Electroweak Precision Observables: the Oblique Parameters $S$, $T$, $U$

Introduction

The standing of this article. The material below is standard theory, recorded in the corpus's language and notation. It is standard input, not a framework result: no number in it is derived from the biquaternion algebra, and the section What the Framework Supplies says exactly which parts are linear algebra the algebra can host and which parts are imported with the particle content. The two citations are Peskin and Takeuchi (1990), where the three parameters are defined, and the electroweak review of the Particle Data Group, where they are fitted.

The claim. Three numbers summarise what any heavy new sector does to the electroweak gauge bosons, and they exist because the vacuum polarisation of those bosons has more freedom than the three inputs of the theory can fix. The freedom is three-dimensional, and that dimension is a count, not a convention:

$$ 8-2-3=3 . $$

The count in words. Four self-energy functions, each with a value and a slope at zero momentum, give eight numbers. Electromagnetic gauge invariance sets two of the values to zero. The three measured inputs $\alpha$, $G_F$ and $M_Z$ absorb three more. Three combinations survive, and they are called $S$, $T$ and $U$.

The definitions, stated once.

$$ S=\frac{4s_W^2c_W^2}{\alpha}\left[\Pi'_{ZZ}(0)-\frac{c_W^2-s_W^2}{s_Wc_W}\Pi'_{Z\gamma}(0)-\Pi'_{\gamma\gamma}(0)\right] , $$

$$ T=\frac{1}{\alpha}\left[\frac{\Pi_{WW}(0)}{M_W^2}-\frac{\Pi_{ZZ}(0)}{M_Z^2}\right] , \qquad U=\frac{4s_W^2}{\alpha}\left[\Pi'_{WW}(0)-c_W^2\Pi'_{ZZ}(0)-2s_Wc_W\Pi'_{Z\gamma}(0)-s_W^2\Pi'_{\gamma\gamma}(0)\right] , $$

with $\Pi_{VV'}(q^2)$ the transverse part of the self-energy and the prime the derivative in $q^2$. All three are shifts: they are defined with the Standard Model of some reference top mass and reference Higgs mass subtracted, and that subtraction is the subject of §The Scheme, Stated Once.

What each measures, in one line each. $S$ measures the slope of the part of the self-energy that breaks weak isospin without breaking the custodial symmetry. $T$ measures the value of the part that breaks custodial symmetry, and it is the $\rho$ parameter written as a deviation. $U$ measures the slope of the part that breaks custodial symmetry, and it is the one the data constrain least, which is why the literature writes "$S$, $T$" more often than "$S$, $T$, $U$".

The measured values, and what they say.

Parameter Fitted value Meaning of a non-zero value
$S$ $0.00\pm0.07$ new states with weak isospin that couple to the transverse modes
$T$ $0.05\pm0.06$ new states with a mass splitting inside an isospin multiplet
$U$ $0.02\pm0.06$ new states whose custodial violation is momentum-dependent

The three are consistent with zero, and the statement that they are is the sharpest single statement in the sector: no new sector that couples obliquely to the gauge bosons sits at the electroweak scale. That sentence is weaker than it sounds, and §What the Parameters Cannot See says why.

Why the corpus needs the article. Because $\rho$ has already appeared twice — as the tree-level number $1$ in article 4 and as the custodial invariant of the companion article on custodial symmetry — and both times the statement "$\rho-1\approx\alpha T$" was used without the definition of $T$. This article supplies the definition, the count that produces it, and the numbers, so that the two earlier uses have something to point at.

The Self-Energy Functions

The object. The one-particle-irreducible two-point function of two gauge fields, in momentum space, carries two Lorentz indices and is written

$$ i\Pi_{VV',\mu\nu}(q)=i\left(\delta_{\mu\nu}-\frac{q_\mu q_\nu}{q^2}\right)\Pi_{VV'}(q^2)+\text{(terms proportional to }q_\mu\text{ or }q_\nu\text{)} . $$

The second piece never contributes to a physical amplitude, for the reason article 7 established: a conserved external current annihilates it. What matters is the transverse function $\Pi_{VV'}(q^2)$.

