Space-Based Universality Tests

An orbiting differential accelerometer can compare two compositions for many free-fall times while modulating the Earth-directed signal through spacecraft spin and orbital motion. Space does not remove systematics: it replaces ground vibration and suspension torque with drag-free control, electrostatic actuation, self-gravity, attitude coupling, thermal drift, data gaps, and colored noise. Establishing universality therefore requires a phase-coherent end-to-end measurement, not a small number in one Fourier bin.

Required background. Equivalence Principles and Gravitational Measurement supplies test-body universality and local frames. Probability and Statistical Inference supplies likelihood and covariance concepts; the explicit Gaussian model below provides the necessary local bridge.

Helpful background. Weak-Field Gravity connects the Earth potential to the predicted orbital acceleration, while Tests of Gravity places the test beside ground and strong-field comparisons.

Differential acceleration in orbit

For coaxial test masses 1 and 2, the measured difference is schematically

$$\boldsymbol\Gamma_d^{\rm meas} =\boldsymbol b_d +A_c\!\left[(T-I)\boldsymbol\Delta -2\Omega\dot{\boldsymbol\Delta} -\ddot{\boldsymbol\Delta} +\delta\,\boldsymbol g_{\rm sat}\right] +2A_d\boldsymbol\Gamma_c^{\rm app} +2C_d\dot{\boldsymbol\Omega} +\boldsymbol n_d.$$

Here $\boldsymbol\Delta$ is the test-mass off-centering, $T$ the Earth gravity-gradient tensor, $I$ the inertial-gradient contribution from spacecraft rotation, $\Omega$ the matrix representing spacecraft angular velocity, $A_c$ and $A_d$ common- and differential-mode sensitivity matrices, $C_d$ an angular-to-linear coupling, and $\delta$ the composition-dependent acceleration ratio. This is the structure of the MICROSCOPE measurement equation in Touboul et al. 2022, Eqs. (1)–(4) and Tables II–III, PDF, translated here only schematically so that no mission-specific matrix element is silently generalized.

For the mission’s spin opposite to orbital motion, spin places the Earth-directed WEP template at $f_{\rm EP}=f_{\rm orb}+f_{\rm spin}$ in the satellite frame. The frequency combination depends on the rotation convention. A genuine signal must have the predicted in-phase component with $\boldsymbol g_{\rm sat}$; a quadrature component is an important null channel. Equal-composition reference masses measure instrumental differential response without the target composition contrast. Their scale factors and off-centerings are separately calibrated: the two sensors do not have identical leakage coefficients Touboul et al. 2022, Eq. (3) and Table III, PDF.

The dominant geometric contamination can be understood without the full matrix model. In a circular orbit about a spherical Earth, $g=GM_\oplus/r^2$ and the radial gradient has magnitude $2g/r$. At illustrative values $g=8\ {\rm m\,s^{-2}}$, $r=7\times10^6\ {\rm m}$, and radial off-centering $\Delta r=20\ \mu{\rm m}$, the differential tidal acceleration is about $4.6\times10^{-11}\ {\rm m\,s^{-2}}$. This is much larger than the $8\times10^{-15}\ {\rm m\,s^{-2}}$ associated with a hypothetical $\delta=10^{-15}$. The separation comes from calibrated geometry and modulation, not from assuming the two masses occupy exactly the same point.

Data, calibration, and selection

The minimum dataset is synchronized differential and common acceleration, spacecraft attitude and orbit, drag-free control commands, temperatures, magnetic monitors, electrostatic actuation, and calibration-session metadata. The final MICROSCOPE article reports eighteen science sessions for the dissimilar-composition sensor and nine for the equal-composition sensor, after specified early nonlinear and rare anomalous sessions were removed Touboul et al. 2022, pp. 2–4 and Fig. 1, PDF. A reusable result must additionally pin the archive identity, calibration version, masks, and any superseding product.

Off-centerings are calibrated through their coupling to known gravity-gradient lines; scale factors and cross-axis terms use dedicated spacecraft excitations. Thermal sensitivity, magnetic response, electrostatic nonlinearity, and spacecraft self-gravity enter as transfer coefficients with shared covariance. Glitch removal is a selection: record thresholds, padding around transients, lost live time, and whether a gap-filling method was fixed before examining the WEP estimate.

