Detection of biological signals from a live mammalian muscle using a diamond quantum sensor
James Luke Webb, Luca Troise, Nikolaj Winther Hansen, Christoffer Olsson, Adam Wojciechowski, Jocelyn Achard, Ovidiu Brinza, Robert Staacke, Michael Kieschnick, Jan Meijer, Axel Thielscher, Jean-Francois Perrier, Kirstine Berg-Sorensen, Alexander Huck, Ulrik Lund Andersen
DDetection of biological signals from a live mammalianmuscle using a diamond quantum sensor
James Luke Webb , Luca Troise , Nikolaj Winther Hansen , ChristofferOlsson , Adam M. Wojciechowski , Jocelyn Achard , Ovidiu Brinza , RobertStaacke , Michael Kieschnick , Jan Meijer , Axel Thielscher ,Jean-François Perrier , Kirstine Berg-Sørensen , Alexander Huck , and UlrikLund Andersen Center for Macroscopic Quantum States (bigQ), Department of Physics, Technical University ofDenmark, Kgs. Lyngby, Denmark Department of Neuroscience, University of Copenhagen, Copenhagen, Denmark Department of Health Technology, Technical University of Denmark, Kgs. Lyngby, Denmark Jagiellonian University, Krakow, Poland Laboratoire des Sciences des Procédés et des Matériaux, Université Sorbonne Paris Nord, 93430Villetaneuse, France Division Applied Quantum System, Felix Bloch Institute for Solid State Physics, Leipzig University,04103, Leipzig, Germany Danish Research Centre for Magnetic Resonance, Centre for Functional and Diagnostic Imagingand Research, Copenhagen University Hospital Hvidovre, Denmark
Abstract
The ability to perform noninvasive, non-contact measurements of electric signals pro-duced by action potentials is essential in biomedicine. A key method to do this is toremotely sense signals by the magnetic field they induce. Existing methods for magneticfield sensing of mammalian tissue, used in techniques such as magnetoencephalographyof the brain, require cryogenically cooled superconducting detectors. These have manydisadvantages in terms of high cost, flexibility and limited portability as well as poor spa-tial and temporal resolution. In this work we demonstrate an alternative technique fordetecting magnetic fields generated by the current from action potentials in living tissueusing nitrogen vacancy centres in diamond. With 50pT/ √ Hz sensitivity, we show the a r X i v : . [ phy s i c s . b i o - ph ] A ug rst measurements of sensing from mammalian tissue with a diamond sensor using mousemuscle optogenetically activated with blue light. We show these measurements can beperformed in an ordinary, unshielded lab environment and that the signal can be easilyrecovered by digital signal processing techniques. Sensing of signals produced by living tissue is an essential tool for both medical diagnosticsand for advancing the fundamental understanding of the structure and functioning of bio-logical systems. Such signals, generated by propagating action potentials, are of particularimportance in excitable cells such as neurons and muscle cells, allowing the cell-to-cell commu-nication and movement that is essential for the functioning of the tissue and the organism asa whole[1]. Action potential can be measured using electrical probes[2], but these are invasiveand can only be positioned to give limited spatial resolution. Magnetic field sensing providesa route towards noninvasive, high resolution, high speed sensing. To date, techniques forsensing the biological magnetic fields have been primarily based on superconducting quantuminterference devices (SQUIDs)[3, 4, 5, 6].This approach requires bulky magnetic shieldingand cryogenic cooling, thus preventing proximity studies of living tissue and delivering poorspatial resolution.Noninvasive stimulation and high-resolution imaging of magnetic fields in an unshielded,ambient environment can be realized by using nitrogen vacancy (NV) centres in diamond formagnetic field sensing[7, 8, 9]. NV centers are quantum defects that can provide broad-band vector magnetic field sensing[10, 11, 12, 13] and imaging with high spatial resolu-tion under ambient conditions using the technique of optically detected magnetic resonance(ODMR)[14, 15]. It has broad applicability in life science[16, 17] particularly due to thehigh biocompatibility of diamond, which can be placed in contact or even within biologicalspecimens[18, 19].Thus far NV sensing has focused on static or slow processes, such as imag-ing magnetotactic bacteria[20, 21]. As yet there has been limited demonstration of sensingbiological electrophysiological signals via magnetic field using diamond, with the most notablework being that by Barry et al.