Why the transverse structure is forced for the photon. Electromagnetic gauge invariance requires $\sum_\mu q_\mu\Pi_{\gamma\gamma,\mu\nu}=0$. With the projectors of article 7,

$$ P_{\mu\nu}=\delta_{\mu\nu}+\frac{q_\mu q_\nu}{\Pi} , \qquad L_{\mu\nu}=-\frac{q_\mu q_\nu}{\Pi} , \qquad \Pi\equiv-\sum_\mu q_\mu q_\mu , $$

the only symmetric tensor built from $\delta$ and $q$ that is annihilated by $q_\mu$ is $P_{\mu\nu}$, since $\sum_\mu q_\mu P_{\mu\nu}=0$ and $\sum_\mu q_\mu L_{\mu\nu}=-q_\nu$. So

$$ \Pi_{\gamma\gamma,\mu\nu}(q)=P_{\mu\nu}\,\Pi_{\gamma\gamma}(q^2) , $$

and the longitudinal projector is absent identically. The statement "the photon self-energy is transverse" and the statement "the longitudinal projector is invisible to a conserved current" are the same statement, and both are the Ward identity.

The four functions. There are four: the three neutral pairs among $\gamma$ and $Z$, and the charged pair $WW$, which mixes with neither:

Function External legs Vanishes at $q^2=0$?
$\Pi_{\gamma\gamma}$ $\gamma\gamma$ yes, and its value is zero
$\Pi_{Z\gamma}$ $Z\gamma$ yes, and its value is zero
$\Pi_{ZZ}$ $ZZ$ no
$\Pi_{WW}$ $WW$ no

Two values are zero and two are not. The two that vanish do so by electromagnetic gauge invariance: a massless photon cannot mix with anything at zero momentum, because a mixing would give it a mass. The two that do not vanish are the masses themselves: $\Pi_{ZZ}(0)=M_Z^2$ and $\Pi_{WW}(0)=M_W^2$ in the reference theory, so the two non-zero values are not free but are fixed by the two measured masses.

The Taylor expansion. Each function is expanded about $q^2=0$,

$$ \Pi_{VV'}(q^2)=\Pi_{VV'}(0)+q^2\,\Pi'_{VV'}(0)+\tfrac12 q^4\,\Pi''_{VV'}(0)+\cdots , $$

and the three parameters retain the first two terms only. That truncation is the definition of "oblique": the new physics is assumed heavy compared with $M_Z$, so that its contribution is a polynomial in $q^2/M^2$ of low order, and the terms beyond the first derivative are dropped.

The four functions give eight numbers. Value and slope, for four functions. That is the starting point of the count.

Why Three and Not More

The arithmetic. Eight numbers, less two, less three.

Step Removed Reason
four functions, value and slope $8$ remain definition
$\Pi_{\gamma\gamma}(0)=0$ and $\Pi_{Z\gamma}(0)=0$ $6$ remain electromagnetic gauge invariance
$\alpha$, $G_F$, $M_Z$ as inputs $3$ remain they fix three normalisations
the survivors $S$, $T$, $U$ by definition

How the three are absorbed. The three inputs are three measurements of three combinations of the same eight numbers: $\alpha$ is the residue of the photon propagator, $G_F$ is the charged-current normalisation at zero momentum transfer, and $M_Z$ is the position of the neutral pole. Which combination each one removes depends on the basis in which the eight are written, so the assignment of inputs to numbers is not one-to-one; the dimension of what is left is.

Why the third input leaves a difference rather than a value. $\alpha$ and $G_F$ together predict $M_W$ in the reference theory, and $M_Z$ is measured. What the three inputs cannot fix is the relative normalisation of the charged and neutral sectors, which is the combination appearing in $T$. The three inputs are three measurements against four normalisations, and the mismatch is one number; the derivatives supply two more.

The count is not a convention. A different set of three inputs would give a different basis for the same three-dimensional space, in the way that a different choice of axes gives different coordinates for the same vector. The dimension is invariant, and the invariance is the content of the statement "three parameters". A reader who doubts the count should try to construct a fourth: any addition would have to be a combination of the eight already counted, so it cannot be independent.

The fourth function is not free either. $\Pi_{WW}$ appears with both its value and its slope, and both are used: the value in $T$, the slope in $U$. The reason the $W$ enters two parameters and the $Z$ enters three is that the $Z$ mixes with the photon and the $W$ does not.

