Singular states of Toda Theory and N=2 Gauge Theories
Yutaka Matsuo (U. Tokyo)
based on arXiv:0911.4787 (PRD) with S. Kanno, S.Shiba, Y. Tachikawa
NCTS‐NTHU 2010/04/16
§1 Introduction
Mystery of M‐theory
Strongly coupled system with strange internal degree of freedom
M2 ~ O(N
3/2) M5 ~ O(N
3)
How to define multiple M‐branes?
M2 : BLG, ABJM...
M5 : ?? probably nonlocal field theory
Multiple M5 compactified on T2 should produce N=4 nonabelian gauge theory. Assuming the theory on M5 is local, standard KK reduction would give
vol(T2) Z
d4x Tr F2 On the other hand, the gauge action is,
g¡2 Z
d4x Tr F2
where the coupling constant g‐2 is proportional to the modulus of T2. It implies that M5 action should not be local.
Why M5 action should be nonlocal?
Direct construction of M5 brane action ((2,0) theory) would need great inspiration
Æ It would be better to have some concrete hint from a simplified set‐up
AGT relation?
Alday, Gaiotto, Tachikawa arXiv:0906.3219
N=2 gauge theory on X
Liouville
Toda theory on Σ
Multiple M5 compactified on X x Σ
Nekrasov’s
Partition function
ZN=2 SY M(a; m; q) = hei~¯1¢~Á(z1) ¢ ¢ ¢ ei~¯n¢~Á(zn)iToda
Correlation function
Nekrasov’s partition function for N=2 quiver gauge theory
Z = ZclassZ1¡loopZinst
Z1¡loop : product of DOZZ factors: OPE coe±cients for vertex
¡2(~ai ¡ ~aj + ²i) etc.
Zinst : product of factors associated with Young diagram X
Y~1;¢¢¢ ;~Yn
Yn k=1
qkj~Ykjzvec(~ak; ~Yk) £ ¢ ¢ ¢ zvec(~a; ~Y ) =
Y
s2Yk
Y
t2Yl
(¡²1LYl(s) + ²2AYi(s) + a)¡1 ¢ ¢ ¢
a: VEV of gauge field (Coulomb branch), m: mass of hyper multiplet ε: deformation parameter,
q,τ: coupling const.
It may be possible to guess M5 brane dynamics through Toda theory, W‐algebra, and N=2 quiver gauge theory
for example,
Geometrical interpretation of gauge symmetry behind W‐
algebra
Æ W gravity: nonlinear higher spin gauge theory in 2D
In order to proceed such relation, we need to establish the link with Toda theory more precisely
§2. Virasoro, Liouville
Toda & W
Main strategy of 2d world
Strong command by symmetry
Virasoro algebra: infinite dim. symmetry
determines the structures (such as correlation functions) of the system completely for
c < 1
Representation theory of Virasoro algebra: Highest weight module over
Lnjhi = 0; (n > 0); L0jhi = hjhi
|h> : highest weight state Å Æ Primary fields Vh(z) [Ln; Lm] = (n ¡ m)Ln+m + c
12(n3 ¡ n)±n+m;0
Among such states, there may exist so called Degenerate states |χ > which satisfies HWC. Such state is called singular state or null state since all the inner product with other
elements in the Hilbert space vanishes. We write,
Hilbert space is spanned by the states like
L¡n1 ¢ ¢ ¢ L¡nrjhi
jÂi = (L
¡n1¢ ¢ ¢ L
¡nr+ ¢ ¢ ¢ )jhi L
njÂi = 0; (n > 0)
jÂi ¼ 0
Classification of the singular vectors is essential to
understand the representation of Virasoro algebra as well as the physics that is associated with the representation.
(cf. Minimal models)
For the application to AGT relation, we need to calculate the correlation functions.
hV
h1(z
1) ¢ ¢ ¢ V
hr(z
r)i
where Vh(z) is the primary field with conformal dimension h There are two different methods to compute such correlation functions
1. Sandwiching complete basis of Virasoro module 1 =
X
jvi2H
jvihvj
where |v> is the orthonormal basis of the Virasoro module straightforward but analytic computation to all order is very difficult.
2. Free boson representation Dotsenko and Fateev
Introduce Liouville field φ S =
Z
d2z £
(@Á)2 + QÁR + ¹ei®Á¤
Rewrite the vertex operator as vertex operator,
V
h$ e
i¯§Á; h = 1
2 ¯
§(Q ¡ ¯
§)
The correlation functions are written as the
correlation functions between vertex operators.
