THE INFRARED SINGULARITIES OF MASSLESS GAUGE THEORIES

Transcription

1 THE INFRARED SINGULARITIES OF MASSLESS GAUGE THEORIES Lorenzo Magnea University of Torino - INFN Torino JGU, Mainz, 07/12/11

2 Outline Infrared divergences to all orders Theory Practice Tools The dipole formula (with E. Gardi) The high-energy limit (with V. Del Duca, C. Duhr, E. Gardi, C.White) Outlook

3 ON INFRARED DIVERGENCES

4 Textbook theory... Singularities arise when propagators go on shell. 2p k = 2p 0 k 0 (1 cos pk ) = 0, k 0 = 0 (IR); cos pk = 1. Emission is not suppressed at long distances. Isolated charged particles are not true asymptotic states of unbroken gauge theories. A serious problem: the S matrix does not exist in the usual Fock space. Possible solutions: construct finite transition probabilities (KLN theorem); construct better asymptotic states (coherent states). Long-distance singularities obey a pattern of exponentiation 1 M = M M = M 0 exp +...

5 Just a formal issue in Quantum Field Theory? Are there practical applications? Higher order QCD calculations at colliders hinge upon cancellation of divergences between virtual corrections and real emission contributions. Cancellation must be performed analytically before numerical integrations. Need local counterterms for matrix elements in all singular regions. State of the art: NLO multileg, NNLO for (some) color-singlet processes. Cancellations leave behind large logarithms: they must be resummed 1 virtual m 2 + (Q 2 dk 2 ) 0 (k 2 ) 1+ real... and Practice = ln(m 2 /Q 2 ) For inclusive observables: analytic resummation to high logarithmic accuracy. For exclusive final states: parton shower event generators, (N)LL accuracy. Resummation probes the all-order structure of perturbation theory. Power-suppressed corrections to QCD cross sections can be studied. Links to the strong coupling regime can be established for SUSY gauge theories.

6 Impact of resummation Z-boson qt spectrum at Tevatron (A. Kulesza et al.) 66 < Q < 116 GeV CDF CDF data on $Z$ production compared with QCD predictions at fixed order (dotted), with resummation (dashed), and with the inclusion of power corrections (solid).

7 Impact of resummation Z-boson qt spectrum at Tevatron (A. Kulesza et al.) 66 < Q < 116 GeV CDF Note shift in the distribution due to non-perturbative corrections extrapolated from all-order resummed result CDF data on $Z$ production compared with QCD predictions at fixed order (dotted), with resummation (dashed), and with the inclusion of power corrections (solid).

8 Impact of resummation Predictions for the Higgs boson qt spectrum at LHC (M. Grazzini) Predictions for the qt spectrum of Higgs bosons produced via gluon fusion at the LHC, with and without resummation, and theoretical uncertainty band of the resummed prediction.

9 Impact of resummation Predictions for the Higgs boson qt spectrum at LHC (M. Grazzini) Note size of nonlogarithmic terms at NLO: the cross section is dominated by infrared effects Predictions for the qt spectrum of Higgs bosons produced via gluon fusion at the LHC, with and without resummation, and theoretical uncertainty band of the resummed prediction.

10 The case of Higgs production The total cross section for gg->h is known to N 3 LL and NNLO, with NLO EW corrections. One of the best-known observables in the SM. A combined analysis (Ahrens et al. 11) gives a 3% (th) + 8% (pdf) + 1% (mq) uncertainty. Ongoing debate on theoretical and pdf uncertainty (Baglio et al. 11). N 3 LL resummed cross section for Higgs production via gluon fusion at LHC The pt distribution for gg->h is known to NNLL and NNLO (M. Grazzini et al. 07, 10; T. Becher et al. 11). Resummation reduces scale uncertainty. A subtle polarization effect uncovered but not implemented yet (Catani, Grazzini, 10). Three-loop coefficient recently computed: theoretically interesting, small impact (Becher, Neubert, 11). NNLL resummed pt distribution for Higgs production via gluon fusion at LHC

