A

dwetSugwP

K

1

orac[cdoXTsoqd

h

0d

Available online at www.sciencedirect.com

Journal of Electron Spectroscopy and Related Phenomena 162 (2008) 49–55

A simple method for determining linear polarization andenergy calibration of focused soft X-ray beams

B. Watts ∗, H. AdeDepartment of Physics, North Carolina State University, Raleigh, NC 27695, USA

Received 12 April 2007; received in revised form 27 August 2007; accepted 28 August 2007Available online 1 September 2007

bstract

Although critical to quantitative linear dichroism studies of molecular orientation, the degree of linear polarization of focused soft X-ray beamselivered by X-ray microscopes has not been previously measured. Here, we present a scaled-down version of a recently developed technique inhich the �∗ near edge X-ray absorption fine structure (NEXAFS) resonance of highly oriented pyrolytic graphite (HOPG) is used to probe the

lectric field intensity in each direction and hence deduce the degree of linear polarization of the incident X-ray beam. Applying this technique tohe soft X-ray microscope at beamline 5.3.2 of the Advanced Light Source in Berkeley, CA, yielded a measured value of 79 ± 11%, for the firsttokes parameter of 0.79 ± 0.11 or as a Stohr P factor of 0. 89 ± 0.06. It is expected that the error margin could be significantly reduced via the

se of an in-vacuum rotation actuator. We have also calibrated the energy of the graphite exciton to be 291.65 ± 0.025 eV, improving the utility ofraphite as an energy calibration standard for NEXAFS and allowing the convenience of both energy calibration and polarization determinationith a single inexpensive sample.ublished by Elsevier B.V.

PG;

irtintnbfiSb

c

eywords: NEXAFS; Linear dichroism; Polarization; Soft X-ray; Graphite; HO

. Introduction

To date, a number of linear dichroism studies of molecularrientation have been carried out using scanning transmission X-ay spectromicroscopes (STXMs) that included materials suchs Kevlar [1,2], linear alkane crystals [3,4], polyethylene [5],rystals of polycyclic aromatic hydrocarbons [6] and phenyl urea7], thermoplastic/liquid crystalline polymer blends [8], block-opolymer, coke [9] and silk [10]. The number of X-ray linearichroism applications has been much smaller than the numberf applications that seek to quantify composition with near edge-ray absorption fine structure (NEXAFS) microscopy [11,12].his is due to a number of factors: (i) the intrinsic number of pos-ible applications, (ii) the non-linearity and non-orthogonality

f the STXMs imaging systems available [1,2,13,14] had madeuantitative analysis of two orthogonal sample orientationsifficult, and (iii) the uncertainty in the degree of linear polar-

∗ Corresponding author.E-mail addresses: benjamin.watts@gmail.com (B. Watts),

wade@ncsu.edu (H. Ade).

Tstnltco

368-2048/$ – see front matter. Published by Elsevier B.V.oi:10.1016/j.elspec.2007.08.008

Carbon; Synchrotron radiation; X-ray microscopy; Photon energy calibration

zation of the focused soft X-ray beam of a STXM and itseproducibility and stability over time also degraded quanti-ative analysis capabilities and decreased productivity. Recentnstrument developments based on laser interferometers haveow vastly improved the linearity and orthogonality of STXMechnology [14]. Furthermore, STXM is a technology that isow implemented at a number of synchrotron facilities atending magnets and undulators. The need to reliably per-orm quantitative dichroism measurement remains and witht, the need to measure the degree of linear polarization of aTXM, particularly those that are operated at bending magneteamlines.

The zone plate optical element utilized in a STXM compli-ates the use of these instruments for dichroic measurements.he zone plate is usually overfilled (see Fig. 1) in order totabilize the transmitted flux and therefore also acts as an aper-ure, spatially truncating the beam. The incident beam is thusot necessarily representative of the transmitted beam and the

ocation of the zone plate relative to the beam greatly affectshe polarization of the transmitted beam. Hence, inevitablehanges in alignment over time lead to changes in the degreef polarization, greatly reducing accuracy and reproducibil-

50 B. Watts, H. Ade / Journal of Electron Spectroscopy and Related Phenomena 162 (2008) 49–55

moun

ifumppmesbeqs

�ntm

•

•

•

P

Mscippm

pNsftiuweea

2

m(wga

Fig. 1. Schematic detailing the

ty. Measurement of the polarization in STXM is challengingor two reasons. Firstly, as discussed above, measurementspstream of the zone plate are not representative of the trans-itted beam, and secondly, the short focal distance of the zone

late requires that the order selecting aperture (OSA) and sam-le be situated within a short distance of the order of a fewillimeters, providing little space for the inclusion of additional

quipment. Thus, conventional macroscopic methods for mea-uring the degree of polarization of a soft X-ray beam cannote easily applied to soft X-ray microscopes. However, knowl-dge of the degree of linear polarization is critical for fullyuantifying the degree of molecular alignment observed in aample.

