2 ( c ) = ω = L − 2 1 H^{(c)}_{2}=\omega=L_{-2}\mathbf{1} H 2 ( c ) = ω = L − 2 1 . For the first nontrivial weights, the literal symmetric representatives are
H 4 ( c ) = L − 2 2 1 , H^{(c)}_{4}=L_{-2}^{2}\mathbf{1}, H 4 ( c ) = L − 2 2 1 ,
H 6 ( c ) = [ L − 2 3 − c + 2 12 L − 3 2 ] 1 , H^{(c)}_{6}=\left[L_{-2}^{3}-\frac{c+2}{12}L_{-3}^{2}\right]\mathbf{1}, H 6 ( c ) = [ L − 2 3 − 12 c + 2 L − 3 2 ] 1 ,
H 8 ( c ) = [ L − 2 4 − c + 8 3 ( L − 3 L − 2 2 + L − 5 L − 3 + 4 3 L − 8 ) + c 2 + 9 c − 61 45 L − 4 2 ] 1. (90) H^{(c)}_{8}=\left[L_{-2}^{4}-\frac{c+8}{3}\left(L_{-3}L_{-2}^{2}+L_{-5}L_{-3}+\frac{4}{3}L_{-8}\right)+\frac{c^{2}+9c-61}{45}L_{-4}^{2}\right]\mathbf{1}. \tag*{(90)} H 8 ( c ) = [ L − 2 4 − 3 c + 8 ( L − 3 L − 2 2 + L − 5 L − 3 + 3 4 L − 8 ) + 45 c 2 + 9 c − 61 L − 4 2 ] 1 . (90)
Indeed, v − j v_{-j} v − j corresponds to T j ω = j ! L − j − 2 1 T^{j}\omega=j!L_{-j-2}\mathbf{1} T j ω = j ! L − j − 2 1 , and the mixed symmetrization in the weight-eight formula is
σ ω ( v v 2 2 ) = ( L − 3 2 L − 2 + L − 5 L − 3 + 4 3 L − 8 ) 1. \sigma_{\omega}(vv_{2}^{2})=\left(L_{-3}^{2}L_{-2}+L_{-5}L_{-3}+\frac{4}{3}L_{-8}\right)\mathbf{1}. σ ω ( v v 2 2 ) = ( L − 3 2 L − 2 + L − 5 L − 3 + 3 4 L − 8 ) 1 .
The corresponding reduced differential polynomials are
F 2 ( c ) = v 2 , F^{(c)}_{2}=v^{2}, F 2 ( c ) = v 2 ,
F 3 ( c ) = v 3 − c + 2 12 v v − 1 2 , F^{(c)}_{3}=v^{3}-\frac{c+2}{12}vv_{-1}^{2}, F 3 ( c ) = v 3 − 12 c + 2 v v − 1 2 ,
F 4 ( c ) = v 4 − c + 8 3 v v − 1 2 + c 2 + 9 c − 61 180 v − 2 2 . (91) F^{(c)}_{4}=v^{4}-\frac{c+8}{3}vv_{-1}^{2}+\frac{c^{2}+9c-61}{180}v_{-2}^{2}. \tag*{(91)} F 4 ( c ) = v 4 − 3 c + 8 v v − 1 2 + 180 c 2 + 9 c − 61 v − 2 2 . (91)
Different ordered representatives can have different lower-factor coefficients. For example, modulo T V c TV^{c} T V c ,
H 6 ( c ) ≡ ( L − 2 3 + c + 2 6 L − 4 L − 2 ) 1 , H^{(c)}_{6}\equiv\left(L_{-2}^{3}+\frac{c+2}{6}L_{-4}L_{-2}\right)\mathbf{1}, H 6 ( c ) ≡ ( L − 2 3 + 6 c + 2 L − 4 L − 2 ) 1 ,
H 8 ( c ) ≡ ( L − 2 4 + c + 8 3 L − 4 L − 2 2 + 2 ( c 2 + 14 c − 21 ) 15 L − 6 L − 2 ) 1. H^{(c)}_{8}\equiv\left(L_{-2}^{4}+\frac{c+8}{3}L_{-4}L_{-2}^{2}+\frac{2(c^{2}+14c-21)}{15}L_{-6}L_{-2}\right)\mathbf{1}. H 8 ( c ) ≡ ( L − 2 4 + 3 c + 8 L − 4 L − 2 2 + 15 2 ( c 2 + 14 c − 21 ) L − 6 L − 2 ) 1 .
These congruences follow by applying T = L − 1 T=L_{-1} T = L − 1 to shorter ordered words. In particular, the difference between the two weight-six representatives is − ( c + 2 ) T ( L − 3 L − 2 1 ) / 12 -(c+2)T(L_{-3}L_{-2}\mathbf{1})/12 − ( c + 2 ) T ( L − 3 L − 2 1 ) /12 . Comparison with cylinder charges also requires the coordinate change discussed in the introduction; equality of the displayed vectors is not asserted in a different coordinate convention.
