Is It a Scalar, and Does It Couple to Mass? · Elementary Particle Physics

The mechanism predicts a functional form rather than a set of numbers, so testing it needs couplings measured across a range of masses — which is why a single plot carries the whole claim.

🎯 Why this matters

These two sections are what retire the phrase “compatible with”. A scalar whose couplings scale as mass is the object the mechanism describes, and nothing else proposed produces both properties at once.

§9.17 established that a boson exists at 125 GeV and that its rates match the Standard Model. Neither fact identifies it. Two properties do, and they are what these two sections measure: that it is a scalar, and that its couplings are proportional to mass.

Both are properties of the mechanism rather than of the particle. Another scalar at 125 GeV could reproduce every rate in §9.17; what it could not reproduce is the coupling pattern §9.12 forces.

§9.18 Down to two hypotheses, then one

Most of the work was already done by the discovery itself:

Bettini p. 422. Only two possibilities survive to the angular analysis.

hypothesis↕	excluded by↕
$J = 1$ , both parities	the observation of $H\to\gamma\gamma$ . A spin-1 particle cannot decay to two photons — the Landau–Yang argument, made once already for the $\pi^0$ in §3.5
most of $J = 2$	the $ZZ^$ decay is not suppressed*. Several $J = 2$ matrix elements carry the CM momentum $q$ to a power greater than 1, and $q$ is small here — so those would be strongly suppressed, and they are not
$J^P = 0^-$	the angular distribution below — and this is the measurement

For $J = 0$ the orbital and spin angular momenta must satisfy $L = S$ , and parity then decides which values are allowed. The book writes the two amplitudes explicitly:

\htmlClass{t-s}{\mathcal{M}_S} = a_1\,\htmlClass{t-d}{\mathbf{e}_1\cdot\mathbf{e}_2} + a_2\,\htmlClass{t-q}{(\mathbf{e}_1\cdot\mathbf{q})(\mathbf{e}_2\cdot\mathbf{q})} \qquad\qquad \htmlClass{t-p}{\mathcal{M}_P} = a_3\,\htmlClass{t-x}{(\mathbf{e}_1\times\mathbf{e}_2)\cdot\mathbf{q}}

(9.132)

Bettini pp. 422–423, Eqs. (9.132) and (9.134). Two amplitudes, one for each parity — and which one nature uses is read off an angle.

Every symbol, one at a time

Hover or tap a symbol above — it lights up in the equation and its meaning, units and type appear here.

⚠️ Why this is not simply §3.5 again — the Z has a longitudinal state

The structure is identical to the $\pi^0\to\gamma\gamma$ analysis of §3.5: two vector particles, two amplitudes of opposite parity, and an observable built from the angle between two decay planes. But one thing has changed, and it is the thing that made most of $J = 2$ hard to exclude.

A real photon is transverse. Its polarization satisfies $\mathbf{e}\cdot\mathbf{q} = 0$ — Eq. (3.23) — so the spin projection along the flight direction can only be ±1, never 0.

A massive vector boson has three polarization states, including the longitudinal one, so $\mathbf{e}\cdot\mathbf{q} \neq 0$ in general. Two consequences follow:

the $a_2$ term exists at all. For two photons it would vanish identically; here it does not, and it must be argued away kinematically rather than by symmetry;
a spin-1 particle can decay to two massive vectors, which is why the Landau–Yang argument works for $\gamma\gamma$ and not for $ZZ^*$ . Excluding $J = 1$ needed the di-photon channel specifically.

So the two analyses share their logic and not their details, and the site is careful to say which is which.

J^P = 0⁺ — parallel polarizations favoured
J^P = 0⁻ — perpendicular polarizations favoured
CMS, 138 fb⁻¹, 118 < m(4ℓ) < 130 GeV

Fig. 9.62 redrawn (CMS). The data follow the solid 0⁺ curve — maxima at φ = 0 and π, where the two decay planes are aligned. Note how much weaker the modulation is than in §3.5: about ±15 % here against ±47 % for the π⁰, which is why this needed 138 fb⁻¹ rather than 112 events.

💡 What this really says — “Offset in phase by π/2” — and what is actually offset

The book closes the section with a sentence that repays unpacking: “Comparison with Fig. 3.1 shows how now the data are offset in phase of π/2.”

The curves are not offset. In both analyses the $0^+$ hypothesis peaks where the two planes are aligned ( $\varphi = 0, \pi$ ) and the $0^-$ hypothesis peaks where they are perpendicular ( $\varphi = \pi/2$ ). The parity–geometry correspondence is the same in both.

What is offset is the data — because the two particles have opposite parity. The $\pi^0$ is a pseudoscalar, so its data sit on the $0^-$ curve, with maxima at $\varphi = 90°$ . The Higgs is a scalar, so its data sit on the $0^+$ curve, with maxima at $0$ and $180°$ . Two datasets, one observable, in antiphase — and the phase is the parity.

