Three-loop matching of the dipole operators for b -> s gamma and b -> s g

This is the matching step of the NNLO program for b -> s gamma, done properly. Misiak and Steinhauser compute the three-loop matching conditions for the dipole operators P7 and P8, which nobody had done before; Ref. [9] stopped at two loops. That is the new content, and it is real: new coefficients C7^(3) and C8^(3) in both the charm and top sectors, with the charm sector essentially exact and the top sector good to a percent or so. The calculation is done twice with independent tool chains (QGRAF/q2e/exp/MATAD and FeynArts plus hand-written code), the gauge-parameter independence of the bare three-loop sum is checked, and the exact x-dependence of the pole parts is recovered from the matching equation and cross-checked against the direct expansions. That is the kind of redundancy I want to see in a three-loop paper, and it is why I trust the pole parts without reservation. The renormalization constants they borrow from Refs. [14, 15] are published and standard; nothing load-bearing is cited to unpublished work. Ref. [12] is a forward reference to the four-quark anomalous dimensions and is not used here. The soft spot is the one the stress-test flags, and it does land. The finite top-sector pieces are known only as expansions in z = MW²/mt² (to z⁴) and w = 1 - z (to w⁸). For Ct7^(3) the two agree nicely, 12.11 versus 12.05. For Ct8^(3) they do not: plugging z = 0.238 into Eq. 6.19 gives about -0.51, against -1.2 from Eq. 6.20. The authors say the w-expansion is the better one at the physical point, and the one- and two-loop panels in Fig. 5 support that, but the quoted ±0.1 in Eq. 6.22 is not what the z-versus-w spread shows. It reads more like the spread of the last few w terms. So the error bar is optimistic, though the authors do acknowledge the truncation openly in Section 6. The saving grace is that this hardly matters: they note that moving Ct8^(3) from -1.2 to -1.3 shifts the decay width by under 0.02%. So the flaw is in the bookkeeping of an uncertainty, not in the physics. Worth a referee note asking them to widen the Ct8^(3) error or state explicitly that it excludes the z-expansion. Otherwise this is careful, honest work that anyone doing NNLO b -> s gamma will need. Send it out.

The paper's central numerical claim for the top-sector three-loop coefficient is Ct8(3)(mu0=mt) = -1.2 ± 0.1 (Eq. 6.22). This value comes from the w-expansion (Eq. 6.20) truncated at w^8. The independent z-expansion (Eq. 6.19) truncated at z^4, evaluated at the physical point z = (MW/mt)^2 ≈ 0.238, appears to give a value near -0.5, because the z^2 and z^3 coefficients (-98.34 and -127.1) are large and the series is dominated by cancellations among terms of order 4-6. If this arithmetic is correct, the two expansions disagree by about 0.7, seven times the quoted uncertainty. The authors acknowledge in Section 6 that the w-expansion works better at the physical point, but they do not fully quantify the discrepancy for Ct8(3). The quoted ±0.1 may therefore be too optimistic. This matters because the paper's claim is that the matching conditions are known to the stated precision; if the error bar is underestimated, the phenomenological reliability of the result is weaker than claimed. The same check for Ct7(3) gives 12.11 (z-expansion) versus 12.05 (w-expansion), consistent with ±0.05, so the concern is specific to Ct8(3).

The paper is a solid, well-executed calculation. The main claim — that the three-loop matching conditions are evaluated — is supported by two independent calculations, gauge-parameter independence checks, and cross-checks of the pole parts. The only soft spot is the precision of the finite parts, which are known only as truncated expansions. The reader correctly identifies this as the weakest assumption. However, the reader states that the uncertainties are estimated by comparing the two expansions at the physical point. For Ct7(3), this is true: the z-expansion gives 12.11 and the w-expansion gives 12.05, consistent with ±0.05. For Ct8(3), the z-expansion appears to give about -0.5 while the w-expansion gives -1.22, a spread of 0.7, much larger than the quoted ±0.1. If this arithmetic is correct, the error bar on Ct8(3) is underestimated. The authors acknowledge that the w-expansion works better, but they do not fully quantify the discrepancy. This is a real but not verdict-changing concern: the phenomenological impact of Ct8(3) is small (the authors note that a change from -1.2 to -1.3 affects the decay width by less than 0.02%), so the paper's main conclusion stands. The concrete test is to compute higher-order terms in the z-expansion and check convergence.