[22] for invertebrates. Difficulties have included reaching suffi-cient sensitivity, keeping the sample alive and undamaged during measurement, interferencefrom stimulation artifacts and the presence of background magnetic noise.In this work, we report the first use of a diamond quantum sensor to measure actionpotentials in vitro from a live mammalian specimen via their magnetic field. We detect theinduced field from the dissected leg muscle of a genetically modified mouse, using optogenetic2timulation of channelrhodopsin to induce the action potential through blue light stimulation.We achieve a magnetic field sensitivity of 50pT/ √ Hz and demonstrate methods that allow thespecimen to remain alive under stimulation for up to 20 hours. By using advanced data post-processing and filtering, we are able to demonstrate the first example of sensing of magneticfield from optogenetic stimulation of a biological system under ambient conditions in a noisy,unshielded laboratory. We consider these measurements an important step towards the goalof in vivo biosensing from living specimens, with the particular end goal of demonstratingsensing from neural networks in the mammalian brain[23, 24] . We used an inverted microscope containing a diamond magnetic field sensor, consisting of asingle crystalline diamond sample with a 20 µ m layer comprising a high density of NV centersat the top facet (Figure 1, see Methods section). The biological specimen was placed nearthe NV surface separated only by a foil/insulator layer, thereby ensuring high proximity of thespecimen to the sensing NV layer. Laser light at 532nm and frequency swept microwaves wereapplied to the NV sensor, while the induced fluorescence from the NV centers was imagedonto a photodetector. The time-varying magnetic field from the specimen was then detectedusing the protocol of optically detected magnetic resonance (ODMR) magnetometry. The diamond sensor was capable of measuring all the ambient background noise up to thekHz frequency range while maintaining maximum sensitivity without the sensor signal outputsaturating. An example of the raw magnetometer signal measured can be seen in Figure2,a). Assuming an ODMR linewidth of 1MHz, the approximate dynamic range without lossof sensitivity was estimated as 42 µ T, comfortably above the 600nT level of the predominant50Hz and 150Hz background noise.We first measured the background noise detected by the magnetometer with deionisedwater in the chamber but without a muscle. Figure 2,b) shows the amplitude spectral density,measured at microwave frequency on resonance (magnetically sensitive). Our noise floor wasapproximately 50pT/ √ Hz , defined by contributions from the electronic noise of the amplifiersand photodetector and the shot noise of the detected fluorescence. Here the shot noiselimited sensitivity was approximately 8pT/ √ Hz . To characterise the noise, we measuredthe magnetometer output over many hours. The result can be seen in the spectrogram in3igure 2,c), showing the range ( < Figure 3,a) shows a sketch of the fundamental biological process to be measured, wherestimulation with light triggers a cascading opening of ion channels, generating an actionpotential (producing current flow and magnetic field) along the muscle. Further details on thisprocess are given in Supplementary Information. Prior to the magnetometry experiment, thisoptogenetic stimulation was tested in a preliminary investigation in a separate setup capableof measuring action potential and muscle extension force. An example of a stimulation,measuring action potential using electrical probes and by measuring the force resulting fromsubsequent muscle contraction can be seen in Figure 3,b) . This test setup was used todetermine the intensity of light required for good stimulation. No saturation in the electricalprobe signal was observed up to the maximum intensity the light source could deliver.Figure 3,c) shows the response measured using an electrical probe contact to a stimulatedmuscle in the magnetometer sample chamber. We measured both the diamond sensor outputand the electrical probe contact simultaneously, to give a complete picture of the musclebehaviour. The maximum biological signal amplitude as measured by the electrical probesversus time is given in Figure 3,d). The signal strength decreased over time as the musclefatigued. This meant that after a certain time, a maximum signal to noise ratio was reachedwhere further averaging would not help resolve the biological signal in the magnetic data. Tofind this point, we calculated the signal to noise ratio of the signal as a function of numberof iterations during postprocessing. The rate of fatigue varied between different muscles,ranging from 8 hours in Figure 3 up to 16-18h.
