ZU-TH 20/24 April 2024
DESY-24-053
LTH 1367

Four-loop splitting functions in QCD
– The quark-to-gluon case –
G. Falcioni${}^{\,a,b}$, F. Herzog${}^{\,c}$, S. Moch${}^{\,d}$, A. Pelloni${}^{\,e}$ and A. Vogt${}^{\,f}$

^aDipartimento di Fisica, Università di Torino, Via Pietro Giuria 1, 10125 Torino, Italy
^bPhysik-Institut, Universität Zürich, Winterthurerstrasse 190, 8057 Zürich, Switzerland
^cHiggs Centre for Theoretical Physics, School of Physics and Astronomy
The University of Edinburgh, Edinburgh EH9 3FD, Scotland, UK
^dII. Institute for Theoretical Physics, Hamburg University
Luruper Chaussee 149, D-22761 Hamburg, Germany
^eInstitute for Theoretical Physics, ETH Zürich, 8093 Zürich, Switzerland

^fDepartment of Mathematical Sciences, University of Liverpool
Liverpool L69 3BX, United Kingdom
Abstract

We present the even-$N$ moments $N\leq 20$ of the fourth-order (N³LO) contribution $P_{\rm gq}^{\,(3)}(x)$ to the quark-to-gluon splitting function in perturbative QCD. These moments, obtained by analytically computing off-shell operator matrix elements for a general gauge group, agree with all known results, in particular with the moments $N\leq 10$ derived before from structure functions in deep-inelastic scattering. Using the new moments and the available endpoint constraints, we construct approximations for $P_{\rm gq}^{\,(3)}(x)$ which improve upon those obtained from the lowest five even moments. The remaining uncertainties of this function are now practically irrelevant at momentum fractions $x>0.1$. The resulting errors of the convolution of $P_{\rm gq}$ at N³LO with a typical quark distribution are small at $x\raisebox{-1.99168pt}{$\>\>\stackrel{{\scriptstyle>}}{{{\scriptstyle\sim}}}\>\>$}10^{-3}$ and exceed 1% only at $x\raisebox{-1.99168pt}{$\>\>\stackrel{{\scriptstyle<}}{{{\scriptstyle\sim}}}\>\>$}10^{-4}$ for a strong coupling $\alpha_{\rm s}=0.2$. The present results for $P_{\rm gq}^{\,(3)}(x)$ should thus be sufficient for most collider-physics applications.

Next-to-next-to-next-to-leading order (N³LO) radiative corrections in QCD form an essential ingredient for precision physics at the Large Hadron Collider (LHC) [1] and the future Electron-Ion Collider [2]. Almost a decade after the pioneering computation of the total cross section for Higgs-boson production via gluon-gluon fusion [3], N³LO partonic cross sections are becoming available for an increasing number of LHC processes. Besides the partonic cross sections, fully consistent N³LO calculations of collider-physics observables require the fourth-order contributions to the splitting functions for the scale dependence (evolution) of the parton distribution functions (PDFs). These N³LO splitting functions are not fully known yet, but substantial progress towards their determination has been made over the past years.

The N³LO flavour non-singlet quark splitting functions $P_{\rm ns}^{\,(3)}(x)$ are fully known in the limit of a large number of colours $n_{c}$ [4]. Analytic results for their $x$-dependence in full QCD are known for the leading [5] and next-to-leading [6] contributions in the limit of a large number of flavours ${\color[rgb]{0,0,0}{n_{\!f}}}$; recently the QED-like ${\color[rgb]{0,0,0}{C^{\,3}_{F}}}{\color[rgb]{0,0,0}{n_{\!f}}}$ contribution has also been completed [7]. For all other colour factors, only the first eight even- or odd-$N$ moments have been published [4]; in the meantime those calculations have reached $N=22$ [8].

So far, the situation is less favourable for the evolution of the flavour-singlet PDFs,

$$\frac{d}{d\ln{\mu^{\,2}}}\;\Big{(}\begin{array}[]{c}\!q_{\rm s}\!\!\\ \!g\!\end{array}\Big{)}\>=\>\left(\begin{array}[]{cc}\!P_{\rm qq}&P_{\rm qg}\!\!\!\\ \!P_{\rm gq}&P_{\rm gg}\!\!\!\end{array}\right)\otimes\Big{(}\begin{array}[]{c}\!q_{\rm s}\!\!\\ \!g\!\end{array}\Big{)}\quad\mbox{with}\quad P_{\,\rm ik}(x,\alpha_{\rm s})\,=\,\sum_{n=0}{a_{\rm s}^{\,n+1}}P_{\,\rm ik}^{\,(n)}(x)\>,$$ (1)

where $a_{\rm s}=\alpha_{\rm s}/(4\pi)=g_{s}^{2}/(4\pi)^{2}$ denote the strong coupling, $q_{\rm s}\,=\,\sum_{\,i=1}^{\,{\color[rgb]{0,0,0}{n_{\!f}}}}\,(q_{i}+\bar{q}_{i})$ and $g$ are the singlet quark and gluon PDFs, and $\otimes$ represents the Mellin convolution in the momentum variable $x$. Only the ${\color[rgb]{0,0,0}{n^{\,3}_{\!f}}}$ leading large-${\color[rgb]{0,0,0}{n_{\!f}}}$ part of the splitting-function matrix in eq. (1) is fully known at N³LO [6]; recently the ${\color[rgb]{0,0,0}{n^{\,2}_{\!f}}}$ contributions to the initial-quark quantities $P_{\rm ps}^{\,(3)}(x)=P_{\rm qq}^{\,(3)}(x)-P_{\rm ns}^{\,(3)+}(x)$ and $P_{\rm gq}^{\,(3)}(x)$ have been derived in refs. [9, 10]. The even moments of $P_{\rm ik}^{\,(3)}(x)$ have been computed to $N=12$ for ${\rm ik}={\rm qq}$ and to $N=10$ for all other entries in eq. (1) using structure functions in deep-inelastic scattering (DIS) [11, 12].

Up to a conventional sign, these even-$N$ moments are identical to the anomalous dimensions of the twist-2 operators of spin $N$,

$$\gamma_{\,\rm ik}(N)\>=\>-\int_{0}^{1}\!dx\>x^{\,N-1}\,P_{\,\rm ik}(x)\;,\quad\gamma_{\,\rm ik}(N,\alpha_{\rm s})\>=\>\sum_{n=0}{a_{\rm s}^{\,n+1}}\,\gamma_{\,\rm ik}^{\,(n)}(N)\;,$$ (2)

which can be computed efficiently by renormalizing off-shell operator matrix elements (OMEs) ${\rm A}_{\rm ik}=\langle\,{\rm k}(p)|O_{\rm i}|\,{\rm k}(p)\rangle$, where $O_{\rm i}$ indicates the quark ($\rm i=q$) or gluon ($\rm i=g$) twist-2 operators. $\gamma_{\,\rm ps}^{\,(3)}$ and $\gamma_{\,\rm qg}^{\,(3)}$ were recently determined in this manner up to $N=20$ in refs. [13, 14]. Here we report on the corresponding results for $\gamma_{\,\rm gq}^{\,(3)}$.

In the non-abelian flavour-singlet cases, the renormalisation of the OMEs involves a mixing between gauge invariant and unphysical operators, also known as aliens [15, 16, 17, 18, 19, 20]. Following ref. [19], the aliens entering the calculation of $\gamma_{\,\rm gq}$ were organised as follows. The first class is given by

$$O_{\!\rm A}^{\,I}\;=\;\eta(N)\,\left(D.F^{a}+{g_{\rm s}}\overline{\psi}\,\hbox to0.0pt{\hbox{$\mskip 1.0mu/$}\hss}\Delta\,t^{a}\psi\right)\,\left(\partial^{\,N-2}A_{a}\right)\,,\quad O_{\rm c}^{\,I}\;=\;-\eta(N)\,\left(\partial\,\bar{c}^{\,a}\right)\left(\partial^{\,N-1}c_{a}\right)\,,$$ (3)

where $\Delta$ is a lightlike vector, $A_{a}=\Delta_{\mu}\,A_{a}^{\mu}\,$, $\partial=\Delta_{\mu}\,\partial^{\,\mu}$ and $D.F^{a}=\Delta_{\nu}(\partial_{\mu}\delta^{\,ab}+{g_{\rm s}}f^{acb}A^{c}_{\mu})\,F^{\mu\nu,b}$. We have determined $\eta(N\,\leq 20)$ to order ${\alpha_{\rm s}^{\,3}}$ and found agreement with ref. [20]. Next we need

$$O_{\!\rm A}^{\,II} \!=\! {g_{\rm s}}\,f^{\,a\,a_{1}a_{2}}\left(D.F^{a}+{g_{\rm s}}\overline{\psi}\,\hbox to0.0pt{\hbox{$\mskip 1.0mu/$}\hss}\Delta\,t^{a}\psi\right)\sum_{\begin{subarray}{c}n_{1},\,n_{2}\,\geq\,0\\ n_{1}+n_{2}\,=\,N-3\end{subarray}}\kappa_{\,n_{1}n_{2}}^{(1)}\left(\partial^{\,n_{1}}A_{a_{1}}\right)\left(\partial^{\,n_{2}}A_{a_{2}}\right)\,,$$ (4)

$$O_{\rm c}^{\,II} \!=\! -{g_{\rm s}}\,f^{\,a\,a_{1}a_{2}}\left(\partial\,\bar{c}^{\,a}\right)\sum_{\begin{subarray}{c}n_{1},\,n_{2}\,\geq\,0\\ n_{1}+n_{2}\,=\,N-3\end{subarray}}\eta^{(1)}_{\,n_{1}n_{2}}\left(\partial^{\,n_{1}}A_{a_{1}}\right)\left(\partial^{\,n_{2}+1}c_{a_{2}}\right)\,.$$ (5)

