Quantum magnetometry of transient signals with a time resolution of 1.

1 nanoseconds

K. Herb,1, ∗ L.A. Völker,1 J. M. Abendroth,1 N. Meinhardt,1 L. van Schie,1, 2 P. Gambardella,2, 3 and C. L. Degen1, 3, †
1
Department of Physics, ETH Zürich, Otto-Stern-Weg 1, 8093 Zürich, Switzerland.
2
Department of Materials, ETH Zürich, Hönggerbergring 64, 8093 Zürich, Switzerland.
3
Quantum Center, ETH Zürich, 8093 Zürich, Switzerland.
(Dated: November 11, 2024)
Quantum magnetometers based on spin defects in solids enable sensitive imaging of various mag-
netic phenomena, such as ferro- and antiferromagnetism, superconductivity, and current-induced
fields. Existing protocols primarily focus on static fields or narrow-band dynamical signals, and are
optimized for high sensitivity rather than fast time resolution. Here, we report detection of fast sig-
nal transients, providing a perspective for investigating the rich dynamics of magnetic systems. We
experimentally demonstrate our technique using a single nitrogen-vacancy (NV) center magnetome-
ter at room temperature, reaching a best-effort time resolution of 1.1 ns, an instantaneous bandwidth
arXiv:2411.05542v1 [quant-ph] 8 Nov 2024

of 0.9 GHz, and a time-of-flight precision better than 20 ps. The time resolution can be extended to
the picosecond range by use of on-chip waveguides. At these speeds, NV quantum magnetometers
will become competitive with time-resolved synchrotron X-ray techniques. Looking forward, adding
fast temporal resolution to the spatial imaging capability further promotes single-spin probes as
powerful research tools in spintronics, mesoscopic physics, and nanoscale device metrology.

INTRODUCTION for representative application scenarios, including time-

resolved imaging of magnetization reversals and domain
Two-level systems with long coherence times have wall propagation in magnetic nanostructures, as well as
found widespread application for sensitive detection of time-of-flight detection of magnetic pulses.
a variety of physical parameters, including magnetic [1–
3] and electric fields [4], temperature [5], pressure [6, 7],
and derived quantities like electrical currents [8, 9] and RESULTS
resistance [10]. While often applied to measure static
fields, quantum sensing can be extended to detect tran- Scheme for time-resolved quantum sensing
sient signals by designing suitable control schemes. No-
tably, coherent control sequences permit the recording of Our sensing scheme is based on the generic concept
noise spectra [11–13], lock-in detection of harmonic test of quantum phase measurements [26]. The basic idea,
signals [14, 15], and the acquisition of arbitrary wave- illustrated in Fig. 1, is to reconstruct an unknown tran-
form patterns [16–18]. Although demonstrations of wide sient magnetic signal B(t) by periodically sampling B(t)
bandwidths have been made [19–22], existing protocols using a spin probe. An example of such a transient sig-
are optimized mainly for high sensitivity rather than high nal is the magnetic stray field generated by a propagat-
temporal resolution. The ability to detect nanosecond or ing magnetic domain wall in proximity of the spin probe
even sub-nanosecond dynamics would significantly widen (Fig. 1a). To mimic such a measurement, we expose
the application range of quantum sensors, especially in the spin qubit to the influence of a transient field using
rapidly evolving fields like spintronics [23] or semicon- a short microwave control pulse, leading to a quantum
ductor metrology [24]. phase pickup ϕ ∝ B(t). This phase pick-up can be read
In this work, we demonstrate detection of transient out by a projective measurement. We then repeat the full
magnetic fields with a time resolution approaching one measurement cycle and use equivalent-time sampling to
nanosecond. Our experiment, carried out with a single reconstruct a trace of the magnetic field transient, equiv-
NV center in a diamond nanopillar, uses a pump-probe alent to pump-probe measurements.
scheme consisting of an electrical start trigger and a de- The quantum phase measurement is executed in three
layed spin manipulation and readout. Fast temporal res- steps (labeled by 1-2-3 in Fig. 1b). First, we prepare the
olution is enabled by combining a speed-optimized pulse spin in a known initial state, denoted by |0⟩. This step is
sequence [25] with efficient microwave pulse delivery. We not time critical and may be applied before triggering the
show that undesired side effects of short pulses, such as transient field. For the NV center, spin initialization is
pulse distortions, non-linear spin driving or spurious ex- typically achieved by optical pumping using a green laser
citation of non-resonant transitions can be deconvolved pulse [27]. Second, the quantum phase measurement is
by numerical simulation of spin dynamics in the labora- performed. A sequence of microwave pulses, precisely
tory frame. We measure transient magnetic test signals timed around t, temporarily generates a coherent super-
position of spin states that interacts with the field B(t).
Third, the resultant final spin state |ψ⟩ is read out via
∗ [email protected] a projective photoluminescence (PL) intensity measure-
† [email protected] ment. Like the initialization, the read out is not time
 2

critical and may be executed once the transient signal Trigger

has passed. Together and after many averaging cycles, B(t) Transient field
the PL readout yields a state probability p = | ⟨0|ψ⟩ |2 a
that contains an estimate of the desired quantum phase ON Detector
ϕ, and thus the field B(t), at time instance t (Fig. 1d). 1 OFF
t Laser,
Initialize 3 Readout Photodetector

In the limit of small signals (|ϕ| ≪ π/2, valid for our b

P1 P2
experiments), the state probability p(t), i.e., the measur- 2
able output signal, is proportional to the input transient Microwaves
τ
magnetic field B(t). The quantitative relation between
p(t) and B(t) can be then expressed by a simple convo-
lution, tmin Convolution
k(t)
Z ∞ c kernel

p(t) = k(t′ − t)γB(t′ )dt′ . (1)

