Tracking detectors & momentum measurement

Every measurement a collider makes begins with a charged particle drawing its own picture. As it crosses matter it strips electrons from atoms, leaving a thin trail of ionisation a detector can sense — in a gas, drifting electrons collected on wires; in silicon, roughly ten thousand electron–hole pairs per wafer crossing swept to a p–n junction. Record where the trail crossed a handful of surveyed layers and you have the trajectory. That answers where it went. The second question — how much momentum it carried — is answered by bathing the tracking volume in a strong axial magnetic field. The $\mathbf{v}\times\mathbf{B}$ Lorentz force bends the trajectory into a helix, and the helix's radius of curvature is a direct measure of momentum: $p_{\mathrm T}=0.3\,BR$ in the working units of GeV/c, tesla, and metres. The remarkable thing is that both answers come from one set of hits: fit a helix to the space-points and you read off direction, production vertex, and curvature at once. But the physics has a sting in the tail. High-momentum tracks barely bend — a 100 GeV/c pion in the 3.8 T CMS solenoid curls with a radius of nearly 90 metres, so across a one-metre tracker it deviates from a straight line by under two millimetres. Measuring that sagitta to a few percent demands space-points good to tens of micrometres, and confronts the two irreducible enemies of precision: the finite resolution of the sensors, which makes the fractional momentum error grow with momentum, and multiple Coulomb scattering in the detector material, which sets a momentum-independent floor. This article builds the momentum-from-curvature relation from the Lorentz force, derives the resolution from the sagitta, and shows where the two error terms cross.

Key takeaways

A charged particle ionises the medium it crosses, leaving a trail that tracking detectors sample as discrete hits across several layers; fitting the hits reconstructs the trajectory and its production vertex.
An axial magnetic field bends the track into a helix through the $\mathbf{v}\times\mathbf{B}$ force, and the curvature radius gives the transverse momentum: $p_{\mathrm T}=0.3\,zBR$ (GeV/c, T, m).
Curvature measures only $p_{\mathrm T}$ ; the full momentum needs the pitch angle, $p=p_{\mathrm T}/\cos\lambda$ .
Resolution has two terms combined in quadrature: a measurement term (Gluckstern) that grows $\propto p_{\mathrm T}$ , and a multiple-scattering floor that is momentum-independent.
Better resolution means a stronger field $B$ , a longer lever arm $L$ (entering as $L^2$ ), finer $\sigma_x$ , more layers $N$ , and less material $X_0$ .
Two technology families dominate: gaseous drift chambers and time projection chambers, and silicon pixel and microstrip detectors resolving hits to $\sim10$ – $25\ \mu$ m.
Key idea: a charged particle's ionisation trail records where it went, and the magnetic bending of that trail records how much momentum it carried — both extracted from a single set of precisely measured hit positions.

The core insight

A charged particle's ionisation trail records where it went, and the magnetic bending of that trail records how much momentum it carried — both read from a single set of precisely measured hit positions, with the curvature of the helix fixing the momentum through $p_{\mathrm T}=0.3\,BR$ .

Where this lives today

The largest silicon trackers ever built sit at the LHC. The CMS tracker alone carries about 200 m² of silicon — roughly 66 million pixels and 10 million strip channels — inside a 3.8 T solenoid, delivering $\sigma(p_{\mathrm T})/p_{\mathrm T}\approx1$ – $2\%$ up to $\sim100$ GeV/c and locating vertices to tens of micrometres. That vertex precision is what makes b-tagging possible: a jet from a b quark contains a hadron that flies a measurable fraction of a millimetre before decaying, and resolving that displaced vertex is central to studying the Higgs decay to $b\bar b$ . ATLAS pairs its silicon with a straw-tube transition-radiation tracker; ALICE uses a large time projection chamber, Charpak's lineage, to reconstruct the thousands of tracks in a heavy-ion collision. The frontier now is the High-Luminosity LHC, where trackers must survive punishing radiation and reconstruct tracks amid up to 200 simultaneous collisions per crossing — driving thinner sensors, smaller pixels, and, for the first time, tracking fast enough to feed the hardware trigger.

