Robust Higher-Order Hamiltonian Engineering for

Quantum Sensing with Strongly Interacting Systems

Hengyun Zhou1 , Leigh S. Martin1 , Matthew Tyler1 , Oksana

Makarova1,2 , Nathaniel Leitao1 , Hongkun Park1,3 , Mikhail D. Lukin1
1
Department of Physics, Harvard University, Cambridge, Massachusetts 02138, USA
2
School of Engineering and Applied Sciences, Harvard University, Cambridge, Massachusetts 02138, USA
3
Department of Chemistry and Chemical Biology, Harvard University, Cambridge, Massachusetts 02138, USA
Dynamical decoupling techniques constitute an integral part of many quantum sensing platforms, often lead-
ing to orders-of-magnitude improvements in coherence time and sensitivity. Most AC sensing sequences involve
a periodic echo-like structure, in which the target signal is synchronized with the echo period. We show that for
strongly interacting systems, this construction leads to a fundamental sensitivity limit associated with imperfect
arXiv:2303.07363v1 [quant-ph] 13 Mar 2023

interaction decoupling. We present a simple physical picture demonstrating the origin of this limitation, and
further formalize these considerations in terms of concise higher-order decoupling rules. We then show how
these limitations can be surpassed by identifying a novel sequence building block, in which the signal period
matches twice the echo period. Using these decoupling rules and the resulting sequence building block, we
experimentally demonstrate significant improvements in dynamical decoupling timescales and magnetic field
sensitivity, opening the door for new applications in quantum sensing and quantum many-body physics.

Introduction — Quantum sensing utilizes quantum particles (a) DROID-60: “dipoles” add
to probe the properties of the surrounding environment [1]. FX F
X
...
Recent advances in quantum sensing technologies have led to
FY F ...
a host of new applications, including probes of magnetism in Y

condensed matter systems [2–6], nuclear magnetic resonance FZ F

Z
...
spectroscopy on the nanoscale [7, 8], as well as in-vivo tem-
perature sensing [9–12].
Key to unlocking new sensing applications is improvements (b) DIRAC: “dipoles” cancel
in metrological sensitivity. A common way to achieve this is FX F
X
...
to utilize dynamical decoupling sequences [13–16], such as
FY F
Y
...
XY-8 for non-interacting spins [17] or DROID-60 for inter-
acting spin systems [18]. Such sequences typically consist FZ F
Z
...
of a train of spin echo pulses synchronized to the target AC
field, as illustrated in Fig. 1, which help to isolate the spin
system from environmental disorder and certain spin-spin in-
teractions, while maintaining sensitivity to the target signal. FIG. 1. Limits of quantum sensing based on decoupling. Illustra-
tion of the “dipole” decoupling rule and sensing structure for conven-
While these techniques are already being actively used across
tional sensing pulse sequences (DROID-60, a) and our new sensing
a range of different experimental platforms, their ultimate per- pulse sequence (DIRAC2, b). F⃗ represents the orientation of the
formance and impact on quantum sensitivity limits are not yet interaction-picture S̃ z spin operator, where a yellow (green) block in
fully understood. row µ = x, y, z signifies S̃ z = +S µ (−S µ ) at the given time point. For
In this Letter, we identify a fundamental limitation asso- XY-8 and DROID-60, maintaining sensitivity to the periodic signal
ciated with the interplay between interaction decoupling and requires dipoles pointing in the same direction (red arrows), such that
sensing in existing pulse sequences, and propose and exper- dipole rules for decoupling cannot be satisfied; this is circumvented
by DIRAC, which allows dipole cancellations while maintaining AC
imentally demonstrate a class of new pulse sequences that field sensitivity.
overcomes this limitation. More specifically, we show that
the full synchronization of the sensing signal with a spin
echo building block inevitably contradicts the cancellation of
higher-order terms in the effective Hamiltonian, which leads
to a fundamental limit on the performance of all existing pulse pulse sequences for not only quantum sensing but also dynam-
sequences. To circumvent this limitation, we develop a differ- ical decoupling and Hamiltonian engineering, and we find sig-
ent paradigm for sensing, in which the target sensing signal nificantly improved performance compared to the best known
period is synchronized with twice the spin echo period, allow- pulse sequences in the disorder-dominated regime of interact-
ing one to realize superior sensitivity in experiments. This is ing spins [18].
achieved by developing concise decoupling rules for higher-
order contributions to the effective Hamiltonian (see Fig. 1 Limits on AC quantum sensing based on conventional de-
for an example), and using them to efficiently screen through coupling — The key idea of this work can be understood by
large design spaces of pulse sequences. This results in better considering a disordered, interacting many-body spin system,
 2