What Each of the Three Measures

The custodial relation, at the level of the self-energies. A sector that respects the custodial symmetry has the same two-point function in its charged direction and in the neutral direction of the same isospin multiplet,

$$ \Pi_{WW}(q^2)=\Pi_{W^3W^3}(q^2)\qquad\text{for all }q^2 , $$

each side being written in the mass eigenstates. At $q^2=0$ the photon pieces vanish and the relation reads $\Pi_{WW}(0)=c_W^2\Pi_{ZZ}(0)$, which is $M_W^2=M_Z^2c_W^2$. One relation, evaluated and differentiated, produces the zeros of $T$ and $U$.

The three parameters as three ways of breaking it.

Parameter Which part of the relation fails
$T$ the relation at $q^2=0$: the two masses are not in the custodial ratio
$U$ the relation in its $q^2$ derivative: the two slopes are not in the custodial ratio
$S$ neither: $S$ is the overall size of the isospin-breaking sector that the relation does not constrain

$T$ against $\rho$. With the definitions above, and at leading order in the deviations,

$$ \alpha T=\frac{M_W^2}{M_Z^2c_W^2}-1=\rho-1 . $$

The verification is direct: $\rho-1$ is the mismatch of the two normalisations, and $T$ is that mismatch divided by $\alpha$ so as to be dimensionless and of order one for a typical heavy sector. $T$ is the oblique parameter of $\rho$, and $\rho$ is the tree-level value of $T$'s numerator. For the measured $T=0.05$, $\rho-1=3.6\times10^{-4}$: the custodial relation holds to four parts in ten thousand.

$S$ is the mixing meter, and the name is half the story. Written in the neutral self-energies the combination is a difference of slopes, and the subtraction is what removes the piece a pure isospin sector would contribute; what is left is the $W^3$–$B$ mixing. But $S$ is not the $W^3$–$B$ function alone. In the definition it appears with the coefficient $-(c_W^2-s_W^2)/(s_Wc_W)$ against $\Pi'_{Z\gamma}$ and with $-1$ against $\Pi'_{\gamma\gamma}$, so a sector whose $\Pi_{3Y}$ vanishes still contributes through $\Pi'_{ZZ}-\Pi'_{\gamma\gamma}$, which is proportional to $\Pi'_{33}-\Pi'_{YY}$. A vector-like singlet or doublet has equal left and right couplings, so those weights cancel and $S=0$; a chiral doublet does not, and it keeps $S=N_c/(6\pi)$ even when its two members are exactly degenerate. That number, $0.053$ for a lepton doublet and $0.16$ for a quark one, is the whole reason a fourth chiral generation is excluded, and it is the value at degeneracy, not away from it. What degeneracy removes is $\Pi_{3Y}$ — the trace $\sum_iT^3_iY_i$ vanishes for any doublet with a common hypercharge — and with it the $Z\gamma$ mixing, not $S$.

$U$ and why it is small. $U$ is the same kind of difference, taken in the derivative. For a heavy sector it is suppressed relative to $S$ by $M_Z^2/M^2$, because the difference of the derivatives brings down a factor of the inverse mass squared. That is why the pairing $S$–$T$ carries nearly all the information and $U$ is quoted last, with the largest relative uncertainty.

The Custodial Limit as the Zero of Two Parameters

The statement. A custodially invariant heavy sector has $T=0$ and $U=0$ exactly, and $S$ unconstrained. The first follows from the value of the custodial relation at $q^2=0$, the second from its derivative.

Verified on the tree-level numbers of article 4. With $M_W=\tfrac12gv$ and $M_Z=\tfrac12\sqrt{g^2+g_{\mathrm Y}^2}\,v$,

$$ \frac{M_W^2}{M_Z^2c_W^2}=\frac{g^2v^2/4}{\frac{v^2}{4}\left(g^2+g_{\mathrm Y}^2\right)\cdot\frac{g^2}{g^2+g_{\mathrm Y}^2}}=1 , $$

identically, so $T=0$ at tree level in the reference theory. That is not a prediction of the framework; it is the custodial symmetry of the doublet, computed in article 4 and interpreted in the companion article on custodial symmetry.

