Ferromagnetic contamination of Ultra-Low-Field-NMR sample containers. Quantification of the problem and possible solutions
Giuseppe Bevilacqua, Valerio Biancalana, Marco Consumi, Yordanka Dancheva, Claudio Rossi, Leonardo Stiaccini, Antonio Vigilante
FFerromagnetic contamination of Ultra-Low-Field-NMR sample containers.Quantification of the problem and possible solutions.
Giuseppe Bevilacqua a , Valerio Biancalana a, ∗ , Marco Consumi b, , Yordanka Dancheva c , Claudio Rossi b ,Leonardo Stiaccini d , Antonio Vigilante a,e a Dept. of Information Engineering and Mathematics - DIISM, University of Siena - Via Roma 56 53100 Siena, Italy b Dept. of Biotechnology, Chemistry and Pharmacy - DBCF, University of Siena - Via A.Moro 2, 53100 Siena, Italy c currently at: Aerospazio Tecnologie srl, Strada di Ficaiole, 53040 Rapolano Terme (SI), Italy d Dept. of Physical Sciences, Earth and Environment - DSFTA, University of Siena - Via Roma 56, 53100 Siena, Italy e Currently at: Department of Physics and Astronomy, University College London, Gower Street, London WC1E 6BT,United Kingdom
Abstract
The presence of a weak remanence in Ultra-Low-Field (ULF) NMR sample containers is investigated onthe basis of proton precession. The high-sensitivity magnetometer used for the NMR detection, enablessimultaneously the measurement of the static field produced in the sample proximity by ferromagneticcontaminants. The presence of the latter is studied by high resolution chemical analyses of the surface,based on X-ray fluorescence spectroscopy and secondary ions mass spectroscopy. Methodologies to reducethe contamination are explored and characterized. This study is of relevance in any ULF-NMR experiment,as in the ULF regime spurious ferromagnetism becomes easily a dominant cause of artefacts.
Keywords:
Magnetic contamination, Volume/surface ferromagnetic contamination, Ultra-Low-field NMR,Ultra-Low-field MRI, Sample containers, optical magnetometry, chemical mapping.
1. Introduction
Most of the magnetic resonance (MR) experi-ments aimed at spectroscopic measurements or atimaging (MRI) suffer from distortions of the mag-netic field. Hence, particular care is necessary toprevent field disturbance induced in the vicinity ofthe sample by parts of the apparatus, including thesample itself.In conventional (high field) NMR experiments,field distortions causing line broadening or misshap-ing (in spectroscopy) as well as image artifacts (inMRI) arise most commonly from susceptibility (seee.g. Refs. [1, 2] and references therein) and con-ductance [3, 4], but also from ferromagnetic terms[2].In earth-field [5], ultra-low-field (ULF) [6, 7],and zero-field [8] MR apparatuses, the ferromag-netic terms may play an important –potentially ∗ Corresponding author
Email address: [email protected] (ValerioBiancalana) dominant– role, and their presence can be detecteddirectly by the same sensor that measures the MRsignal [9, 10]. This aspect is the focus of thepresent work. In particular, we study, character-ize and analyze spurious effects occurring in anULF-NMR apparatus, originating from ferromag-netic contamination of polymeric sample containers(cartridges). These cartridges contain samples fora remote-polarization NMR experiment that usesan optical atomic magnetometer (OAM) as a high-sensitivity, non-inductive detector [11, 12, 13].Magnetic detectors based on OAMs are aninteresting class of sensors that rival with thetop-sensitivity ones based on superconductingquantum-interference devices (SQUIDs), comparedto which they have advantages in terms of main-tenance cost, practicality (no cryogenics needed),and of robustness with respect to strong fields. Insome implementations, OAMs have a broadband re-sponse, extending to static signals. This feature ishere exploited to detect simultaneously both theDC signal due to ferromagnetic impurities and the
Preprint submitted to Elsevier July 1, 2020 a r X i v : . [ phy s i c s . i n s - d e t ] J un ime-dependent signal generated by nuclear preces-sion.This work is about the characterization of spuri-ous ferromagnetic remanence of the cartridges andof the material used for their production. Be-sides magnetometric measurements, surface anal-ysis based on X-ray fluorescence spectroscopy(XRFS) and Time-of-Flight Secondary-Ion Mass-Spectrometry (ToF-SIMS) have been performed. Inaddition, several approaches attempted to removeor reduce the contamination are described and dis-cussed, drawing conclusions about their effective-ness.Detection and analysis of contamination by fer-romagnetic impurities, as well as tiny ferromagneticbehaviour due to specific phenomena, are topics ofinterest among a wide community, spanning frommaterial science [14, 15], to semiconductor technol-ogy [16], nanotechnology [17], and medicine [18]. Arecent paper addresses the problem of characteriz-ing and counteracting ferromagnetic contaminationin a variety of metal-oxide substrates, in which weakextrinsic ferromagnetic behaviour is observed [19],also evidencing possible artifacts due to the mea-surement procedure.Works devoted to investigate extremely weak fer-romagnetic response in nano-structures [17, 15] ordiluted dopants [14] commonly make use of state-of-art magnetometers (typically SQUIDs) and studythe saturation level and the hysteresis that charac-terize the material. On the other hand our mea-surements are more tightly focused to the impli-cations of spurious ferromagnetism in ULF-NMRapparatuses and are mainly concerned with the re-manence.The paper is organized as follows: the Sec. 2 pro-vides a brief description of the OAM used, of thesetup making it suited to detect ULF-NMR signals,and of the instrumentation used to chemically ana-lyze the surface contaminants; the Sec. 3 shows theeffects caused by the spurious magnetization of thecontainers and complementary methodologies usedto evaluate and characterize the contamination; theSecs. 4 and 5 provide the results obtained by the ap-plication of methods aimed at preventing/removingthe ferromagnetic contamination, and the informa-tion inferred about the bulk or surface localizationof the problem. The Sec. 6 concludes the paper,providing a summary of the observations and of theresults as well as a discussion about possible impli-cations and perspectives.