Figure 4,a) shows the amplitude spectral density from Fast Fourier transforming the electricalprobe data. The majority of the signal can be found in a frequency range from DC up to4undreds of Hz (blue histogram plot), thus unfortunately overlapping with the majority of thebackground magnetic noise. We make the reasonable assumption that the magnetic readoutresembles the electrical probe readout since they originate from the same biological process.Therefore to filter the magnetic data, we limited the bandwidth to the range in which weexpect a signal, thereby rejecting the majority of the background noise. Postprocessing thedata collected, we imposed a digital bandpass filter from f low =20Hz to a range of upper cutofffrequencies to determine the minimum at which the filter begins to corrupt the electrical probedata. We chose an upper cutoff of f up =1.5kHz, to include as many of the signal frequenciesas possible. It can be seen clearly from the spectrum in Figure 4,a) that this was more thansufficient to resolve the signal while excluding a significant amount of background noise.In order to remove the background noise within the measurement bandwidth, we Fouriertransformed each 60s iteration dataset, selectively applied frequency domain notch filterscorresponding to the noise peaks and then inverse Fourier transformed the data to recoverthe a filtered version of the timeseries. Due to the frequency overlap between signal andnoise, it was critical to remove only parts of the signal that met two strict conditions: 1) tobe clearly defined as noise (peak sufficiently above the white noise floor) and 2) only at thosefrequencies that did not distort the sought biological signal (on applying the same filter to theelectrical probe data). Meeting only condition 1 would minimise noise while also removingthe sought biological signal, whereas meeting only condition 2 would artificially recover thebiological signal in the magnetic data by selection.We met these conditions by using two threshold values. The first, n th we define as themultiple above the median spectral amplitude a peak must exceed to be classed as noise.To apply this, we divided the spectrum into 40Hz wide windows, taking the median in eachwindow m v and removed only those frequencies in each window that peaked above m v × n th .By windowing, we avoided an excessive biasing of the filtering towards lower frequencies,due to the background 1/f x spectral slope. The second threshold value, m th , we define asthe percentage change in electrical probe signal relative to the unfiltered signal over a 40mswindow which starts at the time of stimulation (t=0). These methods are clarified further inSupplementary Information.We first removed those frequency components with the largest spectral amplitude (mostlikely to be noise) and continued until the SNR for each 60sec iteration was maximised,requiring between 60-200 notched frequencies. In the Supplementary Information, we showhow this process can be simplified by first removing the broad 50/150Hz mains noise throughtime domain filtering, flattening the spectrum in the < < The timeseries for Nx60sec iterations for both electrical probe and magnetic data was thenaveraged. Data was obtained separately from two muscles. The result can be seen inFigure 5,a) and b). We observe √ N scaling (Figure 5,e) reaching an ultimate (rms) noiselevel of 22pT for Muscle 1 and 16pT for Muscle 2. The improved sensitivity for Muscle2 was obtained with 12 hours more measurement averaging. For the second muscle, 2,3-Butanedione monoxime was added to the solution bath in order to inhibit movement withoutaffecting the action potential. For Muscle 1, this compound was absent. A signal wasobserved in the magnetic data for both muscles typical of an action potential propagatingalong the muscle. This signal was present with and without muscle inhibitor, ruling outthe signal being an artifact arising from muscle motion. For Muscle 2, a signal to noiseratio of 1 was reached after 32 iterations (30x32 simulations, 36 minutes measurementtime), defining SNR as the averaged signal strength divided by the standard deviation of theaveraged background magnetic noise. We phenomenologically modeled the expected actionpotential magnetic signal, full details of which are given in the Supplementary Information.The model parameters were within the range provided by literature and yield good agreementto the experimental data[26, 27, 28].We note that for Muscle 1 the diamond was placed approximately 2mm ± ± Using a diamond quantum sensor with pT-scale sensitivity to magnetic field and kHz mea-surement bandwidth, this work provides the first demonstration of sensing of the magneticfield from a signal generated by a living, mammalian biological specimen. We show that thesample can be kept alive for many hours while being measured and that the signal resem-bles that typical of action potentials measured by conventional electrical probes, without thedrawbacks of poor electrical contact adding capacitive distortion. We measure a time delaybetween magnetic and electrical probe signal consistent with signal propagation along themuscle. Using optogenetic activation and comparison