---
title: Forward-Backward Asymmetry
url: https://www.emergentmind.com/topics/forward-backward-asymmetry
type: topic
---

# Forward-Backward Asymmetry

Forward-backward asymmetry is a parity-odd measure of directional imbalance that compares event yields or transition probabilities in a “forward” hemisphere with those in a “backward” hemisphere. In high-energy physics, the forward direction is defined relative to the beam, the incoming quark, or the parent hadron, depending on the process; in decay physics it is often defined in the rest frame of a dilepton or hadronic subsystem; and in stochastic dynamics it can refer to forward versus backward transition times. In all of these settings, the observable isolates the part of a distribution that is odd under a sign reversal of the relevant angle or direction, and it is therefore especially sensitive to vector–axial interference, absorptive phases, nontrivial spin structure, or nonequilibrium driving [1805.09239, 1206.4117, 1709.04894, 2205.04068, 2007.07335].

## 1. General definition and angular origin

The canonical definition is
\[
A_{FB}\equiv \frac{\sigma_F-\sigma_B}{\sigma_F+\sigma_B},
\]
where \(\sigma_F\) and \(\sigma_B\) denote forward and backward cross sections. In counting form this becomes \((N_F-N_B)/(N_F+N_B)\), while in differential form it is obtained by integrating an angular distribution over positive and negative values of a signed angle variable. The observable is therefore not tied to a single process, but to a common kinematic structure: an odd term in an angular or directional distribution [1805.09239, 1709.04894].

In neutral-current Drell–Yan production the odd term appears in the Collins–Soper angle \(\theta^*\),
\[
\frac{d\sigma}{d\cos\theta^*}\propto (1+\cos^2\theta^*)+A_4\cos\theta^*+\dots,\qquad
A_{FB}=\frac{3}{8}A_4,
\]
so \(A_{FB}\) directly isolates the coefficient of \(\cos\theta^*\) [1805.09239]. In rare \(B\to K_J\ell^+\ell^-\) decays the signed angle is \(\theta_\ell\), defined in the dilepton rest frame between the \(\ell^-\) momentum and the parent \(B\) direction, and
\[
A_{FB}(q^2)=
\frac{\int_0^1 d\cos\theta_\ell\, d^2\Gamma/(dq^2\,d\cos\theta_\ell)-\int_{-1}^0 d\cos\theta_\ell\, d^2\Gamma/(dq^2\,d\cos\theta_\ell)}
{d\Gamma/dq^2}
=\frac{3}{4}\frac{J_{6s}(q^2)}{d\Gamma/dq^2},
\]
with
\[
J_{6s}(q^2)=2\beta_\ell\,\mathrm{Re}\!\left[A_{\parallel}^L A_{\perp}^{L*}-A_{\parallel}^R A_{\perp}^{R*}\right].
\]
Here the asymmetry isolates the interference of transversity amplitudes [1206.4117].

In \(e^+e^-\) annihilation near the \(Z^0\) pole the same structure appears in
\[
\frac{d\sigma}{d\cos\theta}\propto (1+\cos^2\theta)+2\mathcal{A}\cos\theta,
\]
so the forward-backward asymmetry is again the normalized odd term. This suggests a unifying viewpoint: \(A_{FB}\) is best regarded as the projection of a distribution onto its first parity-odd angular harmonic, rather than as a process-specific counting ratio [2205.04068].

## 2. Electroweak neutral currents and hadronic final states

In neutral-current Drell–Yan production at the LHC, forward-backward asymmetry originates from the chiral structure of the \(Z\)-fermion couplings and the interference between photon and \(Z\) exchange. The relevant couplings are
\[
g_V^f=T_3^f-2Q_f\sin^2\theta_W,\qquad g_A^f=T_3^f,
\]
with
\[
A_f=\frac{2 g_V^f g_A^f}{(g_V^f)^2+(g_A^f)^2}.
\]
Near the \(Z\) pole, the parton-level asymmetry is approximately
\[
A_{FB}^{q\ell}\simeq \frac{3}{4}A_qA_\ell,
\]
while away from the pole \(\gamma\)–\(Z\) interference enhances sensitivity to electroweak couplings and electric charges [1805.09239].