2. Materials and methods S S α z y M B a) (a) S S z y M B b) (b)Figure 1: Front view of the detection region. The yellowarrow (M) represents the sample and its magnetization andthe circles (S , S ) represent the two sensors: glass cells con-taining Cs vapour. Laser beams cross the cell to opticallypump the Cs atoms and to probe the atomic state evolu-tion, they propagate along x and are not represented, x isalso the direction of the shuttling system. A static (bias)field (red arrow) is oriented along z . Sample and sensors areat different heights, determining the angle α and the conse-quent sample-sensor coupling factor. The dual magnetomet-ric head detects the field generated by the y component ofthe sample magnetization M as a difference-mode term (a).Such field (blue arrows) in this case is parallel/antiparallelto the static (much stronger) bias field, and contains a timedependent term due to the nuclear precession around B and-possibly- a static term due to the cartridge magnetization.The x and z components of M (b), produce a transverse fieldperturbation, so to cause only second-order variations of thefield modulus. Moreover such variations are cancelled in thedifferential measurement, because they appear as a commonmode term. The magnetometric setup is designed to measureNMR signals from samples that have been pre-viously magnetized in a field at the Tesla level,generated by a permanent-magnet Halbach array[20]. The nuclear precession occurs in a ULF (atmicro-Tesla level, corresponding to proton Larmorfrequencies of the order of 100 Hz), so to requirenon-inductive detection. The apparatus containsan OAM [21] that detects in situ the NMR signal, asystem for pneumatic sample transfer [22] from themagnetization- to the detection-region, and coils tocontrol [23] and to stabilize [24, 25] the static mag-netic field and to apply the magnetic pulses neces-sary to manipulate the nuclear spins, making themprecessing around the magnetic field direction.The whole setup (see Fig.1) is built around anunshielded dual OAM operating in a Bell & Bloomconfiguration [26] that is described in Ref. [21] andis adapted to ULF-NMR measurement as describedin Ref. [27]. The OAM is a broadband detec-tor, and it may record NMR signals superimposed2ith- (and modified by-) static terms generated bythe permanent magnetization [9]. The sensor re-sponse is nearly flat from DC to about 30Hz. Afterthis cutoff-frequency, the response decreases witha 6dB/oct roll-off [21], while maintaining a nearlyconstant signal-to-noise ratio for at least 3 octaves.This feature is due to the magnetic noise floor whichconstitutes the actual limit to the system sensitiv-ity: the response decrease acts simultaneously onboth the signal and on that magnetic disturbances.The intrinsic (non-magnetic) noise terms emergeand start to affect relevantly the S/N ratio aboveseveral hundred Hz.Briefly, the dual sensor detects the static andtime-dependent terms of the magnetic field in twolocations nearby the sample position, as sketchedin Fig. 1. The scalar nature of the atomic sen-sors [28], in the presence of a dominant bias fieldoriented along a given direction, makes the systemresponding only to the variations of the field com-ponent along that direction. A small variation of δ ~B over a bias field ~B causes a modulus variationof the total field δB tot ≈ ( δ ~B · ~B ) /B = δB k [29].As shown in Fig. 1a, in which the bias field isoriented along z , a dipolar source –displaced along y over the sensor plane– produces variations of B z when its magnetization is oriented along y . In thiscase the two sensors detect opposite δB tot, so tomaximize the difference-mode response of the mag-netometer output. Disturbances from far locatedsources appear as a common-mode term and areprofitably cancelled by difference. The cancellationis improved by an active compensation system. Thecommon mode signal feeds such system [25], so tobe actively attenuated prior to its cancellation bydifference. Attenuation and cancellation occur alsofor the field induced by sample magnetization along z : as shown in Fig. 1b, the field produced by M z sums in quadrature to the bias field, so to produceequal variations in the two sensors. In conclusion,only the y component of the sample magnetizationproduces a detected signal. It is worth mentioningan important role of the active stabilization system.As the measurement is performed in an unshieldedenvironment, slow drifts of the ambient magneticfield (despite being cancelled by the differerentialmeasurement) would affect the nuclear precession,hindering trace-averaging procedures used to im-prove the NMR S/N. More specifically, uncompen-sated field drifts would cause a T underestimationin the NMR signal. In this work we use water samples which, thanksto the proton abundance and long decay time, facil-itate the NMR characterization of the sample con-tainers. These containers can be used with otherNMR substances, as previously reported [11, 27].The containers are sealed cartridges with screwedor glued caps. The samples are remotely polarized(at about 1 T) and pneumatically shuttled to thedetection region, there –after the application of ap-propriate spin-tipping field pulses– the nuclei pre-cess in a field at microtesla level. To guarantee per-formance and reliability of the transfer system [22],an accurate shaping and a sufficient mechanical ro-bustness of the cartridges are required. To this end,an accurate selection of the material is necessary.Several reasons make the use of metal cartridgesdisadvantageous. Despite the low nuclear preces-sion frequencies typical of the ULF regime, eddycurrents induced in electrically conductive contain-ers may shield both the NMR signal and the spintipping field. In our setup, the shielding effect hin-ders particularly the tipping procedure. The latteris performed by means of sudden (non-adiabatic)rotation of the static field, or by the application ofresonant pulses. In the resonant case, the species-selectivity is enhanced by applying static field atthe hundreds of µ T level together with the ac fieldpulse, which increases the resonant frequency upto the 10 kHz level: at these frequencies the skin-depth in metals is submillimetric. Moreover, sev-eral metallic and metal-alloy materials contain rel-evant traces of ferromagnetic contaminants [30], asit was recently pointed out in a ULF-NMR exper-iment [10]. Additionally, non-metallic containersmay guarantee a wider chemical compatibility withthe NMR samples to be analyzed.The shuttling system