LOCALIZATION AND
VORTICES
Kazutoshi Ohta
Meijigakuin University
The Localization Theorem (Duistermaat-Heckman theorem) essentially works behind the integrability.
Lower dimensional gauge theories
(3-dim Chern-Simons, 2-dim Yang-Mills, etc.) Supersymmetric quantum mechanics
Topological (supersymmetric) gauge theories Reduced matrix model from 3d CS or 2d YM Topological strings
Non-critical strings
There exist various kinds of exact solvable systems:
Introduction
1. Introduction
2. What is the localization?
3. Application to 2d field theory
4. Volume of the moduli space of the Abelian vortex 5. Other applications
6. Wilson loops 7. Conclusion
Plan of my talk
What is the localization?
Originally it appears in the partition function of the statistical model.
The integration formula by Duistermaat and Heckman states that if H determines the symplectic action of U(1) on the phase space the integral localizes to the critical points of H
Z = 1 βn
�
dH=0
� det |ωij|
det |∂ijH|e−βH Z =
�
M
ωn e−βH(�x,�p)
where is a symplectic 2-form on the 2n-dim phase space M.
ω
Outline of the proof
There exists a vector field X associated with H and it satisfies (d + ιX)(H + ω) = dH + ιXω = 0
Z = 1 βn
�
M
e−β(H+ω)
If we add exact term to the “action”, the partition function is independent of the additional term
namely
indep. of λ
(ιXω = Xµωµνdxν) If we define , it satisfies and forms an equivariant cohomology on M.
(d + ιX)2 = LX
dX ≡ d + ιX
(dXΩ = 0 & Ω �= dXψ & LXΩ = 0) dX
∂Z
∂λ ∝ 1 βn
�
M
dX(ψe−β(H+ω)−λdXψ) = 0 Z = 1
βn
�
M
e−β(H+ω)−λdXψ
indep. of λ
Let us now choose
Bi-Hamiltonian structure (H,ω)↔(K,Ω) naturally appears if X is an isometry of M.
which satisfies
dX(H + ω) = 0 dX(K + Ω) = 0
Recall that the partition function is independent of the coupling of the exact term, we can take the large coupling limit
without changing the value of the integral:
dX
Z = 1 βn
�
M
e−β(H+ω)−λ(gµνXνXν+··· )
Z ∼ 1 βn
�
M
√detgδ(X)e1 −β(H+ω)−···
λ → ∞
dXψ = dX(ιXg) = K + Ω, K = ιXιXg, Ω = dιXg
metric
If we carefully integrate out all of the variables and calculate Jacobians in the large coupling limit, we obtain the DH
localization formula
Thus, the integral of the partition function is localized at X=0, that is the fixed points of the Hamiltonian vector flow (dH=0)
Z = 1 βn
�
dH=0
� det |ωij|
det |∂ijH|e−βH
Example Harmonic oscillator
The gaussian integral also can be understood as the sum over the fixed points of the Hamiltonian vector flow
H = p�2
2m + 1
2mω2�x2
Z = �
d�xd�p e−β(2m�p2 +12mω2�x2)
= 1 βn
� 2π ω
�n
e−βH(0,0)
There exists the following correspondence
DH formula Supersymmetric theory
super (BRST) charge Q
isometry gauge symmetry
coordinates and forms bosons and fermions(ghosts) bi-Hamiltonian BRST exact action
dX = d + ιX
The localization is very important to the integrability of the supersymmetric systems (owing to a cancellation of the
functional determinants between boson and fermions).
Relation to SUSY systems
[Witten (1992)]
We first consider the BF-type 2d U(N) gauge theory on a Riemann surface with genus g
Application to 2d Yang-Mills theory
Z =
�
DΦDAe−Tr
R
Σg iΦF
where .F = dA + A ∧ A
Integrating out Φ first, we get Z =
�
DA δ(F )
This means that the partition function of the BF theory calculates the volume of the moduli space (space of solutions) of the flat
connections F=0.
