STA G. Conformal Field Theory in Momentum space. Kostas Skenderis Southampton Theory Astrophysics and Gravity research centre.

Transcription

1 Southampton Theory Astrophysics and Gravity research centre STA G Research Centre Oxford University 3 March 2015

2 Outline 1 Introduction

3 Introduction Conformal invariance imposes strong constraints on correlation functions. It determines two- and three-point functions of scalars, conserved vectors and the stress-energy tensor [Polyakov (1970)]... [Osborn, Petkou (1993)]. For example, O 1 (x 1 )O 2 (x 2 )O 3 (x 3 ) c 123 = x 1 x x2 x x3 x It determines the form of higher point functions up to functions of cross-ratios.

4 Introduction These results (and many others) were obtained in position space. This is in stark contrast with general QFT were Feymnan diagrams are typically computed in momentum space. While position space methods are powerful, typically they provide results that hold only at separated points ("bare" correlators). are hard to extend beyond CFTs The purpose of this work is to provide a first principles analysis of CFTs in momentum space.

5 Introduction Momentum space results were needed in several recent applications: Holographic cosmology [McFadden, KS](2010)(2011) [Bzowski, McFadden, KS (2011)(2012)] [Pimentel, Maldacena (2011)][Mata, Raju,Trivedi (2012)] [Kundu, Shukla,Trivedi (2014)]. Studies of 3d critical phenomena [Sachdev et al (2012)(2013)]

6 References Adam Bzowski, Paul McFadden, KS Implications of conformal invariance in momentum space Adam Bzowski, Paul McFadden, KS Renormalized scalar 3-point functions 15xx.xxxx Adam Bzowski, Paul McFadden, KS Renormalized tensor 3-point functions 15xx.xxxx

7 Conformal invariance Conformal transformations consist of dilatations and special conformal transformations. Dilatations δx µ = λx µ, are linear transformations, so their implications are easy to work out. Special conformal transforms, δx µ = b µ x 2 2x µ b x, are non-linear, which makes them difficult to analyse (and also more powerful). The corresponding Ward identities are partial differential equations which are difficult to solve.

8 Conformal invariance In position space one overcomes the problem by using the fact that special conformal transformations can be obtained by combining inversions with translations and then analyzing the implications of inversions. In momentum space we will see that one can actually directly solve the special conformal Ward identities.

9 Conformal Ward identities These are derived using the conformal transformation properties of conformal operators. For scalar operators: O 1 (x 1 ) O n (x n ) = x 1 /d x x n/d x=x 1 x O 1 (x 1) O n (x n) x=x n For (infinitesimal) dilatations this yields n n 0 = j + O 1 (x 1 )... O n (x n ). j=1 x α j x α j=1 j In momentum space this becomes n n 1 0 = j (n 1)d j=1 O 1 (p 1 )... O n (p n ), p α j p α j=1 j

10 Special conformal Ward identity For (infinitesimal) special conformal transformations this yields ( ) n 0 = 2 j x κ j + 2x κ j x α j x 2 j O 1 (x 1 )... O n (x n ) j=1 x α j In momentum space this becomes x jκ 0 = K µ O 1 (p 1 )... O n (p n ), ( n 1 K µ = 2( j d) p κ 2p α j j j=1 p α j p κ j + (p j ) κ p α j p jα )

11 Special conformal Ward identities To extract the content of the special conformal Ward identity we expand K µ is a basis of linear independent vectors, the (n 1) independent momenta, K κ = p κ 1K p κ n 1K n 1. Special conformal Ward identities constitute (n 1) differential equations.

12 Conformal Ward identities Poincaré invariant n-point function in d n spacetime dimensions depends on n(n 1)/2 kinematic variables. Thus, after imposing (n 1) + 1 conformal Ward identities we are left with n(n 1) 2 n = n(n 3) 2 undetermined degrees of freedom. This number equals the number of conformal ratios in n variables in d n dimensions. It is not known however what do the cross ratios become in momentum space.

13 Outline 1 Introduction

14 Outline 1 Introduction

15 Scalar 2-point function The dilatation Ward identity reads [ 0 = d p ] O 1 (p)o 2 ( p) p The 2-point function is a homogeneous function of degree ( d): O 1 (p)o 2 ( p) = c 12 p 1+ 2 d. where c 12 is an integration constant.