requires precise externalsizing and low fragility, making glass a disadvan-tageous choice. Among other non-conductive ma-terials, we have selected and charactrized polyetherether ketone (PEEK) for its excellent mechanicalproperties (machinability, chemical resistance, me-chanical strength). Some attempts made with otherpolymeric substances –e.g. Iglidur and Acetale– ledto similar observations. We will briefly deal alsowith additional tests performed on other materialsused to produce polymeric samples with 3D print-ing. PEEK finds applications in MRI/NMR setups[1], and –by virtue of its biocompatibility– is oftenselected in medicine to replace titanium in implan-3ations, also to avoid artifacts in the MRI post-surgical evaluations [31].When δB tot is produced by a NMR sample, thetime dependent term is due to nuclear magnetiza-tion that precesses in the xy plane, around the z direction, so to have an y component oscillating atthe nuclear Larmor frequency. Permanent magneti-zation of the sample container appears in the signalas well, with a static difference-mode field variation(DMFV) proportional to the y component of themagnetization. The z component of the magneti-zation produces instead first order effects on theprecession frequency of nuclei inside the container.Summarizing, the broadband response of themagnetometer permits to register directly a staticsignal due to M y , simultaneously with the nuclearprecession signal, whose frequency depends on thebias field and is modified by M z .Beside this NMR frequency shift, the ferromag-netic contamination causes a broadening of the nu-clear resonance [9]. Both NMR shift and broad-ening can be estimated shot by shot by means ofreliable numerical methods [32, 33]. Repeated (cy-cled) NMR measurements are commonly performedto improve the signal-to-noise ratio. In our case,as the cartridge undergoes unpredictable rotationsduring the sample transfer, the effects of the fer-romagnetic term appear as distributed widths andshifts of the NMR. These effects can be quantifiedin terms of: (i) variance of the shot-by-shot fre-quency estimate, (ii) variance of the shot-by-shotdecay-rate estimate, (iii) apparent decay-rate in-crease in the average trace (where T2* decreasesdue to fluctuations of the ferromagnetically inducedshift, which –as said– varies with the cartridge ro-tation angle). More details about these kinds ofanalyses can be found in ref.[9] Previous measurements [9] performed in ourlaboratory, differing from other observations [10],pointed out that PEEK cartridges show indeedan unexpected magnetization level, which can beevidenced – with no spatial resolution – by DCmeasurements and by ULF NMR spectroscopy. Achemical analysis of the sample containers surfaceprovides a complementary insight on the nature ofthis problem. To this end, ToF-SIMS and XRFStechniques are applied, to extract useful informa-tion about the amount and the morphology of thesurface contamination. XRFS are performed using an Olympus DeltaPremium (INNOV-X) instrument to determine ma-jor and trace elements, as previously described [34].The analyses can be performed on the polymericsamples non-destructively.ToF-SIMS measurements are carried out ona TRIFT III spectrometer (Physical Electronics,Chanhassen, MN, USA) equipped with a goldliquid-metal primary ion source by a procedure al-ready reported [35]. There are some geometricalconstraints, which limit the size of the analyzedsample. In particular, NMR cartridges do not fitin the holder, and are sacrificed to undergo ToF-SIMS analysis. As an alternative, smaller polymersamples are produced with the same lathing tools.XRFS and ToF-SIMS analyse small portions ofthe polymer surface, so to make their (local) resultsnot necessarily consistent (in terms of estimatedamounts) with the (global) magnetometric mea-surements. The different kinds of analyses providecomplementary information and may help in con-firming interpretative hypotheses, while perform-ing direct cross-correlation of the observations isnot obvious and straightforward, due to the differ-ent subjects of measurements (small surface por-tions, and global surface- or volume-contaminants,respectively).Compared to ToF-SIMS, the XRFS has a lowersensitivity (13 ppm, instead of 10ppb-1ppm), but adeeper penetration (several microns instead of fewnanometers). ToF-SIMS offers an excellent lateralresolution, which approaches the micrometric level.In addition ToF-SIMS may produce hyperspectralmaps, where different contaminants can be distin-guished with a mass resolution ( m/ ∆ m ) of severalthousands. Different elements are identified on thebasis of the mass to charge ratio ( m/z ).Chemical images relative to positive ion spectraare acquired with a pulsed, bunched 22 keV Au + primary ion beam, by rastering the ion beam overa 300 µ m × µ m area. Static ToF-SIMS con-dition (primary ion dose density < ions/cm )are not requested, so that acquisition time extendsto 10 minutes. Positive ion spectra are calibratedwith CH +3 ( m/z = 15 . H +3 ( m/z = 27 . H +5 ( m/z = 41 . − Pa. The mass resolution ( m/ ∆ m ) is 2000at m/z = 27.4 . Evidence of contamination Ferromagnetic contaminants on the surfaceand/or in the polymer bulk, cause a remanencethat, after the premagnetization in the Halbach ar-ray, makes the cartridge produce a nearly dipolarfield on its exterior and, internally, a field in aver-age antiparallel to the magnetization and scarcelyhomogeneous. We are dealing with a spurious re-manence that is extremely weak, so to produce onlyperturbative effects.Concerning the mentioned dipole approximation,an estimation of the dipolar term and of higher-order multipolar ones, can be derived on the ba-sis of ref.[36] with the assumption of uniformly dis-tributed contaminants. A multipolar expansion ofthe field calculated in the proximity of a uniformly,transversely magnetized cylinder of radius R andsemi-length L, with magnetization M shows that,indicating with z , ρ and ϕ the radial, axial and az-imuthal co-ordinates with respect to the center ofthe cylinder, the field components are B z ≈ µ M LR ρz ( z + ρ ) / (cid:2) L (cid:0) z − ρ (cid:1) − R (cid:0) z − ρ (cid:1) ( z + ρ ) (cid:3) cos( ϕ ) , (1) B ρ ≈ µ M LR ( z + ρ ) / h(cid:0) ρ − z (cid:1) ++ 18 (3 R − L ) (cid:0) z − ρ z + 4 ρ (cid:1) ( z + ρ ) i cos ϕ, (2)and B ϕ ≈ µ M LR ( z + ρ ) / h L − R ) (cid:0) z − ρ (cid:1) ( z + ρ ) i sin( ϕ ) , (3)respectively, where the first lines are dipole terms,quadrupole terms are missing, and the second linesare octupole terms. On the z = 0 (equatorial)plane, the axial component vanishes and the tworemaining simplify to B ρ ≈ µ M LR ρ (cid:20) R − L ρ (cid:21) cos( ϕ ) (4) B ϕ ≈ µ M LR ρ (cid:20) R − L ρ (cid:21) sin( ϕ ) , (5)respectively. In the geometry of our setup, where R/ρ ≈ / L/ρ ≈ /