This “bosonic” 2d BF theory is nicely embedded in the 2d topological (cohomological) field theory:
BRST transformations is given by
The following BF theory action is Q-closed (but not Q-exact) under the above BRST symmetry
QA = iλ, Qλ = −dAΦ, QH = i[Φ, χ], Qχ = H,
Q ¯Φ = iη, Qη = [Φ, ¯Φ]
QΦ = 0
O ≡ Tr
�
Σg
iΦF + 1
2λ ∧ λ
QO = 0, O �= Q(something)
Hamiltonian
The “supersymmetric action” is Q-exact (of course Q-closed).
The partition function is independent of the coupling constant of the SUSY action and localized at the flat connection F=0. So the observable of the cohomological operator gives the partition
function of BF theory.
F = 0 S = 1
h2Q
� Tr
�
η[Φ, ¯Φ] + 1
2χ(H − F ) + λdAΦ¯
�
VEV in TFT
PF of bosonic FT
bi-Hamiltonian
�e−TrR iΦF +12λ∧λ�
TFT = �
DΦDADλ · · · e−TrR iΦF +12λ∧λ+h21 Q{∗}
= �
DΦDADλe−TrR iΦF +12λ∧λ
=
�
DΦDAe−TrR iΦF
This structure of the BF theory means that the localization theorem (DH formula) also works:
After fixing the gauge Φ=diag(φ1,φ2,...,φN), we find
�
DΦDAe−Tr R iΦF
= 1 N !
�
�k∈ZN
� �N
a=1
dφa �
a<b
(φa − φb)2−2ge−i PNa=1 kaφa
= 1 N !
�
�
n∈ZN
�
a<b
(na − nb)2−2g
where ka is the Chern number of a-th Cartan and we have used the Poisson resummation formula:
�
�k∈ZN
e−i Pa kaφa = �
�
n∈ZN
�
a
δ(φa − na)
We can deform the 2d BF theory to 2d Yang-Mills theory without spoiling the localization properties, since we can add the Q-
closed operator:
where .
Deformation to 2d Yang-Mills theory
S = Tr
�
Σg
iΦF + µ
2 Φ2
Integrating out Φ
S = 1
4gYM2
�
Σg
F2 gYM2 = µ/2
The partition function has the expression of the discrete sum Z = 1
N !
�
�n∈ZN
�
a<b
(na − nb)2−2ge−gYM2 A Pa n2a
area [Migdal]
2d BF and YM theory evaluate essentially the volume of the moduli space of the flat connection F=0. We now consider the Abelian vortex equations:
Volume of moduli space of vortices
The BRST transformations:
[A. Miyake, N. Sakai, KO (work in progress)]
F = ω
2 (1 − |h|2)
∂h = ∂¯¯ h = 0
QA = iλ, Qλ = −dφ, Qh = iψ, Qψ = φh, Q¯h = iψ†, Qψ† = −φ¯h, Qφ = 0.
To impose the holomorphic constraint for h (F-term constraints), we introduce extra fields and BRST transformations:
The Q-cohomological (Q-closed but not Q-exact) action:
QYz = iφχz, Qχz = Yz, QYz¯ = −iφχz¯, Qχz¯ = Yz¯.
S0 = �
Σg
�
iφ �
F − ω
2 (1 − |h|2)�
+ 1
2λ ∧ λ + ω
2 ψψ†
�
We can also add the Q-exact action (SUSY action) without changing the partition function, and integrate out all fields exactly. The result of the partition function is
where k is the Chern number: k = 1 2π
�
Σg
F
Zk = 1
(2π)2−2g
� ∞
−∞
dφ � 2π iφ
�k+1−g
eiφ(12A−2πk)
This integral of the partition function reduces to a simple residue integral. So the partition function is localized at a pole. Finally we obtain
For example, if g=0, we have Zk = πk
�g j=0
(4π)j (A − 4πk)k−j g!
j!(k − j)!(g − j)! .
Zk = πk (A − 4πk)k
k! .
This volume is smaller than the moduli space of the point like objects on the sphere:
The vortex has a hard core
vs
Zk < πAk k!
