The discovery of the Higgs boson

The Higgs boson was the last unfilled seat in the Standard Model, and for forty-eight years it was a prediction with no address. Symmetry-breaking demanded a scalar field to give the W, Z, and fermions their mass, but the theory refused to say what the quantum of that field weighed — only that if it existed, a collider of enough energy and enough luminosity would eventually shake one loose. By 2010 the search had been squeezed into a narrow window: LEP's direct hunt had pushed the mass above 114 GeV, while the size of the quantum loop corrections that the Higgs feeds into precision electroweak observables pulled it back below about 160 GeV. The particle, if real, was cornered. The Large Hadron Collider was built to enter that corner. Colliding protons at $\sqrt{s}=7$ and $8$ TeV, it manufactured Higgs bosons chiefly through a quantum loop of virtual top quarks — the largest production rate the Standard Model allows, yet still buried a billion-fold beneath ordinary strong-interaction debris. The discovery could not come from the biggest signal; it came from the cleanest. In two rare but kinematically pristine channels — the diphoton peak $H\to\gamma\gamma$ and the four-lepton "golden channel" $H\to ZZ^*\to 4\ell$ — the Higgs mass shows up as a sharp bump standing over a smooth background. On 4 July 2012 ATLAS and CMS each announced a new boson near 125 GeV at the $5\sigma$ gold standard. This article is where a symmetry-breaking scalar, a top-quark loop, and a statistical bump become one particle.

Key takeaways

Before the LHC the Higgs was boxed in: LEP direct searches required $m_H>114$ GeV, and electroweak loop corrections implied $m_H\lesssim160$ GeV.
The dominant production is gluon–gluon fusion through a virtual top-quark loop, because the Higgs coupling grows with mass and the top is the heaviest quark; the rate is folded over the gluon PDFs.
About $N=\sigma L\approx4\times10^{5}$ Higgs bosons were made, but $\sigma_H$ is $\sim10^{-10}$ of the total pp cross section, so almost every decay drowns in QCD backgrounds.
Discovery came from the rare, clean channels $H\to\gamma\gamma$ ( $\mathrm{BR}\approx0.23\%$ ) and $H\to ZZ^*\to4\ell$ , where the mass is reconstructed event-by-event as a peak over a smooth background.
ATLAS ( $5.9\sigma$ , $126.0\pm0.6$ GeV) and CMS ( $5.0\sigma$ , $125.3\pm0.6$ GeV) announced the boson on 4 July 2012; the combined mass is $m_H=125.09\pm0.24$ GeV.
Key idea: because the Higgs couples to mass, it is copiously produced through a top-quark loop in gluon fusion, yet it can only be cleanly extracted in rare, kinematically distinctive decays where its mass appears as a peak in the invariant mass of its decay products over a small background.

The core insight

Because the Higgs couples to mass, it is copiously produced through a virtual top-quark loop in gluon fusion, yet it can only be cleanly extracted in rare, kinematically distinctive decay channels — $H\to\gamma\gamma$ and $H\to ZZ^*\to4\ell$ — where its mass appears as a peak in the invariant mass of its decay products standing over a small, smoothly falling background.

Where this lives today

The 2012 discovery opened, rather than closed, the Higgs program. The immediate task was to prove the new boson was the Standard-Model Higgs and not an impostor: its spin-parity was pinned to $0^+$ (a spin-1 state is forbidden from decaying to two photons by the Landau–Yang theorem, and the $ZZ^*\to4\ell$ angular distributions ruled out $0^-$ and $2^+$ ), and its couplings were measured channel by channel through the signal-strength $\mu=\sigma/\sigma_{\text{SM}}$ , which sits at $\mu\approx1.0$ across all major modes. By 2018 both experiments had observed the dominant but difficult $H\to b\bar b$ decay and the $t\bar tH$ production mode, directly confirming the top Yukawa coupling that drives gluon fusion; the rare $H\to\mu\mu$ decay, testing the coupling to a second-generation fermion, reached the evidence level around 2020. The current world-average mass is $m_H=125.25\pm0.17$ GeV. The frontier now is the Higgs self-interaction — the shape of the potential that broke the symmetry in the first place — accessible only through the tiny rate of di-Higgs production, a flagship goal of the High-Luminosity LHC, which aims for $3000\ \mathrm{fb^{-1}}$ and will manufacture over a hundred million Higgs bosons. Beyond that, proposed "Higgs factories" — the FCC-ee, ILC, CEPC, and CLIC — would produce Higgs bosons in the clean $e^+e^-\to ZH$ mode, measuring the couplings to sub-percent precision and searching for the small deviations that new physics would imprint. The particle that completed the Standard Model has become the sharpest tool for finding what lies beyond it.

