Fisher Discriminants for Background Suppression in B Meson Decays

19 concepts20 papers1 proposal

In the search for rare particle decays, the primary challenge is not just detecting the signal, but distinguishing it from an overwhelming "continuum" of background noise. In $e^+e^-$ collider experiments like BABAR, Belle, and CLEO, researchers look for specific $B$ meson decays that occur only once in every million events. To isolate these signals, physicists rely on the Fisher discriminant, a multivariate technique that transforms the geometric "shape" of an event into a powerful tool for classification.

The Topology of a Decay

The intuition behind this approach lies in the conservation of momentum and the mass of the particles involved. When an electron and positron collide at the energy of the $\Upsilon(4S)$ resonance, they produce a pair of $B$ mesons. Because the mass of the $B$ meson pair is very close to the total collision energy, the mesons are produced nearly at rest in the center-of-mass frame. As they decay, their daughter particles fly out in all directions, creating a spherical distribution of energy.

In contrast, the dominant background comes from the production of lighter quarks ($u, d, s, c$), known as continuum events. These light quarks are produced with high kinetic energy, causing them to fragment into collimated "jets" of particles. As noted by del Amo Sanchez et al., exploiting this jet-like nature of continuum events is essential for signal isolation.

From Shapes to Scalars

To quantify these shapes, researchers use several "event shape variables." Common examples include:

  • Sphericity: A measure of how 3D-isotropic the particle tracks are.
  • Fox-Wolfram Moments: Ratios of particle momenta that describe the degree of collimation, as utilized by Aubert et al..
  • Legendre Polynomials: Moments of the energy flow relative to the thrust axis of the event, described by Aubert et al..

While any single variable provides some separation, none are perfect. A simple threshold (or "cut") on sphericity might discard too much signal or let in too much background.

The Fisher discriminant solves this by taking a vector of these correlated variables and projecting them onto a single scalar value. As explained by Bergfeld et al., this linear combination is mathematically optimized to maximize the statistical separation between the signal and background classes. By weighting each variable according to its discriminative power, the authors create a more robust decision boundary than individual physical parameters could provide.

Enhancing Sensitivity for Rare Decays

The Fisher discriminant is particularly vital when the signal is extremely faint, such as in "charmless" or leptonic decays. For example, in the search for $B^+ \to l^+ \nu_l$, Aubert et al. employed Fisher discriminants to combine event shape parameters with kinematic measurements, creating a classifier more powerful than any single variable.

This multi-pronged approach often involves:

  1. Kinematic Vetoes: Removing events that match known, common decays (e.g., $\pi^0$ or $\eta$ decays), as seen in Nakao et al..
  2. Flavor Tagging: Identifying the nature of the other $B$ meson in the event to further confirm the signal's origin, a technique used by Aubert et al..
  3. Angular Distributions: Incorporating the helicity and production angles of the particles, which Aubert et al. used to probe resonance properties while suppressing background.

By aggregating these diverse features into a single Fisher score, researchers can effectively "squash" the multidimensional geometry of a particle collision into a one-dimensional axis where signal and noise finally stand apart. This allows for the observation of rare processes—like the decay $B^0 \to a_1(1260)^+ a_1(1260)^-$ reported by Aubert et al.—that would otherwise remain hidden in the continuum.

History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_fisher_discriminant_d6af7216,
  title = {Fisher discriminant},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/fisher-discriminant/rdr_d6af7216ae9a469f9efc7a1d3d60b07b},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-20},
}

This exact version (version 1, frozen text):

@misc{lacuna_fisher_discriminant_d6af7216_v1,
  title = {Fisher discriminant},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/fisher-discriminant/rdr_d6af7216ae9a469f9efc7a1d3d60b07b/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-20},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake