Rb vs Yb Qubit Comparison, AMO Toolkit

01 Why These Two Atoms?

The two leading neutral-atom qubit platforms, and why the choice of atom matters deeply.

Neutral atom quantum computers trap individual atoms in optical tweezers and entangle them via Rydberg blockade. The atom species determines everything from qubit coherence to gate speed to the complexity of the laser system. Two atoms have emerged as the leading candidates for scalable fault-tolerant neutral-atom QC, ⁸⁷Rb (an alkali, I=3/2, used by the Harvard/QuEra group) and ¹⁷¹Yb (an alkaline-earth-like rare earth, I=1/2, J=0 ground state, used by Atom Computing). Their physics are fundamentally different, leading to different strengths.

The core trade-off: ⁸⁷Rb is the mature, well-characterised platform with the strongest published Rydberg-gate and logical-processor record: 99.5% two-qubit CZ gates on up to 60 atom pairs operated in parallel, and 48 logical qubits encoded in 280 physical atoms. ¹⁷¹Yb has a different structural advantage: a J=0 ground state, nuclear-spin qubit, clock-state shelving, and alkaline-earth erasure-conversion protocols. Its best published gate number is 99.40(3)% raw and 99.72(3)% with post-selection. The 6100-atom, 12.6 s coherence result is a Cs-133 benchmark, not an Rb-87 benchmark; it is included below only as adjacent alkali/tweezer scaling context.

99.5%

Rb87 2Q fidelity
(Evered et al. 2023)

48

Rb87 logical qubits
(Bluvstein et al. 2024)

99.72%

Yb171 2Q fidelity
(Muniz et al. 2025, post-sel)

J = 0

Yb171 ground state
electronic protection

Source-confidence rule for this page: peer-reviewed numbers come directly from papers; company roadmap numbers are useful but not equivalent to peer-reviewed benchmarks; rough estimate entries are order-of-magnitude design guidance. The Cs-133 6100-atom result is explicitly marked as adjacent alkali context, not as an Rb87 panel result.

02 Atomic Physics Deep Dive

Electronic structure, relevant transitions, and why the physics dictates platform choices.

🔴 ⁸⁷Rb, Alkali Atom

Rb87 level structure. Qubit: hyperfine states of 5S₁/₂. Two-photon Rydberg excitation via 5P₃/₂ (780 nm + 480 nm). Microwave or Raman coupling for single-qubit gates.

Z = 37, ground config: [Kr] 5s¹, single valence electron
Nuclear spin: I = 3/2
Ground state: 5S₁/₂, electronic spin J = 1/2, so F = 1 or 2
Cooling transitions: D1 at 795 nm (5S→5P₁/₂), D2 at 780 nm (5S→5P₃/₂)
Rydberg excitation: Two-photon via 5P₃/₂ (780 nm + 480 nm)
Natural linewidth: Γ/2π = 6.07 MHz (D2)
Rydberg interaction scale: Strongly state- and field-dependent; use ARC/pair-state calculations for exact C₆. Near n≈70 S-states, C₆ is order 10⁵–10⁶ × 2π MHz·μm⁶ rough estimate

🟢 ¹⁷¹Yb, Alkaline-Earth-Like

Yb171 level structure. Qubit: nuclear spin mI=±½ in ¹S₀ ground state (J=0). Clock transition to ³P₀ at 578 nm. Rydberg excitation: single UV photon (~302 nm) from ³P₀ to Rydberg nP state. MOT on ¹P₁ at 399 nm; narrow-line cooling on ³P₁ at 556 nm.

Z = 70, ground config: [Xe] 4f¹⁴ 6s², two valence electrons (closed shell)
Nuclear spin: I = 1/2 (simplest possible)
Ground state: ¹S₀ (J = 0, critical for qubit protection)
Cooling transitions: ¹P₁ at 399 nm (broad), ³P₁ at 556 nm (intercombination)
Clock transition: ¹S₀ → ³P₀ at 578 nm (< 10 mHz linewidth)
Rydberg excitation: Single UV photon (~302 nm) from metastable ³P₀ state to Rydberg nP levels; overall path from ¹S₀ is two-photon (578 nm + 302 nm)
Clock magic wavelength: near 759 nm for the ¹S₀–³P₀ optical clock transition; not required to cancel the two ¹S₀ nuclear-spin qubit light shifts atomic structure

The J=0 advantage (correctly stated): In Yb171, the ¹S₀ ground state has J=0, so the qubit (nuclear spin mI=±½) couples only at the nuclear magneton scale (~2.6 kHz/G, linear in B). Rb87 clock states |mF=0⟩ have zero first-order Zeeman — their sensitivity is quadratic (575.15 Hz/G² × B² for the clock transition). At a typical bias field B₀≈1 G and noise δB=1 mG, Rb87 has ≈1.15 Hz sensitivity while Yb171 has ≈2.6 Hz sensitivity; the crossover is near B₀≈2.3 G. The primary structural advantages of Yb171's J=0 ground state are: (1) collisional immunity — no electronic spin means spin-exchange relaxation is highly suppressed; (2) spin-independent trapping of the ¹S₀ nuclear-spin qubit to leading scalar order; and (3) erasure conversion via ³P₀.