The physical picture

Two protons cross at the interaction point, but the collision that matters is between their constituents: a gluon from one proton and a gluon from the other. Gluons are massless and the Higgs does not couple to massless particles, so the two cannot meet directly. Instead they annihilate into a triangular loop of virtual top quarks — the heaviest fermion, and therefore the one with the strongest Higgs coupling — and the loop radiates a Higgs. This loop-induced process, gluon–gluon fusion, supplies roughly 87% of all Higgs production at the LHC. The Higgs then lives for about $10^{-22}$ s and decays, most often to the heaviest pair it can reach, $b\bar b$ , but occasionally into the clean channels that give it away. The first figure traces production through a top loop into the branching fractions that decide where to look.

Figure

Higgs production and decay at m_H = 125 GeV. Two gluons fuse through a virtual top-quark loop to make a Higgs with cross section σ(gg → H) ≈ 19 pb at √s = 8 TeV. The boson then decays with the branching fractions shown — dominated by b b̄ (58%) and W W (21%), while the discovery channels ZZ (2.6%) and γγ (0.23%) are rare but clean

Producing the Higgs is only half the problem; finding it is the other half. Roughly $4\times10^{5}$ Higgs bosons were made in the 2011–2012 run, but the LHC recorded some $10^{15}$ inelastic proton collisions in the same period, so a Higgs is one event in a billion. The dominant $b\bar b$ decay is hopeless — it hides beneath an ocean of ordinary QCD jets that produce $b$ quarks a million times more often. The winning strategy inverts the usual instinct: ignore the largest signal and chase the cleanest. In $H\to\gamma\gamma$ the two photons carry the full Higgs energy, and their invariant mass reconstructs to a sharp spike over a smooth diphoton continuum. The second figure is the actual discovery plot — the diphoton mass peak that, with the four-lepton channel, crossed the $5\sigma$ line.

Figure

The diphoton invariant-mass spectrum m_γγ near the discovery. A smooth, steeply falling background from ordinary QED and QCD diphoton production (amber dashed) carries a narrow excess of ≈ 500 signal events (blue) at m_γγ ≈ 125 GeV. The peak width of ≈ 1.7 GeV is the detector resolution, not the natural width Γ_H ≈ 4 MeV — combined with H → ZZ* → 4ℓ this excess reached 5σ

The core equations

The rate of any process at a collider factorises into a machine number and a physics number. The number of Higgs bosons made is the production cross section times the integrated luminosity delivered by the LHC.

N_H=\sigma(pp\to H+X)\int\!\mathcal{L}\,\mathrm{d}t=\sigma\,L_{\text{int}} \qquad (1)

where: $\sigma$ — production cross section ( $\approx22$ pb at $\sqrt{s}=8$ TeV); $L_{\text{int}}=\int\mathcal{L}\,\mathrm{d}t$ — integrated luminosity (in $\mathrm{fb^{-1}}$ , with $1\ \mathrm{fb^{-1}}=10^{3}\ \mathrm{pb^{-1}}$ ); $N_H$ — number of Higgs bosons produced. Reading it: a fixed cross section turns delivered luminosity directly into event count, which is why "more $\mathrm{fb^{-1}}$ " is the entire currency of a discovery run. The production cross section itself is the partonic gluon-fusion rate folded over the probability of finding two gluons in the colliding protons — a partonic cross section convolved with the gluon parton distributions.

\sigma(pp\to H)=\sigma_0\,\tau_H\,\frac{\mathrm{d}\mathcal{L}^{gg}}{\mathrm{d}\tau_H},\qquad \sigma_0=\frac{G_F\,\alpha_s^2}{288\sqrt{2}\,\pi}\Big|\textstyle\sum_q A_{1/2}(\tau_q)\Big|^2 \qquad (2)

where: $\tau_H=m_H^2/s$ — the fraction of $\sqrt{s}$ the Higgs must carry; $\mathrm{d}\mathcal{L}^{gg}/\mathrm{d}\tau_H$ — the gluon–gluon parton luminosity; $G_F$ — Fermi constant; $\alpha_s$ — strong coupling; $A_{1/2}(\tau_q)$ — the quark-loop form factor, which tends to $4/3$ for a heavy quark and to zero for a light one. Reading it: only the top quark contributes meaningfully to the loop sum, so Higgs production is, at its root, a top-quark quantum effect. Once produced, the Higgs decays; its coupling to a fermion pair is proportional to the fermion mass, which fixes the partial widths and hence the branching fractions.