After picking a basis of independent constants $\eta^{(1)}_{\,n_{1}n_{2}}$ and $\kappa^{\,(1)}_{\,n_{1}n_{2}}$ under the relations imposed by BRST and antiBRST symmetry [19], we have determined all the required two-loop mixing constants up to $N=20$. The calculation of the renormalisation constants will be described in detail in a future publication. We found complete agreement for those renormalisation constants which were given in ref. [20]. Furthermore we require the alien operators

$$O_{\!\rm A_{1}}^{\,III} \!=\!\! {g_{\rm s}^{2}}f^{\,a\,a_{1}x}f^{\,a\,_{2}a_{3}x}\,\left(D.F^{a}+{g_{\rm s}}\bar{\psi}\hbox to0.0pt{\hbox{$\mskip 1.0mu/$}\hss}\Delta T^{a}\psi\right)\!\!\!\sum_{\begin{subarray}{c}n_{1},\,n_{2},\,n_{3}\,\geq\,0\\ n_{1}+n_{2}+n_{3}\,=\,N-4\end{subarray}}\!\!\!\kappa^{(1)}_{\,n_{1}n_{2}n_{3}}\,\left(\partial^{\,n_{1}}A_{a_{1}}\right)\left(\partial^{\,n_{2}}A_{a_{2}}\right)\left(\partial^{\,n_{3}}A_{a_{3}}\right)\,,\;\;$$ (6)

$$O_{\!\rm A_{2}}^{\,III} \!=\!\! {g_{\rm s}^{2}}d^{\,a\,a_{1}a_{2}a_{3}}\,\left(D.F^{a}+{g_{\rm s}}\bar{\psi}\hbox to0.0pt{\hbox{$\mskip 1.0mu/$}\hss}\Delta T^{a}\psi\right)\!\!\sum_{\begin{subarray}{c}n_{1},\,n_{2},\,n_{3}\,\geq\,0\\ n_{1}+n_{2}+n_{3}\,=\,N-4\end{subarray}}\!\!\kappa^{(2)}_{\,n_{1}n_{2}n_{3}}\,\left(\partial^{\,n_{1}}A_{a_{1}}\right)\left(\partial^{\,n_{2}}A_{a_{2}}\right)\left(\partial^{\,n_{3}}A_{a_{3}}\right)\,,$$ (7)

$$O_{\rm c}^{\,III} \!=\!\! -{g_{\rm s}^{2}}f^{\,a\,a_{1}x}f^{\,a\,_{2}a_{3}x}\,\left(\partial\bar{c}^{\,a}\right)\!\!\sum_{\begin{subarray}{c}n_{1},\,n_{2},\,n_{3}\,\geq\,0\\ n_{1}+n_{2}+n_{3}\,=\,N-4\end{subarray}}\!\!\eta^{(1)}_{\,n_{1}n_{2}n_{3}}\,\left(\partial^{\,n_{1}}A_{a_{1}}\right)\left(\partial^{\,n_{2}}A_{a_{2}}\right)\left(\partial^{\,n_{3}+1}c_{a_{3}}\right)\,.$$ (8)

We have computed a basis of independent mixing constants $\kappa_{\,ijk}^{\,(1)}$ at one-loop order, finding agreement with ref. [20]. While the gluonic and ghost parts were already presented in ref. [19], the quark contributions to the operators in eq. (6) are presented here for the first time in this form.

The calculation of $\gamma_{\,\rm gq}^{\,(3)}$ requires the OMEs $A_{\rm gq}$ to four loops and $A_{\rm qq}$ to three loops. In addition, the OMEs $A_{\rm iq}$ for all alien operators i are needed at the third order. The Feynman diagrams for the OMEs have been generated using Qgraf [21] and processed, see ref. [22] using a Form [23, 24, 25] program that classifies them according to their colour factors [26] and topologies. The calculation of the two-point functions has been performed by an optimized in-house version of Forcer [27]. By renormalizing the OMEs computed in $4-2\varepsilon$ dimensions, we have obtained the anomalous dimension $\gamma_{\,\rm gq}^{\,(3)}$. For QCD, i.e., the gauge group SU($n_{c}\!=\!3$), this results in the numerical values

$$\gamma_{\,\rm gq}^{(3)}(N\!=\!2)\; =\! -16663.2255+4439.14375\,{\color[rgb]{0,0,0}{n_{\!f}}}-202.555479\,{\color[rgb]{0,0,0}{n^{\,2}_{\!f}}}-6.37539072\,{\color[rgb]{0,0,0}{n^{\,3}_{\!f}}}\>,$$

$$\gamma_{\,\rm gq}^{(3)}(N\!=\!4)\; =\! -6565.75315+1291.00675\,{\color[rgb]{0,0,0}{n_{\!f}}}-16.1461902\,{\color[rgb]{0,0,0}{n^{\,2}_{\!f}}}-0.83976340\,{\color[rgb]{0,0,0}{n^{\,3}_{\!f}}}\>,$$

$$\gamma_{\,\rm gq}^{(3)}(N\!=\!6)\; =\! -3937.47937+679.718506\,{\color[rgb]{0,0,0}{n_{\!f}}}-1.37207753\,{\color[rgb]{0,0,0}{n^{\,2}_{\!f}}}-0.13979433\,{\color[rgb]{0,0,0}{n^{\,3}_{\!f}}}\>,$$

$$\gamma_{\,\rm gq}^{(3)}(N\!=\!8)\; =\! -2803.64411+436.393057\,{\color[rgb]{0,0,0}{n_{\!f}}}+1.81494625\,{\color[rgb]{0,0,0}{n^{\,2}_{\!f}}}+0.07358858\,{\color[rgb]{0,0,0}{n^{\,3}_{\!f}}}\>,$$

$$\gamma_{\,\rm gq}^{(3)}(N\!=\!10) \!=\! -2179.48761+310.063163\,{\color[rgb]{0,0,0}{n_{\!f}}}+2.65636842\,{\color[rgb]{0,0,0}{n^{\,2}_{\!f}}}+0.15719522\,{\color[rgb]{0,0,0}{n^{\,3}_{\!f}}}\>,$$

$$\gamma_{\,\rm gq}^{(3)}(N\!=\!12) \!=\! -1786.31231+234.383019\,{\color[rgb]{0,0,0}{n_{\!f}}}+2.82817592\,{\color[rgb]{0,0,0}{n^{\,2}_{\!f}}}+0.19211953\,{\color[rgb]{0,0,0}{n^{\,3}_{\!f}}}\>,$$

$$\gamma_{\,\rm gq}^{(3)}(N\!=\!14) \!=\! -1516.59810+184.745296\,{\color[rgb]{0,0,0}{n_{\!f}}}+2.78076831\,{\color[rgb]{0,0,0}{n^{\,2}_{\!f}}}+0.20536518\,{\color[rgb]{0,0,0}{n^{\,3}_{\!f}}}\>,$$

$$\gamma_{\,\rm gq}^{(3)}(N\!=\!16) \!=\! -1320.36106+150.076970\,{\color[rgb]{0,0,0}{n_{\!f}}}+2.66194730\,{\color[rgb]{0,0,0}{n^{\,2}_{\!f}}}+0.20798493\,{\color[rgb]{0,0,0}{n^{\,3}_{\!f}}}\>,$$

$$\gamma_{\,\rm gq}^{(3)}(N\!=\!18) \!=\! -1171.29329+124.717778\,{\color[rgb]{0,0,0}{n_{\!f}}}+2.52563073\,{\color[rgb]{0,0,0}{n^{\,2}_{\!f}}}+0.20512226\,{\color[rgb]{0,0,0}{n^{\,3}_{\!f}}}\>,$$

$$\gamma_{\,\rm gq}^{(3)}(N\!=\!20) \!=\! -1054.26140+105.497994\,{\color[rgb]{0,0,0}{n_{\!f}}}+2.39223358\,{\color[rgb]{0,0,0}{n^{\,2}_{\!f}}}+0.19938504\,{\color[rgb]{0,0,0}{n^{\,3}_{\!f}}}\>.$$ (9)

The ${\color[rgb]{0,0,0}{n^{\,2}_{\!f}}}$ and ${\color[rgb]{0,0,0}{n^{\,3}_{\!f}}}$ parts of eqs. (9) agree with the all-$N$ expressions of refs. [6, 10]. The values at $N\leq 10$ agree with those obtained in refs. [11, 12] from structure functions in DIS – a conceptionally much simpler but computationally much more involved approach. Our new exact results at $N\geq 12$, expressed in terms of rational numbers and values of Riemann’s $\zeta$-function, can be found in eqs. (A.3) – (A.7) in the appendix for a general compact simple gauge group.

The all-$N$ expressions of the NⁿLO anomalous dimensions in the $\overline{\mbox{MS}}$ scheme include $\zeta_{m}$ up to $m_{\rm max}=2n-1$ (but not ${\zeta_{2}}$), powers of simple denominators $D_{a}$ and harmonic sums $S_{\vec{w}}(N)$ [28], see also ref. [29]. Up to three loops (N²LO) [30, 31] the anomalous dimensions (2) have the form

$$\gamma_{\,\rm ik}^{\,(n)}(N)\big{|}_{\zeta_{\>\!m}}\! \!=\! \sum_{\;a^{\phantom{a}}}\;\sum_{p=p_{0}}^{2n+1-m}\>\>\sum_{w=0}^{2n+1-m-p}\!\!c_{mapw}^{\,(n)}\,D_{a}^{\,p}\,S_{w}(N)\quad\mbox{where }\quad D_{a}\;=\;\frac{1}{(N+a)}\;,$$ (10)

$$S_{\pm m_{1}}(N)\; =\! \sum_{j=1}^{N}\;(\pm 1)^{j}\,\frac{1}{j^{\,m_{1}}}\;,\quad S_{\pm m_{1},\,m_{2}}(N)\;=\;\sum_{j=1}^{N}\>(\pm 1)^{j}\;\frac{1}{j^{\,m_{1}}}\;S_{m_{2}}(j)\;,\quad\ldots$$ (11)

with ${\zeta_{0}}\equiv 1$, offsets $a=-1,\,0,\,1,\,2$ and minimal powers $p_{0}=0$ for i$\,=\,$k and $p_{0}=1$ for i$\,\neq\,$k. $S_{w}(N)$ in eq. (10) is a shorthand for all $2^{w}\!-\!1$ harmonic sums of a given weight $w$ – defined as the sum of the absolute values of the indices $m_{i}$ ($m_{i}=-1$ does not occur) – with $S_{0}(N)\equiv 1$. The coefficients $c_{mapw}^{\,(n)}$ are integer, modulo small-prime denominators that can be readily removed. Thus Diophantine equations can be used for their determination from a limited number of $N$-values.