What breaks it and by how much. Hypercharge breaks the custodial symmetry explicitly, because it gauges one direction of $SU(2)_R$ and not the other two. The breaking is a loop effect in the reference theory and a tree-level effect for any new sector with a mass splitting inside a multiplet.

Five ways a sector can sit in this classification, and only one silence in it comes from degeneracy.

Sector $S$ $T$ $U$
strictly degenerate chiral doublet, custodial $N_c/(6\pi)$ $0$ $0$
chiral doublet with a mass splitting $N_c/(6\pi)$ at degeneracy, smaller as the splitting grows $\delta\rho/\alpha$ small
vector-like doublet, degenerate (left and right couplings equal) $0$ $0$ $0$
vector-like singlet $0$ $0$ $0$
state with hypercharge and no isospin $0$ $0$ $0$

The rows are ordered by what survives, and the order is the point. A vector-like singlet has no isospin, so it cannot break a relation about isospin, and it is invisible to all three parameters at leading order: it only renormalises the three inputs, which is to say it is absorbed into $\alpha$, $G_F$ and $M_Z$ and leaves no trace. The last row says the same about hypercharge alone: with no isospin there is nothing for the hypercharge to mix with. A vector-like doublet is invisible for the other reason — its left and right couplings are equal, so the weights that make $S$ cancel — and that is the row the word "degenerate" used to be attached to. Degeneracy is not silence. An exactly degenerate chiral doublet has $T=U=0$ by custodial symmetry and $S=N_c/(6\pi)$ at its full value, which is why a degenerate fourth generation is excluded by $S$ alone. $S$, $T$ and $U$ measure the vector-like or chiral character of a multiplet and the breaking of its custodial relation, not merely the existence of new particles — and a sector can be heavy, chiral, present, and still consistent with $(0,0,0)$ only if it is vector-like, which is the sharpest thing that can be said about them.

What a Heavy Sector Would Shift

The formula. For a heavy fermion doublet whose members have masses $m_1$ and $m_2$, the shift of the $\rho$ parameter is

$$ \delta\rho=\frac{N_cG_F}{8\sqrt{2}\,\pi^2}\,f(m_1,m_2) , \qquad f(m_1,m_2)=m_1^2+m_2^2-\frac{2m_1^2m_2^2}{m_1^2-m_2^2}\ln\frac{m_1^2}{m_2^2} . $$

The bracket $f$ vanishes identically at $m_1=m_2$, which is the check that it must pass: a degenerate pair is custodially invariant and cannot shift $\rho$. Both the formula and its limits are standard one-loop results.

Two limits, and they are different in kind. The bracket has two regimes.

  • One mass much lighter. With $m_2\to0$ the bracket tends to $m_1^2$, so $\delta\rho=N_cG_Fm_1^2/(8\sqrt2\pi^2)$. The top–bottom pair is in this regime, and the number is the classical one: with $m_t=172.69$ GeV the bracket is $2.98\times10^{4}$ GeV$^2$ and $\delta\rho=9.3\times10^{-3}$, that is $T=1.27$.
  • Near-degenerate. With $m_2=m_1+\delta m$ the bracket tends to $(\delta m^2)^2/(3m^2)$ with $\delta m^2=m_2^2-m_1^2$, which is quadratic in $\delta m^2$ where the popular estimate is linear. A heavy pair is therefore far safer than the linear estimate suggests, and the popular formula $\delta\rho\simeq N_cG_F\delta m^2/(8\sqrt2\pi^2)$ is not valid here: it overestimates by more than two orders of magnitude in the near-degenerate regime.

The table, exact against the quadratic estimate.

$m_1$, $m_2$ (GeV) $N_c$ $\delta\rho$ exact $T$ exact $T$ from the linear $\delta m^2$ estimate
$300/301$ $3$ $4.2\times10^{-7}$ $0.00006$ $0.026$
$300/310$ $3$ $4.2\times10^{-5}$ $0.006$ $0.26$
$300/400$ $3$ $4.2\times10^{-3}$ $0.57$ $3.0$
$500/510$ $3$ $4.2\times10^{-5}$ $0.006$ $0.43$
$172.69/4.18$ $3$ $9.3\times10^{-3}$ $1.27$ $1.28$

The last row is the reference, and that is the point of the whole scheme. The top–bottom contribution is not new physics: it sits inside the Standard Model prediction, and the reference top mass is defined so that it does. A reader who takes the $1.27$ of the last row as a discrepancy to be explained has forgotten what the subtraction means.