03 Qubit Encoding

How each system defines the two-level qubit, and the physics governing its frequency and noise sensitivity.

Rb87, Hyperfine Qubit

The qubit is encoded in two hyperfine ground states of the electronic 5S₁/₂ manifold, separated by the nuclear-electron hyperfine coupling $A_{\rm hf}$:

$$|0\rangle = |F{=}1, m_F{=}0\rangle, \quad |1\rangle = |F{=}2, m_F{=}0\rangle$$ $$\nu_{\rm qubit} = 6{,}834{,}682{,}610.904\;\text{Hz} \approx 6.834\;\text{GHz}$$ Clock states ($m_F{=}0$): first-order Zeeman is exactly zero.
Quadratic (second-order) Zeeman: $\delta\nu \approx +575.15\;\text{Hz/G}^2 \times B^2$
Sensitivity to fluctuations $\delta B$ at bias field $B_0$: $\delta\nu \approx 1150.3\;\text{Hz/G}^2 \times B_0 \cdot \delta B$

Driven by microwave pulses at 6.834 GHz (or two-photon Raman beams)
Clock states |mF=0⟩: zero first-order Zeeman — symmetry cancels linear Zeeman exactly for mF=0 ↔ mF=0 transition
Residual quadratic Zeeman: 575.15 Hz/G² × B²; at B=1 G → 575.15 Hz total shift, sensitivity dν/dB = 1150.3 Hz/G peer-reviewed
At typical bias field B₀=1 G, noise δB=1 mG: δν ≈ 1.15 Hz, about 2.3× smaller than Yb171's nuclear-spin slope
Differential light shift from tweezers must be calibrated or nulled

Yb171, Nuclear Spin Qubit

The qubit is encoded in the two nuclear spin states of the ¹S₀ ground state (I=1/2, J=0). With no electronic angular momentum, the only B-field coupling is through the nuclear magneton:

$$|0\rangle = |m_I{=}-\tfrac{1}{2}\rangle, \quad |1\rangle = |m_I{=}+\tfrac{1}{2}\rangle$$ $$\nu_{\rm qubit}(B) = g_I \,\mu_N \, B / h \approx -\,2.592\;\text{kHz/G} \times B$$ Zeeman shift: ${\sim}2.6\;\text{kHz/G}$ (nuclear magneton scale; qubit frequency ~26 kHz at $B=10\;\text{G}$)

Driven by RF pulses (kHz–MHz range in typical lab fields) or two-photon Raman
First-order Zeeman shift at nuclear magneton scale (~2.6 kHz/G); Rb87 clock states have zero first-order Zeeman and quadratic 575.15 Hz/G²
For the ¹S₀ nuclear-spin qubit, scalar light shifts are spin-independent to leading order; 759 nm is the ¹S₀–³P₀ clock magic wavelength, not a special cancellation between |mI=±1/2⟩ qubit states
No electronic magnetic moment in ground state → intrinsic noise immunity

B-field sensitivity comparison

$$\underbrace{\delta\nu_{\rm Rb87}^{\rm (clock)}}_{\text{quadratic Zeeman}} \approx 2 \times 575.15\;\text{Hz/G}^2 \times B_0 \cdot \delta B \qquad \text{vs} \qquad \underbrace{\delta\nu_{\rm Yb171}}_{\text{nuclear spin, linear}} \approx 2.6\;\text{kHz/G} \times \delta B$$ At $B_0 = 1\;\text{G}$, $\delta B = 1\;\text{mG}$: $\delta\nu_{\rm Rb87}^{\rm (clock)} \approx 1.15\;\text{Hz}$ vs $\delta\nu_{\rm Yb171} \approx 2.6\;\text{Hz}$ — Rb87 clock is lower-noise at 1 G.
At $B_0 = 10\;\text{G}$: $\delta\nu_{\rm Rb87} \approx 11.5\;\text{Hz}$ vs $\delta\nu_{\rm Yb171} \approx 2.6\;\text{Hz}$ — Yb171 is ~4.4× better.
At $B_0 = 0.1\;\text{G}$: $\delta\nu_{\rm Rb87} \approx 0.115\;\text{Hz}$ vs $\delta\nu_{\rm Yb171} \approx 2.6\;\text{Hz}$ — Rb87 clock is much better at low fields.
Crossover: $B_0 \approx 2.26\;\text{G}$ for the assumed $\delta B=1\;\text{mG}$.

04 Laser Cooling & State Preparation

From room temperature to single-atom quantum control, the cooling chain for each platform.

