Model for the curvature response of the CDF II drift chamber

Abstract: The CDF II experiment at the Fermilab Tevatron used a drift chamber to measure the momenta of charged particles. We present a model for the response of the drift chamber to the curvature of a charged particle's trajectory. Constraints on the model parameters are obtained from cosmic-ray data and from information published by CDF in the context of the W boson mass measurement. Implications for the calibration of the drift chamber measurement of momentum are discussed. The robustness of the CDF calibration procedure is demonstrated. The model provides a framework for the analysis of precision magnetic trackers of high-momentum particles.

• The paper develops a parametric model explaining curvature response in the CDF II Drift Chamber and identifies physical mechanisms like sensor misalignment and ionization loss, enhancing measurement precision to 25 ppb and significantly impacting W-Z mass measurements.
• Key findings show that the cubic expansion of the response function suffices for precision curvature values and all coefficients—except sensor misalignment—are constrained within 30 MeV.
• Validations using cosmic rays highlight charge-antisymmetric parameters like Lorentz-angle cell tilt remain consistent with zero (geminal symmetry verified at sub-% uncertainties)

Overview and motivation

This paper by A. V. Kotwal develops a parametric model for the curvature response of the CDF II central outer tracker (COT), the open-cell drift chamber that provided the momentum measurement underpinning the CDF $W$ boson mass measurement, calibrated to 25 parts per million (ppm) (2604.00318). The central object is the response function relating measured curvature to true curvature, $c^{\rm measured} = c + \delta c$, where $c \equiv q/p_T$. The stated philosophy is explicitly reductionist: rather than high-dimensional fitting or machine learning, each coefficient of a general analytic ansatz is tied to a physical mechanism — sensor misalignment ($a_0$), momentum scale errors ($a_1$), ionization energy loss ($\epsilon$, entering as an effective $b_2$-like term), Lorentz-angle cell tilt asymmetries ($b_1$, $a_2$, $b_3$) — so that calibration can proceed largely from first principles.

The paper's principal claims are strong: (i) the analytic Maclaurin expansion truncated at cubic order is sufficient for all curvature values relevant to precision observables, with higher-order terms absorbed into propagated uncertainties; (ii) every analytic parameter is either already constrained within published uncertainties or contributes negligibly to the $c^{\rm measured} = c + \delta c$0 uncertainty; and (iii) non-analytic (singular) responses are excluded both on physical grounds — the COT is a single, fully instrumented volume with no dead regions — and empirically, with any residual discontinuity bounded at the level of 5 ppb on $c^{\rm measured} = c + \delta c$1, four orders of magnitude below the quoted 25 ppm calibration uncertainty.

The analytic response function

With no loss of generality, the response is expanded about $c^{\rm measured} = c + \delta c$2:

$c^{\rm measured} = c + \delta c$3

where $c^{\rm measured} = c + \delta c$4 is the charge and $c^{\rm measured} = c + \delta c$5 distinguishes outgoing from incoming trajectories. The $c^{\rm measured} = c + \delta c$6 terms capture charge-antisymmetric imperfections; the $c^{\rm measured} = c + \delta c$7 term arises because ionization energy loss shifts $c^{\rm measured} = c + \delta c$8, generating a $c^{\rm measured} = c + \delta c$9-like contribution distinguishable from geometric sources only by comparing incoming and outgoing cosmic-ray legs. Energy loss also induces an $c \equiv q/p_T$0 term indistinguishable from geometrical $c \equiv q/p_T$1. Hard scattering off sense wires is shown via a Rutherford-cross-section estimate to modify the intrinsic resolution at less than the permille level, justifying the simple $c \equiv q/p_T$2 track fit.

Constraints from in-situ cosmic rays

Cosmic-ray muons, acquired by the same trigger paths as the $c \equiv q/p_T$3 and $c \equiv q/p_T$4 data and synchronous with beam crossings within nanoseconds, provide a control sample collected under identical operating conditions. The key observable is $c \equiv q/p_T$5, which isolates $c \equiv q/p_T$6, $c \equiv q/p_T$7, $c \equiv q/p_T$8, $c \equiv q/p_T$9, and $a_0$0; the complementary $a_0$1 is shown analytically to be blind to its coefficients when the dicosmic helix curvature $a_0$2 is used as proxy for $a_0$3, since $a_0$4 by construction. The dicosmic fit, spanning 274 cm versus 96 cm per leg, resolves curvature better than the individual legs by a factor of $a_0$5, making it a valid proxy; the substitution introduces relative inaccuracies below 0.01% for all $a_0$6 coefficients.