The bound $T$ places on a splitting. With $\sigma_T=0.06$ and a colour-triplet doublet at $300$ GeV, the near-degenerate bracket gives $\delta m\lesssim30$ GeV, and the exact bracket agrees: $300/325$ gives $T=0.036$ and $300/330$ gives $T=0.052$. The bound is on the splitting relative to the mass, not on the mass, which is why a heavy degenerate multiplet survives.

$S$ and why it is the harder one to estimate. $S$ survives only through the mixing, so it is sensitive to the difference of the two loop functions rather than to their size, and the difference is what makes the estimates scheme- and splitting-dependent. The large-$N$ number $N_c/(6\pi)$ is $0.053$ per doublet, of the order of the experimental uncertainty, which is the quantitative reason this sector is called precision physics: a single chiral doublet at the electroweak scale would be visible, and none is. The exact one-loop value for a given mass splitting is a two-line computation from the definition, and the literature's tables differ in the third digit for that reason.

The Measured Values

The table. From the electroweak fit to the LEP and SLD data, $M_W$, $m_t$ and the $Z$-pole asymmetries, with the Higgs mass fixed at $125$ GeV:

Parameter Value Uncertainty
$S$ $0.00$ $\pm0.07$
$T$ $0.05$ $\pm0.06$
$U$ $0.02$ $\pm0.06$

The third digit is not meaningful, and the numbers move in the third digit between editions of the review and between fits with different reference masses. What is stable is the size: a few hundredths, which is the same size as the standard estimate $N_c/(6\pi)$ for one chiral doublet.

The correlation is large and it matters. $S$ and $T$ are quoted with a correlation close to $+0.9$, and the correlations involving $U$ are smaller and negative. Two consequences follow, and both are routinely ignored.

  • The errors are not independent. Adding $\pm0.07$ and $\pm0.06$ in quadrature is wrong. The ellipse is long and thin, and its axes are computable: with those two uncertainties and that correlation, the combination $S-1.2\,T$ is measured to $\pm0.020$ while the orthogonal one, $S+0.84\,T$, is measured only to $\pm0.090$ — a factor of four and a half.
  • Which axis a sector sits on decides its bound, and the two axes are not the ones the individual numbers suggest. A heavy scalar doublet raises $S$ and lowers $T$, so it moves along $S-1.2T$: the tight direction, at $\pm0.020$. Such doublets are therefore among the most strongly constrained sectors in the classification, which is why the mass limits on them are tight. The direction that hides is the other one, $S$ and $T$ rising together, which is where a sector with both mixing and custodial violation of the same sign sits.

The ellipse in one sentence. One direction in the $S$–$T$ plane is measured to $0.02$ and the other only to $0.09$, and the well-measured one is the combination the $Z$-pole couplings fix directly; the poorly measured one is where a sector can hide.

What the fit is over-determined by. Some twenty quantities — $M_Z$, $\Gamma_Z$, $\sigma_{\mathrm{had}}$, the three $R_\ell$, the three $A_{FB}^{0,\ell}$, the three $A_\ell$ from SLD, $A_b$, $A_{FB}^{0,b}$, $A_c$, $A_{FB}^{0,c}$, $M_W$, $\Gamma_W$, $m_t$ — against three parameters. The over-determination is the test: a fit with three parameters to twenty numbers is a fit that can fail, and the fact that it does not is the evidence that the oblique approximation captures the data at the current precision.

The strongest objection, and the answer. The objection is that twenty observables reduced to three numbers discards information. The answer is that the discarded directions are not discarded but tested: a sector that couples directly to fermions would show up as a residual in the fit, and the residuals are consistent with zero. The oblique reduction is a hypothesis, and the fit tests it at the same time as it tests the parameters.

The Higgs Mass and the Fit That Predicted It

How the reference theory moves with the Higgs mass. In the Standard Model the Higgs contributes to $S$ logarithmically,

$$ \frac{dS}{d\ln M_H}=\frac{1}{6\pi}=0.053 , $$

so a factor of two in $M_H$ moves $S$ by $0.037$, about half the experimental uncertainty, and a factor of eight moves it by $0.11$. The sign is positive: a heavier Higgs pushes $S$ up.