A recently developed polarimetry technique [15] utilizes the∗ near edge X-ray absorption fine structure (NEXAFS) reso-ance of highly oriented pyrolytic graphite (HOPG) to deducehe degree of linear polarization of the incident X-ray beam. This

ethod may be summarized in three steps:

Firstly, NEXAFS spectra of the sample HOPG crystal arerecorded at normal X-ray incidence, which provide a measureof non-ideality of the sample crystal. The principle behind thismeasurement lies in the fact that in a perfect HOPG crystal,all of the C 1s → �∗ transition moment vectors are alignedwith the surface normal and hence are perfectly orthogo-nal to the electric field vector of a normally incident X-raybeam, resulting in zero NEXAFS resonance intensity. There-fore, any NEXAFS resonance intensity observed near 285eV in a normal incidence geometry can be entirely attributedto defects and contaminants in the HOPG crystal. Note thatsince the resonance intensity is proportional to the squaredsine of the angle between the X-ray beam axis and the crys-tal surface normal (or more precisely, the squared cosine ofthe angle between the electric field vector of the X-ray beamand the transition moment vector [16]), the normal incidencemeasurement is not very sensitive to small errors in sampleorientation.Second, the HOPG crystal is mounted, at some fixed tilt angle,on an azimuthally rotatable stage so that the crystal’s surfacenormal can be precessed about the X-ray beam axis, as shown

in Fig. 1. Recording NEXAFS spectra at a range of azimuthalangles then reveals a trend in the �∗ NEXAFS resonanceintensity, showing maxima and minima when the crystal nor-mal is aligned, respectively parallel and perpendicular to thesynchrotron plane.

niflt

ting of samples in the STXM.

Lastly, the �∗ NEXAFS resonance intensity trend withazimuthal angle is plotted against the squared sine of theazimuthal angle to produce a straightline trend, as demon-strated in Fig. 8. From this plot, we can determine the gradient,m, and intercept, b, of the trend which, when combined withthe resonance intensity observed at normal X-ray incidence,I⊥, can calculate the Stohr polarization factor [16], P, usingthe equation:

= [(m/b − I⊥)] + 1

[(m/b − I⊥)] + 2(1)

easurements performed at a number of beamlines demon-trated that the linear polarization of bending magnet beamlinesan vary significantly, especially due to design and alignmentssues, and demonstrate the need to experimentally determineolarization. Here, we present a scaled-down version of thisolarimeter experiment and use the adapted technique to deter-ine the degree of linear polarization of a STXM instrument.Furthermore, we will also accurately determine the energy

osition of the HOPG exciton resonance peak (observed inEXAFS spectra near 291 eV) by comparison of the NEXAFS

pectra of HOPG and CO2 gas. HOPG is a common standardor calibrating the photon energy scale of NEXAFS spectra, butypically the wider and more asymmetric �∗ peak near 285 eVs used. Use of the more narrow exciton peak will improve thetility of HOPG as a secondary energy standard. In combinationith the polarimetry methods described in this work, a precise

nergy standard for HOPG spectra enables the simultaneousvaluation of both the energy scale and linear polarization ofsoft X-ray beam.

. Experiment

Initial samples were prepared by rubbing a TEM grid (1000esh; Gilder Grids, Grantham, UK) with a piece of HOPG

GRBS/1.2; NanoTech America, Allen, TX). While the methodas found to also work with grids of other sizes, coarser TEMrids were found to result in graphite sections oriented at strongngles to the plane of the mesh so that the crystal quality couldot be accurately assessed (i.e. difficult to achieve normal X-ray

ncidence) and the grid bars of finer TEM grids tended to inter-ere with the measurement. Rubbing with HOPG tends to leave aot of graphite stuck to the TEM grid, but much of it is either toohick for the transmission of soft X-rays or highly crumpled and

scopy

tioartt

gascmw[vnreomit

eAnbefaH

tmdscSruor

Fth

B. Watts, H. Ade / Journal of Electron Spectro

herefore of far too poor crystal quality to be of use in the exper-ment. Different rubbing techniques produced varying amountsf graphite on the TEM grid, however the creation of appropri-tely thin and flat graphite sections appears to be an essentiallyandom occurrence. Excessive rubbing does not tend to improvehe chance of finding an appropriate graphite section since theyend to be fragile and are easily destroyed by successive strokes.

