Sign up with your email address to be the first to know about new products, VIP offers, blog features & more.

Frobenius Structures, the Total Descendent Potential and Gromov – Witten Theory

In this post, continuing from part one on Gromov-Witten invariants, I shall derive two deep properties, the second being central about the genus 1 Gromov-Witten potential, and along the way, discuss some propositions regarding the total ancestor potential. Let me introduce the total ancestor potential

eq1

where the genus g ancestor potential \not \tilde F_t^g is defined by

eq2

with

    \[{\bar \psi _i}: = {\pi ^ * }\left( {{\psi _i}} \right)\]

referring to the pull-backs of the classes {\psi _i}, i = 1, …, m, from {\bar M_{g,m}} relative to the composition

    \[\pi :{X_{g,\,m + l,\,d}} \to {\bar M_{g,\,m + l,\,d}} \to {\bar M_{g,\,m}}\]

Since the sum does not contain the terms with (g, m) = (0, 0),(0, 1),(0, 2) and (1, 0), one can treat {\tilde {\rm A}_t} subject to the dilaton shift

    \[\not q(z) = \not t(z) - z \in {\tilde H_t}\]

as an element in the Fock space depending on the parameter t \in {\rm H}.

Let us implicitly define the operator {S_t} on the Laurent 1/z-series completion of the space {\rm H} defined by

 

eq4

 

with the 2-point gravitational descendent in genus 0 being

 

eq5

 

Now, it is one of the basic facts of quantum cohomology theory that

    \[S_t^ * \left( { - 1/z} \right){S_t}\left( {1/z} \right) = 1\]

that is, given a metaplectic completion of \left( {{{\tilde H}_t},\Omega } \right), {\hat S_t} defines a symplectic transformation depending on the parameter t \in {\rm H}. Put

    \[{\hat S_t}: = \exp {\left( {{\rm{In}}\,{S_t}} \right)^ \wedge }\]

Here is a deep theorem with an excellent proof ‘here’

    \[{\tilde D_{wk}} = {e^{{F^1}\left( t \right)}}{\hat S_t}^{ - 1}{\tilde {\rm A}_t}\]

For the definition of {\tilde D_{wk}}, see my last postGetzler proved the 3g − 2-jet conjecture of Eguchi–Xiong and Dubrovin about genus g descendent potential

    \[\frac{{{{\not \partial }^m}}}{{\not \partial t_{{k_1} + 1}^{{\alpha _1}}...\not \partial t_{{k_m} + 1}^{{\alpha _m}}}}\left( {\tilde F_t^g} \right)\left| {_{t = 0}} \right.\]

if k1 + … + km > 3g − 3.

 

Proposition 1

Any quantized symplectic operator of the form

    \[S\left( {1/z} \right) = \mathord{\buildrel{\lower3pt\hbox{$\scriptscriptstyle\frown$}} \over 1} + {S_1}/z + {S_2}/{z^2} + ...\]

acts on elements G of the Fock space as

 

eq6

 

with {\left[ {{S_{\bar q}}} \right]_ + } the power series truncation of S\left( {{z^{ - 1}}} \right)\bar q\left( z \right) with the quadratic form

    \[W = \sum {\left( {{W_{kl{q_k},q}}} \right)} \]

defined via

eq7

Proposition 2

The genus 0 descendent potential equals

 

eq8

 

which is a deep and famous reconstruction for genus 0 gravitational descendents due to Dubrovin and Dijkgraaf–Witten.

    \[A\left( t \right): = z\left( {{d_t}{S_t}\left( {1/z} \right)} \right)S_t^{ - 1}\left( {1/z} \right)\]

is independent of z and thus defines a linear pencil of connections

    \[{\nabla _z}: = d - {z^{ - 1}}A\left( t \right) \wedge \]

on the tangent bundle T{\rm H} flat for all values of the parameter {z^{ - 1}}

hence

    \[\left\{ {\begin{array}{*{20}{c}}{A \wedge A = 0}\\{dA = 0}\end{array}} \right.\]

and in coordinate form, t = \sum {{t^\alpha }} {\phi _\alpha } the first condition is equivalent to commutativity

    \[\left\{ {\begin{array}{*{20}{c}}{{A_\alpha }{A_\beta } = {A_\beta }{A_\alpha }}\\{A = \sum {{A_\alpha }\left( t \right)d{t^\alpha }} }\end{array}} \right.\]

The correspondence

    \[{\not \partial _\alpha } \Rightarrow {A_\alpha }\left( t \right) = {i_{{{\not \partial }_\alpha }}}A\]

defines commutative associative multiplications { \bullet _t} on the tangent spaces {T_t}{\rm H}, and those are the quantum cup-product, and deeply, we have a coincidence of

    \[\left( {{\phi _\alpha }{ \bullet _t}{\phi _\beta },{\phi _\gamma }} \right)\]

with the third directional derivatives

    \[{\not \partial _\alpha }{\not \partial _\beta }{\not \partial _\gamma }\bar F_X^0\left( t \right)\]

of the genus zero Gromov-Witten potential 

    \[\bar F_X^0\]

which explains why dA = 0 and A_\alpha ^ * = {A_\alpha } yielding hence that the quantum cup-product is Frobenius: \left( {a \bullet b,c} \right) = \left( {a,b \bullet c} \right).

  • A Frobenius structure on a manifold H consists of:

(i) a flat pseudo-Riemannian metric (·, ·),

(ii) a function F whose 3-rd covariant derivatives {F_{abc}} are structure constants \left( {a{ \bullet _t}b,c} \right) of a Frobenius algebra structure: that is, an associative commutative multiplication { \bullet _t} satisfying

    \[\left( {a{ \bullet _t}b,c} \right) = \left( {a,b{ \bullet _t}c} \right)\]

on the tangent spaces {T_t}{\rm H} which depends smoothly on t;

(iii) the vector field of unities 1 of the { \bullet _t}-product which has to be covariantly constant and preserve the multiplication and the metric.

