Spinning Witten Diagrams: From Geodesic to Mellin · 2020. 6. 24. · Spinning Witten Diagrams:...
Transcript of Spinning Witten Diagrams: From Geodesic to Mellin · 2020. 6. 24. · Spinning Witten Diagrams:...
Spinning Witten Diagrams: From Geodesic to Mellinwith En-Rui Kuo and Hideki Kyono, JHEP 1705 (2017) 079 +1712.07991
Heng-Yu ChenNational Taiwan University
OIST Bootstrap Symposium, March 2018
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
In a d-dim. CFT, 1, 2 3 point correlations for the primary operators{O∆i ,li (xi )} are fixed by conformal symmetries, 4 point is the firstnon-trivial one:
where Oi = O∆i ,0, xij = xi − xj , ∆ij = ∆i −∆j , F (u, v) is a function of:
Decomposing F (u, v) into different OPE channels (say s-channel):
where λ12O and λ34O are OPE coefficients, G∆,J(u, v) and W∆,J(xi ) arethe “conformal block” and “conformal partial wave” for O∆,J family.
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
I The scalar conformal block G∆,J(u, v) needs to satisfy the quadraticCasimir equation” [Dolan-Osborn 2003, 2011]
for even d , G∆,J(u, v) is given by hypergeometric functions.
I Generally, the conformal partial wave W∆,J(xi ) has the followingintegral representation:
where Od−∆,J(x0) is the “shadow” operator and integration kernal:
This is sometimes called “spectral/split” representation of CPW.
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
For CFTd with weakly coupled AdSd+1 dual, e.g. N = 4 SYM, or moregeneral large N CFTs, we usually divide {O∆,J} into: [D‘Hoker-Freedman 2002]
where ∆ can receive quantum corrections. In 1/N counting, they giveleading the contributions to correlation functions.In addition at some sub-leading 1/N, we have contributions from:
where ∆(ij)(n, J) = ∆i + ∆j + 2n + J + γij(n, J).
where P(12)n,J and P
(34)m,J are products of OPE coefficients for (12) and (34)
channels.
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
The scalar four point function is also known to enjoy following Mellinrepresentation (Mack, Penedones, Fitzpatrick, Kaplan, Raju, Van Rees, Paulos +....)
Here the integrand M(s, t) is called “Mellin Amplitude” (or reduced?):
with kinematical Mack polynomial Pν,J(s, t) and {δij} are:
We have clean separation among the single trace and double traceoperators ∼ O∆1∂
J(∂2)mO∆2 in Mellin representation, in particularM(s, t) is only fixed by the spectrum of single trace operators.
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
Mellin amplitude M(s, t) shares many similarities with flat space QFTscattering amplitudes, in particular, introducing fictitious momenta {pi}:
We have four constraints, hence two independent Mandelstam variables(s, t) (Can also have u =
∑4i=1 ∆i − (s + t)):
We can now consider expand in either of (s, t, u) variable, say:
where the infinite poles are precisely located at O∆,J with twistτ = ∆− J and its descendants (∂2)mO∆,J .
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
We also consider spinning primary O∆i ,li , transforming in symmetrictraceless rep., their 3 point function is again fixed by conformalsymmetries: [Costa et al 2011]
Here (Pi ,Zi ) are the embedding coordinates and polarization vectors:
where
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
The primary operator O∆i ,li (Pi ,Zi ) is a degree li polynomial in Zi
where O∆i ,A1...Ali(Pi ) is a symmetric traceless transverse (STT) tensor.
Here we have also defined the “box basis” for spinning 3 point functions:
where Pij = −2Pi · Pj , τi = ∆i + li and tensor structures are encoded in:
Recovering the tensor indices using differential operator DZi .
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
I The triplet of non-negative integers {n12, n13, n23} need to satisfy:
and these partitions label the possible resultant tensor structures.
I Another useful tensor basis is the “differential basis”:
where the shift operator Σa,b : (∆1,∆2)→ (∆1 + a,∆2 + b) andD11,D12,D22 and D21 only involve P1,2 and Z1,2:
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
I In differential basis, we explicitly break the permutation symmetryamong {O∆i ,li (Pi ,Zi )}, and differential operators only act on single(0, 0, l3) three point function.