Finding appropriately thin and flat sections of graphite on therids is relatively simple with an optical microscope. Appropri-te thickness can be assessed by the transparency of the graphiteection (when illuminated in transmission) and crystal qualityan be roughly assessed by how uniform it appears when illu-inated in transmission and reflection. Promising candidatesere then introduced into the beamline 5.3.2 STXM instrument

14] at the Advanced Light Source, CA for further inspectionia images at 285.3 eV (corresponding to the HOPG �∗ reso-ance) and 292.3 eV (corresponding to maximum C 1s → �∗esonance) as well as line-spectra covering the full C K-edgenergy range (270–340 eV). Fig. 2 shows representative images

f a thin graphite section viewed with both optical and soft X-rayicroscopes. As discussed in Section 1, any NEXAFS resonance

ntensity observed near 285 eV indicates that the sample crys-al is either not ideal (i.e. contains defects, contaminant species,

mta

ig. 2. Images of the graphite section in (A) optical reflection, (B) optical transmiransmission after TEM grid deformation (tilt angle, θ, of approximately 50◦). The imorizontal line appearing in (D), but not (C), is the result of radiation damage from an

and Related Phenomena 162 (2008) 49–55 51

tc.) or is incorrectly oriented (i.e. not quite at normal incidence).nother possible source of normal incidence NEXAFS reso-ance intensity is significant divergence of the probing X-rayeam. However, the current 5.3.2 STXM zone plate has a diam-ter of 155 �m and focal length of 1225 �m and therefore theocused X-rays are less than 3.63◦ off-axis, which corresponds tomaximum 0.4% increase in �∗ intensity for normal incidenceOPG spectra.The TEM grids holding the most promising graphite sec-

ions were then mounted on a STXM sample plate with rotatableounts (to allow rotation in the final measurement step), intro-

uced into the STXM and the identified high-quality graphiteections were analyzed with line spectra. These line spectraonstitute the normal incidence (I⊥) measurement discussed inection 1 and it is important to acquire spectra of sufficientesolution and statistics for any graphite sections selected forse in the polarization measurement because the following stepf deforming the TEM grid will make it extremely difficult toe-acquire a normal incidence orientation afterwards.

As discussed in Section 1, this method of linear polarizationeasurement requires that the graphite planes be oriented such

hat the surface normal is tilted away from the incident beamxis by a significant angle. To this end, the TEM grid in the area

ssion, (C) X-ray transmission (292.3 eV) at normal incidence and (D) X-rayage widths are coincident with the TEM grid bar spacing of 19 �m. The dark,excessively long duration line-scan.

52 B. Watts, H. Ade / Journal of Electron Spectroscopy and Related Phenomena 162 (2008) 49–55

-prec

odatntcgsWoipeiFs

ttboe6oemtSA

osAccCn

eoeTdactl

3. Results and discussion

All graphite spectra presented here were recorded from thegraphite section illustrated in Fig. 2, using the STXM instrument

Fig. 3. Schematic detailing adaptations to the microscope allowing high

f the graphite section was tilted by about 50◦ by tearing andeforming the grid around the graphite section to create a flappproximately 100 � m wide. The cutting and deformation ofhe TEM grid was performed by pressing the sharp point of aeedle through the grid in a square pattern that created a flaphat could then be bent out of the plane of the TEM grid. Pre-ise control over the positioning of the needle relative to therid was achieved by mounting the needle onto the condensertage of a Leica DM R optical microscope (Leica Microsystems,etzlar GmbH) as shown in Fig. 3. When viewed through the

ptical microscope under rear illumination, with the sample gridn focus, the needle position was clearly observable through theosition and clarity of the needle’s shadow as viewed through theyepiece. Translation of the sample allowed for relative position-ng of the fixed-position needle at any point on the grid surface.ig. 4 shows the deformed grid holding the graphite sectionhown in Fig. 2.