3, the dipolar terms ex-ceed the octupolar ones by more than an order ofmagnitude.The DMFV depends on the z component of thesample field at the position of the sensor, which (seeFig.1) is given by B ρ cos( α ) − B ϕ sin( α ). The ex-treme values of the DMFV observed in large setsof measurements provide a quantitative estimate ofthe dipolar moment and hence of the average poly-mer magnetization.The orientation of the cartridge magnetizationis originally parallel to the field generated by theHalbach array, but undergoes an unpredictable,aleatory rotation θ during the displacement to themeasurement region. Thus ϕ = θ − α is randomlydistributed and the field perturbation due to thecartridge magnetization at the measurement stagevaries on a shot-by-shot basis. As described inref.[9], the response of the magnetometer –which,in turn, is calculated in [21]– to static field vari-ations can be inferred from the differential phaseshift of the polarimetric signals extracted from thetwo magnetometer outputs.Fig.2 shows histograms of characteristic decaytimes, DMFVs and proton precession frequenciesestimated in two large sets of NMR measurements,prior and after having removed the ferromagneticcontamination (as discussed in the next sections).The first row ((a), (c), (e)) refers to a 804 shots mea-surement performed on a contaminated cartridge,while the second row ((b), (d), (f)) shows resultsobtained in 430 shots with a cleaned cartridge (seeSec.5).The characteristic decay time is evaluated bothas an average of estimations performed on singletraces (thick, black-dot line, with the thinner linesindicating the ± standarc deviation interval), or di-rectly on the average trace. The discrepancy be-tween the two estimations is much larger in the caseof the contaminated cartridge. This is a direct con-sequence of the narrower NMR frequency distribu-tion in the clean sample compared to the contam-inated one (cfr. (e) and (f)). A second evidenceis that in (b) the decay rate has narrower distribu-tion compared to (a). It is worth noting that theshortening of the mean decay time observed in (b)is not relevant. It is due to having reloaded the car-tridge with a different water sample after the clean-5 cc u r e n c e s (a) (s) (b) (e) Frequency (Hz) (f)
30 20 10 0 10 20 30010203040506070 (c)
30 20 10 0 10 20 30
DMFV (nT) (d)
Figure 2: Cartridge characterization in terms of relaxation time (a) and (b), DMFV distribution (c) and (d), and NMRfrequency (e) and (f), over two large sets of measurements. The histogram in the first row (a) (c) and (e) are obtained with acontaminated cartridge, while those in the second row (b) (d) and (f) correspond to a clean cartridge. ing process, not to ferromagnetic contamination.As an example of ULF-NMR measurement results,Fig.3 shows averaged traces and corresponding am-plitude spectral density (ASD) plots obtained witha contaminated cartridge and a clean one, respec-tively. These data correspond to the histograms re-ported in Fig.2 (a,c,e) and (b,d,f), respectively. Thedecay-time graphical estimates (red-dashed lines)match the numeric evaluations of Fig.2 (a) and (b).Other tight and quantitative evidences of ferromag-netic contamination emerge from the comparison ofmeasured DMFV distributions (Fig.2 (c) and (d)).The DMFV range decreases in this case from ±
15 to ±
3. Concomitantly, the distribution of nuclear pre-cession frequencies (Fig.2 (e) and (f)), undergoes arange narrowing by a factor of three. In the case ofordinary cartridges, a DMFV of ±
40 nT is typicallymeasured. The DMFV is an indicator of magneticcontamination, which, as discussed in the follow-ing, occurs –to a large extent– when the PEEK islathed. There, uncontrolled and non-reproducibleparameters (such as exact tool positioning, cuttingangles, tool edge sharpness) introduce an importantvariability. However, repeated measurements overmany sample permit to estimate typical DMFV val-ues with reasonable accuracy.The dipole moment, and hence the average mag-netization M , of the material are evaluated from the DMFV value and from a coupling factor deter-mined by the geometry of the arrangement (eqs.5and 4).In the following we will report the directly mea-sured quantity (i.e. the DMFV), reminding that thedipole is proportional to the DMFV, and hence themagnetization is proportional to the DMFV and in-versely proportional to the polymer volume. In ourgeometry, where ρ ≈
50 mm, R=9.3 mm, L=16mm, and α ≈ × times the DMFV in nT, i.e. the ± . µ emu/cm ) (to be doubled, if one considersonly the PEEK volume).Beside the DMFV, the internal field variation in-side the cartridges can be estimated from its effectson the NMR signal: as mentioned in Sec.2.1 anddiscussed in Ref. [9], this feature makes it possibleto estimate both the y and the z components of themagnetization simultaneously. In the following (seeSec. 4) the cartridge magnetization will be linkedto a spread of NMR proton frequencies estimatedin repeated measurements.Sets of magnetometric measurements have beenperformed using cylindrical solid PEEK sampleshaving the same external shape of the cartridges.Hereafter these samples will be referred to as cylin- i g n a l ( a r b . un i t s ) A S D ( a r b . un i t s ) (a) Time (s) (b) (c) Frequency (Hz) (d) Figure 3: Normalized average traces from 804 NMR measurements with a contaminated cartridge (a), and from 430 mea-surements with a clean cartridge (b). The ASD of the corresponding raw data are shown in the black plots (c) and (d),respectively. A 150 Hz notch filter and a band pass filter around the NMR peak are applied to produce the blue ASD plotsand the time-domain traces (a) and (b). In these latter, the green horizontal lines indicate the value 1 /e , and the red-dashedlines indicate the numerically estimates of the decay time: the same as in Fig.2 (a) and (b).Figure 4: The picture shows an empty cartridge and acylindrical sample, they have identical external size. Therelevant dimensions of the cartridge are reported (in mm) inthe drawing. The polymer volume of cartridges is 53% withrespect to cylinders, while the machined surface of cylindersis 57% with respect to cartridges. ders . Both kinds of samples are shown in Fig. 4, to-gether with their relevant sizes. Of course an ULF-NMR characterization is not feasible