On top of the above contributions, terms with $D_{-2}$ (in special combinations with sums $S_{\vec{w}}$, as there is no pole at $N=2$) occur in the coefficient functions for inclusive DIS already at two loops [32, 33, 34]. At the third order in $\alpha_{\rm s}$, positive powers of $N$ enter the DIS coefficient functions with a special weight-5 combination of sums and Riemann-$\zeta$ values [35],

$$f(N)\;=\;5\*{\zeta_{\>\!5}}-2\*{{S}_{-5}}+4\>\!\*{\zeta_{\>\!3}}\>\!\*{{S}_{-2}}-4\>\!\*{{S}_{-2,-3}}+8\*{{S}_{-2,-2,1}}+4\>\!\*{{S}_{3,-2}}-4\>\!\*{{S}_{4,1}}+2\*{{S}_{5}}\;.$$ (12)

If such additional contributions are present in $\gamma_{\,\rm ik}(N)$ beyond three loops, then they are visible already in the terms with $\zeta_{\>\!m\geq 3}$, in the case of $D_{-2}$ by the appearance of delta-function contributions ${\delta}(N-2)$. The comparatively simple $\zeta_{\>\!m\geq 3}$ terms can thus provide useful information about the most complex parts of the functions, the non-$\zeta$ contributions.

Up to terms in addition to eq. (10), we expect the ${\zeta_{5}}$ contributions to $\gamma_{\,\rm gq}^{\,(3)}$ to resemble the ${\zeta_{3}}$ parts of $\gamma_{\,\rm gq}^{\,(2)}$. In particular, $S_{1}\equiv S_{1}(N)$ should enter with the leading-order combination of $1/(N\!+\!a)$ denominators, $p_{\rm{gq}}\,=\,2/(N\!-\!1)-2/N+1/(N\!+\!1)$. The available 10 moments are then sufficient for a direct determination of the coefficients $c_{5apw}^{\,(3)}$ by systems of linear equations, resulting in

$${}\gamma_{\,\rm gq}^{\,(3)}(N)\big{|}_{{\zeta_{5}}}=160\,\*{\color[rgb]{0,0,1}{C^{\,4}_{F}}}\*\Big{(}(N+2)/12-28\,\*{D_{-1}}-239/3\,\*{D_{0}}+473/6\,\*{D_{1}}+52\,\*{D_{0}^{\,2}}+26\,\*{D_{1}^{\,2}}+14\,\*p_{\rm{gq}}\,\*{{S}_{1}}\Big{)}$$

$$\mbox{}+160\,\*{\color[rgb]{0,0,1}{C^{\,3}_{F}}}\*{\color[rgb]{0,0,1}{C_{A}}}\,\*\Big{(}-(N+2)/6+44\,\*{D_{-1}}+442/3\,\*{D_{0}}-428/3\,\*{D_{1}}-92\,\*{D_{0}^{\,2}}-46\,\*{D_{1}^{\,2}}-22\,\*p_{\rm{gq}}\,\*{{S}_{1}}\Big{)}$$

$$\mbox{}+80\,\*{\color[rgb]{0,0,1}{C^{\,2}_{F}}}\*{\color[rgb]{0,0,1}{C^{\,2}_{A}}}\,\*\Big{(}(N+2)/6-44\,\*{D_{-1}}-548/3\,\*{D_{0}}+1007/6\,\*{D_{1}}-8/3\,\*{D_{2}}+102\,\*{D_{0}^{\,2}}+51\,\*{D_{1}^{\,2}}$$

$$\mbox{}-8\,\*{D_{-1}^{\,2}}+25\,\*p_{\rm{gq}}\,\*{{S}_{1}}\Big{)}+80/3\,\*{\color[rgb]{0,0,1}{C_{F}}}\*{\color[rgb]{0,0,1}{C^{\,3}_{A}}}\,\*\Big{(}26\,\*{D_{-1}}+224/3\,\*{D_{0}}-166/3\,\*{D_{1}}+16/3\,\*{D_{2}}-24\,\*{D_{0}^{\,2}}$$

$$\mbox{}-12\,\*{D_{1}^{\,2}}+16\,\*{D_{-1}^{\,2}}-19\,\*p_{\rm{gq}}\,\*{{S}_{1}}\Big{)}+320\,\*{\color[rgb]{0,0,1}{d_{\,R}^{\,abcd}d_{\,A}^{\,abcd}\!/n_{c}}}\*\ \Big{(}-10\,\*{D_{-1}}-202/3\,\*{D_{0}}+391/6\,\*{D_{1}}$$

$$\mbox{}-8/3\,\*{D_{2}}-8\,\*{D_{-1}^{\,2}}+42\,\*{D_{0}^{\,2}}+21\,\*{D_{1}^{\,2}}+8\,\*p_{\rm{gq}}\,\*{{S}_{1}}\Big{)}+320\,\*{\color[rgb]{0,0,1}{n_{\!f}}}\,\*{\color[rgb]{0,0,1}{C^{\,3}_{F}}}\,\*p_{\rm{gq}}-160/3\,\*{\color[rgb]{0,0,1}{n_{\!f}}}\,\*{\color[rgb]{0,0,1}{C^{\,2}_{F}}}\*{\color[rgb]{0,0,1}{C_{A}}}\,\*p_{\rm{gq}}$$

$$\mbox{}+160/9\,\*{\color[rgb]{0,0,1}{n_{\!f}}}\,\*{\color[rgb]{0,0,1}{C_{F}}}\*{\color[rgb]{0,0,1}{C^{\,2}_{A}}}\,\*\Big{(}-26\,\*{D_{-1}}+62\,\*{D_{0}}-49\,\*{D_{1}}-24\,\*{D_{0}^{\,2}}-12\,\*{D_{1}^{\,2}}\Big{)}$$

$$\mbox{}+640/3\,\*{\color[rgb]{0,0,1}{n_{\!f}}}\,\*{\color[rgb]{0,0,1}{d_{\,R}^{\,abcd}d_{\,R}^{\,abcd}\!/n_{c}}}\*\ \Big{(}64\,\*{D_{0}}-68\,\*{D_{1}}+8\,\*{D_{-1}}-48\,\*{D_{0}^{\,2}}-24\,\*{D_{1}^{\,2}}\Big{)}\;.$$ (13)

The contributions of the two quartic colour factors, see eqs. (A.1) and (A.2), have been obtained before [36], the remaining terms are new. The $C_{F}^{\,k}\,C_{A}^{\,4-k}$ contributions at $k\geq 2$ indeed include terms beyond eq. (10); these numerator-$(N\!+\!2)$ terms hint at the appearance of the function $(N\!+\!2)f(N)$ in the ${\color[rgb]{0,0,0}{n^{\,0}_{\!f}}}$-part of $\gamma_{\,\rm gq}^{\,(3)}$. This situation is completely analogous to $\gamma_{\,\rm qg}^{\,(3)}$ in ref. [14], where these terms are of the form $(N\!-\!1){\zeta_{5}}$, cf. the relation between the off-diagonal $\gamma_{\,\rm ik}^{\,(3)}$ in ref. [36]. In both cases these extra contributions vanish for $C_{F}=C_{A}$, which is part of the choice of the colour factors that leads to a ${\cal N}\!=\!1$ supersymmetric theory, for lower-order discussions see refs. [37, 38].

The ${\zeta_{4}}$ parts of $\gamma_{\,\rm ik}$ are special, as all $\pi^{\>\!2}$ terms – see ref. [39] for first five-loop results that include also ${\zeta_{6}}\propto\pi^{\>\!6}$. Using the three-loop DIS coefficient functions of refs. [35, 40], the ${\zeta_{4}}$ part of $\gamma_{\,\rm gq}^{\,(3)}$ can be predicted from the no-$\pi^{\,2}$ theorem [41, 42, 43]; the result is given by eq. (11) of ref. [44].

Ten $N$-values are not sufficient for all-$N$ determinations of the more complex ${\color[rgb]{0,0,0}{n^{\,0}_{\!f}}}\,{\zeta_{3}}$ and ${\color[rgb]{0,0,0}{n^{\,1}_{\!f}}}\,{\zeta_{3}}$ contributions to $\gamma_{\,\rm gq}^{\,(3)}$ – except for the quartic colour factors, which are very closely related [36] to those occurring in $\gamma_{\,\rm qg}^{\,(3)}$ [14]. For one colour factor, ${\color[rgb]{0,0,0}{n_{\!f}}}\,{\color[rgb]{0,0,0}{C^{\,3}_{F}}}$, the diagrams are simple enough to allow the extension of our Forcer calculations to $N=30$. This was sufficient to obtain and check the corresponding coefficients $c_{3apw}^{\,(3)}$ in eq. (10) by a system of Diophantine equations. We hence find

$${}\gamma_{\,\rm gq}(N)\big{|}_{{\zeta_{3}}}\;=\;128\,\*{\color[rgb]{0,0,1}{d_{\,R}^{\,abcd}d_{\,A}^{\,abcd}\!/n_{c}}}\,\*\Big{(}\big{\{}7/12+33/2\,\*\eta-96\,\*{{S}_{1}}\,\*\eta+32\,\*{{S}_{1}}\,\*\nu-24\,\*{{S}_{-2}}-6\,\*{{S}_{1}}\,\*\eta^{2}$$

$$\mbox{}+10\,\*{{S}_{-2}}\,\*\eta-16\,\*{{S}_{-2}}\,\*\nu+43\,\*{{S}_{1}^{\,2}}\,\*\eta-16\,\*{{S}_{1}^{\,2}}\,\*\nu-8\,\*({{S}_{-2,1}}-{{S}_{1,-2}})+8\,\*{{S}_{1}}\,\*{{S}_{-2}}-{{S}_{3}}\big{\}}\,\*p_{\rm{gq}}$$

