DY PRODUCTION AT SMALL QT AND THE COLLINEAR ANOMALY arXiv:1007.4005 + upcoming with Matthias Neubert Thomas Becher University of Bern Seminar at Padua U., Jan. 19, 2011 1 OUTLINE • Introduction • Drell-Yan process • Soft-Collinear • Factorization Effective Theory at low transverse momentum qT • The collinear anomaly and the definition of transverse position dependent PDFs • Resummation of large log’s, relation to the Collins Soper Sterman (CSS) formalism • Expansions from hell: Non-perturbative short-distance physics at low qT. Numerical results 2 Entries/5 GeV 25 ATLAS Preliminary 20 ! L = 229 nb -1 Data 2010 ( s=7 TeV) Z " µµ 15 10 5 0 0 20 40 60 80 100 120 pZ [GeV] T The production of a lepton pair with large invariant mass is the most basic hard-scattering process at a hadron collider. ATLAS has 45 pb-1 of data: ~ 1.5 x107 W’s and 106 Z’s ! This year ~ 1fb-1 should be accumulated. 3 7 1/!Z " d!Z/dpT (1/GeV) due to the smooth10-1 condition imposed DZero rections, statistical tion in each pZ 10-2 T bin atistical uncertainspread in each bin 10-3 ations. As a result, a in any fits as this o an artificially low 10-4 ntral values of the D%, L=1 fb-1 plicative correction -5 10 + ed to compare this Z/ $ * # e e µ factor labeled pT > 15 GeV require10-6 Z/$ *# µ+µ- + corrections for QED FSR; the e measured lepton -7 10 d finally the factor 102 1 10 d mass window to pZT (GeV) 0 electron channel plying only the pµT FIG. 2: Measurements of the normalized differential cross finition as previous section in bins of pZ T for the dielectron [13] and dimuon chanZ nlike pT , the varinels. Both results are shown with combined statistical and • DZero, arXiv:1006.0618, 0712.0803 systematic uncertainties. rements had miniuirement, so a coractors are derived 4 3], as described in “NNLO”), but a NLO differential distribution. The pre- DRELL-YAN PROCESSES The production of a single electroweak boson γ*, Z,W±, H is of great interest for •W mass and width measurements, • PDF determinations, theory benchmarks • Higgs discovery Low transverse momentum qT is particularly relevant • to extract of W mass • to reduce background for Higgs search 5 PERTURBATIVE EXPANSION The perturbative expansion of the qT spectrum contains singular terms of the form (M is the invariant mass of the lepton pair) 2 dσ 1 ! (1) M2 M (1) (2) 2 3 = A α ln + α A + A α ln + ... s s 0 1 3 s 2 2 2 2 dqT qT qT qT " 2 M (n) +A2n−1 αsn ln2n−1 2 + . . . + . . . qT which ruin the perturbative 2expansion† atµ qT ≪ M and must be ψ̄γµ ψ → CV (M , µ) χ̄hc Sn̄ γ Sn χhc resummed to all orders. coll. soft coll. Classic example of an observable which needs resummation! 2 Achieved = (p − p̄)2 by JVµ Collins, Soper and Sterman (CSS) ’84. 6 RESUMMATION A formula which allows for the resummation of the logarithmically enhanced terms at small qT to arbitrary precision was first obtained by Collins, Soper and Sterman (CSS) in ‘84, based on earlier work of Collins and Soper. Their is based on a factorization theorem for the cross section at low qT, which Collins and Soper derived using diagrammatic techniques. We will instead use soft-collinear effective theory to analyze the cross section. 7 SOFT-COLLINEAR EFFECTIVE THEORY Bauer, Pirjol, Stewart et al. 2001, 2002; Beneke et al. 2002; ... In collider processes, we have an interplay of three momentum regions • Hard } high-energy • Collinear • Soft } low-energy part Correspondingly, EFT for such processes has two lowenergy modes: • Collinear fields describing the energetic partons propagating in each direction of large energy, and • soft fields which mediate long range interactions among them. 8 SOFT-COLLINEAR FACTORIZATION Q M1 M2 For M1 ~ M2 ≪Q the cross section factorizes: hard function H(Q2 , µ) J(M12 , µ) J(M22 , µ) jet function S(Λ2s , µ) soft function 9 Λ2s M12 M22 ∼ Q2 ! DIAGRAMMATIC FACTORIZATION The simple structure of soft and collinear emissions forms the basis of the classic factorization proofs, which were obtained by analyzing Feynman diagrams. J.C. Collins, D.E. Soper / Back-to-back lets /. I Collins, Soper, Sterman 80’s ... Advantages of the the SCET approach: • Simpler to exploit gauge invariance on the Lagrangian level • Operator definitions for the soft and collinear contributions • • _J I Fig. 7.2. Dominant integration region for e+e annihilation for small wr. In both fig. 7.1 and Collins and Soper ‘81 the soft gluon subgraphs may be disconnected. Resummation with renormalization group Can include power corrections 10 We begin by considering the slightly simpler process a + b ~ A + B + X, and b are quarks with momenta k~, and k~ respectively. Let k~, be coll defined in subsect. 4.2) in the v~ direction and let k~ be collinear in direction. Then the dominant integration regions are as shown in fig. 7.3. Consider a graph G for this process. A subgraph T of G will be calle SCET APPLICATIONS • Charmless B decays, B→ππ, B→ Xulν, B→ Xs γ • Threshold resummation / soft gluon resummation • DIS at large x, Drell-Yan rapidity spectrum, inclusive Higgs production, top production, direct photon production, single top production, e+e− event shapes, ... • Towards higher-log resummation n-jet processes • IR singularities of multi-loop amplitudes • Jet definitions, beam jets (initial state showering) 11 FACTORIZATION 12 qT0 + A3 =π 2 2 A1 αs"ln 2 + A 2 = 1.78 GeV for M = MZ (28) = q∗ = M exp − dq q q M T T T (n) 2n−1 n# 2C F αs (q∗ ) $ +A2n−1αs ln π 2 + . . . + . . . −1 = 1.78 GeV for M =2MZ (28) $x % = q∗ = M exp − dσ † 2 q µ µ = max (q∗ , TqT ) 2CF αs (q∗ )Tψ̄γµ ψ M CnV (M , µ) χ̄hc S γ= (n)→ 2n−1 ! 