\Gamma(H\to f\bar f)=\frac{N_c\,G_F\,m_H\,m_f^2}{4\sqrt{2}\,\pi}\left(1-\frac{4m_f^2}{m_H^2}\right)^{3/2} \qquad (3)

where: $N_c$ — colour factor ( $3$ for quarks, $1$ for leptons); $m_f$ — fermion mass; the factor $(1-4m_f^2/m_H^2)^{3/2}$ — the $p$ -wave phase-space threshold. Reading it: the width scales as $m_f^2$ , so the Higgs decays preferentially to the heaviest kinematically accessible pair — $b\bar b$ first, then $\tau\tau$ and $c\bar c$ . The fraction of Higgs bosons decaying into a given channel is that channel's partial width over the sum of all of them.

\mathrm{BR}(H\to i)=\frac{\Gamma_i}{\Gamma_H},\qquad \Gamma_H=\sum_i\Gamma_i\approx4.1\ \mathrm{MeV} \qquad (4)

where: $\Gamma_i$ — partial width into channel $i$ ; $\Gamma_H$ — total width, remarkably small for a 125 GeV particle. Reading it: because $\Gamma_H\approx4$ MeV is $10^{-5}$ of $m_H$ , the Higgs is extraordinarily narrow — any observed peak width is set by the detector, not the particle. In the mass-reconstructible channels the Higgs mass is read directly from the four-momenta of its decay products through the invariant-mass relation; for two photons this reduces to their energies and opening angle.

m_{\gamma\gamma}=\sqrt{\Big(\textstyle\sum_i E_i\Big)^2-\Big(\textstyle\sum_i \vec p_i\Big)^2}=\sqrt{2E_1E_2\,(1-\cos\theta_{12})} \qquad (5)

where: $E_1,E_2$ — photon energies; $\theta_{12}$ — angle between them; the second form uses $|\vec p|=E$ for massless photons. Reading it: every signal event drops a point at the same $m_{\gamma\gamma}=m_H$ , while background events smear across the spectrum — a resonance becomes a peak. Finally, an excess is only a discovery once it is unlikely to be a fluctuation; the significance compares signal $S$ to the statistical noise of the background $B$ .

Z=\frac{S}{\sqrt{B}}\ \xrightarrow{\text{exact}}\ Z=\sqrt{2\big[(S+B)\ln(1+S/B)-S\big]},\qquad 5\sigma\leftrightarrow p=2.87\times10^{-7} \qquad (6)

where: $S$ — expected signal events; $B$ — expected background under the peak; $Z$ — significance in Gaussian standard deviations; $p$ — probability the background alone fakes the excess. Reading it: the $5\sigma$ convention (one chance in $3.5$ million) is the particle-physics threshold that separates "evidence" from "discovery".

Derivation AWhy the Higgs is born in a loop of top quarks

The goal is to explain why the largest Higgs production mechanism at a proton collider is not a direct coupling at all but a quantum loop, and why that loop is dominated by the single heaviest quark. We start from the fact that the Higgs coupling is proportional to mass, note that gluons are massless and so cannot couple to the Higgs directly, build the effective gluon–Higgs interaction from a virtual quark triangle, and show that the top quark alone survives the loop sum — turning production into a probe of the top Yukawa coupling.

Step 1. The Higgs couples to every massive particle with a strength set by that particle's mass: to fermions as $y_f=m_f/v$ (with $v=246$ GeV the vacuum expectation value), and to the massless gluon not at all at tree level. So a direct $ggH$ vertex does not exist — production must proceed through particles that do couple. Why: the entire logic of Higgs phenomenology follows from "couples to mass", and it immediately forbids the naive tree-level diagram.

Step 2. Two incoming gluons can each attach to a virtual quark, and the quark line can emit a Higgs, forming a closed triangular loop $gg\to(q\text{-loop})\to H$ . Summing over the quark flavours in the loop gives a form factor $\sum_q A_{1/2}(\tau_q)$ with $\tau_q=m_H^2/4m_q^2$ . Why: the loop is the lowest-order way to connect massless gluons to the Higgs, and every quark flavour circulates in it, weighted by its coupling.

Step 3. Evaluate the form factor. Each quark enters with a Higgs coupling $\propto m_q$ , but a heavy quark ( $m_q\gg m_H/2$ ) contributes $A_{1/2}\to4/3$ — a constant — while a light quark ( $m_q\ll m_H/2$ ) contributes $A_{1/2}\to0$ . The top, with $m_t\approx173$ GeV, dominates; the bottom adds a few percent and everything else is negligible. Why: the coupling growth with mass exactly compensates the loop's decoupling of heavy states, leaving a finite top-dominated amplitude — a non-decoupling that makes the loop a direct window on $m_t$ .