to a muscle where motion had beeninhibited ensured the signal we measure was free of artifacts. The magnetometry technique isnot dependent on optogenetic stimulation and is widely applicable to conventional electricalprobe stimulation, or where stimulation originates from the living specimen itself.Using digital signal processing techniques, we show that a weak magnetic signal can berecovered in a noisy background without magnetic shielding, even in an ordinary laboratoryenvironment with a significant degree of background magnetic noise typical of that in a large,busy building at a university or a hospital. Unlike alternative methods for high-sensitivitymagnetometry, the high dynamic range of the diamond sensor allows the background noiseto be recorded without saturation. Since the sensor does not saturate, the background noisecan be directly detected and can thus be removed by adaptive windowed notch filtering. Weshow that this can also be done to a reasonable degree using fixed-width notch filters atmains harmonics frequencies. This could be implemented in hardware for realtime filteringin a portable sensor device to be used in a research or clinical environment[30]. Futureadvances in sensitivity will only help improve the clear identification of the different sourcesof background noise, and could eventually allow active cancellation of magnetic noise in asmall volume using a second sensor.The capability of operating in an ordinary lab or clinical environment without relyingon superconducting technology, would open the door to many new research and diagnosticpossibilities. A number of competing technologies seek to do this, most notably atomicvapour magnetometers[31, 32, 33, 34]. Although they are thus far superior in sensitivity,compared to diamond NV sensing, they have a number of disadvantages such as lack ofbiocompatibility, low dynamic range, inability to perform vector sensing in a single sensor,7he need to screen from the Earth’s magnetic field to achieve maximum sensitivity and lowbandwidth at maximum sensitivity ( < invivo signals from the mammalian brain. Our setup is designed to be capable of measuringsignals from dissected brain tissue slices and such measurements will take place in the nearfuture. Our setup is also designed to allow magnetic field and thus biological signal imaging.Karadas et al. [24] have calculated the level of magnetic field produced by typical neuronalsignals in the hippocampus to be between 10pT and 1.5nT. This is within the sensitivityrange for bulk sensing using our scheme, but further advancements in sensitivity are requiredfor imaging. Figure 1,a) shows a simplified 3D schematic of our inverted microscope incorporating thediamond biosensor. For optical pumping, up to 2W of horizontally polarised 532nm greenlaser (Coherent Verdi G2) illumination could be delivered from below a raised platform atBrewster’s angle for diamond (67 deg). Polarisation was controlled before incidence on thediamond using a half wave plate to ensure maximum power transmission. Red fluorescencefrom the diamond was collected separately from the incident green light using an aspheric,anti-reflective coated 12mm diameter condenser lens (Thorlabs ACL1210). Fluorescencelight was directed onto an electronically balanced photodetector (New Focus Inc.). 6mWwas the typical power of collected fluorescence for 2W of green laser light. A reference beamfor the photodetector was obtained by splitting off a few-mW portion of the input beam usinga polarising beamsplitter.
The diamond used in this work was a [100] oriented electronic-grade single crystal fromElement Six with dimensions 2 x 2 x 0.5mm overgrown by a 20 µ m thick nitrogen dopedlayer using chemical vapor deposition (CVD). Nitrogen content in the gas phase was optimisedduring the growth to reach a level of 5ppm of nitrogen-14. The diamond was then 2.25MeV8roton irradiated with a fluence of 3x10 protons/cm followed by annealing at 800 ◦ C for 4hours. This gave a NV − density ranging between 0.1 and 1 ppm. The diamond was mountedinto a central hole of a laser cut aluminum nitrate heatsink plate of dimensions 3x3x0.05cm .We measured an ODMR linewidth of 1MHz with a contrast of approximately 1.5 percent foreach nitrogen-14 hyperfine transition. The diamond and aluminium nitride plate were attached using watertight aquarium-safe sili-cone to a custom built broadband microwave antenna fabricated onto a printed circuit boardwith a hole for fluorescence collection from below (see schematic Figure 1,b). On top ofboth antenna and plate was silicone mounted a rectangular 3D-printed, custom designedrectangular plastic sample chamber, which can be seen in Figure 1,c), that could hold a flowbath of solution, fed using a peristaltic pump. The chamber was fully accessible from above,allowing biological samples to be introduced and probe electrodes to contact the sample usingmicromanipulators. The sample was held on a pair of sliding hooks within the bath, directlyabove the top surface of the diamond. To protect the biological sample from laser heating,a 16 µ m thick layer of aluminum foil was placed on the top surface of the diamond, attachedby 50 µ m Kapton tape in order to