2 2 n̄ 2 dσ 1 M αs (1) dq + . .T. ... +A ln (1) (2) 2 3 M 2n−1 p 2 T1 αs ln = A + A + A α ln + 0 3 sq 2 2 2 (29) λq 2= µ =•max (q∗ , qT ) point is the factorization dq q q Starting of the electroweak current in T T T T T † µ CV , anti-coll. soft coll. 2 M ψ̄γµlimit ψ → CV (M , µ) n2 χhc " the Sudakov pT χ̄hc Sn̄ γ S λ =+A(n) αn ln2n−1 M + . . . + . . . (29) M µ2n−1 s 2 † p 2 hard-collinear q 2 2 µ i-coll.p soft coll. T n ∼ ψ̄γ ψ →hcCV (M , µ) χ̄hc Sn̄ γ p̄ M = (p − p̄) p0 µ pµψ → C (M 2 , µ) χ̄ S † γ2µ S χ ! 2 nµ ψ̄γ µ V n hc n̄ hc M dσ CV , 1anti-coll. Mcoll. ∼(1) µ soft (2) 2 (1) 0 3 2 µ p CV , anti-coll. soft coll = (p=− p̄)C A J, Vanti-coll. µA p̄ + soft α ln + A α ln + . . . s 1 0 3 s n̄ ∼ soft2coll. p̄0 dqT2 qT2 V qT qT2 22 hc p̄µ 2 µ p p̄ M p2 p̄=M (p − p̄) = (p − p̄) n̄ ∼ 20 " anti-hard-collinear (n) 2n−1 M p̄ n + ... + ... +A2n−1 αs ln 2 qT (SCET) this can be written in • In Soft-Collinear Effective Theory L⊥ $x−1 T % SOFT-COLLINEAR FACTORIZATION operator form as t coll. ψ̄γµ ψ → CV (M 2 † , µ) χ̄hc Sn̄ 4 13 γ µ Sn χhc SCET hc quark field † µ dq 2 (3) q ψ̄γµ ψ → CV (M 2 , µ) χ̄hc Sn̄† γ µ 2Sn χhc , anti-coll. 2 soft coll.2 µ ψ̄γµ ψ → CV (M , µ) χ̄hc Sn̄ γ Sn χThc p p̄ M = soft (p −coll. p̄) JV CV , anti-coll. C , anti-coll. soft coll. V 2 = (p − p̄)2 J µ p̄The M Drell-Yan cross 2 2 V µ section is obtained from the matrix p̄ M = (p − p̄) JV element of two currents µ p p̄ M = (p µ†− p̄)ν JV 1 (−gµν ) #N1(p) N2(p̄)| JV (x) JV (0) |N1(p) N2(p̄)$ → |CV (M 2 , µ)|2 2Nc n̄ /n , anti-coll. /̄n n n̄ C × ŴDY (x) #N1 (p)| χ̄hc(x) χhc (0) |N1(p)$ #N2 (p̄)| χ̄hc (0) Vχhc (x) |N2(p̄)$ 2 2 2 2 soft 1 µ† ν 2 2 2 1 (x) J (0) |N (p) N (p̄)$ → ) #N (p) N (p̄)| J |C (M , µ)| 1 2 1 2 V 2 2 Vν V µ† gν µν ) #N 1 (p) N2(p̄)| JV (x) JV (0) |N1(p) N2 (p̄)$ →2Nc |CV (M , µ)| 2Nc n and n̄ are light-cone reference vectors along p and p̄ .M = (p − /n /̄ n /̄nχhc (0)a |N (0) /nχhc (x) |N × ŴThe #Nfunction χ̄hcWilson 1 (p)$ #Nof 2(p̄)$ DY (x) 1 (p)| χ̄hc (x) 2 (p̄)| soft contains product four along (x) |N × ŴDY (x) #N1 (p)| χ̄hc(x)2 χhc (0) |N1(p)$ #N2 (p̄)| χ̄hc (0) 2 χlines 2(p̄)$ hc 2 1 the directions of large2energy flow µ† ν # #(x) $$2 (p̄)$ → J (0) |N (p) N (−gµν ) #N1(p) N21(p̄)| J |CV 1 † V † V 1 ŴŴ == #0| TrTrSS (x) SSn̄(x) SSn̄†(0) SSn(0) |0$ (4) n n̄ † 2N DY (x) n c (x) #0| (x) (x) (0) (0) |0$ (4) DY n n̄ n̄ n NN c c /̄n % ' & 0 × ŴDY (x) #N1 (p)| χ̄hc(x) χhc (0) |N1(p)$ #N2 (p̄)| χ̄hc (0) Sn (x) = P exp i ds 2 n · As (x + sn) (5) (−gµν ) #N1(p) N2(p̄) −∞ 14 × Ŵ (x) #N (p hc2 DY 1 hc n / /̄ n π µ 2 n n̄ χ (0) |N (p)$ (0) χ (x) |N (p̄)$ (x) #N (p)| χ̄ (x) #N (p̄)| χ̄ hc 1 2 DY 1 hc 2 hc 2 2 hc # † 1 qT1 # † $ŴDY (x) = #0| Tr S (x † n µ† ν ŴDY (x) = S)n̄(x) n(0)J |0$(x) J (0) |N n̄ (0) (p) N2(p̄)$ → #N1S(p) N2S(p̄)| n (x)µν N1c(4) V V LN⊥c #0| Tr S(−g % & 0 % ' & n /̄ $ # 0 Final step is to expand the matrix elements in small momentum χ (0) |N1(p)$ #N2 (p × Ŵ (x) #N (p)| χ̄ (x) hc DY 1 hc π −1 S (x) = P exp i ds n · A (x + sn) (5) S−n 2(x) = P exp =i 1.75 n s q = M exp components, i.e. to perform expansion. $x %a=derivative DERIVATIVE EXPANSION −∞ T n̄ · k, k⊥) The ∗ 2CF #αs (q∗ ) −∞ 1 † † Ŵ (x) = #0| Tr S (x) S (x) S n̄ · k, k⊥) Nscale as n n̄ ( light-cone components (n · k, n̄DY c µ =2 max (q∗ , qT ) % & (6) expansion parameter phc ∼ M (λ , 1, λ) , phc ∼ M (1, λ2, λ) . 0 Sn (x) = P exp i qTds n · As (x ps ∼ M (λ2, λ2 , λ2) . (n · k, n̄ · k, k⊥) λ =−∞ M (7) while the separation between the two currents scales as x ∼ M −1 (1, 1, λ−1) µ p nµ ∼ 0 p Aµs (x) = Aµs (0) + x · ∂Aµsp(0) +M . . . (λ2 , 1, λ) , hc ∼ 1 φq/N (z, µ) = 2π ! power suppressed, can be dropped phc ∼ M ( µ p̄ µ /̄n 2 2 2 n̄ ∼ −iztn̄·p p ∼ M (λ , λ , λ ). 0 s dt e !N(p)| χ̄(tn̄) χ(0) |N(p)" 15 p̄ 2 1 (n · k, n̄ M · k,−1k(1, ⊥ )1, λ−1 ) x∼ x∼M −1 FACTORIZATION (1, 1,NAIVE λ ) −1 Dropping phc ∼ M (λ2 , 1, λ) , 2ps phc ∼ M (1, λ2, λ) , ∼ M (λ2 , λ2 , λ2 ) . 2 p ∼ M (λ , 1, λ) , p ∼ M (1, λ , λ)to , hc hc power suppressed x-dependence leads ŴDY (0) #N1 (p)| χ̄hc (x+ + x⊥ ) the result (6) (7) /̄ p ∼ M (λ2 , λ2 , λ2 ) . n /n χhcs (0) |N1 (p)$ #N2(p̄)| χ̄hc (0) χhc (x− + x⊥ ) |N2(p̄)$ 2 2 } 1 ŴDY (0) 1 x “transverse/̄n PDF” x =“transverse /nPDF” (8) χhc (0) |N1 (p)$ #N2(p̄)| χ̄hc (0) χhc (x− + x⊥ ) |N2(p̄)$ this 2/̄n spells trouble: well known /nthat2 transverse KLN Ŵ cancellation! ) χnot #N2(p̄)| χ̄hc |N1(p)$ (0) χhc (x DY (x0 ) #N1 (p)| χ̄hc (x+ hc (0) − ) |N2 (p̄)$ PDF well defined w/o additional regulators 2 2 ŴDY (0) = 1 ŴDY (0) #N1 (p)| χ̄hc (x+ + x⊥ ) For comparison: for soft-gluon resummation, the result is 1 /̄ n /n ŴDY (x0 ) #N1 (p)| χ̄hc(x+ ) χhc (0) |N1(p)$ #N2(p̄)| χ̄hc (0) χhc (x− ) |N2 (p̄)$ 2 2 “soft” x “standard PDF” 16 1 x “standard PDF” c T $ Aµ (x) = Aµ (0) + x · ∂Aµ (0) + . . . % & 2 ' s s s qT 2 2 2 , × eq Bq/N1 (ξ1 , xT , µ) Bq̄/N2 (ξ2 , xT , µ) + (q ↔ q̄) + O 2 M q !µ Aµs (0) + x · ∂Aµs (0) /̄ n+ . . . 1 As (x) =−iztn̄·p dt e !N(p)| χ̄(tn̄) χ(0) |N(p)" φq/N (z, µ) = In terms of the hadronic matrix ! elements 2& 2 2 '−Fqq̄ (x2T ,µ) 2π ! 3 2 " "2 1 dσ 4πα " xT M n 21 2 −iq⊥ ·x⊥ /̄ ! " −iztn̄·p (−M d !N(p)| x⊥ e χ̄(tn̄) χ(0) = e φ22q/N C (z,V µ) = , µ) dt 1 /̄n 2 dq 2 dy −2γ|N(p)" E −iztn̄·p 3N M s 4π 4e 2πdt e Bq/N (z,c xT , µ) = !N(p)| χ̄(tn̄ + x2⊥ ) χ(0) |N(p)" T 2π 2 % ! & 2 ' # $2 1 n /̄ qT 2 dt e−iztn̄·p !N(p)| 2 χ̄(tn̄ + x⊥ ) 2 x , µ) = χ(0) |N(p)" T , B (ξ , x , µ) B (ξ , x , µ) + (q ↔ q̄) + O × Bq/N e(z, q/N1 12π T q̄/N2 2 T q 2 2 on then obtains the DY cross section M q # CROSS SECTION ! " d3 σ 1 4πα2 "" 2 2 −iq⊥ ·x⊥ "2 C (−M , µ) = d x e V ⊥ 2 dM 2 dqT√ dy y 3Nc M 2 s √ −y 4π M 2 + qT2 m2⊥ = .