Because the LHC is a \(pp\) collider, the incoming quark direction is not known event by event. Reconstructed observables therefore use kinematic proxies. One standard choice is the Collins–Soper angle
\[
\cos\theta^*=
\frac{2}{M_{\ell\ell}\sqrt{M_{\ell\ell}^2+p_{T,\ell\ell}^2}}
\left(p_1^+p_2^- - p_1^-p_2^+\right),
\]
with \(p_i^\pm=(E_i\pm p_{z,i})/\sqrt{2}\). The paper on Run-II Drell–Yan also adopts an experimentally convenient dilepton asymmetry,
\[
A_{\rm FB}^*=
\frac{\frac{d\sigma}{dM(\ell^+\ell^-)}[\eta(\ell^+)>0]-\frac{d\sigma}{dM(\ell^+\ell^-)}[\eta(\ell^+)<0]}
{\frac{d\sigma}{dM(\ell^+\ell^-)}[\eta(\ell^+)>0]+\frac{d\sigma}{dM(\ell^+\ell^-)}[\eta(\ell^+)<0]},
\]
and emphasizes that rapidity cuts make it sensitive not only to \(\theta_W\) but also to PDFs. At \(|y_{\ell\ell}|>4.5\) and near the \(Z\) peak, the selected region corresponds to \(x_1\approx 0.6\) and \(x_2\approx 7\times 10^{-4}\); in this regime the \(d\bar d\) fraction is strongly suppressed relative to \(u\bar u\), and statistical errors on \(A_{FB}^*\) at \(L=3\ \mathrm{ab}^{-1}\) become smaller than PDF errors up to \(M_{\ell\ell}\lesssim 400\ \mathrm{GeV}\) [1805.09239].

A common misconception is that forward-backward asymmetry ceases to be meaningful in a symmetric \(pp\) initial state. The Drell–Yan analysis shows instead that reconstructed asymmetries remain well defined, provided the quark direction is inferred statistically and the observable is interpreted as a flavour-weighted average over partonic channels [1805.09239].

The same electroweak logic extends to hadronic final states in \(e^+e^-\) annihilation. For inclusive vector-meson production, \(e^+e^-\to V+X\), the hadronic asymmetry is defined by integrating over \(y=(1+\cos\theta)/2\), and for the unpolarized channel one finds
\[
A_{\mathrm{FB},U}(z,Q^2)=
\frac{3\left(\chi\,c_3^e c_3^q+\chi_{\rm int}\,c_A^e c_A^q\right)\,\tilde D_1(z,Q^2)}
{4\left(e_q^2+\chi\,c_1^e c_1^q+\chi_{\rm int}\,c_V^e c_V^q\right)\,D_1(z,Q^2)},
\]
with
\[
\tilde D_1=D_1-\frac{\kappa_M^2}{z}\Big(4\,\mathrm{Re}\,D_{-3dd}+D_{4q}\Big),\qquad \kappa_M=\frac{M}{Q}.
\]
This formulation factorizes electroweak couplings from fragmentation dynamics and makes twist-4 power corrections explicit [2205.04068].

A related semi-inclusive extension for hadron pairs, \(e^+e^-\to h_1+h_2+\bar q+X\), expresses the asymmetry in terms of di-hadron fragmentation functions. The leading asymmetry is
\[
A_{FB}=
\frac{3\left(\chi c_3^e c_3^q+\chi_{int} c_A^e c_A^q\right)\tilde D_1}
{4\left(e_q^2+\chi c_1^e c_1^q+\chi_{int} c_V^e c_V^q\right)\tilde D_1},
\]
while twist-3 asymmetries probe \(G_1^\perp\), \(D^\perp\), \(D^\sphericalangle\), \(G^\perp\), and \(G^\sphericalangle\). This suggests that forward-backward asymmetry can be used not only for electroweak coupling extraction but also to constrain fragmentation-function parameterizations [2210.11956].

## 3. Rare and semileptonic decays

In flavour physics, forward-backward asymmetry is particularly important in semileptonic rare decays because it isolates interference between short-distance Wilson coefficients and hadronic form factors. For \(B\to K_J\ell^+\ell^-\), with \(K_J\) a kaonic resonance, the full angular distribution can be written in terms of transversity amplitudes \(A_{0,\parallel,\perp}^{L,R}\), and only the hadronic helicities \(\lambda=0,\pm1\) contribute for all \(J\ge 1\) because the weak current carries a single Lorentz index [1206.4117].

