A prototype differential atom interferometer for basic physics

Cooling sequence

The cold-atom equipment used on this experiment has beforehand been described in refs. ^42,56. To put together samples of chilly ⁸⁷Sr, the atoms are first collected over 1.5 s in a blue three-dimensional MOT that makes use of the ¹S₀ → ¹P₁ transition at 461 nm and a subject gradient of three.5 mT cm⁻¹. Atoms that leak into the metastable ³P₂ manifold are recycled into the MOT utilizing repump lasers at 679 nm and 707 nm. For environment friendly repumping of ⁸⁷Sr, frequency sidebands at 585 MHz and 487 MHz are utilized to the 707-nm mild utilizing an electro-optic modulator to create frequency elements near-resonant with transitions from all 5 hyperfine manifolds of ³P₂ (ref. ⁵⁷).

When the blue MOT is switched off, the atoms are captured in a pink MOT working on the ¹S₀ F = 9/2 to ({}^{3}P_{{1}},{F}^{{prime} }=11/2) transition at 689 nm, utilizing a subject gradient of 390 μT cm⁻¹. Sidebands at 1,463.265 MHz are utilized to the 689-nm mild utilizing a resonant electro-optic modulator, such that the F = 9/2 to ({F}^{{prime} }=9/2) transition stirs the atoms between Zeeman sublevels of the bottom state, thus mitigating losses into sublevels the place atoms are weakly confined²⁸. During the primary 220 ms within the pink MOT, an depth of 1,800I_sat is used for every of the six MOT beams, the place I_sat = 3 μW cm⁻² is the saturation depth of the 689-nm transition. To seize the big selection of Doppler-shifted atoms launched from the blue MOT, a sawtooth-wave modulation is utilized to the 689-nm mild at a sweep frequency of 20 nm and a peak-to-peak sweep vary of 6 MHz (ref. ⁵⁸). For the next 100 nm, whereas within the ‘narrowband’ pink MOT, the sawtooth frequency modulation is switched off and the intensities of the six MOT beams are ramped linearly from 490I_sat to 40I_sat. To assist help the atoms in opposition to the pressure of gravity, a seventh, unbalanced MOT beam—the ‘up’ beam—is launched within the vertical route through the narrowband MOT. The up beam is important for creating narrowband pink MOTs beneath 100I_sat with out inflicting vital atom loss. Upon completion of the narrowband pink MOT, the atoms have a temperature of two μK and are compressed right into a area comparable in measurement with the optical dipole entice.

Dipole entice and state preparation

Two crossed optical dipole traps, separated vertically by 1 mm, are fashioned by separate 2.5-W horizontal beams at 1,064 nm with horizontal and vertical 1/e² radii of 220 μm and 23 μm, respectively, crossed with a shared 840-mW vertical beam at 813 nm with 1/e² radii of 60 μm in each transverse axes. Overlapping with the highest crossed dipole entice, a 4-mW transparency beam at 488 nm, detuned by 25 GHz from the 5s5p ³P₁ → 5s5d ³D₂ transition, is utilized with a 1/e² radius of 40 μm to guard the atoms from scattered 689-nm mild after they’re loaded into the highest crossed dipole entice area.

Immediately after the free-space pink MOT levels described above, the dipole trapping beams, the transparency beam and repumpers at 679 nm and 707 nm are switched on; the pink MOT is then held for 100 ms in a ‘top-trap loading’ stage, throughout which the bias magnetic fields, beam intensities and detunings of the pink MOT are optimized to load the atoms into the higher of the 2 dipole traps. During the top-trap loading stage, the pink MOT depth is linearly ramped from 20I_sat to 4I_sat to steadily scale back the atom temperature. Next, to load the underside optical dipole entice, the pink MOT is launched for 3 ms by switching off the 689-nm beams. During this time, the chilly atoms already within the high entice are held in place, whereas the warmer atoms fall in direction of the underside entice. While the atoms are falling, the vertical bias magnetic subject is stepped such that the zero of the quadrupole magnetic subject is near the underside dipole entice. After 3 ms of free fall, the pink MOT beams are switched again on for 100 ms in a ‘bottom-trap loading’ stage utilizing the identical parameters because the top-trap loading stage, apart from the completely different bias magnetic subject. All however the hottest atoms within the high entice stay within the high entice through the bottom-trap loading stage, as they’re protected by the 488-nm transparency beam in opposition to scattered 689-nm mild.