Step 4. In the heavy-top limit the loop shrinks to a point and the interaction becomes the effective operator $\mathcal{L}_{\text{eff}}=\dfrac{\alpha_s}{12\pi}\dfrac{H}{v}\,G^a_{\mu\nu}G^{a\mu\nu}$ , whose square gives the partonic cross section $\sigma_0$ of equation (2). Convolving with the steeply-rising gluon PDFs at the small $\tau_H=m_H^2/s\approx2.4\times10^{-4}$ (at 8 TeV) yields $\sigma(gg\to H)\approx19$ pb. Why: the point-like effective vertex is what makes gluon fusion computable and large, and the abundance of low- $x$ gluons in the proton is what makes it the leading channel. *Verified: $\tau_H=(125/8000)^2=2.44\times10^{-4}$ ; the heavy-quark form factor $A_{1/2}\to4/3$ and light-quark $\to0$ ; the effective $ggH$ coupling scales as $\alpha_s/12\pi v$ , reproducing $\sigma_{gg\to H}\approx19$ pb at 8 TeV.*

Go deeper · Derivation B · 2 minCoupling to mass fixes the branching fractions — and why γγ is rare but golden

Derivation BCoupling to mass fixes the branching fractions — and why γγ is rare but golden

The aim is to derive the pattern of Higgs decays from the single principle that the coupling grows with mass, to show why $b\bar b$ dominates the branching fractions yet the discovery came from $\gamma\gamma$ and $ZZ^*$ , and to make explicit the trade of rate against cleanliness that decides where to search. We compute the fermionic partial widths, compare them, add the loop-induced diphoton channel, and weigh each channel's branching fraction against its background.

Step 1. From equation (3), $\Gamma(H\to f\bar f)\propto N_c\,m_f^2$ . For the heaviest open pair, $b\bar b$ , this is large; comparing to $\tau^+\tau^-$ gives $\Gamma_{b\bar b}/\Gamma_{\tau\tau}=N_c(m_b/m_\tau)^2=3\times(2.8/1.777)^2\approx7.4$ , close to the measured branching ratio $0.577/0.063\approx9$ (the residual difference is QCD running of $m_b$ ). Why: the $m_f^2$ scaling alone predicts the hierarchy of fermionic decays, confirming "couples to mass" quantitatively.

Step 2. The vector-boson channels $H\to WW^*$ and $H\to ZZ^*$ are open even below threshold because one boson is virtual; together they take about $24\%$ of decays. The $ZZ^*\to4\ell$ sub-channel is tiny — $\mathrm{BR}(ZZ^*)\times\mathrm{BR}(Z\to\ell\ell)^2\approx1.2\times10^{-4}$ for $\ell=e,\mu$ — but it is nearly background-free. Why: a rare channel with almost no background can outweigh a common channel drowning in QCD; branching fraction is not the figure of merit — signal over background is.

Step 3. The diphoton channel is loop-induced, like production: $H\to\gamma\gamma$ proceeds through a $W$ loop (dominant) interfering destructively with a top loop, giving $\mathrm{BR}\approx0.23\%$ . Despite this small rate, both photons are measured with per-mille energy resolution, so the reconstructed $m_{\gamma\gamma}$ peak is sharp. Why: excellent mass resolution concentrates the signal into a few bins while the background spreads out, converting a small rate into a visible peak.

Step 4. Rank the channels by discovery power, not by rate. $b\bar b$ (58%) is swamped by QCD; $WW^*$ (21%) has two neutrinos and no mass peak, giving only a broad excess; $\gamma\gamma$ (0.23%) and $ZZ^*\to4\ell$ ( $10^{-4}$ ) are rare but reconstruct a sharp $m_H$ . The last two carried the discovery. Why: the article's central irony — the biggest signals are unusable and the discovery rides on the rarest clean ones — falls straight out of comparing rate against background. *Verified: $\Gamma_{b\bar b}/\Gamma_{\tau\tau}=3(2.8/1.777)^2=7.4$ vs measured $\approx9$ ; tree-level $\Gamma(H\to b\bar b)=3G_Fm_H m_b^2/(4\sqrt2\pi)=1.9$ MeV with running $m_b(m_H)\approx2.8$ GeV; $\mathrm{BR}(ZZ^*\to4\ell)\approx1.2\times10^{-4}$ .*

Go deeper · Derivation C · 1 minFrom a cross section to a five-sigma discovery

Derivation CFrom a cross section to a five-sigma discovery

The purpose is to follow a single number — the production cross section — all the way to the statement "a new particle at $5\sigma$ ", so that the leap from "we made $4\times10^{5}$ Higgs bosons" to "the background cannot fake this" is made quantitative. We count the produced Higgs, filter through branching fraction and detector efficiency to a handful of signal events, estimate the background beneath the peak, and form the significance that defines a discovery.