& ' ξ1 = τ e , ξ2 $= τ e , with τ = % # s s qT2 2 2 2 , × eq Bq/N1 (ξ1 , xT , µ) Bq̄/N2 (ξ2 , xT , µ) + (q ↔ q̄) + O 2 M 2 2 2 √ √ q ξ1 = with y τe , ξ2 = −y τe , M + qT m⊥ = . with τ = s s & 2 2 '−Fqq̄ (x2T ,µ) ! 2 " " dσ √ 4πα " √ 2 "2 1 xTM M2 + qT2 2 −iq⊥m ·x⊥ y −y ⊥ CV (−M , µ) d x⊥τe = = ξ = =4e−2γE . τ e , ξ = τ e , with 2 2 2 1 2 dM dqT dy 3Nc M s 4π s s % & 2 ' $ # qT 2 2 2 , × eq Bq/N1 (ξ1 , xT , µ) Bq̄/N2 (ξ2 , xT , µ) + (q ↔ q̄) + O 2 M q 3 (1 (1( (1 (1 (1 2 17 (1 (1 M 2 dqT2 dy = CV (−M 1, µ) d x⊥ e /̄n −2γ 2s −iztn̄·p 3Nc M 4π dt e !N(p)| χ̄(tn̄) 4eχ(0)E |N(p)" φq/N (z, µ) = 2π 2 % & 2 ' $ # ! qT 2 2 2 1 n /̄ , (ξ2 , xT , µ) + (q ↔ q̄) + O × e B 1 (ξ1 , xT , µ) B−iztn̄·p 2 Bq/N (z,q x2T ,q/N µ) = dt e q̄/N2 !N(p)| χ̄(tn̄ + x⊥ ) χ(0) |N(p)" M q (11) ( CROSS SECTION2 2π ! " 2 2 2 d3 σ √ 1 4πα2 "" √ 2 µ "2 2 −iq ·x⊥ M + q m ⊥⊥ µ µ T y −y CV (−M , ,µ) = , e+ x · ∂A (x)4πwith = Ads x(0) = s (0) + . . . (12) ξs2 = τe A τ⊥ = 2 s dM ξ2 1dq=T2 dyτ e 3N M c s s % & 2 ' (1 # $ ! qT 2 2 2 2 m2 √ y× √ 1 (ξ−y M + q n /̄ 1 , , x , µ) B (ξ , x , µ) + (q ↔ q̄) + O e2q Bq/N 1 q̄/N2 −iztn̄·p 2 T ⊥ T T 2 = . ξ1 = τ e , q φq/N ξ2 = τ e , with τ = dt e !N(p)| χ̄(tn̄) χ(0)M|N(p)"(13) (z, µ) = s 2π s 2 ! √ y √ The ξresummation would then M2 + ' qT2 the /̄ m2⊥by& solving nRG −y 1 be!obtained 2 −iztn̄·p = xχ̄(tn̄ (14) τe B , 4πα =T , µ) τ e= , " 1with dt e τ = !N(p)| x. ⊥(x) ,µ)χ(0) |N(p 13 = 2ξ2 "x 2 +−F 2 q/N (z, d σ M 2 equation = T s 2π" 2 "CV (−M 2 , µ) d2 x⊥ e−iq s·x q q̄ dM 2 dqT2 ⊥ M 2s ⊥ 4e−2γE % 2 d −M (1 & 2 ' $ 2 F q ! 2 % # C (M , µ) = Γ (α ) ln + 2γ (α ) C (M , µ) . (15) 3 V 2 V qT " 2 2 s 2 4παcusp" s 2 1 2 d ln µd σ × µ 2 2 −iq ·x , eq Bq/N1 (ξ"1 , xT , µ) Bq̄/N2 (ξ ⊥ O " 2 , xT , µ) + (q ↔ q̄)⊥ + dy 3Nc 4π $ CV (−M , µ) = 2 2 q dqSCET 3NcGao, M sLi, Liu 2005; Idilbi, Ji, Yuan 4π T dy papers: see dM 2 where 2 T d x⊥ e M2 2005; Mantry, Petriello 2009 $ % 2 √ 2 2 2 2 2 √ correct! M + q m B (ξ , x , µ) B (ξ , x , µ) + (q × e be If is q/N 1 q̄/N 2 T ↔ q̄) + O y −y ⊥ T 1 2 T q This cannot = . ξ1 = τ e , ξ2 = τ e , with τ = q s s independent of M, the above cross section is μ dependent! # 3 dσ 4πα 2 18 " " 2 "2 1 " ! 2 −iq⊥ ·x⊥ & x2T M 2 (1 '−Fqq̄ T 2 3 c Bdq/N µ) 2= "" σ (z, x$T ,4πα −iztn̄·p "2 1 !N dt e χ̄(tn̄ + x⊥& ) χ(0) |N(10) (p)" 2 2 (p)| −iq ⊥ ·x⊥ % ' " CV (−M , µ) = d x⊥ e 2q 2 2 2 s 2π dM 2 dq# dy 3N M 4π T , (ξ , x2 , µ) + (q ↔ q̄) + O × T e2 B c (ξ , x2 , µ) B COLLINEAR ANOMALY q/N1 q̄/N2 2 T T %M 2 & 2 ' $ # ! (ξ , x2 , µ) + (q ↔ q̄) + O qT , 2 "e2 Bq/N (ξ1 , x2 ,"µ) Bq̄/N × 2 1 4πα " q 2 T "2 1 2 2 T −iq⊥ ·x⊥ M2 C (−M , µ) dx e q q q d3 σ = 2 2 M dqT dy 3Nc M 2 s 1 V 4π (10) ⊥ & 2 2 '−Fqq̄ (x2T ,µ) ! " " σ 4πα " xT M 2 1 $ %2 of & 2 ' 2 2 −iq⊥ ·x⊥ " RG invariance of cross section implies that the product C (−M , µ) d x e = & ' ! # V ⊥ −2γE 3 2 2 −Fqq̄ (xT ,µ) qT " dqT2 dy d3N 4π 4e σ c M 2 s 4πα 1 x M 22 " 2 2 2 T 2 2 −iq ·x ⊥ ⊥ " " , + (q&↔contain q̄) +O × =PDFs eq 2Bq/N transverse must CV (−M d x2 ⊥(ξe2 , xT , µ) 1 (ξ1 , ,xµ) T , µ) Bq̄/N (11) 2 2 2 −2γ ' % $ E M dM dq# 4π 4e 2 T dy 2 q3Nc M s q T 2 2 hidden (11) Bq/N × Me2q dependence. #1 (ξ1$, xT , µ) Bq̄/N2 (ξ2 , xT , µ) + (q ↔ q̄) + O %M 2 &, q 2 ' T 2 2 2 q 3 2 × eq Bq/N1 (ξ1 , xT , µ) Bq̄/N2 (ξ2 , xT , µ) + (q ↔ q̄) + O M2 , q '−F (x ,µ) Analyzing the relevant diagrams, !one finds that&an2 additional 3 2 " 2 " dσ 4πα " √ 2 1 2·x 2 xT M 2 2m2 −iq " √ M + q regulator is needed to make transverse PDFs well In where2 C (−M , µ) d x = T 2 defined. y −y ⊥e V ⊥ 2 2 2 2 −2γ = m⊥ M4e . + qT (12) ξ1 dy = τ e3N , cM τ= √ξ2s y= τ e , √ with M dq 4π −y T s s the product regulator can but = be removed, . (12) , two ξ2 =PDFs, τ e ,this with τ= ξ1 = ofτ ethe s s ' % & $ 2 2 # anomalous M 2dependence2 remains. Can2 refactorize qT q q̄ ⊥ 2 T ⊥ E µ)qq̄ (x B2Tq̄/N eq Bq/N &1 (ξ21 ,2xT',−F ,µ) 2 (ξ2 , xT , µ) + (q ↔ q̄) + O × 2 & 2 2 '−Fqq̄ (x2T ,µ) 2 x q M T 2 ) xT M Bq/N1 (z1 , xT , µ) Bq̄/N 2 2 (z2 , xT , µ) , 2 −2γ E Bq/N1 (z1 , xT , µ) Bq̄/N2 (z2 , xT , µ)4eM 2 = B (z , x q/N 1 1 T , µ) Bq̄/N2 (z2 , xT , µ) , 4e−2γE ) 2 2 ( q z1 , xT , µ) Bq̄/N22(z2 , xT , µ) q2 =2 ξ1 = 2 2 2 2 2 dF (x , µ) √ q q̄ M + q m T F dF (x , µ) q= q̄ 2Γ T ⊥ T−y (αs ) F (13) with = 2Γ (α ) cusp = . (13) τ ey , ξ = τ e , with τ = s 2 cusp d ln µ d ln µ s s √ Note that√M2y dependence $ % $√ exponentiates! 2 −y 2 2 2 % m2⊥ M + q T d ξ1 = dτ2 e , −M =2 , µ) . . (14) e , τ(M = 2 , µ) 2F ξ2 = Fτ −M , µ)(α =s )Γln (M CV (M , µ)CV=(MΓcusp + 2γ q2with (α+ CqV(α . (14) s) C V s ) 2γ cusp (α2s ) ln s s 19 µ d ln µ d ln µ µ , Regular soft-collinear factorization: H(Q2 , µ) J(M12 , µ) J(M22 , µ) S(Λ2s , µ) Anomalous factorization H(Q2 , µ) Bq̄/N2 (x⊥ , µ) Bq/N1 (x⊥ , µ) F (x⊥ , µ) ln(x2⊥ Q2 ) 20 TRANSVERSE PDFS What God has joined together, let no man separate... The “operator definition of TMD PDFs is quite problematic [...] and is nowadays under active investigation”. quote from Cherednikov and Stefanis ’09. For reviews, see Collins ’03, ’08 Regularization of the individual transverse PDFs is delicate, but the product is well defined, and has specific dependence on the hard momentum transfer M2. Anomaly: Classically, !N1 (p)| χ̄hc(x+ + x⊥ )/̄n χhc(0) |N1 (p)" is invariant under a rescaling of the momentum of the other nucleon N2. Quantum theory needsAµsregularization. be (x) = Aµs (0) + xSymmetry · ∂Aµs (0) + .cannot .. recovered after removing regulator. Not an /̄n 1 −iztn̄·p dt e !N (p)| χ̄(tn̄) χ(0) |N (p)" φ (z, µ) = q/N of QCD, but of the low energy theory. anomaly 2π 2 ! 