Cold Molecule Frontiers

Status: direct laser cooling and magneto-optical trapping of selected diatomic and triatomic species, association of ultracold atoms into rovibrational-ground-state molecules, microwave shielding, molecular quantum degeneracy, and entanglement of molecule pairs are established. Larger polyatomic cooling cycles, scalable defect-free molecular arrays, controlled dipolar many-body phases, and cold-molecule symmetry measurements are active. A generally superior molecular-computing architecture, universally transferable collision shielding, molecular supersolidity, and a detected symmetry-violating signal are not established.

Last reviewed: 26 July 2026. The two 2026 dipolar-gas milestones discussed below are peer-reviewed publications. Array sizes, fidelities, exclusion limits, and publication status are date-sensitive.

Core Question

How can molecular complexity be converted from an uncontrolled source of loss, dephasing, and model uncertainty into a calibrated quantum resource?

The frontier is not simply to make a molecule cold. A useful platform must control a coupled chain:

$$\begin{aligned} \text{source} &\longrightarrow \text{translation and capture} \longrightarrow \text{internal-state preparation} \\ &\longrightarrow \text{survival and shielding} \longrightarrow \text{coherent interaction} \longrightarrow \text{validated inference}. \end{aligned}$$

Every arrow can be the bottleneck. A molecular beam may have high flux but short interrogation time. An assembled gas may occupy one internal state but inherit low end-to-end yield. A tweezer array may support exquisite pair control without yet supporting a many-qubit workload. A condensate may be unambiguously quantum degenerate while the phase diagram generated by its interactions remains unsettled.

The central research problem is therefore resource conversion: use rotation, vibration, parity doublets, permanent electric dipoles, and chemically sensitive short-range dynamics while measuring the leakage and uncertainty that those same structures introduce.

Why It Matters

Long-range interactions with local control

Polar molecules combine long-lived rotational states with electric-dipole matrix elements. Two molecules can interact over micrometre distances without being driven into a short-lived electronic Rydberg state. Geometry, dc polarization, and microwave dressing can change the strength, sign, and anisotropy of the interaction.

This is a distinctive combination, not an automatic advantage. The useful interaction must be compared with differential light shifts, motional dephasing, state-preparation errors, loss, and the time needed to assemble the sample.

Quantum matter with anisotropic interactions

The dipole–dipole interaction is long-ranged and orientation dependent. A stable degenerate molecular gas can therefore address collective behaviour that is difficult to isolate in contact-interacting atomic gases:

• anisotropic Fermi surfaces and collective modes;
• self-bound droplets and droplet arrays;
• long-range spin exchange;
• extended Hubbard and spin models;
• synthetic dimensions built from rotational states; and
• controlled competition between shielding, attraction, and chemical loss.

The 2024 molecular Bose–Einstein condensate and the 2026 observations of self-bound molecular droplets make this an experimental many-body field, not only a proposal programme. They do not, by themselves, establish every phase predicted for dipolar molecules.

Internal amplification for precision tests

Heavy polar molecules can expose electrons or nuclei to large molecule-internal effective fields. Opposite-parity doublets can permit strong laboratory polarization at modest applied fields and can provide internal reversal channels. These properties support sensitive searches for permanent electric dipole moments, parity violation, new spin-dependent forces, and variation of constants.

Cooling can increase interaction time and control, but it can also reduce detected flux and introduce trap shifts. The correct comparison is end-to-end information rate and systematic control, not temperature alone.

A bridge between chemistry and quantum engineering

At ultralow collision energies, a small number of partial waves and quantum states can dominate. State-selected reactants, confinement, electric fields, and microwave dressing can then reshape reaction and loss pathways. The same experiment can test scattering theory, engineer protected interactions, and study chemistry with prepared quantum correlations.

Canonical Boundary

This page owns the dated frontier assessment: what has been demonstrated, what is scaling now, what evidence is still missing, and what changed in the current review cycle. It does not repeat the mature derivations.

• Cold Molecules owns photon-cycle closure, direct cooling versus assembly, state preparation, trapping, polarization, dipolar interaction formulas, collision rates, and platform validation.
• Molecular Quantum Mechanics owns electronic, vibrational, rotational, hyperfine, and parity structure.
• Laser Cooling and Magneto-Optical Traps own radiative force, diffusion, capture, damping, and restoring-force theory.
• STIRAP owns the dark-state and adiabatic-transfer derivation used in ultracold assembly.
• Optical Tweezers and Optical Lattices own confinement, rearrangement, band, and Hubbard concepts.
• Precision Molecular Measurements owns effective fields, parity doublets, electron-EDM inference, nuclear response, protected comparisons, and theory calibration.
• Tests of Fundamental Symmetries owns the broader effective-operator interpretation and current symmetry limits.
• Fundamental Symmetry Frontiers owns the dated comparison of molecular EDM, NSD-PV, radioactive-molecule, nuclear-moment, and fifth-force programs.

Claims on this page should therefore be read as dated statements about platform maturity. Follow the canonical links for derivations and enduring definitions.

Minimal Background

Cold is a vector of resources

No single temperature characterizes a molecular platform. A compact resource vector is

$$\mathcal R = \left( N,\, T,\, \mathcal D,\, p_{\rm int},\, p_{\rm mot},\, \tau,\, \mathcal C,\, \Gamma_{\rm el}/\Gamma_{\rm loss} \right),$$

where:

• $N$ is the usable molecule number;
• $T$ is a translational temperature;
• $\mathcal D$ is phase-space density;
• $p_{\rm int}$ and $p_{\rm mot}$ are internal- and motional-state purities;
• $\tau$ is a relevant lifetime or coherence time;
• $\mathcal C$ is contrast or readout quality; and
• $\Gamma_{\rm el}/\Gamma_{\rm loss}$ compares useful elastic dynamics with loss.

