---
title: Background-Free Rapidity Asymmetry
url: https://www.emergentmind.com/topics/background-free-rapidity-asymmetry
type: topic
---

# Background-Free Rapidity Asymmetry

Background-free rapidity asymmetry denotes a family of observables that isolate an antisymmetric structure in rapidity while removing, canceling, or strongly suppressing contaminating contributions such as Poisson counting noise, short-range dynamical correlations, uncorrelated bulk background, or explicit signal-background admixture. In the literature, the concept appears in several technically distinct settings: event-by-event rapidity-density asymmetry extracted from long-range correlations in heavy-ion collisions, odd rapidity modes generated by saturation-scale fluctuations, rapidity-asymmetry signals of jet-induced diffusion wake, and forward-backward asymmetry measurements in which signal and background are separated statistically rather than geometrically [1107.1215] [1607.03219] [2501.03419] [2508.04194] [0811.1426].

## 1. Observable classes and common logic

The central idea is always to construct an odd-in-rapidity quantity whose symmetric backgrounds either vanish identically or are removed by the measurement procedure. In symmetric collisions, this often means projecting onto the first odd rapidity mode. In jet-hadron observables, it means comparing samples that differ only by the rapidity of the associated jet while holding the trigger rapidity fixed or symmetric. In signal-extraction problems, it means weighting or subtracting background contributions so that the remaining asymmetry is attributable to the signal component.

| Context | Observable | Background treatment |
|---|---|---|
| Long-range rapidity correlations | $D_-^2=\tfrac12[F_{20}-F_{11}]$ | factorial moments remove Poisson noise; forward-backward gap suppresses short-range correlations |
| Saturation-scale fluctuations | $c_n\{n\}$ from $a_1$ | in symmetric collisions, background-symmetric pieces drop out |
| Jet-hadron correlations | $A^{near,away}(\Delta\eta;Y)$ or $A_y(y)$ | subtraction between rapidity classes cancels uncorrelated background |
| Longitudinal asymmetry | ratio of $dN_+/dy$ and $dN_-/dy$ | ratio of two classes cancels many detector inefficiencies and most backgrounds |

A recurring misconception is that “background-free” has a single operational meaning. The cited work uses the phrase in several non-equivalent senses: exact cancellation of an identical background, suppression of short-range sources by a rapidity gap, removal of Poisson noise through factorial moments, or simultaneous determination of signal and background asymmetries through optimal weighting. This suggests that the phrase is methodological rather than universal.

## 2. Long-range rapidity correlations and hidden asymmetry

A particularly direct construction starts from the single-particle rapidity density $\rho(y)$, defined so that in a small rapidity interval $\Delta y$ around $y$ the average number of particles is $\langle n\rangle_{\Delta y}\simeq \rho(y)\Delta y$. The full two-particle density is written as $\rho^{(2)}(y_1,y_2)=\langle \rho(y_1)\rho(y_2)\rangle$, and the connected correlation is
$$
C(y_1,y_2)\equiv \rho^{(2)}(y_1,y_2)-\langle \rho(y_1)\rangle\langle \rho(y_2)\rangle.
$$
After dividing the rapidity axis into bins $\Delta_i$, one assumes that for a fixed underlying density profile $\rho(y;Q)$ the occupancies $n_i$ fluctuate independently and Poisson-distributed about their means. Under this “Poissonian” assumption, the factorial moments
$$
F_{i_1,\ldots,i_B}\equiv \left\langle \frac{n_1!}{(n_1-i_1)!}\cdots \frac{n_B!}{(n_B-i_B)!}\right\rangle
$$
directly equal moments of the dynamical densities in each bin, $\langle n_1^{\,i_1}\cdots n_B^{\,i_B}\rangle_W$ [1107.1215].