t′ =−∞ ∆t p=1
p(t)
Here, γ is the gyromagnetic ratio of the spin probe and p=0 Output signal
k(t) is a convolution kernel (Fig. 1c) that accounts for d
the details and finite duration of the chosen microwave
pulse sequence. The kernel function is computed using a Figure 1. Measurement scheme for transient magnetic
spin density matrix simulation of the pulse sequence in field sensing. a, An electrical trigger initiates play-back of
the transient signal B(t). b, Quantum control sequence for
the laboratory frame (Methods).
probing the transient signal with the NV spin. Laser pulses
and a fast photo detector are used to initialize (1) and read out
To reach a high temporal resolution, the duration τ of (3) the spin state. A composite sequence of two-phase shifted
the microwave pulse sequence (Fig. 1b) must be as short microwave pulses (2, P1-P2), delayed by t with respect to the
as possible. At the same time, the sequence should maxi- start trigger, is used to sample B(t) at time t. τ is the se-
quence duration. c, Convolution kernel k(t) of the microwave
mize ϕ in order to reach adequate sensitivity. These con-
pulse sequence shown in b. tmin is the time resolution defined
flicting requirements necessitate a compromise in mea- by the full width at half maximum. d, Measured qubit output
surement speed and acceptable ratio of signal-to-noise signal, given by the state probability p(t) = | ⟨0|ψ⟩ |2 . p(t) is
(SNR). In theory, the trade-off between speed and sen- the convolution between B(t) and k(t). Each measurement
sitivity is captured by the quantum speed limit (QSL), cycle samples one instance of p(t) (dots). Many repetitions
imposing a bound on how fast a quantum system can are used to construct the full p(t) trace by equivalent-time
be transformed between orthogonal states [28]. In a re- sampling. ∆t is the step size.
lated theory paper [25], we show that for time-resolved
sensing, the QSL can be reached by use of two concate-
nated microwave pulses of equal amplitude and duration Experimental implementation
that are phase shifted by π/2 (P1 and P2 in Fig. 1b).
This sequence is equivalent to a Ramsey interferometry
measurement with zero time delay [29]. The convolution We experimentally demonstrate time-resolved quan-
kernel of the composite sequence, plotted in Fig. 1c and tum sensing of fast magnetic field transients using a sin-
derived analytically in [25], is given by gle NV center loated in a diamond nanopillar photonic
waveguide [30]. The NV center is a S = 1 spin system
( with three spin energy levels, mS = 0 and mS = ±1,
sin[Ω(τ /2 − |t|)] |t| < τ /2
k(t) = . (2) and two allowed spin-flip transitions (frequencies ω and
0 |t| > τ /2 ω ′ ), depicted in Fig. 2a. To form an effective two-level
system, we isolate the mS = 0 to mS = −1 transition
where Ω is the speed of spin rotations (Rabi frequency) (ω) by applying an axial bias field of B0 = 36 mT along
and α = Ωτ /2 the spin rotation angle. The associated the N-to-V symmetry axis. For the microwave driving
time resolution, defined by the full width at half maxi- fields accessible in our setup and the chosen bias field,
mum (FWHM) of the kernel function (Fig. 1c), is given the spin excitation is selective and the ω ′ transition can
by be neglected. However, for even stronger driving, the full
! S = 1 nature of the spin system including the ω ′ transi-
arcsin sin2 α tion and breakdown of the rotating-wave approximation
tmin = τ 1 − . (3)
α would have to be taken into account [25, 31].
Experimentally, we generate control pulses and mag-
Eq. (3) confirms our expectation that to achieve a high netic test waveforms on separate circuits and guide them
temporal resolution, fast spin rotations Ω ∝ τ −1 and to the NV center via a co-planar waveguide (CPW) an-
small rotation angles α are beneficial. tenna (Fig. 2b). The CPW is fabricated by patterning a
 3

b a tive losses are crucial to achieve high Rabi frequencies.

mS = +1
AWG 2 PA mS = –1
�� To maximize the Rabi frequency, the NV center pil-
Laser lar array being used (dashed white box in Fig. 2c) is
�
positioned within ∼ 25 µm of the center conductor.
mS = 0
Trigger The highest Bmw amplitudes realized in this work are
CPW
∼ 6.3 mT, corresponding
√ to Rabi frequencies of approxi-
Delay
mately Ω = γBmw / 2 ∼ 2π × 125 MHz.
AWG 1 PA DIP NV center Because control and test pulses have very short du-
rations, on the order of a few nanoseconds, it is cru-
cial to account for pulse distortions in the excitation
Photodetector
Oscilloscope
path [32, 33]. To correct for the finite rise and ringdown
c times of microwave control pulses, we record the pulse
NV diamond pillar array waveform on an ancillary oscilloscope connected to the
antenna output (Fig. 2b). Pulse imperfections can then
be accounted for by computing the actual kernel function
from the measured pulse shape (Methods) followed by a
numerical deconvolution of the signal p(t) (see below).
CPW center conductor Favorably, even if the actual waveform deviates signifi-
cantly from the desired waveform (Fig. 2d), the two ker-
nels remain very similar (Fig. 2e). The dominant effect
is a slight broadening of the kernel. For the test signal,
d e we apply pre-distortion to the digital input waveform un-
1.0 til the oscilloscope record matches the desired waveform
kernel k(t) (a.u.)

P1 P2
amplitude (a.u.)

1.0 pattern.

0.0 Detection of transient signals

−1.0 0.0 As our first example, we consider the transient mag-

−4 −2 0 2 4 −8 −4 0 4 8 netic field profile associated with magnetization switch-
time (ns) time (ns) ing events in magnetic nanostructures. This scenario is
representative of applications in non-volatile memories
Figure 2. Experimental circuit for NV control and and logic units in spintronic devices [23], and of general
test signal generation. a, NV spin energy levels and al- interest for studying magnetic phase transitions [35]. De-
lowed transitions ω and ω ′ . b, Block diagram of the control terministic switching of magnetization can, for example,
circuit. Microwave pulses and test waveform are generated on be triggered using current or light pulses [34]. The mag-
separate arbitrary waveform generators (AWG1 and AWG2, netization reversals typically occur on a time scale of a
respectively), amplified (PA), combined (DIP), and delivered few nanoseconds [34, 36, 37], and time-resolved detec-
to a common coplanar waveguide (CPW) antenna. A picosec- tion relies on Hall-effect [36] and anisotropic magnetore-
ond digital delay generator is used to fine-adjust the relative sistance measurements [38], the magneto-optical Kerr ef-
timing. See Methods for further details. c, Optical micro-
fect [39, 40], or synchrotron X-ray imaging [34]. More-
graph of the CPW center conductor (gold) and the diamond
nanopillar arrays (black pattern). The diamond nanopillar ar-
over, the switching is highly repeatable [41], which is a
ray is vertically offset from the CPW by ∼ 25 µm. The inset prerequisite for most time-resolved measurement tech-
shows a PL intensity map of a selected pillar array. High- niques (including ours) and required for memory appli-
intensity pillars (one circled) typically contain a single NV cations. Here, sensitive quantum probes with nanoscale
center. Scale bars are 50 µm and 10 µm (inset), respectively. spatial resolution could add important spatio-temporal
d, Waveform of a τ = 4 ns microwave pulse pair. The gray information on the switching dynamics.
oscillation is the design waveform and the blue oscillation is Fig. 3a displays the synthetic input waveform and mea-
the actual waveform measured on the oscilloscope. e, Corre- sured magnetic field for two simulated magnetization re-
sponding sensing kernels k(t) computed using a spin dynamics versals. Clearly, the output signal recorded by the NV
simulation (Methods).
probe closely follows the applied input waveform. Key
details are well resolved, including the transient peak
during the ∼ 3 ns-long reversal (gray shading) and the
gold center conductor onto a quartz coverslip that is nar- ringing before and after the reversal pulses. Despite the
rowed down to 20 µm in the core section to concentrate high time resolution and correspondingly short phase ac-
the microwave current and achieve high microwave field cumulation time, the SNR remains high, with a baseline
amplitudes (Fig. 2c). We use a broad-band (∼ 8 GHz) noise of approximately 30 µTrms . Overall, the result of
design to allow for short pulse lengths. Further, we Fig. 3a demonstrates the feasibility for investigating the
note that precise impedance matching and low absorp- magnetization dynamics on nanosecond time scales.
 4