However, an important fact is that one may introduce so called screening operators
S
§= Z
d³e
i®§Á(³); 1
2 ®
§(Q ¡ ®
§) = 1
Such operators commute with all Virasoro generators: it does not change the conformal properties of the correlation functions. It implies that the correlation functions are obtained in the following form.
hei¯1Á(z1) ¢ ¢ ¢ ei¯nÁ(zn)S+ ¢ ¢ ¢ S¡ ¢ ¢ ¢ i
where the integration of the screening currents is performed along some particular paths, usually between zi’s. It gives Selberg style integral for parameter ζ
Z
dr³+ds³¡(³+i ¡ ³¡j )¡®+®¡(za ¡ ³+)¡®+¯a(za ¡ ³¡)¡®¡¯a
Dotsenko and Fateev has managed to integrate this expression to obtain the exact expression for the four point functions for the some minimal model.
Notice: similar integration appears in the (generalized) matrix model
How about the system with c > 1 ?
Virasoro symmetry is not enough to control the system:
Extended chiral algebras
Kac‐Moody : spin 1 current
Super Virasoro: spin 3/2 curren(s)
W‐algebra: higher spin currents (spin 3,4,5...) Zamolodchikov: add spin 3 ``W‐algebra’’
Fateev‐Lukyanov: add 3, 4, ..., n‐1 ``Wn‐algebra’’
Definition of W‐algebra
Virasoro algebra + primary fields T (z)T (0) » c=2
z4 + 1
z2T (0) + 1 z@T T (z)W (0) = ¢
z2W + 1
z @W
Δ must be (half) integer to keep locality, modular inv. etc.
W (z)W (0) » c
3z6 + 2
z2T + 1
z3@T + 1
z2
μ 3
10@2T + 2q2¤
¶
+ ¢ ¢ ¢
¤ = : T2 : ¡ 3
10@2T; q2 = 3 22 + 5c For Δ=3
Feature of W algebras:
Nonlinear symmetry ( appearance of Λ ) Relation with Lie algebras
higher spin generators can be constructed out of Casimir operators
W3 / dabcJaJbJc : dabc : 3rd Casimir of g
Dimensions of extra generators: dimensions of Casimirs An : 2; 3; ¢ ¢ ¢ ; n
Dn : 2; 4; ¢ ¢ ¢ ; 2n
Free field representation of W
nalgebra
RN =:
YN m=1
μ
Q d
dz + ~hm ¢ @ ~Á
¶ :=
XN k=0
W(k)(z) μ
Q d dz
¶N ¡k
W(0) = 1; W(1) = 0; W(2) = T; ¢ ¢ ¢
RN is called ``Quantum Miura transformation’’
which appeared in the context of integrable system to describe Lax operator
W: KdV potential etc, φ: free bosons
Link with integrable models: KP hierarchy, Toda hierarchy
Minimal models of W
Nalgebra
c = (N ¡ 1) μ
1 ¡ N (N + 1) p(p + 1)
¶
; p = N + 1; N + 1; ¢ ¢ ¢ Classify system with ZN symmetry with c less than N‐1 For applications of N=2 SYM
c = (N ¡ 1) ¡
1 + N (N + 1)Q2¢
; Q = b + 1=b = ²1 + ²2 p²1²2 It matches with chiral anomaly on M5
I8(AN¡1) = (N ¡ 1)I8(1) + N (N2 ¡ 1)p2(N ) 24
observation by Alday, Benini, Tachikawa 0909.4776
Such relation exists for arbitrary ADE Lie algebras
Gauging chiral algebra: every chiral algebra can be derived from gauge symmetry in 2D
(super) Virasoro: 2d (super) gravity Kac‐Moody: 2d Yang‐Mills
Virasoro: Knizhnik‐Polyakov‐Zamolodchikov
S = S0 + Z
d2zhzzT (¹¹ z);
hzz : lightcone component of metric
rzT¹ = @zT ¡ 2@¹ z¹h ¹T ¡ h@z¹T =¹ c
12@z¹3h
! @z3¹h = 0; h = J(¡) + ¹zJ(0) + ¹z2J(+) Anomalous conservation law
J’s satisfy sl(2) current algebra
Extension to W‐algebra
S = S0 + Z
d2z(hzzT (¹¹ z) + AzzzW )¹
OPEs for h and A are described by sl(3,R) current algebra W‐gravity : YM ‘89
Correspondence between the representations of sl(n,R) current algebra and W algebras
for arbitrary A‐D‐E Lie algebra
Taking Conformal gauge
: description by `free’ bosons
2d gravity sl(2,R) KM
KPZ
Liouville
(Distler‐Kawai)
W‐gravity sl(n,R) KM Toda
Toda field theory
S = Z
d2zp g
Ã
(@ ~Á)2 + ¹
X
i
eb~eiÁ~ + Q
4¼R~½~Á
!