 

Now, in Gromov–Witten theory, genus 0 Gromov-Witten invariants of a compact almost Kähler manifold X define on

    \[H = {H^ * }\left( {X;\hat Q\left[ {\left\{ Q \right\}} \right]} \right)\]

a formal structure of a calibrated conformal Frobenius manifold of conformal dimension D = {\rm{di}}{{\rm{m}}_\mathbb{C}}X, and in the coordinate system

    \[t = \sum {{t^\alpha }} {\phi _\alpha }\]

corresponding to a graded basis \left\{ {{\phi _\alpha }} \right\} in {H^ * }\left( {X,\hat Q} \right), the Euler field takes on the form

    \[E = \sum {\left( {1 - {\rm{deg}}\,{\phi _\alpha }/2} \right)} \,{t_\alpha }{\not \partial _\alpha } + \rho \]

where the constant part \rho \in {H^ * }\left( X \right) is the 1-st Chern class of the tangent bundle TX.

So, proposition 3: The equation {\nabla _z}S = 0 in a neighborhood of a semisimple point u has a fundamental solution in the form

    \[{\Psi _u}{R_u}\left( z \right){e^{U/z}}\]

with

    \[{R_u}\left( z \right) = \not 1 + {R_1}\,z + {R_2}\,{z^2} + ...\]

a formal matrix power series satisfying: R_u^ * \left( { - \,z} \right){R_u}\left( z \right) = \not 1

and the series {R_u}\left( z \right) satisfying the content of proposition 3 is unique up to right multiplication by diagonal matrices \exp \left( {{a_1}z + {a_2}{z^3} + {a_3}{z^5} + ...} \right) with {a_k} = {\rm{diag}}\left( {a_k^1,...,a_k^N} \right) constants. And, in the case of conformal Frobenius structures the series is uniquely determined by the homogeneity condition

    \[\left( {z{{\not \partial }_z} + \sum {{u^i}{{\not \partial }_i}} } \right){R_u}\left( z \right) = 0\]

Now, let {\tilde H^L} be the space of Laurent polynomials in z with coefficients in the tangent space {T_u}H to the Frobenius manifold at a semisimple point u. For \bar q\left( z \right) \in {\bar H^L}_ +, let

    \[\Psi _u^{ - 1}\bar q\left( z \right) \equiv \left( {{{\bar q}^1}\left( z \right),...,{{\bar q}^N}\left( z \right)} \right)\]

and let us introduce the direct product

    \[{\Gamma ^ \otimes } = \tau \left( {\hbar ;{{\bar q}^1}} \right)...\tau \left( {\hbar ;{{\bar q}^N}} \right)\]

of N copies of the Witten-Kontsevich tau-function as an element of the Fock space of functions on \bar H_ + ^L. The series {R_u}\left( z \right) defines a symplectic transformation on {\bar H^L} and set

    \[{\hat R_u} = \exp {\left( {{\rm{In}}\,{R_u}} \right)^ \wedge }\]

and let

    \[{\hat \Psi _u} \equiv G\left( {{\Psi ^{ - 1}}\bar q} \right) \Rightarrow G\left( {\bar q} \right)\]

identify the Fock space with its coordinate version, and lastly,

    \[\sum\nolimits_{i = 1}^N {R_1^{ii}} d{u^i}\]

with R_1^{ii} the diagonal entries of the matrix {R_1} in the series

    \[{R_u}\left( z \right) = \not 1 + {R_1}/z + ...\]

which is the 1-form on the Frobenius manifold. Define also the function

    \[C\left( u \right) = \frac{1}{2}\int_L^u {\sum {R_1^{ii}} } d{u^i}\]

of u defined up to an additive constant.

Definition: let the total descendent potential of a semisimple Frobenius manifold be defined by the formula

 

eq9

 

and introduce the total ancestor potential of a semisimple Frobenius manifold

 

eq10

 

Property 1 The total descendent potential {\tilde D_{wk}} of a semisimple Frobenius manifold defined does not depend on the choice of a semisimple point u since both

    \[\left\{ {\begin{array}{*{20}{c}}{{S_u}\left( {1/z} \right)}\\{{\Psi _u}{R_u}\left( z \right){e^{\left( {U/z} \right)}}}\end{array}} \right.\]

satisfy the same equation

    \[{\nabla _z}S = 0\]

with coefficients rational in z. Hence, derivatives of {\hat D_{wk}} in the directions of the parameter u vanish as if

    \[\left\{ {\begin{array}{*{20}{c}}{S_u^{ - 1}}\\{{\Psi _u}{R_u}{e^{\left( {U/z} \right)}}}\end{array}} \right.\]

were inverse to each other.

Consequentially, the main property of this post:

Property 2 

The genus 1 Gromov-Witten potential {F^1}\left( t \right) of a semisimple Frobenius manifold is given by the formula

 

eq11

 

with

    \[\Delta _t^{ - 1}\left( u \right) = \left( {{{\not \partial }_{{u^i}}},{{\not \partial }_{{u^j}}}} \right)\]

being the inner squares of the canonical idempotents {\not \partial _{{u^i}}} of the semisimple Frobenius multiplication { \bullet _u} and the genus 1 descendent potential equals

 

eq12

 

where the partial derivatives are taken with respect to coordinates of {t^0} = \sum {t_0^\alpha } {\phi _\alpha }.

 

rutherfold

 

1 Response