1043

Contributed Paper

Development of Fast Charge Exchange Recombination

Spectroscopy by Using Interference Filter Method

in JT-60U

KOBAYASHI Shinji1), SAKASAI Akira2), KOIDE Yoshihiko2), SAKAMOTO Yoshiteru2), KAMADA Yutaka2),

HATAE Takaki2), OYAMA Naoyuki2) and MIURA Yukitoshi2)

1) Institute of Advanced Energy, Kyoto Univ. Gokasho, Uji 611-0011, Japan2) Naka Fusion Research Establishment, Japan Atomic Energy Research Institute, Ibaraki 311-0193, Japan

(Received 26 December 2002 / Accepted 17 August 2003)

Abstract

Recent developments and results of fast charge exchange recombination spectroscopy (CXRS) usinginterference filter method are reported. In order to measure the rapid change of the ion temperature androtation velocity under collapse or transition phenomena with high-time resolution, two types ofinterference filter systems were applied to the CXRS diagnostics on the JT-60U Tokamak. One candetermine the Doppler broadening and Doppler shift of the CXR emission using three interference filtershaving slightly different center wavelengths. A rapid estimation method of the temperature and rotationvelocity without non-linear least square fitting is presented. The modification of the three-filters systemenables us to improve the minimum time resolution up to 0.8 ms, which is better than that of 16.7 ms forthe conventional CXRS system using the CCD detector in JT-60U. The other system having sevenwavelength channels is newly fabricated to crosscheck the results obtained by the three-filters assembly,that is, to verify that the CXR emission forms a Gaussian profile under collapse phenomena. In a H-modedischarge having giant edge localized modes, the results obtained by the two systems are compared. Theapplicability of the three-filters system to the measurement of rapid changes in temperature and rotationvelocity is demonstrated.

Keywords:

charge exchange recombination spectroscopy, interference filter spectroscopy,fast ion temperature measurement

J. Plasma Fusion Res. Vol.79, No.10 (2003) 1043 - 1050

1. Introduction

Measurements of plasma pressure, rotation velocityand electric field play a key role in understanding theproperties of the improved confinement or themechanism of the collapse phenomenon in fusionplasma, for example, the transition of H-mode,formation of internal transport barrier (ITB), minor

collapse, and disruption. Since these events have timescales of some hundreds of microseconds, theirdiagnostic systems are required to have time resolutionsufficient to observe the evolution of such fastphenomena.

Charge exchange recombination spectroscopy(CXRS) has been developed for the measurements of

author’s e-mail: kobayashi@iae.kyoto-u.ac.jp

This article is based on the invited talk at the 19th Annual Meetings of JSPF (Nov. 2002, Inuyama).

Journal of Plasma and Fusion Research　Vol.79, No.10　October 2003

1044

ion temperature, toroidal/poloidal rotation velocity,impurity density, and radial electric field [1-5]. The iontemperature and rotation velocity have been evaluatedbased on the Doppler broadening and Doppler shift ofspectral lines from the charge exchange reactionsbetween highly ionized impurity ions (Az+) andenergetic hydrogen atoms of the neutral beam (H0) [6],as follows:

H0 + Az+ → H+ + H(z–1)+(n , l ) , (1)

where (n, l) designates an excited state with subsequentdecay with photon emission. The diagnostic system ofCXRS has been based on the combination of amonochromator and a charge coupled device (CCD) forthe time- and space- resolved observation.

Also in the JT-60U Tokamak [7], the multichordalCXRS diagnostic has been developed using a Czerny-Turner monochromator equipped with a CCD detector[5]. The minimum spatial resolution of the system is 0.8cm in the edge region, which is useful in addressing theelectric field or thermal diffusivity for understanding thetransport physics and for analysis of the pedestalstructure. On the other hand, the CCD’s 16.7 msminimum time resolution is barely adequate to observethe evolution of such fast phenomena.

In order to investigate the physical mechanism ofthe transition or collapse phenomena, somedevelopments in the fast measurement by CXRS havebeen attempted in fusion devices, for example,combinations of the Czerny-Turner monochromator withfast CCD detectors [8,9] or the Fabry-Perot spectrometerwith photodiode array [10]. In JT-60U, we have used aspectroscopic system employing interference filters forthe CXRS diagnostics [11]. The system, having threewavelength channels with slightly different centerwavelengths, can determine the Doppler broadening andDoppler shift of the CXR emission. The interferencefilter spectroscopy is one of the simplest ways totransmit a well-defined band of light, thus this methodhas often been used in the Thomson scatteringpolychromator [12] or for the measurement of themotional Stark effect [13].

