Foundations of quantum mechanics · Theory
Trajectory sharpening under quantum reference frame transformations
arXiv:2610.40452v1 ↗PDF ↗Submitted 29 October 2026
- Question
- Quantum reference frame transformations are well developed mathematically but have had little contact with measurable effects. Separately, metrology uses a negative-mass reference to track a trajectory without the usual back-action. Are these the same thing, and how many references can a trajectory be sharp against at once?
- Result
- The two commuting observables behind trajectory sharpening become operators on distinct subsystems in the frame of the negative-mass system, and static and time-dependent transformations give the same trajectory. Two invariant sharpness measures obey a monogamy bound, \(\mathcal A_{B|A}\mathcal A_{B|D}\ge\hbar^2/4\) for every state, although a collective mode across \(A\) and \(D\) can still be an arbitrarily sharp reference.
The paper in detail
An ordinary free particle cannot have a sharp trajectory in the laboratory, because its position at two times does not commute: \[[\hat x(0),\hat x(t)]=\frac{i\hbar t}{m}.\] So \(\Delta\hat x(0)\,\Delta\hat x(t)\ge\hbar|t|/2|m|\). The older metrological trick asks about position relative to a second system \(A\) of mass \(-m\). The relative position then evolves as \[\hat Q(t)=\hat Q(0)+\frac tm\hat\Pi,\qquad\hat Q=\hat x_B-\hat x_A,\quad\hat\Pi=\hat p_B+\hat p_A.\] These are exactly the Einstein–Podolsky–Rosen combinations. They commute, so the whole relative trajectory has no uncertainty-relation floor. Effective negative mass is realistic: a collective spin polarised against a magnetic field behaves as an oscillator of negative mass.
The first result is a translation into frame language. The instantaneous transformation from laboratory \(C\) to frame \(A\) sends \(\hat Q\mapsto\hat x_B^{(A)}\) and \(\hat\Pi\mapsto-\hat p_C^{(A)}\). In \(A\)'s frame the mysterious commuting pair is just the position of one system and the momentum of another, and these commute because they belong to different subsystems. This mapping is kinematical and does not involve the masses. Mass enters through the dynamics, and the paper compares two treatments. Freezing the transformation at \(t=0\) gives a Hamiltonian with a \(B\)–\(C\) cross-coupling under which the position of \(B\) evolves. Keeping the full time dependence preserves the free-particle Hamiltonian and moves the time dependence into the observable. Both give the same trajectory.
Next comes a measure of how sharp a trajectory is. The marginal product \(\mathcal M=\Delta\hat Q\,\Delta\hat\Pi\) changes in time, while the covariance area \(\mathcal A=\sqrt{\det\Gamma}\) of the pair is conserved under opposite-mass evolution and equals \(\min_t\mathcal M(t)\). Both are invariant under the frame change. With a second candidate reference \(D\) of the same mass the position relative to one is conjugate to the momentum combination of the other, which yields the monogamy relation for arbitrary states. Choosing a different best time for each frame does not escape it.
The last result limits how this monogamy should be read. Rotating \(A\) and \(D\) into a collective mode \(F\) and an internal mode \(R\) keeps both at mass \(-m\), so \(F\) is a valid opposite-mass partner for \(B\) and can be arbitrarily sharp. For states whose \(R\) covariances decouple from \((B,F)\), the local references stay bounded below: \(\mathcal A_{B|A}\ge\beta^2\hbar/2\) and \(\mathcal A_{B|D}\ge\alpha^2\hbar/2\). So sharpness can live in a distributed degree of freedom that neither constituent provides alone. What remains open is an experiment, and the extension beyond free particles with exactly opposite masses.
Transferable methods
Reading a commuting pair as locality in another frame
The paper explains why the EPR observables commute by finding the frame in which they act on different subsystems, and uses this to link a metrological effect to a frame transformation.
The transformation \(\hat S_x=\hat{\mathcal P}_{AC}\,e^{\frac i\hbar\hat x_A^{(C)}\hat p_B^{(C)}}\) acts on operators by \[\hat x_A^{(C)}\mapsto-\hat x_C^{(A)},\quad\hat p_A^{(C)}\mapsto-\hat p_C^{(A)}-\hat p_B^{(A)},\quad\hat x_B^{(C)}\mapsto\hat x_B^{(A)}-\hat x_C^{(A)},\quad\hat p_B^{(C)}\mapsto\hat p_B^{(A)}.\]
Substituting gives \(\hat x_B-\hat x_A\mapsto\hat x_B^{(A)}\): the two \(\hat x_C^{(A)}\) terms cancel. Likewise \(\hat p_B+\hat p_A\mapsto-\hat p_C^{(A)}\), because the two \(\hat p_B^{(A)}\) terms cancel. The result acts on \(B\) alone and \(C\) alone, and any two such operators commute.
Since a unitary carries the state along with the operators, all moments are preserved, so variances and the covariance matrix \(\Gamma\) agree in both descriptions. That is why the sharpness measures take the same value in both frames.