$$\mbox{}-\big{\{}40+24\,\*{{S}_{1}}-72\,\*{{S}_{-2}}-10\,\*{{S}_{1}^{\,2}}+72\,\*({{S}_{-2,1}}-{{S}_{1,-2}})+72\,\*{{S}_{1}}\,\*{{S}_{-2}}+36\,\*{{S}_{3}}\big{\}}\,\*{D_{-1}}\,\*{D_{0}}\,\*{D_{1}}\Big{)}$$

$$\mbox{}+32/9\,\*{\color[rgb]{0,0,1}{n_{\!f}}}\,\*{\color[rgb]{0,0,1}{C^{\,3}_{F}}}\,\*\Big{(}-10/3\,\*{\delta}(N-2)-1372/3\,\*{D_{-1}}-2233/2\,\*{D_{0}}+6049/4\,\*{D_{1}}-20/3\,\*{D_{2}}$$

$$\mbox{}-16\,\*{D_{-1}^{\,2}}+1477\,\*{D_{0}^{\,2}}+305\,\*{D_{1}^{\,2}}+32\,\*{D_{2}^{\,2}}-186\,\*{D_{0}^{\,3}}-87\,\*{D_{1}^{\,3}}+216\,\*({D_{0}^{\,4}}-{D_{1}^{\,4}})-\big{\{}30\,\*{D_{-1}}$$

$$\mbox{}-856\,\*{D_{0}}+725\,\*{D_{1}}-32\,\*{D_{2}}-96\,\*{D_{-1}^{\,2}}+288\,\*{D_{0}^{\,2}}+300\,\*{D_{1}^{\,2}}-108\,\*(2\,\*{D_{0}^{\,3}}-{D_{1}^{\,3}})\big{\}}\,\*{{S}_{1}}$$

$$\mbox{}+\big{\{}60\,\*{{S}_{-2}}-78\,\*{{S}_{1,1}}+72\,\*{{S}_{2}}\big{\}}\,\*p_{\rm{gq}}\Big{)}+256\,\*{\color[rgb]{0,0,1}{n_{\!f}}}\,\*{\color[rgb]{0,0,1}{d_{\,R}^{\,abcd}d_{\,R}^{\,abcd}\!/n_{c}}}\,\*\Big{(}\big{\{}-7/12-8\,\*{{S}_{1}}\,\*\eta^{2}+8\,\*{{S}_{-2}}\,\*\eta$$

$$\mbox{}-3\,\*\eta\big{\}}\,\*p_{\rm{gq}}+\big{\{}4-16\,\*{{S}_{-2}}\big{\}}\,\*{D_{-1}}\,\*{D_{0}}\,\*{D_{1}}\Big{)}+64/27\,\*{\color[rgb]{0,0,1}{n^{\,2}_{\!f}}}\,\*{\color[rgb]{0,0,1}{C^{\,2}_{F}}}\,\*\Big{(}52\,\*{D_{-1}}-89\,\*{D_{0}}+76\,\*{D_{1}}+4\,\*{D_{2}}\;$$

$$\mbox{}+12\,\*{D_{-1}^{\,2}}+42\,\*{D_{0}^{\,2}}+9\,\*{D_{1}^{\,2}}-6\,\*{{S}_{1}}\,\*p_{\rm{gq}}\Big{)}+32/9\,\*{\color[rgb]{0,0,1}{n^{\,2}_{\!f}}}\,\*{\color[rgb]{0,0,1}{C_{F}}}\*{\color[rgb]{0,0,1}{C_{A}}}\,\*\Big{(}{\delta}(N-2)/3-103/3\,\*{D_{-1}}+63\,\*{D_{0}}$$

$$\mbox{}-54\,\*{D_{1}}-8/3\,\*{D_{2}}-8\,\*{D_{-1}^{\,2}}-28\,\*{D_{0}^{\,2}}-12\,\*{D_{1}^{\,2}}+6\,\*{{S}_{1}}\,\*p_{\rm{gq}}\Big{)}-128/27\,\*{\color[rgb]{0,0,1}{n^{\,3}_{\!f}}}\*\,{\color[rgb]{0,0,1}{C_{F}}}\,\*p_{\rm{gq}}+\,\ldots$$ (14)

with

$$p_{\rm{gq}}\;=\;2{D_{-1}}-2{D_{0}}+{D_{1}}\;,\quad\eta\;=\;{D_{0}}-{D_{1}}\;,\quad\nu\;=\;{D_{-1}}-{D_{2}}\;.$$ (15)

As above, we have omitted also the argument $N$ of the sums for brevity. A very considerable effort would be required to find the contributions, indicated by ‘$+\>\ldots$’, that are still missing in eq. (S0.Ex17).

Also eq. (S0.Ex17) includes a structure that did not occur in the anomalous dimensions up to the third order in $\alpha_{\rm s}$ but was mentioned above: Both the ${\color[rgb]{0,0,0}{n_{\!f}}}{\color[rgb]{0,0,0}{C^{\,3}_{F}}}$ and ${\color[rgb]{0,0,0}{n^{\,2}_{\!f}}}C_{F}C_{A}$ terms receive a ${\delta}(N\!-\!2)$ contribution. In the latter case, which is fully known [10], this arises from the non-$\zeta$ contribution $(N\!-\!2)^{-1}\,S_{-2}(N\!-\!2)$, corresponding to $\,x^{\>\!-2}\,{\rm H}_{-1,0}(x)=\,x^{\>\!-2}\,[\,\ln(x)\ln(1+x)+\mbox{Li}_{2}(-x)]$ [45, 34]. Exactly the same function occurs in the two-loop coefficient functions for inclusive DIS [32, 33, 34] and in the ${\color[rgb]{0,0,0}{n^{\,2}_{\!f}}}C_{F}C_{A}$ contribution to $\gamma_{\,\rm ps}^{\,(3)}(N)$ recently determined in ref. [9].

The above partial all-$N$ results are not relevant to N³LO analyses of LHC processes. Until the complete functions $P_{\rm ik}^{\,(3)}(x)$ become known, such analyses will have to rely on approximations based on the available moments and information about the large-$x$ (threshold) and small-$x$ (high-energy) limits. With eqs. (9) we are now in the position to improve upon the $N\leq 10$ based approximations of ref. [12], thus putting $P_{\rm gq}^{\,(3)}$ on the same footing as $P_{\rm ps}^{\,(3)}$ and $P_{\rm qg}^{\,(3)}$ in refs. [13, 14].

Up to the presence of a leading-logarithmic BFKL small-$x$ contribution [46, 47], the endpoint structure of $P_{\rm gq}^{\,(3)}(x)$ is the same as that of $P_{\rm qg}^{\,(3)}(x)$ given in eqs. (10) and (11) of ref. [14]. So far the next-to-leading logarithmic (NLL) small-$x$ correction has been calculated [48, 49] and transformed to the $\overline{\mbox{MS}}$ scheme [50, 51] only for $P_{\rm gg}$. At order ${\alpha_{\rm s}^{\,3}}$, the NLL BFKL contributions to $P_{\rm gq}$ and $P_{\rm gg}$ are related by Casimir scaling in the large-${\color[rgb]{0,0,0}{n_{c}}}$ limit. In QCD the breaking of this relation is numerically small, see eq. (4.29) of ref. [31]. In view of the large uncertainties due to the unknown NNLL $x^{\,-1}\ln x$ terms, see below, no relevant bias should be generated by assuming Casimir scaling for the $x^{\,-1}\ln^{\,2}x$ term of $P_{\rm gq}^{\,(3)}$. Under this assumption the known endpoint contributions analogous to eq. (15) of ref. [14] are, to eight significant figures, given by [46, 47, 48, 49, 50, 51, 52, 40, 53, 54]

$$\!\!p_{{\rm gq},0}^{\,({\color[rgb]{0,0,0}{n_{\!f}}})}(x)\;=\;-3692.7188\,\*L_{0}^{3}/x-(47516.440+442.83691\,\*{\color[rgb]{0,0,0}{n_{\!f}}})\,\*L_{0}^{2}/x$$ (16)

$$\mbox{}+(52.235940-7.3744856\,\*{\color[rgb]{0,0,0}{n_{\!f}}})\,\*L_{0}^{6}-(292.21399-1.8436214\,\*{\color[rgb]{0,0,0}{n_{\!f}}})\,\*L_{0}^{5}$$

$$\mbox{}+(7310.6077-378.87135\,\*{\color[rgb]{0,0,0}{n_{\!f}}}-32.438957\,\*{\color[rgb]{0,0,0}{n^{\,2}_{\!f}}})\,\*L_{0}^{4}+(13.443073-0.54869684\,\*{\color[rgb]{0,0,0}{n_{\!f}}})\,\*L_{1}^{5}$$

$$\mbox{}+(375.39831-34.494742\,\,\*{\color[rgb]{0,0,0}{n_{\!f}}}+0.87791495\,\*{\color[rgb]{0,0,0}{n^{\,2}_{\!f}}})\,\*L_{1}^{4}+(22.222222-0.54869684\,\*{\color[rgb]{0,0,0}{n_{\!f}}})\,\*x_{1}\*L_{1}^{5}$$

$$\mbox{}+(662.42163-47.992684\,\*{\color[rgb]{0,0,0}{n_{\!f}}}+0.87791495\,\*{\color[rgb]{0,0,0}{n^{\,2}_{\!f}}})\,\*x_{1}\*L_{1}^{4}\;,$$

where we have used the abbreviations $x_{1}=1\!-\!x$, $L_{1}=\ln(1\!-\!x)$ and $L_{0}=\ln x$. The coefficient of $L_{0}^{\,2}/x$ in the first line of eq. (16) will be referred to as ${\mbox{bfkl}_{1}}$ below.