§3 N=2 Gauge Theories
Quiver Diagrams
Graphic description of gauge group and matter content of N=2 gauge theories
n SU(n) gauge group
n Hypermultiplet with n‐dim.
representation
n
n n Hypermultiplet with
bifundamental representation
2 2
2
2 2
N
f=4 SU(2) gauge theory
Brane configuration in IIA and Seiberg‐Witten curve
NS5
D4
Quiver diagram
Seiberg‐Witten curve
2 2
2
2 2
Brane configuration
(v ¡ ~m1)(v ¡ ~m2)t2 +c(v2 ¡ M v ¡ u)t
+c0(v ¡ m1)(v ¡ m2) = 0
Gaiotto’s curve (G‐curve)
puncture hypermultiplet
cylinder vector multiplet
modulus coupling const.
S‐duality of the system is rewritten as the modular transformation of G‐curve
For SU(N) (N>2), G‐curve is more complicated
n n n1 n2 nr
Configuration of gauge groups can not be arbitrary to keep conformal invariance
At a’s node, one has to attach ka=2na‐na‐1‐na+1
fundamental hypermultiplets to keep symmetry. ka must be positive. So one must attach increasing/decreasing sequence of numbers on the left/right ends. In G‐curve, these sequences are represented by punctures with the label of Young diagram
d1 d2 d3 dr
Young diagram assignment
wa = da+1 ¡ da;
X
wa = N
w0 w1
wl¡1 ... ...
N N
Examples
[1N]
Full puncture
2 3 4 N [N‐1,1]
Simple puncture
G‐curve for linear quiver
...
...
Simple puncture
More nontrivial example...
T
Ntheory
No coupling constants
tri‐fundamental matter which carries O(N3) dof
Building block for general N=2 theories
0 1
AGT conjecture
(Alday, Gaiotto, Tachikawa)Nekrasov’s partition function for SU(2) gauge theory
Liouville conformal block
(correlation function)
Z(q; a; m) = ZtreeZ1¡loopZinst
q: coupling constant
a: VEV of adjoint scalar field m: mass of bifundamental
hypermultiplet
hV®1(1)V®2(1)V®3(z)V®4(0)i Translation rule
αi Å a,m, z Å q
Extension to SU(N)
WyllardSU(N) Nf=2N, N=2 gauge theory Proposal:
simple puncture <‐‐> a singular vertex κωn‐1 (Fateev‐Litvinov) full puncture <‐‐> generic vertex Using the result of correlation function by FL, one may
reproduce Nekrasov’s partition function for SU(N) case also.
§4 Some original results
Construction of degenerate vertex for punctures with arbitrary Young diagram
Kanno‐M‐Shiba‐Tachikawa arXiv:0911.4787
• We construct singular vertex operators that corresponds to the puncture associated to general Young diagram.
• This is necessary to establish general AGT relation for SU(N)
• So far, Nekrasov partition functions for linear quiver are known so one may compare them with Toda correlation function
• In Toda side, it may be possible to calculate correlation function which gives a hint for partition functions of
strongly coupled system (such as TN)
• We will seek degenerate states which has null states at
level one. Only level one null states exist irrespective of the central charge, this is what we need.
Construction of level 1 null states
Screening operator Sj(§) =
Z dz
2¼iVj(§)(z) =
Z dz
2¼i : ei®§~ejÁ~ : General discussion: Kato‐Matsuda, Fateev‐Lukyanov
They commute with all W generators [Wr; Sj(§)] = 0 If we apply them to the highest weight state some times and if they are nonvanishing, they are singular vector
(Sj(§))`j~¯i = WYj~¯0i; j~¯i = ei~¯ ~Á(0)j0i
In order to have level one null states, we need impose
~ei ¢ ~¯ = 0; for some i
Since ~ei = (0; ; ¢ ¢ ¢ ; 1; ¡1; 0; ¢ ¢ ¢ ) , this condition is equivalent to
¯i = ¯i+1; for some i It implies that β takes the following form,
¯ = (¯~ (1); ¢ ¢ ¢ ; ¯(1)
| {z }
l1
; ¯(2); ¢ ¢ ¢ ; ¯(2)
| {z }
l2
; ¢ ¢ ¢ ; ¯(s); ¢ ¢ ¢ ; ¯(s)
| {z }
ls
)
Such states are indeed labeled by Young diagram
l1 l2
ls
Why or how we can claim that this is the desired vertex operator?