Rb87 Cooling Chain

2D/3D MOT, Doppler cooling on D2 line (780 nm) to ~200 μK, loading 10⁸–10⁹ atoms.
Gray Molasses (GM), Sub-Doppler cooling on D1 or D2 line using coherent dark states (Λ-enhanced). Achieves (4.0 ± 0.3) μK without a bias field.
Tweezer loading, Single atoms loaded stochastically into 820–850 nm tweezers. Single-atom detection via fluorescence. ~50% initial loading per site.
Sideband cooling, Raman or resolved-sideband cooling to motional ground state ⟨n⟩ < 0.1 per axis. Required to minimize Doppler dephasing during Rydberg gates.
Optical pumping, Spin initialization to |F=1, mF=0⟩ via σ⁺ + repumper. Fidelity >99.5%.

Key challenge: The D2 line is only 384 THz away from the 5P₃/₂ excited state. Tweezer photon scattering causes qubit dephasing unless the trap is far off-resonance. Typical trap depth ~1 mK.

Yb171 Cooling Chain

Zeeman slower + 2D MOT, Broad ¹P₁ transition at 399 nm (Γ/2π = 29 MHz) for initial deceleration to ~1 m/s, ~10⁹ atoms.
Intercombination MOT, ³P₁ transition at 556 nm (Γ/2π = 182 kHz, narrow linewidth) for sub-Doppler cooling to ~5 μK without a dedicated sub-Doppler stage.
Optical tweezer loading, commonly chosen to balance trap depth, scattering, and clock-state needs. 759 nm is magic for the ¹S₀–³P₀ clock transition; the ground-state nuclear-spin qubit is already nearly spin-independent in scalar traps.
Sideband cooling, Narrow ³P₁ line enables direct sideband-resolved cooling. Achieves ground state ⟨n⟩ < 0.1.
Nuclear spin initialization, Optical pumping via ¹S₀ → ³P₁ with circularly polarized light. Fidelity >99%.

Key advantage: The ¹S₀ ground-state nuclear-spin qubit has almost no electronic magnetic moment and a mostly spin-independent scalar Stark shift. The 759 nm magic condition is most relevant when coherently using ³P₀ clock/shelving states.

05 Coherence & Decoherence

How long quantum information survives, and the dominant dephasing mechanisms for each platform.

⁸⁷Rb — schematic dephasing

¹⁷¹Yb — schematic dephasing

Schematic only: exact T₂ is platform-, encoding-, and sequence-dependent.

Coherence Times (log scale)

Evidence audit: the 12.6(1) s, 6100-atom benchmark is Cs-133 from Manetsch et al. (Nature 2025), not Rb-87. Rb and Yb coherence numbers below are therefore treated cautiously and tagged by source confidence.

Rb87 Coherence

Published Rb array benchmark: 99.5% CZ gates on up to 60 atom pairs in parallel; coherence during algorithms is sufficient but not the 12.6 s Cs number peer-reviewed
Adjacent alkali benchmark: Cs-133, not Rb-87, reached T₂ = 12.6(1) s in a 6100-atom array not Rb
T₁ (energy relaxation): >1 s (thermal photon limited)
Dominant dephasing: Magnetic field gradient noise, laser phase noise, differential light shift in tweezer
Second-order Zeeman: 575.15 Hz/G² × B² for the 87Rb clock transition
Tweezer scattering: Off-resonant photon scattering at rate Γ_sc ∝ I/Δ² limits T₁

$$\text{Rb87 published gate benchmark: } \mathcal{F}_{\rm CZ}=99.5\% \text{ on up to 60 parallel atom pairs}$$ $$\text{Rb87 logical benchmark: } 48\;\text{logical qubits from }280\;\text{physical atoms}$$ The Cs-133 6100-atom coherence result is valuable alkali context, but should not be used as an Rb87 T₂ number.

Yb171 Coherence

T₂* (free precession, optical clock): >100 s demonstrated in optical lattice clocks
T₂ / memory: Atom Computing reports >40 s nuclear-spin coherence/memory in company materials; treat as company-reported until matched to a peer-reviewed large-array benchmark company
T₁ (energy relaxation): Limited by photon scattering from trap, expected comparable to Rb87
Dominant dephasing: Residual differential light shift (trap not perfectly magic), nuclear spin flip from electric quadrupole (suppressed by I=1/2)
B-field dephasing: linear nuclear-spin slope ≈2.6 kHz/G; compare against Rb clock-state quadratic sensitivity at the actual bias field
Clock-state advantage: ¹S₀–³P₀ magic trapping near 759 nm supports clock/shelving operations; ¹S₀ nuclear-spin qubit shifts are scalar to leading order

$$\delta\nu_{\rm Yb171} \approx g_I\mu_N B/h \approx 2.6\;\text{Hz/mG} \times B[\text{mG}]$$ At $B = 1\;\text{G}$: $\delta\nu_{\rm Yb171} \approx 2.6\;\text{kHz}$; Rb87 clock qubit quadratic shift is $\approx 575\;\text{Hz}$
Clock magic wavelength: $\Delta\alpha_{1S0-3P0}(\lambda_{\rm magic}\approx759\;\text{nm}) = 0$ for clock/shelving transitions

Bottom line on coherence: Do not compare Rb and Yb using the 6100-atom, 12.6 s number as an Rb result; it is Cs-133. For Rb87, the strongest peer-reviewed evidence is gate/logical-processor maturity. For Yb171, the strongest evidence is the nuclear-spin/clock-state structure and the 2025 gate benchmark. For both, present-day two-qubit gate errors dominate over idle decoherence in most short circuits.