What that did before 2012. With $S$ measured to be consistent with zero, a heavy Higgs was disfavoured, and the fit with $M_H$ free gave a central value near $90$ GeV with a 95 per cent upper bound near $160$ GeV. That bound was the sector's most-quoted number for a decade.

What happened. The Higgs was found at $125$ GeV, inside the window. This is the one place in the sector where a precision fit made a successful prediction of a mass before the discovery, and it is worth stating what the prediction did and did not assume: it assumed the Standard Model particle content, it used the oblique truncation, and it used the measured top mass. Under those three assumptions the bound was a bound on one number.

Why the success does not generalise. The same fit cannot predict the mass of a new sector, because a new sector enters with three unknown parameters rather than one. What the fit supplies for new physics is an allowed region in the $S$–$T$ plane, not a number.

The number to remember. $1/(6\pi)=0.053$ per unit of $\ln M_H$. The same factor appears as the large-$N$ estimate $N_c/(6\pi)$ for a chiral doublet's contribution to $S$, and the two are easy to confuse: one is a slope in the Higgs mass, the other an estimate of a finite mixing from a fermion loop. They enter different fits and they are not interchangeable.

What the Parameters Cannot See

The definition of oblique. The three parameters describe new physics that couples to the Standard Model only through the gauge-boson self-energies. That is a strong assumption, and it fails in identifiable ways.

What is invisible or misdescribed Why What is needed instead
a $Z'$ with direct fermion couplings it produces four-fermion operators, not self-energies a full four-fermion fit
leptoquarks with flavour structure the couplings carry generation indices flavour-specific observables
new physics lighter than $M_Z$ the expansion in $q^2/M^2$ does not converge a full momentum-dependent analysis
a sector coupling only to the Higgs it shifts $M_H$ and the quartic, not the gauge functions the Higgs couplings
anything with a large width a broad resonance is not a heavy state a line-shape analysis

The one that matters most at the current precision. A $Z'$ is the standard example: it mixes with the $Z$, so it does shift $S$ and $T$, but it also couples to fermions directly, and the direct couplings dominate the $Z$-pole observables. A fit in $S$ and $T$ alone would attribute the shift to the wrong parameter and could miss the state entirely.

The second that matters. The oblique truncation drops terms of order $M_Z^2/M^2$ and higher. For a sector at $1$ TeV those terms are at the per-cent level, which is at the edge of the measurement; for a sector at $300$ GeV they are not small, and the three parameters are a poor summary.

A consistency check that the data pass. If the oblique description were wrong, the fit to twenty observables with three parameters would leave residuals much larger than the quoted uncertainties. It does not, so the description holds at the current precision, and the statement has a date on it: it will fail when the $Z$-pole uncertainties improve by an order of magnitude, or when a $Z'$ is found.

The Scheme, Stated Once

The three choices that change the numbers.

Choice Options Size of the effect
which $\alpha$ $\alpha(M_Z)=1/127.95$ or $\alpha(0)=1/137.04$ $7$ per cent on $S$
reference $m_t$ $172.69$ GeV, or the fitted value shifts $T$ by the top contribution
reference $M_H$ $125$ GeV, or free shifts $S$ by $1/(6\pi)$ per unit $\ln$

The $\alpha$ choice, quantified. The prefactor of $S$ is $4s_W^2c_W^2/\alpha$, which is $91.0$ with $\alpha(M_Z)$ and $97.4$ with $\alpha(0)$. The same physical self-energy therefore gives two different values of $S$, differing by $7$ per cent — about one standard deviation. $S$ without a statement of $\alpha$ is not a number, and this is the same lesson article 5 states about $\sin^2\theta_W$.

The reference choice, quantified. Moving the reference top mass by $1$ GeV moves $T$ by $0.0148$, since $\partial T/\partial m_t$ is the top's contribution to $\delta\rho$ differentiated. That is a quarter of the uncertainty, so the reference mass must be quoted to the GeV.

What does not depend on the scheme. The physical predictions: $M_W$, the $Z$-pole asymmetries and the cross sections. The parameters are coordinates on the space of heavy sectors, and coordinates move when the axes do, which is why the scheme column exists in the table of article 1 and why this article repeats the discipline.