1 /̄n 2 −iztn̄·p Bq/N (z, xT , µ) = dt e21 !N (p)| χ̄(tn̄ + x⊥ ) χ(0) |N (p)" 2π 2 ! RESUMMATION 22 qT ≤QT 1 q 2 i=q,g j=q̄,g q←q s z1 z2 ≥M 2 /s q←g s 3ow perform Calculation of the kernels I a perturbative calculation of the relevantIq←q kernels and Ii←j entering the facto q←g & ' min(Q , z z s−M ) $ −Fqq̄ (xT ,µ) 2 2 # 2 ( ) x q % M 2 2 & operat T 2 at first non-trivial 2 2 n formula (22) order in α . Since we do not have explicit Bq/N1 (z1 , xT , µ) Bq̄/N2 (z2 , xT , µ) q2 = 2 −2γ s B2q/N1 (z , x , µ) B (z q̄/NQCD 2 , xT , µ) , 21 T 2 E T × atransverse dqcalculation Cqq̄→ij (z1 ,ofz2the , qT ,relevant µ) ffkernels µtheenteri + (q, i We now perform perturbative I,i←j ij instead tions of the (good) distribution functions B ,M we ,analyze origin T 4e 2 T 2 SIMPLIFICATION FOR q ' Λ 1 2 2 i/N . z1 z2 s zation formula (22) at order in αs . products Since weofdo exp functions Bi/N defined in 0first (11), non-trivial keeping twonot suchhave function dFqq̄ (x2T , in µ) mindF that only = If 2Γwe (α (13) s) ingFor to different are wellofddefined. write an operator-product expansio cuspperform definitions of thehadrons (good)values transverse functions , we analyze instea perturbative qTdistribution anBi/N operator ln µ we can gous to (19) expansion bad) functions Bi/N defined in (11), keeping in mind that only products of two s product Calculation of the kernels I and I q←q% q←g $are well defined. # eferring to different hadrons If we write an operator-prod 2 d −M "2 1 dz F q 2 C=V (M , µ) = Γ + 2γ (αsµ) ) C , 2µ) . x2 ) , (14) (2 2 2 s )xln V (M cusp (α 2 analogous to (19) B (ξ, x , µ) I (z, , µ) φ (ξ/z, + O(Λ i←j i/N j/N d ln µ µ e now performTa perturbativez calculation of the relevant kernels T QCD I Ti←j entering ξ j tion formula (22) at first non-trivial in αs . Since we do not have(15) expli " # 1qT2dz'order ΛQCD 2 transverse distribution functions 2 2instead 2 finitions of the (good) B , we analyze i/N B (ξ, x , µ) = I (z, x , µ) φ (ξ/z, µ) + O(Λ x owsAgain, that the products of two I functions are well defined and obey a factorizatio only the product of two functions is well i←j i/N j/N i←j T QCD T ) Text p, (p̄) to the power α → 0, β → 0 z p → λp T ξ keeping in mind that only products of two su ad)defined. functionstoBi/N ula analogous (13).defined in (11), j erring to different hadrons are well we write an operator-produc '−F (x If & 2defined. ,µ) 2 ( ) two IxT q functions are well 2 the products 2 2 2 talogous follows that of defined and obey a i←j to (19) = I (z , x , µ) I (z , x , µ) I (z , x , µ) I (z , x , µ) , q←i 1 q̄←j 2 q←i 1 q̄←j 2 T T T T q One-loop results 4e−2γ ormula analogous to (13). # 1 " rbative expansions for the kernels Ii←j from a matching calculation, dz can be derived 2 2 2 2 B (ξ, x , µ) = I (z, x , µ) φ (ξ/z, µ) + O(Λ x Effective theory diagrams for are not well-defined i←j i/N j/N T using external parton states QCDcarryin T), h the matrix elementsTin (10) and (11) zare evaluated 2 ξ 3.1 One-loop results j q q̄ 2 2 T E 2 in dim.ofreg.. Following Smirnov ’83,tree-level we use additional analytical d fraction the nucleon momentum p. The result is obviously given by regularization, whichforisofthe very economical, since it does notand Perturbative expansions I can be derived from a matching i←j 2twokernels follows that the products I functions are well defined obey a(2fc Ii←j (z, xT , µ) =i←j δ(1 − z) δij + O(αs ) . introduce additional scales into the problem. which matrix elements mula the analogous to (13). in (10) and (11) are evaluated using external parton s aelevant fixed fraction the nucleon p.s )The tree-level result is obviously g one-loop of diagrams giving momentum rise to the23O(α corrections to the kernels Iq←q = Iq̄← hown in the first row of Figure 1. There is no need to consider diagrams with external-le 2 ,µ) ' & −F (x 2 q q̄ 2 2 T gure 1: One-loop diagrams contributing to the, µ) matching coefficients Iq←q (top row) and ) dF (x x q q q̄ T F T 2 2 = 2Γ (α ) (bottom row). The vertical lines indicate cut propagators. s = B (z , x , µ) B (z , x , µ) B (z ←gq̄/N2 cusp 2 1 2, q/N q̄/N T 1 T 2 q2 −2γ 4e E d ln µ ANALYTICAL REGULARIZATION fers to hc the running coupling evaluated at the scale µ, unless indicated otherwise. Th hc hc hc hc hc d 2 d = 4$− 2!, and we omit O(!) terms. hc Moreover, % mensional regulator isdF defined by µ denote (x , µ) 2 q q̄ T C F d scaleαdefined −M α e renormalization in the MS scheme. The remaining two diagrams turn ou 2 F q 2 = 2Γ (α ) µ , s + β 2γ (α ) Cd ln cusp C (M , µ) = Γ (α ) ln + (M → β V s s V α d lnregularization µ β cusp be ill-defined in µ dimensional due to light-cone 2singularities. To give meanin d ln µ hc the corresponding loop integrals requires introducinghcadditional regulators. The simples hc hc hc hc hc ossibility is to employ analytic regularization, as is common in the context of asymptoti 2 $ %used q Λhas Figure 2: Matching of an analytically-regularized QCD graph onto SCET pansions [19, 20]. In the context of SCET this method been in diagrams. [15, 44]. One start QCD T2 ' Text p, (p̄) to to the p • −M to Raise 2QCD propagators carrying large F q reconsidering the QCD diagrams contributing the momentum process and raises 2all propagator Cfractional ,the µ)external = Γhard-collinear (α lnlimit 2γ (αtoαs→ )a fractional (M , µ) . Vof(M sin)the cusp diagrams develop singularities β2→ 0 + followed by 0C or Vvice versa, which trough p, (p̄) to power α, (β) powers : which the momentum p flows power, µ µ x x x x x x x x cancel in the sum of the results from both sectors. 