Step 1. Count production with equation (1): $N_H=\sigma L_{\text{int}}=(22\ \mathrm{pb})(20\ \mathrm{fb^{-1}})=(22)(2\times10^{4})\approx4.4\times10^{5}$ Higgs bosons in the combined 7+8 TeV sample. Why: this sets the scale — hundreds of thousands of Higgs bosons exist in the data before any selection.

Step 2. Filter to the diphoton channel: multiply by $\mathrm{BR}(\gamma\gamma)=0.0023$ to get $\approx1000$ diphoton pairs produced, then by a detector acceptance $\times$ efficiency $\varepsilon\approx0.4$ to get $S\approx400$ reconstructed signal events sitting inside the mass peak. Why: only a fraction of the produced Higgs bosons decay to photons, and only a fraction of those are cleanly measured — the usable signal is small.

Step 3. Estimate the background beneath the peak: the smooth QED/QCD diphoton continuum contributes roughly $B\approx10^{4}$ events in the $\pm2$ GeV window around 125 GeV, read directly from the sidebands. The Gaussian significance is $Z=S/\sqrt{B}=400/\sqrt{10^{4}}=400/100\approx4$ . Why: the background is measured, not modelled from theory, so the significance rests on data — but a single channel in one experiment falls just short of discovery.

Step 4. Combine. Add the near-background-free $ZZ^*\to4\ell$ channel and the second experiment: two independent channels in two independent detectors each near $3$ – $4\sigma$ combine in quadrature past the threshold, and indeed ATLAS reported $5.9\sigma$ and CMS $5.0\sigma$ . A $5\sigma$ excess corresponds to a background fluctuation probability $p=2.87\times10^{-7}$ — one in $3.5$ million. Why: independent evidence adds in quadrature, so two moderate signals become one overwhelming one, which is exactly how the 4 July 2012 discovery was assembled. *Verified: $N_H=22\times2\times10^4=4.4\times10^5$ ; $S=4.4\times10^5\times0.0023\times0.4\approx400$ ; $Z=400/\sqrt{10^4}=4.0$ ; exact $Z=\sqrt{2[(S+B)\ln(1+S/B)-S]}=3.97$ for $S=400,B=10^4$ ; $5\sigma\leftrightarrow p=2.87\times10^{-7}$ .*

Limiting cases

Case	Result	Meaning
Very light $m_H$ (LEP era)	excluded, $m_H>114.4$ GeV	direct $e^+e^-\to ZH$ search set the lower edge of the window
Realised case $m_H=125$ GeV	$b\bar b$ dominant, discovery via $\gamma\gamma$ and $ZZ^*\to4\ell$	the Standard-Model Higgs, in the hardest region to find
Heavy Higgs $m_H>2m_Z$	$H\to ZZ\to4\ell$ "golden", both $Z$ on-shell	would have been an easy discovery — but this region was excluded by the electroweak fit
Intrinsic width	$\Gamma_H\approx4.1\ \mathrm{MeV}\ll1$ – $2$ GeV resolution	the observed peak width is instrumental, not the particle's
No top Yukawa	gluon-fusion loop $\to0$	production is a pure quantum-loop effect; remove the top and the main channel vanishes

Common trapThe gluon-fusion channel is the discovery channel

It is not, and believing so inverts the actual logic of the search. Gluon fusion has by far the largest production cross section — about $87\%$ of all Higgs bosons are made this way — but production and discovery are different questions. A Higgs made in gluon fusion still has to decay, and its dominant decay, $H\to b\bar b$ , produces two jets that are utterly indistinguishable from the millions of ordinary QCD jet pairs the LHC makes every second. The signal-to-background ratio in inclusive $gg\to H\to b\bar b$ is hopeless. The discovery instead relied on the rare but clean decays: $H\to\gamma\gamma$ (branching fraction only $0.23\%$ ) and $H\to ZZ^*\to4\ell$ (an effective branching fraction of $\sim10^{-4}$ ), where the final state is so distinctive that the tiny background can be measured from the data sidebands and the Higgs mass reconstructs as a sharp peak. The lesson generalises: at a hadron collider the figure of merit is never the raw rate but $S/\sqrt{B}$ , and a channel a thousand times rarer can be a thousand times more discoverable if its background is small enough. Production picks the biggest door; discovery walks through the cleanest one.

