22
Rational terms in two-loop calculations M. F. Zoller in collaboration with S. Pozzorini and H. Zhang RADCOR 2019 – Avignon – 10.09.2019

M.F.Zoller · Rationaltermsintwo-loopcalculations M.F.Zoller in collaboration with S. Pozzorini and H. Zhang RADCOR2019–Avignon–10.09.2019

  • Upload
    others

  • View
    0

  • Download
    0

Embed Size (px)

Citation preview

  • Rational terms in two-loop calculations

    M. F. Zollerin collaboration with

    S. Pozzorini and H. Zhang

    RADCOR 2019 – Avignon – 10.09.2019

  • Motivation: Towards two-loop automation

    Example of a full NNLO calculation: gg → tt̄g

    ︸ ︷︷ ︸LO

    q1

    ︸ ︷︷ ︸NLO

    q1 q2

    ︸ ︷︷ ︸NNLO

    Tree and one-loop contributions fully automated in OpenLoops 2 [arxiv:1907.13071] for allLHC processes, including real emission → strong numerical stability and CPU efficiencyDownload: https://gitlab.com/openloops/OpenLoops or https://openloops.hepforge.org

    Challenges for numerical two-loop calculation:• Numerical amplitude construction → Numerator of Feynman integral in 4 dimensions• Tensor integral reduction and computation of Master integrals→ computed in d = 4− 2ε dimensions

    • Rational terms stemming from discrepancy between Feynman integrands with 4 and d dimensionalnumerators ⇒ one loop: Rational terms of type R2 [Ossola, Papadopoulos, Pittau]this talk: two-loop rational terms

    1

    https://arxiv.org/pdf/1907.13071.pdfhttps://openloops.hepforge.org/downloads

  • Rational terms at one loopGeneric one-loop diagram (d-dimensional quantities marked with bar)

    Ā1 =∫

    dq̄ N̄ (q̄)D̄0(q̄) · · · D̄N−1(q̄)

    =

    wN−1wN

    w1 w2

    D̄0

    D̄1

    D̄2

    D̄N−1

    q

    (scalar propagators D̄i(q̄) = (q̄ + pi)2 −m2i with loop momentum q̄, mass mi and external momentum pi)

    Splitting 4 and (d− 4) dimensional parts of the numerator yields

    Ā1 = A1 + ∆A1 with ∆A1 = Ā1 −A1 =∫

    dq̄ N̄ (q̄)−N (q)D̄0(q̄) · · · D̄N−1(q̄)

    ,

    where N (q) is constructed from 4-dimensional q, γµ, gµν and N̄ (q̄) from d-dimensionalq̄ = q + q̃, γ̄µ̄ = γµ + γ̃µ̃, ḡµ̄ν̄ = gµν + g̃µ̃ν̃

    Numerator Ñ = N̄ − N is of order ε⇒ only interaction with 1ε UV poles contributes to finite result [Binoth, Guillet, Heinrich]⇒ ∆A1 is a (finite) rational term, that can be written as a local counterterm[Ossola, Papadopoulos, Pittau]

    2

  • Renormalised amplitude from subtraction of divergencies (R-operation [Caswell, Kennedy])R Ā1 = Ā1 + δZĀ1TA1 (counterterm δZĀ1, tree-level structure TA1)

    Rational term can be written as a (local) counterterm ∆A1 = δZR2Ā1TA1

    ⇒ R Ā1 = A1 + (δZĀ1 + δZR2Ā1

    )TA1 =: R4A1

    One-loop diagram A1 with 4-dim numerator ⇒ suitable for numerical implementation

    Generalisation to two loopsRenormalisation procedure: Subtraction of sub-divergencies from one-loop sub-diagrams γ̄,then of local two-loop divergence:

    R Ā2 = Ā2 +∑γ̄

    (δZγ̄

    )· Ā2/γ̄ + δZĀ2TĀ2

    We show that a renormalised d-dim two-loop amplitude can be computed as

    R Ā2 = A2 + ∑γ(δZγ̄ + δZR2γ̄

    )· A2/γ +

    δZĀ2 + δZR2Ā2

    TĀ2 =: R4A2.with (extended) one-loop rational terms δZR2γ̄ and a two-loop rational term δZ