The physical picture

Picture the transverse slice of a collider detector: concentric cylinders of silicon around the beam pipe, inside a solenoid whose field points along the beam. A charged particle born at the collision point spirals outward, depositing a packet of charge in each layer it crosses — those packets are the hits. Seen end-on, the helix projects to a circular arc, and the whole art of momentum measurement is reading the radius of that arc from a few points strung along it, because the radius is momentum in disguise. The first figure pins the geometry to a real number: a 100 GeV/c track in the CMS field barely curves at all.

Figure

A charged track leaves a hit in each concentric silicon layer and curves in the axial field; the transverse momentum follows from the curvature radius through p_T = 0.3 B R, so a 100 GeV/c track in the 3.8 T CMS field bends with R ≈ 88 m and looks almost straight

Because the useful signal is the tiny departure of the arc from a straight chord, precision is set by two competing effects, and a single plot tells the whole engineering story. The finite pitch of the sensors limits how well any one hit is located, and that error, propagated into the curvature, makes the fractional momentum uncertainty climb in direct proportion to momentum — a stiff 1 TeV track is the hardest to measure, not the easiest. Working against it from below, every gram of material scatters the particle by multiple Coulomb scattering, blurring the trajectory by an angle $\propto1/p$ ; since the curvature signal is also $\propto1/p$ , their ratio is momentum-independent — a flat floor. The second figure is what a tracker designer actually optimises: where those two curves live and where they cross.

Figure

Transverse-momentum resolution σ(p_T)/p_T for a CMS-like silicon tracker with B = 3.8 T, L = 1.1 m, N = 13 layers and σ_x = 25 μm: a multiple-scattering floor near 0.7% dominates below ≈ 58 GeV/c, above which the Gluckstern measurement term rising ∝ p_T takes over, reaching ≈ 1.4% at 100 GeV/c

The core equations

The starting point is the momentum–curvature relation. A charged particle in a uniform axial field moves on a helix whose transverse projection is a circle of radius $R$ fixed by the transverse momentum. Balancing the magnetic force against the centripetal requirement gives $p_{\mathrm T}=qBR$ exactly, and the unit conversion produces the ubiquitous factor $0.3$ .

p_{\mathrm T}[\mathrm{GeV}/c]=0.299792458\,z\,B[\mathrm T]\,R[\mathrm m]\;\approx\;0.3\,zBR,\qquad p=\frac{p_{\mathrm T}}{\cos\lambda} \qquad (1)

where: $p_{\mathrm T}$ — momentum transverse to the field; $z$ — charge in units of $e$ ; $B$ — field strength; $R$ — curvature radius of the circular projection; $\lambda$ — pitch (dip) angle, $p=p_{\mathrm T}/\cos\lambda$ recovering the full momentum. Reading it: curvature alone gives the transverse momentum, and a stiffer track has a larger $R$ and bends less. What the tracker actually resolves is not $R$ but the sagitta — the maximum bulge of the arc away from the straight chord joining its ends. Over a lever arm $L$ , geometry gives the sagitta, and equation (1) shows it shrinks as $1/p_{\mathrm T}$ .

s=\frac{L^2}{8R}=\frac{0.3\,BL^2}{8\,p_{\mathrm T}} \qquad (2)

where: $s$ — sagitta; $L$ — lever arm (transverse projected track length); $R$ — curvature radius. Reading it: the measurable signal falls with momentum — why high- $p_{\mathrm T}$ tracks are hard; a 100 GeV/c CMS track bulges only $\approx 1.7$ mm over $L=1.1$ m. Each hit is located only to the granularity of the sensor. For a strip or pixel of pitch $w$ read out in binary mode — a strip either fires or not — the position estimate is the strip centre, and the single-point resolution is the standard deviation of a uniform distribution across the pitch.

\sigma_x=\frac{w}{\sqrt{12}} \qquad (3)

where: $\sigma_x$ — single-point spatial resolution; $w$ — readout pitch. Reading it: an 80 μm strip in binary readout gives $\sigma_x\approx 23\ \mu$ m; analogue charge-sharing between neighbouring channels does substantially better, reaching $\sim 10\ \mu$ m in silicon pixels. Propagating $\sigma_x$ through the sagitta measurement gives the measurement contribution to the resolution. For $N$ uniformly spaced points fitted optimally, this is the Gluckstern formula — the workhorse of tracker design.