This result is previously calculated by Manton and Nasir (1998) in the different way. They need:
1. Details of the solution of the vortex equations 2. Structure of the vortex moduli space metric
3. Integrals over the moduli space with the metric However the localization method does not use the concrete solutions of vortices!
We agin consider the thermodynamical partition function of the vortices (do not confuse with the field theory partition function!):
Thermodynamics of the vortices
For free particles (solitons) , then we have Z = 1
�2k
�
dkpdkx e−βH(p,x)
H = 12gijpipj
We can calculate the free energy F=-T ln Z and equation of state
Z = � 2π2T
�2
�k �
Mk
dkx�
det gij
= � 2π2T
�2
�k
Vol(Mk)
P = kT
A − 4πk
pressure
for g=0 (S2)
Volume of the moduli space of the non-Abelian vortex system (Hitchin system) [Moore, Nekrasov and Shatashvili (1997)]
Other applications 1
Fz ¯z + [Wz, Wz¯] = 0, Dz¯Wz = DzWz¯ = 0,
Localization
Z = 1 N !
�
φ�∈RN
µ(φ)−1+g �
a<b
(φa − φb)2−2g �
(φa − φb)2 + �2�1−g
where is solutions of the following Bethe ansatz equationsRN e2πiφi �
i�=j
φj − φi + i�
φj − φi − i� = 1 and ε is a regulator of the flat direction in W
Volume of the moduli space of the instantons (ADHM equations) [Nekrasov (2002)]
Other applications 2
Localization
Nekrasov’s formula
[B1, B1†] + [B2, B2†] + II† − J†J [B1, B2] + IJ = 0
Prepotential of 4d N=2 SUSY YM theory
Reduced matrix model from 3d Chern-Simons theory on S3 or 2d YM theory on S2 [Ishiki, KO, Shimasaki, Tsuchiya (2008)]
Other applications 3
dimensional reduction
S = Tr
�
S2
iΦF + µ
2 Φ2 S = Tr
�
S3
AdA + A3
dimensional reduction
SMM = 1 g2Tr
�
Mi � 1
2Mi + i
6�ijk[Mj, Mk]
��
i = 1, 2, 3
The partition function of the reduced matrix model is exactly calculated by using the localization:
This partition function is much more complicated rather than the 2d YM theory one. But in the large N limit, we recover the
partition function of U(r) 2d YM theory (or 2d CS theory)
Z2dYM = C �
�n∈Zr
� �r
l=1
dyl �
i<j
((yl − ym)2 − (nl − nm)2)e−g2YMA1 Prl=1(yl2+n2l)
Minahan-Polychronakos form
(Poisson resummation from Migdal’s form)
Zr = �
d�∈Zr
�
Rd�
�r l=1
dyl �
l<m
(yl − ym)2 − (dl − dm)2
(yl − ym)2 − (dl + dm)2e−2g21
Pr
l=1(14dlyl2+121 d3l−121 dl)
Wilson loop
A perfect circle Wilson loop
or some class of Wilson loops is Q-closed (but not Q-exact) in SUSY theory.
This fact is investigated in 4d N =4 SYM theory [Erickson-
Semenoff-Zarembo, Drukker-Gross]. It is also proven by using the localization theorem [Pestun(2007)].
The Q-closed Wilson loop does not change the
localization properties. So we can evaluate the vev of the Wilson loop exactly.
WR(C) = TrR
�
P exp
�� (Aϕ + Φ cos θ)dϕ
��
It is also important that the Q-closed Wilson loop in SUSY reduces to solvable one in equivalent “bosonic” theory
Conclusion
The localization reduces the infinite dimensional problem to the finite one.
The localization plays important role in supersymmetric (exact solvable) field theory.
A special Wilson loop is also solvable by the localization.
I also would like to apply this method to the calculations of non-perturbative dynamics in supersymmetric theories (string theory, field theory, lattice field theory, matrix models and
more!)
Bosonic theory
Supersymmetric field theory
Supergravity/
string theory Holography
Localization