11 IR divergences and Jets Theoretically sound jet algorithms guarantee the cancellation of IR singularities to all orders (SISCone, (anti)-k T, Cambridge-Aachen,... ). Is this important? For Theory: unsafe algorithms give meaningless predictions beyond a given order. For Experiment: unsafe algorithms yield, event by event, a jet content that is unstable against the emission of a soft pion or a collinear decay. What does the divergence mean in practice? 2 = c 1 s + c 2 s c 2 =! 2 = c 1 s + K log s +... = 0 (1 + (c 1 + K) s +...). Q Infrared sensitivity at N k LO destroys the predictivity of a N k-1 LO calculation. Impact depends on the specific algorithm and observable. The single-inclusive jet cross section is least affected: δσ/σ < 5% comparing SISCone and MidPoint Cone algorithms. Multi-Jet cross sections are severely affected. W + n-jets existing NLO predictions (n = 2,3) are not applicable to MidPoint Cone algorithms... the processes are then unreliable as discovery channels... Studies of boosted heavy particles are only feasible with IR safe algorithms.

12 TOOLS

13 Dimensional regularization Exponentiation of infrared poles requires solving d-dimensional evolution equations. The running coupling in d = 4-2 ε obeys µ µ (, ) = 2 + ˆ ( ), ˆ ( ) = 2 The one-loop solution is µ 2, µ = s (µ 2 2 0) µ µ 2 The β function develops an IR-free fixed point, so that the coupling vanishes at μ = 0 for fixed ε < 0. The Landau pole is at µ n=0 b n n. 1 b0 4 s (µ 2 0). µ 2 = 2 Q / 4 b 0 s (Q 2. ) ( s, ) Integrations over the scale of the coupling can be analytically performed. All infrared and collinear poles arise by integration over the scale of the running coupling. ˆ ( s ) For negative ε the beta function develops a second zero, O(ε) from the origin.

14 All factorizations separating dynamics at different energy scales lead to resummation of logarithms of the ratio of scales. Renormalization is a textbook example. Renormalization factorizes cutoff dependence. Factorization: UV G (n) 0 (p i,, g 0 ) = n i=1 Z 1/2 i ( /µ, g(µ)) G (n) R (p i, µ, g(µ)) Factorization requires the introduction of an arbitrarily chosen scale μ. Results must be independent of the arbitrary choice of μ. dg (n) 0 dµ d log G(n) R = 0 d log µ The simple functional dependence of the factors is dictated by separation of variables. Proving factorization is the difficult step: it requires all-order diagrammatic analyses. Evolution equations follow automatically. Solving RG evolution resums logarithms of Q 2 /μ 2 into αs(μ 2 ). = n i=1 i (g(µ)).

15 Factorization: Collinear A textbook example is collinear factorization for DIS structure functions. Collinear factorization separates the dependence on the physical scale Q 2 from the dependence on collinear cutoffs (parton masses m 2 ). For Mellin moments one gets F 2 N, Q2 m 2, s = C N, Q2 µ 2 F Factorization requires the introduction of an arbitrarily chosen scale μf. Results must be independent of the arbitrary choice of μf. The simple functional dependence of the factors is dictated by separation of variables. Proving factorization is the difficult step: it requires all-order diagrammatic analyses, or OPE. Evolution equations for parton distributions follow automatically. Solving Altarelli-Parisi evolution resums logarithms of Q 2 /μf 2 into evolved parton distributions (or fragmentation functions)., s f N, µ2 F m 2, df 2 dµ F = 0 d log f d log µ F = N ( s ). s.