In this study, we report the development of thefilter CXRS system focusing on the following subjects:1) modification of the filter CXRS system, 2) rapidcalculation method of the ion temperature and rotationvelocity, and 3) the applicability of the filter CXRSsystem to the measurement of rapid changes intemperature and rotation velocity.

Because of the poor signal-to-noise (S/N) ratio of

the obtained intensity of the CXR emission, theaccuracy of the ion temperature and rotation velocitydeduced by the three-filters system has been insufficientto analyze the energy transport or determine the radialelectric field. Moreover, the minimum integration timehas been limited to 1.6 ms. Therefore, we modified theoptical system for the improvement of the minimumtime resolution. A fast analytical scheme for the CXRSmeasurement has been needed for the real-timemeasurement and feedback control of the iontemperature and rotation velocity for advanced tokamakoperations or burning plasma experiments [14,15]. Thepossibility of a feedback control based on neural netanalysis of the spectra has been proposed in Ref. [8].The three-filters system, on the other hand, has anability to determine the temperature and rotationvelocity without non-linear least square fitting. Thus, wepresent an analytical scheme for rapidly calculating thetemperature and rotation velocity, which assumes thatthe spectral profile forms a Maxwellian distribution. Theassumption used in the analytical method, however, mayaffect the accuracy of the temperature and rotationmeasurements under the collapse phenomenon wherethe possibility of the non-Maxwellian distribution of thespectra could not be excluded, because the assembly hasminimum measurement points in the spectrum todetermine the temperature and rotation. It is, then,necessary to check the validity of the results under thefast phenomena. In order to solve the problem, wefabricated a new filter CXRS system having sevenwavelength channels. This system has an ability tomeasure the detailed line spectrum under such fastphenomena as edge localized mode (ELM) or minorcollapse. The result observed in a H-mode dischargewith ELMs is compared with that of the three-filterssystem.

2. Experimental Apparatus

2.1 Three-filters and seven-filters

assemblies

The fast CXRS system having three wavelengthchannels was designed to be compact in size, to usesmall optics (25 mm diameter), and to be easy to alignand maintain. A schematic view of the assembly isshown in Fig. 1(a), consisting of a fiber-optic bundle, acollimation lens, beam splitting mirrors, and threeFabry-Perot interference filters mounted in front of thehigh-sensitivity photomultipliers (HAMAMATSU: R-1104). The collimation lens is used to make a parallellight, and one-third and half-mirrors split the light to the

Contributed Paper Development of Fast CXRS Using Interference Filter Method in JT-60U S. Kobayashi et al.

1045

different photomultipliers. The output signal of thephotomultiplier is transferred to an analog-digitalconverter (ADC) of 12 bit accuracy through a pre-amplifier. In order to reduce the loss of the incidentlight, the alignment between the collection optics andthe photomultiplier was optimized because the firstdynode of the photomultiplier was not on its axis. Thesemodifications yielded an improvement in the outputintensity by 60%. A lowpass filter using the Fast FourierTransform and its inverse technique is applied toeliminate the high-frequency noise component.

The passband of each filter was used from 0.2 nmto 1.0 nm at full width at half maximum (FWHM),

which depended on the ion temperature and rotationspeed to be measured. Moreover, the center wavelengthof the three filters was determined so that the differenceof that of the adjacent filter is equivalent to thebandwidth at FWHM, because it widened the dynamicrange of the temperature and rotation. For measurementof the spectral line of CVI (n = 8 → 7 : 529.05 nm) in atemperature range around 1 keV, the differencesbetween the center wavelengths of the three filters,λCWL, and the rest wavelength, λ0, ∆λCWL (i.e. λCWL –λ0) was selected to be – 0.2, 0.0, and 0.2 nm having abandwidth at FWHM of 0.2 nm, respectively.