\left.\frac{\sigma(p_{\mathrm T})}{p_{\mathrm T}}\right|_{\mathrm{meas}}=\frac{\sigma_x\,p_{\mathrm T}}{0.3\,BL^2}\sqrt{\frac{720}{N+4}} \qquad (4)

where: $N$ — number of layers; $L$ — lever arm; $B$ — field; $\sigma_x$ — point resolution ( $p_{\mathrm T}$ in GeV/c, lengths in m). Reading it: the fractional error grows linearly with $p_{\mathrm T}$ and improves as $1/(BL^2)$ and $1/\sqrt N$ — the lever arm, entering squared, is the strongest knob. Finally, multiple Coulomb scattering adds a term that does not vanish at low momentum. The scattering angle scales as $1/(\beta p)$ (Highland) and the curvature signal as $1/p$ , so their ratio is momentum-independent — a floor set by material and field.

\left.\frac{\sigma(p_{\mathrm T})}{p_{\mathrm T}}\right|_{\mathrm{ms}}\approx\frac{0.0136}{0.3\,\beta\,B\,L}\sqrt{\frac{L}{X_0}},\qquad \frac{\sigma(p_{\mathrm T})}{p_{\mathrm T}}=\left.\frac{\sigma(p_{\mathrm T})}{p_{\mathrm T}}\right|_{\mathrm{ms}}\oplus\left.\frac{\sigma(p_{\mathrm T})}{p_{\mathrm T}}\right|_{\mathrm{meas}} \qquad (5)

where: $X_0$ — radiation length of the material; $\beta=v/c$ ; $L/X_0$ — track length in radiation lengths; $\oplus$ — quadrature sum. Reading it: the terms cross near 58 GeV/c for CMS-like numbers — scattering-limited below, resolution-limited above.

Derivation AFrom the Lorentz force to momentum from curvature

The goal is to establish equation (1) — that the transverse momentum of a charged particle equals $qBR$ , and numerically $0.3\,zBR$ in collider units — from nothing but the Lorentz force, and to make explicit why curvature measures only the transverse component so the pitch-angle correction $p=p_{\mathrm T}/\cos\lambda$ is not optional. This is the relation that turns a geometric fit of hit positions into a physical momentum; every number a tracker reports rests on it.

Step 1. Write the force on charge $q$ moving with velocity $\mathbf v$ in a uniform field $\mathbf B=B\hat{\mathbf z}$ : $\mathbf F=q\,\mathbf v\times\mathbf B$ . Split $\mathbf v$ into $v_\parallel$ along $\hat{\mathbf z}$ and $v_{\mathrm T}$ in the transverse plane; the cross product kills the parallel part, so the force is purely transverse and constant, $F=qv_{\mathrm T}B$ . Why: a force always perpendicular to the velocity does no work — it turns the transverse motion in a circle while the axial motion coasts, which is precisely a helix.

Step 2. Impose circular motion in the transverse plane. The transverse force supplies the centripetal acceleration of a particle of relativistic momentum $p_{\mathrm T}=\gamma m v_{\mathrm T}$ on a circle of radius $R$ : $qv_{\mathrm T}B=\gamma m v_{\mathrm T}^2/R$ . Cancel $v_{\mathrm T}$ and multiply by $R$ to get $p_{\mathrm T}=\gamma m v_{\mathrm T}=qBR$ . Why: writing $\gamma m v_{\mathrm T}$ as the relativistic momentum keeps the result exact at collider energies where $\gamma\gg1$ ; the relation $p_{\mathrm T}=qBR$ is fully relativistic despite looking Newtonian.

Step 3. Convert to practical units. In SI, $p_{\mathrm T}=qBR$ gives $p_{\mathrm T}$ in kg·m/s; a momentum in GeV/c equals $p_{\mathrm T}c/(10^9 e)$ , so $p_{\mathrm T}[\mathrm{GeV}/c]=(c/10^9)\,zBR=0.299792458\,zBR$ with $B$ in tesla, $R$ in metres, $z=q/e$ . Why: the famous $0.3$ is nothing but $c$ in units of $10^9$ m/s — a bookkeeping constant, not new physics, that silently assumes a $z$ -fold charged particle.