Worked examples

1 · Counting Higgs bosons and reaching five sigma (confidence builder)

Take a combined 7+8 TeV sample with production cross section $\sigma\approx22$ pb and integrated luminosity $L_{\text{int}}\approx20\ \mathrm{fb^{-1}}$ . Estimate how many Higgs bosons were made, how many diphoton signal events survive, and the resulting significance.

Step 1. Produced Higgs bosons from equation (1), with $1\ \mathrm{fb^{-1}}=10^{3}\ \mathrm{pb^{-1}}$ : $N_H=\sigma L_{\text{int}}=(22\ \mathrm{pb})(20\times10^{3}\ \mathrm{pb^{-1}})=4.4\times10^{5}$ . About 440,000 Higgs bosons were produced.

Step 2. Decay to photons and reconstruct: multiply by $\mathrm{BR}(\gamma\gamma)=0.0023$ to get $\approx1000$ diphoton pairs, then by acceptance $\times$ efficiency $\varepsilon\approx0.4$ : $S=4.4\times10^{5}\times0.0023\times0.4\approx\mathbf{400}$ signal events in the peak.

Step 3. The smooth diphoton background in the $\pm2$ GeV window around 125 GeV is measured from the sidebands as $B\approx10^{4}$ events. The significance is $Z=S/\sqrt{B}=400/\sqrt{10^{4}}=400/100=\mathbf{4.0}$ ; the exact Asimov formula gives $Z=\sqrt{2[(S+B)\ln(1+S/B)-S]}=3.97$ , essentially identical.

Step 4. A single channel falls just short of $5\sigma$ . Adding the near-background-free $ZZ^*\to4\ell$ channel and combining ATLAS with CMS pushes the combined significance past the threshold. A $5\sigma$ result means the probability of a background-only fluctuation is $p=2.87\times10^{-7}$ — one in $3.5$ million. One channel in one experiment reaches $\sim4\sigma$ ; combining channels and experiments crosses $5\sigma$ .

2 · Reconstructing the Higgs mass from a photon pair

An event contains two photons with energies $E_1=70$ GeV and $E_2=60$ GeV, separated by an opening angle $\theta_{12}=149°$ . Compute the diphoton invariant mass and identify the particle.

Step 1. For massless photons the invariant mass is $m_{\gamma\gamma}=\sqrt{2E_1E_2(1-\cos\theta_{12})}$ (equation 5). Evaluate the cosine: $\cos149°=-0.857$ , so $1-\cos\theta_{12}=1.857$ .

Step 2. Insert the energies: $m_{\gamma\gamma}^2=2(70)(60)(1.857)=8400\times1.857=1.560\times10^{4}\ \mathrm{GeV^2}$ . Take the root: $m_{\gamma\gamma}=\sqrt{1.560\times10^{4}}=\mathbf{124.9\ GeV}$ .

Step 3. This single event lands a point at $\approx125$ GeV. Background diphoton events, having no parent resonance, scatter their $m_{\gamma\gamma}$ smoothly across the spectrum; only signal events pile up at the same mass. Accumulate thousands of events and the Higgs announces itself as a peak at $m_H\approx125$ GeV over a smooth continuum. The pair reconstructs to 125 GeV — a Higgs candidate.

3 · Coupling to mass, read off the branching fractions

Verify that the Higgs coupling scales with fermion mass by predicting the ratio of partial widths $\Gamma(H\to b\bar b)/\Gamma(H\to\tau^+\tau^-)$ and comparing to the measured branching-fraction ratio.

Step 1. From equation (3) the phase-space factor is $\approx1$ for both light fermions, so $\Gamma(H\to f\bar f)\propto N_c\,m_f^2$ . The ratio is $\dfrac{\Gamma_{b\bar b}}{\Gamma_{\tau\tau}}=\dfrac{N_c^b\,m_b^2}{N_c^\tau\,m_\tau^2}=\dfrac{3\,m_b^2}{1\cdot m_\tau^2}$ .

Step 2. Insert the running masses at the Higgs scale, $m_b(m_H)\approx2.8$ GeV and $m_\tau=1.777$ GeV: $\dfrac{\Gamma_{b\bar b}}{\Gamma_{\tau\tau}}=3\left(\dfrac{2.8}{1.777}\right)^2=3\times2.48=\mathbf{7.4}$ .

Step 3. Compare to experiment: the measured branching fractions are $\mathrm{BR}(b\bar b)=0.577$ and $\mathrm{BR}(\tau\tau)=0.063$ , giving a ratio $0.577/0.063\approx9.1$ . The prediction $7.4$ agrees to within the QCD-correction level, confirming that the coupling scales as $m_f^2$ with the colour factor $N_c=3$ for quarks. The $m_f^2$ scaling reproduces the observed hierarchy — the Higgs couples to mass.