with the Hamiltonian given by The zeroth-order contribution Eq. (4) naturally suggests a spin
echo structure, since disorder is cancelled when ∑k Fµ,k = 0
H = Hsignal (t) + Hdis + Hdrive (t) + Hint for each µ ∈ {x, y, z}, such that the positive and negative dis-
= B(t) ∑ Siz + ∑ hi Siz + ∑ [Ωx (t)Six + Ωy (t)Siy ] order contributions along each axis balance each other out.
i i i This condition can be interpreted by associating, for each axis
+ ∑ [Jij Si Sj
I z z
+ Jij (Si Sj + Siy Sjy
H x x
+ Siz Sjz )] . (1) µ, a positive charge to a yellow block (+1) and a negative
ij charge to a green block (-1). To cancel disorder, the sum of
This general form encompasses many physical systems, charges along each axis must be 0. Generalizing this analy-
including dipole-dipole interactions, Rydberg atoms, and sis to higher-order terms (see Ref. [22] for more details), we
superexchange-interacting spins. Here, Six,y,z are spin- 21 op- focus on one particular first-order term in Eq. (5), involving
erators, B(t) is the time-dependent sensing target field, hi is a commutator between on-site disorder and Heisenberg inter-
the on-site disorder strength for spin i, Ωx (t), Ωy (t) are time- actions. Under pulse transformations, the disorder term trans-
dependent global Rabi drive strengths for the pulse sequence, forms linearly with Fµ,k , while the Heisenberg term is invari-
I
Jij H
, Jij are the Ising and Heisenberg interaction strengths ant. Consequently, assuming ideal pulses, we find
between spins i, j. The task of quantum sensing with such (1) −i T
Hdis,Heis = ∑ (tk − )τk Fµ,k ∑ [hi Si , Jjl S⃗j ⋅ S⃗l ],
µ H
an interacting quantum many-body system involves the dual 2T µ,k 2 i,j,l
challenge of (6)
1. Decoupling strong disorder and interactions as much as where tk and τk are the center time and duration of the k-th
possible, free evolution period, respectively. Focusing on the frame-
2. Maintaining maximal sensitivity to the target sensing dependent coefficients, we find that the contribution is
field under these constraints. (1) T
Hdis,Heis,µ ∝ ∑ (tk − )τk Fµ,k . (7)
In the interaction picture with respect to the applied drive k 2
pulses {Pk }, due to the secular nature of the Hamiltonian [19], Eq. (7) describes a sum of dipole moments, since it is the
the Hamiltonian can be expressed purely in terms of a polyno- charge τk Fµ,k multiplied by the position tk − T2 . Thus, we con-
mial of the transformed S z operator (the “frame”) at a given clude that cancelling the first order cross-term between disor-
time, i.e. der and Heisenberg interactions requires the sum of dipoles
along each axis to cancel.
H̃(t) = H̃(S̃ z (t)), (2)
In light of this interpretation, we can re-examine the pulse
S̃ (t) = (Pk−1 ⋯P1 ) S (Pk−1⋯ P1 ) = ∑ Fµ,k S ,
z † z µ
(3) sequences previously used for AC field sensing. In Fig. 1(a),
µ
we show the recently-developed sequence for interacting spin
where the time t is in between the (k − 1)-th and k-th pulse. systems [18], DROID-60, and how its frames and target AC
For pulse sequences composed of π/2 and π pulses, we de- field are synchronized. As one can see, the spins flip with the
note these “toggling frame” transformations [20] pictorially same periodicity as the external magnetic field, and the frames
in Fig. 1, where a yellow/green block in row µ = x, y, z, col- along each axis are always paired up to echo out disorder ef-
umn k signifies Fµ,k = +1 or −1 respectively. fects as rapidly as possible. This immediately implies that for
Utilizing this representation, we now show that the conven- a given axis, the dipoles are always oriented in the same direc-
tional way of synchronizing the period of the sensing field tion in order for the phase accumulation to coherently add, see
with the spin echo period necessarily leads to residual first- the red arrows in Fig. 1(a). Thus, with these pulse sequence
order terms in the effective Hamiltonian that, in turn, limit structures, maintaining AC field sensitivity always comes at
the sensing performance. These residual terms can be visu- the expense of introducing first order imperfections, which
alized by the “red dipoles” in Fig. 1(a) for the existing se- will directly affect the coherence time and subsequently the
quence DROID-60, which point from a green block (-1) to a sensitivity of the sequence.
yellow block (+1), and by necessity do not cancel. Such terms, Systematic Higher-Order Sequence Design — To overcome
present in most currently known AC sensing pulse sequences, this conflict, we start by systematically incorporating higher-
result in a fundamental limitation for the sensing protocol. order decoupling conditions (see Ref. [22] for a full deriva-
To understand how these terms arise, we make use of the tion) into sequence design, resulting in dynamical decoupling
Magnus expansion [21], which expresses the effective Hamil- sequences with much better coherence properties, but are not
tonian over one Floquet period T of the pulse sequence as a yet compatible with sensing. We will then present a novel
series summation, with the leading order terms being sequence building block (Fig. 1(b)) that respects the sensing
constraints, which allows us to obtain maximal sensitivity to
1 T
H (0) =∫ H̃(t1 )dt1 , (4) the target sensing field while retaining higher-order decou-
T 0 pling performance.
(1) −i T t1
We focus on the regime where the on-site disorder is dom-
H = ∫ dt1 ∫ dt2 [H̃(t1 ), H̃(t2 )]. (5)
2T 0 0 inant over spin-spin interactions, as is typically the case for
 3