Step 4. Restore the third dimension. The helix has a pitch angle $\lambda$ between the velocity and the transverse plane, so the circular projection gives $p_{\mathrm T}=p\cos\lambda$ and the full momentum is $p=p_{\mathrm T}/\cos\lambda$ , with $\lambda$ from the same helix fit. Why: forgetting this is the commonest blunder in reading the relation — a track at $45^\circ$ carries $\sqrt2$ times the momentum its curvature naively suggests.

Go deeper · Derivation B · 2 minThe sagitta and the Gluckstern resolution

Derivation BThe sagitta and the Gluckstern resolution

The aim is to turn the geometry of a shallow arc into the measurement term of the momentum resolution, equation (4), and to expose why the fractional error grows with momentum and why the lever arm enters squared. We compute the sagitta the hits determine, relate its uncertainty to the point resolution, and quote the optimal $N$ -point Gluckstern result, checking that the crude three-point estimate reproduces its coefficient.

Step 1. For a circular arc of radius $R$ subtending a chord of length $L$ , the sagitta is $s=R-R\cos(\theta/2)$ with $\theta=L/R$ the arc angle. In the shallow-arc limit relevant to stiff tracks, $\cos(\theta/2)\approx1-\theta^2/8$ , giving $s=L^2/(8R)$ . Why: at high momentum $R$ is enormous and $\theta$ tiny, so the whole measurement lives in the second-order term of the cosine — the sagitta is a small difference of large numbers, which is exactly why precision matters.

Step 2. Insert $R=p_{\mathrm T}/(0.3B)$ to get $s=0.3\,BL^2/(8p_{\mathrm T})$ , equation (2). Since $s\propto 1/p_{\mathrm T}$ , a fractional error in $s$ equals the fractional error in $p_{\mathrm T}$ : $\sigma(p_{\mathrm T})/p_{\mathrm T}=\sigma_s/s$ . Why: this is the pivot of the argument — momentum is measured through the reciprocal-momentum signal $s$ , so a fixed uncertainty $\sigma_s$ becomes a fractional error that climbs linearly with $p_{\mathrm T}$ .

Step 3. Estimate $\sigma_s$ from three equally spaced points — two ends and a middle. The sagitta is the middle point minus the mean of the ends, $s=x_2-\tfrac12(x_1+x_3)$ , so with independent errors $\sigma_x$ per point, $\sigma_s=\sigma_x\sqrt{1+\tfrac14+\tfrac14}=\sigma_x\sqrt{3/2}$ . Then $\sigma(p_{\mathrm T})/p_{\mathrm T}=\sigma_x\sqrt{3/2}\big/[0.3BL^2/(8p_{\mathrm T})]=8\sqrt{3/2}\,\sigma_x p_{\mathrm T}/(0.3BL^2)$ , with coefficient $8\sqrt{3/2}\approx9.80$ . Why: three points are the minimum that can define a curvature, and this back-of-envelope coefficient lets us sanity-check the full formula.

Step 4. For $N$ equally spaced points fitted by least squares — using every hit — Gluckstern (1963) showed the coefficient becomes $\sqrt{720/(N+4)}$ , giving equation (4). For $N=3$ this is $\sqrt{720/7}\approx10.14$ , within a few percent of the three-point estimate $9.80$ . Why: real trackers have ten or more layers, and the $1/\sqrt{N}$ buried in $\sqrt{720/(N+4)}$ is why adding well-placed layers buys resolution — until multiple scattering from the added material takes over.

Go deeper · Derivation C · 1 minThe multiple-scattering floor

Derivation CThe multiple-scattering floor

The purpose here is to explain the flat, momentum-independent term in equation (5): why traversing detector material sets a resolution floor that no amount of point precision can beat, and why that floor depends on field, lever arm, and radiation length but not on momentum. We propagate the Highland scattering angle into an apparent sagitta error and take its ratio to the true sagitta signal.