in the case ofcylinders: the magnetometric analysis is limited tothe DMFV measurements.Typically, cylinders produce half DMFV thancartridges. Being the cylinders about double inpolymer volume and half in machined surface com-pared to cartridges, such DMFV is attributable to a ferromagnetic surface contamination.This consideration highlights the importance ofcomplementing the magnetometric measurementswith spatially and chemically resolved analyses.XRFS performed on ordinary cartridges andcylinders has provided an immediate evidence offerromagnetic contaminants, e.g. the first analyzedcylinder (machined with conventional tools) showedthe presence of several metallic elements (includingferromagnetic ones), among which: Fe (42 ppm),Cr (21 ppm), Cd (18 ppm), Zn (13 ppm). XRFSis useful to obtain initial indications about the con-taminants present on the polymer surface, howeverits poor sensitivity makes it unsuitable to analyseweakly contaminated samples. Thanks to its highersensitivity and spatial resolution, ToF-SIMS anal-yses provide a deeper insight, which include alsomorphological information. Images produced byToF-SIMS confirm the presence of metal impuri-ties implanted on the surface of several polymericsamples.Repeated ToF-SIMS maps performed on similarsamples confirm the presence of iron (and othermetals) contamination, often appearing in micro-metric fragments (see Fig5a). They also show thatthe contamination may occur with a scarce repro-ducibility. As an example, in Fig.5b, a spread Fe7ontamination interests an area about 200 µ m in di-ameter, around a localized micrometric spot with ahigh concentration of Al and Si. (a) (b)Figure 5: ToF-SIMS images of submillimetric portionsof PEEK surface machined with conventional tools (high-speed-steel, HSS). The data bars measure 100 µ m. Theimage at left shows the total ion concentration (in red) inoverlay together with the iron image (in green). Two welldefined iron spots appear, denoting the presence of localizedparticles about 15-20 µ m in size. In other instances differentiron distribution are recorded. In the case shown in the im-age at right, the iron contaminant is spread over a wider area(about 200 µ m in diameter), around a 20 µ m spot containinga high concentration of Al (in blue) and Si (in red). In conclusion, ToF-SIMS analyses confirm thesurface nature of the ferromagnetic contamination,also in accordance with information provided by thePEEK producer [37]: no ferromagnetic substancesare used in the PEEK production process, so thattheir presence in traces can only be attributed toresidual impurities.However, as discussed in the following (see Secs.4and 5), there is also an evidence that some vol-ume contamination level is also present. Comparedto surface contamination, such bulk term is muchweaker, to the point of making it tolerable for ourapplications.Several approaches have been attempted to pre-vent or to remove the surface contamination, andtheir effectiveness have been evaluated on the ba-sis of the above mentioned techniques. They aredescribed extensively in the Secs.4 and 5.
4. Preventing surface contamination: non-ferromagnetic tools
The surface contamination most probably occursat the machining stage. To prove that, titaniumtools have been crafted and used at the lathingstage to avoid direct transfer of contaminants fromthe blade to the PEEK surface. In this way two additional evidences emergedproving that surface contamination occurs duringthe manufacturing process. The first one is indirect:polymeric samples machined with titanium toolsinstead of ordinary HSS tools give much weakerDMFV. The second evidence is direct and is pro-vided by ToF-SIMS maps, where no micrometriciron spots appear, while a barely detectable, homo-geneously spread iron contamination is sometimesvisible. (a) (b)Figure 6: (a) Iron map from ToF-SIMS of a submillimetricportion of PEEK surface machined with a Ti tool. The databar measures 100 µ m. The spread iron contamination ap-pears at ppb level, and might be due just to instrumentationnoise, in analogy with Al and Ti maps. (b) The Ti tool bladespoiled after having machined two PEEK cylinders. Titanium-machined samples result in a muchweaker static field perturbation (DMFV ≈± ± ≈ ±
200 pT). This is the low-est DMFV value obtained in term of magnetic con-tamination and it is the same value measued withanother method that removes the cartridge contam-ination discussed in the next section. This fact,combined with the observed reversed ratio betweenDMFV from low contaminated cartridge and cylin-der, is a reliable indication that the residual levelof ferromagnetism comes from the material bulk.Concerning the ToF-SIMS measurements, no ironmicroparticles appear in Ti-worked samples. A veryweak concentration (at ppb level, close to the in-strumental sensitivity) of sub-pixel size iron sig-nal is recorded, as shown in Fig.6a. Due to the8oorer hardness and resilience, the Ti tools sharp-ness is spoiled in a much shorter time with respectto the steel (see Fig.6b) [38], suggesting that muchmore particles are ripped out. Surprisingly, poly-meric samples worked with Ti tools result uncon-taminated also in terms of Ti particles, when an-alyzed chemically (both Al and Ti maps do notpresent spots in Ti machined samples). A possibleexplanation is that the morphology of the rippedTi fragments does not facilitate their penetrationin the PEEK surface.Summarizing, the use of non-magnetic machiningtools is a promising and effective approach to pre-vent the ferromagnetic contamination of cartridges.The titanium tools do not allow to make screwedcaps, making necessary the use of glue to seal thecartridges. From a practical point of view, the ex-cellent thermal and mechanical properties of PEEKfacilitate the re-use of glued-cap cartridges. Thecap can be removed heating up the aqueous contentin a commercial microwave oven, so to produce anover pressure and to explode the container.Glass lathing tools were also tested, with resultsas good as those obtained with titanium in terms ofcontamination, but with the drawback of an evenworse resilience. Another attempt was made with atungsten carbide tool as specifically recommendedfor PEEK machining [38]. This tool did not suf-fer from the wearing shown by the Ti one, but themagnetic contamination resulted similar to the HSScase.
5. Removing surface contamination: clean-ing methods
An alternative procedure to reduce the surfacecontamination is the use of mechanical or chemicalmethods to remove residues from HSS-machining.