Physics of the portrait

The discovery of the Higgs boson is one of the rare cases where a single theoretical prediction, made when the predictor was thirty-five, was confirmed in his presence forty-eight years later. Peter Higgs sat in the CERN auditorium on 4 July 2012 and wept.

Origins · 1 minPeter Higgs and how this was discovered

Peter Higgs

1929–2024 · the 1964 prediction of a massive scalar boson and its 2012 confirmation

Peter Higgs was the British theoretical physicist whose name attached, half by accident, to the mechanism that gives elementary particles their mass. In 1964, working at the University of Edinburgh, he wrote two short papers showing how a scalar field with a non-zero vacuum value could break a gauge symmetry while leaving the theory consistent — and, crucially, he pointed out in a sentence added at a referee's prompting that the mechanism implied a new massive spin-0 particle, an excitation of the field itself. Robert Brout and François Englert had reached the mechanism independently and slightly earlier, as had Gerald Guralnik, Carl Hagen, and Tom Kibble; but Higgs alone spelled out the boson, and so the particle carries his name. For decades it was a theorist's bookkeeping device that many doubted would ever be seen: too light and LEP would have found it, too heavy and it would have spoiled the precision electroweak fits. The 2012 discovery at exactly 125 GeV — inside the narrow window the theory allowed — was the vindication of a lifetime of quiet conviction. Higgs, famously modest and press-shy, shared the 2013 Nobel Prize in Physics with Englert (Brout having died in 2011, and the prize not awarded posthumously). He disliked the tabloid nickname "the God particle" and insisted the physics belonged to many hands. His deeper point outlasts the celebration: the Higgs field is not just one more particle but the reason the vacuum is not empty — a condensate filling all of space, without which electrons would be massless and atoms could not exist.

History

1964

— Englert–Brout and Higgs predict a scalar field that gives particles mass

1990

— LEP direct searches begin; they will set m_H > 114.4 GeV

2008

— the LHC starts up as a dedicated Higgs-hunting machine

2012

— ATLAS (5.9σ) and CMS (5.0σ) announce a 125 GeV boson on 4 July

2013

— Higgs and Englert receive the Nobel Prize in Physics

Going further

Texts. Thomson, Modern Particle Physics, §17.7 (the discovery of the Higgs boson) · Halzen & Martin, Quarks and Leptons, Ch. 15 (the Higgs mechanism) · Griffiths, Introduction to Elementary Particles, 2nd ed., §11 · Peskin & Schroeder, An Introduction to Quantum Field Theory, §20.1 (Higgs mechanism) and §22 · Schwartz, Quantum Field Theory and the Standard Model, Ch. 29 · Djouadi, The Anatomy of Electroweak Symmetry Breaking, Vol. I (Higgs production and decay).

Provenance. The mass-generating scalar and its couplings are owned by aq-pp-higgs-mechanism and aq-pp-spontaneous-symmetry-breaking; the event-rate relation $N=\sigma L$ comes from aq-pp-luminosity-event-rates; the gluon parton luminosity from aq-pp-parton-distributions; and the loop constraints that fixed the mass window from aq-pp-electroweak-radiative-corrections. The measured parameters (ATLAS $126.0\pm0.6$ GeV at $5.9\sigma$ ; CMS $125.3\pm0.6$ GeV at $5.0\sigma$ ; combined $m_H=125.09\pm0.24$ GeV, world average $125.25\pm0.17$ GeV) are the 2012 discovery papers (Aad et al., Phys. Lett. B 716 (2012) 1; Chatrchyan et al., Phys. Lett. B 716 (2012) 30) and the ATLAS+CMS combination. A verifier should confirm: (1) $N_H=\sigma L\approx4\times10^{5}$ from $\sigma=22$ pb, $L=20\ \mathrm{fb^{-1}}$ ; (2) $m_{\gamma\gamma}=124.9$ GeV from $E_1=70$ , $E_2=60$ GeV, $\theta=149°$ ; (3) $\Gamma_{b\bar b}/\Gamma_{\tau\tau}=3(m_b/m_\tau)^2=7.4$ ; (4) $5\sigma\leftrightarrow p=2.87\times10^{-7}$ .

Prerequisites. The Higgs Mechanism · Spontaneous Symmetry Breaking · Luminosity and Event Rates · Parton Distributions · Electroweak Radiative Corrections. Leads to. Higgs Couplings · Supersymmetry · Open Questions of the Standard Model.

