\midheading{Diphoton production}

\label{subsec:gamgam}
Process 285 represents the production of a pair of real photons. Since     
this process includes two real photons, the cross section diverges when    
one of the photons is very soft or in the direction of the beam. Thus in   
order to produce sensible results, the input file must supply values for  
both ptmin$_{photon}$ and etamax$_{photon}$. This will ensure that the cross   
section is well-defined.

The calculation of process 285 may be performed using either the           
Frixione algorithm or standard cone isolation. Since version 10.1 also a   
fixed cone size can be specificed as well as a simple hybrid cone          
isolation, see ref.~\cite{Neumann:2021zkb}.           

This process also includes the one-loop gluon-gluon contribution as        
given in ref.~\cite{Bern:2002jx}. The 
production of a photon via parton fragmentation is included at NLO and     
can be run separately by using the frag option in part. This option        
includes the contributions from the integrated photon dipole subtraction   
terms and the LO QCD matrix element multiplied by the fragmentation        
function.

The phase space cuts for the final state photons are defined in
{\tt{input.ini}}, for multiple photon processes such as {\tt 285 -
287} the $p_T$'s of the individual photons (hardest, second hardest
and third hardest or softer) can be controlled independently.The
remaining cuts on $R_{\gamma j}$, $\eta_{\gamma}$ etc. are applied
universally to all photons. Users wishing to alter this feature should
edit the file {\tt{photon\_cuts.f}} in the directory {\tt{src/User}}.

This process can be calculated at LO, NLO, and NNLO.
NLO calculations can be performed by subtraction, zero-jettiness slicing and $q_T$-slicing. 
NNLO calculations can be performed by  zero-jettiness slicing and $q_T$-slicing. 
Input files for these 6 possibilities are given in the  link below.

The fixed-order NNLO calculation has been implemented in ref.
\cite{Campbell:2016yrh}. Transverse
momentum resummation at the level of $\text{N}^3\text{LL}+\text{NNLO}$
has been implemented in ref. \cite{Becher:2020ugp}. By including the three-loop hard
\cite{Caola:2020dfu} and beam functions
\cite{Luo:2020epw},\cite{Ebert:2020yqt},\cite{Luo:2019szz} it has been upgraded to $\text{N}^3\text{LL}'$ in ref. \cite{Neumann:2021zkb}.



\subsection{Transverse momentum resummation}

Transverse momentum resummation can be enabled for process {\tt 285} at
highest order $\text{N}^3\text{LL}'+\text{NNLO}$ with `part=resNNLOp`.
The setting `part=resNNLO` resums to order
$\text{N}^3\text{LL}+\text{NNLO}$ ($\alpha_s^2$ accuracy in improved
perturbation theory power counting) and `part=resNLO` to order
$\text{N}^3\text{LL}+\text{NLO}$. Note that process 285 with resummation
only includes the $q\bar{q}$ channel. The $gg$ channel enters at an
increased relative level of $\alpha_s$, so has to be added with process
number 2851 at order `part=resNLO` for overall
$\text{N}^3\text{LL}'+\text{NNLO}$ precision. For an overall consistent
precision of $\text{N}^3\text{LL}+\text{NNLO}$ the $gg$ channel can be
added with `part=resLO`.

Note that at fixed-order the $gg$ channel is included at NNLO
automatically at the level of $\alpha_s^2$.

The fixed-order NNLO calculation has been implemented in ref.
\cite{Campbell:2016yrh}. Transverse
momentum resummation at the level of N$^3$LL+NNLO
has been implemented in ref. \cite{Becher:2020ugp}. By including the three-loop hard
\cite{Caola:2020dfu} and beam functions
\cite{Luo:2020epw},\cite{Ebert:2020yqt},\cite{Luo:2019szz} it has been upgraded to N$^3$LL' in ref. \cite{Neumann:2021zkb}.