The amplitudes depend on \(C_7^{\mathrm{eff}}, C_9^{\mathrm{eff}}, C_{10}\) and effective form factors. In the large-recoil limit, seven independent \(B\to K_J^*\) form factors reduce to two universal functions, \(\xi_\perp^{K_J^*}\) and \(\xi_\parallel^{K_J^*}\), through LEET relations. The numerator of \(A_{FB}\) is then dominated by transverse interference and takes the schematic form
\[
A_{FB}(q^2)\propto
\mathrm{Re}\!\left\{\big[C_9^{\mathrm{eff}}(q^2)+\kappa(q^2)C_7^{\mathrm{eff}}\big]\,C_{10}^*\right\},
\]
with \(\kappa_\perp\simeq\kappa_\parallel\) in the large-recoil limit. This yields the well-known zero-crossing condition
\[
\mathrm{Re}\!\left[C_9^{\mathrm{eff}}(q_0^2)+\kappa(q_0^2)C_7^{\mathrm{eff}}\right]=0.
\]
Because the form-factor dependence largely cancels in \(\kappa(q^2)\), the zero of \(A_{FB}\) is theoretically clean and only mildly dependent on the resonance spin \(J\) through kinematics [1206.4117].

The analysis covers \(K^*(892)\), \(K_1(1270)\), \(K_1(1410)\), \(K^*(1410)\), \(K^*(1680)\), \(K_0^*(1430)\), \(K_2^*(1430)\), \(K_3^*(1780)\), and \(K_4^*(2045)\), with predicted branching ratios of order \(10^{-6}\) to \(10^{-7}\). The central phenomenological conclusion is that new physics in \(C_9\) and \(C_{10}\) can sizably deform the \(A_{FB}(q^2)\) shape and shift the zero-crossing point \(q_0^2\), while modifications in \(C_{10}\) alone rescale the asymmetry without changing \(q_0^2\) at leading order for real Wilson coefficients [1206.4117].

A closely related semileptonic observable appears in \(\tau^-\to K\pi\nu_\tau\), where
\[
\frac{d^2\Gamma}{ds\,d\cos\theta}=\mathcal N(s)\,[A(s)+B(s)\cos\theta+C(s)\cos^2\theta],
\qquad
A_{FB}(s)=\frac{B(s)}{2A(s)+\tfrac{2}{3}C(s)}.
\]
In the Left–Right Inverse Seesaw analysis, \(B(s)\) is dominated by vector–scalar interference,
\[
B(s)= -\,\frac{2m_\tau}{\sqrt{s}\,|\vec p_K|}\,
\mathrm{Re}\!\left[(1+g_V)g_S^*F_+(s)F_0^*(s)\right],
\]
so the asymmetry is controlled by the interplay of \(F_+(s)\), \(F_0(s)\), and the scalar operator \(g_S\). The paper finds that the integrated CP asymmetry remains tiny, whereas the differential forward-backward CP asymmetry becomes most sensitive near the \(K^*(892)\) and \(K_0^*(1430)\) resonances [2604.24924].

## 4. Top-antitop production at hadron colliders

In \(t\bar t\) production at the Tevatron, forward-backward asymmetry exploits the asymmetric \(p\bar p\) initial state. The standard production-level definition is
\[
A_{\mathrm{FB}}^{t\bar t}=
\frac{N(\Delta y>0)-N(\Delta y<0)}{N(\Delta y>0)+N(\Delta y<0)},
\qquad
\Delta y=y_t-y_{\bar t}.
\]
At leading order in QCD the asymmetry vanishes; at higher order it is generated primarily by interference between the tree-level \(q\bar q\to t\bar t\) amplitude and one-loop box diagrams, with additional contributions from initial-state and final-state radiation, and with non-negligible electroweak corrections [1709.04894, 1107.4995].

The final Tevatron combination by CDF and D0 gives
\[
A_{\mathrm{FB}}^{t\bar t}=0.128\pm0.025,
\]
with corresponding lepton-based asymmetries
\[
A_{\mathrm{FB}}^\ell=0.073\pm0.020,\qquad
A_{\mathrm{FB}}^{\Delta\eta}=0.108\pm0.046.
\]
The combined differential asymmetry rises with invariant mass and with \(|\Delta y|\); for example,
\[
A_{\mathrm{FB}}(350\text{--}450~\mathrm{GeV})=0.081\pm0.037,\quad
A_{\mathrm{FB}}(>650~\mathrm{GeV})=0.319\pm0.147,
\]
and the fitted slopes are
\[
\alpha_{m_{t\bar t}}=(9.71\pm3.28)\times10^{-4}\ \mathrm{GeV}^{-1}c^2,\qquad
\alpha_{\Delta y}=0.187\pm0.038.
\]
These combined results are consistent with NNLO QCD plus NLO EW predictions within about \(1.3\sigma\) to \(1.6\sigma\), which is an important correction to the earlier impression of a large, persistent anomaly [1709.04894].