Which $\alpha$ this article's numbers actually use, and why the answer is $\alpha(0)$. Every number here is normalised with $\alpha(0)=1/137.036$, not $\alpha(M_Z)$. That is not a slip of the same kind the table above warns about; it is the convention the definitions were written in, and it is visible in two of the article's own numbers. $T=\delta\rho/\alpha$ at the top row gives $1.27$ with $\alpha(0)$ and $1.19$ with $\alpha(M_Z)$. And $\alpha T=\rho-1$ at $T=0.05$ gives $3.6\times10^{-4}$ with $\alpha(0)$ and $3.9\times10^{-4}$ with $\alpha(M_Z)$. Using $\alpha(M_Z)$ throughout would rescale $S$ and $T$ by $0.934$ — the $7$ per cent of the table, and nothing else changes.

The corpus's own convention is $\alpha(M_Z)$, and the two must be kept apart. Article 5 states its angle at $M_Z$ in the $\overline{\mathrm{MS}}$ scheme, articles 4 and 5 use the couplings at $M_Z$, and the corpus is consistent in doing so. The oblique parameters are the exception, because their normalisation is inherited from Peskin and Takeuchi and is a statement about the reference theory rather than about the $Z$ pole. So this article uses $\alpha(0)$ inside $S$, $T$ and $U$ and says so once here, and a reader comparing with a table that used $\alpha(M_Z)$ must accept the $7$ per cent before comparing digits. A statement of the form "$S=0.00$" without an $\alpha$ is not a number, and neither is "$T=1.27$".

What the Framework Supplies

What is derived and what is imported.

Item Status
Transversality of the self-energy from the Ward identity the algebra's, via article 7's projector
The projector algebra $P^2=P$, $L^2=L$, $P+L=\delta$ article 7, recomputed there
The count $8-2-3=3$ linear algebra; available to any framework
The custodial relation $\Pi_{WW}=\Pi_{W^3W^3}$ and its $q^2=0$ case article 3's group, article 4's masses
The definitions of $S$, $T$, $U$ imported, Peskin–Takeuchi
The measured values imported, PDG
The one-loop Veltman bracket, and the estimate $N_c/(6\pi)$ for a degenerate chiral doublet imported
The particle content that fixes $N_c$ and the multiplicities imported; a no route yet row of that article

What the algebra contributes that is not a transcription. Two things, and they are the reason the article can be written in this language at all.

  • Transversality is structural. The self-energy must lie in the projector $P$, and $P$ is the object article 7 already built to make the vector propagator invertible. So the same algebra that fixed the gauge produces the tensor structure of the vacuum polarisation, and no new machinery is needed.
  • The count is a dimension. The three parameters are the dimension of a quotient of the eight Taylor coefficients by five constraints, and a dimension is exactly what a linear-algebra framework can state without knowing the dynamics.

What the algebra does not contribute. The numerical values, and the existence of the three parameters as measurements. The framework can say why three combinations survive; it cannot say what they are worth, because their worth is set by the particle content, and the particle content is the item that article classifies as having no route yet.

The strongest objection, and the answer. The objection is that this article is standard theory with the author's notation on top, and the framework could be deleted without loss. The answer is that the deletion would lose the two structural statements above, and that those two are what make the article consistent with article 7 rather than merely adjacent to it: the same $P$ and $L$, the same Ward identity, the same $\Pi$. A transcription that did not use them would be a transcript; this one is a continuation.

One honest limitation. The oblique parameters are a summary of heavy physics, not a theory of it. Nothing in the framework or in the algebra produces the heavy sector to be summarised, so the article's subject is what the Standard Model's precision data say about the absence of such a sector, and it is silent on what such a sector would be.

Summary

Three numbers summarise heavy new physics in the gauge sector, and their number is forced: four self-energy functions with value and slope give eight numbers, two vanish by electromagnetic gauge invariance, three are absorbed by the measured inputs $\alpha$, $G_F$ and $M_Z$, and three survive. The count is $8-2-3=3$ and it is a dimension, not a convention.