The sample TEM grid and rotatable STXM mount plate werehen introduced to the beamline 5.3.2 STXM instrument [14] athe Advanced Light Source, California, as detailed in Fig. 1. Car-on K-edge spectra of the graphite section shown in Fig. 2 werebtained at azimuthal angles over a range of about 130◦. Thentrance, dispersive exit and non-dispersive exit slits were set to0 �m, 30 �m and 30 �m, respectively, corresponding to normalperation of the 5.3.2 STXM. Note that only the non-dispersivexit slit impacts the vertical acceptance angle of the bendingagnet source and hence the polarization of the accepted radia-

ion. Since the sample mount was limited to manual rotation, theTXM chamber was vented between each recorded spectrum.ll spectra were normalized as discussed by Watts et al. [17].Further spectra were acquired in the presence of a few mTorr

f CO2 gas in order to accurately calibrate the photon energycale of the HOPG spectra. Since the sharpest peak in the NEX-FS spectrum of graphite, the exciton peak near 291 eV, is

overed by the large C O �∗ resonance of CO2, the energyalibration required the acquisition of separate I0, graphite andO2 spectra as an interlaced series to allow isolation of the reso-ance peaks while ensuring against a possible drift in the photon

Fa

ision cutting and bending of the TEM grid around the graphite section.

nergy between spectra. This is a modified version of a previ-usly described STXM energy calibration method successfullymployed to calibrate NEXAFS spectra of polymers [18,19].he energy calibration spectra were collected with the entrance,ispersive exit and non-dispersive exit slits set to 25 �m, 10 �mnd 60 �m, respectively, which was previously determined toorrespond to a photon energy resolution of 0.075 eV throughhe Gaussian part of Voight line-shapes fitted the vibrationalines of the CO2 carbon K-edge spectrum [20].

ig. 4. Optical light image of a deformed TEM grid, with the flap bent to a tiltngle, θ, of about 50◦. As in Fig. 2, the 6 �m grid bars are spaced 19 �m apart.

B. Watts, H. Ade / Journal of Electron Spectroscopy and Related Phenomena 162 (2008) 49–55 53

Fos

ozitttotwa

atnraapdsdcmfs�bipbtmpAap

Fig. 6. Overlay of measured HOPG spectrum (red), showing the �∗peak, andthe measured spectrum of CO2 gas. The spectra were energy calibrated usingliterature values for the CO2 peak positions [22] indicated with vertical lines.A(r

sr

2TasThe slope and intercept of the trend in Fig. 8 were then deter-mined and used to calculate the linear polarization of the X-raybeam, via Eq. (1), as 79 ± 11% (Stohr P factor of 0.89 ± 0.06or Stokes parameter of 0.79 ± 0.11). The 11% error quoted

∗

ig. 5. HOPG spectra recorded in the ALS-5.3.2 STXM instrument at a varietyf azimuthal angles, φ (and fixed tilt angle, except for the normal incidencepectrum).

n ALS beamline 5.3.2. The different intensity levels of the largeones seen in Fig. 2(C) and (D) were found to indicate variationsn sample thickness only, since no variation in spectral fea-ures were observed. While such variations do not interfere withhe polarization measurement, measurements were restricted tohe large, thicker zone to the right of centre. The dark spotsbserved in these images were found to vary in spectral features,herefore representing contaminant material that could interfereith the polarization measurement and hence were carefully

voided.NEXAFS spectra measured at a range of azimuthal angles

re shown in Fig. 5. As expected for the HOPG carbon K-edge,hese spectra show strongly angle-dependent �∗ and �∗ reso-ance peaks at photon energies of 285.3 eV and 292–310 eV,espectively. The sharp exciton resonance peak near 291.5 eVnd broad EXAFS oscillations in the higher energy region arelso characteristic of the HOPG NEXAFS spectrum. The smalleaks observed near 288 eV suggest a small amount of oxi-ation of the graphite [21], which is not extensive enough toignificantly influence the polarization measurement. Further, asiscussed in the previous polarimetry article [15], the spectralontamination of the oxidation would only affect the polarizationeasurement if it created bonds that were aligned in some pre-

erred direction, which is unlikely in highly symmetric materialsuch as HOPG. Note that the black spectrum with the weakest∗ resonance intensity was measured at normal X-ray incidenceefore the supporting TEM grid was deformed. Further, utiliz-ng the spectra in Fig. 6 and reference values for the spectraleaks of CO2 [22], the position of the HOPG exciton peak haseen determined to be 291.65 ± 0.025 eV, in agreement withhe value previously reported by Bruhwiler et al. [23]. This new

easurement of the exciton peak position represents the most

recise calibration to date of the often referenced graphite NEX-FS spectrum. Bruhwiler et al. did not quote any precision to

ccompany their exciton peak position measurement, merely ahoton energy resolution of 0.3 eV was quoted. For compari-