The use of appropriate (magnetically tested)sand-paper to polish the cylinder surface after itsHSS-machining helps in reducing the ferromagneticresponse. The procedure is barely reproducible,and (crucial point) sand paper polishing is not prac-ticable in the inner surface of cartridges, for which,consequently, alternative methods have to be devel-oped in view of application. However, the effect isevident and this suggests that an intensive mechan-ical abrasion might constitute an valid method toproduce clean containers. In this perspective, a mechanical (abrasion)cleaning approach based on a tumbling techniquewas attempted and tested. Both cylinders and car-tridges with threaded caps were treated in a tum-bling machine with corundum (Al O ) sand, whichis commercially available and is commonly used,e.g., for jewelry polishing. Different tumbling dura-tions were applied, ranging from ten hours to sev-eral days.The tumbling technique produces evident re-sults, however the contaminant reduction does notachieve a satisfactory level. In the case of cylinders,20 hours tumbling reduce the signal by 10% (DMFV ≈ ±
18 nT), and further 20% (DMFV ≈ ±
15 nT) isobtained after additional 20 hours. A long last-ing (250 h) tumbling, resulted in a 40% reductionDMFV (from ≈ ±
16 nT to ≈ ±
10 nT): an appre-ciable but yet improvable level, despite an evidentpolishing of the surface, that after such long treat-ment appeared glossy to the naked eye and perfectlysmooth to the touch.These results are consistent with the ToF-SIMSanalysis performed on a tumbled sample. As shownin Fig.7, there is an evidence of residual iron con-tamination but no large size (micrometric) parti-cles are visible anymore, an additional, persistentSi and Al contamination is pointed out. The latteris more evident where the mechanical action of thetumbling is stronger (in the proximity of the edges).This indicates that some corundum particles pen-etrate the surface as to produce Al contamination.In conclusion, both magnetometric and ToF-SIMS results indicate that the tumbling approachis not a promising method. It definitely providessome degree of cleaning, but it requires a long-lasting treatment and the final DMFV maintainsunsatisfactorily high levels.
Chemical cleaning based on acids has been at-tempted as well. Cylinders have been treatedfor different durations with several kinds of acids,namely • H SO • H SO
8M / H O • HNO • oxalic acid (H C O ),9 a) (b)(c) (d)Figure 7: ToF-SIMS Chemical maps of cartridge cleanedby abrasion. The image (a) shows the total ion image, the(b) is the Al map, the (c) is the Si map, and the (d) is theFe map. The data bars measure 100 µ m. Iron contamina-tion persists, but no microparticles appear. Distributed Siand Al contamination is evident all over the surface. Notice-ably, the near-edge regions are more heavily contaminatedby Al, likely due to corundum residues, implanted where thetumbling is more effective. These cylinders have then been analysed bothmagnetometrically and by XRFS or ToF-SIMS. Assummarized in Table 1, the magnetometric mea-surements showed evident yet unsatisfactory decon-tamination levels, with the only exception of theoxalic acid. Oxalic acid exhibited the best capacityto remove Fe from the polymeric surface, consis-tently with several previous studies available in theliterature [39, 40, 41].An overnight treatment in oxalic acid associ-ated with sonication removes the contamination be-low the ToF-SIMS detection level and reduces theDMFV down to ± treatment time DMFVreduction H SO
8M 3h 50%H SO / H O
10% 3h 35%HNO C O Table 1: Signals from different PEEK cylinders beforeand after treatments of different duration. The only acidtreatment leading to satisfying results is with oxalic acid(H C O ).(a) (b) (c)Figure 8: Iron maps of PEEK disks lathed by ordinary HSStools to fit the ToF-SIMS holder; the data bars measure 100 µ m. The maps are recorded without any purification (a),after 1 hour sonication in 0.3M oxalic acid (b), and afteran overnight treatment (c). In the last case a reduction ofthe iron contamination down to the detection threshold isachieved. similar results are obtained with Ti machined car-tridges.The fact that the two techniques produce similarresidual DMFV reinforces the hypothesis formu-lated in Sec.4 that this persistent level is most likelydue to a volume contamination.In this respect, for comparison, we have mea-sured the DMFV in cylinders made of other dielec-tric materials, observing smaller -however, still welldetectable- values. In particular, an Acrylonitrile-butadiene-styrene (ABS) cylinder produced by a 3Dextrusion printer and an Acetale cylinder lathedwith amagnetic tool show about a ±
800 pT and ±
6. Conclusion
The problem of residual permanent magneti-zation of polymeric sample containers used in aremote-detection ULF-NMR experiment has beenstudied by means of magnetometric measurementand chemical analyses of the polymer surface.10ome degree of magnetic remanence has beenpointed out, and has been attributed to polymersurface contamination occurring during the con-tainer production, and (at a much lower extent, asto have negligible consequences in our applications)to bulk contamination of the material. The latterevidence emerges, at comparable levels, also withother polymeric materials. Consistently, presenceof contaminants in extruded polymers has been re-cently observed [42].Despite the demonstrated existence of ferromag-netic behaviours in polymers [43], in this case theresidual bulk ferromagnetism is attributable to di-luted contaminants (at ppm concentration), dis-persed in the volume during the machining stages(extrusion) of the production of the PEEK rods. Itcan be matter of micro-particles dispersed in thevolume, which are not detected by the ToF-SIMSanalyses of clean samples, due to the extremely lowprobability of appearing in the small surface por-tions analysed or to nano-particles in extremely di-lute concentration, which appear in ToF-SIMS closeto the detectable level.The cartridge magnetization causes spuriousfields both outside and inside the cartridges. Thesefields are evaluated magnetometrically both di-rectly as a static dipolar field and via their effectson the contained NMR sample. Complementaryinformation is inferred from XRFS and ToF-SIMSanalyses.Several approaches have been proposed to coun-teract the ferromagnetic contamination and theireffectiveness has been tested, by means of DC-magnetometry, ULF-NMR, XRFS and ToF-SIMS.Good results have been obtained by using non-magnetic tools at the machining stage. Thismethod poses some limitations in the cartridge con-struction. It reduces the ferromagnetic contamina-tion to a level that is yet detectable, but eventu-ally meets the requirements for accurate ULF-NMRspectroscopy.Several post-production cleaning procedures havebeen tested, based on mechanical or chemical ap-proaches.Positive results have been obtained with sand-paper polishing (not applicable to thread and innersurfaces of the cartridges) and (to an unsatisfactorylevel) using tumbling techniques. Similarly, an ap-preciable contaminant reduction is observed withchemical treatments based on several strongly re-active acids, but the effect results insufficient evenafter long-lasting procedures. Excellent results are instead obtained withovernight oxalic acid treatment associated with son-ication. In this case, the residual average re-manence is the same as that achieved with non-magnetic machining, and there is an evidence thatsuch level is attributable to ferromagnetic contam-inants dispersed in the polymer volume at a subppm concentration.