    R2Ā2⇒ Two and one-loop diagrams A2 and A2/γ constructed with 4-dim numerators

    3

  • Outline

    I. Computation of rational terms from tadpole integrals with one scale→ Rational terms of one-loop diagrams and sub-diagrams

    II. Rational terms of two-loop diagrams

    – without a global divergence

    – with a global divergence

    III. Full set of QED rational terms at two loops

    IV. Summary and Outlook

    4

  • Rational terms from tadpole integrals: The one-loop case

    Ā1 =∫

    dq̄ N̄ (q̄)D̄0(q̄) · · · D̄N−1(q̄)

    =

    wN−1wN

    w1 w2

    D̄0

    D̄1

    D̄2

    D̄N−1

    q R Ā1 = Ā1 + δZĀ1TA1

    Define difference between diagram with d-dim and 4-dim numerator: ∆A1 = Ā1−A1

    Use exact decomposition of propagator denominators D̄i = (q̄ + pi)2 −m2i [Chetyrkin, Misiak, Münz]

    1D̄i

    = 1q̄2 −M2

    + ∆iq̄2 −M2

    1D̄i

    = +

    with ∆i = −p2i − 2q̄ · pi + m2i −M2 recursively up to a fixed order Z + 1:

    1D̄i

    =Z∑n=0

    (∆i)n

    (q̄2 −M2)n+1+ (∆i)

    Z+1

    (q̄2 −M2)Z+11D̄i

    e.g. for Z = 2: = + + +

    ⇒ Isolate divergencies + rational terms in tadpole integrals with one auxiliary scale M2

    Terms with a negative degree of divergence X (naive power counting) cancel in ∆A1.

    5

  • One-loop exampleD = dimension of the loop momenta, metric and γ-matrices in the numerator ⇒ D ∈ {4, d}

    ∆A1 = D = d − D = 4 = D = d − D = 4 + D = d − D = 4

    ︸ ︷︷ ︸=O(ε)

    + D = d − D = 4

    ︸ ︷︷ ︸=O(ε)

    + D = d − D = 4

    ︸ ︷︷ ︸=O(ε)

    + D = d − D = 4

    ︸ ︷︷ ︸=O(ε)

    + D = d − D = 4

    ︸ ︷︷ ︸=O(ε)

    + D = d − D = 4

    ︸ ︷︷ ︸=O(ε)

    + D = d − D = 4

    ︸ ︷︷ ︸=O(ε)

    =∫

    dq̄ N̄ (q̄)−N (q)(q̄2 −M2)3

    +O(ε)

    6

  • One-loop rational terms from tadpoles

    In general: ∆A1 =N+X∑n=N

    ∫dq̄

    (N̄ (q̄)−N (q)

    )Pn(M2, p2i , q̄ · pj)

    (q̄2 −M2)npolynomial Pn from decom-position of denominators

    Decompose numerators:N̄ (q̄)Pn(M2, p2i , q̄ · pj) =

    R∑r=0N̄µ̄1···µ̄r q̄

    µ̄1 · · · q̄µ̄r

    N (q)Pn(M2, p2i , q̄ · pj) =R∑r=0Nµ1···µrq

    µ1 · · · qµr

    using q̄ · pj = q · pj for external 4-momenta pj. With AµiBµ̄i = AµiBµi for any vectors A,B

    ⇒ ∆A1 =N+X∑n=N

    R∑r=0

    (N̄µ̄1···µ̄r −Nµ1···µr

    )I µ̄1···µ̄rn with I µ̄1···µ̄rn =

    ∫dq̄ q̄

    µ̄1 · · · q̄µ̄r(q̄2 −M2)n

    • Generic method to compute ∆A1 from tadpole integrals with one (auxiliary) scale M2.• Result independent of M2 since the denominator decomposition is exact and only terms which

    cancel in ∆A1 are discarded.• Dependence on external momenta and masses resides only in N̄µ̄1···µ̄r and Nµ1···µr⇒ ∆A1 is a rational term