electronic spin ensembles [18, 23, 24]. We numerically sim-

0.2
ulate the performance of dynamical decoupling pulse se- (a) Zeroth order
quences designed with different numbers of decoupling rules 0.18 Higher order
Symmetrized higher order
imposed, using parameters drawn from the experimental sys- 0.16
tem in Ref. [18]. We simulate the decay of a polarized ini- 0.14 1

Probability in Bin
tial state along x̂, ŷ or ẑ after different numbers of repeti- DROID-60
0.12 DROID-R2D2

Polarization
tions of the full pulse sequence. The fitted characteristic decay
0.1
timescales after subtracting out any long-time plateaus (for ex- 0.5
ample, due to residual disorder pinning) are histogrammed in 0.08
Fig. 2(a) for different sequence design methods, highlighting 0.06
the progressive improvement in performance as higher-order 0
0.04 0 2 4 6
terms are included. Time (ns)
0.02 10 5
We start by randomly generating pulse sequences with 24
free evolution frames, where all zeroth-order robust Hamil- 0
0 2 4 6 8 10
tonian engineering rules (see Ref. [19]) have been included, Decay Time (ns) 10 4
but no higher-order rules, as illustrated by the blue bars in (b)
10 5

Inverse Sensitivity (a.u.)

Fig. 2(a), resulting in a typical decay timescale on the order
of 5 µs. Next, we enforce that various higher-order terms are XY-8
4
zero (see Ref. [22]), with the most crucial ones being first- 10 DROID-60
DIRAC2
order cross-terms between disorder and Heisenberg interac-
tions (the “dipole rule” described above), as well as second- 10 3
order terms originating purely from disorder. This signifi-
cantly reduces the sequence search space, allowing us to ex-
10 2
haustively search through pulse sequences up to length 24. 0 0.5 1 1.5 2 2.5
This extends the decay time out to the orange bars. Cru- Time/Tdecay
cially, it removes the sequences that had relatively short decay
times, revealing the longer lived ones. Finally, we apply a fur- FIG. 2. Simulation evaluation and pulse sequence search. (a) Nor-
ther layer of symmetrization to the pulse sequence, where the malized histogram of coherence time, averaged over x̂, ŷ, ẑ initial
frame ordering is reversed and sign flipped in the second half states (averaging performed over decay rates), for different sets of de-
coupling rules under consideration. Simulations are performed with
of the sequence. This further improves the decay times, as
exact diagonalization of 6 spins. As more rules are sequentially in-
seen in the yellow distribution of Fig. 2(a), with a long tail ex- cluded, the distribution shifts further to the right, eliminating poorly
tending to the right. The longer timescales also imply that cer- performing pulse sequences. Inset: Comparison of simulated decou-
tain higher-order imperfections that may be dominant when pling performance of the existing sequence DROID-60 and a new
performing general Hamiltonian engineering are also system- sequence DROID-R2D2 that incorporates higher-order terms. (b)
atically removed. Simulated sensitivities for different pulse sequences, where our new