After each dipole traps are loaded, the MOT beams are switched off, a horizontal bias subject is utilized and the trapped atoms are optically pumped into the stretched state M_F = 9/2 by making use of a horizontal bias subject of 38 μT and delivering a 20-ms pulse of circularly polarized mild at 689 nm, resonant with the ¹S₀ F = 9/2 to ³P₁ ({F}^{{prime} }) = 9/2 transition. During the optical pumping, sawtooth-wave frequency modulation is utilized to the 689-nm mild at a price of 30 okHz over a variety of 6 MHz. Finally, all beams besides the dipole entice are switched off, and the bias magnetic subject is adiabatically ramped to the ultimate subject used for atom interferometry: 31 μT aligned with the linear polarization of the vertical 698-nm clock beam.

Velocity choice on the clock transition

The clock beam at 698 nm propagates vertically upwards by means of each dipole entice areas with a waist of 600 μm. The clock laser linewidth is verified in opposition to an unbiased cavity-stabilized laser to make sure that it’s beneath 2 Hz earlier than supply of the sunshine to atoms by means of an uncompensated 10-m fibre⁴². Clock spectroscopy sequences are carried out instantly after atoms are launched from each dipole traps. The excitation fraction is detected utilizing a 200-μs fluorescence pulse at 461 nm to detect the variety of atoms within the floor state ¹S₀, which is adopted by 3.5-ms repumping pulses at 679 nm and 707 nm and one other 200-μs fluorescence pulse at 461 nm to detect atoms which can be within the ³P₀ state after the interferometer sequence. Scattered mild from every 461-nm spectroscopy pulse is gathered in separate exposures of an electron-multiplying charge-coupled gadget (EMCCD) digital camera (Andor iXon Ultra 897), and a separate EMCCD picture with out atoms current is used to subtract background counts.

At the utmost out there clock energy of 640 mW, a Rabi π-pulse time of 44 μs is measured. However, the clock transition was noticed to have a peak excitation fraction of 0.3 and a Doppler-broadened linewidth of 60 okHz, which is significantly bigger than the 20-kHz Fourier restrict. To enhance the constancy of the Rabi pulses within the atom-interferometer sequence, a velocity choice process is used. The clock beam is pulsed on for 200 μs at 20 mW, which implements a π pulse that excites the slowest atoms to the higher clock state ³P₀. The atoms within the floor state are then pushed away utilizing a 500-μs pulse at 461 nm, leaving solely the gradual atoms within the ³P₀ state to enter the interferometer sequence. After this velocity choice sequence, a resonant, 44-μs Rabi π pulse yielded a peak de-excitation fraction of 90%.

Clock atom interferometry

The clock atom interferometry consists of a sequence of three resonant pulses on the 698-nm clock transition, with pulse areas π/2 − π − π/2, a π-pulse time t_π = 44 μs and a darkish time T = 200 μs between every consecutive pulse. For the info in Fig. 4, the section of the clock mild is all the time stepped deterministically through the darkish instances such that the phases of the primary, second and third pulses are 0, ϕ and 4ϕ, respectively, with ϕ starting from 0 to 2π in 100 steps in a randomized order. Each knowledge level within the right-hand facet of Fig. 4 is the results of 2 × 100 samples, interleaved between HLN and LLN samples. For the HLN samples, additional section steps had been utilized through the interferometer darkish instances (Fig. 3). The HLN samples had been drawn independently from a Gaussian distribution with a typical deviation of 4π rad and imply of 0 rad.