    7

  • One-loop sub-diagrams in two-loop diagramsOne loop: All external momenta in 4 dimensionsTwo loop: External momentum to a sub-diagram γ̄ can be d-dim loop momentum q̄1• Decomposition of d-dim denominators yields

    1(q̄1 + q̄2)2 −m2

    = 1q̄22 −M2

    + −q̄21 − 2q̄1 · q̄2 + m2 −M2

    q̄22 −M21

    (q̄1 + q̄2)2 −m2

    ⇒∫

    dq̄2[N (q2, q1)Pn(q̄21, q̄1q̄2)

    ]/(q̄22 −M2)n has terms ∝ q21 (from N construction in 4-dim),

    ∝ q̄21 (from denominator decomposition)

    • Difference of d-dim and 4-dim tensor structures t̄ᾱ(q̄1)− tα(q1) multiplied by one-loop UVcounterterm δZγ̄ not negligible due to insertion into a loop diagram

    Extended one-loop rational term for one-loop sub-diagram γ̄ with external indices α = (α1α2):

    (∆γ)α :=[γ̄ᾱ − γα +

    (t̄ᾱ − tα

    )δZγ̄

    ]ᾱ→α

    This has the structure ∆γ = R2(γ) + ∆R2(γ)ε with the well-known one-loop R2(γ).

    Additional rational term ∆R2(γ) ∝ q̃21 with q̃1 = q̄1 − q1 requires superficial degree ofdivergence = 2 in sub-diagram (e.g. gauge boson self-energy).

    8

  • Example: One-loop self-energy of the photonOne-loop amplitude in d and 4 dimensions from denominator decomposition and tadpole method:

    γ̄µ̄ν̄ =∫

    dq̄2N̄ µ̄ν̄sub(q̄1, q̄2)

    D̄(2)0 (q̄2)D̄

    (3)0 (q̄1 + q̄2)

    = ie2

    16π2

    −(4−2ε) q̄21

    3εgµ̄ν̄ + 4

    3εq̄µ̄1 q̄ν̄1 + C

    µ̄ν̄finite

    ,

    γµν =∫

    dq̄2N µνsub(q1, q2)

    D̄(2)0 (q̄2)D̄

    (3)0 (q̄1 + q̄2)

    = ie2

    16π2

    −4q2 − 2q̃213ε

    gµν + 43εqµ1 qν1 + C

    µνfinite

    ,

    where the constant parts Cµ̄ν̄finite − Cµνfinite = O(ε).

    For the difference of the UV counterterms(t̄µ̄ν̄ − tµν

    )δZγ̄ =

    [(q̄µ̄1 q̄ν̄1 − q

    µ1 qν1)−

    (q̄21g

    µ̄ν̄ − q21gµν)] ie2

    16π2−4

    ∝ q̃21εgµν for µ̄→ µ, ν̄ → ν

    This yield the extended rational term

    q̄1 q1 q̄1 q1

    (∆γ)µν =

    µ̄

    ν̄

    D = d −

    µ

    ν

    D = 4

    +

    µ̄

    ν̄

    µ

    ν

    δZγ̄

    µ̄→ µν̄ → ν

    = ie2

    16π2

    q21

    23︸ ︷︷ ︸

    known R2

    + q̃212

    3 ε︸ ︷︷ ︸new ∆R2(γ)/ε

    gµν

    9

  • Two-loop diagramsGeneric irreducible two-loop diagram consists of three chains and two connecting vertices V0, V1:

    Ā2 =

    w(1)1

    w(1)2

    w(1)N1

    D̄(1)0

    D̄(1)1

    D̄(1)2

    D̄(1)N1

    w(3)1

    w(3)N3

    D̄(3)0

    D̄(3)1

    D̄(3)N3

    w(2)1

    w(2)2

    w(2)N2

    D̄(2)0

    D̄(2)1

    D̄(2)2

    D̄(2)N2

    V0

    V1

    q1 q2−q2−q1

    =∫

    dq̄1∫

    dq̄2N̄ ᾱ11 (q̄1) N̄

    ᾱ22 (q̄2) N̄

    ᾱ33 (q̄3) Γ

    V0V1ᾱ1ᾱ2ᾱ3 N1∏

    i1=0D̄

    (1)i1 (q̄1)