Our procedure for the construction of approximate expressions for $P_{\rm gq}^{\,(3)}(x)$ is identical to that used for its counterpart $P_{\rm qg}^{\,(3)}(x)$ in ref. [14]. Also here it is not possible, even with ten moments, to make a ‘direct fit’ for the coefficient of $x^{\,-1}\ln x$. In fact, knowing five additional moments on top of those employed in ref. [12] leads to considerable improvements elsewhere, see below, but does not enable us to assign a smaller uncertainty to this important coefficient.

The approximations selected to represent our error bands for the physically relevant numbers ${\color[rgb]{0,0,0}{n_{\!f}}}=3,\,4,\,5$ of light flavours read (with ${\mbox{bfkl}_{1}}$ as defined below eq. (16))

$$P_{\rm gq,\,A}^{\,(3)}({\color[rgb]{0,0,0}{n_{\!f}}}=3,x)\;=\;p_{{\rm gq},0}^{\,({\color[rgb]{0,0,0}{n_{\!f}}}=3)}(x)+6\,{\mbox{bfkl}_{1}}\,\*L_{0}/x-744384\,\*x_{1}/x+2453640-1540404\,x\*(2+x)$$

$$\mbox{{}}+1933026\,\*L_{0}+1142069\,\*L_{0}^{2}+162196\,\*L_{0}^{3}-2172.1\*L_{1}^{3}-93264.1\*L_{1}^{2}-786973\,\*L_{1}+875383\,\*x_{1}\*L_{1}^{2},$$

$$P_{\rm gq,\,B}^{\,(3)}({\color[rgb]{0,0,0}{n_{\!f}}}=3,x)\;=\;p_{{\rm gq},0}^{\,({\color[rgb]{0,0,0}{n_{\!f}}}=3)}(x)+3\,\*{\mbox{bfkl}_{1}}\,\*L_{0}/x+142414\,\*x_{1}/x-326525+2159787\,x\*(2-x)$$

$$\mbox{{}}-289064\,\*L_{0}-176358\,\*L_{0}^{2}+156541\,\*L_{0}^{3}+9016.5\,\*L_{1}^{3}+136063\,\*L_{1}^{2}+829482\,\*L_{1}-2359050\,\*L_{0}\*L_{1}\,,$$

$$P_{\rm gq,\,A}^{\,(3)}({\color[rgb]{0,0,0}{n_{\!f}}}=4,x)\;=\;p_{{\rm gq},0}^{\,({\color[rgb]{0,0,0}{n_{\!f}}}=4)}(x)+6\,{\mbox{bfkl}_{1}}\,\*L_{0}/x-743535\,\*x_{1}/x+2125286-1332472\,x\*(2+x)$$

$$\mbox{{}}+1631173\,\*L_{0}+1015255\,\*L_{0}^{2}+42612\,\*L_{0}^{3}-1910.4\,\*L_{1}^{3}-80851\,\*L_{1}^{2}-680219\,\*L_{1}+752733\,\*x_{1}\*L_{1}^{2}\,,$$

$$P_{\rm gq,\,B}^{\,(3)}({\color[rgb]{0,0,0}{n_{\!f}}}=4,x)\;=\;p_{{\rm gq},0}^{\,({\color[rgb]{0,0,0}{n_{\!f}}}=4)}(x)+3\,\*{\mbox{bfkl}_{1}}\,\*L_{0}/x+160568\,\*x_{1}/x-361207+2048948\,x\*(2-x)$$

$$\mbox{{}}-245963\,\*L_{0}-171312\,\*L_{0}^{2}+163099\,\*L_{0}^{3}+8132.2\,\*L_{1}^{3}+124425\,\*L_{1}^{2}+762435\,\*L_{1}-2193335\,\*L_{0}\*L_{1}\,,$$

$$P_{\rm gq,\,A}^{\,(3)}({\color[rgb]{0,0,0}{n_{\!f}}}=5,x)\;=\;p_{{\rm gq},0}^{\,({\color[rgb]{0,0,0}{n_{\!f}}}=5)}(x)+6\,{\mbox{bfkl}_{1}}\,\*L_{0}/x-785864\,\*x_{1}/x+285034-131648\,x\*(2+x)$$

$$\mbox{{}}-162840\,\*L_{0}+321220\,\*L_{0}^{2}+12688\,\*L_{0}^{3}+1423.4\,\*L_{1}^{3}+1278.9\,\*L_{1}^{2}-30919.9\,\*L_{1}+47588\,\*x_{1}\*L_{1}^{2}\,,$$

$$P_{\rm gq,\,B}^{\,(3)}({\color[rgb]{0,0,0}{n_{\!f}}}=5,x)\;=\;p_{{\rm gq},0}^{\,({\color[rgb]{0,0,0}{n_{\!f}}}=5)}(x)+3\,\*{\mbox{bfkl}_{1}}\,\*L_{0}/x+177094\,\*x_{1}/x-470694+1348823\,x\*(2-x)$$

$$\mbox{{}}-52985\,\*L_{0}-87354\,\*L_{0}^{2}+176885\,\*L_{0}^{3}+4748.8\,\*L_{1}^{3}+65811.9\,\*L_{1}^{2}+396390\,\*L_{1}-1190212\,\*L_{0}\*L_{1}\,.$$

At $N=22$ we predict, with the brackets indicating a conservative uncertainty of the last digit:

$$-\gamma_{\,\rm gq}^{\,(3)}(N\!=\!22)\;=\;662.7981(3)\,,\>\>549.2876(3)\,,\>\>426.6266(2)\quad\mbox{for}\quad{\color[rgb]{0,0,0}{n_{\!f}}}\,=\,3,4,5\;.$$ (20)

The selection procedure and its results are illustrated in fig. 1 for ${\color[rgb]{0,0,0}{n_{\!f}}}=4$; the two functions in eq. (S0.Ex39) are shown by the red (solid) lines. A comparison with the corresponding $N\!\leq\!10$ results of ref. [12], shown by the blue (dashed) lines, reveals an impressive improvement down to $x\approx 10^{-2}$. $P_{\rm gq}^{\,(3)}(x)$ can now be considered as sufficiently well constrained at $x$-values above 0.1.

Figure 1: Two sets of 80 trial functions for the four-loop (N³LO) contribution to the quark-to-gluon splitting function at ${\color[rgb]{0,0,0}{n_{\!f}}}=4$. The two cases selected for eq. (S0.Ex35) are shown by the solid (red) lines. Also shown, by the dashed (blue) lines, are the selected approximations of ref. [12] based on only the first 5 even moments.

Figure 2: Left: the perturbative expansion of the splitting functions $P_{\rm gq}$ to N³LO for ${\color[rgb]{0,0,0}{n_{\!f}}}=4$ and $\alpha_{\rm s}=0.2$, using eq. (S0.Ex35) for the four-loop contribution. Right: the resulting N²LO and N³LO convolutions with the reference quark distribution (21) normalized to the NLO results.

The uncertainties of $P_{\rm gq}^{\,(3)}(x)$ ‘as such’ are still rather large at $10^{-2}\raisebox{-1.99168pt}{$\>\>\stackrel{{\scriptstyle<}}{{{\scriptstyle\sim}}}\>\>$}x<10^{-1}$. However, as shown in the left panel of fig. 2, the resulting uncertainties of the expansion of the splitting function $P_{\rm gq}$ to N³LO appear to be perfectly tolerable also in this range, down to (for LHC purposes) rather low scales. Of course, the splitting functions enter physical quantities only via the convolution with the respective quark or gluon PDF. This convolution is illustrated in the right panel of fig. 2 using a sufficiently realistic order-independent model input [31] for the singlet quark PDF in eq. (1),

$$xq_{\rm s}(x,\mu_{0}^{\,2})\;=\;0.6\,x^{\,-0.3}(1-x)^{3.5}\left(1+5.0\,x^{\,0.8}\right)\,,$$ (21)

together with a strong coupling of $\alpha_{\rm s}(\mu_{0}^{\,2})\,=\,0.2$ corresponding to a scale $\mu_{0}^{\,2}\,\approx\,30\ldots 50$ GeV². Here the uncertainties of $P_{\rm gq}^{\,(3)}(x)$ do not appear to have a relevant impact at $x\raisebox{-1.99168pt}{$\>\>\stackrel{{\scriptstyle>}}{{{\scriptstyle\sim}}}\>\>$}10^{-3}$. An overall uncertainty of $\pm 1\%$ of the NLO result for $P_{\rm gq}\otimes q_{\rm s}$ is reached only at $x\approx 10^{-4}$.

To summarize, we have computed the even-$N$ moments $N\leq 20$ of the fourth-order (N³LO) quark-to-gluon splitting function $P_{\rm gq}^{\,(3)}(x)$ in the framework of the operator-product expansion. These results facilitate the determination of the all-$N$ form of the ${\zeta_{5}}$ contribution and of parts of the corresponding ${\zeta_{3}}$ terms. These expressions show deviations from the functional form of the anomalous dimensions up to three loops which should be of interest to future research.

Together with (approximately) known endpoint constraints at $x\rightarrow 1$ and $x\rightarrow 0$, our additional moments – only the results up to $N=10$ had been obtained before [11, 12] – have been employed to construct improved approximations for $P_{\rm gq}^{\,(3)}(x)$ that should be sufficiently accurate for a wide range of phenomenological applications in collider physics. Their uncertainties are still large at $x\raisebox{-1.99168pt}{$\>\>\stackrel{{\scriptstyle<}}{{{\scriptstyle\sim}}}\>\>$}10^{-3}$. A drastic improvement in this region would be obtained if the next-to-leading and next-to-next-leading small-$x$ contributions, $x^{-1}\ln^{2}x$ and $x^{-1}\ln x$, became known.

Appendix A Mellin moments of $P^{\,(3)}_{\,\rm gq}$

The exact results for the four-loop anomalous dimensions $\gamma^{\,(3)}_{\,\rm gq}(N)$ at even $2\leq N\leq 10$ have already been presented, for a general compact simple gauge group, in eqs. (11) - (13) of ref. [11] and eqs. (9) and (10) of ref. [12]. We agree with the results given in those articles, which were obtained in a different theoretical framework. Here we report the corresponding results at $12\leq N\leq 20$. The numerical values for QCD, i.e., SU(${\color[rgb]{0,0,0}{n_{c}}}=3$), have been given in eqs. (9) above.