For a symmetric collision at $y=0$, the event-by-event density can be expanded on $[-Y,Y]$ in Legendre polynomials:
$$
\rho(y;Q)=\rho_0(y)+\sum_{\ell=1}^\infty a_\ell(Q)\,P_\ell(y/Y),
$$
with $\rho_0(y)=\langle \rho(y)\rangle$. Even $\ell$ correspond to symmetric fluctuations, whereas odd $\ell$ correspond to antisymmetric fluctuations and therefore measure violation of boost invariance event by event. The two-particle cumulant decomposes as
$$
C(y_1,y_2)=\sum_{\ell,m}\langle a_\ell a_m\rangle_W P_\ell(y_1/Y)P_m(y_2/Y).
$$

The first odd mode is isolated by the dynamical variance of the forward-backward difference. Using two narrow bins centered at $\pm y_0$ and separated by a rapidity gap, with occupancies $n_f$ and $n_b$, one defines
$$
D_-^2\equiv \frac14\langle (n_f-n_b)^2\rangle_W=\frac12\,[F_{20}-F_{11}],
$$
where $F_{20}=\langle n_f(n_f-1)\rangle$ and $F_{11}=\langle n_f n_b\rangle$. Because the bins are separated by $|2y_0|\gg \Delta y$, short-range correlations such as resonance decays or jet fragmentation are strongly suppressed, while factorial moments remove pure statistical fluctuations. In this sense, $D_-$ is a direct, background-free measurement of rapidity-density asymmetry [1107.1215].

Applied to STAR Au+Au data at $\sqrt{s}=200\,\mathrm{GeV}$ with bins at $y=\pm 0.9$ and $\Delta y=0.2$, the reported values for top-central collisions are
$$
D_{ff}^2=350\pm 17,\qquad D_{fb}^2=202\pm 17,\qquad \langle n_f\rangle=96\pm 5,
$$
which give
$$
D_-=5.1\pm 1.2,\qquad D_+=15.1\pm 0.4.
$$
With $D_-/\sigma(D_-)\simeq 4.3$, the asymmetric component is clearly nonzero. The interpretation is that the rapidity profile in single events is not an even function of $y$, so boost invariance is violated on the event-by-event basis even at central rapidity. The paper further notes that this effect may seriously influence the hydrodynamic expansion of the system [1107.1215].

## 3. Odd rapidity modes from saturation-scale fluctuations

A closely related but more explicitly microscopic formalism writes the event-by-event single-particle distribution as
$$
\frac{dN}{dy}=\left\langle \frac{dN}{dy}\right\rangle [\,1+a_0+a_1 y+\cdots\,],
$$
with $\langle a_0\rangle=\langle a_1\rangle=\cdots=0$ by construction. Dropping $a_0$ and higher even terms isolates the antisymmetric fluctuation in the coefficient $a_1$. In a KLN-type model with proton saturation scales $Q_1^2$ and $Q_2^2$, one finds in the limit of small $\rho$-fluctuations
$$
a_1[\rho_1,\rho_2]=\frac{\lambda(\rho_1-\rho_2)}{2}+O((\rho_1-\rho_2)^3),
$$
where $\rho_i\equiv \ln(Q_{o,i}^2/\bar Q_0^2)$ [1607.03219].

The connected two-particle rapidity correlation satisfies
$$
\frac{C_2}{\langle dN/dy_1\rangle \langle dN/dy_2\rangle}
=
\langle a_1^2\rangle\, y_1 y_2+\cdots,
$$
and in symmetric collisions all background-symmetric pieces such as $\langle a_0^2\rangle$ and $\langle a_0 a_1\rangle$ drop out. More generally, the genuine $n$-particle cumulant obeys
$$
\frac{C_n}{\langle dN/dy_1\rangle\cdots \langle dN/dy_n\rangle}
=
\langle a_1^n\rangle_{[n]}\, y_1\cdots y_n+\cdots.
$$
The first cumulants are
$$
c_2\{2\}\equiv \langle a_1^2\rangle,\qquad
c_4\{4\}\equiv \langle a_1^4\rangle-3\langle a_1^2\rangle^2,
$$
with analogous expressions given for $c_6\{6\}$ and $c_8\{8\}$ [1607.03219].

The underlying fluctuation model takes
$$
\rho\equiv \ln(Q^2/\bar Q^2),\qquad
P(\rho)=\frac{1}{\sqrt{2\pi\sigma^2}}\exp[-\rho^2/(2\sigma^2)],
$$
so that $\mathrm{Var}(\rho)=\sigma^2$. For small $\sigma$,
$$
\langle a_1^2\rangle \simeq (\lambda^2/2)\sigma^2,\qquad
\langle a_1^4\rangle \simeq (3/4)\lambda^4\sigma^4.
$$
In the color-domain picture, the variance of the mean $\ln Q_s^2$ decreases with the number of domains, giving
$$
\sigma^2(N)\simeq \sigma_{mb}^2\,(N_{mb}/N)=\sigma_{mb}^2\,(N_{ch}^{mb}/N_{ch}),
$$
and therefore
$$
\sqrt{\langle a_1^2\rangle}\simeq \lambda \sigma(N)\propto 1/\sqrt{N_{ch}}.
$$
This was reported to be in excellent accord with the preliminary ATLAS result $\sqrt{\langle a_1^2\rangle}\sim N_{ch}^{-1/2}$ [1607.03219].