    N1∏i2=0

    D̄(2)i2 (q̄2)

    N1∏i3=0

    D̄(3)i3 (q̄3)

    with q̄3 = −(q̄1 + q̄2)

    Renormalised diagram in d dimensions(what we need)

    Splitting in pieces with 4-dim numerators(what we can implement in OpenLoops)∆γ: one-loop rational term insertion

    R Ā2 = Ā2 +∑γ̄

    (δZγ̄

    )· Ā2/γ̄

    ︸ ︷︷ ︸=:R Ā2

    +δZĀ2

    R Ā2 = A2 +∑γ

    (δZγ̄ + ∆γ

    )· A2/γ

    ︸ ︷︷ ︸=:R4A2

    +δZĀ2 + ∆A2

    Our task: Compute ∆A2 = R Ā2 − R4A2 and show that it is rational.

    10

  • The case with no global but one subdivergenceUse factorisation and split sub-diagram (chains 2+3) and rest (chain 1) into d-dim and 4-dim parts:

    α1

    α2

    N̄γN̄1

    ᾱ = (ᾱ1ᾱ2)Ñ = N̄ − Nᾱ = d-dim indexα = 4-dim indextα = tensor, e.g.qα11 q

    α21 − q21gα1α2

    R Ā2 =∫

    dq̄1N1α(q1) + Ñ1 ᾱ(q̄1)

    D̄(1)0 · · · D̄

    (1)N1︸ ︷︷ ︸

    no divergence

    dq̄2N̄ ᾱγ (q̄1, q̄2)

    D̄(2)0 · · · D̄

    (2)N2D̄

    (3)0 · · · D̄

    (3)N3

    + t̄ᾱδZγ̄

    ︸ ︷︷ ︸

    finite → discard Ñ1 ᾱ

    =∫

    dq̄1N1α(q1)

    D̄(1)0 · · · D̄

    (1)N1

    dq̄2Nαγ (q1, q2)

    D̄(2)0 · · · D̄

    (3)N3

    + tαδZγ̄ + (∆γ)α +O(ε)

    with (∆γ)α =∫

    dq̄2Ñαγ (q1, q2)

    D̄(2)0 · · · D̄

    (3)N3

    +(t̄ᾱ − tα

    )δZγ̄ = R2(γ) +

    ∆R2(γ)ε

    Diagram can be computed with 4-dim numerators:R Ā2 = A2 +A2/γ

    (δZγ̄ + ∆γ

    )=: R4A2

    • No two-loop rational term, i.e. ∆A2 = 0• One-loop rational terms R2(γ) needed only to order ε0

    • Additional q̃21 term only in one-loop integrals → need to be computed

    11

  • The case with a global divergence

    • Define for each chain i = 1, 2, 3 the maximum degree of divergence of the full diagram(X ≤ 0) and any sub-diagram constructed from chains i, j 6= i (degree of sub-divergence Xij):

    Z1 = Max (X,X12, X13) , Z2 = Max (X,X12, X23) , Z3 = Max (X,X23, X13)

    • Use exact decomposition of propagator denominators on each chain i up to order Zi+ 1:1

    D(i)0 (q̄i) · · ·D

    (i)Ni

    (q̄i)=

    S(i)Zi + F(i)Zi

    1D

    (i)0 (q̄i) · · ·D

    (i)Ni

    (q̄i),

    with S(i)Zi selecting all terms contributing to a (sub-)divergence,

    S(i)Z1

    D(i)0 (q̄i) · · ·D

    (i)Ni

    (q̄i)=

    Zi∑n=0

    P(i)n (q̄i, p(i)j )(

    q̄2i −M2)N+1+n ← tadpoles with scale M2

    and F(i)Zi selecting all other terms, especially those with at least one original propagator D(i)j

    • In particular, terms selected by F(i)Zi have no global divergence ⇒ Apply previous case!

