Extensions to the charge exchange recombination spectroscopy diagnostic suite at ASDEX Upgrade

Extension of the core CXRS in AUG AIP/123-QED Extensions to the charge exchange recombination spectroscopy diagnostic suite at ASDEX Upgrade R. M. McDermott,1, a) A. Lebschy,1, 2 B. Geiger,3C. Bruhn,1, 2 M. Cavedon,1M. Dunne,1R. Dux,1R. Fischer,1A. Kappatou,1T. P¨utterich,1E. Viezzer,4and the ASDEX Upgrade Team5, b) 1)Max-Planck-Institut f¨ur Plasmaphysik Boltzmannstr. 2, 85748 Garching, Germany 2)Physik-Department E28, Technische Universit¨at M¨unchen, D-85748 Garching, Germany 3)Max-Planck-Institut f¨ur Plasmaphysik,Wendelsteinstr. 1, 17491 Greifswald, Germany 4)Dept. of Atomic, Molecular and Nuclear Physics, University of Seville, Avda. Reina Mercedes, 41012 Seville, Spain 5)Max-Plank-Institut f¨ur Plasmaphysik, D-85748, Garching, Germany (Dated: 11 July 2017) 1 Extension of the core CXRS in AUG A new core charge exchange recombination spectroscopy (CXRS) diagnostic has been installed in the ASDEX Upgrade tokamak that is capable of measuring the impurity ion temperature, toroidal rotation, and density on both the low field side (LFS) and high field side (HFS) of the plasma. The new system features 48 lines-of-sight (LOS) with a radial resolution that varies from ±2cm on the LFS down to ±0.75cm on the HFS and has sufficient signal to run routinely at 10ms and for special circumstances down to 2.5ms integration time. The LFS-HFS ion temperature profiles provide an additional constraint on the magnetic equilibrium reconstruction and the toroidal rotation frequency profiles are of sufficiently high quality that information on the poloidal velocity can be extracted from the LFS-HFS asymmetry. The diagnostic LOS are coupled to two flexible-wavelength spectrometers such that complete LFSHFS profiles from two separate impurities can be imaged simultaneously, albeit with reduced radial coverage. More frequently, the systems measure the same impurity providing very detailed information on the chosen species. Care has been taken to calibrate the systems as accurately as possible and to include in the data analysis any effects that could lead to spurious temperatures or rotations. PACS numbers: 52.55.-s, 52.70.-m, 07.60.Rd, 42.79.e Keywords: Tokamak, plasma diagnostics, optical systems, spectroscopy, CXRS, ion temperature, poloidal rotation, toroidal rotation, impurity density, neutral attenuation, charge exchange cross-sections, electric field a)Electronic mail: Rac[email protected] b)For authors’ list, see H. Zohm et al., NF 55 104010 (2015) 2 Extension of the core CXRS in AUG I. INTRODUCTION Charge exchange recombination spectroscopy (CXRS) is a routine diagnostic on most fusion experiments to measure the temperature, rotation and density of impurity ions (typically low-Z) in the plasma1,2. The technique relies on the transfer of electrons from neutral particles to impurity ions, which subsequently de-excite and emit characteristic line radiation (the measured CX spectra). In most fusion plasmas, due to the high temperature and densities, the natural population of neutral particles is extremely low making the charge exchange signal negligibly small. However, if a population of neutral particles is introduced into the plasma via a gas puff, a neutral beam, or a pellet then the charge exchange signal can become sufficiently large to diagnose the impurity temperature and rotation from the Doppler width and shift of the measured emission line and the impurity density from the measured intensity coupled with knowledge of the neutral particle population. In addition, the injection of a population of neutral particles yields the added benefit of localizing the CXRS measurements to the intersection of the lines-of-sight (LOS) with the cloud of neutral particles. Regardless of the source, the number of neutral particles decreases with distance into the plasma due to ionisation and charge exchange processes and the measured CXRS signal decreases accordingly. For this reason, and often for geometry and hardware considerations, CXRS systems in fusion devices have traditionally focused on measurements in the outer half of the plasma, in tokamaks the so-called low field side (LFS), and relied on the assumption that these parameters are constant on magnetic flux surfaces to map them to other locations in the plasma. This was the situation, for example, at ASDEX Upgrade (AUG) until recently3. Similar to systems existing at TCV4and DIII-D5, a