16 Factorization: Sudakov Sudakov logarithms are remainders of infrared and collinear divergences. Divergences arise in scattering amplitudes from leading regions in loop momentum space. Soft gluons factorize both form hard (easy) and from collinear (intricate) virtual exchanges. Jet functions J represent color singlet evolution of external hard partons. The soft function S is a matrix mixing the available color representations. In the planar limit soft exchanges are confined to wedges: S is proportional to the identity. Leading integration regions in loop momentum space for Sudakov factorization In the planar limit S can be reabsorbed defining jets as square roots of elementary form factors. Beyond the planar limit S is determined by an anomalous dimension matrix ΓS.

17 Color flow In order to understand the matrix structure of the soft function it is sufficient to consider the simple case of quark-antiquark scattering. At tree level Tree-level diagrams and color flows for quark-antiquark scattering For this process only two color structures are possible. A basis in the space of available color tensors is c (1) abcd = ab cd, c (2) abcd = ac bd The matrix element is a vector in this space, and the Born cross section is M abcd = M 1 c (1) abcd + M 2 c (2) abcd color M 2 = J,L M J M L tr c (J) abcd c(l) abcd Tr HS 0 A virtual soft gluon will reshuffle color and mix the components of this vector QED : M div = S div M Born ; QCD : [M div ] J = [S div ] JL [M Born ] L

18 Sudakov factorization: pictorial A pictorial representation of Sudakov factorization for fixed-angle scattering amplitudes

19 Operator Definitions The precise functional form of this graphical factorization is M L p i /µ, s(µ 2 ), = S LK i j, s(µ 2 ), pi p j H K µ 2, (p i n i ) 2 n 2, s(µ 2 ) i µ2 n (pi n i ) 2 J i n 2, s(µ 2 ( i n i ) 2 ), J i, s(µ 2 ),, i µ2 i=1 Here we introduced dimensionless four-velocities βi μ = Q pi μ, βi 2 = 0, and factorization vectors ni μ, ni 2 0 to define the jets, n 2 i J (p n) 2 n 2 µ 2, s(µ 2 ), u(p) = 0 n(, 0) (0) p. where Φn is the Wilson line operator along the direction n μ, 2 Note: Wilson lines represent fast particles, n( 2, 1) = P exp ig d n A( n). not recoiling against soft radiation 1 The vectors n μ : Ensure gauge invariance of the jets. Separate collinear gluons from wide-angle soft ones. Replace other hard partons with a collinear-safe absorber.

20 Soft Matrices The soft function S is a matrix, mixing the available color tensors. It is defined by a correlator of Wilson lines. (c L ) {ak } S LK ( i j, )=h0 The soft function S obeys a matrix RG evolution equation ny k=1 [ k (1, 0)] b k a k 0i (c K ) {bk }, µ d dµ S LK ( i j, )= S LJ ( i j, ) S JK ( i j, ) NOTE: Γ S is singular due to overlapping UV and collinear poles. S is a pure counterterm. In dimensional regularization, using αs(μ 2 = 0, ε < 0) = 0, S i j, s(µ 2 ), 1 µ 2 d 2 = P exp S 2 2 i j, s( 2, ),. The determination of the soft anomalous dimension matrix Γ S is the keystone of the resummation program for multiparton amplitudes and cross sections. It governs the interplay of color exchange with kinematics in multiparton processes. It is the only source of multiparton correlations for singular contributions. Collinear effects are `color singlet and can be extracted from two-parton scatterings. 0

21 THE DIPOLE FORMULA

22 Surprising Simplicity The matrix ΓS can be computed from the UV poles of S. Computations can be performed directly for the exponent: the relevant diagrams are called webs. ΓS appears highly complex at high orders. g-loop webs directly correlate color and kinematics of up to g+1 Wilson lines: color multipoles are expected The two-loop calculation (Aybat, Dixon, Sterman, 06) leads to a surprising result: for any number of external massless partons (2) S = 2 (1) S = (i) K ( s) = C (i) 2 s + 10 (2) C A 9 T F C F. No new kinematic dependence; no new matrix structure: only color dipoles survive κ is the two-loop coefficient of the cusp anomalous dimension, γk (αs), rescaled by the appropriate quadratic Casimir eigenvalue, s 2 + O 3 s. A web contributing to the soft anomalous dimension matrix Note: the cusp (Korchemsky et al., 86-89) governs IR singularities for two-parton processes.