    12

  • The case with a global divergenceExact decomposition ⇒ isolate divergent parts ⇒ Discard terms which cancel from calculation:

    ∆A2 =S(1)Z1 + F

    (1)Z1

    S(2)Z2 + F(2)Z2

    S(3)Z3 + F(3)Z3

    ∆A2 = ∆A2,rat + 3∑i=1

    ∆A(i)2,sub + ∆A2,fin ,

    with ∆A2,rat = S(1)Z1S

    (2)Z2S

    (3)Z3∆A2 ,

    ∆A(i)2,sub = F(i)ZiS(j)ZjS

    (k)Zk

    ∆A2 ,

    ∆A2,fin = F(1)Z1F

    (2)Z2F

    (3)Z3∆A2

    +3∑i=1

    S(i)ZiF(j)ZjF(k)Zk∆A2 .

    ← only one-scale tadpoles

    ← no global divergence, at most onesub-divergence ⇒ O(ε)

    ← no global or sub-divergence ⇒ O(ε)

    ⇒ ∆A = ∆A2,rat is rational! Compute as in one-loop case from

    ∆A2 =Ni+1+Zi∑ni=Ni+1

    ∫dq̄1

    ∫dq̄2

    (N̄ (q̄1, q̄2)−N (q1, q2)

    )Pn1n2n3(M2, p2i , qi · pj)

    (q̄21 −M2)n1 (q̄22 −M2)n2 (q̄23 −M2)n3

    =Ni+1+Zi∑ni=Ni+1

    R∑r=0

    S∑s=0

    (N̄µ̄1···µ̄rν̄1···ν̄s −Nµ1···µrν1···νs

    )I µ̄1···µ̄rν̄1···ν̄sn1n2n3

    with I µ̄1···µ̄rν̄1···ν̄sn1n2n3 =∫dq̄1∫dq̄2 q̄

    µ̄11 ···q̄

    µ̄r1 q̄

    ν̄12 ···q̄

    ν̄s2

    (q̄21−M2)n1 (q̄22−M2)n2 (q̄23−M2)n3

    13

  • Example: QED vertex correctionLet D ∈ {4, d} be the numerator dimension. Decomposing chain 1 (with external fermion line):

    ∆A2 =

    D = d + D = d δZγ

    D = 4 + D = 4 (δZγ + ∆γ)

    =

    D = d + D = d δZγ

    D = 4 + D = 4 (δZγ + ∆γ)

    +

    D = d + D = d δZγ

    D = 4 + D = 4 (δZγ + ∆γ)

    ︸ ︷︷ ︸

    =0+O(ε) (case without global divergence)

    +more terms with original propagator denominators along outer chain︸ ︷︷ ︸=0+O(ε) (case without global divergence)

    black lines: original propagators D̄(i)j ; red lines: factors in denominator decomposition ∝ (q̄2i −M2).

    14

  • Example: QED vertex correctionDecomposing chains 2,3 (photon self-energy sub-diagram):

    ∆A2 =

    D = d + D = d + D = d + D = d + D = d

    + D = d + D = d δZγ

    D = 4 + D = 4 + D = 4

    + D = 4 + D = 4 D = 4 + D = 4 (δZγ + ∆γ)

    + D = d − D = 4

    ︸ ︷︷ ︸=0 (case without global divergence)

    + D = d − D = 4

    ︸ ︷︷ ︸=0 (case without global divergence)

    + . . .︸ ︷︷ ︸=0

    Only massive tadpoles contribute to difference ∆A2.