The quadratic Casimir invariants are ${\color[rgb]{0,0,0}{C_{A}}}={\color[rgb]{0,0,0}{n_{c}}}$ and ${\color[rgb]{0,0,0}{C_{F}}}=({\color[rgb]{0,0,0}{n_{c}^{\,2}}}-1)/(2{\color[rgb]{0,0,0}{n_{c}}})$ in SU$({\color[rgb]{0,0,0}{n_{c}}})$. The quartic group invariants are products of two symmetrized traces of four generators $T_{r}^{a}$ of the fundamental ($R$) or adjoint ($A$) representation,

$$d_{r}^{\,abcd}\;=\;\frac{1}{6}\>{\rm Tr}\,(\,T_{r}^{a\,}T_{r}^{b\,}T_{r}^{c\,}T_{r}^{d\,}+\,\mbox{ five $bcd$ permutations}\,)\;,$$ (A.1)

which leads to

$${\color[rgb]{0,0,0}{{d_{RA}^{(4)}\over n_{c}}}}\;=\;\frac{({\color[rgb]{0,0,0}{n_{c}^{\,2}}}-1)\,({\color[rgb]{0,0,0}{n_{c}^{\,2}}}+6)}{48}\;\,,\quad{\color[rgb]{0,0,0}{{d_{RR}^{(4)}\over n_{c}}}}\;=\;\frac{({\color[rgb]{0,0,0}{n_{c}^{\,2}}}-1)\,({\color[rgb]{0,0,0}{n_{c}^{\,4}}}-6\,{\color[rgb]{0,0,0}{n_{c}^{\,2}}}+18)}{96\,{\color[rgb]{0,0,0}{n_{c}^{\,3}}}}\;.$$ (A.2)

Their values in QCD are ${\color[rgb]{0,0,0}{{d_{RA}^{(4)}\over n_{c}}}}\equiv{\color[rgb]{0,0,0}{d_{\,R}^{\,abcd}d_{\,A}^{\,abcd}\!/n_{c}}}\,=\,5/2$ and ${\color[rgb]{0,0,0}{{d_{RR}^{(4)}\over n_{c}}}}\equiv{\color[rgb]{0,0,0}{d_{\,R}^{\,abcd}d_{\,R}^{\,abcd}\!/n_{c}}}\,=\,5/36$.

$${}\gamma_{\,\rm gq}^{\,(3)}(N=12)\,=\,{\color[rgb]{0,0,1}{C^{\,4}_{F}}}\,\*\Big{(}-\frac{25682557611275387813242057003995677}{135916765937789612439776112000000}-\frac{78873286642153277627}{313680161535741000}\,\*{\zeta_{3}}$$

$$\mbox{{\hskip 14.22636pt}}+\frac{12335483}{193050}\,\*{\zeta_{4}}+\frac{17406716}{42471}\,\*{\zeta_{5}}\Big{)}+{\color[rgb]{0,0,1}{C_{A}}}\,\*{\color[rgb]{0,0,1}{C^{\,3}_{F}}}\,\*\Big{(}\frac{422699180143530974480654228020604813}{1143819017242696998194479488000000}$$

$$\mbox{{\hskip 14.22636pt}}+\frac{100064481514290063961}{313680161535741000}\,\*{\zeta_{3}}-\frac{760772056049}{9591882300}\,\*{\zeta_{4}}-\frac{237624788}{351351}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{C^{\,2}_{A}}}\,\*{\color[rgb]{0,0,1}{C^{\,2}_{F}}}\,\*\Big{(}-\frac{224250321575347263949905289161744199}{1455769658308887088611155712000000}-\frac{2010183037743051616}{39210020191967625}\,\*{\zeta_{3}}$$

$$\mbox{{\hskip 14.22636pt}}+\frac{3810092749}{532882350}\,\*{\zeta_{4}}+\frac{1211419205}{3864861}\,\*{\zeta_{5}}\Big{)}+{\color[rgb]{0,0,1}{C^{\,3}_{A}}}\,\*{\color[rgb]{0,0,1}{C_{F}}}\,\*\Big{(}-\frac{605797324486893500585306535221627}{18906099458556975176768256000000}$$

$$\mbox{{\hskip 14.22636pt}}-\frac{1052537460264167687}{25347891841272000}\,\*{\zeta_{3}}+\frac{79289578229}{9591882300}\,\*{\zeta_{4}}-\frac{23156863}{1054053}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{{d_{RA}^{(4)}\over n_{c}}}}\,\*\Big{(}-\frac{113527295175322043699}{2082683928124800000}-\frac{66259676094769}{321401840760}\,\*{\zeta_{3}}+\frac{1159790344}{3864861}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{n_{\!f}}}\,\*{\color[rgb]{0,0,1}{C^{\,3}_{F}}}\,\*\Big{(}-\frac{332043482084304364183967274124153}{19063650287378283303241324800000}-\frac{393343213522261}{33948069430275}\,\*{\zeta_{3}}-\frac{6070628}{495495}\,\*{\zeta_{4}}$$

$$\mbox{{\hskip 14.22636pt}}+\frac{12640}{429}\,\*{\zeta_{5}}\Big{)}+{\color[rgb]{0,0,1}{n_{\!f}}}\,\*{\color[rgb]{0,0,1}{C_{A}}}\,\*{\color[rgb]{0,0,1}{C^{\,2}_{F}}}\,\*\Big{(}\frac{2936117949737136383600864384611}{157550828821308126473068800000}-\frac{3457752318647119}{67896138860550}\,\*{\zeta_{3}}$$

$$\mbox{{\hskip 14.22636pt}}+\frac{8619754063}{251215965}\,\*{\zeta_{4}}-\frac{6320}{1287}\,\*{\zeta_{5}}\Big{)}+{\color[rgb]{0,0,1}{n_{\!f}}}\,\*{\color[rgb]{0,0,1}{C^{\,2}_{A}}}\,\*{\color[rgb]{0,0,1}{C_{F}}}\,\*\Big{(}-\frac{3409052330654655340645389731}{1469005396935273906508800000}$$

$$\mbox{{\hskip 14.22636pt}}+\frac{25514383657902523}{407376833163300}\,\*{\zeta_{3}}-\frac{5541945667}{251215965}\,\*{\zeta_{4}}-\frac{119360}{5577}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{n_{\!f}}}\,\*{\color[rgb]{0,0,1}{{d_{RR}^{(4)}\over n_{c}}}}\,\*\Big{(}-\frac{14322664324006372519}{260335491015600000}-\frac{33040603052}{2608781175}\,\*{\zeta_{3}}+\frac{421760}{5577}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{n^{\,2}_{\!f}}}\,\*{\color[rgb]{0,0,1}{C^{\,2}_{F}}}\,\*\Big{(}\frac{169033294355320918596314039}{79352523673735777985520000}+\frac{81408784}{15810795}\,\*{\zeta_{3}}-\frac{5056}{1287}\,\*{\zeta_{4}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{n^{\,2}_{\!f}}}\,\*{\color[rgb]{0,0,1}{C_{A}}}\,\*{\color[rgb]{0,0,1}{C_{F}}}\,\*\Big{(}-\frac{1990546832712974636980331}{1109825505926374517280000}-\frac{171649172}{57972915}\,\*{\zeta_{3}}+\frac{5056}{1287}\,\*{\zeta_{4}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{n^{\,3}_{\!f}}}\,\*{\color[rgb]{0,0,1}{C_{F}}}\,\*\Big{(}\frac{2265689177355445577}{3387745744586002800}-\frac{5056}{11583}\,\*{\zeta_{3}}\Big{)}\;,$$ (A.3)

$${}\gamma_{\,\rm gq}^{\,(3)}(N=14)\,=\,{\color[rgb]{0,0,1}{C^{\,4}_{F}}}\,\*\Big{(}-\frac{1175858514183978179181088212329622671}{5838242900509599261617655720000000}$$

$$\mbox{{\hskip 14.22636pt}}-\frac{32053546295279752556}{124759155156260625}\,\*{\zeta_{3}}+\frac{10403997892}{165540375}\,\*{\zeta_{4}}+\frac{5234629744}{12297285}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{C_{A}}}\,\*{\color[rgb]{0,0,1}{C^{\,3}_{F}}}\,\*\Big{(}\frac{435029240281577148104394100536180709}{1077829458555618325221721056000000}+\frac{172973892630128742931}{499036620625042500}\,\*{\zeta_{3}}$$

$$\mbox{{\hskip 14.22636pt}}-\frac{5900838170593}{71016820875}\,\*{\zeta_{4}}-\frac{160919504}{223587}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{C^{\,2}_{A}}}\,\*{\color[rgb]{0,0,1}{C^{\,2}_{F}}}\,\*\Big{(}-\frac{62684886023029557261923348044015457}{331639833401728715452837248000000}-\frac{173609406082076933}{2268348275568375}\,\*{\zeta_{3}}$$

$$\mbox{{\hskip 14.22636pt}}+\frac{231212408549}{15781515750}\,\*{\zeta_{4}}+\frac{4199718508}{12297285}\,\*{\zeta_{5}}\Big{)}+{\color[rgb]{0,0,1}{C^{\,3}_{A}}}\,\*{\color[rgb]{0,0,1}{C_{F}}}\,\*\Big{(}-\frac{4272807969344509309643357544788243}{255107564155175934963720960000000}$$

$$\mbox{{\hskip 14.22636pt}}-\frac{926544367976770183}{25591621570515000}\,\*{\zeta_{3}}+\frac{794134472909}{142033641750}\,\*{\zeta_{4}}-\frac{177252716}{7378371}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{{d_{RA}^{(4)}\over n_{c}}}}\,\*\Big{(}-\frac{21335213154567939744419}{414584269442343000000}-\frac{613323018929398364}{3198952696314375}\,\*{\zeta_{3}}+\frac{3460252064}{12297285}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{n_{\!f}}}\,\*{\color[rgb]{0,0,1}{C^{\,3}_{F}}}\,\*\Big{(}-\frac{490379446041031040879323234108051}{40826873430137057773550040000000}-\frac{33894392681638}{2403420508125}\,\*{\zeta_{3}}-\frac{46632848}{4099095}\,\*{\zeta_{4}}$$