7. Acknowledgments
The authors are pleased to thank Dr. TommasoLisini from SirsLab at DIISM, for providing the 3Dprinted samples, and Dr. Sophie Versavaud - Vic-trex Technical Support for the kind and effectiveassistance.
8. Compliance with ethical standardsConflict of Interest
The authors declare thatno conflict of interest exists, concerning the findingsand the results presented in this paper.
References [1] M. C. Wapler, J. Leupold, I. Dragonu, D. von Elver-feld, M. Zaitsev, U. Wallrabe, Magnetic properties ofmaterials for MR engineering, micro-MR and beyond,Journal of Magnetic Resonance 242 (2014) 233 – 242.[2] F. D. Doty, G. Entzminger, Y. A. Yang, Magnetism inhigh-resolution NMR probe design. I: General methods,Concepts in Magnetic Resonance 10 (3) (1998) 133–156.[3] L. H. Bennett, P. S. Wang, M. J. Donahue, Artifactsin magnetic resonance imaging from metals, Journal ofApplied Physics 79 (8) (1996) 4712–4714.[4] M. Mößle, S.-I. Han, W. R. Myers, S.-K. Lee, N. Kelso,M. Hatridge, A. Pines, J. Clarke, SQUID-detectedmicro-Tesla MRI in the presence of metal, Journal ofMagnetic Resonance 179 (1) (2006) 146 – 151.[5] V. Zampetoulas, D. J. Lurie, L. M. Broche, Correc-tion of environmental magnetic fields for the acquisitionof nuclear magnetic relaxation dispersion profiles belowearth’s field, Journal of Magnetic Resonance 282 (2017)38 – 46.[6] G. Bevilacqua, V. Biancalana, Y. Dancheva, L. Moi,Chapter three - Optical atomic magnetometry for ultra-low-field NMR detection, in: G. A. Webb (Ed.), An-nual Reports on NMR Spectroscopy, Vol. 78 of AnnualReports on NMR Spectroscopy, Academic Press, 2013,Ch. 3, pp. 103 – 148.[7] M. C. D. Tayler, T. Theis, T. F. Sjolander, J. W. Blan-chard, A. Kentner, S. Pustelny, A. Pines, D. Budker,Invited review article: Instrumentation for nuclear mag-netic resonance in zero and ultralow magnetic field, Re-view of Scientific Instruments 88 (9) (2017) 091101.
8] D. A. Barskiy, M. C. Tayler, I. Marco-Rius,J. Kurhanewicz, D. B. Vigneron, S. Cikrikci, A. Ay-dogdu, M. Reh, A. N. Pravdivtsev, J.-B. Hövener, et al.,Zero-field nuclear magnetic resonance of chemicallyexchanging systems, Nature communications 10 (1)(2019) 1–9.[9] G. Bevilacqua, V. Biancalana, Y. Dancheva, L. Stiac-cini, A. Vigilante, Spurious ferromagnetic remanencedetected by hybrid magnetometer, Review of ScientificInstruments 90 (4) (2019) 046106.[10] M. C. D. Tayler, J. Ward-Williams, L. F. Gladden,Ultralow-field nuclear magnetic resonance of liquidsconfined in ferromagnetic and paramagnetic materials,Applied Physics Letters 115 (7) (2019) 072409.[11] G. Bevilacqua, V. Biancalana, Y. Dancheva, A. Vigi-lante, A. Donati, C. Rossi, Simultaneous detection of Hand D NMR signals in a micro-Tesla field, The Journalof Physical Chemistry Letters 8 (2017) 6176–6179.[12] G. Bevilacqua, V. Biancalana, Y. Dancheva, A. Vig-ilante, Restoring narrow linewidth to a gradient-broadened magnetic resonance by inhomogeneous dress-ing, Phys. Rev. Applied 11 (2019) 024049.[13] G. Bevilacqua, V. Biancalana, Y. Dancheva, A. Vig-ilante, Sub-millimetric ultra-low-field MRI detectedin situ by a dressed atomic magnetometer, AppliedPhysics Letters 115 (17) (2019) 174102.[14] M. Sepioni, R. R. Nair, I.-L. Tsai, A. K. Geim,I. V. Grigorieva, Revealing common artifacts due toferromagnetic inclusions in highly oriented pyrolyticgraphite, EPL (Europhysics Letters) 97 (4) (2012)47001.[15] Y. Wang, X. Chen, L. Li, A. Shalimov, W. Tong,S. Prucnal, F. Munnik, Z. Yang, W. Skorupa, M. Helm,S. Zhou, Structural and magnetic properties of irradi-ated SiC, Journal of Applied Physics 115 (17) (2014).[16] K. Shinji, N. Nozomi, T. Koki, M. Masanori, B. Yoshio,O. Krzysztof, T. Dietl, Origin and control of high-temperature ferromagnetism in semiconductors, NatureMaterials 6 (2007) 440–446.[17] G. Zhao, P. Beeli, Observation of an ultrahigh-temperature ferromagnetic-like transition in iron-contaminated multiwalled carbon nanotube mats, Phys.Rev. B 77 (2008) 245433.[18] M. Muluaka, Assessment of lung particle accumulationin factory workers by magnetic field measurement, In-ternational Journal Occupation Medicine and Environ-mental Health (2003) 209–213.[19] L. M. C. Pereira, J. P. Araújo, M. J. V. Bael, K. Temst,A. Vantomme, Practical limits for detection of ferro-magnetism using highly sensitive magnetometry tech-niques, Journal of Physics D: Applied Physics 44 (21)(2011) 215001.[20] K. Halbach, Design of permanent multipole magnetswith oriented rare earth cobalt material, Nuclear In-struments and Methods 169 (1) (1980) 1 – 10.