Two platforms should not be ranked by one component while silently changing the others. A $6\ \mathrm{nK}$ condensate of hundreds of molecules, a millikelvin magneto-optical trap of thousands of heavy polyatomics, and a pair of ground-state molecules with second-scale coherence serve different questions.

End-to-end yield

For assembly from ultracold atoms, a transparent first model is

$$p_{\rm mol} = p_{\rm pair}\, p_{\rm assoc}\, p_{\rm tr}\, p_{\rm surv}\, p_{\rm det},$$

where pair preparation, association, coherent transfer, survival, and detection are distinct conditional probabilities. In an array with independent site success $p_{\rm mol}$, the probability of obtaining an already defect-free $M$-site array is

$$P_{\rm full}=p_{\rm mol}^{M}.$$

This exponential is why reservoir loading, nondestructive error detection, rearrangement, and repeated assembly matter. A high one-way STIRAP fidelity does not, by itself, imply a high array yield.

For direct cooling, the analogous product includes source flux, slowing, capture, optical-cycle survival, transfer to a conservative trap, internal state preparation, and readout. The relevant output is usable molecules per unit time, not only the temperature of those that survive.

Useful collisions versus loss

Evaporative cooling requires rethermalizing collisions to outrun inelastic or reactive loss. Define

$$\gamma = \frac{\beta_{\rm el}}{\beta_{\rm loss}} \simeq \frac{\Gamma_{\rm el}}{\Gamma_{\rm loss}},$$

for a fixed density and collision-energy distribution. In a simple independent-event picture, the probability that the next collision is useful rather than lossy is

$$p_{\rm el} = \frac{\gamma}{1+\gamma}.$$

Large $\gamma$ is necessary for efficient evaporation, but it is not a complete stability certificate. One-body loss, three-body loss, microwave technical noise, field-linked resonances, heating, and density evolution can become limiting.

Microwave shielding works by dressing rotational channels so that an incoming molecular pair encounters a repulsive long-range adiabatic potential before reaching lossy short range. Its performance depends on detuning, Rabi frequency, polarization purity, species, collision energy, and the nearby channel structure. “Microwave on” is not a universal guarantee of shielding.

Dipolar interaction and exchange scale

For induced laboratory-frame dipoles $d_1$ and $d_2$ separated by $\mathbf r=r\hat{\mathbf r}$,

$$V_{\rm dd} = \frac{ \mathbf d_1\!\cdot\!\mathbf d_2 -3 (\mathbf d_1\!\cdot\!\hat{\mathbf r}) (\mathbf d_2\!\cdot\!\hat{\mathbf r}) }{ 4\pi\epsilon_0 r^3 }.$$

A useful scale is

$$\frac{C_3}{h r^3} = \frac{d_{\rm eff}^{\,2}} {4\pi\epsilon_0 h r^3} \simeq 151\ \mathrm{Hz} \left(\frac{d_{\rm eff}}{1\ \mathrm D}\right)^2 \left(\frac{1\ \mu\mathrm m}{r}\right)^3,$$

before angular and state-matrix-element factors. The $r^{-3}$ dependence rewards close spacing but also makes motional spread, tweezer displacement, and geometric calibration part of the gate or simulator Hamiltonian.

If resonant exchange in a two-state subspace is described by

$$H_{\rm ex} = J \left( |01\rangle\langle10| +|10\rangle\langle01| \right),$$

then an initial $|01\rangle$ evolves into a Bell state after

$$t_{\sqrt{\mathrm{iSWAP}}} = \frac{\pi\hbar}{4|J|} = \frac{1}{8|J|/h},$$

and reaches a full iSWAP after twice that time. Leakage, motion, imperfect occupancy, and readout must be reported separately from fidelity conditioned on detecting both molecules.

Degeneracy and phase identification

For a trapped Bose gas, condensation can be supported by a bimodal momentum distribution, condensate-fraction scaling, and phase-space density crossing the appropriate threshold. For a Fermi gas, $T/T_F$ is the natural degeneracy measure. These are stronger claims than “nanokelvin.”

A self-bound droplet requires evidence that the cloud remains localized when external confinement is removed or made insufficient to bind it. That observation does not automatically establish:

• off-diagonal long-range order;
• superfluid response;
• crystalline translational order;
• simultaneous superfluid and crystalline order; or
• the microscopic stabilization mechanism without model dependence.

A claim of molecular supersolidity would require evidence for both phase coherence and density ordering, together with controls against fragmented, heated, or dynamically arrested droplets.

Precision response and null inference

A symmetry-sensitive molecular phase can be written schematically as

$$\phi_s = \frac{\Delta E_s\,\tau}{\hbar}, \qquad \Delta E_s = K_{se}d_e +K_{sS}C_S +\sum_j K_{sj}g_j.$$

The measured channel generally constrains a combination of effective operators. A quoted single-parameter bound is conditional on the remaining coefficients being fixed or marginalized according to an explicit model. The molecular response coefficients $K_{sj}$ are theoretical calibration inputs with uncertainties and correlations.

Cooling can increase $\tau$, but statistical sensitivity also depends on detected number, contrast, duty cycle, and technical noise. A useful shot-noise scaling is

$$\sigma_{\Delta E} \propto \frac{\hbar} {\mathcal C\,\tau\sqrt{N_{\rm det}}},$$

not $\tau^{-1}$ alone.

What Is Known

Direct laser cooling reaches trapped diatomics

Established. Quasi-closed optical cycling, radiation-pressure slowing, magneto-optical trapping, sub-Doppler cooling, loading into conservative traps, and single-molecule tweezer imaging have been demonstrated for a selected set of molecules, prominently including SrF, CaF, and YO.

