b-tagging

When a jet of particles sprays out of a collision, a natural question is: which kind of quark started it? Most of the time you cannot tell — a jet from an up quark looks much like a jet from a down quark. But jets started by a bottom quark (a 'b quark') carry a subtle fingerprint, and b-tagging is the set of tricks for spotting that fingerprint. Being able to say 'this jet came from a b quark' is enormously useful, because many of the most interesting particles — the top quark and the Higgs among them — preferentially decay into b quarks.

The fingerprint comes from the b quark's modestly long lifetime. After a b quark hadronizes into a B hadron, that hadron lives just long enough — about a trillionth of a second — to travel a measurable distance, typically a few millimetres, before decaying. So a b-jet contains a secondary vertex: a little knot of tracks that originate not at the collision point but a small, detectable distance away. Modern b-taggers use precise silicon trackers to find these displaced vertices and tracks, and increasingly use machine learning to combine many subtle clues into a single probability that a jet is a b-jet.

B-tagging transformed experimental particle physics. The top quark was discovered partly by demanding b-jets, since top almost always decays to a b quark; and observing the Higgs decaying to a pair of b quarks — its single most common decay — was only possible because b-tagging could pull that signal out of a vast background of ordinary jets. The honest limitations: b-tagging is never perfect. It has an efficiency (the fraction of true b-jets it catches, perhaps 70 percent) and a mistag rate (the fraction of light-quark jets wrongly tagged), and both must be measured carefully, because a mis-estimated tagging rate feeds straight into systematic uncertainty.