The dynamics is a shear. With \(\hat\Pi\) conserved, \[\begin{pmatrix}\hat Q(t)\\\hat\Pi(t)\end{pmatrix}=\begin{pmatrix}1&t/m\\0&1\end{pmatrix}\begin{pmatrix}\hat Q(0)\\\hat\Pi(0)\end{pmatrix},\] and this matrix has determinant 1, so \(\det\Gamma\) is constant. Since \(\mathcal M^2=\mathcal A^2+\mathrm{Cov}(\hat Q,\hat\Pi)^2\), and the covariance vanishes at one time \(t_\ast\), \(\mathcal A\) is the minimum of \(\mathcal M\).
Editorial check: with \(\hbar=1\), \(m=1.3\) and an initial covariance matrix with entries \(0.7,-0.3,0.5\), the area stayed at \(0.5099\) for all tested times, and \(\mathcal M\) touched that value at \(t_\ast=0.78\), as the paper's formula for \(t_\ast\) predicts.
How to reuse it
Editorial suggestion: when observables that seem jointly delocalised commute, look for a unitary, possibly state-dependent, that makes them local to separate factors. This applies to stabilizers and to relational observables generally. Note that this particular map assumes translations; other frame changes will have other images.
Monogamy from crossed canonical pairs
Each relational pair commutes internally, so no uncertainty relation limits it. The constraint comes from pairing observables of different frames, which turns out to be canonically conjugate.
With references \(A,D\) and probe \(B\), direct calculation gives \[[\hat x_B-\hat x_A,\ \hat p_B+\hat p_D]=i\hbar,\qquad[\hat x_B-\hat x_D,\ \hat p_B+\hat p_A]=i\hbar,\] since only the \(B\) terms survive. All other commutators between the two pairs vanish.
Robertson's inequality on each crossed pair gives \(\Delta\hat Q_{B|A}\Delta\hat\Pi_{B|D}\ge\hbar/2\) and \(\Delta\hat Q_{B|D}\Delta\hat\Pi_{B|A}\ge\hbar/2\). Multiplying and regrouping the four factors gives \(\mathcal M_{B|A}\mathcal M_{B|D}\ge\hbar^2/4\).
For the stronger form, set \(\hat{\mathbf X}=(\hat Q_{B|A},\hat\Pi_{B|A})\) and \(\hat{\mathbf Y}=(\hat\Pi_{B|D},-\hat Q_{B|D})\), so that \([\hat X_i,\hat Y_j]=i\hbar\delta_{ij}\). For any real vectors \(\mathbf u,\mathbf v\), Robertson applied to the projections gives \[(\mathbf u^{\mathsf T}\Gamma_X\mathbf u)(\mathbf v^{\mathsf T}\Gamma_Y\mathbf v)\ge\frac{\hbar^2}4(\mathbf u^{\mathsf T}\mathbf v)^2.\] Optimising over \(\mathbf u\) gives the matrix inequality \[\Gamma_Y\succeq\frac{\hbar^2}4\Gamma_X^{-1}.\]
Taking determinants gives \(\det\Gamma_X\det\Gamma_Y\ge\hbar^4/16\), and \(\hat{\mathbf Y}\) is a determinant-one rotation of \(D\)'s pair, so \(\det\Gamma_Y=\det\Gamma_{B|D}\). Therefore \(\mathcal A_{B|A}\mathcal A_{B|D}\ge\hbar^2/4\). Because \(\mathcal A\le\mathcal M\), this implies the marginal bound, and because \(\mathcal A\) is the minimum over time, it holds with a separate best time for each frame.
Editorial check: we confirmed these commutators by symplectic algebra and evaluated the product on 20,000 random three-mode Gaussian states with \(\hbar=1\). The smallest value found was about \(0.62\), above the bound of \(0.25\). This sampling tests the bound but does not show how tight it is. For the collective mode with \(\alpha=\beta=1/\sqrt2\), the commutator \([\hat x_B-\hat X_F,\hat p_B+\hat P_F]=i\hbar(1-\alpha^2-\beta^2)\) vanishes, as the construction needs.
How to reuse it
Editorial suggestion: to find a trade-off among quantities that are each individually unconstrained, compute commutators between the sets. Then use the matrix form \(\Gamma_Y\succeq(\hbar^2/4)\Gamma_X^{-1}\) to keep correlations. It needs only finite second moments and fixed commutators, not Gaussian states.
Context & limitations
Why it matters. This gives an abstract frame change an interpretation in terms of an effect already seen in atomic-ensemble and mechanical experiments. The monogamy bound also says something new and testable: sharpness relative to a reference is a limited resource that cannot be shared between two separate systems.
Limitations. The model is a toy one of free particles with exactly opposite masses, and experiments realise negative mass only effectively, in oscillators. The collective-mode statement holds for a stated class of states whose covariances separate the internal mode. The paper proposes platforms but reports no measurement of the monogamy relation.