TTP12-046

SFB/CPP-12-94
ZU-TH 25/12
LPN12-130

Complete three-loop QCD corrections

to the decay H → γγ

P. Maierhöfer a, P. Marquard b
a Institut für Theoretische Physik, Universität Zürich, 8057 Zürich, Switzerland
b Institut
für Theoretische Teilchenphysik,
Karlsruhe Institute of Technology (KIT), 76128 Karlsruhe, Germany

Abstract

We present the result for the three-loop singlet QCD corrections to the decay of a
Higgs boson into two photons and improve the calculation for the non-singlet case.
With the new result presented, the decay width Γ(H → γγ) is completely known
at O(GF α2 α2s , GF α3 ).

Key words: Perturbative calculations, Quantum Chromodynamics, Higgs

1 Introduction

After the discovery of a new boson at about 126 GeV at the LHC [1,2] it is
imperative to study all decay modes of the Standard Model Higgs boson as
precisely as possible. In this paper we are concerned with the decay of the
Higgs boson into two photons which is the decay channel with the largest
significance for the discovery of the Higgs boson at the LHC.

The decay of a Higgs boson into two photons is mediated through either
charged gauge bosons or top quarks at one-loop order. The decay rate can be
cast into the form
M3
Γ(H → γγ) = H |AW + At |2 , (1)
64π

Preprint submitted to Elsevier 14 January 2013

with the leading order values

q√
   
(0) α 2GF 3 3 1 √
AW = − 2+ + 2− arcsin2 τW ,
2π τW τW τW
(0) 3
 
1

√ 
At = Ât 1+ 1− arcsin2 τ , (2)
2τq τ
√
2α 2GF 2
Ât = Nc Qt ,
3π

where τW = MH2 /(4MW 2

) and τ = MH2 /(4m2t ). The former contribution is
larger than the latter (by a factor 4.5) and opposite in sign. Both contributions
have been investigated in great detail in the literature. The two-loop QCD
corrections to the decay have first been evaluated in the heavy-top limit in
Refs. [3–5] and later, keeping the full top-mass dependence, in Refs. [6–8]. The
two-loop electroweak corrections have been investigated in [9–12]. Combining
the two-loop QCD and electroweak corrections one observes a nearly complete
cancellation between these two contributions for MH = 126 GeV as discussed
below.

At next-to-next-to-leading order (NNLO) the non-singlet QCD contributions

(cf Fig. 1 (c)-(e)) have been calculated in the heavy top limit, including ad-
ditional terms in an expansion in τ = MH2 /(4m2t ), in Ref. [13]. At this order
a new class of diagrams, so-called singlet diagrams (cf Fig. 1 (f)-(g)), con-
tributes for the first time. They can be characterized by the property, that the
external lines are coupled to different fermion loops. The contribution from
this kind of diagrams has not been taken into account up to now and there is
no formal argument, that it should be suppressed compared to the non-singlet
contribution, when considering the same order in αs . In this letter we present
the calculation of this last missing piece to obtain a complete NNLO QCD
prediction for the decay of a Higgs boson into two photons. In addition, to
improve the existing prediction and to check our setup, we recalculated the
non-singlet contribution and added more terms in the expansion in τ .

This letter is organized as follows: In Section 2 we describe the necessary

steps of the calculation and in Section 3 we present the full three-loop QCD
corrections to the decay width. In Section 4 we give the numerical result and
our conclusions.

2
 γ γ
H H
γ γ
(a) (b)

γ γ γ
H H H
γ γ γ
(c) (d) (e)

γ γ
q
H t q H t
γ γ
(f ) (g)

Fig. 1. Sample diagrams of the 1-loop (a), 2-loop (b), 3-loop non-singlet (c)-(e), and
3-loop singlet (f)-(g) top quark induced contribution to H → γγ.
2 Calculation

We start from the decay amplitude of a Higgs boson into two photons
µ
q1
Aµν = H . (3)
q2
ν
Decomposing the Lorentz structure in the most general way the amplitude
Aµν can be written as

Aµν = q1ν q2µ A + g µν B + q1µ q2ν C + q1µ q1ν D + q2µ q2ν E , (4)

with scalar form factors A, B, C, D, and E, which can be calculated by

applying the projectors

(d − 1)q1µ q2ν + q1ν q2µ − q1 q2 g µν

PAµν = ,
(d − 2)(q1 q2 )2
q1 q2 g µν − q1µ q2ν − q1ν q2µ
PBµν = ,
(d − 2)q1 q2
(d − 1)q1ν q2µ + q1µ q2ν − q1 q2 g µν
PCµν = , (5)
(d − 2)(q1 q2 )2
q2µ q2ν
PDµν = ,
(q1 q2 )2
q1µ q1ν
PEµν =
(q1 q2 )2

on the amplitude (e. g. A = PAµν Aµν ). Due to gauge invariance the contractions
q1µ Aµν and q2ν Aµν of the amplitude with the momenta of the external photons