That is a satisfying thing to be able to say. The same measurement, made 60 years apart on particles differing in mass by a factor of a thousand, gives opposite answers to the same question, and the answers differ by exactly the half-period that distinguishes even from odd.

One caution the plot makes visible: the modulation here is ±15 %, against ±47 % for the $\pi^0$ . The discrimination per event is three times weaker, which is why 112 bubble-chamber events sufficed in 1963 and 138 fb⁻¹ was needed in 2021.

Finally, a point easy to skip: both amplitudes would be present if parity were violated in the decay. The data show no evidence of that, so within the accuracy of the measurement the decay conserves parity — which is itself a result, given that the $Z$ s doing the decaying belong to the one interaction that violates parity maximally.

§9.19 The couplings, and the one plot that tests the mechanism

Bettini p. 425, citing Workman et al. 2022 (the PDG). See the erratum below on the gg entry.

channel↕	BR▼	measured?↕
$H\to b\bar b$	58.2 %	yes — but only in associated production, where a lepton from the $W$ or $Z$ tags the event out of the QCD background
$H\to WW^*$	21.4 %	yes — a broad bump, not a peak, because of the two neutrinos
$H\to gg$	8.6 %	no — indistinguishable from QCD dijets. And see the erratum
$H\to\tau^+\tau^-$	6.27 %	yes — reconstructed by a kinematic fit, at ~15 % resolution
$H\to c\bar c$	2.89 %	not yet — charm tagging is much harder than beauty tagging
$H\to ZZ^*$	2.62 %	yes — the golden channel
$H\to\gamma\gamma$	0.227 %	yes — the other discovery channel
$H\to\mu^+\mu^-$	0.0218 %	barely — 2σ (ATLAS) and 3σ (CMS), and statistics-limited, so Run 3 will settle it

Erratum — Table 9.2’s entries sum to more than 100 %

Add the eight branching ratios as printed:

21.4 + 58.2 + 8.6 + 6.27 + 2.62 + 2.89 + 0.227 + 0.0218 = \mathbf{100.23\,\%}

That is impossible: these are eight of the Higgs decay channels, not all of them — $H\to Z\gamma$ (0.153 %) and $H\to s\bar s$ (0.034 %) are missing, so the listed subset must sum to less than 100 %, by about 0.19 %.

At least one entry is therefore too large, and the candidate is $H\to gg$ . The table cites Workman et al. 2022 — the PDG — which gives 8.18 % for that channel. Substituting it:

21.4 + 58.2 + \mathbf{8.19} + 6.27 + 2.62 + 2.89 + 0.227 + 0.0218 = 99.82\,\%

and adding the two omitted channels gives 100.01 %. Everything closes.

Nothing on this page depends on it — $H\to gg$ is the one large channel that cannot be measured, being indistinguishable from QCD dijet production. But a table of branching ratios that sums past 100 % is worth catching. Confirmed on the render of PDF p. 443.

The couplings themselves are the real test, and the way the result is presented deserves care because it is easy to misread:

the couplings, and why they fall on one line

v = 246.22
print("Sec. 9.12 predicts TWO different power laws for the raw couplings:")
print("  g_Hff  = m_f / v            proportional to the fermion MASS")
print("  g_HVV  = 2 M_V^2 / v        proportional to the boson mass SQUARED")
print("\nplotting those against mass would give two separate curves.  so CMS")
print("defines REDUCED coupling modifiers that absorb the difference:")
print("  bosons:   y_V = sqrt(kappa_V) x m_V / v")
print("  fermions: y_F =      kappa_F  x m_F / v")
print("\nwith every kappa = 1 (the Standard Model) BOTH reduce to m/v, so the")
print("prediction becomes a SINGLE straight line of slope 1 through the origin.")
print(f"\n  v = {v} GeV\n")
print("  particle     mass (GeV)     y = m/v")
P = [('muon', 0.1056584), ('tau', 1.77686), ('b quark', 4.18),
     ('W', 80.377), ('Z', 91.1875), ('top', 172.69)]
for nm, m in P:
    print(f"  {nm:11s} {m:9.4f}      {m/v:.6f}")
print(f"\n  the plot spans a factor {P[-1][1]/P[0][1]:.0f} in mass -- three decades -- and the")
print( "  measured points sit on that line throughout.")
print("\nwhat the line being straight actually tests:")
print( "  NOT that the couplings are large or small, but that ONE number, v,")
print( "  reproduces six couplings spanning three orders of magnitude.  any")
print( "  scalar could give a 125 GeV peak; only the BEH scalar gives THIS.")
print("\nthe overall agreement:")
print( "  mu = 1.002 +- 0.057   -- the Standard Model, to 5.7%")

prints

Sec. 9.12 predicts TWO different power laws for the raw couplings:
g_Hff  = m_f / v            proportional to the fermion MASS
g_HVV  = 2 M_V^2 / v        proportional to the boson mass SQUARED

plotting those against mass would give two separate curves.  so CMS
defines REDUCED coupling modifiers that absorb the difference:
bosons:   y_V = sqrt(kappa_V) x m_V / v
fermions: y_F =      kappa_F  x m_F / v

with every kappa = 1 (the Standard Model) BOTH reduce to m/v, so the
prediction becomes a SINGLE straight line of slope 1 through the origin.