For more detailed measurement of the spectralprofile, we fabricated a new filter CXRS system havingseven wavelength channels. Figure 1(b) illustrates aschematic diagram of the seven-filters system. Theincident light is cascaded between the interference filtersusing mirrors and lenses to reduce the loss of the light.Each filter has a 0.3 nm passband at FWHM in the caseof measurement in the CVI line (529.05 nm). Thedifferences in the wavelength ∆λCWL are also shown inthe figure.

The CXR spectrum induced by the neutral beamshould be separated from that of the cold component bymeans of electron excitation and/or the charge exchangereaction of background thermal neutrals in the plasma.We adopted two ways of separating the cold componentby means of twelve assemblies of the three-filterssystem. One is simultaneous observation using off-beamobjective optics. Half of the twelve assemblies are usedfor measurements of the neutral beam. The other sixassemblies are used for only the cold component. Thedetailed explanation for the objective optics is presentedin Ref. [5]. The second method is using a shotaccompanying with the probe beam breakdown, whichenables us to measure the ion temperature and rotationvelocity at 12 maximum spatial points.2.2 Calibration experiment

The spectral transmission of the Fabry-Perot filterused in the fast CXRS systems is affected by theincident angle of the light to the filter and the ambienttemperature. A shortening of the center wavelength by0.1 nm results from a misalignment of the incident angleby 1 degree. The wavelength shift also appears when thefilter is used with non-collimated light. The centerwavelength of the filter corresponding to 40 km/s of therotation velocity shifts due to a 3 °C change in theambient temperature. The ambient temperature,therefore, should be kept constant during theexperiment.

Fig. 1 Schematic view of the fast CXRS systems, (a)three-filters and (b) seven-filters assemblies. Thedifferences between the center wavelength ofeach filter and the rest wavelength are shown inthe figures.

Journal of Plasma and Fusion Research　Vol.79, No.10　October 2003

1046

For the purpose of accurate measurement of the iontemperature and rotation velocity, we carried out acalibration experiment to obtain the spectral sensitivityof the three-filters assembly. The setup of the calibrationexperiment is illustrated in Fig. 2(a). A monochromator(JASCO : CT-50) whose spectral resolution was set tobe 0.05 nm was used, a tungsten lamp being employedas a light source. The spectral sensitivity of eachchannel was measured by scanning the wavelength ofthe monochromator using a stepping motor. Thereference wavelength was determined by the mercurylamp. Figure 2(b) shows an experimental result of thespectral sensitivity of the three-filters assembly. Theseven-filters assembly was also calibrated using thesame experimental setup.

2.3 Analytical method for three-filters

assembly

The analytical method for the determination of thetemperature and rotation velocity used in the three-filters assembly is as follows. The measured intensity ofthe CXR emission of i-th channel Ii is represented by

I i ∝ ∫ ηi T i (λ ) f (λ ) dλ , (2)

where ηiTi(λ) is the spectral sensitivity of i-th channelincluding the interference filter and photomultiplier. Inthis formula, we assume that the spectral profile of theCXR emission f (λ) forms a Maxwellian distributionfunction. Then, the intensity ratios, which are defined asthe ratios of the intensity at the central wavelengthchannel to that of the shorter wavelength channel I1/I2

and to the longer wavelength channel I3 / I2, depend onthe temperature and rotation. Prior to the experiment, wecalculated the relation between the intensity ratios andthe ion temperature and rotation velocity. Figure 3shows the relation in the case of the filter combinationshown in Fig. 2(a). The solid lines represent thecontours for the ion temperature of 0.2, 1.0, 1.8, and 2.6keV, respectively, and the dotted lines are those for therotation velocity with a contour interval of 50 km/s.When the intensity ratios are measured based on adischarge experiment, the temperature and rotationvelocity can be determined using the relation. Forexample, if the intensity ratios I1/I2 and I3 /I2 aremeasured to be 0.76 and 0.44, the ion temperature androtation speed are 1 keV and –50 km/s, respectively.

Fig. 2 (a) A setup for the calibration experiment and (b)the spectral sensitivity of the three-filtersassembly.

Fig. 3 Contour plot for the ion temperature and rotationvelocity as a funcitions of intensity ratios I1/I2 andI3/I2.

Contributed Paper Development of Fast CXRS Using Interference Filter Method in JT-60U S. Kobayashi et al.