I This is particularly useful when considering only the contribution ofsymmetric traceless operator O∆,J to the spinning 4 point function:
I Using the differential basis to express the 3 point functions, thespinning CPW can be related to scalar CPW as:
where Dn10,n20,n12
Left and Dn30,n40,n34
Right are the composite differentialoperators formed from Dij and Hij , i , j = 1, 2, 3, 4 plus shiftoperator Σa,b.
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
I We can also easily obtained the integral representation:[Simmons-Duffin 2012, HYC, Kuo, Kyono]
where τ (not twist ∆− J) include the effects of shift operator Σa,b. Wewill explain how identical expressions can be obtained holographically.
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
In AdSd+1/CFTd correspondence, CFT correlation functions are mappedto so-called Witten diagrams, the first non-trivial one is the 3pt vertex:
For our purpose, we can consider the following interaction vertex:
corresponding to holographic dual of scalar-scalar-tensor correlator.This yields a contribution to the 3pt correlation function as: [Costa et al 2014]
where W A is the bulk polarization vector and KA is its projector for indexcontractions.
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
Here the bulk-boundary propagator for massive spin-J particle is:
where τ = ∆ + J. Like in CFTd , AdS-isometries almost fix I0,0J :
where the overall factor b(∆1,∆2,∆, J) is the result of computation:
and the poles in ∆ capture double trace operators contributions, once weuse them to construct the four point functions.
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
Similarly, for general three spin interaction vertex, we have:[Sleight-Taronna 2016]
where the operators are given by:
Other combinations are equivalent up to integration by parts.
The integrated results can now be expressed in terms of linearcombination of box tensor structures (Diagonalization Problem):
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
Now for the 4 pt functions, we normally consider both contact andexchange diagrams:
They respectively yield the following contributions:
where Π∆,J(X , X ) is the spin J bulk to bulk propagator.
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
AdSd+1 analogue of splitting 4 point into product of 3 point functions isgiven by “split representation”: [Costa et al 2014]
We split the bulk-bulk propagator into products of bulk-boundary ones:
where aJ−k(ν), k = 0, 1, . . . , J are meromorphic functions in ν.
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
I We can now study the 3-point interaction vertices at X and X usingthe results earlier, and express them in terms of CFT structures:
I However when integrating with aJ−k(ν), the pre-factors from3-point functions, 4-point Witten diagram of spin-J exchange gives:
where the spectral function bl(ν) captures the physical spectrum:
Now we have both single trace and double trace contributions.
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
Q: What computes individual conformal blocks/partial waves?A: Geodesic Witten Diagrams. [Hijano et al 2015]
Quick Justification: What else? Instead of integrating over entireAdSd+1, we consider the trajectories which still preserve theAdS-isometries hence conformal symmetries.
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
More explicitly, we consider following geodesics:
and restrict the exchanged spinning particle to move along them.Such a 4-point geodesic Witten diagram gives the following:
This double integral along the geodesics can be shown such that:
by direct integration for low J exchange or by verifying that it satisfiesquadratic Casimir equation.
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
We can have an alternative proof by cutting 4-point geodesic Wittendiagram and consider the contributions from geodesic 3-point diagrams:[HYC-Kuo-Kyono]
Mathematically, this means we applied the following identity:
between the bulk to bulk and bulk to boundary propagators.
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
Generic 3-point interaction vertex along the geodesic, hC1...CJis divergent
free:
We have the following contribution, precisely the building block of CPW:
Fusing them back together using the kernal ν2
ν2+(h−∆)2 , we recovered
precisely the integral representation of scalar conformal block.
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
Scalar 4-point geodesic Witten diagrams can be regarded asdecomposition of Witten diagrams, by realizing their respective 3-pointbuilding elements only differ by ν-dependent overall factors, i .e.
Identical relation also exists for lower l spin, l = 0, . . . , J − 1 terms in thesummation, substitute back to P0 and ν integrations, we have:
We decompose the spin-J exchange Witten diagram into sum of geodesicWitten diagrams for single trace operator (Spin-J) and double traceoperator (Spin-l ≤ J). After non-trivial cancelation residues from thespurious poles.
