Physical Calibration as the Bridge Between an Identifiable Model and a Correct Measurement

The highlighted sentence—“Physical calibration concerns the validity and absolute scale of that model”—is a compact warning about what self-calibration can and cannot do. In the parent document, the authors separate three ideas: identifiability, statistical efficiency, and physical calibration. The first two are largely internal to a mathematical observation model. Physical calibration asks a different question: whether that model is the right description of the apparatus, and whether its numerical output has the correct physical scale.

In the Rydberg receiver model used in the paper, the ideal squared-separation observation at LO setting $j$ and phase step $k$ has the form

$$ Y_{jk} = D+\left|L_j e^{i\phi_{jk}}+s\right|^2, $$

or, after writing the unknown signal as $s=a e^{i\theta}$ and $P=a^2$,

$$ Y_{jk} = D+L_j^2+P+2L_j\sqrt{P}\cos(\phi_{jk}-\theta). $$

Here $D=\delta^2$ is the squared microwave detuning, $L_j$ is the LO amplitude at the atoms, and $P$ is the signal power in squared Rabi-frequency units. Identifiability asks whether the unknowns can be uniquely recovered from noiseless values of $Y_{jk}$ under this equation. Statistical efficiency asks how much uncertainty remains when the observations are noisy. Physical calibration asks whether the measured spectra really obey this equation, and whether the recovered $P$ corresponds to the true microwave field amplitude in the laboratory.

That last question is not automatically answered by algebra.

Internal correctness versus physical correctness

A self-calibrating protocol can be internally successful while still physically wrong. Suppose the measured data are well fit by a sinusoid in phase, and the two LO settings allow the nuisance parameters $L_1$, $L_2$, and $D$ to be eliminated. Then, within the assumed model, the signal amplitude and phase may be identifiable. If the noise model is also accurate, the Cramér–Rao bound can describe the best possible local variance for unbiased estimators under that same model [Kay 1993; Van Trees et al. 2013].

But physical calibration asks whether the fitted quantities mean what the model says they mean. For Rydberg electrometry, that includes several nontrivial links: the optical spectral feature must be correctly converted into a microwave dressed-state separation; the dressed-state separation must be correctly related to a Rabi frequency; the Rabi frequency must be correctly related to electric field through the relevant atomic dipole matrix element and polarization component; and the phase and amplitude at the atoms must correspond to the intended signal and LO fields. These are experimental and atomic-physics issues, not merely inversion issues. Rydberg-EIT electrometry is attractive partly because atomic transition properties can provide traceable field measurements, but systematic uncertainties remain central to practical metrology [Holloway et al. 2014; Holloway et al. 2017; Schlossberger et al. 2024].

This is why the sentence says “validity and absolute scale.” Validity means that the functional form of the model is physically adequate. Absolute scale means that the units and numerical conversion factors are correct. A model may have the right shape but the wrong scale; for example, it may recover a signal proportional to the true electric field but with an incorrect calibration factor. Conversely, it may have an accurate scale factor for a simplified regime while being invalid outside that regime because additional spectral shifts or line-shape effects are present.

What “absolute scale” means in this setting

The parent document works in cyclic Rabi-frequency units. In an ideal two-level description, a microwave electric field component $E$ couples two atomic states through a dipole matrix element. In angular-frequency notation one often writes a Rabi frequency proportional to

$$ \Omega_{\mathrm{ang}} \propto \frac{d E}{\hbar}, $$

where $d$ is the relevant transition dipole matrix element and $\hbar$ is the reduced Planck constant. If cyclic frequencies are used instead, factors of $2\pi$ must be handled consistently. The parent document explicitly states its convention, which is important: a hidden angular-versus-cyclic mismatch would not necessarily destroy identifiability, but it would change the absolute scale of the inferred field.

In the paper’s inversion, the estimated signal power $P=a^2$ is not initially an electric-field power in volts squared per meter squared. It is a squared Rabi-frequency quantity inside the analytical model. To turn it into a physical field amplitude, one needs the atomic dipole calibration, polarization projection, line-shape interpretation, and any apparatus-dependent transfer factors. If any of those are wrong, the mathematical estimate can be precise and reproducible while still reporting the wrong field amplitude.

This distinction is especially important for the phrase “self-calibrating.” In the paper, self-calibration means that some nuisance quantities, such as unknown LO amplitudes or common detuning terms, can be inferred or eliminated from the same observations. It does not mean that every physical assumption has been experimentally validated by the protocol itself. Self-calibration is internal to the chosen model; physical calibration tests the model against the apparatus and external standards.