electrically insulate the foil and diamond from the sample.The resulting tens of micrometer separation between sample and diamond was undesirabledue to reduction in magnetic field strength, but was taken as a precaution against sampleheat damage based on previous experimental experience. The microwave field was generated using a three-frequency drive scheme [36] using two ra-diofrequency (RF) generators (Stanford SG394) feeding a balanced mixer and then amplified(Minicircuits ZHL-16W-43+). One generator drove the transition between the m s = 0 and m s = ± of the ground state of the NVs with a frequency in the range of 2.7-3GHz andfrequency modulated at 33kHz to implement lock-in detection. The second generator pro-vided a fixed frequency of 2.16MHz to drive multiple hyperfine transitions. Two rare-earthmagnets were aligned parallel to the (110) direction in the diamond and perpendicular to themain direction of signal current propagation, generating a DC bias field of ∼ magnetic data . Weused a lock-in time constant of 30 µ s, giving a magnetic field measurement bandwidth ofapproximately 4.8kHz. The muscle was surface contacted by an electrical probe consisting oftwo L-shaped AgCl coated silver wires positioned 3 mm apart under the muscle mounted ona micromanipulator. The recording electrode voltage was amplified (Axon Cyberamp 320).This was then digitised at the same rate and simultaneously with the magnetic data. Weterm this channel the electrical probe data . The muscle was stimulated optogenetically using blue light from a 470nm LED. Experi-ments were performed on genetically modified mice in which Channelrhodopsin 2 (ChR2),a light-gated cation channel, was used to create an action potential in the muscle. Ani-mals were euthanized by cervical dislocation and extensor digitorum longus (EDL) musclesfrom both hind limbs were dissected in carbogen-saturated (95 % O /5 % CO ) cold artificialcerebrospinal fluid (ACSF). Small suture loops were tied on distal and proximal tendons forlater suspension in the recording chamber. Until use, EDL muscles were stored in a hold-ing chamber continuously bubbled with carbogen. For some muscles, the myosin ATPaseinhibitor 2,3-Butanedione monoxime (5mM in ACSF; Sigma) was added in order to uncoupleexcitation from contraction, ensuring that we measure only action potential and removingany possible artifacts arising from sample motion. Full details of the biological preparationare given in Supplementary Information with this work.Prior to the experiment, the sample chamber and connecting tubing were cleaned by pump-ing heavily diluted household bleach through the system, followed by flushing with deionisedwater. This was then replaced with ACSF solution, carbogenated in a 500ml bottle and form-ing a closed circuit with the sample chamber. Temperature was measured in the chamber as34 ◦ C with laser and microwave power on. The mouse muscle was held in the chamber bysuture loops on hooks just above (but not in contact with) the diamond.10 .6 Stimulation Protocol
The muscle was optically stimulated every 2sec, with a light pulse length of 5ms. The absenceof contamination of the recording by a photovoltaic effect (Becquerel effect) induced by lightwas confirmed by taking traces recorded with the same protocols in the absence of muscle.Data from the magnetometer and from the electrodes was recorded from the ADC for 60secdata acquisition iterations during stimulation, giving 30 stimulations per iteration. The fulldata from both magnetic and electrical probe channels (2x 80kSa/sec x60sec) was storedfor postprocessing. Many hours of data aquisition was possible. Postprocessing ensured thatunexpected transient noise could be captured. A random delay time was implemented between60sec iterations (length between 10-30sec), ensuring each iteration began with a differentmains phase to assist averaging. The absolute start time of each iteration was recorded andthis measurement timeseries is used in the relevant plots in the results section. A fast ODMRsweep for selecting the optimal MW frequency was performed every five minutes during themeasurement to compensate for any thermal drift. [1] Kandel, E. R., Schwartz, J. H., Jessell, T. M., andA. J. Hudspeth, S. A. S. & Mack, S.Principles of neural science, fifth edition (2013).[2] Scanziani, M. & Häusser, M. Electrophysiology in the age of light.
Nature , 930–939(2009).[3] Jaklevic, R. C., Lambe, J., Silver, A. H. & Mercereau, J. E. Quantum interferenceeffects in josephson tunneling.
Physical Review Letters , 159–160 (1964).[4] Clarke, J., Lee, Y.-H. & Schneiderman, J. Focus on SQUIDs in biomagnetism. Super-conductor Science and Technology , 080201 (2018).[5] Hämäläinen, M., Hari, R., Ilmoniemi, R. J., Knuutila, J. & Lounasmaa, O. V.Magnetoencephalography—theory, instrumentation, and applications to noninvasivestudies of the working human brain. Rev. Mod. Phys. , 413–497 (1993). URL https://link.aps.org/doi/10.1103/RevModPhys.65.413 .116] Cohen, D. Magnetic fields around the torso: Production by electrical activity of thehuman heart. Science , 652–654 (1967).[7] Gruber, A. Scanning confocal optical microscopy and magnetic resonance on singledefect centers.