In the inset of Fig. 2(a), we show a direct comparison be- sequence DIRAC2 consistently outperforms all existing sequences.
The x-axis is normalized by the decay time of the sequence for com-
tween the previous best sequence in this parameter regime, parison.
DROID-60 [18], designed with zeroth-order rules and the
same symmetrization methods, and a new sequence DROID-
R2D2 (Disorder RObust Interaction Decoupling - Robust To
Disorder 2nd order) that accounts for higher-order rules. As of “dipole-rule” violations in Eq. (7).
one can see, the best symmetrized higher-order pulse se- To address this challenge, we devise a new strategy for the
quences show almost an order of magnitude improvement in design of AC sensing pulse sequences that overcomes this
decay time compared to prior sequences that only include conflict between sensing and decoupling. Instead of requiring
zeroth-order terms (for the simulation parameters mentioned the frame flips to be commensurate with the AC signal, the
above, DROID-R2D2 has 1/e decay time 390 µs, compared key idea is to design the frame flips to be at twice the rate of
to 64 µs for DROID-60). This highlights the power of system- the AC signal. By moving between frames on different axes,
atically including higher-order rules in the sequence design. it is possible to continue to coherently accumulate phase for
Surpassing the AC Sensing Limit with Higher-Order Se- interaction-decoupling AC sensing sequences in this case.
quence Design — Although incorporating higher-order de- Fig. 1(b) illustrates a representative sequence designed in
coupling conditions improves coherence times for dynamical this way, which we name DIRAC2 (DIsorder Robust AC sens-
decoupling, it does not fully overcome the conflict between ing of period 2). In this sequence, the yellow and green blocks
sensing and interaction decoupling. This is because as de- for the frame matrix are lined up with positive and negative
scribed above, when the spins flip with the same periodic- values of the target AC signal, respectively, indicating coher-
ity as the magnetic field, maintaining AC field sensitivity al- ent phase addition and maximal sensitivity. In addition, the
ways results in residual first-order imperfections in the form dipoles along each axis cancel each other out, indicating that
 4

first-order disorder-Heisenberg interaction cross-terms men- (a) 1

tioned in Eq. (7) are fully suppressed. Moreover, the faster DROID-60
DROID-R2D2

Polarization
flipping rate of the frames relative to the AC signal has the
added benefit that for the same target signal, this pulse se- 0.5
quence is more effective in decoupling time-varying noise,
which can lead to further performance improvements. Finally,
this sequence also incorporates a number of other higher-order 0
considerations and symmetrizations described in the previous 0 0.5 1 1.5 2 2.5 3
Time (ns) 4
section to further boost performance. Indeed, in Fig. 2(b), we 10
(b) 0.02
simulate the performance of different decoupling sequences