200 �T

0 20 40 60 80
time (ns)

b c
�
90°
z 60°
45°
vd

200 �T
tpeak
200 �T �
d 5.0
1.0
0.5
0.2
0.001

0 20 40 60 −15 −10 −5 0 5 10 15
time (ns) relative time (ns)

Figure 3. Experimental demonstration of transient magnetic field sensing. a, Scenario for detecting current-induced
magnetization reversals in a magnetic microdot [34]. Two reversals are simulated with the magnetization configurations (up
or down, indicated by an arrow) shown as insets. The gray curve is the input waveform and blue dots show the measured
sensor output signal p(t), averaged over ∼ 107 repetitions per point and converted to units of magnetic field. The p(t) curve
is low-pass filtered at τ3dB = 0.8 ns. Shaded vertical bars indicate the duration of the ∼3-ns-long switching events. The total
measurement time is ca. 1.5 h. b, Scenario for detecting the stray field spike generated near a passing magnetic domain wall
(inset). The gray curve is the input waveform and the blue dots the measured sensor output. No filtering is applied. tpeak ∼ 3 ns
is width of the stray field spike. c, Improved time resolution through shortening of the rotation angle α. Corresponding pulse
durations are τ = 4, 3, 2 ns. The gray curve shows the input waveform and the blue curves the experimental data. The α = 90◦
curve is the same as in b. d, Improved time resolution using signal deconvolution by inverse Wiener filtering (see text). λ is
the regularization parameter. The gray curve shows the input waveform and the blue curves the experimental data. For a-d,
Ω/2π = 125 MHz and α = 90◦ , unless noted otherwise.

As a second scenario, we consider the transient mag- tion speed [34, 45, 46]. Fig. 3b shows the measured NV
netic field spike produced by a propagating magnetic do- signal (dots) together with the simulated input wave-
main wall. This scenario is representative for magnetic form. The simulation assumes a domain wall veloc-
racetrack devices [42, 43], where domain wall injection ity of vd = 100 m/s and a distance to the spin probe
and movement is controlled by an electrical current [44]. of z = 100 nm (see inset), leading to a magnetic field
Time-resolved imaging of moving domain walls could, spike with typical magnitude ∼ 0.5 mT and a duration
for example, yield information about the dynamic dis- tpeak of a few nanoseconds (Methods). Again, the NV
tortion of the wall or local variations in the propaga- measurement clearly reproduces the main spike feature
 5

of the transient signal. However, contrary to Fig. 3a, a b

cross. corr. (a.u.)

tToF
c �t
some signal broadening is evident. This broadening is z

expected, since the microwave pulse duration (τ = 4 ns)

is longer than the magnetic field spike (tpeak ∼ 3 ns). In trigger 1.0

norm. signal (a.u.)

Figs. 3a,b, the microwave pulse duration τ ∼ π/Ω is de-
termined by the maximum Rabi frequency available with p(t) -2ns 0 2ns
0.5
our setup (Ω/2π ∼ 125 MHz). tToF
�t p(t)
0.0 p(t-500 ps)

p(t-�t)
−0.5

-8 -4 0 4 8
Time resolution and signal deconvolution relative time (ns)

Next, we discuss two approaches to further optimize Figure 4. Time-of-flight (ToF) detection. a, Concept
the time resolution for a given maximum Ω. Both ap- of ToF detection of two time-shifted waveforms, such as mag-
netic domain walls propagating with different velocities (in-
proaches rely on trading SNR for better temporal res-
set). b, Experimental ToF demonstration for a time shift of
olution. A first method is to reduce the spin rotation δt = 500 ps. c, Cross-correlated signals ⟨p(t)p(t − δt)⟩ for
angle α (set to α = 90◦ in Figs. 3a,b), equivalent to waveforms with nominal delays of 250, 500 and 750 ps. Mea-
a narrowing of the kernel function and hence smaller sured ToF values are determined by the positions of the peak
tmin [see Eq. (3)]. Fig. 3c shows signal transients with maxima (marked with triangles) and are 249(1) ps, 523(2) ps
α = 45 − 90◦ recorded at maximum Ω/(2π) = 125 MHz. and 729(2) ps.
Decreasing α leads to a small but noticeable improve-
ment in peak width and for α = 45◦ , the broadening is
essentially eliminated. However, an increase in the noise Time-of-flight detection
level is also evident. The time resolution for the short-
est pulse (τ = 2 ns and α = 45◦ ) is tmin = 1.1 ns, and As a further application scenario, we consider time-
the corresponding instantaneous frequency bandwidth is of-flight (ToF) detection of magnetic pulse events. In
ΩBW ≈ τ −1 ≈ 0.9 GHz [25]. These figures are an order this scenario, rather than mapping the waveform of the
of magnitude faster than previous targeted approaches at transient signal, we are interested in precisely timing the
fast waveform detection [21, 47], and two orders of mag- arrival of a magnetic field pulse with respect to a start
nitude faster than methods based on Walsh and Haar re- trigger. Such a capability would be useful, for exam-
construction [17, 18]. (Note that these latter experiments ple, to measure the propagation velocity and dispersion
were optimized for sensitivity, not time resolution.) of domain walls in magnetic race tracks. To demon-
Alternatively, the time resolution can be optimized in strate ToF detection, we generate two identical wave-
post-processing by numerical deconvolution of the ker- forms with a pre-determined delay δt (Fig. 4a). Fig. 4b
nel function k(t) using inverse filtering. This approach shows recordings of two such waveforms that are time-
has the added advantage of removing artifacts caused by shifted by δt = 500 ps. The two transients can be eas-
pulse distortions in the control circuit, because the true ily distinguished, despite a much slower rise time and
measured kernel function k(t) can be used as an input overall duration of the waveform. To quantify the ToF
(Figs. 2d,e). Further, deconvolution allows taking into precision, we compute the regularized cross-correlation
account strong spin driving effects (Methods). (via self-deconvolution, see Methods), shown in Fig. 4c
for four waveforms with δt = 0, 250, 500, and 750 ps. A
Fig. 3d shows the measured waveform (α = 90◦ ) to- least-squares fit yields a fit precision of better than 5 ps.
gether with a series of reconstructed waveforms, using The absolute timing error is approximately ±20 ps and
Wiener deconvolution (Methods). Here, λ is a unit-less limited by jitter in our triggering pulse. Both numbers
regularization parameter that controls noise suppression. could be improved to ≲ 1 ps, if desired, by using op-
For weak filtering (λ < 1), the sharp signal spike is accu- timized signal averaging and employing hardware with
rately resolved at the expense of overall increased noise. lower jitter.
By contrast, strong filtering (λ > 1) efficiently removes
noise but broadens the signal peak. Comparing Figs. 3c
and 3d, both optimization methods – shortening of the
rotation angle and inverse filtering – lead to comparable
improvements in peak resolution; however, the inverse
filtering approach is more powerful as it can account for
experimental imperfections and non-linear spin driving,
and is applied in post-processing. A quantitative discus-
sion of the trade-off between sensitivity and time resolu-
tion is given in the Methods section.
 6