3
vanish. This imposes the relations B = −q1 q2 A and D = E = 0 on the form
factors. In order to verify these relations explicitly we calculate A, . . . , E
separately. The coefficient C, although non zero, does not contribute to the
decay rate of the Higgs boson because of the transversality of the photon wave
functions. Therefore the Lorentz structure of the physical amplitude simplifies
to

Aµν = (q1ν q2µ − q1 q2 g µν )A. (6)

To obtain the decay width, the combination A = AW + At is inserted into

eq. (1) for the W and the top quark loop induced contributions.

Since the calculation of the decay amplitude including the full dependence
on Higgs-boson and top-quark mass is not possible at the moment, only an
expansion in τ = MH2 /(4m2t ) can be obtained. The expansion of the amplitude
in τ can be performed following two conceptually different strategies. The first
option, which is only applicable for the calculation of a few terms in the ex-
pansion, relies on the direct expansion of each Feynman diagram in τ and the
subsequent reduction of the appearing integrals to master integrals. Since this
method becomes very time-consuming when calculating higher terms in the
expansion, a second approach has been devised. In this alternative approach,
the full mass dependence is kept as long as possible, meaning that the reduc-
tion to master integrals is performed keeping the full mass dependence and
only the master integrals are calculated in an expansion in τ . In the following
we sketch the steps necessary to perform the calculation and explain how the
results for the needed master integrals in an expansion in τ can be obtained.

The Feynman diagrams are generated with qgraf [14] and processed by a
Mathematica program to map the integrals on 4 singlet and 15 non-singlet
topologies. A sample of the contributing diagrams is depicted in Fig. 1. All
appearing integrals are reduced to master integrals using Crusher [15]. To
expand the master integrals Mi (τ ) in τ = MH2 /(4m2t ) a series ansatz
X (k) (k) (k)

Mi (τ ) = Mi,0 + Mi,1 (−τ )−ǫ + Mi,2 (−τ )−2ǫ τ k (7)
k

is inserted into the differential equation [16]

 
∂ ∂ 1
MH2 2
+ m2t 2
− D̂ Mi (τ ) = 0, (8)
∂MH ∂mt 2

where D̂ applied to Mi (τ ) returns the mass dimension of Mi . The differen-

tial equation follows from the fact that Mi (τ ) is a homogeneous function in
its dimensionful parameters MH2 and m2t . Quarks except for the top quark
are treated as massless. Using the series ansatz results in a system of alge-
(k)
braic equations which are systematically solved for the coefficients Mi,j by a
Mathematica program which uses Fermat [17] to simplify rational functions.

4
 (a) (b) (c) (d)

(e) (f ) (g) (h)

Fig. 2. Integrals which serve as boundary conditions for the solutions of the differ-
ential equations for the vertex master integrals. Dashed internal lines are massless,
solid lines carry the mass mt . External lines on the right side of the diagrams (a)-(e)
are massless on-shell, the external line on the left side is on-shell with mass MH .
These diagrams only contribute to the singlet case.
(k)
By this procedure all coefficients Mi,j are expressed as linear combinations of
the integrals which are depicted in Fig. 2 and serve as boundary conditions.
(k) (k)
Coefficients Mi,1 and Mi,2 of non-integer powers of τ arise only in the singlet
contribution due to the massless cuts of the singlet diagrams [18].

3 Results

The top quark loop induced amplitude At is expressed as a perturbative se-

ries in the strong coupling constant αs with the one-, two-, and three-loop
(0) (1) (2)
contributions At , At , and At ,
  2 
(0) αs (1) αs (2)
At = Ât At + At + At + ... . (9)
π π
The three-loop contribution is, furthermore, split into the non-singlet part
(2) (2)
At,0 , the singlet part with two top-quark loops At,t , and the singlet part with
(2)
a top-quark and light-quark loop At,q
(2) (2) (2) X (2)
At = At,0 + At,t + (Q−2
t Q2q )At,q (10)
q6=t

Qt and Qq are the electromagnetic charges of the top quark and the light
quarks (q ∈ {u, d, s, c, b}).