new core CXRS diagnostic capable of measuring across the magnetic axis of the machine and providing both LFS and high field side (HFS) measurements was implemented at ASDEX Upgrade for the 2014 experimental campaign. This system is capable of providing up to 48 point CXRS profiles across the mid-plane of the plasma from the pedestal top on the LFS to the pedestal top on the HFS with high temporal and radial resolution. This is possible due to a combination of improved optical design of the in-vessel system, low f-number spectrometers, and high neutral beam energy (93 keV). With this system, impurity ion temperature, rotation and density profiles 3 Extension of the core CXRS in AUG can be obtained at two different positions on a magnetic flux surface, thus enabling the study of poloidal asymmetries in these quantities and the determination of the poloidal flow in the core plasma4–6. This paper is organized as follows: In section II the design, geometry, and hardware of the new LFS-HFS CXRS diagnostics are presented. The details of the calibrations of these systems are given in III. Section IV demonstrates the capabilities of the new systems, showing examples of the measured spectra and deduced impurity ion temperature, rotation and impurity density profiles. Charge exchange cross-section effects and the effects of other atomic processes on the measured CXRS spectra are discussed in section V and, finally, a summary of this work is presented in section VI. II. NEW CXRS DIAGNOSTICS AUG is equipped with two neutral beam boxes separated toroidally by 180 degrees. Each box is comprised of four sources capable of delivering 2.5MW of power each (in D) for a total of 20MW of available neutral beam injection (NBI) power. The beam boxes have maximum acceleration voltages of 60 keV (box I, sources 1-4) and 93 keV (box II, sources 5- 8) and the majority of the pre-existing CXRS diagnostics utilise source 3 of NBI box I. This has the advantage that the inclusion of just one NBI source in a plasma discharge enables measurements from the complete suite of CXRS diagnostics. However, this has occasionally been problematic as CXRS measurements were not available when beam box I was not in operation due to technical reasons or the constraints of a given experiment. For this reason, it was decided that the new CXRS diagnostics would view one of the sources from beam box II. In addition, NBI box II provides a second advantage with respect to box I. Namely, the higher energy beam suffers less from attenuation leading to improved CXRS signals in the plasma core and on the HFS. While both neutral beam boxes deliver the same power, the sources in box II inject fewer neutral particles in comparison to box I, due to the differences in acceleration voltage. This would lead to a reduction in the CXRS signal, except the lower particle numbers are very well compensated (to within ∼10%) by the increase in the CXRS cross-sections, which peak around 100keV (∼50keV/amu) for most of the standard CXRS transitions (e.g. B n=7-6, C n=8-7) . Therefore, the CXRS systems viewing the 60keV and 4 Extension of the core CXRS in AUG -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 16123 4 5 6 7 8 9 10 11 14 15 12 13 NBI BOX II Sources 5-8 NBI BOX I Sources 1-4 CMR: NBI S3 (edge CXRS) CER: NBI S3 (LFS CXRS) COR: NBI S8 (LFS/HFS CXRS) 4.0 X (m) Y (m) 1.00 1.25 1.50 1.75 2.00 2.25 2.50 -1.0 -0.5 0.0 0.5 1.0 R (m) Z (m) NBI 2/3 NBI 1/4 NBI 5/8 NBI 6/7 NBI 8 NBI 5 NBI 7 NBI 6 COR 1 COR 2 COR 3 Magnetic axis FIG. 1. Top-down (left) and poloidal (right) cross-sections of AUG showing the geometries of the NBI sources as well as a subset of the LOS of the core and edge toroidal CXRS diagnostics. For clarity, only sources from NBI box II and a subset of the LOS from the new diagnostics are shown in the poloidal projection. The equilibrium is taken from discharge 33096 at 2s. 93KeV NBI sources see very similar intensities at the plasma LFS edge. With distance into the core, however, the higher energy of box II provides an advantage. The geometries of the four neutral beam sources in NBI box II are shown in Fig. 1. Sources 5 and 8 are injected radially with small, symmetric vertical tilts above and below the mid-plane of the machine. Sources 6 and 7 are injected tangentially significantly above and below the mid-plane. A standard magnetic equilibrium of an AUG plasma is overplotted in light blue. From the poloidal cross-section one can see that the center point of the