Two results stand out numerically. First, the fit yields $a_0$7 MeV, consistent with the ab initio value of ~9 MeV used in the CDF analyses, and minimally correlated with other parameters (maximum correlation coefficient 8%). Second, after constraining the post-alignment $a_0$8, the coefficients $a_0$9, $a_1$0, and $a_1$1 are statistically consistent with zero, bounding these imperfections directly. The fits exhibit very large anticorrelations ($a_1$2–$a_1$3 at $a_1$4%, $a_1$5–$a_1$6 at $a_1$7%), which the paper exploits: removing redundant parameters per the Akaike Information Criterion stabilizes the fits and reduces the effective uncertainty on $a_1$8 by a factor of five.

Propagation to invariant-mass biases

For two-body decays into massless daughters, the first-order fractional mass bias reduces to $a_1$9, which separates cleanly into charge-symmetric terms $\epsilon$0 (with $\epsilon$1) that bias the mass, and charge-antisymmetric terms $\epsilon$2 that cancel. At second order the bias is approximately $\epsilon$3, since $\epsilon$4 is negligible. This decomposition has direct consequences: the $\epsilon$5 parameters are invisible in inclusive mass fits but measurable through charge-asymmetry observables, while the $\epsilon$6 parameters are calibrated by the $\epsilon$7 and $\epsilon$8 mass fits.

Evaluating $\epsilon$9 with the cosmic-ray constraints gives second-order biases of order 0.2–0.3 ppm at the $b_2$0 scale ($b_2$1 GeV), dominated by $b_2$2 and $b_2$3; at the $b_2$4 scale the largest term is $b_2$5, contributing ~2 ppm, reduced to a combined bound of 1.6 ppm once the $b_2$6% anticorrelation with $b_2$7 is included. A dedicated fit with $b_2$8 alone finds a small but discernible value $b_2$9 GeV$b_1$0, corresponding to only ~1 ppm on the $b_1$1 mass.

Bounds on $b_1$2 and $b_1$3 from quarkonium data

Because the cosmic-ray sample spans only $b_1$4 GeV$b_1$5, the paper derives independent bounds from the $b_1$6 and $b_1$7 data, which extend to $b_1$8 GeV$b_1$9. The 20 ppm consistency between the momentum calibrations extracted from the two mesons bounds the extrapolated effect of $a_2$0 at $a_2$1 to 15 ppm; the deviation observed in the highest-curvature $a_2$2 bins — identified as the largest single systematic in the CDF momentum calibration — bounds it to 19 ppm. Combined, these constrain $a_2$3 to 12 ppm, already included in the published 25 ppm uncertainty. Analogous reasoning bounds $a_2$4 to 4 ppm from the meson comparison and 5 ppm from cosmic rays, combining to 3 ppm. Notably, the pulls from the $a_2$5 and $a_2$6 data on both $a_2$7 and $a_2$8 act in opposite directions, suggesting their combined central values vanish.

Application to the $a_2$9 measurement

Exploiting the charge symmetry of Tevatron $b_3$0 production, the half-difference of fractional mass biases between $b_3$1 and $b_3$2 equals exactly $b_3$3, while the charge average equals $b_3$4. Two observables therefore pin down the model: $b_3$5, measured consistent with zero at 0.2‰ statistical precision, and $b_3$6, the positron–electron half-difference of $b_3$7, calibrated to zero within 43 ppm. Both constrain precisely the low-curvature $b_3$8 terms ($b_3$9, $c^{\rm measured} = c + \delta c$00). The second-order bias on $c^{\rm measured} = c + \delta c$01 then evaluates to $c^{\rm measured} = c + \delta c$02 ppm (or 0.002 ppm using $c^{\rm measured} = c + \delta c$03), plus vanishing contributions from $c^{\rm measured} = c + \delta c$04 and the cross term. The paper concludes that the most general analytic response function is "completely pinned down": $c^{\rm measured} = c + \delta c$05 for all resonances after calibration, with $c^{\rm measured} = c + \delta c$06 tuned to zero within 34 keV and $c^{\rm measured} = c + \delta c$07 bounded as above.