$$\mbox{{\hskip 14.22636pt}}+\frac{6784}{273}\,\*{\zeta_{5}}\Big{)}+{\color[rgb]{0,0,1}{n_{\!f}}}\,\*{\color[rgb]{0,0,1}{C_{A}}}\,\*{\color[rgb]{0,0,1}{C^{\,2}_{F}}}\,\*\Big{(}\frac{40759120798908335841181047497491}{3140528725395158290273080000000}-\frac{992731197367267}{26437625589375}\,\*{\zeta_{3}}$$

$$\mbox{{\hskip 14.22636pt}}+\frac{65246478712}{2152024875}\,\*{\zeta_{4}}-\frac{3392}{819}\,\*{\zeta_{5}}\Big{)}+{\color[rgb]{0,0,1}{n_{\!f}}}\,\*{\color[rgb]{0,0,1}{C^{\,2}_{A}}}\,\*{\color[rgb]{0,0,1}{C_{F}}}\,\*\Big{(}-\frac{958430136662854610670607569547}{322105510296939311822880000000}$$

$$\mbox{{\hskip 14.22636pt}}+\frac{9023658476762282}{174488328889875}\,\*{\zeta_{3}}-\frac{40764233512}{2152024875}\,\*{\zeta_{4}}-\frac{4416}{245}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{n_{\!f}}}\,\*{\color[rgb]{0,0,1}{{d_{RR}^{(4)}\over n_{c}}}}\,\*\Big{(}-\frac{86703311715954607957}{1850822631439031250}-\frac{244011856707712}{22370298575625}\,\*{\zeta_{3}}+\frac{615424}{9555}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{n^{\,2}_{\!f}}}\,\*{\color[rgb]{0,0,1}{C^{\,2}_{F}}}\,\*\Big{(}\frac{649724993957136005232607141}{339884061189952195917000000}+\frac{178574416}{42567525}\,\*{\zeta_{3}}-\frac{13568}{4095}\,\*{\zeta_{4}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{n^{\,2}_{\!f}}}\,\*{\color[rgb]{0,0,1}{C_{A}}}\,\*{\color[rgb]{0,0,1}{C_{F}}}\,\*\Big{(}-\frac{2493111180935243429476603}{1493995873362427234800000}-\frac{83451856}{36891855}\,\*{\zeta_{3}}+\frac{13568}{4095}\,\*{\zeta_{4}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{n^{\,3}_{\!f}}}\,\*{\color[rgb]{0,0,1}{C_{F}}}\,\*\Big{(}\frac{321519364624645823}{538959550275045900}-\frac{13568}{36855}\,\*{\zeta_{3}}\Big{)}\;,$$ (A.4)

$${}\gamma_{\,\rm gq}^{\,(3)}(N=16)\,=\,{\color[rgb]{0,0,1}{C^{\,4}_{F}}}\,\*\Big{(}-\frac{91333794681488605852642048633486608590877607351}{431323535597978497478425306351127101440000000}$$

$$\mbox{{\hskip 14.22636pt}}-\frac{15461805996412648565848919}{58626866062381256640000}\,\*{\zeta_{3}}+\frac{105621100681}{1715313600}\,\*{\zeta_{4}}+\frac{2466482078}{5579145}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{C_{A}}}\,\*{\color[rgb]{0,0,1}{C^{\,3}_{F}}}\,\*\Big{(}\frac{69709103720766489723595997659712410526245706881}{161746325849241936554409489881672663040000000}$$

$$\mbox{{\hskip 14.22636pt}}+\frac{561806270069913957735019}{1503252975958493760000}\,\*{\zeta_{3}}-\frac{181207395103216571}{2126662954416000}\,\*{\zeta_{4}}-\frac{2706171203}{3550365}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{C^{\,2}_{A}}}\,\*{\color[rgb]{0,0,1}{C^{\,2}_{F}}}\,\*\Big{(}-\frac{2905887663991964454299544802600912038519944459}{13478860487436828046200790823472721920000000}$$

$$\mbox{{\hskip 14.22636pt}}-\frac{14062285015933017481879}{142125735908803046400}\,\*{\zeta_{3}}+\frac{14216141064343763}{708887651472000}\,\*{\zeta_{4}}+\frac{93724168}{255255}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{C^{\,3}_{A}}}\,\*{\color[rgb]{0,0,1}{C_{F}}}\,\*\Big{(}-\frac{317835545673384782764570229718975673012928297}{53915441949747312184803163293890887680000000}$$

$$\mbox{{\hskip 14.22636pt}}-\frac{1262927249130586239359921}{39084577374920837760000}\,\*{\zeta_{3}}+\frac{475554692179667}{132916434651000}\,\*{\zeta_{4}}-\frac{23407688153}{937296360}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{{d_{RA}^{(4)}\over n_{c}}}}\,\*\Big{(}-\frac{11773700142311179915702121}{240974431125652224000000}-\frac{101169081992668980737}{563503134009816000}\,\*{\zeta_{3}}+\frac{4137449651}{15621606}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{n_{\!f}}}\,\*{\color[rgb]{0,0,1}{C^{\,3}_{F}}}\,\*\Big{(}-\frac{57298339123449136718474401962230424513361}{7181544049250391233407014757761024000000}-\frac{87782813700795439}{5546240318619000}\,\*{\zeta_{3}}$$

$$\mbox{{\hskip 14.22636pt}}-\frac{16614041449}{1562160600}\,\*{\zeta_{4}}+\frac{1096}{51}\,\*{\zeta_{5}}\Big{)}+{\color[rgb]{0,0,1}{n_{\!f}}}\,\*{\color[rgb]{0,0,1}{C_{A}}}\,\*{\color[rgb]{0,0,1}{C^{\,2}_{F}}}\,\*\Big{(}\frac{93344344164963857590569964407611700892711}{10772316073875586850110522136641536000000}$$

$$\mbox{{\hskip 14.22636pt}}-\frac{2475119092628220127}{88739845097904000}\,\*{\zeta_{3}}+\frac{643872057559}{23605982400}\,\*{\zeta_{4}}-\frac{548}{153}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{n_{\!f}}}\,\*{\color[rgb]{0,0,1}{C^{\,2}_{A}}}\,\*{\color[rgb]{0,0,1}{C_{F}}}\,\*\Big{(}-\frac{2033308368213027173938918062479854873193}{633665651404446285300618949214208000000}$$

$$\mbox{{\hskip 14.22636pt}}+\frac{15529936726131895757}{354959380391616000}\,\*{\zeta_{3}}-\frac{3535338880967}{212453841600}\,\*{\zeta_{4}}-\frac{121513}{7803}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{n_{\!f}}}\,\*{\color[rgb]{0,0,1}{{d_{RR}^{(4)}\over n_{c}}}}\,\*\Big{(}-\frac{405223386945234740808587}{9940195283933154240000}-\frac{14956670451523}{1563722760600}\,\*{\zeta_{3}}+\frac{145816}{2601}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{n^{\,2}_{\!f}}}\,\*{\color[rgb]{0,0,1}{C^{\,2}_{F}}}\,\*\Big{(}\frac{9044520444067665064750987439226131}{5275271823213838538966191718400000}+\frac{4119326989}{1171620450}\,\*{\zeta_{3}}-\frac{2192}{765}\,\*{\zeta_{4}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{n^{\,2}_{\!f}}}\,\*{\color[rgb]{0,0,1}{C_{A}}}\,\*{\color[rgb]{0,0,1}{C_{F}}}\,\*\Big{(}-\frac{60504685061989103630936679672439}{38788763405984106904163174400000}-\frac{3125731858}{1757430675}\,\*{\zeta_{3}}+\frac{2192}{765}\,\*{\zeta_{4}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{n^{\,3}_{\!f}}}\,\*{\color[rgb]{0,0,1}{C_{F}}}\,\*\Big{(}\frac{34108337373135323374301}{63317015347371757171200}-\frac{2192}{6885}\,\*{\zeta_{3}}\Big{)}\;,$$ (A.5)

$${}\gamma_{\,\rm gq}^{\,(3)}(N=18)\,=\,{\color[rgb]{0,0,1}{C^{\,4}_{F}}}\,\*\Big{(}-\frac{199441346676346613221100951050417553229794351172058063}{903627934773265885372610492741460297127036416000000}$$

$$\mbox{{\hskip 14.22636pt}}-\frac{9734182701413671124997104021}{35813961619291817562100500}\,\*{\zeta_{3}}+\frac{55191569546}{916620705}\,\*{\zeta_{4}}+\frac{555009226736}{1208442807}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{C_{A}}}\,\*{\color[rgb]{0,0,1}{C^{\,3}_{F}}}\,\*\Big{(}\frac{1703079039178221518412667538347685102671198180332127}{3752088310477228589782465713805371475960012800000}$$

$$\mbox{{\hskip 14.22636pt}}+\frac{58151956931435413049516329}{144995796029521528591500}\,\*{\zeta_{3}}-\frac{208211747676885206}{2413254243364965}\,\*{\zeta_{4}}-\frac{618848170736}{769009059}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{C^{\,2}_{A}}}\,\*{\color[rgb]{0,0,1}{C^{\,2}_{F}}}\,\*\Big{(}-\frac{123319626596597732563913557669030322668202344255007}{521123376455170637469786904695190482772224000000}$$

$$\mbox{{\hskip 14.22636pt}}-\frac{97421341297611056710882606}{813953673165723126411375}\,\*{\zeta_{3}}+\frac{7737412238766461}{321767232448662}\,\*{\zeta_{4}}+\frac{1104257588788}{2819699883}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{C^{\,3}_{A}}}\,\*{\color[rgb]{0,0,1}{C_{F}}}\,\*\Big{(}\frac{115832786415748476638280610626833293796768491507}{55177769271723949849742142850078992293529600000}$$