Step 1. A particle crossing thickness $x$ of radiation length $X_0$ acquires a Gaussian spread in projected angle, the Highland formula $\theta_0=\dfrac{13.6\ \mathrm{MeV}}{\beta c p}\,z\sqrt{x/X_0}\,[1+0.038\ln(x/X_0)]$ . Over the lever arm this random kick displaces the track, mimicking an extra sagitta $\sigma_s^{\mathrm{ms}}\sim L\,\theta_0$ . Why: the scattering is genuine deflection — the tracker cannot tell a field-induced curvature from a scatter, so it contaminates the very quantity being measured.

Step 2. Divide the scattering-induced sagitta by the true sagitta $s=0.3BL^2/(8p_{\mathrm T})$ . The momentum cancels: $\theta_0\propto1/(\beta p)$ and $s\propto1/p_{\mathrm T}$ , so for $\beta\approx1$ the ratio is momentum-independent. Why: this cancellation is the whole point — at low momentum the track bends a lot (large signal) but scatters a lot too (large noise), and the two track together, leaving a constant floor.

Step 3. Collecting constants gives $\left.\sigma(p_{\mathrm T})/p_{\mathrm T}\right|_{\mathrm{ms}}\approx\dfrac{0.0136}{0.3\,\beta B L}\sqrt{L/X_0}\approx\dfrac{0.045}{\beta BL}\sqrt{L/X_0}$ (GeV, T, m). The two terms add in quadrature because point errors and scattering are independent, giving the canonical form $\sigma(p_{\mathrm T})/p_{\mathrm T}=a\oplus b\,p_{\mathrm T}$ . Why: the floor $a$ is beaten only by a stronger field, a longer lever arm, or a lighter tracker — which is why the LHC experiments worked so hard to make silicon and supports thin, since $X_0$ sets a limit no readout can improve.

Limiting cases

Case	Result	Meaning
High momentum ( $p_{\mathrm T}\to\infty$ )	$R\to\infty$ , $s\to0$ , $\sigma(p_{\mathrm T})/p_{\mathrm T}\propto p_{\mathrm T}$	measurement-limited; the sagitta vanishes and stiff tracks are the hardest, degrading toward a "charge-sign only" limit at the TeV scale
Low momentum ( $p_{\mathrm T}\lesssim 58$ GeV/c here)	$\sigma(p_{\mathrm T})/p_{\mathrm T}\to$ const $\approx0.7\%$	multiple-scattering floor; resolution stops improving because the material scatters as much as the field bends
Very low $p_{\mathrm T}$	$2R<r_{\max}$ : track curls up inside the tracker	below $p_{\mathrm T}\approx0.3\,B\,r_{\max}/2$ ( $\approx0.6$ GeV/c for CMS) the helix never reaches the outer layers
Strong field / long arm	$\sigma(p_{\mathrm T})/p_{\mathrm T}\propto1/(BL^2)$	doubling $L$ quarters the error; the lever arm is the most powerful design knob
Many layers	$\sigma(p_{\mathrm T})/p_{\mathrm T}\propto\sqrt{720/(N+4)}$	resolution improves as $1/\sqrt N$ — but only if the extra layers do not add scattering material
Thin tracker ( $X_0\to\infty$ )	$\left.\sigma/p\right\|_{\mathrm{ms}}\to0$	the floor is set by material alone; an ideal massless tracker is limited only by the measurement term

Common trapTrack curvature measures the particle's total momentum

It does not — the curvature fixes only the transverse momentum $p_{\mathrm T}$ , the component perpendicular to the field. A particle travelling steeply along the beam makes a tightly wound helix whose circular projection is small, so its curvature looks low-momentum even though $p=p_{\mathrm T}/\cos\lambda$ may be large; you must combine the curvature with the pitch angle $\lambda$ from the same fit to recover $p$ . Two further traps hide in the innocent $0.3$ : it bakes in the units GeV/c, tesla, metres, so feeding it MeV or centimetres gives nonsense; and it assumes unit charge — a doubly charged fragment bends twice as hard, so $p_{\mathrm T}=0.3\cdot2\cdot BR$ . Finally, more layers do not always help: each is material, and past the scattering-dominated point an extra plane can worsen the low-momentum resolution it was meant to improve.