The full results with the expansion in τ up to τ 20 including generic SU(N)

colour factors and renormalisation scale dependence are available online. 1 For
the presentation of the results in this section we truncate the expansion after
τ 5 and insert the SU(3) colour factors explicitly. The top-quark mass is renor-
malized in the MS scheme. We use the abbreviations Lµ = log µ2 /m2t in the
1 http://www-ttp.particle.uni-karlsruhe.de/Progdata/ttp12/ttp12-046/

5
non-singlet and Lτ = log(−4τ ) in the singlet contribution. For completeness
also the αs0 and αs1 contributions are given as an expansion in τ .

         
(0) 7 2 26 512 1216
At =+1+τ + τ2 + τ3 + τ4 + τ5 (11)
30 21 525 17325 63063
   
(1) 38 7 1664 8
At =−1+τ − Lµ + τ 2 − Lµ

135 15 
14175 21
 
3 12626 52 4 40664 4096
+τ − Lµ + τ − − Lµ

496125 175 2338875

17325
9128671204 12160
+ τ5 − − Lµ (12)
245827456875 63063
(2) 31 7
At,0 = − − Lµ
24
 4 
22326329 4116067 769 7 2
+τ − + ζ3 − Lµ + Lµ
 622080 138240 1080 120 
68094821183 508541309 1241 3
+ τ2 − + ζ3 − Lµ + L2µ
 2612736000 23224320 1575 7
102458003430188113 190929277363
+ τ3 − + ζ3
122903101440000 275251200
256363 221 2
− Lµ + L

496125 350 µ
489470768471800920451 547186023461087
+ τ4 − + ζ3
202790117376000000

272498688000
14462 512 2
− Lµ + L
 66825 693 µ
114742890543235030923359893 1766782778485181879
+ τ5 − + ζ3
6430815309390151680000

119027426918400
307117801 1520 2
+ Lµ + L (13)
7449316875 1911 µ
(2) 1
At,t = +
8 
28777 749
+τ − + ζ3

207360 3072 
2 34183679 18935 2
+τ + + ζ3 − Lτ

522547200 221184 45 
3 2665898377 1419929 41
+τ − + ζ3 − Lτ
 390168576000 23592960 2025 
4 37746440955049 11964631 598
+τ − + ζ3 − Lτ
 1931334451200000 235929600 42525 
5 2381500300071333647 862749257 74924
+τ − + ζ3 − Lτ
231991894278144000000 28311552000 7016625
(14)

6
 (2) 13 2 1
At,q = − + ζ3 + Lτ
12
 3 6 
3493 7 19
+τ − + ζ3 + Lτ
 48600 45 1620 
3953 4 1
+ τ2 − + ζ3 − Lτ

396900 63 3780 
3 3668899 52 1696
+τ − + ζ3 − Lτ

2679075000 1575 1063125 
414962 1024 136
+ τ4 + + ζ3 − Lτ

2210236875 51975 91125 
5 611578464557 2432 7571576
+τ + + ζ3 − Lτ (15)
1409328810264375 189189 6257426175

(2)
The results for At,0 agree with the results presented in Ref. [13]. In Tab. 1
we give the results for the first 20 coefficients of the series in τ in numerical
form. For the singlet contribution we show the results for the constant and the
logarithmic part proportional to Lτ separately. To illustrate the convergence of
the series we show all 20 terms of the expansion of the three-loop contribution
in graphical form in Fig. 3. Around τ = 0.14, corresponding to the Higgs-
boson mass favoured by the LHC experiments, the first four (five) terms in
the expansion are sufficient to obtain an accurate result with a relative error of
10−5 wrt. the total 3-loop amplitude for the singlet (non-singlet) contribution.
At larger values of τ more terms of the expansion have to be taken into account,
e.g. at τ = 0.5 nine (sixteen) terms are needed to obtain similar accuracy.

Collecting all available information the amplitude can be cast in the form
 2
(0) (0) α (1) αs (1) αs (2)
AH→γγ = AW + Ât At + AEW + Ât At + Ât At , (16)
| {z } π | {z } π | {z } π | {z }
ALO ANLO−EW ANLO−QCD ANNLO

(1)
where AEW denotes the electroweak corrections to the W and top-quark in-
duced processes combined. The partial decay width is then given by

MH3  2 α αs
ΓH→γγ = ALO + (2ALO ANLO−EW ) + (2ALO ANLO−QCD )
64π π  2
π
αs 
+ (2ALO Re(ANNLO ) + A2NLO ) . (17)
π

For a Higgs boson with a mass of MH = 126 GeV (τ = 0.14) this evaluates
numerically to

ΓH→γγ = (9.398 · 10−6 − 1.48 · 10−7 + 1.68 · 10−7 + 7.93 · 10−9 ) GeV (18)
= 9.425 · 10−6 GeV,