$$\mbox{{\hskip 14.22636pt}}-\frac{395558870241334146068201903}{13482903197851037199849600}\,\*{\zeta_{3}}+\frac{9748563515611381}{4826508486729930}\,\*{\zeta_{4}}-\frac{214104770108}{8459099649}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{{d_{RA}^{(4)}\over n_{c}}}}\,\*\Big{(}-\frac{1534254855736463458285860487597}{32941608367048794488275200000}-\frac{263063336077755987407089}{1556198430038208356400}\,\*{\zeta_{3}}$$

$$\mbox{{\hskip 14.22636pt}}+\frac{705610821248}{2819699883}\,\*{\zeta_{5}}\Big{)}+{\color[rgb]{0,0,1}{n_{\!f}}}\,\*{\color[rgb]{0,0,1}{C^{\,3}_{F}}}\,\*\Big{(}-\frac{154871348508574160288273783965897373976839491}{31860889749307135877273684882463194351308800}$$

$$\mbox{{\hskip 14.22636pt}}-\frac{18100451697075180817}{1062928800827568675}\,\*{\zeta_{3}}-\frac{84572797384}{8459099649}\,\*{\zeta_{4}}+\frac{55040}{2907}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{n_{\!f}}}\,\*{\color[rgb]{0,0,1}{C_{A}}}\,\*{\color[rgb]{0,0,1}{C^{\,2}_{F}}}\,\*\Big{(}\frac{494213149462187941188004051358795571999087887}{93708499262668046697863779066068218680320000}$$

$$\mbox{{\hskip 14.22636pt}}-\frac{197585480953102482619}{9566359207448118075}\,\*{\zeta_{3}}+\frac{59943194135296}{2410843399965}\,\*{\zeta_{4}}-\frac{27520}{8721}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{n_{\!f}}}\,\*{\color[rgb]{0,0,1}{C^{\,2}_{A}}}\,\*{\color[rgb]{0,0,1}{C_{F}}}\,\*\Big{(}-\frac{936013096859225678199723846435281783145737}{290119192763678163151281049740149283840000}$$

$$\mbox{{\hskip 14.22636pt}}+\frac{122057786649260319779}{3237844654828593810}\,\*{\zeta_{3}}-\frac{35839946880856}{2410843399965}\,\*{\zeta_{4}}-\frac{20446720}{1491291}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{n_{\!f}}}\,\*{\color[rgb]{0,0,1}{{d_{RR}^{(4)}\over n_{c}}}}\,\*\Big{(}-\frac{2376495270782792811304983137}{65883216734097588976550400}-\frac{9726278907372152}{1143120431067615}\,\*{\zeta_{3}}+\frac{24663040}{497097}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{n^{\,2}_{\!f}}}\,\*{\color[rgb]{0,0,1}{C^{\,2}_{F}}}\,\*\Big{(}\frac{2076988113916115506387549621182589219}{1350630257073663390768162168309360000}+\frac{13478091680}{4478346873}\,\*{\zeta_{3}}-\frac{22016}{8721}\,\*{\zeta_{4}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{n^{\,2}_{\!f}}}\,\*{\color[rgb]{0,0,1}{C_{A}}}\,\*{\color[rgb]{0,0,1}{C_{F}}}\,\*\Big{(}-\frac{51603554118665492025456278063355347}{35310594956174206294592474988480000}-\frac{543557988224}{380659484205}\,\*{\zeta_{3}}+\frac{22016}{8721}\,\*{\zeta_{4}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{n^{\,3}_{\!f}}}\,\*{\color[rgb]{0,0,1}{C_{F}}}\,\*\Big{(}\frac{25322735830563737715837587}{51572104731780217289424720}-\frac{22016}{78489}\,\*{\zeta_{3}}\Big{)}\;,$$ (A.6)

$${}\gamma_{\,\rm gq}^{\,(3)}(N=20)\,=\,{\color[rgb]{0,0,1}{C^{\,4}_{F}}}\,\*\Big{(}-\frac{18011842890292937504270512411100571563413211395232599}{78747532442114674106545576709495450730024960000000}$$

$$\mbox{{\hskip 14.22636pt}}-\frac{27623613597125033101695488051}{98312835817663812915570000}\,\*{\zeta_{3}}+\frac{57192203098}{972173475}\,\*{\zeta_{4}}+\frac{923526447854}{1935088155}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{C_{A}}}\,\*{\color[rgb]{0,0,1}{C^{\,3}_{F}}}\,\*\Big{(}\frac{742257701799954265313171119485669995895179080771700597}{1566661434901018253277590947167856861892075520000000}$$

$$\mbox{{\hskip 14.22636pt}}+\frac{304638383107094264933207}{710296405760118870000}\,\*{\zeta_{3}}-\frac{570786749229593761}{6585879026727000}\,\*{\zeta_{4}}-\frac{149116082954}{175917105}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{C^{\,2}_{A}}}\,\*{\color[rgb]{0,0,1}{C^{\,2}_{F}}}\,\*\Big{(}-\frac{13595171979898487398112366060650503619924613365949}{53542769477136645703266949663973235197952000000}$$

$$\mbox{{\hskip 14.22636pt}}-\frac{13706603212061214884924310847}{98312835817663812915570000}\,\*{\zeta_{3}}+\frac{1068938712970319969}{39515274160362000}\,\*{\zeta_{4}}+\frac{178657350101}{430019590}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{C^{\,3}_{A}}}\,\*{\color[rgb]{0,0,1}{C_{F}}}\,\*\Big{(}\frac{8128052126051322869747930125993834197558195023}{996276944019023178954993747070209054720000000}$$

$$\mbox{{\hskip 14.22636pt}}-\frac{4468686144075569343986511569}{165579512956065369120960000}\,\*{\zeta_{3}}+\frac{494114186999099}{627226573974000}\,\*{\zeta_{4}}-\frac{10876666809}{430019590}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{{d_{RA}^{(4)}\over n_{c}}}}\,\*\Big{(}-\frac{9934247787263135793881744121471193}{223044635561617655771521536000000}-\frac{20055818840172257506705213}{125439024966716188728000}\,\*{\zeta_{3}}$$

$$\mbox{{\hskip 14.22636pt}}+\frac{153050440696}{645029385}\,\*{\zeta_{5}}\Big{)}+{\color[rgb]{0,0,1}{n_{\!f}}}\,\*{\color[rgb]{0,0,1}{C^{\,3}_{F}}}\,\*\Big{(}-\frac{941248780576101712647404838543624320767081907}{394884456817228610955910317701869858560000000}$$

$$\mbox{{\hskip 14.22636pt}}-\frac{6339794916091394356451}{354748373774649855000}\,\*{\zeta_{3}}-\frac{91361957041}{9675440775}\,\*{\zeta_{4}}+\frac{6752}{399}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{n_{\!f}}}\,\*{\color[rgb]{0,0,1}{C_{A}}}\,\*{\color[rgb]{0,0,1}{C^{\,2}_{F}}}\,\*\Big{(}\frac{102664753783997698300409838737001288852130553}{40379162501611346684213385870266391552000000}$$

$$\mbox{{\hskip 14.22636pt}}-\frac{5347393544742883698797}{354748373774649855000}\,\*{\zeta_{3}}+\frac{2584501408063}{112880142375}\,\*{\zeta_{4}}-\frac{3376}{1197}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{n_{\!f}}}\,\*{\color[rgb]{0,0,1}{C^{\,2}_{A}}}\,\*{\color[rgb]{0,0,1}{C_{F}}}\,\*\Big{(}-\frac{465634445017831591325444801806925052804292331}{148765335532252329889207211100981442560000000}$$

$$\mbox{{\hskip 14.22636pt}}+\frac{3171458341939781449}{96255155006010000}\,\*{\zeta_{3}}-\frac{4555835727754}{338640427125}\,\*{\zeta_{4}}-\frac{1539424}{125685}\,\*{\zeta_{5}}\Big{)}$$

$$\mbox{{\hskip 14.22636pt}}+{\color[rgb]{0,0,1}{n_{\!f}}}\,\*{\color[rgb]{0,0,1}{{d_{RR}^{(4)}\over n_{c}}}}\,\*\Big{(}-\frac{2704920902087866223629472523102433}{83641738335606620914320576000000}-\frac{108318373596246439}{14144672000584125}\,\*{\zeta_{3}}$$

$$\mbox{{\hskip 14.22636pt}}+\frac{1863808}{41895}\,\*{\zeta_{5}}\Big{)}+{\color[rgb]{0,0,1}{n^{\,2}_{\!f}}}\,\*{\color[rgb]{0,0,1}{C^{\,2}_{F}}}\,\*\Big{(}\frac{4773708126683882394527205560488250927}{3460438305705072216870062548871040000}+\frac{4000788628}{1527701175}\,\*{\zeta_{3}}$$

$$\mbox{{\hskip 14.22636pt}}-\frac{13504}{5985}\,\*{\zeta_{4}}\Big{)}+{\color[rgb]{0,0,1}{n^{\,2}_{\!f}}}\,\*{\color[rgb]{0,0,1}{C_{A}}}\,\*{\color[rgb]{0,0,1}{C_{F}}}\,\*\Big{(}-\frac{11325534673549584438954822499666967}{8259788293841919601074263155200000}-\frac{11266215274}{9675440775}\,\*{\zeta_{3}}$$

$$\mbox{{\hskip 14.22636pt}}+\frac{13504}{5985}\,\*{\zeta_{4}}\Big{)}+{\color[rgb]{0,0,1}{n^{\,3}_{\!f}}}\,\*{\color[rgb]{0,0,1}{C_{F}}}\,\*\Big{(}\frac{23642031766087097685137941}{52433512436087366888304000}-\frac{13504}{53865}\,\*{\zeta_{3}}\Big{)}\;.$$ (A.7)

Form files with the results for $\gamma_{\,\rm gq}(N)$ at even $N\leq 20$ and the partial all-$N$ expressions in the main text have been deposited at the preprint server http://arXiv.org together with a Fortran subroutine of our approximations for the splitting function $P_{\,\rm gq}^{\,(3)}(x)$ in eqs. (16) - (S0.Ex39). These files are also available from the authors upon request.