These results establish that molecular vibrational and rotational structure does not categorically prevent laser cooling. They do not imply that an arbitrary molecule can be cooled by adding repump lasers. Successful species combine favourable electronic structure with measured branching ratios, rotational closure, manageable dark states, and accessible wavelengths.

CaF tweezer experiments have additionally demonstrated Raman sideband cooling to high motional ground-state occupation. That milestone closes a specific control gap between laser-cooled molecular arrays and coherent dipolar gates. It is not yet a demonstration of a large, fault-tolerant molecular processor.

Polyatomic cooling is experimentally established

Established for selected triatomics; active for broader chemical complexity. CaOH has been magneto-optically trapped and cooled below the Doppler limit. It has also been trapped and coherently controlled in optical tweezers with single-molecule imaging fidelity above $90\%$ in the reported experiment.

In 2025, a magneto-optical trap of the heavier polyatomic SrOH was reported with $2000(600)$ molecules, temperature $1.2(3)\ \mathrm{mK}$, and lifetime $91(9)\ \mathrm{ms}$. The lifetime was limited by decay into unaddressed vibrational states. This is a platform milestone for precision-oriented polyatomics, not yet an ultracold symmetry result.

The frontier is moving from “can a polyatomic molecule scatter enough photons?” to “can a heavier, more complex species be trapped, purified, coherently interrogated, and read out with a competitive information rate?”

Assembly produces state-selected ultracold molecules

Established. Magnetoassociation followed by coherent optical transfer has created rovibrational-ground-state samples of several bi-alkali species, including KRb, RbCs, NaK, NaRb, and NaCs. The method inherits low entropy from ultracold atoms and can deliver a selected hyperfine and rotational state.

Assembly has also reached the single-particle level in optical tweezers. A 2024 RbCs experiment began with arrays of up to eight Rb and eight Cs atoms, reported an overall ground-state molecule-assembly efficiency of $48(2)\%$, and combined multistate readout with formation-error detection and rearrangement.

The established fact is coherent molecular assembly and small-array control. Deterministic large-array preparation is still an engineering and scaling problem.

Microwave shielding enables quantum degeneracy

Established in multiple bi-alkali platforms. In 2022, microwave-shielded fermionic $^{23}\mathrm{Na}^{40}\mathrm K$ molecules were evaporatively cooled to $21\ \mathrm{nK}$, or $T/T_F=0.36$, with an elastic-to-inelastic collision ratio reported to exceed $460$ under favourable conditions.

In 2024, bosonic $^{23}\mathrm{Na}^{133}\mathrm{Cs}$ molecules were evaporatively cooled through Bose–Einstein condensation. The reported condensates reached $6(2)\ \mathrm{nK}$, condensate fraction $60(10)\%$, and a lifetime close to $2\ \mathrm s$.

In July 2026, $^{23}\mathrm{Na}^{87}\mathrm{Rb}$ molecules were condensed using dual microwave shielding. The reported condensates contained about $500$ molecules, and tuning the dressed interactions produced both gas-phase condensates and a self-bound droplet.

These experiments show that collisional instability is not an absolute bar to molecular quantum degeneracy. They do not show that one shielding recipe works unchanged across species, field geometries, temperatures, and many-body densities.

Dipolar molecular droplets have been observed

Established observation; active interpretation. In March 2026, self-bound droplets and droplet arrays were observed starting from a microwave-dressed NaCs molecular condensate. The experiment tuned interaction strength and anisotropy and reported densities up to about $100$ times that of the initial condensate. Ramp rate selected between robust one-dimensional arrays and fluctuating two-dimensional structures.

The observation establishes self-binding and structured droplet formation in a strongly dipolar molecular gas. The paper describes quantum-liquid and crystalline states as possibilities. It does not report a completed demonstration of a molecular crystal or supersolid.

The independent NaRb result published in July 2026 reported a gas-to-droplet transition identified by time-of-flight expansion. Together, the two platforms make droplet physics a reproducible molecular research direction, while leaving the equilibrium phase diagram and coherence properties open.

Pair entanglement and molecular gates are real

Established at the two-molecule scale. Independent 2023 experiments generated Bell states using dipolar exchange between individually trapped molecules. One used CaF molecules produced by direct laser cooling; another used assembled NaCs molecules.

A NaCs experiment published in the 2025 issue of Nature implemented a two-qubit iSWAP gate. At $1.9\ \mu\mathrm m$ separation, a $664\ \mu\mathrm s$ interaction generated a maximally entangled Bell state with reported fidelity $94(3)\%$ in trials conditioned on both molecules being present.

A separate RbCs experiment in rotationally magic tweezers reported a Bell state fidelity

$$0.924^{+0.013}_{-0.016},$$

with a leakage-corrected value

$$0.976^{+0.014}_{-0.016},$$

and second-scale entanglement lifetime limited by detectable leakage in the reported setting.

The correct conclusion is that controllable molecular entanglement and a molecular two-qubit gate have been demonstrated. Large connected arrays, parallel high-fidelity gates, repeated circuits, logical encoding, and fault-tolerant operation remain active goals.

Molecules set powerful null constraints on new physics

Established null tests. Molecular experiments provide some of the strongest constraints on an electron electric dipole moment. The trapped HfF$^+$ result reported

$$|d_e| < 4.1\times10^{-30}\ e\,\mathrm{cm} \qquad (90\%\ \text{confidence}),$$

under its stated single-source interpretation. The independent ACME ThO beam result reported

$$|d_e| < 1.1\times10^{-29}\ e\,\mathrm{cm} \qquad (90\%\ \text{confidence}).$$

Both results are consistent with zero. Neither is a detection of symmetry violation beyond the Standard Model.

Cold and trapped polyatomic programmes seek longer interaction time, opposite-parity internal comagnetometers, and sensitivity to electron- and nucleus-sector operators. Their projected reach is active, not a published exclusion limit until a blinded analysis, systematic budget, response calibration, and confidence construction are complete.