7
 4

3
singlet
2 non-singlet

1
(2)

0
At

−1

−2

−3

−4
0 0.2 0.4 0.6 0.8 1
τ
Fig. 3. Numerical results for the non-singlet and singlet contribution at three-loop
order. In the plot the first 20 terms in the expansion in τ are shown. The singlet plot
(2) P (2)
shows the sum At,t + q6=t Q2q /Q2t At,q . For the plot of the non-singlet contribution
(2)
At,0 we set the renormalization scale µ = MH . The vertical line at τ = 0.14 marks
the value of τ for MH = 126 GeV. The convergence of the expansion for τ → 1 can
be improved by using the on-shell scheme for the top-quark mass.
where we used mt (MH ) = 166 GeV 2 , αs (MH )/π = 0.0358, GF = 1.16637 ·
10−5 GeV−2 , and α = α(0) = 1/137 as input parameters. The value for the
two-loop electroweak correction was taken from Ref. [10]. It has to be noted,
that there is a partial cancellation between the two-loop QCD and electroweak
corrections. To assess the influence of the singlet diagrams the next-to-next-
to-leading order term can be further decompsed as

ΓH→γγ |NNLO = (7.5 · 10−10 |NLO2 + 1.73 · 10−9 |non−singlett + 5.45 · 10−9 |singlet ) GeV ,

which shows, that the singlet diagrams are a factor of three larger than the
non-singlet ones and thus the most important three-loop contribution.

4 Conclusion

We presented new results for the singlet diagrams which contribute to the
decay of a Higgs boson into two photons at next-to-next-to-leading order,
2 The MS mass mt (MH ) was obtained from the on-shell mass Mt = 172.64 GeV
using the MS-on-shell relation at 3-loop accuracy [19–21].

8
 (2) (2) (2) (2) (2)
n At,0 At,q [0] At,q [Lτ ] At,t [0] At,t [Lτ ]
0 -1.29166667 -0.28196206 0.16666667 0.12500000 0.00000000
1 -0.09881144 0.11511420 0.01172840 0.15430166 0.00000000
2 0.25870689 0.06636139 -0.00026455 0.16832244 -0.04444444
3 0.16373113 0.03831749 -0.00159530 0.06551243 -0.02024691
4 0.08688592 0.02387041 -0.00149246 0.04141534 -0.01406232
5 0.05709451 0.01588624 -0.00121001 0.02636532 -0.01067807
6 0.05798951 0.01114079 -0.00095670 0.01555892 -0.00622762
7 0.07545680 0.00814321 -0.00075859 0.01272091 -0.00587319
8 0.10086885 0.00615332 -0.00060811 0.00751470 -0.00333698
9 0.12935841 0.00477756 -0.00049383 0.00708618 -0.00353585
10 0.15827468 0.00379378 -0.00040622 0.00423275 -0.00200757
11 0.18622498 0.00307013 -0.00033822 0.00437893 -0.00229504
12 0.21252358 0.00252489 -0.00028474 0.00264217 -0.00130721
13 0.23687981 0.00210551 -0.00024213 0.00291761 -0.00157907
14 0.25922151 0.00177712 -0.00020779 0.00177454 -0.00090205
15 0.27959438 0.00151591 -0.00017979 0.00205621 -0.00113694
16 0.29810454 0.00130527 -0.00015673 0.00125855 -0.00065080
17 0.31488570 0.00113331 -0.00013755 0.00151292 -0.00084866
18 0.33008045 0.00099137 -0.00012148 0.00093081 -0.00048637
19 0.34383005 0.00087306 -0.00010789 0.00115159 -0.00065225
20 0.35626884 0.00077356 -0.00009631 0.00071159 -0.00037400
Table 1
Numerical values of the first 20 terms in the expansion in τ for the three-loop non-
(2) (2) (2)
singlet At,0 and singlet contribution At,t and At,q . For the non-singlet part we set
(2)
µ = mt . In case of the singlet contribution the results for the constant part (At [0])
(2)
and the logarithmic part (At [Lτ ]) are shown separately.

which up to now have not been considered. The corrections to the decay rate
due to singlet diagrams are about a factor of three larger than the non-singlet
ones. An improved prediction for the non-singlet contribution reduces the
error of the three-loop contribution, which can be neglected for a Higgs-boson
mass of 126 GeV. The total partial decay width is ΓH→γγ = ΓLO + ΓNLO−EW +
ΓNLO−QCD +ΓNNLO = (9.398·10−6 −1.48·10−7 +1.68·10−7 +7.93·10−9) GeV =
9.425 · 10−6 GeV.