06 Gate Performance

Single-qubit and two-qubit gate fidelities, Rydberg excitation physics, and error sources.

Gate Fidelity Comparison (2019–2025)

Two-qubit gate fidelities, best published values. Rb87: Evered et al. (Nature 2023). Yb171: Muniz et al. (PRX Quantum 2025, post-selection and raw). Surface code fault-tolerance threshold shown as dashed line.

Rb87 Gate Results

Best 2Q CZ fidelity: 99.5% on 60 qubit pairs simultaneously Evered 2023
Parallel entanglement: 60 Bell pairs in parallel, same 99.5% per pair
1Q gate fidelity: >99.9% (randomized benchmarking, 2024–25)
Gate mechanism: Rydberg blockade CZ gate via two-photon (780+480 nm) excitation to |70S₁/₂⟩
Gate speed: 0.5–2 μs per two-qubit gate (limited by Rabi frequency)
Key innovation: Optimal-control pulse shaping + atomic dark-state scattering suppression
Logical qubit demo: 48 logical qubits, fault-tolerant operations (Bluvstein 2024)
Array size for gate demos: Up to 280 physical qubits (Bluvstein 2024)

$$\mathcal{F}_{\rm CZ} = 1 - \varepsilon_{\rm Ryd} - \varepsilon_{\rm scatt} - \varepsilon_{\rm D} - \varepsilon_{\rm \Omega}$$ $\varepsilon_{\rm Ryd} \sim (\Gamma_{\rm Ryd} \tau_g)$, Rydberg decay error
$\varepsilon_D = \tfrac{1}{2}(k\, v_{\rm rms}\, \tau_g)^2$, Doppler dephasing (mitigated by sideband cooling)
Best: 1 − 99.5% = 0.5% total error budget

Yb171 Gate Results

Best 2Q CZ fidelity: 99.72(3)% with post-selection (Muniz et al. PRX Quantum 2025) Muniz 2025
Without post-selection: 99.40(3)%, still competitive with Rb87
Rydberg gate fidelity (F=1/2 Rydberg): 99.4(1)% (Ma et al. 2024), 3.3× improvement over earlier results
1Q gate fidelity: >99.5% (Muniz 2025)
Gate mechanism: Rydberg blockade via single UV photon (~302 nm) from ³P₀ (metastable) to Rydberg nP state
Gate speed: ~1–5 μs per two-qubit gate (similar to Rb87)
Unique feature: Can encode qubit in both nuclear spin AND clock state (³P₀), dual encoding protocols
Array size for gate demos: 1180-qubit Atom Computing Phoenix; gate fidelity demos on smaller subsets

$$\mathcal{F}_{\rm CZ}^{\rm Yb} = 99.72\%\;\text{(post-sel)}, \quad 99.40\%\;\text{(raw)}$$ Nuclear spin qubit: $\varepsilon_B \propto (g_I \mu_N \delta B \, \tau_g / \hbar)^2$, 270× smaller than Rb87
Magic wavelength: $\varepsilon_{\rm light\text{-}shift} \approx 0$, entire error term eliminated

Gate fidelity status (2025): Both platforms now demonstrate or approach the surface-code fault-tolerance threshold of 99.9% for two-qubit gates. Rb87 leads in parallel gate demonstrations (60 simultaneous pairs at 99.5%), while Yb171 shows slightly higher per-pair fidelity (99.72% with post-selection). The error budget for Yb171 is expected to shrink further as the platform matures, given its structural noise advantages. The race to consistent 99.9%+ without post-selection is the key 2025–2027 target for both.

07 Readout & SPAM

State preparation and measurement fidelity, the bookend errors of every circuit.