23 M d d ln µ Z p i µ, s(µ 2 ), p i µ, s(µ 2 ), = Z p i, s(µ 2 f ), µ f H = Z p i µ, s(µ 2 ), The Dipole Formula The two-loop result led to an all-order understanding. For massless partons, the soft matrix obeys a set of exact equations that correlate color exchange with kinematics. The simplest solution to these equations is a sum over color dipoles (Becher, Neubert; Gardi, LM, 09). It leads to an ansatz for the all-order singularity structure of all multiparton fixed-angle massless scattering amplitudes: the dipole formula. All soft and collinear singularities can be collected in a multiplicative operator Z p i µ, µ f µ, s(µ 2 ), Z contains both soft singularities from S, and collinear ones from the jet functions. It must satisfy its own matrix RG equation p i µ, s(µ 2 ) The finite matrix Γ inherits the dipole structure from the soft matrix. It reads,. dip pi µ, s(µ 2 ) = 1 4 K s(µ 2 ) j=i ln 2 pi p j µ 2 T i T j + n i=1 Ji s(µ 2 ). The color generators Ti act as gluon insertion operators on the hard amplitude

24 All known results for IR divergences of massless gauge theory amplitudes are recovered. The absence of multiparton correlations implies remarkable diagrammatic cancellations. The color matrix structure is fixed at one loop: path-ordering is not needed. All divergences are determined by a handful of anomalous dimensions. The cusp anomalous dimension plays a very special role: a universal IR coupling. The planar solution is naturally generalized to the non-planar case. Can this be the definitive answer for IR divergences in massless non-abelian gauge theories? There are precisely two sources of possible corrections. Features of the dipole formula Quadrupole correlations may enter starting at three loops: they must be tightly constrained functions of conformal cross ratios of parton momenta. pi µ, s(µ 2 pi ) = dip µ, s(µ 2 ) + ijkl, s(µ 2 ), ijkl = p i p j p k p l p i p k p j p l The cusp anomalous dimension may violate Casimir scaling beyond three loops. (i) K ( s) = C i K ( s ) + (i) K ( s) The functional form of Δ is further constrained by: collinear limits, Bose symmetry and transcendentality bounds (Becher, Neubert; Dixon, Gardi, LM, 09). Four-loop analysis indicates Casimir scaling holds (Becher, Neubert, Ahrens, Vernazza).

25 THE HIGH-ENERGY LIMIT

26 Reggeization Studies of the high-energy limit predate the modern era of quantum field theory. In the t/s 0 limit particles exchanged in the t-channel (may) `Reggeize. Large logarithms of s/t are generated by a simple replacement of the t-channel propagator, 1 t 1 t (t) s t Gluon-gluon scattering: the t-channel gluon Reggeizes The Regge trajectory has a perturbative expansion, with IR divergent coefficients (t) = s ( t, ) 4 (1) + s( t, ) 4 2 (2) + O s 3 The gluon has been shown to Reggeize at NLL, and the two-loop Regge trajectory is known. For example, for gluon-gluon scattering the matrix element obeys Regge factorization M gg!gg a 1 a 2 a 3 a 4 (s, t) =2g 2 s s t apple (T b ) a1 a 3 C 1 3 (k 1,k 3 ) (t) apple s (T b ) a2 a t 4 C 2 4 (k 2,k 4 ) with the perturbative coefficients K (1) = C A b (1) = C A 2 (2) = C A apple b0 2 + b(2) K 2 + C A n f 27

