Skip to content
Back to search results

PDF Fitting

Introduced by
James Bjorken and collaborators
Year
1969
Type
QCD framework
Family
Hadron structure
Font size

What does it do?

Parton Distribution Function (PDF) fitting is the process of determining the probability distributions of quarks and gluons inside hadrons using high-energy collision data.

PDFs are fundamental inputs to all hadron collider phenomenology, essential for predicting cross-sections, designing triggers, and interpreting new physics searches at the Large Hadron Collider.

The intuition behind the method

Hadrons contain quarks and gluons that cannot be directly observed due to color confinement. Their distributions as functions of momentum can be inferred from deep inelastic scattering experiments and hadron collider data using QCD. PDFs encode the internal structure of protons and neutrons. Different parton flavors have different distributions: valence quarks carry most momentum, while gluons become important at high energies. Fitting PDFs to precision measurements provides constraints on the parton content and tests QCD predictions.

History

Bjorken discovered asymptotic scaling in deep inelastic scattering in 1969, leading to the parton model. The DGLAP evolution equations, formulated in the 1970s, connected PDFs across scales. PDF fitting began systematically in the 1990s (CTEQ, MRS) and has evolved into a mature science. Modern PDF collaborations (CT, NNPDF, MMHT) use machine learning and Bayesian approaches for flexible parameterization and uncertainty quantification. Recent fits include precision electroweak data and Higgs measurements, improving constraints on sea quarks and gluons.

When to use it

PDF fitting is mandatory for all hadron collider predictions. Use when calculating cross-sections for Standard Model processes, estimating backgrounds, designing triggers, or predicting new physics signatures. Different PDF sets optimized for different purposes: inclusive cross-sections use CT or NNPDF; precision electroweak uses NNPDF; dedicated sets for heavy quark production or Higgs. Always evaluate PDF uncertainties on your prediction.

Limitations and considerations

  • Inherent uncertainty in extrapolating from measured region to extreme kinematics (high-Q scales, small-x regime)
  • Correlation between PDF parameters and other QCD parameters (alphas, heavy quark masses) complicates interpretation
  • Limited sensitivity to flavor structure; strange and charm quark PDFs poorly constrained by traditional data
  • Tensions between different experimental datasets can introduce fitting bias
  • High-precision fits computationally expensive, limiting iterations and uncertainty studies

Source conditions do not replace independent scientific review or validation by an analysis engine.

Strengths

  • Quantitatively connects hadron structure to collider observables using QCD framework
  • Enables precision predictions of Standard Model processes for calibration and systematics
  • Uncertainty quantification through PDF error sets guides experimental precision needed
  • Sensitivity to new physics at high parton momentum fractions (Bjorken-x near 1)
  • Continuous improvement as new high-precision data becomes available

Applications

  • Higgs boson production cross-section predictions at LHC for ggF and VBF production modes
  • Standard Model background estimation for new particle searches at all energies
  • Dijet and multijet production cross-sections for trigger optimization and systematics estimation
  • Rare process predictions (diboson, top quark pair) for precision measurements
  • Constraints on quark flavor content from W/Z asymmetries and heavy flavor production

How it works

How it works

1

Collect measurements from deep inelastic scattering (HERA), fixed-target experiments, and hadron collider data sensitive to parton distributions.

2

Apply QCD corrections (NLO, NNLO) to theoretical predictions to match the precision of experimental measurements; account for nuclear effects and correlations.

3

Parameterize PDFs at a reference scale using flexible functional forms (e.g., polynomials) with adjustable coefficients.

4

Evolve PDFs to higher scales using DGLAP equations; vary parameters to minimize chi-squared between predictions and data.

5

Extract best-fit PDFs and uncertainty bands; validate using independent datasets and apply to new physics predictions at colliders.

Common pitfalls

  • Using outdated PDF sets; recent measurements (especially NNLO) can shift predictions significantly
  • Neglecting PDF correlations; treating PDF uncertainties as independent leads to underestimated systematics
  • Applying PDFs outside their fitted kinematic region; extrapolated predictions in extreme phase space unreliable
  • Forgetting that PDF uncertainties are not simple addition to other systematics; correlations matter for precision measurements

Frequently asked questions

What is the difference between NNPDF and CT PDFs?

CT (CTEQ) uses traditional parameterization with analytic forms. NNPDF (Neural Network PDF) uses neural networks for flexible parameterization. NNPDF generally provides more flexibility but requires careful training; CT is simpler but less flexible. Both are valid; differences quantify PDF uncertainty.

How do I propagate PDF uncertainties to my analysis?

Use the PDF uncertainty sets provided by the collaboration (e.g., 100 eigenvector sets from NNPDF). Calculate your observable with each set, extracting the RMS variation as the uncertainty. This correctly accounts for correlations between parameters.

Can PDFs constrain new physics?

Yes, indirectly. New physics at production or decay can distort PDF constraints inferred from data. Anomalous forward-backward asymmetries or unexpected kinematic distributions hint at new physics. PDFs and new physics fits must be performed simultaneously for unbiased results.

Why is the small-x region problematic?

At very small x (high parton density), saturation and unitarity effects become important, invalidating the linear DGLAP approximation. The Color Glass Condensate (CGC) describes physics in this regime but is not yet incorporated into standard PDF fitting.

Related methods

Sources

  1. Bjorken, J. D. (1969). Asymptotic sum rules at infinite momentum. Physical Review, 179(5), 1547.

    Open DOI reference
    Copied.
  2. Alekhin, S., et al. (2014). PDF4LHC recommendations for LHC Run II. The European Physical Journal C, 75(7), 304.

    Open publication reference
    Copied.
  3. Bailey, S., et al. (2020). Parton distributions for the LHC Run II. The European Physical Journal C, 76(7), 391.

    Open publication reference
    Copied.