Worked examples

1 · A 100 GeV/c pion in the CMS tracker (fully worked)

The CMS solenoid provides $B=3.8$ T over a silicon tracker of outer radius $L\approx1.1$ m with $N=13$ measurement layers of point resolution $\sigma_x=25\ \mu$ m. Take a 100 GeV/c pion emerging transverse to the beam. Find its curvature radius, its sagitta, and the transverse-momentum resolution.

Step 1. Curvature radius from equation (1): $R=p_{\mathrm T}/(0.3B)=100/(0.3\times3.8)=100/1.14=\mathbf{87.7\ m}$ . The track's radius of curvature is nearly eighty times the size of the detector — over the tracker it is almost a straight line.

Step 2. Sagitta from equation (2): $s=L^2/(8R)=1.1^2/(8\times87.7)=1.21/701.6=1.72\times10^{-3}$ m $=\mathbf{1.72\ mm}$ . This millimetre-scale bulge, measured against a metre-long chord, is the entire momentum signal.

Step 3. Measurement resolution from Gluckstern, equation (4): $\dfrac{\sigma(p_{\mathrm T})}{p_{\mathrm T}}\Big|_{\mathrm{meas}}=\dfrac{(25\times10^{-6})(100)}{0.3\times3.8\times1.1^2}\sqrt{\dfrac{720}{13+4}}=\dfrac{2.5\times10^{-3}}{1.379}\times\sqrt{42.35}=1.81\times10^{-3}\times6.51=\mathbf{1.18\%}$ .

Step 4. Add the scattering floor. Taking a material budget $L/X_0\approx0.4$ and $\beta\approx1$ , equation (5) gives $\left.\sigma(p_{\mathrm T})/p_{\mathrm T}\right|_{\mathrm{ms}}=\dfrac{0.0136}{0.3\times3.8\times1.1}\sqrt{0.4}=\mathbf{0.69\%}$ . Combining in quadrature: $\sqrt{0.69^2+1.18^2}=\sqrt{1.87}=\mathbf{1.37\%}$ . A 100 GeV/c track is measured to about 1.4%, dominated by sensor resolution rather than scattering.

2 · Where scattering gives way to measurement

Using the same tracker, find the crossover momentum at which the measurement term equals the multiple-scattering floor, and check what limits the resolution far above and far below it.

Step 1. From Example 1 the two terms are $\left.\sigma/p\right|_{\mathrm{ms}}=0.69\%$ (constant) and $\left.\sigma/p\right|_{\mathrm{meas}}=1.18\%\times(p_{\mathrm T}/100\ \mathrm{GeV})$ — a slope $b=1.18\times10^{-4}$ per GeV/c, linear in $p_{\mathrm T}$ .

Step 2. Set them equal: $b\,p_{\mathrm T}=0.0069$ , so $p_{\mathrm T}=0.0069/(1.18\times10^{-4})=\mathbf{58\ GeV/c}$ . Below this the tracker is scattering-limited (resolution flat near 0.7%); above it, resolution-limited (resolution rising with $p_{\mathrm T}$ ).

Step 3. Sanity-check the extremes. At $p_{\mathrm T}=1$ GeV/c the measurement term is a negligible $0.0012\%$ , so the resolution sits on its floor at $\approx\mathbf{0.7\%}$ . At $p_{\mathrm T}=1$ TeV/c it is $\mathbf{11.8\%}$ and dominates — the sagitta has shrunk to $172\ \mu$ m, only a few point-resolutions wide. Scattering-limited below ≈ 58 GeV/c, measurement-limited above; a 1 TeV/c muon is known only to about 12%.

3 · Binary strip resolution and low-momentum curl-up

A silicon microstrip sensor has 80 μm pitch read out in binary mode. What single-hit resolution does it give, and what is the lowest transverse momentum that still reaches the 1.1 m outer layer of the CMS tracker?

Step 1. Binary resolution from equation (3): $\sigma_x=w/\sqrt{12}=80/3.464=\mathbf{23.1\ \mu m}$ . Analogue readout that interpolates the charge shared between neighbouring strips beats this, reaching $\sim10\ \mu$ m — why pixel detectors interpolate rather than threshold.