For higher-order cumulants, the small-$\sigma$ behavior is
$$
c_4\{4\}\simeq -\,2\lambda^4\sigma^5/\sqrt{\pi},\qquad
c_6\{6\}\simeq +\,3\lambda^6\sigma^7/(2\sqrt{\pi}),\qquad
c_8\{8\}\simeq -\,3\lambda^8\sigma^9/\sqrt{\pi},
$$
up to overall factors of order one. Since $\sigma(N)$ decreases with multiplicity, the paper predicts that $c_4\{4\}$ and $c_8\{8\}$ change sign at an intermediate multiplicity where $\sigma(N)\simeq 0.5$. This sign change is presented as a sharp, background-free signature of genuine saturation-scale-fluctuation-induced rapidity asymmetry [1607.03219].

## 4. Jet-hadron rapidity asymmetry and diffusion wake

In jet-medium studies, the relevant object is the per-trigger jet-hadron correlation in pseudorapidity and azimuth, defined relative to the leading jet by
$$
C_X(\Delta\eta,\Delta\phi)\equiv \frac{1}{N_{jet}}\frac{d^2N_{jet-h}^X}{d\Delta\eta\, d\Delta\phi},
$$
with $X=\mathrm{AA}$ or $\mathrm{pp}$, $\Delta\eta\equiv \eta_h-\eta_{jet_1}$, and $\Delta\phi\equiv \phi_h-\phi_{jet_1}$. Projection onto near and away regions yields $C_X^{near}(\Delta\eta)$ and $C_X^{away}(\Delta\eta)$. The medium-induced modification is
$$
\Delta C^{near,away}(\Delta\eta;|\Delta\eta_{jj}|)=
C_{AA}^{near,away}(\Delta\eta;|\Delta\eta_{jj}|)-C_{pp}^{near,away}(\Delta\eta),
$$
and the rapidity asymmetry for a large gap $Y$ relative to a reference small gap $Y_0$ is
$$
A^{near,away}(\Delta\eta;Y)\equiv
\Delta C^{near,away}(\Delta\eta;|\Delta\eta_{jj}|>Y)-
\Delta C^{near,away}(\Delta\eta;|\Delta\eta_{jj}|<Y_0).
$$
This observable was proposed as a robust signal of the diffusion wake induced by di-jets in Pb+Pb collisions [2501.03419].

Two background-removal schemes were presented. In the theoretical subtraction, CoLBT-hydro is run twice event by event, with and without the embedded di-jet, and the hydro-only charged-hadron spectrum is subtracted bin-by-bin from the jet+hydro spectrum. In the mixed-event subtraction, one forms the same-event pair distribution $S(\Delta\eta,\Delta\phi)$ and a mixed-event reference $M(\Delta\eta,\Delta\phi)$, with the trigger jet paired with charged hadrons from a different event of the same centrality class, and defines $C=S-M$. After either subtraction, the projected asymmetries are very similar in the simulation [2501.03419].

For centrality 0–10% Pb+Pb at $\sqrt{s_{NN}}=5.02\,\mathrm{TeV}$, with leading jet $p_T^{jet_1}>120\,\mathrm{GeV}/c$, sub-leading jet $p_T^{jet_2}>90\,\mathrm{GeV}/c$, $|\eta_{jet_{1,2}}|<1.6$, and soft hadrons $1<p_T^h<2\,\mathrm{GeV}/c$, the predicted near-side asymmetry for $|\Delta\eta_{jj}|>1.0$ shows a negative dip of about $-0.3$ per trigger at $\Delta\eta\approx -1.5$ and a positive excess of $+0.15$ near $\Delta\eta\approx 0.5$. For $0.5<|\Delta\eta_{jj}|<1.0$, the dip moves to $\Delta\eta\approx -1.0$ with amplitude about $-0.2$. The away-side asymmetry has a similar shape but a larger peak-trough difference of about $0.6$ per trigger. The physical interpretation is that a finite rapidity separation shifts the depletion region created by the diffusion wake of one jet relative to the enhancement associated with the other jet [2501.03419].