Rb87 Readout

Method: Fluorescence imaging, atoms in |F=2⟩ fluoresce on D2 cycling transition; |F=1⟩ atoms are dark (shelved or blown away)
SPAM/readout: high-fidelity fluorescence readout is well established in Rb arrays; exact numbers depend strongly on protocol and whether loss is included protocol-dependent
Adjacent alkali readout benchmark: Cs-133 Manetsch 2025 reports 99.98952(1)% imaging survival and >99.99% imaging fidelity; not an Rb number
Mid-circuit measurement: Demonstrated with atom transport (Bluvstein 2022), move ancilla atoms to readout zone, measure, return data atoms
Limitation: Reading destroys the measured atom unless shelving to F=1 is performed before fluorescence
Non-destructive readout: Active area of research; clock-shift measurements explored

Yb171 Readout

Method: Shelve one spin state to ³P₀ (dark), image the other with broad ¹P₁ transition at 399 nm, scattered photons distinguish |↑⟩ from |↓⟩
Repetitive readout: state discrimination fidelity 0.993(4) with state-averaged survival 0.994(3) in Huie et al. PRX Quantum 2023 peer-reviewed
Advantage: ³P₀ metastable state (20 s lifetime) enables truly non-destructive qubit storage during readout, image one logical basis without disturbing atoms in the other
Clock-state readout: State-selective fluorescence on the narrow ³P₁ transition (556 nm) enables high-SNR readout
Mid-circuit measurement: Structurally enabled by the ³P₀ shelving protocol, single-shot QND-like readout of subsets of qubits
Dual encoding advantage: Can store qubits in ³P₀ (protected from imaging light) while others are measured

Yb171 structural readout advantage: The ³P₀ metastable state (lifetime ~20 s) acts as a long-lived memory state that is far off resonance from the ¹S₀ imaging transition. By shelving qubits to ³P₀ before imaging, Yb171 enables site-selective, low-crosstalk readout that is naturally suited for quantum error correction cycles where only syndrome qubits are measured per round. This is significantly harder to achieve non-destructively in Rb87.

08 Array Size & Scalability

Current demonstrations and the path to fault-tolerant arrays of 10,000+ qubits.

Array Size Milestones (physical qubits/atoms)

Demonstrated physical qubit/atom counts. Rb87 milestones emphasize published gate and logical-processor results. Atom Computing's 1180-qubit Yb result is company-reported. The 6100-atom Manetsch benchmark is Cs-133, shown only as adjacent alkali context.

Rb87 Scaling

2021: QuEra Aquila, 256 atoms (analog mode)
2022: Bluvstein et al., 24-qubit toric code, surface code 19 qubits
2023: Evered et al., 60 qubits with 99.5% 2Q gates
2024: Bluvstein et al., 280 physical qubits, 48 logical qubits (color code)
2025 adjacent alkali: Manetsch et al. demonstrated 6100 coherent Cs-133 atoms, not Rb87; useful scaling context but not counted as an Rb milestone
Rearrangement: AOD-based atom sorting achieves >99% defect-free loading
Roadmap: 10,000+ qubit arrays, parallel Rydberg gates, photonic interconnects

Yb171 Scaling

2022: Jenkins et al., 10×10 array, 92.73% filling efficiency
2023: Atom Computing Phoenix, 1,180 qubits (1,225 sites)
2024: Atom Computing, 40 s quantum memory coherence announced
2025: Muniz et al. (Atom Computing collab.), PRX Quantum, 99.72% gate fidelity (post-sel)
Gate fidelity at scale: Not yet demonstrated at 1000+ qubits simultaneously
Rearrangement: Similar AOD-based sorting; magic-wavelength tweezer enables defect-free loading
Roadmap: Scale to 10,000+ qubits, demonstrate surface code, integrate photonic network

Current gap: Rb87 leads this comparison in peer-reviewed logical-processor maturity (48 logical qubits, 280 physical atoms) and parallel gate demonstrations (60 pairs at 99.5%). Yb171 leads in the best reported per-pair CZ number when post-selection is allowed (99.72(3)%; raw 99.40(3)%) and has structural erasure/shelving advantages. The largest published coherent neutral-atom array is Cs-133 (6100 atoms), not Rb87 or Yb171.

09 Erasure Conversion, Yb171's Structural Advantage

A new paradigm for quantum error correction that alkaline-earth atoms enable especially naturally.

In standard quantum error correction, errors are Pauli errors (X, Y, Z) that the code must detect and correct. Erasure errors, errors where we know which qubit failed, are much cheaper to correct: an [[n,k,d]] code corrects d-1 erasures but only ⌊(d-1)/2⌋ Pauli errors. If we can convert Pauli errors into known erasures, the overhead for fault tolerance drops dramatically.

Wu, Kolkowitz, Puri et al. (Nature Communications 2022) showed that for Yb171 and Sr87, ~98% of gate errors can be converted to erasures using the ³P₀ metastable state. When a qubit undergoes an error during a Rydberg gate, it often ends up in the ³P₀ state (detectable as "atom is in wrong sublevel") rather than a Pauli error. Detecting whether an atom is in ³P₀ vs ¹S₀ after each gate round , which is straightforward with site-selective imaging, heralds the error location without collapsing the logical qubit.

$$\text{Standard QEC:} \quad d_{\rm eff} = d \quad \rightarrow \quad \varepsilon_{\rm logical} \propto \left(\frac{p}{p_{\rm th}}\right)^{\lceil d/2\rceil}$$ $$\text{Erasure QEC:} \quad d_{\rm eff} = d-1 \quad \rightarrow \quad \varepsilon_{\rm logical} \propto \left(\frac{p}{p_{\rm th,\, erasure}}\right)^{d-1}$$ With 98% error-to-erasure conversion in the Wu et al. Yb/Sr proposal: simulated surface-code threshold increases from 0.937% to 4.15%.
Practical consequence: a distance-5 surface code with erasures ≈ distance-9 surface code with Pauli errors.