plasma is (often) positioned at slightly positive Z-values making NBI 8 the only NBI source of box II that can support CXRS measurements in the plasma core. Taking this into account, three new optical heads comprising the COR diagnostic have been installed between sectors 5 and 6 and are aligned such that the entire path-length of NBI 8 in the plasma volume can be imaged. A subset of the LOS from these optical heads is shown in Fig. 1. Fig. 2 shows the design drawings as well as the manufactured optical 5 Extension of the core CXRS in AUG COR 1 COR 2 COR 3 Magnetic shutter FIG. 2. Three new optical heads have been installed in the AUG vacuum vessel between sectors 5 and 6 behind the adjacent limiter shadow. A photograph of the optical heads in shown on the left, while on the right techincal drawings for these heads are shown. heads installed in the vessel. COR 1 is equipped with 20 fibre optics and images the LFS of the plasma from a major radius of 1.89m to 2.13m. The second optical head, COR 2, with 23 lines of sight, covers the plasma centre from 1.57m to 1.86m and is shifted vertically downward (away from the centre of source 8) by several centimetres in order to obtain better coverage across the magnetic axis of the plasma. Lastly, COR 3 has 27 fibres and covers the HFS from R=1.2m to 1.54m. Together these three heads provide 70 measurement positions across the full minor radius of the plasma (pedestal top to pedestal top) with a channel to channel separation of roughly 1.5 cm. The optical heads employ a fairly simple design in which a single aspherical lens with a 50mm diameter and 80mm focal length (F/1.6) is used to focus the light from the plasma onto an array of 400µ(F/2.2) silica-clad-silica fiber optics arranged at the focal distance behind the lens. The aspherical lenses were found to have improved imaging properties with respect to similar plano-convex lenses resulting in sharper images and improved radial resolution. A comparison of the images produced using the two lenses is shown in Fig. 3. Here, diodes were used to measure the light intensity across the diameter of the image. The vertical axis shows the measured current in the diode. The radial region from which light 6 Extension of the core CXRS in AUG distance from center [cm] -4 -2 0 2 4 10-1 100 101 102 103 current [nA] Fit spherical lens Fit aspherical lens Spherical lens Aspherical lens FIG. 3. Comparison of spot-size using a spherical and aspherical lens for the new CXRS diagnostic design. The aspherical lenses improve the focus and reduces the effective spot-size to 1.5cm. is collected is significantly reduced when using the aspherical lenses leading to an effective spot-size of 1.5 cm. In the vessel, the lens of each head is protected from glow discharges and boronizations by a magnetic shutter actuated by a piece of iron welded to each lens cover and oriented such that when a toroidal magnetic field is present the shutters are pulled open and otherwise remain closed. The orientation was chosen such that the shutters will open for both forward and reversed field magnetic configurations. These shutters, visible in Fig. 2, were also tested in the lab prior to installation to ensure that they would open even for the lowest toroidal magnetic fields employed in the AUG tokamak (∼1T). The LOS geometries of all of the fibers in the optical heads are determined to high accuracy by back-lighting the fibre optics and measuring the intersection coordinates of the LOS with planar surfaces (specifically installed for this purpose) with a robotic arm and a three-dimensional model of the AUG vacuum vessel. Multiple positions are measured along each LOS enabling the geometry and optical head positions to be determined to an accuracy of ∼1mm. Here, the “optical head” position refers to the position at which all of the LOS from a given head converge. This information is then used to determine the measurement positions and spatial resolution as well as the angles of the LOS to the magnetic field lines and to the NBI trajectories. The measured charge exchange intensities are proportional to the sum of the neutral particle populations integrated along the LOS, weighted by the corresponding effective emission rate, see Eqn. 1. 