Explicit total-derivative identities
The conservation identities can be verified before passing to the quotient by displaying vectors W k , l ( c ) W^{(c)}_{k,l} W k , l ( c ) such that
H 2 k ( c ) ( 0 ) H 2 l ( c ) = T W k , l ( c ) . H^{(c)}_{2k}(0)H^{(c)}_{2l}=TW^{(c)}_{k,l}. H 2 k ( c ) ( 0 ) H 2 l ( c ) = T W k , l ( c ) .
Since ω ( 0 ) = T \omega(0)=T ω ( 0 ) = T , one may take W 1 , l ( c ) = H 2 l ( c ) W^{(c)}_{1,l}=H^{(c)}_{2l} W 1 , l ( c ) = H 2 l ( c ) . For the weight-four pair, an explicit choice is
W 2 , 2 ( c ) = [ 4 L − 2 3 + 2 ( c + 5 ) 3 L − 4 L − 2 + 4 − c 6 L − 3 2 + 19 c − 22 15 L − 6 ] 1. (92) W^{(c)}_{2,2}=\left[4L_{-2}^{3}+\frac{2(c+5)}{3}L_{-4}L_{-2}+\frac{4-c}{6}L_{-3}^{2}+\frac{19c-22}{15}L_{-6}\right]\mathbf{1}. \tag*{(92)} W 2 , 2 ( c ) = [ 4 L − 2 3 + 3 2 ( c + 5 ) L − 4 L − 2 + 6 4 − c L − 3 2 + 15 19 c − 22 L − 6 ] 1 . (92)
This follows from (37) after inverse scaling and symmetrization; it can also be checked directly using the Virasoro commutator and [ T , L − n ] = ( n − 1 ) L − n − 1 [T,L_{-n}]=(n-1)L_{-n-1} [ T , L − n ] = ( n − 1 ) L − n − 1 . For the distinct pair of weights four and six, one may take
W 2 , 3 ( c ) = [ 9 2 L − 2 4 + ( c + 11 ) L − 4 L − 2 2 − ( c + 2 ) L − 3 2 L − 2 + 3 ( c + 4 ) L − 6 L − 2 − c 2 + 13 c − 14 6 L − 5 L − 3 + c 2 + c − 29 18 L − 4 2 − 19 c 2 − 227 c + 226 42 L − 8 ] 1. (93) \begin{aligned}
W^{(c)}_{2,3}=\Biggl[&\frac{9}{2}L_{-2}^{4}+(c+11)L_{-4}L_{-2}^{2}-(c+2)L_{-3}^{2}L_{-2}+3(c+4)L_{-6}L_{-2}\\
&-\frac{c^{2}+13c-14}{6}L_{-5}L_{-3}+\frac{c^{2}+c-29}{18}L_{-4}^{2}-\frac{19c^{2}-227c+226}{42}L_{-8}\Biggr]\mathbf{1}. \tag*{(93)}
\end{aligned} W 2 , 3 ( c ) = [ 2 9 L − 2 4 + ( c + 11 ) L − 4 L − 2 2 − ( c + 2 ) L − 3 2 L − 2 + 3 ( c + 4 ) L − 6 L − 2 − 6 c 2 + 13 c − 14 L − 5 L − 3 + 18 c 2 + c − 29 L − 4 2 − 42 19 c 2 − 227 c + 226 L − 8 ] 1 . (93)
Substitution in the zero-mode sum verifies H 4 ( c ) ( 0 ) H 6 ( c ) = T W 2 , 3 ( c ) H^{(c)}_{4}(0)H^{(c)}_{6}=TW^{(c)}_{2,3} H 4 ( c ) ( 0 ) H 6 ( c ) = T W 2 , 3 ( c ) . Both identities use the literal representatives in (A.1).
Evaluating the finite inverse
To exhibit the construction itself, return to the rescaled coordinates u − j u_{-j} u − j over Q [ ℏ ] \mathbb{Q}[\hbar] Q [ ℏ ] . The reduced weight-eight polynomial is
F 4 = u 4 − ( 2 + 8 ℏ 3 ) u u − 1 2 + ( 1 5 + 3 ℏ 10 − 61 ℏ 2 180 ) u − 2 2 . (94) F_{4}=u^{4}-\left(2+\frac{8\hbar}{3}\right)uu_{-1}^{2}+\left(\frac{1}{5}+\frac{3\hbar}{10}-\frac{61\hbar^{2}}{180}\right)u_{-2}^{2}. \tag*{(94)} F 4 = u 4 − ( 2 + 3 8ℏ ) u u − 1 2 + ( 5 1 + 10 3ℏ − 180 61 ℏ 2 ) u − 2 2 . (94)
Here the inverse in (55) has just three nonzero summands:
− M u 4 = − ( 2 + 8 ℏ 3 ) u u − 1 2 + ( 3 ℏ 20 − 11 ℏ 2 60 ) u − 2 2 , (95) -Mu^{4}=-\left(2+\frac{8\hbar}{3}\right)uu_{-1}^{2}+\left(\frac{3\hbar}{20}-\frac{11\hbar^{2}}{60}\right)u_{-2}^{2}, \tag*{(95)} − M u 4 = − ( 2 + 3 8ℏ ) u u − 1 2 + ( 20 3ℏ − 60 11 ℏ 2 ) u − 2 2 , (95)
M 2 u 4 = ( 1 5 + 3 ℏ 20 − 7 ℏ 2 45 ) u − 2 2 , M 3 u 4 = 0. M^{2}u^{4}=\left(\frac{1}{5}+\frac{3\hbar}{20}-\frac{7\hbar^{2}}{45}\right)u_{-2}^{2},\qquad M^{3}u^{4}=0. M 2 u 4 = ( 5 1 + 20 3ℏ − 45 7 ℏ 2 ) u − 2 2 , M 3 u 4 = 0.