OUTLOOK Finally, a potential issue of our technique will be per-

turbation of the system by microwave and optical pulses,
In summary, we experimentally demonstrate time- which could lead to magnetic, electrical or optical excita-
resolved detection of nanosecond magnetic waveforms us- tion, absorptive heading or thermal drift. Whether such
ing the electronic spin of a single NV center in a diamond effects are present and how they can be mitigated, for ex-
nanoprobe. Employing microwave control pulses as short ample, by reducing the duty cycle or adjusting antenna
as 2 ns, we achieve a best-effort time resolution of 1.1 ns, design, will need to be considered on a case-by-case basis.
an instantaneous bandwidth of 0.9 GHz, and a time-of-
flight precision of better than 20 ps. The temporal reso-
lution can be optimized by decreasing the pulse rotation METHODS
angle or by numerical post-processing. The method does
not make any assumptions on the temporal shape of the Experimental apparatus
transient apart from that the signal must be repeatable
and allow synchronization with the NV detection (see Experiments were performed using a home-built confo-
below). cal microscope equipped with a custom green λ = 520 nm
Technical upgrades to the microwave delivery should diode laser [64] and a single-photon avalanche photo
allow improving the time resolution well into the pi- diode (APD, APCM-AQRH, Excelitas). The laser was
cosecond regime. While our demonstration operated at directly modulated (on-off) with ∼ 10 ns resolution. Ar-
Ω/2π = 125 MHz, on-chip CPWs have enabled Rabi fre- riving photons were separated using a dichroic beamsplit-
quencies of 440 MHz [48], and nanoscale antenna designs ter (Semrock FF526 Di01), time-tagged (National Instru-
should permit frequencies beyond 1 GHz [49], a roughly ments DAQ 6363) and time-binned in software.
10-fold improvement compared to this study. At such A schematic of the microwave and test signal circuit is
high frequencies, a number of effects including Bloch- shown in Fig. 2b. Microwave pulses for NV center control
Siegert shifts, violation of the rotating-wave approxima- were synthesized using IQ modulation of a local oscillator
tion and spurious excitation of the ω ′ transition (Fig. 2a) (LO, NI QuickSyn FSW-0020). The IQ modulation was
will need to be considered [25, 31, 48, 50]. Promisingly, performed with a home-built IQ modulator circuit based
these effects are accounted for in the deconvolution step on a HMC1097 chip (Analog Devices). The intermediate
(Methods), because the sensing kernel is calculated by a frequency (IF) was provided by the primary arbitrary
laboratory-frame simulation of spin dynamics. Further, waveform generator (AWG1, Tektronix AWG5014C) at
NV spin initialization and readout in the mS = −1 or carrier frequencies between 200 − 400 MHz.
mS = +1 state, rather than the mS = 0 state, could Test signals were synthesized using direct sampling
be used to mitigate partial excitation of the ω ′ transi- on a second waveform generator (AWG2, Keysight
tion [25]. M8190A). The AWG offers an instantaneous bandwith
Looking forward, combining our technique with scan- of 5 GHz ensuring that test signals requiring high band-
ning probe imaging [1, 3, 51, 52] will allow study- width can be generated accurately. For fast demonstra-
ing the nanosecond dynamics of magnetic systems with tion waveforms with frequency components 0.35 GHz, we
nanoscale spatial resolution. Such imaging experiments split, in software, the waveform into two streams contain-
have recently opened new avenues for analyzing spin tex- ing the < 0.35 GHz and > 0.35 GHz frequency compo-
tures and spin excitations in nanoscale magnetic systems, nents, respectively. This splitting was necessary because
ranging from thin-film ferromagnets to ferrimagnets and of the frequency bandwidths of the power amplifiers.
antiferromagnets [53–61]. Since our method relies on AWG1 was further used to synchronize the remaining
equivalent-time sampling and requires averaging, prereq- instrumentation through marker outputs, including mod-
uisites are that measurements can be repeated and trig- ulation of the diode laser, timing of the photon counting,
gered. This limits the range of systems and dynamics and triggering of AWG2.
that may be studied. These restrictions are shared with To fine-adjust the relative timing between AWG1 and
other time-resolved techniques that rely on pump-probe AWG2, we either used a programmable delay line (DL1-
schemes, including X-ray and magneto-optical imaging, 544-1600PS, MTS Systemtechnik, 5 ps step size) that was
which have been very successful at studying magnetiza- interfaced via a home-built interface (ice40HX8K-EVB,
tion dynamics. Examples given include magnetization Olimex), or the channel skew calibration of AWG1 (50 ps
reversals and domain wall propagation; the former has step size). This allowed sampling of the magnetic field
been demonstrated for up to 1011 repetitions [34] and transient at time increments that were much finer than
with jitter below 0.2 ns [41]. In addition to probing mag- the timing resolution of AWG1 (1 ns).
netization dynamics, time-resolved sensing also offers the We used two linear power amplifiers for signal ampli-
possibility to map time-varying currents allowing, e.g., fication. The high-frequency channel (> 0.35 GHz) con-
the in situ calibration of rf transmitters [47], study cur- taining the microwave pulses and high-frequency stream
rent dynamics in integrated circuits [62], light-induced of the test waveform used a 80 W, 0.8 − 4.2 GHz module
phase transitions in magnetic materials [35], or transient (80S1G4, Amplifier Research). The low-frequency chan-
photocurrents in light-sensitive electronic devices [63]. nel (< 0.35 GHz) containing the low-frequency stream
 7