16 Scalar 2-point function The special conformal Ward identity reads 0 = K O 1 (p)o 2 ( p), K = d2 dp d 1 d p dp Inserting the solution of the dilatation Ward identity we find that we need 1 = 2

17 Scalar 2-point function The general solution of the conformal Ward identities is: O (p)o ( p) = c 12 p 2 d. This solution is trivial when = d + k, k = 0, 1, 2,... 2 because then correlator is local, O(p)O( p) = cp 2k O(x 1 )O(x 2 ) k δ(x 1 x 2 ) Let φ 0 be the source of O. It has dimension d =d/2 k. The term φ 0 k φ 0 has dimension d and can act as a local counterterm.

18 Position space [Petkou, KS (1999)] In position space, it seems that none of these are an issue: O(x)O(0) = C x 2 This expression however is valid only at separated points, x 2 0. Correlation functions should be well-defined distributions and they should have well-defined Fourier transform. Fourier transforming we find: d d x e ip x 1 x 2 = πd/2 2 d 2 Γ Γ( ) ( d 2 ) 2 p 2 d, This is well-behaved, except when = d/2 + k, where k is a positive integer.

19 Strategy Regularize the theory. Solve the Ward identities in the regulated theory. Renormalize by adding appropriate counterterms. The renormalised theory may be anomalous.

20 Regularization We use dimensional regularisation to regulate the theory d d + 2uɛ, j j + (u + v)ɛ In the regulated theory, the solution of the Ward identities is the same as before but the integration constant may depend on the regulator, O(p)O( p) reg = c(ɛ, u, v)p 2 d+2vɛ.

21 Regularization and Renormalization Now, in local CFTs: This leads to O(p)O( p) reg = c(ɛ, u, v)p 2 d+2vɛ. c(ɛ, u, v) = c( 1) (u, v) ɛ O(p)O( p) reg = p 2k [ c ( 1) We need to renormalise... ɛ + c (0) (u, v) + O(ɛ) + c ( 1) v log p 2 + c (0) + O(ɛ) ].

22 Renormalization Let φ 0 the source that couples to O, S[φ 0 ] = S 0 + d d+2uɛ x φ 0 O. The divergence in the 2-point function can be removed by the addition of the counterterm action S ct = a ct (ɛ, u, v) d d+2uɛ x φ 0 k φ 0 µ 2vɛ, Removing the cut-off we obtain the renormalised correlator: [ ] O(p)O( p) ren = p 2k C log p2 µ 2 + C 1

23 Anomalies The counter term breaks scale invariance and as result the theory has a conformal anomaly. The 2-point function depends on a scale [Petkou, KS (1999)] A 2 = µ µ O(p)O( p) = cp2 d, The integrated anomaly is Weyl invariant A = d d x φ 0 k φ 0 On a curved background, k is replaced by the "k-th power of the conformal Laplacian", P k.

24 Outline 1 Introduction

25 We would now like to understand 3-point functions at the same level: What is the general solution of the conformal Ward identities? What is the analogue of the condition = d + k, k = 0, 1, 2,... 2 Are there new conformal anomalies associated with 3-point functions and if yes what is their structure?

26 Conformal Ward identities Dilatation Ward identity 3 0 = 2d t + j=1 p j O 1 (p 1 )O 2 (p 2 )O 3 (p 3 ) p j t = The correlation is a homogenous function of degree (2d t ). The special conformal Ward identities give rise to two scalar 2nd order PDEs.

27 Special conformal Ward identities Special conformal WI 0 = K 12 O 1 (p 1 )O 2 (p 2 )O 3 (p 3 ) = K 23 O 1 (p 1 )O 2 (p 2 )O 3 (p 3 ), where K ij = K i K j, K j = 2 p 2 j + d j p j p j, (i, j = 1, 2, 3). This system of differential equations is precisely that defining Appell s F 4 generalised hypergeometric function of two variables. [Coriano, Rose, Mottola, Serino][Bzowski, McFadden, KS] (2013).

28 There are four linearly independent solutions of these equations. Three of them have unphysical singularities at certain values of the momenta leaving one physically acceptable solution. We thus recover the well-known fact that scalar 3-point functions are determined up to a constant.

29 Scalar 3-pt functions and triple-k integrals The physically acceptable solution has the following triple-k integral representation: O 1 (p 1 )O 2 (p 2 )O 3 (p 3 ) = C 123 p 1 d p 2 d 2 2 p 3 d 2 3 dx x d 2 1 K 1 d (p 1 x)k 2 d (p 2 x)k 3 d (p 3 x), where K ν (p) is a Bessel function and C 123 is an constant. This is the general solution of the conformal Ward identities.

30 Triple K-integrals Triple-K integrals, I α{β1 β 2 β 3 }(p 1, p 2, p 3 ) = 0 dx x α 3 j=1 are the building blocks of all 3-point functions. The integral converges provided 3 α > β j 1 j=1 p β j j K β j (p j x), The integral can be defined by analytic continuation when α + 1 ± β 1 ± β 2 ± β 3 2k, where k is any non-negative integer.