Ask the tutor: Ask me to walk through why gluon fusion runs through a top-quark loop, or to work any discovery-significance number from a cross section, branching fraction, and luminosity.

If the Higgs is produced mostly by gluon fusion, why didn't they just look at that?

They did look at gluon-fusion production — it is where almost all the Higgs bosons come from — but production and detection are separate steps. A Higgs made in gluon fusion still decays, and its most common decay, $H\to b\bar b$ , yields two jets that look exactly like the enormous background of ordinary QCD di-jet events. There is no way to tell a Higgs $b\bar b$ pair from the millions of non-Higgs ones. So even though gluon fusion supplies the signal, you can only see it in decay channels clean enough to stand out. Both discovery channels, $\gamma\gamma$ and $ZZ^*\to4\ell$ , are fed mainly by gluon-fusion production — the point is that they are identified by their distinctive, low-background decays, not by how the Higgs was made.

What exactly does a "5σ" discovery mean?

It is a statement about how unlikely the background alone is to produce the observed excess. If there were no Higgs, the number of events in the signal region would still fluctuate statistically around the expected background. A $5\sigma$ excess means the observed number is five standard deviations above that background expectation — a fluctuation that would happen by chance with probability $p=2.87\times10^{-7}$ , about one in $3.5$ million. Particle physics adopted this stringent threshold because searches scan many mass values and many channels, so moderate excesses appear all the time by chance; demanding $5\sigma$ guards against calling a fluctuation a discovery. It is not a claim that the signal is five times the background, nor a measurement error — it is purely the probability that "nothing there" could mimic what was seen.

Why is the measured peak so much wider than the Higgs's natural width?

The Standard-Model Higgs has a total width of only $\Gamma_H\approx4.1$ MeV, because its couplings are modest and $\Gamma\sim m^3$ -type suppression keeps the decay rate small for a 125 GeV particle. That intrinsic width corresponds to a lifetime of about $1.6\times10^{-22}$ s. But the diphoton peak in the data is about $1$ – $2$ GeV wide — hundreds of times broader. The difference is entirely instrumental: the width you see is the detector's energy resolution smearing each photon's measured energy, not the particle's true width. To actually resolve $\Gamma_H$ you would need a mass measurement precise to a few MeV, far beyond a calorimeter's reach. This is why the Higgs width is instead inferred indirectly, from combinations of on-shell and off-shell rates, rather than read off the peak.

How could the discovery mass be so close to what the theory expected?

The Higgs mass is a free parameter of the Standard Model — the theory does not predict it directly. But the Higgs contributes to quantum loop corrections in other precisely measured quantities, notably the $W$ and $Z$ masses and the weak mixing angle. Fitting all of those measurements together constrained $m_H$ to lie below about $160$ GeV, with a best fit near $90$ GeV, while LEP's failure to find it directly required $m_H>114.4$ GeV. The allowed window was therefore a narrow band roughly between $114$ and $160$ GeV. Finding the boson at $125$ GeV, squarely inside that band, was a striking self-consistency check: the same theory that needed a Higgs also correctly forecast roughly where it had to be. Had it turned up at, say, $300$ GeV, the electroweak fit would have been in serious trouble.

Why did it take two experiments to announce the discovery?

ATLAS and CMS are two independent detectors, built by separate collaborations with different technologies, analysing the same LHC collisions without sharing results before publication. Requiring both to see the same signal guards against a mistake or a subtle bias in one apparatus faking a particle. When two independently designed experiments, each blind to the other's numbers, both find an excess at the same mass and near the same significance, the case becomes overwhelming — a detector artefact would have to be reproduced by chance in two different machines. That is why the 4 July 2012 seminar presented ATLAS and CMS back to back, and why the community treated the coincidence of two $\sim5\sigma$ results at $125$ – $126$ GeV as decisive rather than either result alone.

Is the discovered particle definitely the Standard-Model Higgs?

It is consistent with the Standard-Model Higgs to within current precision, but "definitely" is stronger than the data yet allow. Everything measured so far fits: the spin-parity is $0^+$ , the couplings to $W$ , $Z$ , top, bottom, and tau scale with mass as predicted, and the overall signal strength is $\mu\approx1$ . But the measurements have uncertainties of order $5$ – $10\%$ , and many extensions of the Standard Model (supersymmetry, composite-Higgs models, extended Higgs sectors) predict a boson that looks almost identical at this precision while differing in the couplings to lighter fermions, in the invisible width, or in the self-coupling. Pinning those down is the mission of the High-Luminosity LHC and proposed Higgs factories. For now the honest statement is: it is a Higgs boson, and it is the Standard-Model one as far as we can currently measure.