Yb171 Erasure Protocol

After each gate round, check if atom is in ¹S₀ (ground) or ³P₀ (metastable) via state-selective fluorescence
Atom in ³P₀ = detected error → erasure flag. Atom in ¹S₀ = passed gate correctly
The ³P₀ state is stable on the timescale of gate operations (τ ~ 20 s), no spontaneous decay during check
98% of dominant gate errors (Rydberg decay, Doppler) map to ³P₀ state
Only 2% are "dark" Pauli errors that the erasure protocol misses
Result: up to ~10× reduction in physical qubit overhead for fault-tolerant computation

Rb87, Limited Erasure

In Rb87, gate errors are mostly Rydberg decay or atomic motion errors, they deposit the atom back in the ground state without a detectable "flag"
No equivalent long-lived metastable state accessible from the qubit manifold to flag errors
Some error conversion possible via Rydberg state monitoring (atom loss detection), but less efficient than Yb171
Atom loss from the trap is detectable as an erasure (~99% efficiency for atom-loss erasure detection)
But Pauli errors from coherent control noise are not detectable → remain as standard errors
Net erasure fraction: Architecture-dependent and not a universal published Rb number; atom loss is heraldable, but most coherent Pauli errors are not naturally converted to erasures rough estimate

Long-term significance: Erasure conversion may be the most important structural advantage Yb171 has over Rb87 for large-scale fault-tolerant computation. A 10× reduction in qubit overhead means the difference between needing 10,000 physical qubits per logical qubit vs 1,000. Over millions of physical qubits, this could represent years of roadmap advantage.

10 Full Head-to-Head Comparison

Every major dimension, side by side. Advantage marked in bold color.

Category	Metric	⁸⁷Rb	¹⁷¹Yb	Edge
Atomic Physics	Electronic structure	Alkali, 1 valence e⁻, 5S₁/₂ ground	Alkaline-earth-like, 2 val. e⁻, ¹S₀ ground	Yb J=0 protection
	Nuclear spin	I = 3/2	I = 1/2	Yb simpler
	Ground-state J	J = 1/2 (electronic moment)	J = 0 (no electronic moment)	Yb
	Clock transition	None (microwave HF, 6.834 GHz)	578 nm (¹S₀→³P₀, <10 mHz)	Yb
Qubit Encoding	Qubit type	Hyperfine (F=1↔F=2, 6.834 GHz)	Nuclear spin (mI=±½, ~kHz scale)	Tie
	B-field sensitivity	~0 first-order; quadratic 575.15 Hz/G² × B² (clock states mF=0)	~2.6 Hz/mG (linear, nuclear magneton scale)	Yb at B > 2.3 G; Rb at lower fields
	Light-shift cancellation	Requires active calibration	¹S₀ nuclear-spin scalar shift nearly spin-independent; 759 nm is clock magic for ¹S₀–³P₀	Yb
Coherence	T₂ (best published)	No 6100-atom Rb T₂ claim; use Rb gate/logical benchmarks instead	>40 s storage/memory claimed by Atom Computing company	Yb (structural)
	T₂* (free precession)	Not quoted here without an Rb87 array source	Not fully characterized in tweezers	No direct comparison
	Primary dephasing	B-field noise, light shift	Magnetic noise plus residual tensor/vector/scattering effects; clock-state operations benefit from magic trapping	Yb
Gate Performance	Best 2Q fidelity	99.5% raw, parallel (Evered 2023)	99.72% with post-sel (Muniz 2025)	Yb (slightly)
	2Q fidelity raw	99.5% (no post-selection, 60 pairs)	99.40% (no post-selection)	Rb (current)
	Parallel gates	60 qubit pairs simultaneously at 99.5%	Not yet demonstrated at scale	Rb
	Gate speed	0.5–2 μs (2Q Rydberg)	1–5 μs (2Q Rydberg)	Rb (slightly)
Readout	SPAM/readout fidelity	High; protocol-dependent. Cs 6100-array imaging benchmark is not Rb	0.993(4) state discrimination; 0.994(3) survival (Huie 2023)	Protocol-dependent
Readout	Non-destructive readout	Limited, active research	³P₀ shelving enables partial QND readout	Yb
Scalability	Largest Rb/Yb array benchmark	280 physical atoms in logical-processor demo; 256-atom analog arrays	1,180 qubits (Atom Computing 2023)	Rb peer-reviewed logic; Yb company count
Scalability	Logical qubit demos	48 logical qubits (Bluvstein 2024)	Not yet demonstrated at scale	Rb
Error Correction	Erasure conversion fraction	Architecture-dependent; no universal Rb fraction	~98% (Wu et al. 2022)	Yb major
Error Correction	Qubit overhead reduction	Standard surface code overhead	Surface-code threshold 0.937% → 4.15% in Wu et al. simulation	Yb
Laser System	Complexity	Simpler: D2 at 780 nm (diode laser)	More complex: 399, 556, 578, 759 nm lasers	Rb
Laser System	Rydberg UV laser	480 nm (diode-pumped solid-state)	~302 nm UV (single photon from ³P₀); harder than Rb 480 nm diode	Rb
Maturity	Platform maturity	10+ years of tweezer work; most-cited demos	~5 years in tweezers; rapid progress 2022–25	Rb