1047

Finally, the ion temperature Ti and rotation velocity Vr

are summarized as a function of the intensity ratios, asfollows:

T i = 10 ,

Vr = bi jΣi , j = 0

4

log I1 / I2

ilog I3 / I2

j, (3)

where aij and bij are the coefficients for the fittingfunctions, respectively. Because the temperature androtation can be estimated without complicated nonlinearleast squares fitting, we are planning for the feedbackcontrol of the ion temperature profile on the JT-60Udischarge using the real time processor (RTP) system[14-16].

The detectable range of the temperature androtation velocity depend on the filter passband, centerwavelength, and the accuracy of the intensity ratio. Inthe case of the combination of three filters shown in Fig.2(b), for example, the detectable range of the iontemperature is from 0.2 keV to 2.6 keV, whichcorresponds to the Doppler broadening from 2/3 to 2.5of ∆λFWHM. For the rotation velocity, the range from-150 km/s to 150 km/s can be measured, that is, the shift

of the CVI line can be detected between ∆λCWL of theshorter wavelength channel and that of the longerwavelength channel.

3. Results

3.1 Improved three-filters CXRS system

An example of the rapid time response of the three-filters assembly is shown in Figs. 4(a)-(c), where thetime evolution of the stored energy, Dα intensity, andintensities of the CVI line are shown during a minorcollapse in reversed shear discharge on the JT-60UTokamak (a toroidal magnetic field of 3.7 T and aplasma current of 1.5 MA). The differences inwavelength of the three filters from the rest wavelengthare also shown in the figure. The locations ofmeasurement by two sets of the three-filters assemblyare at the core region, ρ ∼ 0.3 (ρ : normalized minorradius), and the peripheral region (ρ ∼ 0.85). Tointensify the incident light, two chords which wereobserved at the same location were bundled together.The time integration of these data was 0.8 ms. Thebackground component of the CVI line, measured usingthe objective optic viewing off the neutral beam, wassubtracted for extracting the spectrum by the CXRemission at the beamline. As a result of a minor

Fig. 4 An example of the rapid change in a minor collapse of the reversed shear discharge in JT-60U. Temporal profilesof (a) stored energy and Dα intensity, the line spectra of CVI at (b) the core region (ρ ∼ 0.3) and (c) the edge (ρ ∼0.85) region by the three-filters assembly and (d)-(g) deduced temperature and rotation velocity at the twolocations.

a i jΣi , j = 0

4

log I1 / I2

ilog I3 / I2

j

Journal of Plasma and Fusion Research　Vol.79, No.10　October 2003

1048

collapse, a sudden drop in the stored energy and a burstof the Dα intensity was observed. At the core region of ρ∼ 0.3, a burst of the intensity near the centralwavelength was also observed. This phenomenonsuggests a narrowness of the Doppler broadening. Theincrease in the intensity at the longer wavelengthchannel indicates the Doppler shift of the CVI line whosedirection is the same as that of the plasma current.

Figures 4(d)-(g) show the time evolution of the CVI

ion temperature and toroidal rotation velocity at twolocations. The rotation velocity is not converted into thatfor the bulk ion. The integration time for the iontemperature and rotation velocity deduced by the three-filters assembly (bold lines) was 0.8 ms. In this case, theintegration time of the CCD system (squares) was 50 mssince three spectra were summed up to improve the S/Nratio. The results of the three-filters assembly werealmost consistent with that provided by the CCDsystem, except for the timing of the minor collapse. Theion temperature at ρ ∼ 0.3 decreased from 3.5 keV to 1keV simultaneously with the occurrence of the minorcollapse. The toroidal rotation velocity at ρ ∼ 0.3changed from –120 km/s to – 40 km/s across thecollapse. The rotation in the peripheral region (ρ ∼ 0.85)shows similar behavior. Under this condition, nosignificant slowing down in the toroidal rotation at theperipheral region was observed before the minorcollapse. Although the modification of the three-filtersassembly improved in the minimum integration time bya factor of 2, the scatter of the temperature and rotationremained. The relationship between estimated error andthe S/N ratio is discussed in Sec. 4. The development ofthe fast CXRS system using the three-filters assemblyenabled us to measure the fast phenomena in the JT-60Uplasma that couldn’t be detected by the standard CXRSsystem with the CCD detector.3.2 Seven-filters assembly and comparison

with the three-filters system

In this subsection, we describe the result ofmeasurement of the CXR spectrum using the seven-filters system in a H-mode plasma having giant (Type I)ELMs. The result of the three-filters system is thencompared with those obtained by the seven-filtersassembly.