A later formulation generalizes the construction to dijets and $\gamma/Z^0$-jets and removes the background by design rather than by subtraction. With trigger rapidity fixed or restricted to a symmetric interval around $y=0$, one defines the opposite-side hadron rapidity distribution
$$
D(y;\eta_{\rm assoc})\equiv \frac{1}{N_{\rm trig}}\frac{\partial N_{\rm opp}(y;\eta_{\rm assoc})}{\partial y},
$$
and the rapidity asymmetry
$$
A_y(y)\equiv D(y;\eta_{\rm assoc}^{(1)})-D(y;\eta_{\rm assoc}^{(2)}).
$$
If
$$
D(y;\eta_{\rm assoc})=S(y;\eta_{\rm assoc})+B(y),
$$
with the background $B(y)$ independent of $\eta_{\rm assoc}$, then
$$
A_y(y)=S(y;\eta_{\rm assoc}^{(1)})-S(y;\eta_{\rm assoc}^{(2)}),
$$
so the background cancels exactly and no mixed-event subtraction or event-plane correction is required [2508.04194].

In CoLBT-hydro simulations of Pb+Pb at $\sqrt{s_{NN}}=5.02\,\mathrm{TeV}$, the $\gamma$-jet asymmetry for $1<p_T^h<2\,\mathrm{GeV}/c$ exhibits a valley at $y\approx 1.0$ for forward jets and a peak at $y\approx 0.0$ for central jets, while for $2<p_T^h<4\,\mathrm{GeV}/c$ the asymmetry is negligible. In dijets, selecting soft hadrons gives a negative dip at $y\approx -1.0$ and a positive peak at $y\approx +1.0$; again, the asymmetry tends to zero for $2$–$4\,\mathrm{GeV}/c$. Subtracting a pp baseline has a negligible effect on $A_y(y)$ [2508.04194].

## 5. Longitudinal asymmetry and participant-zone rapidity shift

In collisions of identical nuclei, the numbers of participants from the two nuclei can differ event by event. Writing these as $A$ and $B$, the participant asymmetry is
$$
\alpha\equiv \frac{A-B}{A+B},
$$
and the participant-zone rapidity shift is
$$
\Delta y\equiv y_0=\frac12\ln\frac{1+\alpha}{1-\alpha}=\mathrm{atanh}\,\alpha.
$$
For small $|\alpha|$,
$$
y_0\approx \alpha+\tfrac13\alpha^3+\mathcal O(\alpha^5).
$$
This longitudinal asymmetry causes a shift in the center-of-mass rapidity of the participant zone and therefore modifies the final-state rapidity distribution [1608.01428].

The analysis strategy is to bin events by $\alpha$, or by the experimentally accessible spectator asymmetry, compute the average $\langle y_0\rangle$ in each bin, and fit
$$
\langle y_0\rangle=c_1\alpha+c_2\alpha^2+c_3\alpha^3.
$$
Analytically one expects $c_1=1$, $c_2=0$, and $c_3=1/3$, while finite binning, fluctuations in $A+B$, and details of the rapidity profile generate effective deviations. Toy simulations with shifted Gaussian rapidity distributions show that the ratio of two distributions shifted by $\pm y_0$ expands into a polynomial whose linear coefficient directly measures $y_0$. Full event generators—HIJING, AMPT default, AMPT String-Melting, and DPMJET—were then used to fit the ratio of positive- and negative-asymmetry samples with a third-order polynomial in $y$, with the linear coefficient $c_1$ remaining dominant [1608.01428].

Experimentally, one cannot count $A$ and $B$ directly, but one can measure spectator neutrons in the two Zero-Degree Calorimeters and define
$$
\alpha_{\rm spec}=
\frac{E_{\rm left}-E_{\rm right}}{E_{\rm left}+E_{\rm right}}
\approx
\frac{(N-A)-(N-B)}{(N-A)+(N-B)}.
$$
A Monte Carlo Glauber response matrix is then used to unfold $\alpha_{\rm spec}\to \langle \alpha\rangle$ or directly $\langle y_0\rangle$. The final-state rapidity or pseudorapidity distributions for $\alpha_{\rm spec}>0$ and $\alpha_{\rm spec}<0$ are ratioed and fit with a third-order polynomial. Because the measurement is a ratio of two event classes, many detector inefficiencies, acceptance effects, and most backgrounds cancel, yielding an effectively background-free measure of the participant-zone rapidity shift [1608.01428].