What Is Actively Developing

Direct cooling beyond a short species list

The immediate questions are no longer purely spectroscopic:

• Can branching measurements close all channels at the photon number needed for slowing, trapping, imaging, and repeated readout?
• Can dark states be destabilized without excessive diffusion?
• Can a molecular MOT load a conservative trap with high phase-space density?
• Can heavier species retain optical closure while supplying large symmetry-enhancement factors?
• Can polyatomic vibrational structure provide useful parity doublets and qudits without an unmanageable leakage graph?

Progress should be reported as an end-to-end molecule rate with state purity, not only as a newly identified cycling transition.

Deterministic assembly and array scaling

Small-array molecule assembly has all essential primitives: atom loading, pairing, association, coherent ground-state transfer, state-resolved readout, error detection, motion, and rearrangement. Scaling requires those primitives to work in parallel without correlated loss or calibration drift.

Important frontier metrics include:

$$\begin{aligned} R_{\rm useful} &= \frac{N_{\rm accepted}}{T_{\rm cycle}}, \\ \epsilon_{\rm site} &= 1-p_{\rm mol}, \\ \epsilon_{\rm corr} &= \Pr(\text{multi-site error}), \\ T_{\rm rebuild} &= \text{time to restore an accepted array}. \end{aligned}$$

A protocol that produces a defect-free array once is not yet scalable if rebuild time grows too rapidly, if errors are spatially correlated, or if the molecule state cannot be checked without destruction.

Shielding, controlled chemistry, and universality

Microwave shielding has enabled spectacular gains, but it remains a multichannel scattering problem. Active work targets:

• simultaneous suppression of two- and three-body loss;
• robust shielding against polarization impurities;
• adiabatic preparation of dressed states;
• control near field-linked resonances;
• tunable elastic scattering without heating;
• shielding in lower-dimensional traps and lattices; and
• extension beyond the currently demonstrated bi-alkali species.

A trustworthy universality claim would compare dimensionless parameters and channel structure across species, rather than relying on similar microwave frequencies or dipole moments.

Dipolar quantum matter

The 2026 droplet results open several experimentally separable questions:

1. What stabilizes each droplet in the explored regime?
2. Is the observed array an equilibrium state or a ramp-generated pattern?
3. Does phase coherence extend across droplets?
4. Is there static translational order in a structure factor?
5. Are the collective modes consistent with a liquid, crystal, or supersolid response?
6. How do three-body loss and microwave noise reshape the apparent phase boundary?

Future phase claims need observables beyond density images. Interference, matter-wave coherence, Bragg response, structure factors, compressibility, collective modes, and ramp-reversal tests probe different parts of the classification.

Many-body spin models and synthetic dimensions

Rotational states can encode spins or sites in a synthetic dimension. Dipolar exchange supplies long-range couplings, while microwave fields set local energies and transitions. Active experiments are moving from pair dynamics to larger arrays and ensembles.

The validation burden grows with system size. A fitted magnetization trace does not uniquely establish the intended Hamiltonian. Stronger evidence includes:

• independent calibration of pairwise $J_{ij}$;
• geometry and polarization sweeps;
• conservation-law checks;
• local correlation functions;
• held-out observables;
• finite-size and boundary-condition tests; and
• comparison with exact calculations in tractable subregions.

Molecular qubits, qudits, and error handling

Molecular rotation supplies many long-lived levels and strong microwave connectivity. This supports qubits, qudits, synthetic dimensions, memories, and erasure-aware encodings. The same multilevel structure creates leakage channels and spectral crowding.

The near-term frontier is not simply a larger Hilbert space. It is a well-characterized computational subspace with:

• high-fidelity state preparation and measurement;
• parallel local and global control;
• fast entangling gates relative to decoherence;
• leakage detection or conversion to a located erasure;
• low motional excitation;
• stable magic or near-magic trapping; and
• credible scaling of molecule supply.

Conditional Bell-state fidelity, unconditional accepted-event rate, and leakage probability should be reported separately. Each answers a different question.

Precision symmetry measurements with cold polyatomics

Polyatomic molecules offer parity-doublet structures that can polarize at small fields and support internal comagnetometry. Heavy species such as SrOH and YbOH are being developed to combine optical control with enhanced sensitivity to symmetry-violating operators.

The frontier measurement must close a long chain:

$$\text{molecules per shot} \times \text{state purity} \times \text{contrast} \times \text{coherence} \times \text{duty factor} \longrightarrow \text{information rate}.$$

It must then control field-correlated shifts, geometric phases, trap inhomogeneity, leakage, detection asymmetry, and theory response coefficients. A trapped sample is not automatically more sensitive than a beam if the gain in interrogation time is lost in count rate or systematics.

What Is Debated or Uncertain

Direct cooling versus assembly

There is no species-independent winner.

Direct cooling can offer rapid repetition, access to chemically diverse radicals, and naturally reconfigurable single particles. Its costs are photon-cycle closure, repumping, source capture, and optical scattering.

Assembly can offer nanokelvin motion and exceptional internal purity by inheriting atomic control. Its costs are two-species preparation, pair loading, resonant association, coherent transfer, and multiplicative yield.

The comparison should be made for a declared task and include cycle time, number, state purity, lifetime, and control overhead.

How universal microwave shielding will be

The repulsive dressed potential is conceptually general, but practical performance depends on molecular constants and technical fields. Near-resonant channel structure can introduce both protection and loss. It remains uncertain how broadly current ratios of elastic to inelastic scattering will transfer to new species and dense many-body regimes.

Whether molecular complexity is a computing advantage

Many rotational and hyperfine states can encode more information per particle, enable qudits, or create synthetic dimensions. They also increase calibration load, leakage opportunities, and control crosstalk.