This completes the matching step of the NNLO program for B -> X_s gamma, which is needed to remove the charm-quark mass renormalization ambiguity in the NLO prediction. The two independent calculations and the gauge-parameter checks give confidence in the result. The exact pole parts are a strong result, and the numerical values are useful for phenomenology. The paper also notes that the effect of the NNLO matching alone on the decay width is small (within 2% in the MS scheme with MW < mu0 < mt).

The paper aims to complete the matching step for the next-to-next-to-leading QCD prediction of the $bar B to X_s gamma$ branching ratio. The motivation is that the Standard Model prediction's main theoretical uncertainty comes from the charm-quark mass ambiguity in the two-loop amplitude, and that ambiguity can only be removed by a three-loop, NNLO calculation. The authors compute the three-loop matching conditions for the dimension-five dipole operators $P_7$ and $P_8$, which are the only matching conditions for this process that were still missing at this order. They report numerical values for the top and charm sectors and argue that the matching contribution alone shifts the decay width by at most about 2% for $M_W < mu_0 < m_t$, smaller than the estimated higher-order uncertainty.

What carries the argument. The load-bearing device is the matching equation that equates the amputated one-particle-irreducible $b to s gamma$ and $b to s g$ Green functions in the full Standard Model with those in the effective theory, order by order in $tilde alpha_s$. The calculation uses the background-field 't Hooft-Feynman gauge, expands the integrands in external momenta and the $b$-quark mass, regulates spurious infrared divergences dimensionally, and reduces the diagrams to three-loop vacuum integrals evaluated with MATAD. Because MATAD handles only one non-zero mass, the authors use two expansions, around $m_t=M_W$ and $m_t gg M_W$, and combine them at the physical mass ratio; projection operations pick out only the two Dirac structures needed for $C_7$ and $C_8$.

Core claim. The central claim is that the three-loop matching coefficients $C_7^{(3)}$ and $C_8^{(3)}$ for the dipole operators are now determined. In the top sector the paper quotes $C_7^{t(3)}(mu_0=m_t)=12.05 pm 0.05$ and $C_8^{t(3)}(mu_0=m_t)=-1.2 pm 0.1$; in the charm sector it gives expansions around $M_W$ with $C_7^{c(3)}(mu_0=M_W) simeq 1.5$ and $C_8^{c(3)}(mu_0=M_W) simeq -1.9$. The finite parts are obtained as truncated expansions in $z=M_W^2/m_t^2$ and $w=1-M_W^2/m_t^2$, whose overlap at the physical mass ratio supplies the quoted uncertainties. The paper also shows that all $1/epsilon$ poles cancel in the matching equation and that the pole parts, though computed as expansions, have exact $x$-dependence fixed by that cancellation.

The premise it rests on. The finite three-loop coefficients at the physical mass ratio are reliably obtained from two truncated expansions that overlap there; if the expansions converge more slowly than their overlap suggests, the quoted central values and errors could shift.

The matching contribution to the $bar B to X_s gamma$ width stays within about 2% for $M_W < mu_0 < m_t$ in the $overline{rm MS}$ scheme, below the estimated higher-order perturbative uncertainty.

The calculation completes the first of the three steps, matching, mixing, and matrix elements, needed for the NNLO QCD prediction.

The quoted top-sector values are accurate enough for phenomenology: changing $C_7^{t(3)}$ from 12 to 13 would affect the decay width by only about 0.02%.

The charm-sector expansions can be used as they stand because their uncertainty is smaller than the experimental error in $m_t$.

The remaining work is the three-loop mixing and matrix-element calculation; until those are done, the NNLO prediction is not complete.

Editorial inference: the two-expansion overlap strategy is a reusable template for other three-loop matching problems with two widely separated mass scales, provided the expansions overlap in the physical region.

Editorial inference: the small phenomenological sensitivity to $C_8^{t(3)}$ means that even a modest shift in its central value would not by itself rescue or exclude new physics in $B to X_s gamma$; the decay is more sensitive to $C_7$.

Editorial inference: a useful stress test would be to push both expansions one or two orders higher and check that the physical-point values and their estimated errors are stable, or to obtain the finite parts exactly by an independent method.

This review was written by a machine. No person has reviewed this paper on Pith yet. A scientist listed on Pith can review it for a fee, and their review is published here whatever it concludes.

Request a human review