Figures 5(a) and (b) plot the time evolution of theintensities of the divertor Dα and CVI line of the seven-filters CXRS system, respectively. In this discharge, theseven-filters system used a chord of the poloidalobjective optics viewing the separatrix whose spatialresolution is 1.5 cm and integration time was 2.08 ms.

The probe beam was turned off at t = 8.92 s. As can beseen by comparing with the Dα signal, the edge plasmavariation during the ELM modulated both the Dα andthe CVI signal. Figure 6 shows the line spectra at 5.4 ms,1.2 ms before, and 2.9 ms after the ELM event at t =8.887 s. In this figure, the background component wassubtracted referring the relative time of the Dα burst ateach ELM event. Unfortunately, the result of thecalibration experiment revealed that the spectraltransmission of the channel at ∆λCWL = – 0.52 nm hasdiffered from the designed Gaussian shape. Hence, theintensity of this channel is not plotted in Fig. 6. Thehorizontal error bars represent the passband of the filterat FWHM. Results of the Gaussian fittings for three

Fig. 5 Temporal evolution of (a) Dα intensity and (b) CVI

line intensity obtained by the seven-filters systemin a H-mode plasma having giant ELMs.

Fig. 6 Line spectral profiles of CVI obtained by the seven-filters system at 5.4 ms (�), 1.2 ms (™) before, and2.9 ms (�) after an ELM event.

Contributed Paper Development of Fast CXRS Using Interference Filter Method in JT-60U S. Kobayashi et al.

1049

spectral profiles are also shown in the figure. It wasfound that the line spectrum broadened slightly justbefore the Dα burst (solid line with open circles).

The comparison of the results deduced by the three-and seven- filters systems is shown in Fig. 7 togetherwith the time evolution of the ion temperature. Thetemperature determined by the seven-filters system wasobtained using the Gaussian fitting of the spectra,assuming that the instrumental response function wasrepresented by a Gaussian function with the width of thefilter passband. In order to improve the S/N ratio of thethree-filters assembly, viewing at 1 cm inside theseparatrix, the temporal profile of the CXR emission forseven giant ELMs was piled up based on the time of theDα bursts. About 3 ms before the Dα bursts, a slightincrease in the ion temperature of the seven-filterssystem was observed. Then, the ion temperaturerecovered 2 ms after the burst. This temporal behavioracross the ELM events was also similar to that seen inthe three-filters assembly. The inset of Fig. 7 shows theradial profile of the ion temperature under a non-collapse phase (t = 8.889 s) obtained by the filtersystems and the conventional CCD system. The resultsfrom two filters assemblies are almost consistent withthose of the CCD system. Moreover, as shown in Fig. 6,no clear difference in the spectra between themeasurement and the Gaussian fittings was observedunder the time scale of a few milliseconds, whichsuggested the validity of the assumption of a Gaussianline shape of the radiated emission. These results

demonstrate that the three-filters system can be appliedto many other experiments in which the rapid changes intemperature and rotation velocity are expected even inthe collapse phenomena.

Unfortunately, the ion temperature in the wholepedestal region was not measured using the fast CXRSsystems; however, the pedestal ion pressure structure onthis fast time scale is interesting and such measurementcan help further our understanding of the physics ofpedestal formation.

4. Discussion

This section describes the measurement error of thethree-filters assembly. The uncertainty of thetemperature and rotation, as shown in Fig. 4, is causedby the weak intensity of the detected charge exchangeemission and the insufficient S/N ratio. The noisecomponent is due mainly to the dark current of thephotomultiplier and the scatter of the number of incidentphotons. In the three-filters CXRS system, a smalldegree of ambiguity of the intensity ratio creates arelatively large error in calculating the temperature androtation because of the minimum measurement point inspectrum. We estimated the relationship of the error ofthe deduced ion temperature and rotation velocity to theS/N ratio of the output signal. The S/N ratio of theoutput signal before the collapse shown in Fig. 4 was tobe about 4, which corresponded to the estimated errorsof 30% in temperature and 25 km/s in rotation. Anotherambiguity regarding the estimations of temperature androtation is caused by the non-Gaussian profile of theemission such as two components of the iontemperature. In the case of a calculation having a tailcomponent of three percent of the density and threetimes higher in temperature than the bulk ion, theoverestimation of temperature will be less than 4%.Therefore, this effect is considered to be negligible.