2α diagrams in Figure 1 can now be With the analytic regulators 1 in place, the remainingνtwo 1 → computed and both give the same result. For their sum, we obtain 2 1+α , 2 −(p − ! k)2 " −−αiε! QCD "[−(p − k) − iε] ! 2 2 "!+α T 2 2 −β µ q 2z xT µ C α Γ(−% − α) F s b+c e!γE Iq←q (z, x2T , µ) = . (32) 2π ν12 ν22 (1 − z)1−α+β Γ(1 + α) 4 similarly for the anti-hard-collinear propagators, but with a different regulator q 'Λ (30 d β and •power Limit is trivial for QCD, but effective theory he α → 0, β → 0 sociated scale ν2QCD, . Forthe QCD diagrams, such first graph in taking Figurethe2,limit the% modification Like in full analytic regulators mustasbethe taken to zero before → 0. have which the ofascollinear The result depends thelimits order in the limits → 0 in and β → sum 0asarelong performed. trivialdiagrams in the sense thatonpoles, the α which → 0only and βcancel →α02 are smooth ! is Exkept finite panding first analytic in β and then in α, the light-cone are regulated by the in α parameter, and anti-collinear diagrams. owever, with the regulators in placesingularities also the SCET diagrams the (anti-)hard and we find for the sum of all four one-loop diagrams llinear sectors are now# also well-defined. The" %! diagrams in2 "each sector involve $! & divergence 2 1 1+z 2 µ CF αs # the analytical regulator, of2 all If the momentum L⊥ sum − ln 2contributions. δ(1 − z) + Iq←q (z, x2T , µ)# which = − cancel in+ the 2π % α ν1 (1 − z)+ α reg. 24 in (30) is hard-collinear, as in the first SCET!diagram in Figure 2, the " ' regularization in T F 2Γcusp (α ntributions to the expansion of the integral. One finds that only the hard, the= hard-collinear 2 ,µ) & are2evaluated d2ln'µ−F q q̄ (xTthen nd anti-hard-collinear regions contribute. If the diagrams off-shell, also a ) x q T verifies that the contribution 2 arises in the individual2 diagrams, but one easily ft contribution (z1 , xTmode , µ) scaling Bq̄/N2as(zM(λ, , µ) q2 = Bq/N1 2, x Na1semi-hard −2γ λ,Tλ) vanishes. E 4e $ regulators and is obtained Since the original diagram is well defined without the analytical 2 d −M y addinghcup the contributions hc limits hc fromhcthe different2regions, hc Fguaranteed that the hc we are C (M , µ) = Γ independent (α + hc s ) ln → 0 and β → 0 can be taken in the sumV and that the 2result iscusp of the regula2 d ln µ µ dF (x , µ) q q̄ α T F r. Individually, however, in each sector involve divergences in the analytical β α the diagrams + → = 2Γ (α ) β α s cusp β gulators. If the momentum k in (31) is hard-collinear, as in the first SCET diagram in d ln µ hc gure 2, hc the regularization inhcthe effective theory takes the in QCD. If, on the hc same hc hc form hc qT2as ' ΛQCD her hand, the momentum k is anti-hard-collinear, then the propagator is far off-shell and in Figure 2: Matching of anline, analytically-regularized graph onto SCET diagrams. CET is represented by a Wilson as shown in the QCD second diagram in Figure 2. Using the • Text p,and (p̄)performing to thethe power → 0, β → 0find that Regulators play double role. E.g. hc $ α regulates placement rule (31) appropriate expansions, we propagators 2 the Feynman d from diagrams develop singularities followed line by α W → 0 −M or vice versa, whichq hcemission Wilson line 2in the limit β → 0FWilson le forand aofgluon the anti-hard-collinear hc in the current operator (M from , µ) Γcusp (αs ) ln 2 + 2γ (αs ) cancel in the sum ofC theVresults both= sectors. ) gets replaced by With d theln analytic two diagrams in Figure µ regulators µ 1 can now be 2α µ µ in place, the remaining ν n n̄ · p n 1 computed and both give the same result. sum, we obtain → For their . (32) 1+α ! 2 2 "!+α n · k !− 2iε"−α !(n2 "· k −βn̄ · p − iε) µ q 2z xT µ C α Γ(−% − α) F s b+c 2 e!γE Iq←q (z, x2T , µ) = . (32) ote that, as mentioned earlier 2.2, regularized rule 2π in Section ν12 ν22 the (1 − z)1−α+β Γ(1 α) QCD 4 for the anti-hardqT Feynman '+ Λ llinear Wilson is the no analytic longer invariant under the rescaling transformation λp. As • Regulator Like in fullline QCD, regulators must taken to zero before takingsector the limitp %→ → 0. breaks invariance of be anti-hard-collinear under The result depends on the α order in which the limits αit→regularizes 0 and β → the 0 arefermion performed. Exin aFigure 2, the regulator plays a→ double role: 2. propagators xtenhard-collinear p, (p̄) to the power α 0, β → 0 p → λp rescaling of the hard-collinear momentum panding first in β andand then the in α,Wilson the light-cone singularities are regulated diagrams. by the α parameter, diagrams lines in anti-hard-collinear Both classes and we find for the sum of all four one-loop diagrams diagrams develop singularities in the$! limit β → 0 followed by versa, which " %! " α → 0 or vice & # 2 2 1sectors. 