An advantage must be demonstrated at the algorithm or error-corrected primitive level. State count alone is not a computational benchmark.

The phase of molecular droplets

Self-binding is observed. The labels “quantum liquid,” “crystal,” and “supersolid” require additional, distinct evidence. The present data motivate those hypotheses but do not collapse them into one established phase label.

How theory uncertainty limits symmetry inference

Molecular effective fields and nuclear response coefficients are calculated, not directly applied voltages. Correlated electronic-structure uncertainty can matter when several experiments are combined. The best experimental limit under a single-source assumption need not be the best constraint in a multi-operator global fit.

What counts as scalable molecular control

Two-molecule gates, small defect-free arrays, and second-scale coherence are necessary milestones. They do not determine the eventual scale at which loading, gate parallelism, calibration, crosstalk, leakage, and reconstruction remain manageable.

Key Experimental Platforms

Platform	Established capability	Frontier bottleneck	Evidence that matters next
laser-cooled diatomics	MOTs, conservative traps, tweezers, motional cooling, pair entanglement	species range, loading, repeated imaging, larger arrays	end-to-end rate, ground-state fraction, unconditional gate and readout errors
laser-cooled polyatomics	CaOH MOT and tweezers; SrOH MOT	vibrational leakage, heavy-species cooling, precision interrogation	resolved branching, trap lifetime, coherence, reversal-channel systematics
assembled bi-alkalis	selected ground states, degenerate gases, small tweezer arrays	multiplicative yield, rebuild time, correlated defects	large accepted arrays with state-resolved occupancy and calibrated duty cycle
microwave-shielded gases	degenerate NaK, NaCs BEC, NaRb BEC, molecular droplets	technical robustness, dense-gas loss, phase identification	collision maps, lifetimes, coherence, structure factors, collective response
lattice molecular spins	long-range exchange and site-resolved correlations	entropy, defects, Hamiltonian validation	local correlations and held-out observables across calibrated geometries
molecular tweezer qubits	Bell pairs, iSWAP, magic trapping, multistate readout	parallel gates, supply, leakage, motion, crosstalk	unconditional process benchmarks and repeated-circuit performance
molecular beams and trapped ions	strongest electron-EDM null limits	flux–time tradeoff and systematic control	blinded results, reversal closure, response covariance, independent replication
cold heavy polyatomics	control primitives for future symmetry tests	count rate, trapping, coherence, theory coefficients	a complete uncertainty budget and competitive measured limit

Key Theoretical Tools

Coupled-channel scattering

Molecular collisions mix rotation, partial waves, hyperfine structure, dc fields, microwave photons, and short-range loss. Coupled-channel calculations are the natural language for shielding, resonances, elastic scattering, and state-changing loss.

An absorbing boundary condition can model unit short-range loss, but that is a model choice. Agreement with measured rate coefficients over field, temperature, and polarization sweeps is more informative than a fit at one operating point.

Floquet and dressed-state methods

Microwave shielding and periodically driven spin models are naturally described in a dressed or Floquet basis. The approximation must retain enough photon sectors and rotational levels to converge the observables of interest. Adiabatic following can fail during ramps or collisions even when the static quasienergy diagram appears protected.

Open-system and leakage models

Molecular qubits are multilevel open systems. A minimal model separates:

$$\mathcal H = \mathcal H_{\rm comp} \oplus \mathcal H_{\rm leak} \oplus \mathcal H_{\rm loss}.$$

Dephasing within $\mathcal H_{\rm comp}$, coherent leakage, irreversible loss, and erasure detection have different operational consequences. Post-selection can diagnose a high-quality conditional operation while simultaneously reducing the unconditional computation rate.

Many-body effective Hamiltonians

Rotational states can realize XY, XXZ, extended Hubbard, and synthetic dimension models. Deriving an effective Hamiltonian requires a controlled projection and scale separation. Corrections can include:

• state-dependent tunnelling;
• longer-range and anisotropic $J_{ij}$;
• density-assisted terms;
• off-resonant microwave couplings;
• motion–spin coupling;
• loss and dephasing; and
• inhomogeneous fields.

Parameter identifiability should be checked. Several correction terms may produce similar low-order observables.

Beyond-mean-field dipolar theory

Gross–Pitaevskii and extended Gross–Pitaevskii descriptions provide useful starting points for condensates and droplets. Strong interactions, finite range, loss, microwave dressing, and non-equilibrium formation can challenge local beyond-mean-field approximations.

Theoretical labels should track the evidence:

• a density fit can support a model;
• a collective spectrum tests its dynamics;
• a structure factor tests ordering;
• interference and transport probe coherence and superfluid response; and
• agreement across these observables is stronger than agreement with one image.

Molecular electronic-structure calibration

Symmetry experiments need calculated effective fields and response coefficients. Reliable inference uses basis-set and correlation studies, relativistic effects, comparison with ordinary spectroscopic observables, uncertainty estimates, and covariance where several operators or species are combined.

Statistical decision models

Frontier claims often mix conditioning, post-selection, nuisance parameters, and multiple scans. A trustworthy analysis declares:

• the sample space and accepted events;
• whether fidelity is conditional on survival;
• confidence or credible interval construction;
• treatment of systematic nuisance parameters;
• look-elsewhere corrections for broad searches; and
• which analysis choices were fixed before unblinding.