Step 2. A track reaches radius $r$ only if $2R\geq r$ , so $R\geq r/2$ . With $R=p_{\mathrm T}/(0.3B)$ , the minimum is $p_{\mathrm T}\geq0.3\,B\,r/2=0.3\times3.8\times1.1/2=\mathbf{0.63\ GeV/c}$ . Softer tracks spiral back before leaving the tracker, appearing as short curls near the beam line.

Physics of the portrait

Tracking has two founding revolutions — the visual era of cloud and bubble chambers, in which a physicist scanned photographs by eye, and the electronic era, in which the detector counts its own hits and hands them to a computer. The bridge between them was one instrument, and one man.

Origins · 1 minGeorges Charpak and how this was discovered

Georges Charpak

1924–2010 · the multiwire proportional chamber and electronic tracking

Georges Charpak survived Dachau to become the physicist who made particle detectors electronic. Before 1968 the finest tracking used bubble chambers — beautiful, but slow: each event was a photograph, scanned and measured by hand at a few per second. Charpak, working at CERN, asked whether a plane of closely spaced anode wires strung in a gas could each act as an independent proportional counter, so an avalanche on a given wire localised the passing particle electronically and instantly. His multiwire proportional chamber (MWPC), reported in 1968, did exactly that: thousands of channels read out a thousand times faster than any chamber before it, with millimetre resolution and no film to develop. The idea spawned an entire family — drift chambers, which time the electrons' arrival to reach tens of micrometres, and Nygren's time projection chamber, which turns a whole gas volume into a three-dimensional track camera. Nearly every major experiment for four decades used a descendant of the MWPC, and the technique reached far beyond physics into medical imaging. Charpak received the 1992 Nobel Prize in Physics — unusually, to a single person for inventing a detector rather than for a discovery made with one. The silicon trackers now at the heart of ATLAS and CMS are the direct heirs of his insight: sample the ionisation trail electronically, in many layers, and let the computer fit the helix.

History

1912

— Wilson's cloud chamber photographs the first particle tracks

1952

— Glaser invents the bubble chamber, the workhorse of the visual era

1968

— Charpak builds the multiwire proportional chamber, and tracking goes electronic

1974

— Nygren conceives the time projection chamber, a 3-D track camera

1980

— the first silicon microstrip detectors resolve short-lived decay vertices

2009

— the LHC silicon trackers of ATLAS and CMS begin recording collisions

Going further

Texts. Thomson, Modern Particle Physics, §1.2 (interactions and detection of charged particles) · Grupen & Shwartz, Particle Detectors, 2nd ed., Chs. 7–8 (tracking, momentum measurement) · Kleinknecht, Detectors for Particle Radiation, 2nd ed., Ch. 2 · Green, The Physics of Particle Detectors, Chs. 8–10 · Particle Data Group, Review of Particle Physics, "Particle Detectors at Accelerators" (Gluckstern resolution, radiation length). Standard field references — Peskin & Schroeder, Griffiths, Halzen & Martin, Thomson, Schwartz — supply the relativistic kinematics ( $p=\gamma m v$ ) and the ionisation physics on which the whole chapter rests.

Provenance. The momentum relation $p_{\mathrm T}=0.3\,BR$ follows from the Lorentz force and the relativistic momentum $p=\gamma m v$ recalled from aq-pp-special-relativity-kinematics; the ionisation that makes the trail visible is governed by the Bethe–Bloch law owned by aq-pp-ionisation-energy-loss and aq-pp-particle-matter-interaction (recalled by name, not re-derived here). The Gluckstern coefficient $\sqrt{720/(N+4)}$ is from Gluckstern (1963); the multiple-scattering floor uses the Highland form of the scattering angle. A verifier should confirm: (1) $p_{\mathrm T}[\mathrm{GeV}/c]=0.2998\,zBR$ from $c/10^9$ ; (2) $R=87.7$ m for 100 GeV/c at 3.8 T and sagitta $1.72$ mm over $L=1.1$ m; (3) the Gluckstern resolution $1.18\%$ and multiple-scattering floor $0.69\%$ for the stated CMS-like parameters, crossing near 58 GeV/c; (4) $\sigma_x=w/\sqrt{12}=23.1\ \mu$ m for $w=80\ \mu$ m and curl-up threshold $p_{\mathrm T}\approx0.63$ GeV/c at $r=1.1$ m.