Why identifiability does not guarantee calibration

Identifiability is a uniqueness property. If the map from parameters to noiseless observations is one-to-one, then in principle the parameters can be recovered exactly from perfect data. For the sinusoidal observation model, multiple phase samples can identify a background, a modulation depth, and a phase; with two LO settings, the additional structure can identify the signal amplitude while treating LO amplitudes and detuning as nuisance parameters.

However, this uniqueness is conditional on the model. If the true apparatus produces

$$ Y_{jk} = D_j+\left|L_j e^{i\phi_{jk}}+s\right|^2 $$

with $D_j$ depending on LO setting, but the inversion assumes a common $D$, then the recovered signal power may be biased. The parent document later emphasizes this kind of observational ambiguity: an LO-power-dependent contribution to squared detuning can be indistinguishable from a different signal amplitude. In that case, the data can look perfectly consistent with the wrong model. A small residual is not enough to prove physical correctness, because the wrong physical effect may be absorbed into the model’s parameters.

This is the core meaning of “validity.” The model must exclude or correctly incorporate all effects that can mimic the signal. In Rydberg spectroscopy, such effects might include imperfect isolation of the transition, AC Stark shifts, line pulling, nonuniform fields over the vapor cell, polarization mismatch, optical-power-dependent shifts, or fitting biases in extracting a spectral separation. The parent document does not claim that all these effects are present; rather, it warns that the analytical model’s success must be physically qualified.

Why statistical efficiency does not guarantee calibration

Statistical efficiency is also conditional. A Cramér–Rao bound is derived from a probability model for the data. If the assumed noise is homoscedastic Gaussian noise in squared separations, then the Fisher information and the resulting bound describe that mathematical experiment. The bound can correctly show, for example, that the signal phase is orthogonal to certain nuisance directions under uniform phase sampling in the stated model.

But a statistically efficient estimator under a misspecified model can be efficiently wrong. If the spectral extraction noise is not homoscedastic, if the errors are correlated across phase steps, or if the mean response differs from the assumed sinusoid, then the formal variance calculation no longer answers the physical question. It may still be useful as a benchmark, but it is not a substitute for calibration of the measurement chain. Standard estimation theory makes this distinction clear: optimality claims depend on the assumed statistical model [Kay 1993; Van Trees et al. 2013].

Thus, physical bias and statistical variance are separate. A receiver may have very small random uncertainty and still have a large systematic error. Conversely, a physically well-calibrated receiver may have poor precision if the design is ill-conditioned or the signal-to-noise ratio is low.

How to read the highlighted sentence in the paper’s argument

The highlighted sentence serves as a boundary marker for the paper’s theoretical claims. The paper analyzes identifiability, acquisition complexity, information loss, and exact model ambiguities. Those are strong statements about the analytical model. But the authors explicitly state that the work is theoretical and does not claim experimental validation of a particular vapor-cell spectral model. Therefore, when the paper says physical calibration concerns validity and absolute scale, it is identifying the remaining bridge between theory and apparatus.

A careful verification of the parent document should therefore ask two kinds of questions. First, inside the model: are the identifiability proofs, Fisher-information calculations, and invariance claims mathematically correct? Second, outside the model: does an experimental Rydberg receiver produce observations close enough to the assumed squared-separation law, with known enough scale factors, for the inferred amplitude to be physically meaningful? The highlighted passage is about the second question. It does not weaken the mathematical analysis; it states the condition under which that analysis becomes a calibrated measurement of a real microwave field.

References

[Holloway et al. 2014] C. L. Holloway, J. A. Gordon, S. Jefferts, A. Schwarzkopf, D. A. Anderson, S. A. Miller, N. Thaicharoen, and G. Raithel, “Broadband Rydberg atom-based electric-field probe for SI-traceable, self-calibrated measurements,” IEEE Transactions on Antennas and Propagation, vol. 62, no. 12, pp. 6169–6182, 2014.

[Holloway et al. 2017] C. L. Holloway, M. T. Simons, J. A. Gordon, A. Dienstfrey, D. A. Anderson, and G. Raithel, “Electric field metrology for SI traceability: Systematic measurement uncertainties in electromagnetically induced transparency in atomic vapor,” Journal of Applied Physics, vol. 121, Article 233106, 2017.

[Kay 1993] S. M. Kay, Fundamentals of Statistical Signal Processing, Volume I: Estimation Theory. Prentice Hall, 1993.

[Schlossberger et al. 2024] N. Schlossberger et al., “Rydberg states of alkali atoms in atomic vapour as SI-traceable field probes and communications receivers,” Nature Reviews Physics, vol. 6, pp. 606–620, 2024.

[Van Trees et al. 2013] H. L. Van Trees, K. L. Bell, and Z. Tian, Detection, Estimation, and Modulation Theory, Part I, 2nd ed. Wiley, 2013.