It is vital to tell apart between the 2 sorts of randomization employed on this work. For each the LLN and HLN datasets, the clock laser section is scanned deterministically by means of 100 values in randomized order; this scan-order randomization ensures that any spurious time-oscillatory indicators, similar to 50 Hz from room lights, are usually not aliased to appear to be obvious fringes. For the HLN dataset, we moreover utilized giant, uncorrelated section jumps between pictures, which absolutely randomize absolutely the section of every particular person interferometer on a shot-by-shot foundation. This per-shot section randomization mimics the regime anticipated in long-baseline atom interferometers, the place built-in laser frequency noise over multi-second interrogation instances will produce section excursions of many radians (see ‘Laser phase noise estimate for a kilometre-scale detector’ part). Under these situations, a single atom interferometer retains no recoverable section data, so this offers a stringent check of the noise rejection functionality of differential measurements. The section randomization absolutely masks the fringes in every particular person interferometer however doesn’t have an effect on the measurement of the relative section of the 2 interferometers.

Laser section noise estimate for a kilometre-scale detector

The section noise imparted onto the atoms by the laser can typically be calculated from the spectral density of the frequency fluctuations within the laser beam⁵⁹. In our prototype, the laser section imprinted on every atom interferometer in a single repetition of the interferometer sequence starting at time t is roughly ϕ_laser = φ(t) − 2φ(t + T) + φ(t + 2T), the place φ(t) is the time-dependent section of the laser subject oscillating as (cos (kz-{omega }_{0}t+varphi (t))). This approximation holds within the restrict of brief beam-splitter and mirror pulses separated by a darkish time T (ref. ³⁵). Treating φ(t) as a stationary noise course of with a one-sided energy spectral density S_φ(f) and making use of the optical Wiener–Khinchin theorem⁶⁰, we observe a variance within the interferometer laser section:

For a future long-baseline atom interferometer, we mannequin the clock laser as a thermal-noise-limited, cavity-stabilized laser⁶¹ with a flicker frequency noise spectrum of the shape ({S}_{varphi }(f)={S}_{varphi }(f=1,{rm{Hz}})instances {(1{rm{Hz}}/f)}^{3}). Propagating this purposeful kind by means of the above equation, the usual deviation of the interferometer laser section simplifies as (sqrt{langle {phi }_{mathrm{laser}}^{2}rangle }=4{rm{pi }}Tsqrt{mathrm{ln}(2)}sqrt{{S}_{varphi }(1,mathrm{Hz})}). To present an optimistic numerical estimate of the laser section, we assume a laser noise spectrum on the restrict of present laser know-how, with fractional frequency noise S_y(f) = (10⁻³³/f)/Hz (ref. ⁶²). For the ⁸⁷Sr clock transition at 698 nm, the corresponding noise spectral density of the clock laser section fluctuations could be (sqrt{{S}_{varphi }(1,{rm{Hz}})}) = 14 mrad/(sqrt{{rm{Hz}}}), leading to a typical deviation for the interferometer laser section (sqrt{langle {phi }_{{rm{laser}}}^{2}rangle }=710,{rm{mrad}}) for T = 5 s, the interferometer time projected for a kilometre-scale detector¹. Even for an interferometer repetition price of a number of pictures per second, the laser section noise imprinted on every particular person interferometer is, subsequently, far above the extent wanted to achieve the final word goal section decision of (1{0}^{5},{rm{rad}}/sqrt{{rm{Hz}}}) (ref. ¹), highlighting the necessity for laser noise cancellation within the differential section δϕ.

Compounding the necessities for laser section noise cancellation, a big momentum switch of n ≈ 10⁴ photon recoils is focused for long-baseline detectors¹, which boosts detector sensitivity however imprints laser section noise n instances onto every atom interferometer³⁵. Taking into consideration the big momentum switch, long-baseline interferometers will in all probability be within the absolutely phase-randomized regime explored by the HLN dataset on this work.

Differential bias section

To induce a constant relative section offset between the highest and backside atom interferometers, one other, horizontal 689-nm Stark-shifting pulse is utilized to solely the highest interferometer for 30 μs through the hole between the primary π/2 pulse and the center π pulse. The Stark-shifting beam is detuned by −80 MHz from the ¹S₀ F = 9/2 to ³P₁ F′ = 11/2 transition, with a waist of 500 μm and an influence of 1 mW, which induces a section shift particularly on atoms within the floor state (the decrease arm) of the highest interferometer. For the info on this paper, the Stark-shifting pulse is used to generate a bias differential section ϕ_Stark between the highest and backside interferometers, such that the info lie on a Lissajous ellipse (Fig. 4b) slightly than a straight line and, thus, comprise extra details about the differential section δϕ. A non-zero differential bias section is required for environment friendly, low-error extraction of δϕ, whether or not δϕ is extracted utilizing a maximum-likelihood estimator or the ellipse-fitting methodology. In a long-baseline detector, a darkish matter or gravitational-wave sign would induce fluctuations within the ellipse becoming angle, on high of the static bias.