Evidence Required for the Next Claims

Proposed claim	Minimum persuasive evidence
scalable molecular array	accepted occupancy and state maps, rebuild time, correlated-error analysis, repeated preparation over calibration times
high-fidelity molecular gate	unconditional process or logical benchmark, leakage and erasure channels, SPAM separation, motional and geometric sensitivity
universal shielding	dimensionless cross-species comparison, polarization and detuning robustness, elastic and all relevant loss channels
molecular crystal	static translational order and finite-size controls, not density modulation alone
molecular supersolid	simultaneous phase coherence and density order, plus dynamical or transport evidence
quantum advantage from qudits	task-level resource comparison including control, leakage, and error correction
improved symmetry limit	blinded estimator, complete systematics, response-coefficient uncertainty, explicit operator assumptions
new symmetry violation	replication, alternative reversals or species, exhaustive correlated-systematic tests, and global operator consistency

Common Misconceptions

“Cold” means internally pure

Translational temperature and internal-state entropy are different observables. A slowed beam can occupy many internal states; an assembled sample can be internally pure before it is collisionally stable.

A diagonal Franck–Condon matrix guarantees laser cooling

Vibrational branching is only one part of optical closure. Rotational branching, parity, hyperfine structure, dark states, electronic leakage, and available repumps all matter.

A high STIRAP fidelity is the molecule yield

STIRAP is one conditional step. Pair loading, association, survival, and detection multiply with it. Round-trip recovery also does not equal a one-way transfer probability without a model of both directions.

A long lifetime proves shielding

Lifetime can be limited or enhanced by density, one-body background loss, temperature, trap geometry, or state preparation. Shielding is supported by field-dependent elastic and inelastic collision measurements and comparison with channel calculations.

A condensate proves the target many-body phase

Condensation establishes macroscopic occupation and coherence properties of the gas. It does not establish a crystal, spin liquid, or supersolid without the observables specific to that phase.

A self-bound droplet is automatically a supersolid

Self-binding and supersolidity answer different questions. Supersolidity requires both density order and superfluid phase coherence.

Conditional fidelity is the same as end-to-end success

A conditional fidelity describes accepted trials. A processor or simulator also cares how frequently those trials occur, whether rejected events are located erasures, and how the conditioning scales with system size.

More molecular levels always help

Extra levels are resources only when they can be initialized, addressed, protected, and read out. Otherwise they are leakage channels.

A null EDM result rules out new physics

A null result excludes a region of an operator-dependent parameter space under stated assumptions. It neither proves exact symmetry nor eliminates all models beyond the Standard Model.

Colder is always better for precision

Lower velocity or trapping can increase interrogation time, but detected flux, duty cycle, contrast, trap shifts, and systematic reversals determine the final information rate.

Connections to Other Volumes

• Rotations and Orbital Angular Momentum supplies the tensor, coupling, and selection-rule language for rotational states.
• Open Quantum Systems gives master equations, quantum trajectories, loss channels, and decoherence tools for molecular control.
• Scattering Theory develops partial waves, threshold laws, resonances, and multichannel scattering.
• Effective Hamiltonians and Scale Separation explains projection, perturbative elimination, and model-validity criteria.
• Identical Particles and Exchange Symmetry underlies bosonic condensation, Fermi degeneracy, and exchange constraints.
• Lattice Models Overview provides Hubbard and spin-model language for molecular arrays.
• Entanglement in Quantum Information supplies fidelity, channels, erasures, gates, and error-correction benchmarks.
• Symmetry Principles gives the parity, time-reversal, and operator structure behind precision null tests.
• Fundamental Symmetry Frontiers separates molecular control milestones from completed symmetry measurements and current limits.

Within this volume, Spectroscopy connects observed structure to the state graph used for control, while Cold Molecules connects collision and loss measurements to the platform-level evidence ledger.

What Changed This Year

18 March 2026: self-bound NaCs droplets

Zhang et al. reported self-bound droplets and droplet arrays of microwave-dressed NaCs molecules in Nature. The work began from a molecular Bose–Einstein condensate, tuned dipolar strength and anisotropy, and observed density increases up to roughly two orders of magnitude. This changes the frontier from “can molecular dipoles support a self-bound many-body object?” to “what phase and stabilization mechanism does each droplet regime realize?”

Status change: self-bound molecular droplets moved from conjectural to established observation. Molecular crystalline order and supersolidity remain active hypotheses.

9 July 2026: NaRb Bose–Einstein condensation and a tunable droplet

Shi et al. reported a Bose–Einstein condensate of ground-state NaRb molecules using dual microwave shielding, with about $500$ molecules, and observed a gas-to-droplet transition by tuning the dressed interactions.

Status change: molecular Bose condensation and droplet formation now span more than one bosonic bi-alkali platform and more than one microwave dressing strategy. Broad shielding universality is still not established.

No symmetry discovery

No peer-reviewed cold-molecule result in this review cycle established a nonzero electric dipole moment or another symmetry-violating signal beyond the Standard Model. Platform advances should not be rewritten as discovery claims.

Control milestones retained from 2025

The 2025 literature established a heavy-polyatomic SrOH MOT and long-lived RbCs pair entanglement in rotationally magic tweezers. These remain current platform baselines. They are not new 2026 records, and this review does not silently redates them.

Representative Papers and Reviews

Field orientation

1. L. D. Carr, D. DeMille, R. V. Krems, and J. Ye, “Cold and ultracold molecules: science, technology and applications,” New Journal of Physics 11, 055049 (2009).
2. J. L. Bohn, A. M. Rey, and J. Ye, “Cold molecules: progress in quantum engineering of chemistry and quantum matter,” Science 357, 1002–1010 (2017).
3. M. R. Tarbutt, “Laser cooling of molecules,” Contemporary Physics 59, 356–376 (2018).
4. T. P. Softley, “Cold and ultracold molecules in the twenties,” Proceedings of the Royal Society A 479, 20220806 (2023).