11 Companies & Ecosystem

Who is building what, and how the two platforms map to commercial quantum computing ventures.

⁸⁷Rb Ecosystem

QuEra Computing ↗

Harvard/MIT spin-out (Lukin/Greiner groups). Aquila (256 atoms, 2021). Cloud access via AWS Braket. Dual analog+digital capability. Key partner for the Harvard group's gate fidelity demonstrations. Leading neutral-atom company by publication record and cloud reach.

Pasqal ↗

French spin-out from Browaeys/Lahaye group (IOGS, Paris). Fresnel processor (100 Rb atoms, 2D programmable). Hybrid analog-digital. Enterprise partnerships: EDF, BASF, Thales. Merged with Qu&Co (2022). Focus: optimization, combinatorial problems, HPC integration.

Infleqtion (ColdQuanta) ↗

Uses Rb (and Cs). Acquired SuperTech. Cloud quantum and quantum networking products. Boulder, Colorado-based.

Harvard/MIT Academic Groups

Lukin group (Harvard): All major Rb gate fidelity/logical qubit demonstrations (Evered 2023, Bluvstein 2024). Greiner group (Harvard): Quantum simulation. Vuletic group (MIT): Tweezer array methods. Most of the landmark papers originate here and license to QuEra.

Endres Group, Caltech

Manuel Endres group at Caltech. Published the 6100-atom Cs-133 array with T₂=12.6(1) s (Manetsch et al., Nature 2025), the largest coherent neutral-atom array demonstrated. Included here as adjacent alkali scaling context, not as an Rb87 result.

¹⁷¹Yb Ecosystem

Atom Computing ↗

Berkeley-based startup. Phoenix (1,225-site, 1,180-qubit Yb171 system, 2023; company announcement). Uses nuclear spin qubits for extended coherence. Atom Computing reports >40 s quantum memory in company materials. Focus: fault-tolerant digital QC with mid-circuit measurement and scalable arrays.

Kaufman Group (JILA/CU Boulder)

Adam Kaufman's group at JILA was the first to demonstrate nuclear-spin Yb171 qubits in optical tweezers (Jenkins et al. PRX Quantum 2022). Key academic partner for Atom Computing. Also works on Sr87 and other alkaline-earth species. The 99.72% gate fidelity result (Muniz et al. PRX Quantum 2025) is an Atom Computing collaboration with JILA and other groups.

NIST / Optical Clock Groups

Several NIST groups (Jun Ye, Andrew Ludlow) have deep Yb171 expertise from optical lattice clock work. The ultranarrow clock transition knowledge transfers directly to tweezer QC. Coherence >100 s demonstrated in optical lattice clocks informs Atom Computing's coherence roadmap.

Also: Sr87 (similar physics)

Fermi National Lab (Jonathan Simon group), Stanford (M. Endres group). Sr87 has I=9/2 (more complex spin than Yb171's I=1/2) but the same J=0 ground state protection and erasure conversion advantage. Nuclear clock groups exploring qubit applications.

12 Key Papers & References

Essential reading, all major experimental demonstrations, theory papers, and review articles.

⁸⁷Rb, Landmark Papers

High-fidelity parallel entangling gates on a neutral-atom quantum computer

Evered, Bluvstein, Kalinowski et al., Nature 622, 268–272 (2023)

99.5% CZ gate fidelity on 60 qubit pairs simultaneously using Rydberg blockade with dark-state-assisted scattering reduction. Crossed the NISQ/fault-tolerance boundary for neutral atoms.

nature.com ↗ · arXiv:2304.05420 ↗

Logical quantum processor based on reconfigurable atom arrays

Bluvstein, Evered, Geim et al., Nature 626, 58–65 (2024)

280 physical qubits, 48 logical qubits encoded in transversal color codes. Fault-tolerant operations with mid-circuit measurement and feed-forward. First demonstration of logical gate advantage in neutral atoms.