Inverse Sensitivity (Hz1/2 /nT)

with the same parameters as the preceding section, where we
find that DIRAC2 can outperform the best known decoupling 0.015
sequences for interacting spin ensembles (DROID-60) by an
order of magnitude.
0.01
Experimental Performance of Higher-Order Sequences —
To verify the performance of our new methods, we experi-
mentally implement these decoupling and sensing sequences 0.005 XY-8
DROID-60
in a high-density ensemble of nitrogen-vacancy (NV) centers DIRAC2
in diamond, see Ref. [18, 25] for more details of the exper- 0
imental system. First, we compare the performance of our 0 2000 4000 6000 8000 10000 12000 14000
Total Phase Accumulation Time (ns)
new dynamical decoupling sequence DROID-R2D2 against
the previous best sequence DROID-60 in Fig. 3(a), where we
FIG. 3. Experimental improvements for dynamical decoupling
prepare an x̂ initial state, decouple with either the DROID-60 and sensing with higher-order pulse sequence design. (a) Exper-
or DROID-R2D2 sequence for a period of time, and measure imental coherence decay curves under dynamical decoupling with
the final polarization along the x̂ axis. The total measurement different sequences, where the new sequence DROID-R2D2 signif-
window is held constant to normalize out charge dynamics icantly outperforms the existing sequence DROID-60. (b) Exper-
and T1 decay effects, and focus on decoupling performance. imental AC magnetic field sensitivities, where our new sequence
The higher-order sequence DROID-R2D2 that we design here DIRAC2 is significantly better than DROID-60 and XY-8, improv-
considerably outperforms the best known sequence DROID- ing quantum sensing with strongly-interacting spin ensembles.
60 (15 µs vs. 9 µs), although technical imperfections still limit
the achievable coherence times to be shorter than simulations
for both sequences.
Having shown that higher-order sequence design improves higher-order Hamiltonian engineering method to overcome
dynamical decoupling, let us now compare the performance of these limitations. We implemented these sequences experi-
different sensing sequences. For the comparison, we choose a mentally, resulting in significant improvements in AC mag-
target signal frequency of (2π) × 7 MHz, a π pulse time of 10 netic field sensitivities using these techniques.
ns, and a free evolution time τ = 61 ns for XY-8, DROID-60, These gains in sensitivity can be immediately translated to
τ = 25 ns for DIRAC2, in order to synchronize the sequence nanoscale NMR experiments, as recently demonstrated with
and sensing signal. In Fig. 3(b), we show the inverse sen- DROID-60 in Ref. [26]. In addition to extending coherence
sitivity (the higher the better) for each of the sequences, as times for sensing sequences, these techniques are also impor-
a function of the total phase accumulation time. The sensi- tant in extending coherence times for dynamical decoupling
tivities under optimal measurement conditions in this sample, and removing systematic artifacts when engineering desired
i.e. √the highest point on each curve, were ηXY-8 = √ 151 ± 2 target many-body Hamiltonians [22, 27].
nT/ Hz for the XY-8 sequence, ηDROID = 90 ± 2 nT/ √ Hz for The analytical insights developed here provide simple ge-
the DROID-60 sequence, and ηDIRAC = 58 ± 4 nT/ Hz for the ometric intuitions for various higher-order decoupling rules,
DIRAC2 sequence. Thus, while DROID-60 outperforms the which significantly simplify the design and optimization pro-
non-interacting sensing sequence XY-8, DIRAC2 achieves yet cess. We expect that our techniques can be readily extended
another significant improvement over DROID-60, achieving to even higher-order contributions. Moreover, it will also
close to a factor of 3 improvement in sensitivity over the con- be interesting to explore the application of these ideas to
ventional sensing sequence XY-8. This is because the combi- spin systems in other parameter regimes, such as Rydberg
nation of suppressed higher-order terms and faster decoupling atoms [28, 29], nuclear magnetic resonance [30, 31], and
results in longer coherence times, while maintaining a similar trapped ions [32, 33], or higher spin systems [34, 35]. Fi-
rotation rate under the target field. nally, our results provide an important tool for the reliable en-
Discussions and Outlook — In this work, we identified gineering of many-body Hamiltonians, free of higher-order ar-
a key limitation of existing pulse sequences for dynamical- tifacts, opening the door to exploration of exotic driven phases
decoupling-based quantum sensing, and devised a novel of matter and creation of entangled quantum states for quan-
 5