Experimental management

Electronic management indicators are produced by means of the experimental management platform ARTIQ, which makes use of a field-programmable gate array⁶³. The management software program is written in Python and is offered as open supply at ref. ⁶⁴.

Phase extraction

We extract each fixed differential phases (used to quantify laser noise cancellation) and oscillatory-signal elements utilizing a unified unbinned maximum-likelihood evaluation. For every experimental shot i, we modelled the measured excitation fractions (y_A,i, y_B,i) from the 2 interferometers A and B as noisy observations of sinusoidal interferometer responses that share a shot-dependent widespread section ϕ_i however differ by a differential section δϕ(t_i). The widespread section ϕ_i is handled as a nuisance parameter and marginalized to yield a chance that relies upon solely on the differential section. In apply, we use this marginalized chance for inference: we report level estimates from the maximization of the marginal chance and compute uncertainties from repeated Monte Carlo simulations carried out with matching parameters and analysed utilizing the identical evaluation pipeline, following a hybrid Bayesian–frequentist strategy generally utilized in precision measurements and particle physics.

The per-shot chances are obtained by numerical integration over the widespread section utilizing a uniform prior on [−π, π]:

the place ({mathcal{N}}(cdot | mu ,{sigma }^{2})) denotes a Gaussian likelihood density. The response features p_A and p_B are sinusoidal fringe fashions of the shape ({p}_{{rm{A}}}(phi )={p}_{0,{rm{A}}}+frac{{{mathcal{C}}}_{{mathcal{A}}}}{2}cos ,phi ) and ({p}_{{rm{B}}}(phi )={p}_{0,{rm{B}}}+frac{{{mathcal{C}}}_{{mathcal{B}}}}{2}cos ,(phi +{rm{delta }}phi )), parameterized by offsets p_0,{A,B} and contrasts ({{mathcal{C}}}_{{{rm{A}},{rm{B}}}}), with noise variance ({sigma }_{{{rm{A}},{rm{B}}}}^{2}={p}_{{{rm{A}},{rm{B}}}}(1-{p}_{{{rm{A}},{rm{B}}}})/{N}_{{{rm{A}},{rm{B}}}}) describing the SQL ensuing from the measured N_{A, B} atoms within the two interferometers. This marginalization permits strong inference, even when particular person interferometer fringes are absolutely washed out by laser section noise.

Mode 1: differential-phase stability evaluation

For the steadiness evaluation (Allan deviation) introduced in Fig. 4c, we estimated a piecewise fixed δϕ over consecutive blocks of 141 pictures.

Mode 2: oscillatory-signal evaluation

For the oscillatory-signal searches introduced in Fig. 5, we parameterized the differential section as ({rm{delta }}phi (t)={rm{delta }}{phi }_{0}+S,sin (omega t)+C,cos (omega t)). This parameterization captures the leading-order differential-phase response anticipated from each gravitational waves and ultralight darkish matter fields, which might induce coherent oscillations by modulating the efficient mild propagation time or the atomic transition frequency. Signal significance is quantified utilizing a likelihood-ratio check statistic that compares the best-fitting mannequin of the sign with the null speculation (C = S = 0). When scanning over frequency, we calibrate the null distribution of the check statistic with Monte Carlo simulations to account for the trials issue. In the absence of an injected sign, the framework appropriately favours the null speculation. It, thus, offers a statistically well-defined reference for future sensitivity research. C and S could be transformed to amplitude A and section χ utilizing the formulation

The resolvable frequency band within the prototype is set by the efficient sampling interval within the experiment (set by the typical shot cycle time) and the commentary length. At low frequencies, the sensitivity is proscribed by the finite run length; at increased frequencies, it’s restricted by the shot price and useless time. The injected-signal exams subsequently probe the band the place the prototype has statistical energy over hour-to-day information. In a long-baseline detector, the identical evaluation framework applies, however the efficient response and optimum band are engineered by means of the interrogation time, repetition price and baseline to shift the instrumental peak sensitivity into the mid-frequency regime. Accordingly, the resolvable frequency band is instrument-dependent: the frequency band of the prototype implementation doesn’t signify an intrinsic limitation of differential atom interferometry nor of the evaluation framework itself.