31 Renormalization and anomalies If the equality holds, α + 1 ± β 1 ± β 2 ± β 3 = 2k, the integral cannot be defined by analytic continuation. Non-trivial subtractions and renormalization may be required and this may result in conformal anomalies. Physically when this equality holds, there are new terms of dimension d that one can add to the action (counterterms) and/or new terms that can appear in T µ µ (conformal anomalies).

32 Scalar 3-pt function For the triple-k integral that appears in the 3-pt function of scalar operators the condition becomes d 2 ± ( 1 d 2 ) ± ( 2 d 2 ) ± ( 3 d 2 ) = 2k There are four cases to consider, according to the signs needed to satisfy this equation. We will refer to the 4 cases as the ( ), ( +), ( + +) and (+ + +) cases. Given 1, 2 and 3 these relations may be satisfied with more than one choice of signs and k.

33 Procedure To analyse the problem we will proceed by using dimensional regularisation d d + 2uɛ, j j + (u + v)ɛ In the regulated theory the solution of the conformal Ward identity is given in terms of the triple-k integral but now the integration constant C 123 in general will depend on the regulator ɛ, u, v. We need to understand the singularity structure of the triple-k integrals and then renormalise the correlators. We will discuss each case in turn.

34 The ( ) case = 2d + 2k This the analogue of the = d/2 + k case in 2-point functions. There are possible counterterms S ct = a ct (ɛ, u, v) d d x k 1 φ 1 k 2 φ 2 k 3 φ 3 where k 1 + k 2 + k 3 = k. The same terms may appear in T µ µ as new conformal anomalies. After adding the contribution of the countertrems one may remove the regulator to obtain the renormalised correlator.

35 Example: 1 = 2 = 3 = 2, d = 3 The source φ for an operator of dimension 2 has dimension 1, so φ 3 has dimension 3. Regularizing: ( π ) 3/2 O(p 1 )O(p 2 )O(p 3 ) = C 123 dx x 1+ɛ e x(p 1+p 2 +p 3 ) 2 0 ( π ) [ ] 3/2 1 = C ɛ (γ E + log(p 1 + p 2 + p 3 )) + O(ɛ).

36 Renormalization and anomalies We add the counterterm S ct = C ( 123 π ) 3/2 d 3+2ɛ x φ 3 µ ɛ 3!ɛ 2 This leads to the renormalized correlator, ( π ) 3/2 p 1 + p 2 + p 3 O(p 1 )O(p 2 )O(p 3 ) = C 123 log 2 µ The renormalized correlator is not scale invariant µ ( π ) 3/2 µ O(p 1)O(p 2 )O(p 3 ) = C There is a new conformal anomaly: T = φ O + 1 ( π ) 3/2 3! C 123 φ 3. 2

37 The ( +) case = d + 2k In this case the new local term one can add to the action is S ct = a ct d d x k 1 φ 1 k 2 φ 2 O 3 where k 1 + k 2 = k. In this case we have renormalization of sources, φ 3 φ 3 + a ct k 1 φ 1 k 2 φ 2 The renormalised correlator will satisfy a Callan-Symanzik equation with beta function terms.

38 Callan-Symanzik equation The quantum effective action W (the generating functional of renormalised connected correlators) obeys the equation ( µ µ + d d δ ) x β i W = d d x A, δφ i i ( x) This implies that for 3-point functions we have µ µ O i(p 1 )O j (p 2 )O j (p 3 ) = β j,ji ( Oj (p 2 )O j ( p 2 ) + O j (p 3 )O j ( p 3 ) ) + A (3) ijj, β i,jk = δ2 β i δφ j δφ, k {φl }=0 A(3) ijk ( x δ 3 1, x 2, x 3 ) = δφ i ( x 1 )δφ j ( x 2 )δφ k ( x 3 ) d d x A({φ l ( x)})

39 Example: 1 = 4, 2 = 3 = 3 in d = = 10 = 2d + 2k, which satisfies the ( )-condition with k = 1. There is an anomaly d d xφ 0 φ 1 φ = 4 = d + 2k, which satisfies the ( +) condition with k = 0. The following counterterm is needed, d 4 xφ 0 φ 1 O 3 There is a beta function for φ 1.

