The D⁰−D̄⁰ mass difference from a dispersion relation

The part worth your time is Sec. II, not the number in the abstract. Falk et al. show that by stripping the heavy quark mass out of the correlator — writing the charm field in HQET and keeping only the term that survives mc → ∞ — you get a function Σ_v(q) with a physical meaning at every q, not just at q = p_D. That is what the earlier dispersion treatments (their Ref. [13]) could not do, and it is what makes Eq. (19) possible: Δm as a principal-value integral of ΔΓ(E) over heavy meson mass. The analyticity argument is textbook and the derivation is explicit about what gets dropped, including the oscillating term replaced by local ΔC=2 operators, which they argue sits below 10⁻³. That is a genuine construction and it will get cited on its own merits regardless of the numbers that follow it. The numerical claim is much softer, and they say so. r(4P) lands between roughly -0.1 and -1 depending on whether you use g_4P or g'_4P in Eq. (30), and for the PP case they tell you outright not to trust the result. The honest reading is that the sign and rough magnitude of x/y are what the model gives, not what the theory predicts. On the stress-test point: the paper asserts in Sec. IV.B that the 4P integral is dominated by E near m_D without arguing it. But the left panel of Fig. 4 is the argument — r(4P) is flat as m1 runs from 0.2 to 1 GeV, which is exactly what you expect if the low-E region contributes little. Contrast Fig. 2, where r_PP swings by a factor of several over the same range. So the claim is supported by their own numerics; they just did not connect the dots in the text. The residual worry is the uncomputed O(Λ_QCD/E) correction, which they flag as O(1) at small E. If the low-E weight really is small, that correction is weighted down too, but nobody has checked it. Minor, and fixable with a sentence. The interpretive payoff is what matters for the field: a percent-level x is not by itself evidence of new physics, and the only clean signal would be CP violation. Useful to have on record. Who should read it: anyone quoting D-mixing bounds or building a new-physics case on x ≫ y. Cite the derivation; cite the number with the caveat attached. Send it out.

The reader's weakest_assumption is correct: the headline number r(4P) ~ -1 in Fig. 4 is set by the assumed form of g_4P(E) in Eq. (30), and the two models differ by a factor of two. The authors acknowledge this repeatedly (Sec. IV.B, Sec. V: 'we can trust it only at the order of magnitude level'). I do not present that as a new finding. What remains under-quantified is a specific technical point on which the reliability of the 4P result rests: the paper asserts that 'because the decay is close to threshold, the dispersion integral is dominated by E near m_D' (Sec. IV.B), which is what distinguishes the 4P case from the PP case where the authors themselves say 'we should not trust this result.' That assertion is not demonstrated. The dispersion relation Eq. (19) carries an explicit O(Λ_QCD/E) correction which the paper says is O(1) at small E; if the 4P integrand of Eq. (23) has non-negligible support below ~1 GeV, the 4P result inherits the same low-energy unreliability as the PP result, and the quoted range |x| ~ 10^-3 to 10^-2 is correspondingly less secure. This is a checkable gap, not a re-statement of the acknowledged model dependence.

The derivation of Eq. (19) is the solid part of the paper: stripping the heavy-quark mass from the correlator via HQET gives a function with a physical interpretation at every E, and the unsubtracted dispersion relation is justified by the 1/E^2 falloff of ΔΓ(E) in the model. I checked the factor-of-two in Eq. (23) and the analytic structure of the correlator; both are consistent. The numerical claim is softer, as the reader says, and the authors are unusually candid about it. My only addition is that the claim 'the 4P integral is dominated by E near m_D' is asserted rather than shown, and it is the load-bearing reason the 4P result is treated as more trustworthy than the PP result. The proposed test would settle whether that reason holds. If it does, the reader's CONDITIONAL verdict stands as written; if it does not, the verdict should remain CONDITIONAL but with the headline range explicitly widened. Either way, the reader's identification of the modeled g_F(E) as the weakest assumption is correct, and I would not move the verdict.

The dispersion relation itself is the durable contribution: it is derived without free parameters and converts a model of ΔΓ(E) — an observable that can in principle be computed or measured at other heavy-meson masses — into a prediction for Δm, which is not directly calculable. It yields a falsifiable, if crude, prediction (x and y comparable in magnitude and of opposite sign) that bears directly on how the current and forthcoming D-mixing bounds should be interpreted, and it sharpens the argument that a percent-level x need not by itself signal new physics. The authors are explicit that the numerical estimate is order-of-magnitude only and that the PP channel is not under control; that candor is appropriate and should be preserved in revision.

Neutral D mesons can turn into their antiparticles, but in the Standard Model this mixing vanishes in the limit of exact SU(3) flavor symmetry, so its size depends entirely on how badly that symmetry is broken. The authors ask whether the mass difference $x = 2M_{12}/2m_D$-driven quantity $x = 2M_{12}/2m_D$ can be as large as the width difference $y = 2M_{12}/2m_D$...

What carries the argument. The machinery is a correlator of two weak Hamiltonians sandwiched between heavy-meson states, $g_F(E)$, the energy dependence of the partial width to a given final state.

Core claim. The paper's central claim is a new relation plus a numerical consequence drawn from it. The relation, Eq. (19), states that the neutral D mass difference is a principal-value integral over the width difference of a heavy meson taken at varying heavy-meson mass, $x$ is comparable in magnitude to $y$ and of opposite sign, with $|x|$ between roughly $10^{-3}$ and $10^{-2}$.

The premise it rests on. The numerical answer rests entirely on the hand-modeled energy dependence of the partial widths $g_F(E)$ in Eqs. (27) and (30), whose free parameters and two alternative functional forms move the result by about a factor of two, and the authors say plainly that for the two-pseudoscalar case 'we should not trust this result.'

A percent-level measurement of $x$ in D mixing would no longer be, by itself, a signal of physics beyond the Standard Model.

The sign of $x$ relative to $y$ becomes a diagnostic: the phase-space mechanism predicts opposite signs, so a same-sign percent-level result would point elsewhere.

The result gives experimental searches a concrete target region, $|x| \sim 10^{-3}$ to $10^{-2}$, rather than a generic 'very small' expectation.

It identifies threshold and multi-body final states as the place where the Standard Model's D mixing is generated, which is precisely where operator-product-expansion treatments are blind.

Because the relation is an integral transform, a lattice or dispersive determination of how $\Delta\Gamma$ falls with heavy-meson mass would convert this estimate into a real prediction without any new theoretical idea.

The same transform could be applied to the $B_s$ system as a consistency check: there $\Delta\Gamma$ is measured and $\Delta m$ is measured, so the relation's convergence and its neglected $O(\Lambda_{\rm QCD}/E)$ corrections could be tested where both sides are known.

If the sign correlation between $x$ and $y$ survives better modeling, it is a cheap discriminant: a same-sign measurement at the percent level would be hard to accommodate without new physics or a very different SU(3)-breaking pattern.

The paper's logic suggests that any charm mixing prediction built purely on short-distance operator products will underestimate $x$, since the effect lives at thresholds that the expansion cannot resolve.

This review was written by a machine. No person has reviewed this paper on Pith yet. A scientist listed on Pith can review it for a fee, and their review is published here whatever it concludes.

Request a human review