of the test waveform used a 250 W, 0.009 − 300 MHz where B(t) is the (unknown) transient magnetic field that
module (BBA150, Rohde Schwarz). On the low-power we want to estimate, k(t) the convolution kernel, and n(t)
side, the microwave pulses were combined with the high- is the readout noise. Our goal is to construct a filter c(t)
frequency stream of the test waveform using a power such that
combiner (ZN2PD2-14W-S+, Mini-Circuits). On the
high-power side, signals were recombined using a custom γ B̃(t) = c(t) ∗ p(t) → γB(t) (5)
band diplexer (Universal Microwave Components Corpo-
ration). is close to B(t) in the least-squares sense. In the fre-
The diplexer output was connected via an ultra-low quency domain, this is achieved by
loss SMA cable (MaxGain 300, Amphenol Times Mi-
crowave Systems) to the input of a coplanar waveguide K̂ † (ω)|γ B̂(ω)|2
Ĉ(ω) = . (6)
(CPW) acting as the local antenna for exciting the NV |K̂(ω)|2 |γ B̂(ω)|2 + |N̂ (ω)|2
center. The CPW was photo-lithographically defined on
a quartz cover slip and placed between the objective and where hat symbols denote the Fourier transform of the
the diamond sample. The CPW had a 3 dB-bandwidth respective time-domain function.
of ∼ 7.5 GHz. The output of the CPW was connected Assuming N̂ (ω) has a flat (white noise) spectrum,
to a high-power attenuator (Meca 697-30-1, 30 dB, 50Ω) we can introduce the dimensionless parameter λ−1 =
and further connected to a fast microwave oscilloscope γ B̂(ω)/N̂ (ω) to parametrize the SNR and obtain an es-
(Rohde Schwarz, RTO2064). We used the digitized sig- timate of the input magnetic field,
nal of the oscilloscope to correct for imperfections of the
pulse and test signal delivery system (see text). ˆ K̂ † (ω)
γ B̃(ω) = P̂ (ω). (7)
The bias field B0 was generated by a NdFeB permanent |K̂(ω)|2 + λ2
magnet. The magnet was mounted on a XYZ translation
stage to adjust the magnitude of the bias field and align λ is adjusted manually and plays the role of a regulariza-
the field vector along the NV symmetry axis. tion parameter that controls the trade-off between SNR
and time resolution. When decreasing λ, the deconvolu-
tion allows fast features to be reconstructed more accu-
rately while becoming more sensitive to the noise present
NV centers in a measurement trace.
While the Wiener deconvolution can capture linear dis-
Experiments were performed on electronic-grade di- tortions as produced by pulse distortions, any non-linear
amond single crystals (Element6 Ltd.) with a natu- effects, such as phase accumulation violating |ϕ| ≪ π/2
ral isotope composition. NV centers were created by or the presence of a strong off-axis field component, can-
15 +
N ion implantation at an energy of 5 keV with a not be reverted. For this, the Wiener deconvolution has
fluence of 109 cm−2 , followed by annealing at 1200 ◦ C to be expanded to incorporate a non-linear function. This
for 4h in high vacuum and cleaning in a 1:1:1 tri-acid is known as the Wiener-Hammerstein model.
mixture of H2 SO4 : HClO4 : HNO3 at 120 ◦ C. Nanopil-
lar waveguide arrays were fabricated on membrane sam-
ples using electron-beam lithography and RIE etching
Kernel function and spin simulations
(QZabre AG) to increase photon yield. The continuous
wave (CW) photon count rate of single NV centers was
I0 ∼ 1 − 2 Mcts/s and the optical spin contrast ϵ between To calculate the sensing kernel from an input mi-
30% and 40%. crowave pulse waveform B1 (t), we performed a density
matrix simulation of the spin evolution. In the exper-
iment, B1 (t) was obtained by direct sampling of the
waveform with the downstream oscilloscope, as shown in
Wiener deconvolution
Fig. 2d. The simulation was performed in the laboratory
frame using the Hamiltonian
We used Wiener deconvolution to improve the time
resolution of our measurements. Wiener deconvolution is H = DŜz2 + γB0 Ŝz + γB1 (t)Ŝx + γBstim (t)Sz , (8)
closely related to Wiener filtering [65]. The Wiener filter
is a linear filter that minimizes the mean square error where γ = 28.0345 GHz/T is the gyromagnetic ratio and
between the estimated signal and the original signal. In D = 2π·2.87 GHz the zero-field splitting parameter of the
the situation that the convolution kernel function and NV spin, respectively, and Ŝx and Ŝz are spin-1 matrices.
the measurement SNR are perfectly known, the Wiener The first two terms in Eq. (8) define the energy spectrum
filter is the optimal filter. illustrated in Fig. 2a. The third term is the oscillating
Consider a noisy signal output p(t), given by microwave field generated by the microwave antenna; the
conversion factor between oscilloscope voltage signal and
p(t) = k(t) ∗ γB(t) + n(t), (4) magnetic field experienced by the NV spin is calibrated
 8

using Rabi oscillations. The fourth term is a δ-like stim-

a
Sensitivity
ulus for sampling the kernel function, implemented as a
Gaussian pulse with a line-width much smaller than the tmin
(expected) kernel width.
The kernel function k(t) was then obtained by time b
propagating the spin density matrix under the Hamilto-
nian of Eq. (8) [66] for a series of stimuli Bstim (t − t′ ),

Decreased sensitivity
each value of t′ providing one point of the kernel func-
tion. An example for a measured B1 (t) and calculated
k(t) is shown in Fig. 2d,e.
Importantly, because the simulation is executed in the
laboratory frame and involves all three spin states of the
NV center (mS = 0, mS = ±1), the simulation nat-
urally captures excitation of the non-resonant transition
and Bloch-Siegert-shifts due to the counter-rotating term
associated with B1 (t)Sˆx .
Improved time resolution

Calibration of magnetic signal amplitude Figure 5. Trade-off between sensitivity and time reso-
lution. a, Kernel function k(t) for decreasing rotating angles
The measured photon counts were converted to transi- α. The time resolution tmin is given by the FWHM and the
tion probabilties p(t) by normalizing them with reference sensitivity (given by Bmin ) is proportional to the area under
values for mS = 0 and mS = −1 recorded concurrently. k(t). b, Bmin plotted against tmin for the experimental pa-
rameters given with Eq. (9).
To calibrate the signal amplitude, pulsed ODMR mea-
surements were performed where a constant test signal
was applied using the diplexer. The carrier frequency of From this equation we deduce the minimum detectable
the < 1 us long π pulse was varied. The extracted fre- field (the “sensitivity”) by solving SNR = 1 for Bmin
quency shift was used to obtain a calibration for the data using ϕ = γBmin τ . This results in
shown in Figs. 3,4. q
α 4 − 2ϵ sin2 (α) p
Bmin = C0 . (11)
ϵγe τ (1 − cos (α)) sin (α)
Sensitivity
For a rotation angle of α = π/2, a kernel duration of
Starting from Eq. (6) in Ref. [25], we estimate the pho- τ = 2 ns and the above measurement parameters,
√ we find
ton counts C associated with a certain transition proba- a nominal sensitivity of Bmin ∼ 35 µT/ Hz. To inspect
bility p(ϕ = γe Bτ ) by how sensitivity can be traded for improved time resolu-
tion, we plot Bmin [given by Eq. (11)] against tmin [given
I0 T tint by the FWHM of the sensing kernel, Eq. (3)], shown in
C(ϕ) = [1 − ϵp(ϕ)] C0 = [1 − ϵp(ϕ)] , (9)
tseq Fig. 5.

where ϵ is the optical contrast of the NV center, I0

the count rate in continuous-wave illumination, T the Time-of-flight measurements
total measurement time, tint the integration time in
photon binning, and tseq the duration of the sensing To estimate the time-of-flight (ToF) between two sam-
sequence (given by the time between subsequent trig- pled waveforms, we needed to estimate the time delay.
gers). Representative values for our measurements are Numerous methods exist in the field of signal process-
I0 ∼ 1 − 2 Mcts/s, ϵ = 0.3 − 0.4, tint ∼ 300 − 500 ns, ing for this purpose. Here, we choose an approach that
and tseq ∼ 2.5 µs. We typically averaged until reaching does not require precise a priori knowledge of the sens-
C0 ∼ 0.1 − 1 Mcts, corresponding to total measurement ing kernel or waveform shape. In an ideal, free-space
times T ∼ 1 − 10 s, scenario, a time lag δt can be seen as a convolution with
From Eq. 9, we estimate the shot-noise-limited SNR to a kernel k(t) = δ(t − δt). Therefore, by deconvolving one
measured signal with the other, one obtains the shared
C(ϕ) − C(ϕ = 0) convolution kernel. For an acquisition with finite sam-
SNR = p
C(ϕ = 0) pling, this can be modeled by a sinc function and fitted.
ϵϕ (1 − cos (α)) sin (α) p (10) By using the Wiener deconvolution method, we can again
= q C0 . regularize the solution allowing for a solution optimal for
α 4 − 2ϵ sin2 (α) the given signal-to-noise ratio.
 9