Science , 2012–2014 (1997).[8] Doherty, M. W. et al.
The nitrogen-vacancy colour centre in diamond.
Physics Reports , 1–45 (2013).[9] Taylor, J. M. et al.
High-sensitivity diamond magnetometer with nanoscale resolution.
Nature Physics , 810–816 (2008).[10] Schloss, J. M., Barry, J. F., Turner, M. J. & Walsworth, R. L. Simultaneous broadbandvector magnetometry using solid-state spins. Physical Review Applied (2018).[11] Fescenko, I. et al. Diamond magnetometer enhanced by ferrite flux concentrators.
Physical Review Research (2020).[12] Wolf, T. et al. Subpicotesla diamond magnetometry.
Physical Review X (2015).[13] Chatzidrosos, G. et al. Miniature cavity-enhanced diamond magnetometer.
PhysicalReview Applied (2017).[14] Delaney, P., Greer, J. C. & Larsson, J. A. Spin-polarization mechanisms of the nitrogen-vacancy center in diamond. Nano Letters , 610–614 (2010).[15] Wojciechowski, A. M. et al. Precision temperature sensing in the presence of magneticfield noise and vice-versa using nitrogen-vacancy centers in diamond.
Applied PhysicsLetters , 013502 (2018).[16] Schirhagl, R., Chang, K., Loretz, M. & Degen, C. L. Nitrogen-vacancy centers in dia-mond: Nanoscale sensors for physics and biology.
Annual Review of Physical Chemistry , 83–105 (2014).[17] Wu, Y., Jelezko, F., Plenio, M. B. & Weil, T. Diamond quantum devices in biology. Angewandte Chemie International Edition , 6586–6598 (2016).[18] McGuinness, L. P. et al. Quantum measurement and orientation tracking of fluorescentnanodiamonds inside living cells.
Nature Nanotechnology , 358–363 (2011).[19] Kucsko, G. et al. Nanometre-scale thermometry in a living cell.
Nature , 54–58(2013). 1220] Sage, D. L. et al.
Optical magnetic imaging of living cells.
Nature , 486–489 (2013).[21] Davis, H. C. et al.
Mapping the microscale origins of magnetic resonance image contrastwith subcellular diamond magnetometry.
Nature Communications (2018).[22] Barry, J. F. et al. Optical magnetic detection of single-neuron action potentials usingquantum defects in diamond.
Proceedings of the National Academy of Sciences ,14133–14138 (2016).[23] Hall, L. T. et al.
High spatial and temporal resolution wide-field imaging of neuronactivity using quantum NV-diamond.
Scientific Reports (2012).[24] Karadas, M. et al. Feasibility and resolution limits of opto-magnetic imaging of neuralnetwork activity in brain slices using color centers in diamond.
Scientific Reports (2018).[25] Webb, J. et al. Sensing of magnetic fields from biological signals using diamond nitrogenvacancy centres. arXiv (2020).[26] Cannon, S., Brown, R. & Corey, D. Theoretical reconstruction of myotonia and paralysiscaused by incomplete inactivation of sodium channels.
Biophysical Journal , 270–288(1993).[27] Nikolic, K. et al. Photocycles of channelrhodopsin-2.
Photochemistry and Photobiology , 400–411 (2009).[28] Yizhar, O., Fenno, L. E., Davidson, T. J., Mogri, M. & Deisseroth, K. Optogenetics inneural systems. Neuron , 9–34 (2011).[29] Juel, C. Muscle action potential propagation velocity changes during activity. Muscle &Nerve , 714–719 (1988).[30] Webb, J. L. et al. Nanotesla sensitivity magnetic field sensing using a compact diamondnitrogen-vacancy magnetometer.
Applied Physics Letters , 231103 (2019).[31] Boto, E. et al.
A new generation of magnetoencephalography: Room temperature mea-surements using optically-pumped magnetometers.
NeuroImage , 404–414 (2017).[32] Boto, E. et al.
Moving magnetoencephalography towards real-world applications with awearable system.
Nature , 657–661 (2018).1333] Jensen, K. et al.
Magnetocardiography on an isolated animal heart with a room-temperature optically pumped magnetometer.