Prerequisites. Ionisation Energy Loss · Interactions of Particles with Matter · Relativistic Kinematics. Leads to. Collider Experiments · Cherenkov Radiation · Electromagnetic Showers.

Ask the tutor: Ask me to work the Gluckstern formula for your own detector's field, radius, and layer count, or to show how vertexing turns the same hits into a b-tag.

Why is the constant in p_T = 0.3 B R exactly 0.3?

Because it is the speed of light in disguise. Starting from the exact relation $p_{\mathrm T}=qBR$ in SI units and converting the momentum into the particle-physics unit GeV/c, the conversion factor works out to $c$ divided by $10^9$ , i.e. $2.998\times10^8/10^9=0.2998\approx0.3$ . So the "0.3" is not an approximation of any physics — it is simply $c$ measured in units of $10^9$ m/s, and it is exact to the precision you keep on $c$ . The corollary is that the formula only works when $B$ is in tesla, $R$ in metres, and $p_{\mathrm T}$ in GeV/c; use any other units and you must recompute the constant.

Why are high-momentum tracks harder to measure, not easier?

Because the quantity you actually measure — the sagitta, the bulge of the arc away from a straight line — shrinks in inverse proportion to momentum. A 1 GeV/c track curls sharply and its sagitta over a metre is centimetres; a 1 TeV/c track is almost perfectly straight, bulging only a fraction of a millimetre. Since your sensors have a fixed position resolution of tens of micrometres, that fixed error is a larger and larger fraction of an ever-smaller signal. This is why the fractional resolution grows linearly with $p_{\mathrm T}$ , and why at the TeV scale a tracker eventually can barely do better than determine the sign of the charge.

What sets the momentum-independent resolution floor?

Multiple Coulomb scattering in the detector material. As the particle crosses each layer of silicon, gas, cooling, and support, small-angle nuclear scatters randomly deflect it, and the tracker cannot distinguish a deflection caused by the magnetic field from one caused by a scatter. The scattering angle scales as $1/p$ and so does the curvature signal, so their ratio — the fractional momentum error — is independent of momentum. It depends only on the field, the lever arm, and the amount of material measured in radiation lengths $X_0$ . This is why LHC trackers obsess over being "low-mass": every gram of material you remove lowers the floor that no readout electronics can beat.

What is the difference between transverse momentum and total momentum here?

The magnetic field points along the beam axis, so it only bends the component of motion in the transverse plane; the component along the beam coasts undisturbed. The curvature of the circular projection therefore measures only $p_{\mathrm T}$ , the transverse part. To get the total momentum you combine it with the helix's pitch angle $\lambda$ (equivalently the polar angle $\theta$ ) through $p=p_{\mathrm T}/\cos\lambda=p_{\mathrm T}/\sin\theta$ . In collider physics $p_{\mathrm T}$ is often the more useful quantity anyway, because the interesting products of a hard collision emerge with large transverse momentum while the beam remnants stream down the pipe — which is exactly why detectors are built cylindrically around the beam.

Gaseous chambers or silicon — which is better?

They trade different things. Gaseous detectors — drift chambers and time projection chambers — instrument a huge volume cheaply with very little material (a low scattering floor) and can give three-dimensional tracking, but their point resolution is coarser (tens to hundreds of micrometres) and they are slower. Silicon pixels and strips give superb point resolution (down to $\sim10\ \mu$ m) and fast, radiation-hard readout, ideal for pinpointing decay vertices, but they are costly and add material. Modern experiments use both: silicon in the innermost layers where vertex precision matters most, and either silicon or a large gaseous tracker further out for the long lever arm. ALICE's TPC and CMS's all-silicon tracker are the two extremes of this choice.

How does the tracker find the collision point and displaced decay vertices?

By extrapolating fitted tracks back toward the beam line. Each track's helix, continued inward, points to where the particle was produced; the common intersection of many tracks marks the primary vertex, located to tens of micrometres by the innermost pixel layers. A b hadron lives long enough to travel a fraction of a millimetre before decaying, so its daughter tracks trace back not to the primary vertex but to a slightly displaced secondary one. Resolving that offset — the impact parameter — is what lets an experiment "tag" b jets, a technique central to Higgs studies and to searches for new heavy particles.