Magnetic test waveforms magnetization of 75 µB /nm2 . The reversal was simulated

by a domain wall that passed across the disk [34] with
Domain wall propagation – The simulation assumed a velocity of 100 m/s. The domain wall was assumed to
a magnetic domain wall in a thin magnetic film with a have a width of 50 nm and a left Néel chirality; however,
surface magnetization of 25 µB /nm2 . The time trace was smaller domain walls and other chiralities do not much
calculated by first computing the spatial stray field profile affect the magnetic stray field. The NV center was po-
B(x) across a domain wall extending along y and located sitioned at z = 100 nm above the center of the disk. Its
at x = 0, and then converting the spatial profile into a vector orientation was along θ = 54◦ , ϕ = 90◦ . Note that
temporal profile by setting t = x/vd , where vd = 100 m/s real world signals such as the magnetization reversal in
is the domain wall velocity. Here, B(x) is the vector magnetic dots might be affected by jittering in the order
component projected onto the NV symmetry axis, given of a few nano seconds [36, 41], which might induce fur-
by θ = 54◦ and ϕ = 0◦ , where θ is the polar angle and ther blurring. For these devices, jitter could for example
ϕ the azimuth, respectively. The stand-off distance was be reduced by higher current densities or inplane mag-
assumed to be z = 150 nm. netic fields or adjusting the device geometry. However,
Magnetization reversal of magnetic disk – The sim- this limitation is inherent to all measurement approaches
ulation assumed a disk of 1 µm diameter and a surface that cannot capture the transient in a single shot.

[1] G. Balasubramanian, I. Y. Chan, R. Kolesov, M. Al- [10] A. Ariyaratne, D. Bluvstein, B. A. Myers, and A. C. B.
Hmoud, J. Tisler, C. Shin, C. Kim, A. Wojcik, P. R. Hem- Jayich, Nanoscale electrical conductivity imaging using a
mer, A. Krueger, T. Hanke, A. Leitenstorfer, R. Brats- nitrogen-vacancy center in diamond, Nature Communi-
chitsch, F. Jelezko, and J. Wrachtrup, Nanoscale imaging cations 9, 2406 (2018).
magnetometry with diamond spins under ambient condi- [11] J. Bylander, S. Gustavsson, F. Yan, F. Yoshihara,
tions, Nature 455, 648 (2008). K. Harrabi, G. Fitch, D. G. Cory, Y. Nakamura, J. S.
[2] J. R. Maze, P. L. Stanwix, J. S. Hodges, S. Hong, J. M. Tsai, and W. D. Oliver, Noise spectroscopy through dy-
Taylor, P. Cappellaro, L. Jiang, M. V. G. Dutt, E. Togan, namical decoupling with a superconducting flux qubit,
A. S. Zibrov, A. Yacoby, R. L. Walsworth, and M. D. Nat. Phys. 7, 565 (2011).
Lukin, Nanoscale magnetic sensing with an individual [12] S. Schmitt, T. Gefen, F. M. Sturner, T. Unden, G. Wolff,
electronic spin in diamond, Nature 455, 644 (2008). C. Muller, J. Scheuer, B. Naydenov, M. Markham,
[3] C. L. Degen, Scanning magnetic field microscope with a S. Pezzagna, J. Meijer, I. Schwarz, M. Plenio, A. Ret-
diamond single-spin sensor, Appl. Phys. Lett. 92, 243111 zker, L. P. McGuinness, and F. Jelezko, Submillihertz
(2008). magnetic spectroscopy performed with a nanoscale quan-
[4] F. Dolde, H. Fedder, M. W. Doherty, T. Noebauer, tum sensor, Science 356, 832 (2017).
F. Rempp, G. Balasubramanian, T. Wolf, F. Reinhard, [13] D. R. Glenn, D. B. Bucher, J. Lee, M. D. Lukin, H. Park,
L. C. L. Hollenberg, F. Jelezko, and J. Wrachtrup, and R. L. Walsworth, High-resolution magnetic reso-
Electric-field sensing using single diamond spins, Nat. nance spectroscopy using a solid-state spin sensor, Na-
Phys. 7, 459 (2011). ture 555, 351 (2018).
[5] V. M. Acosta, E. Bauch, M. P. Ledbetter, A. Waxman, [14] S. Kotler, N. Akerman, Y. Glickman, A. Keselman, and
L. S. Bouchard, and D. Budker, Temperature dependence R. Ozeri, Single-ion quantum lock-in amplifier, Nature
of the nitrogen-vacancy magnetic resonance in diamond, 473, 61 (2011).
Phys. Rev. Lett. 104, 070801 (2010). [15] J. M. Boss, K. S. Cujia, J. Zopes, and C. L. Degen, Quan-
[6] M. Lesik, T. Plisson, L. Toraille, J. Renaud, F. Occelli, tum sensing with arbitrary frequency resolution, Science
M. Schmidt, O. Salord, A. Delobbe, T. Debuisschert, 356, 837 (2017).
L. Rondin, P. Loubeyre, and J. Roch, Magnetic measure- [16] E. Magesan, A. Cooper, H. Yum, and P. Cappellaro, Re-
ments on micrometer-sized samples under high pressure constructing the profile of time-varying magnetic fields
using designed nv centers, Science 366, 1359 (2019). with quantum sensors, Phys. Rev. A 88, 032107 (2013).
[7] S. Hsieh, P. Bhattacharyya, C. Zu, T. Mittiga, T. J. [17] A. Cooper, E. Magesan, H. N. Yum, and P. Cappellaro,
Smart, F. Machado, B. Kobrin, T. O. Hohn, N. Z. Rui, Time-resolved magnetic sensing with electronic spins in
M. Kamrani, S. Chatterjee, S. Choi, M. Zaletel, V. V. diamond, Nat. Commun. 5, 3141 (2014).
Struzhkin, J. E. Moore, V. I. Levitas, R. Jeanloz, and [18] N. Xu, F. Jiang, Y. Tian, J. Ye, F. Shi, H. Lv,
N. Y. Yao, Imaging stress and magnetism at high pres- Y. Wang, J. Wrachtrup, and J. Du, Wavelet-based fast
sures using a nanoscale quantum sensor, Science 366, time-resolved magnetic sensing with electronic spins in
1349 (2019). diamond, Phys. Rev. B 93, 161117 (2016).
[8] K. Chang, A. Eichler, J. Rhensius, L. Lorenzelli, and [19] N. D. Zanche, C. Barmet, J. A. Nordmeyer-Massner, and
C. L. Degen, Nanoscale imaging of current density with a K. P. Pruessmann, Nmr probes for measuring magnetic
single-spin magnetometer, Nano Letters 17, 2367 (2017). fields and field dynamics in mr systems, Magn. Reson.
[9] J. P. Tetienne, N. Dontschuk, D. A. Broadway, A. Stacey, Med. 60, 176 (2008).
D. A. Simpson, and L. C. L. Hollenberg, Quantum imag- [20] M. Chipaux, L. Toraille, C. Larat, L. Morvan, S. Pez-
ing of current flow in graphene, Science Advances 3, zagna, J. Meijer, and T. Debuisschert, Wide bandwidth
e1602429 (2017). instantaneous radio frequency spectrum analyzer based
 10