v = 246.22 GeV

particle     mass (GeV)     y = m/v
muon           0.1057      0.000429
tau            1.7769      0.007217
b quark        4.1800      0.016977
W             80.3770      0.326444
Z             91.1875      0.370350
top          172.6900      0.701365

the plot spans a factor 1634 in mass -- three decades -- and the
measured points sit on that line throughout.

what the line being straight actually tests:
NOT that the couplings are large or small, but that ONE number, v,
reproduces six couplings spanning three orders of magnitude.  any
scalar could give a 125 GeV peak; only the BEH scalar gives THIS.

the overall agreement:
mu = 1.002 +- 0.057   -- the Standard Model, to 5.7%

Standard Model: y = m/υ, one line, slope 1
fermions — y_F = κ_F m_F/υ
bosons — y_V = √κ_V m_V/υ

Fig. 9.64 redrawn. Six couplings, three decades of mass, one straight line — and one parameter, υ = 246 GeV, fixing all of them. The muon point is the newest and weakest: 2σ at ATLAS and 3σ at CMS, statistics-limited, so Run 3 settles it.

⚙️ Engineer’s bridge — a straight line on a log–log plot is a claim about a law, not a fit

It is worth being precise about what Fig. 9.64 tests, because the plot is constructed to make the Standard Model look like a straight line and that could be mistaken for circularity.

What is put in. The definition of the reduced coupling modifiers — one square root for bosons, none for fermions — is deliberately chosen to absorb the known difference between the two power laws of §9.12. That part is bookkeeping.

What is not put in. The vertical positions. Each $\kappa$ is a free parameter extracted from the data, one per particle. Nothing forces $\kappa_\mu = \kappa_\tau = \kappa_b = \kappa_t = \kappa_W = \kappa_Z = 1$ ; six independent measurements could have come out anywhere. That they land on one line means one number reproduces six couplings across three orders of magnitude.

An engineer will recognise the move as the reason log–log plots are drawn at all: you rectify the expected relation so that departures from it are visible rather than buried in curvature, and then you look for wiggle. Fitting a straight line to two points proves nothing; six points over three decades with independent errors is a different kind of statement.

And the departures are exactly where you would want to look. The muon point is the newest and least significant; the charm coupling is not measured at all; and the first two generations remain almost untested — which is precisely what §9.20 lists as unfinished.

Where it breaks: a straight line through four points over three decades is a weaker claim than it looks, and the axis hides that. Logarithmic axes compress disagreement: a coupling wrong by 30 % is barely off the line by eye. The fitted points are also not independent of the hypothesis — they are extracted as modifiers κ relative to the Standard Model prediction, so the plot cannot distinguish “the couplings scale with mass” from “we assumed they do and measured the residual”. The genuine test is whether a point falls off the line, and the places where that could still happen — charm, and the first two generations — are precisely the places with no measurement at all.

🔑 If you remember only three things

Spin 1 was excluded before spin 0 was established. Ruling out is faster than ruling in, and the sequence matters because the second step assumes the first.
The technique is fifty years old. The analysis that fixed the pion’s parity in 1962 is the one that fixed the Higgs’s, on a particle nobody had seen when it was invented.
The slope is the mechanism. Couplings proportional to mass is a claim about a law, and the data run over two orders of magnitude in mass to test it.

Where this goes next

The particle is a scalar, its parity is even, and its couplings are proportional to mass over three decades with an overall signal strength of $\mu = 1.002 \pm 0.057$ . The BEH mechanism of §9.12 is confirmed in both of its parts — the gauge-boson masses from the kinetic term and the fermion masses from the Yukawa terms.

§9.20 closes the chapter with the global fit over 21 measurements, and with the list of what remains untouched: the first two generations’ Yukawa couplings, and the shape of the potential itself — the self-coupling $\lambda$ , which no measurement has yet reached.

✅ Check yourself — spin, parity and the couplings

0/6 answered · 0 correct

1.J = 1 was excluded by the γγ observation but NOT by the ZZ* one. Why the difference?
2.The scalar amplitude has two terms, a₁(e₁·e₂) and a₂(e₁·q)(e₂·q). Why does only the first matter here?
3.The book says the Higgs data are 'offset in phase by π/2' relative to Fig. 3.1's π⁰ data. What exactly is offset?
4.The φ modulation is ±15 % here against ±47 % for the π⁰. What follows?
5.Table 9.2's eight branching ratios sum to 100.23 %. Why is that impossible, and what is the likely culprit?
6.Fig. 9.64 plots reduced coupling modifiers against mass and the Standard Model is a straight line. Is that circular?