Direct cooling and polyatomic control

5. N. B. Vilas et al., “Magneto-optical trapping and sub-Doppler cooling of a polyatomic molecule,” Nature 606, 70–74 (2022).
6. C. Hallas et al., “Optical trapping of a polyatomic molecule in an $\ell$-type parity doublet state,” Physical Review Letters 130, 153202 (2023).
7. N. B. Vilas et al., “An optical tweezer array of ultracold polyatomic molecules,” Nature 628, 282–286 (2024).
8. Y. Lu et al., “Raman sideband cooling of molecules in an optical tweezer array,” Nature Physics 20, 389–394 (2024).
9. Z. D. Lasner et al., “Magneto-optical trapping of a heavy polyatomic molecule for precision measurement,” Physical Review Letters 134, 083401 (2025).

Assembly, shielding, and quantum matter

10. K.-K. Ni et al., “A high phase-space-density gas of polar molecules,” Science 322, 231–235 (2008).
11. A. Schindewolf et al., “Evaporation of microwave-shielded polar molecules to quantum degeneracy,” Nature 607, 677–681 (2022).
12. N. Bigagli et al., “Observation of Bose–Einstein condensation of dipolar molecules,” Nature 631, 289–293 (2024).
13. S. Zhang et al., “Observation of self-bound droplets of ultracold dipolar molecules,” Nature 651, 601–606 (2026).
14. Z. Shi et al., “Bose–Einstein condensate of ultracold sodium–rubidium molecules with tunable dipolar interactions,” Nature Physics (2026).

Molecular arrays and quantum logic

15. C. M. Holland, Y. Lu, and L. W. Cheuk, “On-demand entanglement of molecules in a reconfigurable optical tweezer array,” Science 382, 1143–1147 (2023).
16. Y. Bao et al., “Dipolar spin-exchange and entanglement between molecules in an optical tweezer array,” Science 382, 1138–1143 (2023).
17. D. K. Ruttley et al., “Enhanced quantum control of individual ultracold molecules using optical tweezer arrays,” PRX Quantum 5, 020333 (2024).
18. L. R. B. Picard et al., “Entanglement and iSWAP gate between molecular qubits,” Nature 637, 821–826 (2025).
19. D. K. Ruttley et al., “Long-lived entanglement of molecules in magic-wavelength optical tweezers,” Nature 637, 827–832 (2025).
20. T. R. Hepworth et al., “Long-lived multilevel coherences and spin-1 dynamics encoded in the rotational states of ultracold molecules,” Nature Communications 16, 7131 (2025).

Precision and symmetry tests

21. ACME Collaboration, “Improved limit on the electric dipole moment of the electron,” Nature 562, 355–360 (2018).
22. T. S. Roussy et al., “An improved bound on the electron’s electric dipole moment,” Science 381, 46–50 (2023).
23. N. R. Hutzler, “Polyatomic molecules as quantum sensors for fundamental physics,” Quantum Science and Technology 5, 044011 (2020).
24. M. S. Safronova et al., “Search for new physics with atoms and molecules,” Reviews of Modern Physics 90, 025008 (2018).

Exercises

Exercise 1: Assembly yield and array scaling

A molecular tweezer platform has independent end-to-end site success $p_{\rm mol}=0.48$ before rearrangement.

1. Find the probability of obtaining already full arrays with $M=8$ and $M=20$ sites.
2. How many independent 20-site attempts are required on average for one already full array?
3. Explain why this calculation does not predict the performance of a reservoir-and-rearrangement protocol.

Solution

For independent sites,

$$P_{\rm full}(M)=0.48^M.$$

Thus

$$\begin{aligned} P_{\rm full}(8) &=2.82\times10^{-3},\\ P_{\rm full}(20) &=4.22\times10^{-7}. \end{aligned}$$

The mean waiting count for independent Bernoulli attempts is

$$\langle n\rangle = \frac{1}{P_{\rm full}(20)} \simeq 2.37\times10^6.$$

This is precisely why independent one-shot filling is not the scaling strategy. A reservoir-and-rearrangement protocol conditions on measured successes, moves molecules into target sites, and can repeat failed assembly. Its performance depends on reservoir size, imaging fidelity, move survival, correlated defects, and rebuild time. It is not described by $p_{\rm mol}^M$.

Exercise 2: A shielding collision budget

Suppose the ratio of useful elastic to lossy collisions is $\gamma=460$ and model each collision event as independently elastic or lossy.

1. What is the loss probability per event?
2. What is the probability of surviving 100 events?
3. Why is this not a complete evaporation model?

Solution

The per-event probabilities are

$$p_{\rm loss} = \frac{1}{1+\gamma} = \frac{1}{461} = 2.17\times10^{-3},$$

and

$$p_{\rm el} = \frac{460}{461}.$$

The simple survival probability is

$$P_{100} = \left(\frac{460}{461}\right)^{100} \simeq 0.805.$$

Real evaporation is not a fixed sequence of identical binary events. Density, temperature, trap depth, elastic cross section, one-body loss, three-body loss, and microwave dressing evolve. Evaporation also removes selected high-energy particles, so rethermalization and truncation efficiency must be modeled. The estimate is a screening calculation, not a prediction of final phase-space density.

Exercise 3: Dipolar range and gate time

Take an effective dipole matrix element $d_{\rm eff}=1.0\ \mathrm D$ and ignore angular reduction factors.

1. Estimate $C_3/(h r^3)$ at $r=2.0\ \mu\mathrm m$.
2. If this scale equals $J/h$, estimate the Bell-state and iSWAP times.
3. By what factor does the interaction increase if the separation is reduced from $2.0\ \mu\mathrm m$ to $0.8\ \mu\mathrm m$?