on nitrogen vacancy centers in diamond, Applied Physics [37] G. Sala, C. H. Lambert, S. Finizio, V. Raposo, V. Kriza-
Letters 107, 233502 (2015). kova, G. Krishnaswamy, M. Weigand, J. Raabe, M. D.
[21] J. Zopes and C. Degen, Reconstruction-free quantum Rossell, E. Martinez, and P. Gambardella, Asynchronous
sensing of arbitrary waveforms, Phys. Rev. Applied 12, current-induced switching of rare-earth and transition-
054028 (2019). metal sublattices in ferrimagnetic alloys, Nature Materi-
[22] D. A. Anderson, R. E. Sapiro, and G. Raithel, Rydberg als 21, 10.1038/s41563-022-01248-8 (2022).
atoms for radio-frequency communications and sensing: [38] M. Hayashi, L. Thomas, C. Rettner, R. Moriya, and
atomic receivers for pulsed rf field and phase detection, S. S. P. Parkin, Direct observation of the coherent preces-
IEEE Aerospace and Electronic Systems Magazine 35, sion of magnetic domain walls propagating along permal-
48 (2020). loy nanowires, Nature Physics 3, 21–25 (2006).
[23] A. Manchon, J. Železný, I. M. Miron, T. Jungwirth, [39] D. Atkinson, D. A. Allwood, G. Xiong, M. D. Cooke,
J. Sinova, A. Thiaville, K. Garello, and P. Gambardella, C. C. Faulkner, and R. P. Cowburn, Magnetic domain-
Current-induced spin-orbit torques in ferromagnetic and wall dynamics in a submicrometre ferromagnetic struc-
antiferromagnetic systems, Rev. Mod. Phys. 91, 035004 ture, Nature Materials 2, 85–87 (2003).
(2019). [40] G. S. D. Beach, C. Nistor, C. Knutson, M. Tsoi, and
[24] N. G. Orji, M. Badaroglu, B. M. Barnes, C. Beitia, B. D. J. L. Erskine, Dynamics of field-driven domain-wall prop-
Bunday, U. Celano, R. J. Kline, M. Neisser, Y. Obeng, agation in ferromagnetic nanowires, Nature Materials 4,
and A. E. Vladar, Metrology for the next generation 741–744 (2005).
of semiconductor devices, Nature Electronics 1, 532–547 [41] E. Grimaldi, V. Krizakova, G. Sala, F. Yasin, S. Couet,
(2018). G. S. Kar, K. Garello, and P. Gambardella, Single-shot
[25] K. Herb and C. L. Degen, Quantum speed limit in quan- dynamics of spin-orbit torque and spin transfer torque
tum sensing, arXiv:2406.18348 (2024). switching in three-terminal magnetic tunnel junctions,
[26] C. Degen, F. Reinhard, and P. Cappellaro, Quantum Nature Nanotechnology 15, 111 (2020).
sensing, Rev. Mod. Phys. 89, 035002 (2017). [42] S. S. P. Parkin, M. Hayashi, and L. Thomas, Magnetic
[27] F. Jelezko, T. Gaebel, I. Popa, A. Gruber, and domain-wall racetrack memory, Science 320, 190 (2008).
J. Wrachtrup, Observation of coherent oscillations in a [43] S. Vélez, J. Schaab, M. S. Wörnle, M. Müller,
single electron spin, Phys. Rev. Lett. 92, 076401 (2004). E. Gradauskaite, P. Welter, C. Gutgsell, C. Nistor, C. L.
[28] L. I. Mandelstam and I. E. Tamm, The uncertainty rela- Degen, M. Trassin, M. Fiebig, and P. Gambardella, High-
tion between energy and time in nonrelativistic quantum speed domain wall racetracks in a magnetic insulator,
mechanics, 9, 249 (1945). Nature Communications 10, 4750 (2019).
[29] J. M. Taylor, P. Cappellaro, L. Childress, L. Jiang, [44] T. P. Dao, M. Müller, Z. Luo, M. Baumgartner,
D. Budker, P. R. Hemmer, A.Yacoby, R. Walsworth, A. Hrabec, L. J. Heyderman, and P. Gambardella, Chiral
and M. D. Lukin, High-sensitivity diamond magnetome- domain wall injector driven by spin–orbit torques, Nano
ter with nanoscale resolution, Nat. Phys. 4, 810 (2008). Letters 19, 5930 (2019).
[30] T. Zhu, J. Rhensius, K. Herb, V. Damle, G. Puebla- [45] A. Malozemoff and J. Slonczewski, Magnetic Domain
Hellmann, C. L. Degen, and E. Janitz, Multicone dia- Walls in Bubble Material, Applied solid state science
mond waveguides for nanoscale quantum sensing, Nano (Academic Press, 1979).
Letters 23, 10110 (2023). [46] A. Thiaville and Y. Nakatani, Domain-wall dynamics in
[31] P. Kairys, J. C. Marcks, N. Delegan, J. Zhang, D. D. nanowiresand nanostrips, in Spin Dynamics in Confined
Awschalom, and F. J. Heremans, Quantifying the limits Magnetic Structures III , edited by B. Hillebrands and
of controllability for the nitrogen-vacancy electron spin A. Thiaville (Springer Berlin Heidelberg, Berlin, Heidel-
defect, arXiv:2309.03120 (2023). berg, 2006) pp. 161–205.
[32] S. Gustavsson, O. Zwier, J. Bylander, F. Yan, F. Yoshi- [47] K. Herb, J. Zopes, K. S. Cujia, and C. L. Degen, Broad-
hara, Y. Nakamura, T. P. Orlando, and W. D. Oliver, band radio-frequency transmitter for fast nuclear spin
Improving quantum gate fidelities by using a qubit to control, Review of Scientific Instruments 91, 113106
measure microwave pulse distortions, Phys. Rev. Lett. (2020).
110, 040502 (2013). [48] G. D. Fuchs, V. V. Dobrovitski, D. M. Toyli, F. J. Here-
[33] F. Chan, J. Pauly, and A. Macovski, Effects of rf am- mans, and D. D. Awschalom, Gigahertz dynamics of a
plifier distortion on selective excitation and their correc- strongly driven single quantum spin, Science 326, 1520
tion by prewarping, Magnetic Resonance in Medicine 23, (2009).
224–238 (1992). [49] W. Rose, H. Haas, A. Q. Chen, N. Jeon, L. J.
[34] M. Baumgartner, K. Garello, J. Mendil, C. O. Avci, Lauhon, D. G. Cory, and R. Budakian, High-resolution
E. Grimaldi, C. Murer, J. Feng, M. Gabureac, C. Stamm, nanoscale solid-state nuclear magnetic resonance spec-
Y. Acremann, S. Finizio, S. Wintz, J. Raabe, and troscopy, Phys. Rev. X 8, 011030 (2018).
P. Gambardella, Spatially and time-resolved magnetiza- [50] F. Bloch and A. Siegert, Magnetic resonance for nonro-
tion dynamics driven by spin-orbit torques, Nature Nan- tating fields, Phys. Rev. 57, 522 (1940).
otechnology 12, 980 (2017). [51] F. Casola, T. V. der Sar, and A. Yacoby, Probing con-
[35] A. Kirilyuk, A. V. Kimel, and T. Rasing, Ultrafast optical densed matter physics with magnetometry based on
manipulation of magnetic order, Rev. Mod. Phys. 82, nitrogen-vacancy centres in diamond, Nat. Rev. Mat. 3,
2731 (2010). 17088 (2018).
[36] G. Sala, V. Krizakova, E. Grimaldi, C. H. Lambert, [52] J. Rovny, S. Gopalakrishnan, A. C. B. Jayich,
T. Devolder, and P. Gambardella, Real-time hall-effect P. Maletinsky, E. Demler, and N. P. d. Leon, New oppor-
detection of current-induced magnetization dynamics in tunities in condensed matter physics for nanoscale quan-
ferrimagnets, Nature Communications 12, 656 (2021). tum sensors, arXiv:2403.13710 (2024).
 11