$T$ is the $\rho$ parameter written as a deviation, $\alpha T=\rho-1$; with $T=0.05$ the custodial relation holds to $3.6\times10^{-4}$. $S$ is what survives the subtraction in the neutral sector — the $W^3$–$B$ mixing together with the difference $\Pi'_{ZZ}-\Pi'_{\gamma\gamma}$ — and it is largest for a degenerate chiral doublet, at $N_c/(6\pi)$; $U$ is the same kind of difference taken in the derivative, which is why $U$ is the smallest and the least measured. A custodially invariant sector has $T=U=0$ exactly, and the statement is the value and the derivative of one relation, $\Pi_{WW}(q^2)=\Pi_{W^3W^3}(q^2)$, whose $q^2=0$ case is $M_W=M_Zc_W$.

The measured values are $S=0.00\pm0.07$, $T=0.05\pm0.06$ and $U=0.02\pm0.06$, with an $S$–$T$ correlation near $+0.9$. The strong correlation is not a detail: it means the two uncertainties cannot be combined in quadrature and that a sector moving along the ellipse is less constrained than either bound suggests.

A mass splitting shifts $T$ through the Veltman bracket, which is quadratic in $\delta m^2$ for a near-degenerate pair and linear in the heavy mass for a hierarchical one; the popular linear estimate $\delta\rho\simeq N_cG_F\delta m^2/(8\sqrt2\pi^2)$ overestimates the near-degenerate case by more than two orders of magnitude, and the table above shows the exact values beside it. The top–bottom pair contributes $T=1.27$ and sits inside the reference, which is what the subtraction means. An exactly degenerate chiral pair has $T=U=0$ by custodial symmetry and $S=N_c/(6\pi)$ at full strength; it is the vector-like pair, with equal left and right couplings, that vanishes from all three. $S$ is therefore a meter of chiral versus vector-like structure, not of particle count.

The standard large-$N$ estimate for a chiral doublet's contribution to $S$, $N_c/(6\pi)=0.053$, is the size of the experimental uncertainty. That coincidence is what makes the sector a precision test: one chiral doublet at the electroweak scale would be visible, and none is. A vector-like singlet, and any state with no isospin, shifts nothing at leading order, so the parameters measure multiplet structure and its breaking rather than the presence of heavy states. The oblique description is itself a hypothesis, tested by the fit's residuals, with a date on it.

The framework supplies the transversality of the self-energy through article 7's projector and the dimension count as linear algebra. It does not supply the definitions, the numbers, or the particle content, and the particle content is the article's no route yet item.

Summary of Notation

Symbol Meaning
$\Pi_{VV'}(q^2)$ Transverse part of the $VV'$ self-energy
$\Pi'_{VV'}(0)$ Its derivative in $q^2$ at zero momentum
$P_{\mu\nu}$, $L_{\mu\nu}$ Article 7's transverse and longitudinal projectors
$\Pi$ $-\sum_\mu q_\mu q_\mu$, the positive momentum invariant
$s_W$, $c_W$ $\sin\theta_W$, $\cos\theta_W$, article 5
$\alpha$ $1/137.036$ inside $S$, $T$ and $U$ — the definitions' normalisation; $\alpha(M_Z)=1/127.95$ elsewhere in the corpus
$\rho$ $M_W^2/(M_Z^2c_W^2)$, equal to $1+\alpha T$ at leading order
$\Pi_{33}$, $\Pi_{3Y}$, $\Pi_{YY}$ The $W^3W^3$, $W^3B$ and $BB$ self-energies, in terms of which $\Pi_{ZZ}$, $\Pi_{Z\gamma}$ and $\Pi_{\gamma\gamma}$ are written
$q^2$ The momentum transfer squared of the self-energy
$M_W$, $M_Z$ $80.377$ and $91.1876$ GeV
$G_F$ $1.1664\times10^{-5}$ GeV$^{-2}$
$S$ The $W^3$–$B$ mixing parameter
$T$ The custodial-breaking value parameter, $\alpha T=\rho-1$
$U$ The custodial-breaking slope parameter
$\delta\rho$ The shift of $\rho$, equal to $\alpha T$ at leading order
$f(m_1,m_2)$ The Veltman bracket, vanishing at $m_1=m_2$
$N_c$ Colour multiplicity: $1$ for leptons, $3$ for quarks
$m_t$, $M_H$ Reference top and Higgs masses of the subtraction

Further Reading