1. It should correctly reproduce the Nekrasov partition function Æ It requires the
computation of correlation function. We are working on it (to be published soon, hopefully)
2. It can correctly reproduce SW curve Æ Yes, we could do it at least in the semi‐
classical limit
3. At least one can show that such operators has real eigenvalues for W0. Æ OK. we did it.
Reality of eigenvalues of W
0looks like a trivial mathematical problem?
NOT REALLY SO since we have W‐system with c > N‐1!
β’s are not real nor pure imaginary but general complex number.
From the experience of Liouville, we expect to have
where p and ρ are real. However, it breaks the reality condition of eigenvalues of W generators. Correct prescriptions seems to be
¯~ = ~p ¡ iQ~½
¯ = ~p ¡ iQ(~½ ¡ ½Y ); ½Y = (½l1; ¢ ¢ ¢ ; ½lr)
Eigenvalues (weight) of W‐generators can be written out of power sum of βi. With above combination, such power sums becomes real.
In particular, vertex for massless matter fields corresponds to p=0 limit in above formula.
Derivation of Seiberg‐Witten curve in semiclassical limit
Definition of W‐generators
RN =:
YN m=1
μ Q d
dz + ~hm ¢ @ ~Á
¶ :=
XN k=0
W(k)(z) μ
Q d dz
¶N ¡k
SW curve is expected to be described by <RN V V... V>=0.
This is, however, difficult since we have factors of higher derivatives in the def. of Ws. So we take a `semi‐classical’
(or dispersionless) limit where we drop higher derivative terms. We may put
Q d
dz ! ix (c-number)
This limit can be justified when all mass parameters (or p in V) is very large
Semiclassical analysis
rN =:
YN m=1
³
x ¡ i~hm ¢ @ ~Á
´ :=
XN k=0
w(k)(z)xN ¡k
We evaluate the correlation function near the singular vertex operator at z=0. For ¯ = (¯~ (1); ¢ ¢ ¢ ; ¯(1)
| {z }
l1
; ¯(2); ¢ ¢ ¢ ; ¯(2)
| {z }
l2
; ¢ ¢ ¢ ; ¯(s); ¢ ¢ ¢ ; ¯(s)
| {z }
ls
)
h¢ ¢ ¢ ~rN(z)V¯~(0) ¢ ¢ ¢ >
= c0
Y
k
(v ¡ ¯(k))ll + c1z
Y
k
(v ¡ ¯(k))ll¡1(vd1 + ¢ ¢ ¢ ) + ¢ ¢ ¢ where r~N = zNrN; v = zx
This is precisely the same as SW curve at z=0
In particular, for
¯ = (¯~ (1); ¢ ¢ ¢ ; ¯(1)
| {z }
l1
; ¯(2); ¢ ¢ ¢ ; ¯(2)
| {z }
l2
)
l1 l2
We have very simple relation among level one states (mw0(m)w¡1(n) ¡ nw0(n)w¡1(m))j~¯i ¼ 0
One may confirm this relation for the insertion of vertex at arbitrary point
Toward proof of AGT relation
1. Formulae in the gauge side is completely known (after Nekrasov)
2. Nontrivial issue is how to compute the correlation functions in 2d side.
3. Two computation methods (insertion of complete basis &
Dotsenko‐Fateev integral) give some algorithm to give the correlation function.
4. For the Virasoro case (N=2), this seems to be established soon (some groups seem to be very close to it)
5. For the W case (N>2), there seems to be a technical difficulty (Fateev‐Litvinov: treatment of W‐1) which will need some inspiration from W‐gravity
§5 Summary and Speculation
We give (somewhat lengthy) review of Virasoro, W‐
algebra, W‐gravity etc with a hope that they will be useful to understand M5 dynamics in the future.
We proposed the degenerate vertex of Toda theory which should be used to derive AGT relation for
SU(N) quiver gauge theories.
We proved, in particular, that Seiberg‐Witten curve can be derived as VEV of quantum Miura
transformation.
A difficulty for the W case (N>2)
We can not replace W‐1 in the correlation function for the generic case (except for the linear quiver where we can use the null state condition)
We need extra coordinates to describe such generators The number of coordinates increases as N. What is the interpretation?
It may be possible to interpret such extra coordinates as coming from the string dynamics instead of particles on M5.