40 O 4 O 3 O 3 ( O 4 (p 1 )O 3 (p 2 )O 3 (p 3 ) = α 2 p 1 p 1 + α 8 [ ( (p 2 2 p 2 3)log p2 1 µ 2 log p2 3 µ 2 log p2 2 µ 2 (p p 2 3)log p2 2 µ 2 log p2 3 µ 2 ) I (non-local) (p 2 1 p 2 2)log p2 3 µ 2 + (p2 1 p 2 3)log p2 2 µ 2 + p2 1 ) ]

41 O 4 O 3 O 3 I (non-local) = 1 [ J 2 π log ( X p ) 2 p 3 Li 2 ( X p 2 p log p 1 log p 2 p 3 p 3 ( log Y p ) 1 p 3 ) ( Li 2 Y p )] 1, p 3 J 2 = (p 1 + p 2 p 3 )(p 1 p 2 + p 3 )( p 1 + p 2 + p 3 )(p 1 + p 2 + p 3 ), X = p2 1 p2 2 p2 3 + J 2, Y = p2 2 p2 1 p2 3 + J 2. 2p 2 p 3 2p 1 p 3

42 Callan-Symanzik equation It satisfies µ µ O 4(p 1 )O 3 (p 2 )O 3 (p 3 ) = ( α p 2 2 log p2 2 2 µ 2 + p2 3 log p2 3 µ 2 p ) 2 (p2 2 + p 2 3). This is indeed the correct Callan-Symanzik equation. (Recall that O 3 (p)o 3 (p) = p 2 log p2 µ 2 )

43 The (+ + +) and ( + +) cases In these cases it is the representation of the correlator in terms of the triple-k integral that is singular, not the correlator itself, O 1 (p 1 )O 2 (p 2 )O 3 (p 3 ) = C 123 p 1 d 2 1 p 2 d 2 2 p 3 d 2 3 I d/2 1,{ 1 d/2, 3 d/2, 3 d/2} Taking the integration constant C 123 ɛ m for appropriate m and sending ɛ 0 results in an expression that satisfies the original (non-anomalous) Ward identity. In other words, the Ward identities admit a solution that is finite.

44 The (+ + +) case = d 2k For example, for k = 0 the finite solution to the Ward identities is O 1 (p 1 )O 2 (p 2 )O 3 (p 3 ) = cp ( ) 1 p ( ) 2 p ( ) 3 When the operators have these dimensions there are "multi-trace" operators which are classically marginal where k 1 + k 2 + k 3 = k. O = k 1 O 1 k 2 O 2 k 3 O 3

45 The ( + +) case = 2k For k = 0, there are "extremal correlators". In position, the 3-point function is a product of 2-point functions c 123 O 1 (x 1 )O 2 (x 2 )O 3 (x 3 ) = x 2 x x3 x In momentum space, the finite solution to the Ward identities is O 1 (p 1 )O 2 (p 2 )O 3 (p 3 ) = cp (2 2 d) 2 p (2 3 d) 3 When the operators have these dimensions there are "multi-trace" operators of dimension 1 where k 2 + k 3 = k. O = k 2 O 2 k 3 O 3

46 Outline 1 Introduction

47 New issues arise for tensorial correlation functions, such as those involving stress-energy tensors and conserved currents. Lorentz invariance implies that the tensor structure will be carried by tensors constructed from the momenta p µ and the metric δ µν. After an appropriate parametrisation, the analysis becomes very similar to the one we discussed here. In particular, these correlator are also given in terms of triple-k integrals.

48 Diffeomorphism and Weyl Ward identities The fact that classically a current and the stress-energy tensor are conserved implies that n-point functions involving insertions of α J α or α T αβ can be expressed in terms of lower-point functions without such insertions. The same holds for correlation functions with insertions of the trace of the stress-energy tensor. The first step in our analysis is to implement these Ward identities. We do this by providing reconstruction formulae that yield the full 3-point functions involving stress-energy tensors/currents/scalar operators starting from expressions that are exactly conserved/traceless.

49 Example: T µν OO Ward identities p ν 1 T µν (p 1 )O(p 2 )O(p 3 ) = p 3µ O(p 3 )O( p 3 ) + (p 2 p 3 ) T µ µ (p 1 )O(p 2 )O(p 3 ) = O(p 3 )O( p 3 ) + (p 2 p 3 ) Reconstruction formula T µν (p 1 )O(p 2 )O(p 3 ) = t µν (p 1 )O(p 2 )O(p 3 ) [ + p α 2 T α µν (p 1 ) ] d 1 πµν (p 1 ) O(p 2 )O( p 2 ) + (p 2 p 3 ), where t µν (p 1 )O(p 2 )O(p 3 ) is transverse-traceless and π µν (p) = δ µν pµ p ν p 2, T µν α (p) = 1 [ ( p 2 2p (µ δ ν) α pα δ µν + (d 2) pµ p ν )] d 1 p 2

50 Tensorial decomposition It remains to determine the transverse-traceless part of the correlator which is undetermined by the Weyl and diffeomorphism Ward identities. We now use Lorentz invariance to express the transverse-traceless correlator in terms of scalar form factors.