With residual vector $\mathbf r(\delta,\boldsymbol\nu)$ and colored covariance $C(\boldsymbol\lambda)$, the likelihood is

$$\ln\mathcal L =-\frac12\mathbf r^{\mathsf T}C^{-1}\mathbf r -\frac12\ln\det C+\text{constant}.$$

The nuisance vector $\boldsymbol\nu$ contains calibration and systematic amplitudes; $\boldsymbol\lambda$ describes the noise. Priors must be stated, especially when external calibration data constrain the same coefficients across sessions. Segment selection and gap treatment can change the effective covariance and may not be hidden inside a nominal error bar.

A bounded computational benchmark

No new mission analysis was executed for this page. A reviewable benchmark would generate synchronized synthetic channels from a measured orbit and attitude, inject a blinded $\delta$ plus correlated noise and systematic templates, apply exactly the declared masks, and fit the same model used for the null data. It should be repeated at three time steps or frequency resolutions and at three filtering bandwidths. The recovered $\delta$ and its uncertainty must stabilize well below the claimed statistical precision.

The benchmark passes only if all of the following hold:

1. the injection is recovered with calibrated coverage;
2. split-orbit and split-session estimates agree within their covariance;
3. the equal-composition and quadrature channels remain consistent with their null models;
4. deliberately increased common-mode leakage is recovered as leakage, not $\delta$; and
5. omitting a relevant thermal or off-centering template produces a detectable residual failure.

A numerical reproduction should identify its actual interpreter and dependency versions, seeds, input hashes, precision, tolerances, and fit command. It should report failed convergence, coverage, and null-channel checks explicitly. This benchmark design does not supply a mission-data pipeline; the simpler gradient calculation above can be checked directly, whereas an independent MICROSCOPE reanalysis requires the mission inputs and calibrated response.

Evidence and limitations

Evidence assessed through 2026-08-19 includes the MICROSCOPE final primary result and the companion descriptions of systematic errors and data analysis. The primary article’s Eqs. (1)–(5), Fig. 1, and Tables II–III support the observable and reported summary; detailed conclusions about calibration and inference require the companion studies, especially Bergé et al. 2022, §§2–5, Rodrigues et al. 2022, §§3–6, and Touboul et al. 2022, §§3–5.

The mission result is a test of weak equivalence for its platinum–rhodium and titanium–aluminium–vanadium compositions in the Earth field. It is not a strong-equivalence test, does not cover interactions screened in orbit, and does not constrain charges to which those compositions have nearly equal sensitivity. A later calibrated time-series product, corrected gap model, or revised common-mode/thermal covariance should trigger reanalysis rather than a date-only update.

Exercise

Suppose common acceleration $\Gamma_c(t)$ leaks into the differential channel with unknown coefficient $\epsilon$, while the WEP template is $g_x(t)$. Explain how a blinded injection and an equal-composition channel distinguish $\delta$ from $\epsilon$.

Solution

Fit both templates, $d(t)=\delta g_x(t)+\epsilon\Gamma_c(t)+n(t)$, using their full covariance. Repeated blinded injections test bias and interval coverage; one recovered injection cannot establish coverage. If the two templates are proportional over the observed interval, the science channel alone measures only a combination of $\delta$ and $\epsilon$. The equal-composition sensor tests whether instrumental effects can produce an Earth-phase signal, but it has its own leakage coefficient. Therefore transferring its constraint to the science sensor requires a calibrated relation between the responses. Dedicated common-acceleration excitations and a change of spin frequency can supply the missing independent information.

References

• Bergé, Joël, et al. “MICROSCOPE Mission: Data Analysis Principle.” Classical and Quantum Gravity 39 (2022): 204007. DOI; Open PDF.
• Rodrigues, Manuel, et al. “MICROSCOPE: Systematic Errors.” Classical and Quantum Gravity 39 (2022): 204006. DOI; Open PDF.
• Touboul, Pierre, et al. “MICROSCOPE Mission: Final Results of the Test of the Equivalence Principle.” Physical Review Letters 129 (2022): 121102. DOI; Open PDF.
• Touboul, Pierre, et al. “Result of the MICROSCOPE Weak Equivalence Principle Test.” Classical and Quantum Gravity 39 (2022): 204009. DOI.