Beam emittance and luminosity design

Use transverse position and paraxial slope at a fixed reference momentum, with their means subtracted. The two-by-two covariance includes the correlation between displacement and direction. Its determinant gives the squared rms geometric emittance. A strongly correlated distribution can have large separate position and slope variances yet a relatively small determinant. Omitting the correlation term would then overstate the beam’s phase-space area and obscure how a waist differs from a tilted phase-space ellipse.

This rms convention does not multiply by pi and does not mean the area containing a stated percentage of particles. Such areas depend on distribution shape. Projecting a coupled four-dimensional beam onto one plane can also hide correlations with the other plane. In uncoupled, fixed-momentum linear symplectic transport the covariance evolves by the transport matrix and its determinant is preserved. Nonlinear filamentation can change projected rms measures even when fine-grained Hamiltonian phase-space volume remains conserved.

Define normalized rms emittance as geometric rms emittance times relativistic beta gamma. This removes the leading adiabatic shrinkage of angular spread when longitudinal momentum increases without corresponding transverse heating. Under suitable paraxial, slowly accelerating Hamiltonian conditions it is approximately conserved. Radiation damping, quantum excitation, scattering, collective forces and nonlinear coupling can invalidate that simple conservation statement; it is not an exact rule for every storage ring.

The Twiss beta function is an optical length parameter, unrelated to the speed ratio despite sharing the same letter. At negligible dispersion and in an uncoupled matched beam, position variance is geometric emittance times this optical beta. Squeezing beta reduces spot size while increasing appropriate angular spread; it does not reduce normalized emittance. Momentum spread also contributes a dispersive position term when dispersion is nonzero, so inferring emittance from spot size alone needs that contribution removed or modeled.

For two identical transverse Gaussian profiles colliding head-on, luminosity is collision frequency times the product of bunch populations divided by four pi times the two rms widths. Collision frequency already counts colliding bunch pairs; multiplying by the bunch number again double-counts it. Unequal widths require the convolution widths rather than the identical-beam expression. Crossing angle, bunch length, beam-beam forces and imperfect alignment introduce further geometric or dynamical corrections.

Multiplying luminosity by a process cross section gives its produced event rate, before selection. Acceptance and reconstruction efficiency then reduce the expected selected rate if their definitions avoid overlap and correlations are handled. More tightly focusing a beam therefore does not automatically increase useful data without limit: backgrounds, beam disruption, hourglass effects and detector occupancy can offset the nominal overlap gain. The simple formula is a design benchmark under declared assumptions, not a prediction for a named collider operating point.