Data filtering

The 461-nm, 689-nm and 698-nm laser locks had been monitored all through the experiment. Experiment runs wherein a number of of those locks failed or wherein the noticed variety of atoms in both entice was beneath a manually set threshold close to 60% of the median variety of atoms had been thought-about invalid and excluded from the info.

Number of atoms

Atoms are detected on the finish of atom-interferometer sequences by means of fluorescence imaging on an EMCCD digital camera. Under the belief that fluorescence scales linearly with the variety of atoms, the fluorescence sign could be transformed to the variety of atoms utilizing a calibration derived from absorption imaging of clouds of atoms ready beneath similar situations as these used for the atom interferometry. The variety of atoms N within the calibration dataset is extracted from the uncooked absorption pictures by means of the relation Nσ(ω) = ∫ OD(x, y) dx dy (ref. ⁶⁵), the place OD(x, y) is the optical depth of the pattern at transverse place (x, y) within the absorption probe beam and σ(ω) is the absorption cross part of the ⁸⁷Sr atoms on the laser frequency ω.

As the hyperfine shifts of the states ¹P₁ F = 7/2, 9/2 and 11/2 are respectively +37 MHz, −23 MHz and −6 MHz (ref. ⁶⁶), that are vital in contrast with the 30.5-MHz pure linewidth of the ¹P₁ state⁶⁷, the absorption cross part σ(ω) in ⁸⁷Sr typically relies on the polarization and M_F. To keep away from any reliance on direct measurements of the polarization of our absorption probe mild and the M_F state of the atoms, we, as a substitute, measured the absorption amplitudes of the three strains from ¹S₀ to ¹P₁ F = 7/2, 9/2 and 11/2 by finishing up spectroscopy over a ±120-MHz vary of detunings utilizing samples of atoms pumped into M_F = 9/2 with the identical preparation sequence used for calibrating the variety of atoms and for the atom-interferometry datasets. We fitted the height amplitudes σ_7/2, σ_9/2 and σ_11/2 of the three Lorentzians to the absorption spectroscopy knowledge utilizing mounted literature values for the linewidths and the hyperfine splittings between the Lorentzians^66,67. Finally, we calibrated the optical depth per unit atom utilizing the id that the sum of the height absorption cross sections should match the resonant absorption cross part for the less complicated isotopes with zero nuclear spin: ∑_Fσ_F = σ₀ = 3λ²/2π (ref. ⁶⁵). We obtained an uncertainty for the whole variety of atoms of 8% for the atom-interferometry datasets. This uncertainty is dominated by the uncertainty within the distinction within the variety of atoms between the calibration dataset and the fluorescence dataset.

For the mixed HLN and LLN dataset, the median variety of atoms within the high entice was 3,100(210) and within the backside 2,040(160). The variety of atoms in every entice fluctuated through the 61.9 h when the dataset was utilized, with a most deviation of 15% from the median. No vital distinction within the variety of atoms was noticed between pictures with and with out induced section noise.