Fcaai

wider energy range showing the full C K-edge spectra is displayed inset.For interpretation of the references to colour in this figure legend, the reader iseferred to the web version of the article.)

on the present measurement was made with a photon energyesolution of 0.075 eV.

Plotting the measured values of �∗ resonance intensity at85.3 eV against φ produce a sinusoidal trend, as shown in Fig. 7.his data was fitted to obtain the azimuthal offset, φ′ = 22◦,nd the same �∗ resonance intensity data then plotted againstin2(φ − φ′) to demonstrate a linear trend, as shown in Fig. 8.

ig. 7. Plot of the normalized HOPG C 1s → � resonance peak intensity (bluerosses) and a fitted sine squared function (solid black line) against azimuthalngle, φ. The angular position of the maximum intensity gives φ′ = 22◦, thengular scale offset. The dotted line across the bottom represents the resonancentensity at normal X-ray incidence.

54 B. Watts, H. Ade / Journal of Electron Spectroscop

Fig. 8. Plot of the normalized HOPG C 1s → �∗ resonance peak intensity (bluecla

f(eoaot

mbrbbthawrro

4

dmmrHoe5dsw

Siamasm

mstcci

A

EDC

R

[

[

[[

[

[

[[

rosses) and a straight line fit (solid black line) against sin2(φ − φ′). The dashedines represent the 50% confidence level of the straight line fit and the dotted linecross the bottom represents the resonance intensity at normal X-ray incidence.

or this measurement is derived from the 50% confidence levelassuming a normal distribution) of the least squares fit to thexperimental data shown in Fig. 8. Note that this measurementnly probes the linear polarization of the delivered X-ray beamnd does not measure the circular polarization (though the designf the ALS beamline 5.3.2 makes significant circular polariza-ion unlikely).

Conducting the experiment on a manually rotated sampleount introduced many complications. Firstly, the STXM cham-

er had to be vented and the mount plate removed in order tootate the sample, then re-introduce it and pump down the cham-er again. This meant that the sample position and focus had toe re-acquired, which was more onerous than usual owing tohe fact that the unusual depth of the sample increased the likeli-ood of collision with the OSA. Utilizing an in-vacuum rotationctuator, such as that described by Hernandez-Cruz et al. [24],ould circumvent these issues, significantly reducing the time

equired to conduct the measurement and increasing its accu-acy through improved sample stability and a greater utilizationf beamtime for improved statistics.

. Conclusions

A recently reported method for the determination of theegree of linear polarization of synchrotron radiation has beeniniaturized and adapted for use in STXM instruments. Theethod uses the highly aligned C 1s → �∗ transition vector

esonance from a thin graphite section, produced by rubbingOPG against a TEM grid, to probe the polarization statef the focused X-ray beam near 285 eV. The degree of lin-ar polarization of the focused X-ray beam delivered by the

.3.2 STXM instrument at the ALS in Berkeley, California wasetermined to be 79 ± 11%. The large, 11% error of this mea-urement is thought to originate from complications associatedith being limited to manually rotating the sample outside the

[

[

[

y and Related Phenomena 162 (2008) 49–55

TXM chamber and it is expected that the incorporation of ann-vacuum rotation actuator could significantly reduce this errors well as the amount of beamtime required to perform theeasurement. This method could be feasibly implemented asroutine polarimetry measurement for quantitative soft X-ray

pectromicroscopic studies of molecular orientation and align-ent.Also, the position of the graphite exciton peak has been deter-

ined to be 291.65 ± 0.025 eV by comparison to a CO2 gastandard. This precise measurement of the narrow spectral fea-ure makes it an excellent secondary reference for energy scalealibration and furthermore allows the convenience of energyalibration and linear polarization determination via a singlenexpensive sample.

cknowledgements

Work at NCSU is supported by the U.S. Department ofnergy DE-FG02-98ER45737. The ALS is supported by theirector of the Office of Science, Department of Energy, underontract No. DE-AC02-05CH11231.

eferences

[1] H. Ade, B. Hsiao, Science 262 (1993) 1427–1429.[2] A.P. Smith, H. Ade, Appl. Phys. Lett. 69 (1996) 3833–3835.[3] J.X. Fu, S.G. Urquhart, J. Phys. Chem. A 109 (2005) 11724.[4] Y. Zou, T. Araki, G. Appel, A.L.D. Kilcoyne, H. Ade, Chem. Phys. Lett.