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
For holographic reconstruction of spinning conformal blocks, we firstconsider geodesic 3-point functions for three primary operators with spins:
Arbitrary three point interaction vertex at X (λ) involving contractionswith metric and covariant derivatives yield sum of integrals like:
where τi = ∆i + li and Ql1,l2,l3 (Pi ,Zi ,X (λ)) is a transverse polynomial inZi of degree li .
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
I It is known that from earlier analysis: [Costa et al 2011]
the extra powers of e±λ can be absorbed as shift in ∆1,2, onlyscalar 3-point geodesic integral to do. Final result can be expressedin terms of linear combination of box tensor basis.
I Converting various box tensor basis into differential basis:
the fusion procedure then goes through identically as the scalargeodesic Witten diagram.
We have established that the spinning geodesic Witten diagrams generallyproduces linear combination of spinning conformal blocks for same O∆,J .
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
To provide dictionary between the bulk 3-point interaction vertices andthe spinning conformal blocks, adopt the parameterization of three spininteraction vertex earlier, with one important modification:
We have explicitly broken the cyclic properties among {O∆i ,li}, butnevertheless we still have the same number of independent vertices andbox tensor basis, as expected for linear combinations.
I For (l1 − l2 − 0), we have following vertices:
Direct computation then yields following contributions:
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
More general however, we have matrix diagonalization problem.
I For (1− 1− 2) there are five possible box tensor basis:
we label them in this ordering as {[Ir ]}, r = 1, 2, 3, 4, 5.
I There are also five possible vertices for 3-point geodesic WD:
their respective contributions are denoted as: {[Ir ]}, r = 1, 2, 3, 4, 5.
They are related by an invertible 5× 5 matrix [Ir ] = Trs [Is ], and for∆1 = ∆2,∆0 = ∆:
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
The Witten Diagrams also give us natural derivation for the Mellin partialamplitudes, or precisely the Mack polynomials(Slight change of notation for dynamical prefactors):
The kinematical integral Iν,J(Pi ) contributes to most of the scalar Mellinamplitudes after applying Symanzik star-formula:
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
I Stripping off the Γ-functions responsible for the double traceoperators and Pν,J(s, t) is given by: (Mack, HYC+Kuo+Kyono)
which is proportional to Mack polynomial Pν,J(s, t) up to spuriouspoles.
I The dynamical parts are encoded in the remaining pre-factors forthree point Witten functions and cutting kernel, and we can expressthem into ων,J(t) and the spectral function bJ(ν).
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
I For the derivation of spinning Mellin amplitude, we take similarstrategy and parameterize the spinning four point functions as:
The dynamical b(kL,R ,nL,R) depends on the specific basis forinteraction vertices, and determine the spectrum of double traceoperators.
I The kinematical I(nL,nR )ν,J (Pi ,Zi ) can be computed by generalizing
the Symanzik formula to include the polarization vectors {Zi}:(HYC+Kuo+Kyono)
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
This is a complicated long expression! But the salient features are:
I It now contains alternative basis of tensor structures Vij and Hij
which are linear combinations of Vi,jk and Hij .
I The parameters {δij} and non-negative integers {aij , bij} satisfy theconstraints:
I We can subsequently identify Mellin variables (s, t) in this case as:
I The generalized Mack polynomial Pν,J(s, t; aij , bij) does not containany poles in (s, t) variables.
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
Okay.....in case you are still wondering:
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
The proposal for the Mellin representation of the Spinning WittenDiagrams is:
We can show that the poles corresponding to the double trace operatorsin b(kL,R , lL,R) can be canceled by the zeros from denominators
Γ(δ12)Γ(δ34), hence M(nL,nR )ν,J (s, t; aij , bij) does not contain double trace
singularities in (s, t).
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin
I Further constraining the vertices by imposing conservation/on-shellconditions?
I Generalization to other exchange channels with mixed symmetries?Fermionic geodesic Witten diagrams?
I Generalization to superconformal blocks? Compact directions?
I Useful for Mellin bootstrap for external operators with spins?
Heng-Yu Chen National Taiwan University Spinning Witten Diagrams: From Geodesic to Mellin