[21] G. Bevilacqua, V. Biancalana, P. Chessa, Y. Dancheva,Multichannel optical atomic magnetometer operatingin unshielded environment, Applied Physics B 122 (4)(2016) 103.[22] V. Biancalana, Y. Dancheva, L. Stiaccini, Note: A fastpneumatic sample-shuttle with attenuated shocks, Re-view of Scientific Instruments 85 (3) (2014).[23] V. Biancalana, G. Bevilacqua, P. Chessa, Y. Dancheva,R. Cecchi, L. Stiaccini, A low noise modular currentsource for stable magnetic field control, Review of Sci- entific Instruments 88 (3) (2017) 035107.[24] J. Belfi, G. Bevilacqua, V. Biancalana, R. Cecchi,Y. Dancheva, L. Moi, Stray magnetic field compen-sation with a scalar atomic magnetometer, Review ofScientific Instruments 81 (6) (2010) 065103.[25] G. Bevilacqua, V. Biancalana, Y. Dancheva, A. Vig-ilante, Self-adaptive loop for external-disturbance re-duction in a differential measurement setup, Phys. Rev.Applied 11 (2019) 014029.[26] W. E. Bell, A. L. Bloom, Optically driven spin preces-sion, Phys. Rev. Lett. 6 (1961) 280–281.[27] G. Bevilacqua, V. Biancalana, A. Ben Amar Baranga,Y. Dancheva, C. Rossi, Micro-Tesla NMR J-couplingspectroscopy with an unshielded atomic magnetometer,Journal of Magnetic Resonance 263 (2016) 65–70.[28] D. Budker, M. Romalis, Optical magnetometry, NaturePhysics (2007) 227–234.[29] G. Bevilacqua, V. Biancalana, Y. Dancheva, L. Moi,All-optical magnetometry for NMR detection in amicro-Tesla field and unshielded environment, Journalof Magnetic Resonance 201 (2009) 222–229.[30] L. Zhang, J. Gao, L. N. W. Damoah, D. G. Robertson,Removal of iron from aluminum: A review, MineralProcessing and Extractive Metallurgy Review 33 (2)(2012) 99–157.[31] S. M. Kurtz, J. N. Devine, PEEK biomaterials intrauma, orthopedic, and spinal implants, Biomaterials28 (2007) 4845–4869.[32] M. Bertocco, C. Offelli, D. Petri, Analysis of dampedsinusoidal signals via a frequency-domain interpolationalgorithm, IEEE Transactions on Instrumentation andMeasurement 43 (2) (1994) 245–250.[33] I. Yoshida, T. Sugai, S. Tani, M. Motegi, K. Minamida,H. Hayakawa, Automation of internal friction measure-ment apparatus of inverted torsion pendulum type,Journal of Physics E Scientific Instruments 14 (1981)1201–1206.[34] G. Leone, A. D. Vita, M. Consumi, G. Tamasi,C. Bonechi, A. Donati, C. Rossi, A. Magnani, Compari-son of original and modern mortars at the herculaneumarchaeological site, Conservation and Management ofArchaeological Sites 21 (2) (2019) 92–112.[35] G. Leone, M. Consumi, S. Lamponi, A. Magnani, Com-bination of static time of flight secondary ion massspectrometry and infrared reflection-adsorption spec-troscopy for the characterisation of a four steps built-upcarbohydrate array, Applied Surface Science 258 (17)(2012) 6302 – 6315.[36] A. Caciagli, R. J. Baars, A. P. Philipse, B. W. Kuipers,Exact expression for the magnetic field of a finite cylin-der with arbitrary uniform magnetization, Journal ofMagnetism and Magnetic Materials 456 (2018) 423 –432.[37] V. E. GmbH, Technical Service - Victrex Technical Sup-port, Private communication (2018).[38] Victrex, Victrex Finishing Brochure (2016).URL [39] D. Panias, M. Taxiarchou, I. Paspaliaris, A. Kontopou-los, Mechanisms of dissolution of iron oxides in aqueousoxalic acid solutions, Hydrometallurgy 42 (2) (1996) 257– 265.[40] S. O. Lee, T. Tran, B. H. Jung, S. J. Kim, M. J. Kim,Dissolution of iron oxide using oxalic acid, Hydromet-allurgy 87 (3) (2007) 91 – 99.
41] R. Salmimies, M. Mannila, J. Kallas, A. Häkkinen,Acidic dissolution of magnetite: Experimental study onthe effects of acid concentration and temperature, Claysand Clay Minerals 59 (2) (2011) 136–146.[42] M. F. Franco, R. Gadioli, M. A. De Paoli, Presence ofiron in polymers extruded with corrosive contaminantsor abrasive fillers, Polímeros 29 (2019).[43] J. Mahmood, J. Park, D. Shin, H.-J. Choi, J.-M. Seo, J.-W. Yoo, J.-B. Baek, Organic ferromagnetism: Trappingspins in the glassy state of an organic network structure,Chem 4 (10) (2018) 2357–2369.41] R. Salmimies, M. Mannila, J. Kallas, A. Häkkinen,Acidic dissolution of magnetite: Experimental study onthe effects of acid concentration and temperature, Claysand Clay Minerals 59 (2) (2011) 136–146.[42] M. F. Franco, R. Gadioli, M. A. De Paoli, Presence ofiron in polymers extruded with corrosive contaminantsor abrasive fillers, Polímeros 29 (2019).[43] J. Mahmood, J. Park, D. Shin, H.-J. Choi, J.-M. Seo, J.-W. Yoo, J.-B. Baek, Organic ferromagnetism: Trappingspins in the glassy state of an organic network structure,Chem 4 (10) (2018) 2357–2369.