Earlier Tevatron-era measurements and model-building papers nonetheless treated the asymmetry as a potential new-physics signal. Explicit proposals included flavour-changing \(Z'\) models, \(W'\) exchange, scalar-mediated \(t\)-channel exchange, singlet-extended MSSM scenarios, and Little Higgs \(Z_H\) exchange [1209.4354, 1205.3311, 1107.4350, 1111.4488, 1309.5419]. A more conservative effective-field-theory analysis emphasized that the odd term in the partonic angular distribution is proportional to
\[
\Delta C\equiv C_1-C_2,
\]
so tree-level new-physics generation of \(A_{FB}\) requires chiral couplings to both light quarks and top quarks [1301.1410].

At the LHC, a true inclusive forward-backward asymmetry vanishes because the initial state is symmetric. This does not imply that the underlying physics disappears; rather, it must be encoded in derived observables. One standard option is the charge asymmetry
\[
A_C=
\frac{N(|y_t|>|y_{\bar t}|)-N(|y_t|<|y_{\bar t}|)}
{N(|y_t|>|y_{\bar t}|)+N(|y_t|<|y_{\bar t}|)},
\]
while a more process-level decomposition introduces collider-independent partonic asymmetries
\[
A_q=
\frac{\sigma_q(\Delta y>0)-\sigma_q(\Delta y<0)}
{\sigma_q(\Delta y>0)+\sigma_q(\Delta y<0)},
\qquad q\in\{u,d\},
\]
with
\[
FB=A_uF_u+A_dF_d,\qquad
A_C=A_uF_uD_u+A_dF_dD_d.
\]
This framework shows explicitly how a positive Tevatron asymmetry can coexist with a small LHC charge asymmetry because the \(pp\) weights \(F_qD_q\) differ from the \(p\bar p\) weights \(F_q\) [1205.1898].

A different LHC adaptation is the one-side forward-backward asymmetry,
\[
A_{\mathrm{OFB}}=\frac{F_-+B_-}{F_++B_+},
\]
with \(F_\pm\) and \(B_\pm\) defined by cuts on the longitudinal boost \(P^z_{t\bar t}\) and, optionally, on \(M_{t\bar t}\). By selecting events with \(P^z_{t\bar t}\) above a threshold, one statistically tags the quark direction and suppresses symmetric \(gg\to t\bar t\) production. This was proposed both for tops and for other final states such as charged leptons and \(b\bar b\), precisely to recover forward-backward information in a \(pp\) environment [1008.2685, 1011.1428].

## 5. CP-odd and spin-sensitive extensions

Forward-backward asymmetry admits several refinements that retain its basic odd-under-reversal structure while probing additional dynamics. One important extension is the CP-odd forward-backward asymmetry in \(\tau^\mp\to K\pi\nu\),
\[
A_{\rm CP}^{\rm FB}(s)\equiv A_{FB}^{\tau^-}(s)-A_{FB}^{\tau^+}(s).
\]
In the Left–Right Inverse Seesaw model this quantity isolates the imaginary part of vector–scalar interference,
\[
A_{\rm CP}^{\rm FB}(s)\propto
\frac{\kappa_{VS}(s)}{2A(s)+\tfrac{2}{3}C(s)}
\,\mathrm{Im}\!\big[(1+g_V)g_S^*\big]\,
\mathrm{Im}\!\big[F_+(s)F_0^*(s)\big],
\]
and therefore requires both a weak phase and a strong-phase difference. The predicted signal peaks around \(\sqrt{s}\approx1.4\ \mathrm{GeV}\), whereas the integrated CP asymmetry remains far smaller because the odd-in-\(\cos\theta\) interference term vanishes upon full angular integration [2604.24924].