[53] L. Rondin, J. P. Tetienne, S. Rohart, A. Thiaville, T. Hin- of a photocurrent vortex in monolayer mos2 using dia-
gant, P. Spinicelli, J. F. Roch, and V. Jacques, Stray-field mond quantum sensors, Phys. Rev. X 10, 011003 (2020).
imaging of magnetic vortices with a single diamond spin, [64] P. Welter, Microscopy of magnetic fields by scanning di-
Nat. Commun. 4, 2279 (2013). amond magnetometry, PhD Thesis, ETH Zurich, 2022
[54] P. Appel, B. J. Shields, T. Kosub, N. Hedrich, R. Hubner, (2022).
J. Fassbender, D. Makarov, and P. Maletinsky, Nano- [65] W. H. Press, S. A. Teukolsky, W. T. Vetterling, and B. P.
magnetism of magnetoelectric granular thin-film antifer- Flannery, Numerical Recipes 3rd Edition: The Art of Sci-
romagnets, Nano Lett. 19, 1682 (2019). entific Computing, 3rd ed. (Cambridge University Press,
[55] A. Jenkins, M. Pelliccione, G. Yu, X. Ma, X. Li, USA, 2007).
K. L. Wang, and A. C. B. Jayich, Single-spin sensing [66] K. Herb and P. Welter, Parallel time integration using
of domain-wall structure and dynamics in a thin-film batched blas (basic linear algebra subprograms) routines,
skyrmion host, Phys. Rev. Materials 3, 083801 (2019). Computer Physics Communications 270, 108181 (2022).
[56] Y. Dovzhenko, F. Casola, S. Schlotter, T. X. Zhou,
F. Buttner, R. L. Walsworth, G. S. D. Beach, and
A. Yacoby, Magnetostatic twists in room-temperature
skyrmions explored by nitrogen-vacancy center spin tex- ACKNOWLEDGEMENTS
ture reconstruction, Nature Communications 9, 2712
(2018).
[57] B. A. McCullian, A. M. Thabt, B. A. Gray, A. L. Me- The authors thanks Hugo Karras and Prof. Gun-
lendez, M. S. Wolf, V. L. Safonov, D. V. Pelekhov, V. P. nar Jeschke for providing the Keysight M8190A AWG
Bhallamudi, M. R. Page, and P. C. Hammel, Broadband for the measurements and their help during setup, and
multi-magnon relaxometry using a quantum spin sensor Dr. Theodor Prosch for fruitful discussions. This work
for high frequency ferromagnetic dynamics sensing, Na- has been supported by the SNSF through Project Grant
ture Communications 11, 5229 (2020). No. 212051, and by the SBFI through Project QMetMu-
[58] I. Bertelli, J. J. Carmiggelt, T. Yu, B. G. Simon, C. C. FuSP” under Grant. No. UeM019-8, 215927. J.M.A. and
Pothoven, G. E. W. Bauer, Y. M. Blanter, J. Aarts, and
L.A.V. acknowledge funding from SNSF Project Grant
T. V. der sar, Magnetic resonance imaging of spin-wave
transport and interference in a magnetic insulator, Sci. No. 201590.
Adv. 6, eabd3556 (2020).
[59] Q. Sun, T. Song, E. Anderson, A. Brunner, J. Forster,
T. Shalomayeva, T. Taniguchi, K. Watanabe, J. Grafe, AUTHOR CONTRIBUTIONS
R. Stohr, X. Xu, and J. Wrachtrup, Magnetic domains
and domain wall pinning in atomically thin CrBr3 re-
vealed by nanoscale imaging, Nature Communications
K.H. and C.L.D. conceived the experiment with input
12, 1989 (2021). from N.M., L.v.S. and P.G.. K.H. carried out the ex-
[60] A. Finco, A. Haykal, R. Tanos, F. Fabre, S. Chouaieb, periments with the help of L.A.V.. J.M.A. assisted in
W. Akhtar, I. Robert-Philip, W. Legrand, F. Ajejas, the preparation of diamond samples. K.H. and C.L.D.
K. Bouzehouane, N. Reyren, T. Devolder, J. Adam, performed the data analysis. All authors discussed the
J. Kim, V. Cros, and V. Jacques, Imaging non-collinear results.
antiferromagnetic textures via single spin relaxometry,
Nature Communications 12, 767 (2021).
[61] S. Velez, S. R. Gomez, J. Schaab, E. Gradauskaite, M. S. COMPETING INTERESTS
Wornle, P. Welter, B. J. Jacot, C. L. Degen, M. Trassin,
M. Fiebig, and P. Gambardella, Current-driven dynamics
and ratchet effect of skyrmion bubbles in a ferrimagnetic The authors declare no competing interests.
insulator, Nature Nanotechnology 17, 834 (2022).
[62] A. Nowodzinski, M. Chipaux, L. Toraille, V. Jacques,
J. F. Roch, and T. Debuisschert, Nitrogen-vacancy cen- DATA AVAILABILITY
ters in diamond for current imaging at the redistributive
layer level of integrated circuits, Microelectron. Reliab.
55, 1549 (2015). The data that support the findings of this work are
[63] B. B. Zhou, P. C. Jerger, K. Lee, M. Fukami, F. Mujid, available from the corresponding authors upon reason-
J. Park, and D. D. Awschalom, Spatiotemporal mapping able request.