2 1+z µ CF αboth # s 2 ncel in the sum of xthe results from + L − 2 ln δ(1 − z) + Iq←q (z, , µ) = − # ⊥ T 2 2π % remaining α two νdiagrams (1 − z)+1 can now be α reg. 25 1 With the analytic regulators in place, the in Figure ANALYTICAL REGULARIZATION x x x x x x x x ! " ' 2 F C (M q δ(1 2 I (z, x , µ) = − z) − + , µ) = Γ (α ) ln + 2γ (α ) q←q V s s T (M , µ) = Γ (α ) ln + 2γ (α ) C (M , µ) . cusp V s s V 2 2π 2 $ ! d lncusp µ µ µ " " # # 1-LOOP RESULT 2 2 qT ' ΛQCD q ' ΛQCD × Cqi z1 , αs (µb ) Cq̄j z2 , αs (µb ) φi/N1 (ξ1 /z1 , µb ) φj/N2 (ξ2 /z2 , µ −γE T µb = b0 /xT , and we adopt the standard choice b0 = 2e + δ( x2T µ2 L⊥ = ln −2γ 4e E !" # $" # % 2 2 !" # $" # % C α 2 1 + z 1 µ F s Iq←q2(z, x2T , µ) = δ(1 − z) + z2 + L⊥ 2 −µ22 ln 2 δ(1 − z) CF− αs 1+ 1 2π + L$⊥ ν1 − z) + (1 − z)+ Iq←q (z, xT , µ) = δ(1 − z) − − 2αln 2 δ(1 2 ,µ) 2π2 2 $ −Fqq̄ (xα2 " ν (1 − z) + 1 −F (x # & T ,µ) q q̄ 2 2 2 T & " 2 22 # π T 2 2 2 2 + δ(1 − z)2 − 22 + Lπ⊥ +T − (1 − z) (16) 2 ←j q←i 2 1−6(1 T q̄←j + − z) (16) 2 q←i + δ(1 T1 q̄←j −2γ 2E T q2 T − z) −q 2$2 +2L$ ⊥q←i F T s 6 −2γE T then β acts as the analytic If the expansions are performed in the q̄←q̄ opposite order, regulator, 2 dep. If the expansions are performed in the opposite order, then β acts as the analyticMregulator, anomalous and we obtain and we obtain !" # $" # 2 !" # $" # 1 2 q C α F s 1 M22 ln 2 δ(1 − z) + L⊥ 2 − + Iq̄←q̄2 (z, x2T , µ) = δ(1 − Cz)F α −s − + 2α ln 2 δ(1 Iq̄←q̄ (z, xT , µ) = δ(1 − z) − 2π + L⊥$ ν1 − z) 2π $ α ν1 % & 2 % & 1+z 1 ++z 2 − (1 − z) (17) − z) + (17) (1 −− z)(1 + (1 − z)+ finds ower →first 0, βthe → power 0 , thenpα→→λp0,, one xtTaking p,α(p̄) to β→ 0 ( p → λp) If the expansions are performed in the oppo and ' & we obtain ' & ) x q ) x q !µ (z , x , µ) = I (z , x , µ) I (z , x , I (z , x , µ) I 4e(z , x , µ) = C Iα ( I (z, x , µ)4e= δ(1 − z) − 2π 2 2 In the product the 1/α divergences vanish, but anomalous M2 dependence remains. 26 % &− z)+ (1 2 1/α 1+z − (1 − z) + (17) (1 − z)+ 1/α CF αs • L⊥ + O(αs2 ) Fqq̄ (L⊥ , αs ) = π $ " CF αs 2 C α F s L⊥ + IO(α(z, ) Fqq̄ (L⊥ , αs ) = L2⊥ + q←q s L⊥ , αs ) = Iq̄←q̄ (z, L⊥ , αs ) = δ(1 − z) 1 + π CF αs $ " #% 4π 2 2 L⊥ + O(αs ) CF αs' 2 π ( q̄ (L⊥ , αs ) = Iπq←q (z, L⊥ , αs ) = Iq̄←q̄ (z, L⊥ , αs ) = δ(1 − z) 1− + CF αs LLP + 3L − 2 ⊥ ⊥ ⊥ q←q (z) − (16− z) + O(αs ) 4π $ " #% 2π 2 (18) C α π F s 2 ' ( ' z, L⊥ , αs ) = Iq̄←q̄ (z, L⊥ , αs ) = δ(1 − L⊥ + 3L⊥ − CFz) αs 1 + 4π TF αs 2 6 L⊥IP (z) − (1 − z) + O(α ) − (z, L , α ) = I (z, L , α ) = − L⊥ Pq←g (z) − 2z(1 q←q s q←g ⊥ s q̄←g ⊥ s (18) 2π 2π ( CF αs ' ( 2T α ' L⊥ Pq←q (z) − (1 − z) + O(α ) − F s For coefficients, we + obtain Iq←g , αsthe ) =first −s two expansion L⊥ Pq←g (z) − 2z(1 − z) O(αs2 ) 2π(z, L⊥ , αs ) = Iq̄←g (z, L⊥ 2π ' ( ) α *2 $ ΓF β TF αs αs2 F s 0 0 2 z, L⊥ , αs ) = Iq̄←g (z, L⊥ , αs ) = − L⊥ Pq←g (z) − 2z(1 − ⊥z), αs+ O(αs Γ ) 0 L⊥ + L⊥ + ΓF1 L⊥ Fqq̄ (L )= 2π • 4π 4π 2 $ " # % first two expansion coefficients, we obtain 808 224 dq2 = CF CA − 28ζ3 − TF nf 27 27 $ % ) α *2 ΓF β αs F s 0 0 2 Γ0 L⊥ + L⊥ + ΓF1 L⊥ + dq2 (19) Fqq̄ (L⊥ , αs ) = 4π 4π 2 Fqq̄ (L⊥ , αs ) Fgg (L⊥ , αs ) $ " # % = 808 224 CF (20) CA dq2 = CF CA − 28ζ3 − TF nf 27 27 For the functions I and F, one then obtains Solving its RG and using Davies, Stirling and Webber ’84 and de Florian and Grazzini ’01, we extract the two-loop F Casimir scaling F (L , α ) F (L , α ) All the necessary = input for NNLL resummation! (21) q q̄ ⊥ CF s gg ⊥ s CA 27 Fqq̄ (L⊥ , αs ) Fgg (L⊥ , αs ) = CF CA (21) RESUMMED RESULT , , d3 σ 4πα2 + 2 + + 1 dz1 1 dz2 = e dM 2 dqT2 dy 3Nc M 2 s q q i=q,g j=q̄,g ξ1 z1 ξ2 z2 # % $ " ξ1 ξ2 2 , , qT , M 2 , µ φi/N1 (z1 , µ) φj/N2 (z2 , µ) + (q, i ↔ q̄, j) . × Cqq̄→ij z1 z2 3 The hard-scattering kernel is ! !2 1 2 2 2 ! Cqq̄→ij (z1 , z2 , qT , M , µ) = CV (−M , µ)! 4π " d2 x⊥ e−iq⊥ ·x⊥ # x2T M 2 4e−2γE $−Fqq̄ (x2T ,µ) × Iq←i (z1 , x2T , µ) Iq̄←j (z2 , x2T , µ) • Two (22) sources of M dependence: hard function and anomaly • Fourier transform can be evaluated numerically or analytically, if higher-log terms are expanded out. 28 (23) n ∼ p0 COLLINS SOPER STERMAN FORMULA p̄µ n̄ ∼ 0 p̄ µ " ' ' ' " 1 dz1 " 1 dz2 d3 σ 4πα2 1 2 −iq⊥ ·x⊥ 2 = d x e e ⊥ q 2 dM 2 dqT dy 3Nc M 2 s 4π z1 ξ2 z2 q i=q,g j=q̄,g ξ1 ( " 2 ) *+ M 2 2 % & % & dµ̄ M × exp − ln 2 A αs (µ̄) + B αs (µ̄) 2 2 µ̄ µ̄ µb ) * % & % & s12 (ξ2b/z q̄j(q, 2 j/N qi1b/z1 ,1µb )i/N φj/N q̄, j)s2 , 2b 2 , αs (µ 2b ) φi/N s 1 (ξ 1 2 , µ1b ) + b i↔ qi× C1qi z1 ,sαs (µbb) Cq̄j zq̄j ! ! " # "# # " " # )2 (µ )z φ, α (µ (ξ )/zC, µ )zφ, α (µ (ξ /z × C z , α (µ ) C z×, α C 4 −γstandard E. The low scale is , and we set µ = b /x , and we adopt the cho = b0•/x , and we adopt the standard choice b = 2e b 0 T T 0 • Landau-pole singularity in the Fourier transform. To use the formula, one needs additional prescription to deal with this. 29 s cusp ! s" β(α s ) dgd 1 F 2 A αs = Γcusp (αs ) − −γE − d µ = µb = 2e2 xβ⊥ ! " B! αs" = 2γ q (αs ) + g1 (αs ) − β µq= B αs = 2γ (αsq)T+ g1 (αs ) − ! " # ! "# ! −γE 2 # # µ = µb = 2e /x⊥C"ijin! z, αs (µ CV !#− µ µb"2#$ Ii←j If adopt the choice our result to −F!q b )" =reduces # b , CSS 2 ! ! 2 x M #·x # Ii←j z 2 1 Cij z,2αs (µb )−iq = C − µ , µ T 2 2 2 V b ⊥ ⊥ b ! ! we identity z2 ,formula, qT , M provided , µ) = C d x⊥ e where V (−M , µ) ∞ −2γE % &n 4π 4e$ α where s anomaly contribution ∞ dq % &n g (α ) = F (0, α ) = $ 1 s s n α ! " s β(αs ) dg1 (αs ) 2, 2= F (0, αs ) = n=1 dqn 4π g (α ) A αs = ΓFcusp (αs ) − 1 s × Iq←i2 (z1dα ,x s T , µ) Iq̄←j (z2 , xT , µ) 4π n=1 ∞ $ ! " # # β(α ) dg (α ) s 2 s 2 (72) # 2 # B αs = 2γ q (αs ) + g1 (αs ) − , ∞ e g (α ) = ln C (−µ , µ) = $ # V # 2 dαs −γE 2 −1s 2 2 # # µ = µ = 2e x b g (α ) = ln C (−µ , µ) = n=1 e 2 ⊥s V ! " # ! "# ! " RELATION TO CSS Cij z, αs (µb) = #CV − µ2b , µb # Ii←j z, 0, αs (µb ) , The one-loop coefficients are dq1 = 0 and q µ=q ∞ The one-loop % α &n Tcoefficients are d1 = 0 and $ s n=1 ' 2 ( g1these (αs ) = Frelations (0, αs ) = todqnderive, unknown three-loop coefficient, Use necessary 7π q −γE ' (. 