A different extension appears in top physics through transverse polarization. In broad axigluon scenarios, the same new-physics dynamics that generate \(A_{FB}\) also generate a transverse top polarization \(P_\perp\) normal to the production plane, because the complex propagator supplies the required absorptive phase. In the purely axial case discussed in the axigluon analysis,
\[
A_{FB}(t)=
\frac{\big(\tfrac{3}{2}\big)a_q a_t\,\beta\,(t^2-M_A^2)\,t^4}
{\big[(t^2-M_A^2)^2+\Gamma_A^2M_A^2\big](t^2+2m_t^2)+a_q^2a_t^2\beta^2t^6},
\]
while
\[
P_\perp(t)=
\frac{\big(-\tfrac{3\pi}{4}\big)a_q a_t\,\beta\,m_t\,M_A^2\,t^3}
{\big[(t^2-M_A^2)^2+\Gamma_A^2M_A^2\big](t^2+2m_t^2)+a_q^2a_t^2\beta^2t^6}.
\]
Their ratio is
\[
\frac{P_\perp(t)}{A_{FB}(t)}=
\frac{-\tfrac{\pi}{2}m_tM_A}{t(t^2-M_A^2)}.
\]
This relation makes clear that \(A_{FB}\) and transverse polarization are correlated but not redundant: the former measures the odd real part of the interference, whereas the latter measures the absorptive part [1303.1200].

Lepton-based top asymmetries provide a further spin-sensitive variant. The observable
\[
A_{\mathrm{FB}}^\ell=
\frac{N(q_\ell\eta_\ell>0)-N(q_\ell\eta_\ell<0)}
{N(q_\ell\eta_\ell>0)+N(q_\ell\eta_\ell<0)}
\]
is experimentally robust because it requires only acceptance corrections and is closely related to the mean shift of the signed pseudorapidity variable \(x=q_\ell\eta_\ell). The empirical form
\[
A(x)=a\,\tanh(x/2)
\]
was shown to reproduce the inclusive asymmetry to better than a percent in simulation, while a double-Gaussian model explains why the dominant contribution comes from \(|x|\approx1\) rather than from the far forward tail [1403.7565].

## 6. Nonequilibrium transition-time asymmetry

Outside collider physics, forward-backward asymmetry can be defined for transition times in stochastic processes. In the two-lane random-walk model studied in nonequilibrium statistical mechanics, a particle hops right and left on each lane with rates \(u,w\) and \(u',w'\), and switches lanes with rates \(a,b\). The master equations are
\[
\frac{d P^{(1)}_{n}}{dt}=uP^{(1)}_{n-1}+wP^{(1)}_{n+1}+aP^{(2)}_n-(u+w+b)P^{(1)}_n,
\]
\[
\frac{d P^{(2)}_{n}}{dt}=u'P^{(2)}_{n-1}+w'P^{(2)}_{n+1}+bP^{(1)}_n-(u'+w'+a)P^{(2)}_n.
\]
One then defines forward and backward conditional mean first-passage times \(T_+\) and \(T_-\), and a convenient normalized asymmetry
\[
A_{FB}=\frac{T_+-T_-}{T_++T_-}.
\]
Unlike the particle-physics observable, this quantity compares transition times rather than cross sections, but it plays the same structural role: it measures directional nonequivalence [2007.07335].

The crucial contrast is with the single-lane one-dimensional random walk, where microscopic reversibility implies
\[
T_+^{(1\mathrm{D})}=T_-^{(1\mathrm{D})}=\frac{1}{u+w}
\]
even in the presence of bias. In the two-lane model, multiple pathways with different local rates break this equality whenever the system is out of equilibrium. The natural thermodynamic control parameter is the cycle affinity
\[
A=k_{\mathrm B}T\ln\!\left(\frac{u\,w'}{u'\,w}\right),
\]
and the symmetry is restored only when
\[
u\,w'=u'\,w,
\]
that is, when detailed balance holds. The sign of the forward-backward transition-time asymmetry tracks the direction of the net loop current, so the asymmetry becomes a direct measure of deviation from equilibrium rather than of parity violation [2007.07335].

This broader usage clarifies the conceptual core of forward-backward asymmetry. Whether formulated as an angular coefficient in scattering, a zero-crossing diagnostic in flavour physics, a CP-odd interference term in semileptonic decay, a spin-sensitive observable in top production, or a transition-time imbalance in stochastic transport, the observable quantifies the failure of a process to be symmetric under reversal of a physically distinguished direction.

Source: https://www.emergentmind.com/topics/forward-backward-asymmetry