7 Extension of the core CXRS in AUG LCX,Z (λ)∝nZX nX j hσn,j,Z,λvjieff ZLOS n0,n,j (l)dl (1) Here, L is the measured charge exchange intensity for a particular element, Z, and transition wavelength, λ, and nZis the impurity density, which is assumed to be constant along the intersection of the LOS and the neutral beam. The summations on nand jare over the principle quantum number of the donor neutral and the neutral beam energy components, respectively, and hσn,j,Z,λvjieff is the appropriate effective emission rate with σthe charge exchange cross-section and vjthe collision velocity. The effective position of the CXRS measurement is then given by the center of mass of the total distribution along the LOS as a function of Rand Z. For the AUG beams, the neutral distributions of the different energy components and excited states have slightly different divergences along the beam. However, the differences are quite small and the CXRS intensity (for NBI Box 2) is dominated by the ground state of the first energy component. Therefore, the effective measurement positions are typically calculated using only the neutral distribution from this component of the beam. RCOM =RLOS RLOSn0,0,0(l)dl RLOS n0,0,0(l)dl (2) This has also been compared to calculations using all energy components and excited states and been found to be the same well within the error bars on the calculations. Please note that for the purposes of these calculations n0,n,j is modelled using a beam attenuation code, which requires a magnetic equilibrium in order to map the necessary plasma profiles (ne,Te,Ti) onto the LOS positions. In all cases a standard AUG magnetic equilibrium has been used. In principle, the determined measurement positions can change slightly for different plasma shapes, positions, and parameters. However, due to AUG’s relatively limited shaping capability, these corrections are also well within the error bars for the vast majority of AUG plasmas. Small changes in the three dimensional model for the neutral beam trajectories used for the calculations, however, can make a significant impact on the calculated measurement positions. The beam model used in this work has been constrained by beam emission spectroscopy (BES) measurements and images of the beam impact on the inner wall of the machine. All of the neutral beams at AUG have a full width half maximum of approximately 20cm. Care has been taken to ensure that the grids used for the neutral density integrals along the LOS encompass the entire beam (typically 80cm in 8 Extension of the core CXRS in AUG the directions perpendicular to the beam trajectory) and are sufficiently dense to describe the beam geometry well (∼1.5cm resolution perpendicular to the beam). Examples of the LOS integrals through the beam as a function of major radius are shown in the top row of Fig. 4 for three LOS corresponding to the centre fiber from each optical head. Magenta arrows indicate the direction along the LOS away from the optical head position. The centre of mass of each distribution is shown by a red dashed line and the radial and vertical error bars are taken at the positions encompassing 67% of the total neutral distribution. In the bottom row the radial and vertical error bars for all 70 LOS are shown. The LOS from the innermost two heads are very tangent to magnetic flux surfaces yielding a radial spread of signal less than 1cm after integrating through the neutral beam. Therefore, for these LOS, the radial resolution is effectively determined by the spot-size of the optics (±0.75cm). The LOS from COR 1 (the LFS head) are less tangential and the integration yields radial resolutions of up to ±2cm at the edge. Error bars on the Z-positions of the measurements are similar for all LOS, ∼ ±1.6cm. The radial and vertical error bars of these systems are included in the analysis of the data whenever the profiles are mapped via magnetic flux surfaces. On the HFS, the different NBI sources in box II are well separated vertically and the LOS only measured emission from NBI source 8. On the LFS, however, some emission from NBI sources 5 and 7 is also observed and is taken into account in the calculated radial positions of the measurements, radial resolution, and in the analysis of the impurity densities when these beams sources are in use. Of the 70 LOS from the three optical heads, 48 LOS are dedicated to CXRS measurements. The remaining 22 LOS are dedicated to BES measurements, which are important to the CXRS systems for two reasons. First, the BES measurements help to constrain the three dimensional geometries used to describe the beams and calculate the effective measurement positions. Second, accurate knowledge of the absolute neutral