430 (2006) 287–292.[5] Y. Wang, M. Rafailovich, J. Sokolov, D. Gersappe, T. Araki, Y. Zou, A.L.D.

Kilcoyne, H. Ade, G. Marom, A. Lustiger, Phys. Rev. Lett. 96 (2006)028303.

[6] G.D. Cody, H. Ade, S. Wirick, J. Waldbauer, C-XANES linear dichroismstudies of single crystals of polycyclic aromatic hydrocarbons, NationalSynchrotron Light Source Activity Report, 1997, p. B28.

[7] L.M. Croll, J.F. Britten, C. Morin, A.P. Hitchcock, H.D.H. Stover, J. Syn-chrotron Rad. 10 (3) (2003) 265–268.

[8] A.P. Smith, C. Bai, H. Ade, R.J. Spontak, C.M. Balik, C.C. Koch, Macro-mol. Rapid Commun. 19 (1998) 557–561.

[9] G.D. Cody, Soft X-ray imaging and C-XANES linear dichroism of nematicphase transitions in coal derived liquids, National Synchrotron Light SourceActivity Report, 1997, p. B27.

10] D. Hernandez-Cruz, M.E. Rousseau, M.M. West, M. Pezolet, A.P. Hitch-cock, Biomacromolecules 7 (2006) 836–843.

11] H. Ade, X. Zhang, S. Cameron, C. Costello, J. Kirz, S. Williams, Science258 (1992) 972–975.

12] H. Ade, Trends Polym. Sci. 5 (2) (1997) 58–66.13] C. Jacobsen, S. Wirick, G. Flynn, C. Zimba, J. Microsc. 197 (2) (2000)

173–184.14] A.L.D. Kilcoyne, T. Tyliszczak, W.F. Steele, S. Fakra, P. Hitchcock,

K. Franck, E. Anderson, B. Harteneck, E.G. Rightor, G. Mitchell, A.P.Hitchcock, L. Yang, T. Warwick, H. Ade, J. Syncrotron Rad. 10 (2003)125–136.

15] B. Watts, L. Thomsen, P.C. Dastoor, J. Electron Spectrosc. Relat. Phenom.151 (2005) 208–214.

16] J. Stohr, NEXAFS Spectroscopy, Springer, Heidelberg, 1992.17] B. Watts, L. Thomsen, P.C. Dastoor, J. Electron Spectrosc. Relat. Phenom.

151 (2005) 105–120.

18] H. Ade, A.P. Smith, H. Zhang, G.R. Zhuang, J. Kirz, E. Rightor, A. Hitch-

cock, J. Electron Spectrosc. Relat. Phenom. 84 (1997) 53–72.19] O. Dhez, H. Ade, S. Urquhart, J. Electron Spectrosc. Relat. Phenom. 128

(2003) 85–96.20] T. Tyliszczak, unpublished data.

scopy

[

[

B. Watts, H. Ade / Journal of Electron Spectro

21] A. Kuznetsova, I. Popova, J.T. Yates, M.J. Bronikowski, C.B. Huffman, J.Liu, R.E. Smalley, H.H. Hwu, J.G. Chen Jr., J. Am. Chem. Soc. 123 (2001)10699–10704.

22] Y. Ma, C.T. Chen, G. Meigs, K. Randall, F. Sette, Phys. Rev. A 44 (3)(1991) 1848–1858.

[

[

and Related Phenomena 162 (2008) 49–55 55

23] P.A. Bruhwiler, A.J. Maxwell, C. Puglia, A. Nilsson, S. Andersson, N.Martensson, Phys. Rev. Lett. 74 (4) (1995) 614–617.

24] D. Hernandez-Cruz, A.P. Hitchcock, T. Tyliszczak, M.E. Rousseau, M.Pezolet, Rev. Sci. Instrum. 78 (2007) 033703.