The calculation applies (30) , reduction, and extraction at each step. Inverse scaling by (87) gives F 4 ( c ) F^{(c)}_{4} F 4 ( c ) in (A.2).
The mode calculation at weight four
We record the details of the calculation leading to (37) . For the action on u 2 u^{2} u 2 , we use ordered products of the U − j U_{-j} U − j . The terms with at least two factors are as follows; the omitted terms in this table have at most one factor:
Table 1.
The additional nonzero r = 7 r=7 r = 7 contribution has only one factor. The exact PBW conversions, including their lower-factor terms, are
σ ( u 2 u − 1 ) = ( U − 1 U 0 2 + ℏ 6 U − 3 U 0 + ℏ 2 120 U − 5 ) 1 , \sigma(u^2u_{-1})=(U_{-1}U_0^2+\frac{\hbar}{6}U_{-3}U_0+\frac{\hbar^2}{120}U_{-5})1, σ ( u 2 u − 1 ) = ( U − 1 U 0 2 + 6 ℏ U − 3 U 0 + 120 ℏ 2 U − 5 ) 1 ,
σ ( u u − 3 ) = ( U − 3 U 0 + 3 ℏ 40 U − 5 ) 1 , \sigma(uu_{-3})=(U_{-3}U_0+\frac{3\hbar}{40}U_{-5})1, σ ( u u − 3 ) = ( U − 3 U 0 + 40 3ℏ U − 5 ) 1 ,
σ ( u − 1 u − 2 ) = ( U − 2 U − 1 + ℏ 120 U − 5 ) 1 , σ ( u − 5 ) = U − 5 1. \sigma(u_{-1}u_{-2})=(U_{-2}U_{-1}+\frac{\hbar}{120}U_{-5})1,\qquad\sigma(u_{-5})=U_{-5}1. σ ( u − 1 u − 2 ) = ( U − 2 U − 1 + 120 ℏ U − 5 ) 1 , σ ( u − 5 ) = U − 5 1.
They follow by reordering the two- and three-factor permutation averages. Including the one-factor contributions, the full mode sum is
( 12 U − 1 U 0 2 + ( 2 + 11 3 ℏ ) U − 3 U 0 + 3 ℏ U − 2 U − 1 + ( 3 10 ℏ + 1 20 ℏ 2 ) U − 5 ) 1. (12U_{-1}U_0^2+(2+\frac{11}{3}\hbar)U_{-3}U_0+3\hbar U_{-2}U_{-1}+(\frac{3}{10}\hbar+\frac{1}{20}\hbar^2)U_{-5})1. ( 12 U − 1 U 0 2 + ( 2 + 3 11 ℏ ) U − 3 U 0 + 3ℏ U − 2 U − 1 + ( 10 3 ℏ + 20 1 ℏ 2 ) U − 5 ) 1.
Inverting these symmetric PBW identities gives (37) .
AI-assisted research and writing
The problem studied in this paper was proposed by the author, who set the research objectives and guided the scope, conventions, and successive revisions. The work developed through sustained interaction with OpenAI’s ChatGPT (GPT-6 Pro, web interface) and Codex (GPT-6 Astra). The author contributed mathematical questions, choices of formulation, and critical revisions; the AI systems suggested intermediate constructions, possible lines of argument, and alternative ways to organize proofs. Some ideas were developed from the author’s proposals and corrections, while others began as AI suggestions and were refined through subsequent exchanges. AI assistance thus contributed to the development of mathematical ideas and arguments as well as to their presentation.
ChatGPT and Codex were also used to draft and revise exposition, write code for exact symbolic calculations, assist with checking intermediate arguments and formulas, and locate literature for verification. AI suggestions are not treated as independent validation of mathematical claims. The conclusions rest on the proofs and cited results, supplemented by exact computational checks. The author retains responsibility for the final content.
Hongbo Zhao
School of Mathematics and Statistics, Chongqing University, Chongqing, P.R.C. China
E-mail: hongbozhao@cqu.edu.cn