27 Multi-Regge kinematics Reggeization follows form the dominance of t-channel ladder diagrams as t/s 0. By unitarity, multi-gluon emission must similarly simplify in the high-energy limit. Regge factorization extends to multi-particle emission in `Multi-Regge kinematics. y 3 y 4... y L, k? i ' k? j, 8i, j s s 12 ' k? 3 k? L e y 3 y L e i s ij ' k? i k? j e y i y j e i, 3 apple i<japple L Multi-gluon emission and Multi-Regge kinematics Large logarithms of s/ti are generated by the Reggeization of t-channel propagators, as apple gg! (L 2)g Ma 1...a L = 2gs 3 s (T b ) a1 a 3 C 1 3 (k 1,k 3 ) apple (T a 4 ) bc V 4 (q 1,q 2 ) " The impact factors C and the Lipatov vertices V are universal and independent of s. 1 t 2 " s45 1 t 1 t 2 s34 t 1 (t1 ) # (t2 ) #... apple (T c ) a2 a L C 2 L (k 2,k L )

28 The dipole formula at high energy Introducing Mandelstam color operators, and using color and momentum conservation T s = T 1 + T 2 = (T 3 + T 4 ), it is easy to see that the infrared dipole operator Z factorizes in the high-energy limit pi Z µ, s(µ 2 ), = Z e s t t, s(µ 2 ), Z 1 µ 2, s(µ 2 ), The operator Z1 is s-independent and proportional to the unit matrix in color space. Color dependence and s dependence are collected in the factor ( s ez t, s(µ 2 ), =exp K s (µ ), " #) s 2 ln T 2 t +i T 2 s, t where the coupling dependence is (once again!) completely determined by the cusp anomalous dimension and by the β function, through the function (Korchemsky 94-96) K s (µ 2 ), 1 4 Z µ 2 The simple structure of the high-energy operator governs Reggeization and its breaking. 0 d 2 s + t + u =0 T t = T 1 + T 3 = (T 2 + T 4 ), T u = T 1 + T 4 = (T 2 + T 3 ) T 2 s + T 2 t + T 2 u = 2 b K s ( 2, ) 4X i=1 C i

29 At leading logarithmic accuracy, the (imaginary) s-channel contribution can be dropped, and the dipole operator becomes diagonal in a t-channel basis. M pi µ, s(µ 2 ), =exp Reggeization of leading logarithms K s (µ 2 ), If, at LO and at leading power in t/s, the scattering is dominated by t-channel exchange, then the hard function is an eigenstate of the color operator Tt 2 T 2 t H gg!gg Leading-logarithmic Reggeization for arbitrary t-channel color representations follows ln s t T 2 t Z 1 H t/s!0! C t H gg!gg t pi µ, s(µ 2 ), M gg!gg = CA K( s s (µ 2 ), ) t Z 1 H gg!gg t The LL Regge trajectory is universal and obeys Casimir scaling. Scattering of arbitrary color representations can be analyzed Example: let 1 and 2 be antiquarks, 4 a gluon and 3 a sextet; use 3 6 = a = LL Reggeization of the 3 and 15 t-channel exchanges follows. Scattering for generic color exchange

30 The high-energy infrared operator can be systematically expanded beyond LL, using the Baker-Campbell-Hausdorff formula. At NLL one finds a series of commutators s ez t, s, NLL = ( K( s, ) T s 2 t 1+i K ( s, ) t + K2 ( s, ) 3! ln 2 s t Beyond leading logarithms The real part of the amplitude Reggeizes also at NLL for arbitrary t-channel exchanges. At NNLL Reggeization generically breaks down also for the real part of the amplitude. At two loops, terms that are non-logarithmic and non-diagonal in a t-channel basis arise apple T 2 s K ( s, ) ln 2! ) h T 2 t, T 2 t, T 2 s i +... s T 2 t, T 2 s t E 0 ( s, ) K 2 s, T 2 s 2 At three loops, the first Reggeization-breaking logarithms of s/t arise, generated by s E 1 t, s, 2 s h 3 K3 ( s, )ln T 2 t s, T 2 t, T 2 i s NOTE In the planar limit (NC ) all commutators vanish and Reggeization holds also beyond NLL (as perhaps expected from string theory). Possible quadrupole corrections to the dipole formula cannot come to the rescue.