tum metrology [36–39]. [18] H. Zhou, J. Choi, S. Choi, R. Landig, A. M. Douglas, J. Isoya,
Acknowledgements — We thank J. Choi, H. Gao, F. Jelezko, S. Onoda, H. Sumiya, P. Cappellaro, H. S. Knowles,
N. Maskara for helpful discussions. This work was supported H. Park, and M. D. Lukin, Physical Review X 10, 031003
in part by CUA, ARO MURI, DARPA DRINQS, Moore Foun- (2020).
[19] J. Choi, H. Zhou, H. S. Knowles, R. Landig, S. Choi, and M. D.
dation GBMF-4306, NSF PHY-1506284. Lukin, Physical Review X 10, 031002 (2020).
[20] P. Mansfield, Journal of Physics C: Solid State Physics 4, 1444
(1971).
[21] W. Magnus, Communications on Pure and Applied Mathemat-
ics 7, 649 (1954).
[1] C. L. Degen, F. Reinhard, and P. Cappellaro, Reviews of Mod- [22] M. Tyler et al., in preparation.
ern Physics 89, 035002 (2017). [23] C. Zu, F. Machado, B. Ye, S. Choi, B. Kobrin, T. Mittiga,
[2] F. Casola, T. van der Sar, and A. Yacoby, Nature Reviews Ma- S. Hsieh, P. Bhattacharyya, M. Markham, D. Twitchen, A. Jar-
terials 3, 17088 (2018). mola, D. Budker, C. R. Laumann, J. E. Moore, and N. Y. Yao,
[3] L. Rondin, J.-P. Tetienne, T. Hingant, J.-F. Roch, P. Maletinsky, Nature 597, 45 (2021).
and V. Jacques, Reports on Progress in Physics 77, 056503 [24] B. Merkel, P. Cova Fariña, and A. Reiserer, Physical Review
(2014). Letters 127, 030501 (2021).
[4] C. Du, T. van der Sar, T. X. Zhou, P. Upadhyaya, F. Ca- [25] G. Kucsko, S. Choi, J. Choi, P. C. Maurer, H. Zhou, R. Landig,
sola, H. Zhang, M. C. Onbasli, C. A. Ross, R. L. Walsworth, H. Sumiya, S. Onoda, J. Isoya, F. Jelezko, E. Demler, N. Y. Yao,
Y. Tserkovnyak, and A. Yacoby, Science 357, 195 (2017). and M. D. Lukin, Physical Review Letters 121, 023601 (2018).
[5] I. Gross, W. Akhtar, V. Garcia, L. J. Martı́nez, S. Chouaieb, [26] N. Arunkumar, K. S. Olsson, J. T. Oon, C. Hart, D. B.
K. Garcia, C. Carrétéro, A. Barthélémy, P. Appel, P. Maletinsky, Bucher, D. Glenn, M. D. Lukin, H. Park, D. Ham, and
J.-V. Kim, J. Y. Chauleau, N. Jaouen, M. Viret, M. Bibes, R. L. Walsworth, arXiv preprint arXiv:2203.12501 (2022),
S. Fusil, and V. Jacques, Nature 2017 549:7671 549, 252 10.48550/arxiv.2203.12501.
(2017). [27] L. S. Martin, H. Zhou, N. T. Leitao, N. Maskara, O. Makarova,
[6] M. J. Ku, T. X. Zhou, Q. Li, Y. J. Shin, J. K. Shi, C. Burch, L. E. H. Gao, Q.-Z. Zhu, M. Park, M. Tyler, H. Park, S. Choi,
Anderson, A. T. Pierce, Y. Xie, A. Hamo, U. Vool, H. Zhang, and M. D. Lukin, arXiv preprint arXiv:2209.09297 (2022),
F. Casola, T. Taniguchi, K. Watanabe, M. M. Fogler, P. Kim, 10.48550/arxiv.2209.09297.
A. Yacoby, and R. L. Walsworth, Nature 583, 537 (2020). [28] D. Bluvstein, A. Omran, H. Levine, A. Keesling, G. Semeghini,
[7] N. Aslam, M. Pfender, P. Neumann, R. Reuter, A. Zappe, F. F. S. Ebadi, T. T. Wang, A. A. Michailidis, N. Maskara, W. W. Ho,