9
Acknowledgements

We are grateful to S. Uccirati and C. Sturm for providing the exact numerical
value for the two-loop electroweak contribution. We like to thank N. Zerf
for crosschecks on the non-singlet part and J.H. Kühn, M. Steinhauser and
T. Kasprzik for carefully reading the manuscript. This work was supported
by the DFG through the SFB/TR 9 “Computational Particle Physics”. The
work of Ph.M. was supported by the Swiss National Science Foundation.

References

[1] G. Aad et al. [ATLAS Collaboration], Phys. Lett. B 710 (2012) 49

[arXiv:1202.1408 [hep-ex]].

[2] S. Chatrchyan et al. [CMS Collaboration], Phys. Lett. B 710 (2012) 26

[arXiv:1202.1488 [hep-ex]].

[3] H. - Q. Zheng and D. - D. Wu, “First order QCD corrections to the decay of
the Higgs boson into two photons,” Phys. Rev. D 42 (1990) 3760.

[4] A. Djouadi, M. Spira, J. J. van der Bij and P. M. Zerwas, “QCD corrections
to gamma gamma decays of Higgs particles in the intermediate mass range,”
Phys. Lett. B 257 (1991) 187.

[5] S. Dawson and R. P. Kauffman, “QCD corrections to H → gamma gamma,”

Phys. Rev. D 47 (1993) 1264.

[6] J. Fleischer, O. V. Tarasov and V. O. Tarasov, “Analytical result for the two
loop QCD correction to the decay H → 2 gamma,” Phys. Lett. B 584 (2004)
294 [hep-ph/0401090].

[7] R. Harlander and P. Kant, “Higgs production and decay: Analytic results at
next-to-leading order QCD,” JHEP 0512 (2005) 015 [hep-ph/0509189].

[8] U. Aglietti, R. Bonciani, G. Degrassi and A. Vicini, “Analytic Results for

Virtual QCD Corrections to Higgs Production and Decay,” JHEP 0701 (2007)
021 [hep-ph/0611266].

[9] S. Actis, G. Passarino, C. Sturm and S. Uccirati, “NNLO Computational

Techniques: The Cases H → gamma gamma and H → g g,” Nucl. Phys. B
811 (2009) 182 [arXiv:0809.3667 [hep-ph]].

[10] G. Passarino, C. Sturm and S. Uccirati, “Complete Two-Loop Corrections to H

→ gamma gamma,” Phys. Lett. B 655 (2007) 298 [arXiv:0707.1401 [hep-ph]].

[11] G. Degrassi and F. Maltoni, “Two-loop electroweak corrections to the Higgs-

boson decay H → gamma gamma,” Nucl. Phys. B 724 (2005) 183 [hep-
ph/0504137].

10
[12] F. Fugel, B. A. Kniehl and M. Steinhauser, “Two loop electroweak correction
of O(G(F)M(t)**2) to the Higgs-boson decay into photons,” Nucl. Phys. B 702
(2004) 333 [hep-ph/0405232].

[13] M. Steinhauser, “Corrections of O (α2s ) to the decay of an intermediate mass

Higgs boson into two photons,” In *Tegernsee 1996, The Higgs puzzle* 177-185
[hep-ph/9612395].

[14] P. Nogueira, “Automatic Feynman graph generation,” J. Comput. Phys. 105

(1993) 279.

[15] D. Seidel and P. Marquard, unpublished.

[16] E. Remiddi, “Differential equations for Feynman graph amplitudes,” Nuovo

Cim. A 110 (1997) 1435 [hep-th/9711188].

[17] R. H. Lewis, “Fermat’s User Guide,” http://www.bway.net/˜lewis.

[18] A. Maier, P. Maierhöfer and P. Marquard, “Higher Moments of Heavy Quark

Correlators in the Low Energy Limit at O(α2s ),” Nucl. Phys. B 797 (2008) 218
[arXiv:0711.2636 [hep-ph]].

[19] K. Melnikov and T. v. Ritbergen, “The Three loop relation between the MS-bar
and the pole quark masses,” Phys. Lett. B 482 (2000) 99 [hep-ph/9912391].

[20] P. Marquard, L. Mihaila, J. H. Piclum and M. Steinhauser, “Relation between

the pole and the minimally subtracted mass in dimensional regularization and
dimensional reduction to three-loop order,” Nucl. Phys. B 773 (2007) 1 [hep-
ph/0702185].

[21] K. Melnikov and T. van Ritbergen, Nucl. Phys. B 591 (2000) 515 [hep-
ph/0005131].

11