51 Example: T µν OO The tensorial decomposition of t µν (p 1 )O(p 2 )O(p 3 ) involves one form factor: t µν (p 1 )O(p 2 )O(p 3 ) = Π µν αβ (p 1)A 1 (p 1, p 2, p 3 )p α 2 p β 2, where Π µν αβ (p 1) is a projection operator into transverse traceless tensors. An inefficient parametrization can lead to proliferation of form factors. The state-of-the-art decomposition of T µ1 ν 1 T µ2 ν 2 T µ3 ν 3 [Cappelli et al (2001)] prior to this work involved 13 form factors, while the method described here requires only 5.

52 Dilatation and special conformal Ward identities It remains to impose the dilatation and special conformal Ward identities. The dilatation Ward identity imply that the form factors are homogeneous functions of the momenta of specific degree. The special conformal Ward identities (CWI) imply that that the form factors satisfy certain differential equations. These split into two categories: 1 The primary CWIs. Solving these determines the form factors up to constants (primary constants). 2 The secondary CWIs. These impose relation among the primary constants.

53 Example: T µν OO primary CWI The primary CWI is K ij A 1 = 0, i, j = 1, 2, 3. This is precisely the same equation we saw earlier in the analysis of OOO. The general solution in given in terms of a triple-k integral A 1 = α 1 I d/2+1{ d/2, d/2, d/2}, where α 1 is constant (primary constant).

54 Example: T µν OO secondary CWI The secondary CWI is ( c 1 (p) + c 2 (p) ) + c 3 (p) A 1 OO p 1 p 2 where c 1 (p), c 2 (p), c 3 (p) are specific polynomials of the momenta. This equation then determines the primary constant α 1 in terms of the normalization of the 2-point function of O. T µν OO is completely determined, including constants, by conformal invariance.

55 The general case If we have n form factors then the structure of the primary CWI is K 12 A 1 = 0, K 13 A 1 = 0, K 12 A 2 = c 21 A 1, K 13 A 2 = d 21 A 1, K 12 A 3 = c 31 A 1 + c 32 A 2, K 13 A 3 = d 31 A 1 + d 32 A 2, K 12 A n = n 1 j=1 c nja j K 13 A n = n 1 j=1 d nja j where c ij, d ij are lower triangular matrices with constant matrix elements. These equations can be solved in terms of triple-k integrals. The solution depends on n primary constants, one for each form factor. The secondary Ward identities yield relations between the

56 Example: T µ1 ν 1 T µ2 ν 2 T µ3 ν 3 In d > 3 there are 5 form factors and thus 5 primary constants. The secondary CWI impose additional constraints and we are left with the normalization c T of the 2-point function of T µν and two additional constants. In d = 4 the normalization of the 2-point function and one constant can be traded for the conformal anomaly coefficients, c and a. Thus, in d = 4 the conformal anomaly determines T µ1 ν 1 T µ2 ν 2 T µ3 ν 3 up to one constant.

57 Example: T µ1 ν 1 T µ2 ν 2 T µ3 ν 3 in d = 3 In d = 3 there are only 2 form factors and thus 2 primary constants. The secondary Ward identities relates one of them with the normalization of the 2-point function.

58 Remarks In the same manner one can obtain all three-point functions involving the stress-energy tensor T µν, conserved currents J a µ and scalar operators. In odd dimensions the triple-k integrals reduce to elementary integrals and can be computed by elementary means. In even dimensions the evaluation of the triple-k integrals is non-trivial. In special cases, which include all 3-point functions of T µν and J a µ in even dimensions, non-trivial renormalization is needed.

59 Higher-point functions? The complexity of the analysis increases because the number of possible tensor structure and thus form factors increases. The form factors now depend also on the number of independent scalar products p ij = p i p j, i, j = 1, 2,..., n, i < j The number of independent such scalar products is n(n 3)/2. This is equal to the number of independent cross-ratios.

60 Outline 1 Introduction

61 /Outlook We obtained the implications of conformal invariance for three-point functions working in momentum space. We discussed renormalization and anomalies. The presence of "beta function" terms in the Callan - Symanzik equation for CFT correlators is new. It would be interesting to understand how to extend the analysis to higher point functions. What is the momentum space analogue of cross-ratios? Bootstrap in momentum space?