De Oliveira, A. Denisenko, H. Sumiya, S. Onoda, J. Isoya, and S. Choi, M. Serbyn, M. Greiner, V. Vuletić, and M. D. Lukin,
J. Wrachtrup, Science 357, 67 (2017). Science 371, 1355 (2021).
[8] D. R. Glenn, D. B. Bucher, J. Lee, M. D. Lukin, H. Park, and [29] S. Geier, N. Thaicharoen, C. Hainaut, T. Franz, A. Salzinger,
R. L. Walsworth, Nature 555, 351 (2018). A. Tebben, D. Grimshandl, G. Zürn, and M. Weidemüller, Sci-
[9] R. Schirhagl, K. Chang, M. Loretz, and C. L. Degen, Annual ence 374, 1149 (2021).
Review of Physical Chemistry 65, 83 (2014). [30] D. G. Cory, J. B. Miller, and A. N. Garroway, Journal of Mag-
[10] G. Kucsko, P. C. Maurer, N. Y. Yao, M. Kubo, H. J. Noh, P. K. netic Resonance (1969) 90, 205 (1990).
Lo, H. Park, and M. D. Lukin, Nature 500, 54 (2013). [31] P. Peng, X. Huang, C. Yin, L. Joseph, C. Ramanathan,
[11] J. Choi, H. Zhou, R. Landig, H. Y. Wu, X. Yu, S. E. von Stetina, and P. Cappellaro, arXiv preprint arXiv:2102.13161 (2021),
G. Kucsko, S. E. Mango, D. J. Needleman, A. D. Samuel, P. C. 10.48550/arxiv.2102.13161.
Maurer, H. Park, and M. D. Lukin, Proceedings of the Na- [32] J. Zhang, G. Pagano, P. W. Hess, A. Kyprianidis, P. Becker,
tional Academy of Sciences of the United States of America H. Kaplan, A. V. Gorshkov, Z. X. Gong, and C. Monroe, Nature
117, 14636 (2020). 551, 601 (2017).
[12] M. Fujiwara, S. Sun, A. Dohms, Y. Nishimura, K. Suto, [33] C. Kokail, C. Maier, R. van Bijnen, T. Brydges, M. K. Joshi,
Y. Takezawa, K. Oshimi, L. Zhao, N. Sadzak, Y. Umehara, P. Jurcevic, C. A. Muschik, P. Silvi, R. Blatt, C. F. Roos, and
Y. Teki, N. Komatsu, O. Benson, Y. Shikano, and E. Kage- P. Zoller, Nature 2019 569:7756 569, 355 (2019).
Nakadai, Science Advances 6, eaba9636 (2020). [34] H. Zhou et al., in preparation.
[13] D. Suter and G. A. Álvarez, Reviews of Modern Physics 88, [35] N. Leitao et al., in preparation.
041001 (2016). [36] E. Davis, G. Bentsen, and M. Schleier-Smith, Physical Review
[14] L. Viola, E. Knill, and S. Lloyd, Physical Review Letters 82, Letters 116, 053601 (2016).
2417 (1999). [37] O. Hosten, R. Krishnakumar, N. J. Engelsen, and M. A. Kase-
[15] G. de Lange, Z. H. Wang, D. Ristè, V. V. Dobrovitski, and vich, Science 352, 1552 (2016).
R. Hanson, Science 330, 60 (2010). [38] P. Cappellaro and M. D. Lukin, Physical Review A 80, 032311
[16] D. Farfurnik, A. Jarmola, L. M. Pham, Z. H. Wang, V. V. Do- (2009).
brovitski, R. L. Walsworth, D. Budker, and N. Bar-Gill, Physi- [39] G. Goldstein, P. Cappellaro, J. R. Maze, J. S. Hodges, L. Jiang,
cal Review B 92, 060301 (2015). A. S. Sørensen, and M. D. Lukin, Physical Review Letters 106,
[17] T. Gullion, D. B. Baker, and M. S. Conradi, Journal of Mag- 140502 (2011).
netic Resonance 89, 479 (1990).