31 The dipole formula applies for any number of particles: we expect similar simplifications in Multi-Regge kinematics, and similar results concerning Reggeization. Indeed, one can prove recursively that the dipole operator Z factorizes in MR kinematics, as Z The Multi-Regge high-energy operator has again a simple structure. ez MR pl µ, s(µ 2 ), y k, s (µ 2 ), = Z emr y k, s (µ 2 ), =exp ( K Multi-Regge kinematics Z MR 1 " LX 1 s (µ 2 ), k=3 k? i µ, s(µ 2 ), T 2 t k 2 y k +i T 2 s #) Color structure in Multi-Regge kinematics We have defined the t-channel color operators T tk = T 1 + A t-channel basis of common eigenstates of Ttk can be constructed using Clebsch-Gordan coefficients. The operators Ttk thus commute, and each color representation contributing to the hard function in the high-energy limit Reggeizes separately at LL. kx p=1 T p+2

32 Known results on the high-energy limit of QCD amplitudes imply new constraints on quadrupole corrections to the dipole formula at three loops and beyond. Previous analyses using collinear constraint, Bose symmetry and transcendentality bounds could not exclude a class of correction, including for example (212) ( ijkl, s )= where Lijkl = log(! ijkl). A three-loop diagram for Δ s 3 T a 1 T b 2T c 3T d ( s 12)( s 34 ) ( s 13 )( s 24 ) = 1342 ( s 13)( s 24 ) ( s 14 )( s 23 ) = 1423 ( s 14)( s 23 ) ( s 12 )( s 34 ) = Constraining quadrupoles h f ade f cbe L L 1423 L L L s e 2i ; L 1234 = 2(L i ); t 2 t ; L 1342 ' 2L ; s + t 2 s + t e 2i ; L 1423 ' 2i, s i +cycl. Previously admissible corrections display superleading high-energy logarithms at three loops. (212) ( ijkl, s )) = s In the high-energy limit one finds (with L = log s/t ) 3 T a 1 T b 2T c 3T d 4 32 i h L 4 i L 3 2 L 2 i 3 L f ade f cbe +... No known explicit example of admissible quadrupole correction survives. A complete proof is still lacking: linear combinations might restore the proper Regge behavior. i

33 OUTLOOK

34 Summary After years, soft and collinear singularities in gauge theories amplitudes are still a fertile field of study. A definitive solution may be at hand. We are probing the all-order structure of the nonabelian exponent. All-order results constrain, test and complement fixed-order calculations. Understanding singularities has phenomenological applications through resummation. Factorization theorems Evolution equations Exponentiation. Sudakov factorization soft-gluon resummation. Multiparton processes require anomalous dimension matrices. A simple dipole formula may encode all infrared singularites for any massless gauge theory, a natural generalization of the planar limit. The study of possible corrections to the dipole formula is under way. The high-energy limit of the dipole formula provides insights into Reggeization and beyond, at least for divergent contributions to the amplitude. Leading logarithmic Reggeization is proved for generic color representations exchanged in the t channel, and for any number of partons in Multi-Regge kinematics. Regge factorization generically breaks down at NNLL, with computable corrections. The high-energy limit further constrains quadrupole corrections to the dipole formula: no known examples survive.

35 M = M 0 apple 1+apple s CATASTROPHE?

36 dip pi µ, s(µ 2 ) = 1 4 K s(µ 2 ) j=i ln 2 pi p j µ 2 T i T j + n i=1 Ji s(µ 2 ). SOLUTION?

37 CATASTROPHE!

38 SOLUTION?

39 THANK YOU!