4π e = C − 16 n=1 2 e for/xNNLL F ⊥ 7π resummation ∞ (73) q1 3 # #2 $ q % αs &n 2 . g2 (αs ) = ln #CV (−µ , µ)# = en 4π q (3) TheF two-loop coefficient n=1 e1 = CF − 16 . 3 q Fin (49), while eq d has been given 2 A = Γ2 + 2β0 d2 = 239.2q2− 652.9 #= Γ2 q3 q The two-loop coefficient been giventoinB(α (49), while efficients are d1 = 0 and compiled in [31], howeverd2ithas contributes O(αes2 s ) at 3 ( in compiled in [31], however it contributes to B(α ) at O(α relations (72) are compatible with our perturbative resu s Not equal toq the 'cusp anom. dim. as was usually assumed! 2 s 7π e1 = CF relations (74) − 16 . in (72) are compatible our perturbative resuC according to (72) with the coefficient A in the 3 Note that 30 Note that according to (72) the coefficient A in the C anomalous q q dimension starting at three-loop order, and the efficient d2 has been given in (49), while e2 can be extracted from the results DIVERGENT EXPANSIONS, AND OTHER SURPRISES 31 TRANSVERSE MOMENTUM SPECTRUM The spectrum has a number of quite remarkable features wich we now discuss in turn: • Expansion in αs : strong factorial divergence • qT-spectrum: • calculable, even • expansion around qT = 0: worst expansion, ever • Long-distance • small, but near qT = 0 effects associated with ΛQCD OPE breaks down 32 CV (−M K(η, 0, r) =, µ) r d x⊥ e 2(η−1)γ E Γ(η) 4π e . 4e 4eE−2γE (33) (23) −2γ & 1 1 e 2 2 2−iq⊥ ·x⊥ −ηL⊥ − 4 aL2 ⊥ ≡ × Iq←i (z , x , µ) I (z , x , µ) d x e e K η, 1 q̄←j 2 T T ⊥ LEADING MOMENTUM DEPENDENCE a = αs4π (µ) × O(1) (34) 2 µ % ' & % E −2γE−1 ∞ −2γE 1 q e 2 T ·x⊥ (1−η)#− 4 a# #/2 µ = µ = 2e x ⊥ e d% J e b e = ≡ 0 at 0 b0 Up = 2e , and in the case hand to corrections suppressed by powers of α , the q 2 s T- µ 2 (24) µ µ −∞ 2 −ηL⊥−γ − 14EaL b⊥ −γ e E = αqT(µ) of dependence our formula result has the form aµ = × O(1) s (35) ) αs (µ) ( F 2 % a = Γ + &(M q, 2µ) ' β0 . −2γ E ηF 0 1 1C α 2 M−iq e 2 2 ⊥ ·x⊥ −ηL⊥ − 2π aL⊥ 4 F ds x⊥ e e K η, a, T , ≡ = O(1) qln F 2 = ηΓF2=4π + 2β d = 239.2 − 652.9 = # Γ π0 2 µ 2 µ2 µ2 (36) (25) 2e−γE , and in the 2case 2 at hand % xT µ2 & K(η, 0, r) = r η−1 Γ(1 − η) = ln with L⊥exp , and the two quantities −αs cL (26) 2(η−1)γE Γ(η) ⊥ −2γ E e ) 4e αs (µ) ( F 2 a= 2π Γ0 + ηF (M , µ) β0 . CF αs M 2 η= ln 2 = O(1) and π µ K(η, 0, r) = r η−1 a = αs (µ) × O(1) . Γ(1 − η) . 2(η−1)γ E e Γ(η) 33 (27) . 4e K π E 2 µ CF αs M η= ln 2 = O(1) x2T µ2 'µ∞ a = α Ls⊥ =× ln O(1) (µ) π % &E % 2& q −2γ q 4e PERTURBATIVE EXPANSION = dL J e b e η, a, T µ2 L⊥ /2 T (1−η)L⊥ − 41 aL2⊥ 0 0 2 2⊥ & % xT µ qC α −∞ 2 µ 1 2 term can M T µ ≈ q F s L /2 (1−η)L − aL L = ln Since is not large as long as , the T ⊥ ⊥ ⊥ 'J e 4% ⊥ 2 & −2γ ln = O(1) η = dLformally b e −2γ 4e ⊥ 0 0 2e qT 1 away π µ 2 be expanded −iq ·x −ηL − aL e µ d x⊥ e K η, a, 2 , ≡ 2 ! $ E ⊥ ⊥ ⊥ 4π 1 4 E 2 ⊥ µ µ % & 2 a 1 a 2 4 γE K(η, a, r) = 1 − ∂ + ∂ . . K(η, 0, r) % , and in the case at−2γ hand 2 .& η η + E qT e 4 2! 4 −ηL⊥ − 1 aL2 η,22 a, , ≡ ) αs (µ) ( 2F K x 2 2 µ µ µ a= Γ + η F (M T , µ) β0 . 0 L = the ln integal explicitly: and for a=0, we can perform 4 ⊥ 2π hand )( F Γ0 + ⊥ 4e−2γE 2− 2η) Γ(1 x µ η−1 T K(η, 0, r) =Lr⊥ = ln . 2(η−1)γ E Γ(η) E e 4e−2γ ) 2 5 ηF (Ma = , µ) β . αs (µ)µ0×≈O(1) qT 34 (36 K(η, a, r) − n! the critical4 n=0 = η is close to exp − n! then has 4 n=0 ∂η K(η, 0, r) = value 1. One ∂η r e2(η−1)γE Γ(η) , (65) FACTORIAL DIVERGENCE t is not difficult to see that series is factorially divergent. To illustrate we Γ(1the − η) 1 2ζ 2ζ5 this point, 3 2 4 = − to the (1default − η) scale − choice (1 −µη) .. , onsider the special case2(η−1)γ where Er = 1 (corresponding = qT+) .and e Γ(η) 1−η 3 5 is close to the critical value 1. One then has 2n+1 and taking 2n Γ(1 derivatives of1 the 2ζ leading term generates (2n)!/(1 − η) . A m − η) 2ζ 3 5 2 4 = − (1 − η) − (1 − η) + . . . , (66) Unfortunately, this series is 3strongly factorially divergent: E Γ(η) e2(η−1)γ 1−η 5 analysis reveals that ' ( . A$ ∞ ∞ the leading and taking 2n derivatives of term generates (2n)!/(1 − η)2n+1 more careful % & $ n # a 1 (2n)! −2γE n # = analysisK(η, reveals that − + k a + O(1 − − e a, 1) n + ... 2n+1 exp n! ' 4 (1 − η) ( ∞ ∞ n=0 n=0 $ $ (2n)! % a &n # 1 −2γE n # − + k a + O(1Nason, − η) , Ridolfi (67)‘99 − e K(η, a, 1) exp = first noted by: Frixione, n 2n+1 n! k do 4 not (1 − η) where the coefficients exhibit the strongn=0factorial growth of the terms n=0 n sum. is badly the fact thatofitthe hasterms alternating Can Borelthisresum it, which leads nonperturbative andfirstsign im where theWhile coefficients kseries exhibit divergent, the strongmakes factorial growth in the n do not can be Borel-summed. obtainthe factexplicit um. While this series is badly divergent, that it has alternating sign implies that it highly nontrivial a We dependence ) * ' + ' + ,( , an be Borel-summed. We obtain 2 # ) * π ' (1−η) + 1 1−η ' 1 + ,( −2γ E + ,(# a a √ +1 − e K(η, a, 1 − Erf √ # 1) Borelπ= (1−η)2 e 11−−ηErf −2γ 1 E # a K(η, a, 1) Borel = e aa 1 − Erf √ − e a a 1 − Erf √ a a a ∞ (68) $ ∞ $ n n k a + − η) , n k+ a + O(1 − η) ,O(1the In practice, it+is simplest, to use exact expression and n + ... n=0 n=0 evaluate K-function numerically. where Erf(x) is