Extracting noise ranges

To estimate some other type of noise in our measurement of δϕ attributable to injecting laser noise, we utilized the maximum-likelihood phase-extraction methodology independently to each the LLN and HLN datasets. The time collection of phases extracted from 141-shot blocks is modelled as

the place ({mathcal{N}}(mu ,{sigma }^{2})) denotes a standard distribution with imply μ and normal deviation σ. We use a No-U Turn Markov-chain Monte Carlo methodology carried out within the PyMC bundle⁶⁸ to pattern from the posterior distribution of σ_δϕ. The imply and 68% credible intervals had been σ_LLN = 3.69(19) mrad and σ_HLN = 3.89(20) mrad. To examine these with the usual deviations of Monte Carlo simulations with solely SQL current (see ‘Monte Carlo SQL’ part), we calculated from these per-block normal deviations the usual error on the imply over the entire dataset, giving ({sigma }_{langle {rm{delta }}{phi }_{mathrm{LLN}}rangle }=260(13),mathrm{mu rad}) and ({sigma }_{langle {rm{delta }}{phi }_{mathrm{HLN}}rangle }=275(14),mathrm{mu rad}).

Theoretical SQL

We outlined the SQL because the Cramer–Rao certain to the per-shot section noise σ_δϕ, calculated utilizing the straightforward chance mannequin in equation (3) wherein quantum projection noise is the one noise course of included. The Cramer–Rao certain is a decrease restrict to σ_δϕ for any unbiased estimator of δϕ, whatever the δϕ extraction approach used. It is used as a rigorous benchmark in differential atom interferometers and quantum sensors^7,47. For the Cramer–Rao SQL calculation, we differentiate the log-likelihood with respect to variations in δϕ round a central parameter set, akin to the contrasts, variety of atoms and imply δϕ extracted from a maximum-likelihood match to the complete dataset in Fig. 4. We additionally enter the median for the measured variety of atoms into the chance mannequin. The dominant supply of uncertainty within the Cramer–Rao SQL is the roughly 7% uncertainty within the calibration of the variety of atoms. We calculated a typical deviation of 43.5(16) mrad per shot. Over the entire dataset of 28,312 pictures, this ends in a decrease certain for the uncertainty in δϕ of 258(10) μrad.

Monte Carlo SQL

We validated the unbinned maximum-likelihood phase-extraction methodology and established the SQL reference utilizing Monte Carlo simulations that replicated the experimental sampling and noise funds. Synthetic pictures included the measured contrasts, imply numbers of atoms and their fluctuations, and the projection-noise-limited excitation read-out, with the identical estimator utilized as in the true knowledge evaluation. These exams verified that the estimator was unbiased and that the noticed section variance was according to quantum projection noise beneath the statistics for the measured variety of atoms.

We generated 5,100 artificial datasets such that the variation within the variety of atoms was according to the uncertainty of the imply from the absorption methodology described above, the recognized shot-to-shot variation inside datasets and 0 different noise sources, as proven in Extended Data Fig. 1. Each dataset consisted of 28,312 simulated interferometer pictures, every of which skilled contrasts of 0.81 and 0.84 for the 2 traps, and the median numbers of atoms had been 3,100(210) and a couple of,040(160), respectively, which matches our actual knowledge. We included gaps within the simulated datasets to match the distribution of gaps in our true knowledge. These gaps are attributable to numerous experimental calibrations and outages. We verified that the recovered δϕ values are unbiased throughout the statistical uncertainty and that the nominal 68% intervals have the right frequentist protection. By calculating an overlapping Allan deviation for every simulation run after which contemplating the distribution of Allan deviations throughout all generated datasets, we report the 68% and 95% credible intervals for a differential interferometer restricted solely by atom shot noise (the SQL), as proven in Fig. 4c. In distinction to the theoretical calculation, the Monte Carlo ensembles reproduce the complete experimentally noticed distributions for the variety of atoms, distinction and projection-noise-limited excitation read-out slightly than solely their imply values. The ensuing SQL reference is, subsequently, a prediction analysed with the identical estimator as the info.

Statistical compatibility with the SQL prediction was assessed utilizing two complementary exams utilized to the Allan deviation in log-space. A worldwide check statistic evaluating the measured values with Monte Carlo ensembles at every averaging time yielded p = 0.82 for the HLN dataset and p = 0.65 for the LLN dataset, indicating that there was no vital deviation from the Monte Carlo SQL prediction. Additionally, the measured Allan deviation slopes (s = −0.465 for HLN and s = −0.463 for LLN) are according to the Monte Carlo SQL ensemble, which itself displays white-noise scaling (s =−0.5), with p = 0.45 and p = 0.43, respectively.