The construction is not entirely model-independent. The paper states that the mapping $\alpha_{\rm spec}\to \langle y_0\rangle$ relies on Glauber-MC tuning, that ZDC resolution is poor in very central collisions, and that the small nonlinear coefficients are sensitive to the shape of $dN/dy$. This suggests that the observable is robust at the level of cancellation but still calibration-dependent.

## 6. Statistical extraction, collider applications, and interpretive limits

A general statistical framework for background-free rapidity asymmetry treats signal and background asymmetries simultaneously. Let $x$ be a discriminating variable and $y$ the rapidity. If $S(x)$ and $B(x)$ are the signal and background shapes, and $I_i=+1$ for $y_i>0$ and $I_i=-1$ for $y_i<0$, one defines per-event weights
$$
S_i=\frac{S(x_i)}{S(x_i)+B(x_i)},\qquad
B_i=\frac{B(x_i)}{S(x_i)+B(x_i)}.
$$
In the small-asymmetry limit, the weighted moments satisfy a $2\times 2$ linear system for the signal and background asymmetries. When one assumes $A_b=0$, the optimal estimator reduces to
$$
\hat A_s=\frac{\sum_i S_i I_i}{\sum_i S_i^2},\qquad
\mathrm{Var}(\hat A_s)=\frac{1}{\sum_i S_i^2}.
$$
For vanishing asymmetries, this estimator reaches the minimal variance bound given by the Cramér-Rao inequality [0811.1426].

A complementary treatment includes explicit background subtraction, acceptance-luminosity asymmetry, and unfolding. In a binned forward-backward analysis with counts $N_i^\pm$, background estimates $B_i^\pm$, relative acceptance$\times$luminosity factor $r$, and normalization $P$, the asymmetry estimator is
$$
A_i=\frac{1}{P}\,
\frac{(N_i^+-rN_i^-)-(B_i^+-rB_i^-)}
{(N_i^++rN_i^-)-(B_i^+ + rB_i^-)}.
$$
An unbinned extended-likelihood form replaces discrete bin counts by a signal PDF $S(y;A)$ and background PDF $B(y)$, while detector smearing is handled through a response matrix $R$ or a folded reconstruction-level PDF. This formulation makes the separation between background subtraction and unfolding explicit [2602.02325].

A collider application with a different meaning of “background-free” is the Tevatron top-antitop forward-backward asymmetry. There, the rapidity difference is
$$
\Delta y\equiv y_t-y_{\bar t},
$$
or in reconstructed lepton+jets events,
$$
\Delta y=q\,(y_{t,\mathrm{lep}}-y_{t,\mathrm{had}}),
$$
and
$$
A_{FB}=\frac{N(\Delta y>0)-N(\Delta y<0)}{N(\Delta y>0)+N(\Delta y<0)}.
$$
Matrix-element-level studies showed that the $W+4$ jets background has a positive slope in $\Delta y$, with $A_{FB}^{bkg}\approx +6\%$ for $M_{t\bar t}<450\,\mathrm{GeV}$ and $A_{FB}^{bkg}\approx +11\%$ for $M_{t\bar t}>450\,\mathrm{GeV}$, while the leading-order $t\bar t$ signal sample has $A_{FB}^{sig}=0$ by construction. The proposed remedies were tighter kinematic cuts, template fits in $\Delta y$, and bin-by-bin subtraction using the predicted background fractions and asymmetries [1205.5173].

A plausible implication is that “background-free rapidity asymmetry” spans a spectrum of constructions. At one end are observables such as $A_y(y)$ in which the background cancels exactly by symmetry and sample selection. At the other are statistically optimal extractions in which a background model remains necessary but the asymmetry estimator is unbiased, variance-efficient, or explicitly corrected for background and detector response. Across these uses, the common target is the same: an odd rapidity structure that can be interpreted dynamically rather than as an artifact of counting noise, combinatorial contamination, or acceptance imbalance.

Source: https://www.emergentmind.com/topics/background-free-rapidity-asymmetry