DPD tutorial

Before you start

Build or install the DPD-enabled PDFxTMD library. Download either set from the DPD page, choose a directory for PDFxTMD data sets, and extract the archive there.

mkdir -p /path/to/PDFxTMDSets
tar --use-compress-program=unzstd \
  -xf ~/Downloads/MSTW2008lo68cl_GSDPDF_Hybrid_256x3.tar.zst \
  -C /path/to/PDFxTMDSets

On Linux and macOS, edit ~/.PDFxTMDLib/config.yaml so that paths includes the data directory:

paths:
  - /path/to/PDFxTMDSets

On Windows, edit C:\ProgramData\PDFxTMDLib\config.yaml and use the same paths entry with a Windows directory path.

Use one of these directory names exactly:

• MSTW2008lo68cl_GSDPDF_PDFxTMD for the dense grid
• MSTW2008lo68cl_GSDPDF_Hybrid_256x3 for the hybrid set

Python

import pdfxtmd

set_name = "MSTW2008lo68cl_GSDPDF_Hybrid_256x3"
dpd = pdfxtmd.GenericCDPDFactory().mkCDPD(set_name, 0)

value = dpd.dpd(
    pdfxtmd.PartonFlavor.g, pdfxtmd.PartonFlavor.g,
    1.0e-2, 100.0, 2.0e-2, 400.0,
)
print(value)

The arguments are the two parton flavours, x1, mu1², x2, and mu2². The Python bindings must have been built with DPD support.

C++

#include "PDFxTMDLib/GenericPDF.h"
#include <iostream>

int main() {
  using namespace PDFxTMD;
  CollinearDPD dpd("MSTW2008lo68cl_GSDPDF_Hybrid_256x3", 0);
  const auto value = dpd.dpd(g, g, 1.0e-2, 100.0, 2.0e-2, 400.0);
  std::cout << value << '\\n';
}

Next steps

The repository includes the complete C++ DPD example and Python DPD example. To train and package a new hybrid set, follow the DPDFSurrogate guide.