In order to reduce the temperature and rotationerrors deduced by the three-filters system down to 10%and 10 km/s, respectively, an improvement in the S/Nratio by factor of 2.5 is required. Note that the S/N ratioof the photomultiplier output is almost proportional tothe square root of the intensity of the incident light. Weare, then, planning further modifications of the opticalsystem for the filter assembly.

5. Summary

The filter spectroscopy was applied to CXRS in JT-60U for the purpose of measuring the ion temperatureand rotation velocity with high-time resolution. The

Fig. 7 Temporal profile of the ion temperature deducedby the seven-filters assembly (�), three-filtersassembly (�), and conventional CCD system (™),The measured locations of the three- and seven-filters assemblies are shown in the inset togetherwith the radial profile of the ion temperaturededuced from the CCD system.

Journal of Plasma and Fusion Research　Vol.79, No.10　October 2003

1050

modification of the three-filters CXRS system improvedthe minimum integration time up to 0.8 ms, whichenabled us to measure such a rapid time response as theminor collapse that couldn’t be detected by the standardJT-60U CXRS system with the CCD detector. Bycomparing the results obtained by the conventionalCXRS and those by the seven-filters systems, weconfirmed the applicability of the three-filters system tomeasurements of rapid changes in the ion temperatureand rotation velocity.

Acknowledgments

The authors acknowledge the members of the JapanAtomic Energy Research Institute (JAERI) who havecontributed to the JT-60U projects. The authors wouldlike to thank Dr. Ushigusa of JAERI for his discussionsand suggestions. This work was supported by acollaboration program between JAERI and KyotoUniversity.

References

[1] R.J. Fonck, D.S. Darrow and K.P. Jaehnig, Phys.Rev. A 29, 3288 (1984).

[2] R.P. Seraydarian, K.H. Burrell, N.H. Brooks et al.,Rev. Sci. Instrum. 57, 155 (1986).

[3] K. Ida, S. Kado and Y. Liang, Rev. Sci. Instrum.71, 2360 (2000).

[4] P. Gohil, K.H. Burrell, R.J. Groebner et al., Proc.of Fourteenth IEEE/NPSS Symposium on FusionTechnology (IEEE, Princeton, NJ, 1992), Vol. II,

p.1199.[5] Y. Koide, A. Sakasai, Y. Sakamoto et al., Rev. Sci.

Instrum. 72, 119 (2001).[6] R.C. Isler, Phys. Rev. Lett. 38, 1359 (1977).[7] Y. Miura and the JT-60 Team, Phys. Plasmas 10,

1809 (2003).[8] D.M. Thomas, K.H. Burrell, R.J. Groebner et al.,

Rev. Sci. Instrum. 68, 1233 (1997).[9] K.H. Burrell, D.H. Kaplan, P. Gohil et al., Rev.

Sci. Instrum. 72, 1028 (2001).[10] K. Ida, J. Xu, K.N. Sato et al., Fusion Eng. Des.

34-35, 219 (1997).[11] J.S. Koog, A. Sakasai, Y. Koide et al., Rev. Sci.

Instrum. 70, 372 (1999).[12] T.N. Carlstrom, G.L. Cambell, J.C. DeBoo et al.,

Rev. Sci. Instrum. 63, 4901 (1992).[13] T. Fujita, H. Kubo, T. Sugie et al., Fusion Eng.

Des. 34-35, 289 (1997).[14] Y. Kawano et al., Proc. 30th EPS Conf. On

Controlled Fusion and Plasma Physics (St.Petersburg, 2003), vol 27B in press.

[15] Y. Kusama, “Requirements for Diagnostics inControlling Advanced Tokamak Modes”, inAdvanced Diagnostics for Magnetic and InertialFusion (ed. by P. Stott et al., Kluwer Academic/Plenum Publishers, New York 2002) p31.

[16] S. Sakata, M. Koiwa, T. Aoyagi and T. Matsuda,JAERI, JAERI-Tech 2000-043 July 2000 [inJapanese].