nature.com ↗ · arXiv:2312.03982 ↗

A large-scale quantum processor with fast, full-coherence operations

Manetsch, Nomura, Bataille et al., Nature 647, 60–67 (2025)

6100-atom Cs-133 array on a 12,000-site grid. T₂ = 12.6(1) s, 99.98952(1)% imaging survival, and >99.99% imaging fidelity. This is not an Rb87 result; it is listed as adjacent alkali/tweezer scaling context.

nature.com ↗ · arXiv:2403.12021 ↗

Coherent transport and entanglement generation using rearrangeable atom arrays

Bluvstein et al., Nature 604, 451–456 (2022)

Mid-circuit measurement and feed-forward via atom transport. Surface code on 19 qubits, toric code on 24 qubits. First programmable neutral-atom quantum error correction demonstration.

nature.com ↗ · arXiv:2112.03923 ↗

Parallel implementation of high-fidelity multiqubit gates with neutral atoms

Levine, Keesling, Semeghini et al., PRL 123, 170503 (2019)

First high-fidelity (97.4%) parallel Rydberg gates on Rb87 qubits. Introduced the basis for the Harvard group's Rydberg gate protocol.

PRL ↗ · arXiv:1908.06101 ↗

256-atom quantum simulator (QuEra Aquila)

Ebadi, Wang, Levine et al., Nature 595, 227–232 (2021)

256-qubit analog Rb87 quantum simulator studying frustrated magnetism and quantum phase transitions. First 200+ qubit neutral-atom demonstration. Basis for QuEra Aquila product.

nature.com ↗

¹⁷¹Yb, Landmark Papers

Ytterbium nuclear-spin qubits in an optical tweezer array

Jenkins, Lis, Senoo, McGrew, Kaufman, Phys. Rev. X 12, 021027 (2022)

First demonstration of Yb171 nuclear spin qubits in optical tweezers. 10×10 array, 92.73% filling, sub-100 ns single-qubit manipulation. Founded the Atom Computing roadmap. Kaufman group, JILA/CU Boulder.

PRX ↗ · arXiv:2112.06732 ↗

High-fidelity universal gates for ¹⁷¹Yb nuclear-spin qubits

Muniz, Stone, Stack et al. (Atom Computing collaboration, 53 authors), PRX Quantum 6, 020334 (2025)

99.72(3)% two-qubit CZ gate fidelity with post-selection, 99.40(3)% raw. Universal gate set with individually controlled and parallel single- and two-qubit gates. Demonstrates >200 CZ gates on atom pairs. Establishes Yb171 ground-state nuclear spin qubit as a top-tier gate platform.

PRX Quantum ↗ · arXiv:2411.11708 ↗

Rydberg excitation spectroscopy of ¹⁷¹Yb and improved Rydberg gate fidelity

Ma, Peper, Li, Knapp, Liu, Peng, Zhang, Horvath, Burgers, Thompson (Princeton), arXiv:2406.01482 (2024)

CZ fidelity 99.4(1)% using F=1/2 Rydberg series (3.3× better than F=3/2). MQDT spectroscopy of ¹⁷¹Yb Rydberg states. Single-photon UV (~302 nm) excitation from ³P₀ to Rydberg nP levels. Thompson group, Princeton.

arXiv:2406.01482 ↗

Erasure conversion for fault-tolerant quantum computing with alkaline-earth atoms

Wu, Kolkowitz, Puri, Thompson, Nature Commun. 13, 4657 (2022)

Theoretical framework showing that 98% of gate errors in Yb171/Sr87 can be converted to detectable erasures using the ³P₀ metastable state. ~10× reduction in fault-tolerance overhead. Highly cited theory paper.

Nature Commun. ↗

Atom Computing Phoenix: 1225-site ¹⁷¹Yb nuclear spin array

Atom Computing, Technical whitepaper + announcement (2023)

Industry announcement of Phoenix: 1,225 tweezer sites, 1,180 qubits loaded. Nuclear spin qubit. Atom Computing also reports >40 s coherence/memory in company materials; treat these scale/memory numbers as company-reported until fully peer-reviewed.

atom-computing.com ↗

Two-qubit encoding with ¹⁷¹Yb: clock + nuclear spin dual encoding

Multiple authors, npj Quantum Information (arXiv:2402.13134, 2024)

Proposes encoding logical qubits in both the nuclear spin (fast gates) and clock-state (³P₀, long coherence) of Yb171. Dual-encoding enables gate operations followed by coherent storage, relevant for fault-tolerant architectures.

arXiv:2402.13134 ↗

Reviews & Theory

Tweezer arrays of individual atoms for quantum applications

Kaufman & Ni, Nature Physics 17, 1324–1333 (2021)

Comprehensive review of optical tweezer approaches to neutral-atom quantum computing. Covers both alkali and alkaline-earth platforms. Excellent entry point for the field.