theiserror Note thatNote the singularity η = 1 has disappeared where Erf(x) the function. error function. that theatsingularity at η = 1 after has Borel summation. Expressions for the first few kn coefficients can be readily derived in terms of disapp Borel summation. Expressions for the first 35 few kn coefficients can be readily derived exp −αs cL2⊥ CLOW α M qT. VERY η= ln F 2 s µ2 π For moderate qT, the natural scale L choice ⊥ is μ = qT. However, detailed analysis shows that near qT ≈ 0 the Fourier integral is dominated by # $x−1 T % = q∗ = M exp − x (qwhich ∗ , qT ) π 2CF αs (q∗ ) $ = 1.75 GeV for M = MZ corresponds to η=1. pT λ= computed M → Spectrum can be methods down to qT=0! µ p nµ ∼ 0 p 36 with short-distance INTERCEPT AT QT=0 • Dedicated analysis of qT → 0 limit yields: N −#/αs dσ ∼√ e (1 + c1 αs + . . . ) 2 dqT αs Parisi, Petronzio 1979; Collins, Soper, Sterman 1985; Ellis, Veseli 1998 • Were for the first time we are able to compute the normalization N and NLO coefficient c1 • Expression singularity) cannot be expanded about αs = 0 (essential 37 K(η, a, r) = 1 + ∂η + xT µ∂η + . . . K(η, 0, r) = ln4 −2γ 4 L⊥ 2! 4e E x2T µ2 T L⊥ = ln −2γ 4e E % ' try to 1obtain & we ∞ −2γ −2γE E Given our for the intercept, can also 2 2 1 result q e e 2 2 T q⊥ ·x⊥ −ηL⊥ − 4 aL⊥ (1−η)#− 4 a# x0T µe#/2 b0 2 e d% J e K = ≡ derivatives with respect to q . Leading term is obtained by T L = ln 2 2 µ ⊥ −∞ 4e−2γE µ µ expanding SLOPE AT Q =0? 1 4π % µ ≈ qT −ηL⊥ − 14 aL2⊥ d2 x⊥ e−iq⊥ ·x⊥ e qT2 ' e−2γE & K η, a, 2 , ≡ 2 µ µ η=1 = 2e , and in the case at hand Yields extremely violently divergent series ) αs (µ) ( F 2 a= Γ0 + ηF (M , µ) β, 0 . 2 -n−1 ∞ n+1 + * 2π (−1) qT n2 /a * √ K(η = 1, a, qT ) exp ∼ e 2 a q ∗ n=1 Γ(1 − η) η−1 K(η, 0, r) = r . 2(η−1)γ E e Γ(η) 38 5 −γE NUMERICAL RESULTS • Spectrum can be calculated numerically, even though power expansion in qT2 is absolutely meaningless (not even Borel summable)! • Find smooth behavior down to very small qT Z production, Tevatron 10 25 8 20 6 4 LO+NLL 2 0 dqT �pb�GeV� NLO+NNLL dΣ dqT 2 dΣ �pb�GeV� Z production, Tevatron 15 10 5 0 1 2 3 4 5 6 qT 0 0 5 10 15 qT 39 20 25 30 INTERPRETATION Z production, Tevatron 10 area = total cross section = perturbative! non-perturbative dqT 2 dΣ �pb�GeV� 8 6 non-perturbative 4 perturbative 2 0 0 1 2 3 4 5 6 qT • Spectrum is short-distance dominated, but essential features are non-perturbative! 40 of the W boson. We emphasize that the above relation only holds for xT & 1/M , because otherwise the power corrections to our factorization formula become large. It can therefore not be used to study the xT → 0 limit of the functions Fqq̄ or Bi/N . LOW-ENERGY QCD EFFECTS 2.3 Simplifications at large qT2 For given transverse momentum qT , the Fourier integral in (18) receives important contributions from transverse separations xT ! qT−1 only. For large transverse momenta in the • perturbative domain, qT2 & Λ2QCD , we therefore need the xT -dependent PDFs at transverse separation xT # Λ−1 QCD . In this case these functions obey an operator-product expansion of the form [4, 34, 35] % ! 1 dz Bi/N (ξ, x2T , µ) = Ii←j (z, x2T , µ) φj/N (ξ/z, µ) + O(Λ2QCD x2T ) . (20) z ξ j Related question is that about the impact of longdistance power correction in matching relation In the context of SCET, generalized PDFs defined in terms of hadron matrix elements in • collinear fields are separated by distances that are not light-like are referred to as beam which functions. For such functions an analogous expansion was considered in [17], and an expression for the one-loop kernel of the quark beam function was derived in [18]. The evolution equations for the new kernels DGLAP equations 2 with the standard −Λ2 x2T 2 Ii←j follow when we combine2(16) T T % ! 1 du d φi/N (z, µ) = Pi←j (z/u, µ) φj/N (u, µ) . (21) d ln µ u z j Find that these cannot be analyzed order by order, but only numerically using functions that vanish at large x , such as θ(1 − Λ x ) or e • Fixed-order OPE in xT2 is again extremely 9 divergent. Same as expansion of K around qT=0 41 LONG-DISTANCE EFFECTS • Yet resummed behavior is smooth and rather insensitive to the way in which the cutoff is introduced: Z production, Tevatron, Gaussian cutoff Z production, Tevatron, Gaussian cutoff 10 25 �NP �0 �NP �0.1GeV �NP �0.3GeV �NP �0.5GeV �NP �0.7GeV dΣ 4 2 0 dqT 2 �pb�GeV� 20 6 dqT 2 dΣ �pb�GeV� 8 15 �NP �0 �NP �0.1GeV �NP �0.3GeV �NP �0.5GeV �NP �0.7GeV 10 5 0 1 2 3 4 5 0 6 qT 0 1 2 3 4 qT • Indications that long-distance effects are very small already above qT=2 GeV 42 5 6 Resulting power correction has a complicated shape, but is approximately linear: 10 ∼ �pb�GeV� 8 � ΛNP 1 − 0.7 × qT � dqT 2 q dΣ T �0 6 4 2 0 0.0 0.5 1.0 1.5 2.0 �NP Cannot be described in terms of a single operator matrix element 43 CONCLUSION • Have derived resummed result for Drell-Yan qT spectrum • • Reduces to the known CSS result for a special scale choice. • • • Naive factorization broken by collinear anomaly. Only the product of two transverse PDFs is well defined, but has anomalous dependence on the large momentum transfer Obtain three-loop coefficient A(3) , the last missing piece needed for NNLL accuracy. Many surprising features: • emergence of nonperturbative scale q*~2GeV: spectrum is shortdistance dominated, even at very low qT • strongly divergent expansions in αs, qT/q*, ΛQCD/q*. Phenomenological analysis at NNLL+NLO is in progress. 44
© Copyright 2026 Paperzz