Scientific Reports (2018).[34] Jensen, K. et al. Non-invasive detection of animal nerve impulses with an atomic mag-netometer operating near quantum limited sensitivity.
Scientific Reports (2016).[35] QuSpin. Technical documentation on QZFM Gen-2 OPM sensor, QuSpin Inc (2020).[36] El-Ella, H. A. R., Ahmadi, S., Wojciechowski, A. M., Huck, A. & Andersen, U. L. Op-timised frequency modulation for continuous-wave optical magnetic resonance sensingusing nitrogen-vacancy ensembles. Optics Express , 14809 (2017). We would like to thank Carmelo Bellardita for helping us do the immunohistochemistry. Weacknowledge the Core Facility for Integrated Microscopy, Faculty of Health and MedicalSciences, University of Copenhagen for using their confocal microscope for immunohisto-chemistry image acquisition. We acknowledge the assistance of Kristian Hagsted Rasmussenfor fabrication and diamond processing and Mursel Karadas (former DTU Heath Technol-ogy, currently New York University) for contributions and prior experimental and theoreticalmodeling work.
The project was conceived by AH and ULA. Methodology development, investigation andanalysis were performed by JLW, LT, NWH and AMW. NWH performed all animal dissections.Modeling work was performed by CO. Diamond growth and irradiation was performed byJA, OB, RS, MK and JM. This manuscript was written by JLW with editing and reviewcontribution by all other authors. The overall supervision was performed by AT, JFP, KBS,AH and ULA.
The authors declare that they have no known competing interests that would influence thework reported here. 14
Figures and Legends
Figure 1
Figure 1: Experiment schematic and photograph. a) Simplified 3D schematic of themagnetometer setup, showing the laser and blue LED illumination and fluorescence (FL)collection directions and sample chamber orientation, the direction of maximum magneticfield sensivity (B) and the direction of current flow (I) in the muscle. b) Side view schematic(not to scale) of the chamber/diamond sensor/MW antenna stack, joined and affixed to amovable plate with silicone. c) Photograph from above of the chamber, showing solutioninflow connections and the mouse muscle, below which the diamond lies separated by a layerof Kapton tape and aluminium foil acting as a heatsink.15igure 2
Figure 2: Demonstration of high dynamic range, bandwidth and sensitivity to mag-netic field. a) (upper pane) The raw unfiltered magnetic signal for (left) a full 60sec iterationand (right) for a zoomed 0.1sec segment of the same iteration. The signal was dominated bylow frequency and DC laser power drift ( < > ± µ T, well within the dynamicrange of the magnetometer. (lower pane) Spectral density in pT/ √ Hz for b) a single 60seciteration and c) a spectrogram of repeated 60sec acquisitions over 10 hours. The sensitivityfloor is approximately 50pT/ √ Hz with f(-3dB)=4.8Hz defined by the lock-in amplifier lowpass filter. Also indicated are calculations of the total noise, which includes electronic andshot noise, and of the estimated shot noise level alone. Many sources of background magneticnoise can be seen to peak well above this floor.16igure 3 Figure 3: Mouse muscle electrophysiology and signal variation over time a) Sketchof the biological signal generation process. In the muscle cell bi-lipid membrane (1) channel-rhodopsin (2) opening triggers influx of Na + ions (3), creating an action potential runningalong the muscle. b) Preliminary measurements taken on a separate setup of a single stim-ulation and readout via electrical probes (mV) and via muscle contraction force (mN). Thestrength of the signal as a function of light intensity is also shown. c) Example of the readoutof the biological signal in the magnetometer setup from a muscle (Muscle 1) by the electricalcontact probe. Here t=0ms is when the stimulation light is applied. The red trace shows theaverage signal observed over all stimulations. d) (left axis) Maximum size of the initial peakin the signal, which steadily drops by a factor of 2 over time as the muscle becomes fatigued.17igure 4 Figure 4: Frequency spectrum of the biological signal and defining optimum fil-ter thresholding. a) Spectrogram of the normalised Fourier transform amplitudes of theelectrical probe voltage data, showing that the majority of the signal frequency components(shaded blue region) are under 1.5kHz. b) Percentage deviation from the unfiltered signalas a function of upper bandpass cutoff frequency f up . The signal begins to be significantlycorrupted below 1.5kHz, as can be seen in the inset example for f up =400Hz where t=0 is thestimulation time. 18igure 5 Figure 5: Simultaneous electrical and magnetometer readout of the biological signal.