Solution

Using the scale above,

$$\frac{J}{h} \simeq 151\ \mathrm{Hz} \left(\frac{1}{2.0}\right)^3 = 18.9\ \mathrm{Hz}.$$

Therefore

$$t_{\sqrt{\mathrm{iSWAP}}} = \frac{1}{8J/h} \simeq 6.62\ \mathrm{ms},$$

and

$$t_{\mathrm{iSWAP}} = 2t_{\sqrt{\mathrm{iSWAP}}} \simeq 13.2\ \mathrm{ms}.$$

The geometric enhancement is

$$\frac{J(0.8\ \mu\mathrm m)} {J(2.0\ \mu\mathrm m)} = \left(\frac{2.0}{0.8}\right)^3 = 15.625.$$

Reducing spacing accelerates the gate, but the $r^{-3}$ law also amplifies position noise and motion–rotation coupling. The matrix element and angular factor must be calibrated for a real gate.

Exercise 4: Conditional fidelity and useful rate

A pair experiment produces a molecule in each tweezer independently with probability $p=0.80$. Conditioned on both molecules being present, its Bell state fidelity is $F_{\rm cond}=0.94$. The experiment repeats at $2.0\ \mathrm{Hz}$.

1. What fraction of trials contains a molecule pair?
2. What is the rate of accepted pair trials?
3. Compute $p^2F_{\rm cond}$ and explain what it does and does not mean.

Solution

The pair probability is

$$p_{\rm pair}=p^2=0.64.$$

The accepted-pair rate is

$$R_{\rm pair} = (2.0\ \mathrm{Hz})(0.64) = 1.28\ \mathrm{s^{-1}}.$$

The product

$$p_{\rm pair}F_{\rm cond} = (0.64)(0.94) = 0.602$$

is a useful scalar estimate of high-quality Bell-state weight per raw trial under this simplified model. It is not an unconditional state fidelity unless empty and single-occupancy outcomes are embedded in a declared Hilbert space and scored by a specified channel metric. If absence is detected, it may be a located erasure rather than an unlocated gate error. Reporting occupancy, conditional fidelity, leakage, and accepted-event rate separately preserves that distinction.

Exercise 5: Classify a droplet claim

An experiment releases a molecular cloud from its optical trap. For one microwave setting it expands; for another it remains localized and develops a repeatable density modulation. No interference, structure-factor, collective-mode, or transport measurement is reported.

Classify the strongest justified claims among:

1. self-bound droplet;
2. crystalline order;
3. superfluidity;
4. supersolidity.

State one additional observable needed for each claim not yet justified.

Solution

Persistence without sufficient external confinement supports the self-bound droplet claim, provided residual trapping and imaging artifacts are excluded.

A density modulation alone does not establish equilibrium crystalline order. A static structure factor with finite-size and shot-to-shot controls would be one relevant test.

Superfluidity is not established. Phase coherence, interference, quantized circulation, or a suitable transport and collective-response measurement would add evidence.

Supersolidity is not established because it requires both density order and superfluid phase coherence. The experiment would need both classes of observable in the same regime, along with controls showing that the modulation is not a non-equilibrium fragmentation pattern.

Exercise 6: Electron-EDM phase scale

Use the published HfF$^+$ bound as a scale:

$$|d_e|=4.1\times10^{-30}\ e\,\mathrm{cm}.$$

Take $E_{\rm eff}=23\ \mathrm{GV\,cm^{-1}}$ and interrogation time $\tau=3.0\ \mathrm s$. Estimate the phase magnitude

$$|\phi| = \frac{2|d_e|E_{\rm eff}\tau}{\hbar}.$$

Use $e=1.602\,176\,634\times10^{-19}\ \mathrm C$ and $\hbar=1.054\,571\,817\times10^{-34}\ \mathrm{J\,s}$.

Solution

Convert the dipole and field:

$$\begin{aligned} |d_e| &= (4.1\times10^{-30}) (1.602\,176\,634\times10^{-19}) (10^{-2})\ \mathrm{C\,m}\\ &= 6.57\times10^{-51}\ \mathrm{C\,m}, \\ E_{\rm eff} &= 23\times10^9\times10^2\ \mathrm{V\,m^{-1}} = 2.3\times10^{12}\ \mathrm{V\,m^{-1}}. \end{aligned}$$

The phase scale is

$$|\phi| = \frac{ 2(6.57\times10^{-51}) (2.3\times10^{12}) (3.0) }{ 1.0546\times10^{-34} } \simeq 8.6\times10^{-4}\ \mathrm{rad}.$$

This small phase is extracted statistically with reversals and an interferometric estimator. The factor of two depends on the convention for the two compared orientation states. A limit also depends on systematics and on the operator assumptions used to convert the measured channel to $d_e$.

Exercise 7: Build a status ledger

Assign established, active, conjectural, or unsupported to each statement as of this review:

1. Bosonic bi-alkali molecules can form a Bose–Einstein condensate.
2. Self-bound droplets of dipolar molecules have been observed.
3. The observed NaCs droplets are a molecular supersolid.
4. Molecule pairs can be entangled by dipolar interactions.
5. Molecular tweezer arrays have demonstrated fault-tolerant computation.
6. Molecular experiments have detected a nonzero electron EDM.
7. Heavy polyatomic MOTs can support future precision measurements.

Solution

1. Established. NaCs and NaRb condensates are peer-reviewed results.
2. Established. Self-bound NaCs droplets and a NaRb droplet transition have been reported.
3. Unsupported as a present fact. Supersolidity needs simultaneous coherence and density-order evidence not supplied by the droplet observation alone.
4. Established. Independent pair experiments and later gate work provide direct evidence.
5. Unsupported. Pair gates and small arrays are important primitives, not fault-tolerant computation.
6. Unsupported. The leading molecular measurements are null results.
7. Active. SrOH trapping establishes a platform primitive and motivates projected precision work; a competitive symmetry measurement is still to be demonstrated.

The exercise illustrates why the object and scope of each claim matter. “Observed droplets” and “observed supersolid” are not interchangeable, and “platform for a measurement” is not “completed measurement.”