    15

  • Rational terms for QED in MS scheme (ξ = 0, m = 0)Calculation in the GEXCOM [Chetyrkin, M.Z.] framework: QGRAF [Noguira] → Q2E+EXP[Seidesticker, Harlander, Steinhauser] → FORM [Vermaseren] code → MATAD [Steinhauser]

    p

    ∆A1,e = ie2

    16π2 [−1] /p, ∆A2,e = ie4

    (16π2)2[ 1918ε +

    247108

    ]/p

    p

    µ ν ∆Aµν1,γ = ie2

    16π2[23p

    2 + 23εp̃2]gµν, ∆Aµν2,γ = ie

    4

    (16π2)2[(pµpν − gµνp2)

    ( 23ε −

    7118)

    + gµνp2(−1112

    )]

    µ

    ∆Aµ1,eeγ = ie3

    16π2 [−2] γµ, ∆Aν2,eeγ = ie

    5

    (16π2)2[139ε +

    19127

    ]γµ

    µ ν

    ρσ ∆Aµνρσ1,4γ = ie4

    16π2[43]tµνρσ, ∆Aµνρσ2,4γ = ie

    6

    (16π2)2 [−3] tµνρσ

    with tµνρσ = gµνgρσ + gµρgνσ + gµσgνρ

    • ∆Al are polynomial in p and independent of M 2 ⇒ Rational terms

    • One-loop rational term for photon self-energy extended with term ∝ p̃2/ε

    • Full dependence on ξ and electron mass m in upcoming paper16

  • Summary and Outlook

    • Renormalised diagrams in d-dimensions can be split into objects with 4-dimensional numerators:

    R Ā2 =A2 + ∑γ

    (δZγ̄ + ∆γ

    )· A2/γ

    + ∆A2with (extended) one-loop rational terms ∆γ and a two-loop rational term ∆A2.⇒ Numerical implementation in automated tools, e.g. OpenLoops, possible

    • We present a generic method to compute ∆Al from tadpoles with one auxiliary scale M2,which also serves as a proof that ∆Al is rational

    • Full set of QED rational terms at two-loop level

    • Further tasks:

    – Compute necessary one-loop integrals involving q̃2 to order ε

    – Implementation in OpenLoops

    17

  • Backup

    18

  • Reducible two-loop diagrams

    w(1)1

    w(1)2

    w(1)N1−2 w(1)N1−1

    D̄(1)2

    D̄(1)N1−2

    D̄(1)N1−1

    D̄(1)0

    D̄(1)1

    q1Pσ1σ2

    w(2)N2−2

    w(2)N2−1

    w(2)1

    w(2)2

    D̄(2)N2−1

    D̄(2)0

    D̄(2)1

    D̄(2)2

    D̄(2)N2−2

    q2

    The one-loop procedure is directly applicable for reducible two-loop diagrams (two one-loop diagramsconnected by a bridge P ) due to the factorisation

    RĀ2 = R

    dq̄1Tr

    N̄ (1)(q̄1)σ1

    D̄(1)0 (q̄1) · · · D̄

    (1)N1(q̄1)

    Pσ1σ2R∫

    dq̄2Tr

    N̄ (2)(q̄2)σ2

    D̄(2)0 (q̄2) · · · D̄

    (2)N2(q̄2)

    = R4

    dq̄1Tr

    N (1)(q1)σ1

    D̄(1)0 (q̄1) · · · D̄

    (1)N1(q̄1)

    Pσ1σ2R4∫

    dq̄2Tr

    N (2)(q2)σ2

    D̄(2)0 (q̄2) · · · D̄

    (2)N2(q̄2)

    and the fact that both terms R [...] are finite.

    19

  • Example with two sub-divergencesγ1: upper photon-fermion loop; γ2: lower photon-fermion loop.Contributions with one original propagator D̄(i)j (black) in sub-diagram γ̄2:

    0 =

    D = d + D = d + D = d + D = d

    + D = d δZγ̄1

    D = 4 + D = 4

    + D = 4 + D = 4 + D = 4 (δZγ̄1 + ∆γ1)

    ,

    20

  • Example with two sub-divergences

    Contributions with one original propagator D̄(i)j (black) in sub-diagram γ̄2:

    0 =

    D = d + D = d + D = d + D = d

    + D = d δZγ̄2

    D = 4 + D = 4

    + D = 4 + D = 4 + D = 4 (δZγ̄2 + ∆γ2)

    ,

    Contributions with one original propagator D̄(i)j (black) in both sub-diagrams γ̄1, γ̄2:

    0 = D = d − D = 4 .

    21