A pointed methodological claim appears here regarding four low-$c^{\rm measured} = c + \delta c$08 points in the published $c^{\rm measured} = c + \delta c$09 data that deviate from the linear model by $c^{\rm measured} = c + \delta c$10 if combined. The paper argues there is no justification for a non-linear fit to accommodate this fluctuation, since the only viable non-linear term $c^{\rm measured} = c + \delta c$11 is eliminated by alignment quality and the $c^{\rm measured} = c + \delta c$12 constraint, and would have negligible impact in the relevant $c^{\rm measured} = c + \delta c$13 range. This is a defensible but assumption-laden position: it rests on the completeness of the analytic model and the absence of unmodeled systematics at low curvature.

Singular response functions

The paper systematically excludes non-analytic extensions. Laurent terms $c^{\rm measured} = c + \delta c$14 are forbidden physically: as $c^{\rm measured} = c + \delta c$15 they imply indeterminate straight-line trajectories, inconsistent with the COT's fully active volume whose wire planes are tilted $c^{\rm measured} = c + \delta c$16 azimuthally symmetrically, leaving no dead space. Empirically, hit efficiency actually increases slightly (~1‰) as tracks straighten, superlayer efficiency is constant at 99.95% down to $c^{\rm measured} = c + \delta c$17 with a charge-dependent discontinuity parameter $c^{\rm measured} = c + \delta c$18%, and the inefficiency is stable at $c^{\rm measured} = c + \delta c$19 ppm over the full ten-year operation — supporting smoothness four orders of magnitude more stringently than the 25 ppm calibration uncertainty.

Puiseux terms $c^{\rm measured} = c + \delta c$20 with $c^{\rm measured} = c + \delta c$21 are addressed next. Fits including $c^{\rm measured} = c + \delta c$22 return $c^{\rm measured} = c + \delta c$23, i.e., redundancy with $c^{\rm measured} = c + \delta c$24 (correlation $c^{\rm measured} = c + \delta c$25%), and such terms are additionally rejected as unphysical since they imply zero curvature is distinguishable from vanishing curvature. The charge-dependent limit collapses to a step function $c^{\rm measured} = c + \delta c$26 — the worst case, since it cancels in $c^{\rm measured} = c + \delta c$27, $c^{\rm measured} = c + \delta c$28, and $c^{\rm measured} = c + \delta c$29 by charge symmetry, yet biases masses by $c^{\rm measured} = c + \delta c$30 and cannot be extrapolated away from low-$c^{\rm measured} = c + \delta c$31 calibrations. Cosmic-ray studies of left-right drift asymmetry and average drift displacement find no discontinuity within 1.2 μm, implying a bound of 5 ppb on any induced $c^{\rm measured} = c + \delta c$32 bias. No physical model generating $c^{\rm measured} = c + \delta c$33 from chamber geometry exists; the appendix offers synchrotron radiation as the sole particle-interaction mechanism, yielding a vanishingly small coefficient ($c^{\rm measured} = c + \delta c$34 ppb).

Limitations and open questions

Several caveats bear directly on the conclusions. The claim that the analytic model is complete rests on the AIC-based truncation and on the assumption that spatial non-uniformities of the coefficients are adequately averaged; residual polar-angle dependence from imperfect wire-shape modeling persists and is handled by a quadratic correction whose twist coefficient carries a 0.8 MeV systematic on $c^{\rm measured} = c + \delta c$35. The exclusion of non-linear fits to the $c^{\rm measured} = c + \delta c$36 data presumes no unmodeled low-curvature systematic, an assumption justified internally but not independently testable with the presented data. The bound on $c^{\rm measured} = c + \delta c$37 relies on the absence of any plausible detector-level mechanism; a dedicated simulation of drift-cell physics could strengthen this, and the paper notes this remains undone. Finally, the framework's applicability elsewhere is asserted rather than demonstrated: whether trackers with fragmented geometries (e.g., silicon trackers, or LHC detectors where lepton charge distributions are asymmetric) admit equally tight constraints is left open, as is the extension to the polar-angle response, which the paper deliberately does not treat.

Conclusion

This work provides a closed-form, physically interpretable model of the COT curvature response, demonstrating that every parameter of the general analytic ansatz is either measured, calibrated within published uncertainties, or negligible, and that singular behavior in the $c^{\rm measured} = c + \delta c$38 limit is excluded at the ppb level using in-situ cosmic-ray performance studies. The result substantiates the robustness of the CDF momentum calibration underlying the 25 ppm $c^{\rm measured} = c + \delta c$39 measurement without recourse to black-box methods, and establishes a transferable framework — contingent on a single, unfragmented active tracking volume — for precision magnetic trackers at future collider and fixed-target experiments.