densities of all of the beam energy components and excited states is needed to determine impurity densities from the CXRS measurements. BES measurements of the Dαspectrum provide a measurement of the n=3 populations of all beam energy components. With the help of a radiative model, this can be translated into LOS integrated measurements of the neutral densities in the ground and first excited state7,8. This information is exactly what is needed to calculate CXRS impurity densities, see Eqn. 1, and can also be used to benchmark the neutral density 9 Extension of the core CXRS in AUG Unfortunately, this is only possible before and after experimental campaigns when access to the torus is possible. Degradation of the optics by damage or coating of the first optical component of the system does occur during the course of plasma campaigns. This is typically observed to be on the 10% level and, therefore, an additional 10% uncertainty is included in the absolute intensity calibrations of the diagnostic when analysing impurity densities. An absolute intensity calibration of both the entire system (optical head through to the spectrometer) and the spectrometer alone are performed such that the transmissions of the individual LOS (optical heads + relay fibers) can be determined independent of their position on the spectrometer. Here the transmission refers to the fraction of photons that are transmitted through the combination of the optical heads and the relay fibers. It does not include the spectrometers or spectrometer fibers. The transmissions of the system have a small wavelength dependence and vary from ∼50% at 400nm to ∼65% at 650nm. This information is useful for several reasons. First, it enables any changes in the intensity calibrations to be isolated to either the optical heads and relay-fibers or to the spectrometer and spectrometer fibres. These changes are tracked over multiple experimental campaigns and help us to identify problems and damaged components. Second, should the need arise, it enables us to change the connections of the relay fibers with respect to the spectrometer fibers (where a given LOS is measured on the camera), while maintaining the absolute intensity calibration to within the aforementioned 10-15%. Lastly, knowing the absolute intensity calibrations of the spectrometers alone, allows direct comparisons to be made between the various spectrometers installed at AUG. Fig. 9 shows the spectrometer intensity calibrations for one of the new systems (COR F/2.8) in comparison to one of the older systems (CER F/4) as a function of channel on the CCD camera at a fixed wavelength. Both of these systems employ the same Princeton Instruments CCD chip. EM gain was not used for either set of data. In the case of the CER the controller gain was set to 2 and for the COR the controller gain was set to 4. Therefore, the data from the COR has been divided by 2 to enable a more direct comparison. Hence, these curves correspond to an effective controller gain of 2 for both systems. For otherwise identical cameras and settings, a factor of two improvement in intensity sensitivity was expected based on the F/# of the spectrometers. However, the measured intensities show on average less than this, only a factor of ∼1.5 at the camera center increasing to ∼1.8 near the edges of the camera due to the reduced vignetting with the larger lenses. It is not 16 Extension of the core CXRS in AUG Spectrometer intensity calibrations (567nm) COR (F/2.8) CER (F/4) 5 10 15 20 25 Channel # (position on camera) 0 1e11 (cnts/ph)/(m2*str*px) 15 10 0 5 FIG. 9. Intensity calibrations of the new COR (F/2.8) spectrometer compared to the older CER (F/4) diagnostic as a function of position on the camera at fixed wavelength. Both systems feature identical cameras and no EM gain was used. The data shown here correspond to effective controller gains of 2 for both systems. entirely clear where this “loss” of light originates. It is possible that the larger lenses have lower transmissions compared to those used in the F/4 systems. IV. ANALYSIS OF CXRS SPECTRA At AUG the majority of the plasma facing surfaces are either solid W or W-coated. There are no longer any graphite surfaces and the carbon content in the machine is typically 0.3% or less, making C CXRS measurements challenging. Therefore, the standard AUG CXRS impurity is boron, which is introduced into the machine through regular boronizations. Additional impurities, such as He, Li, C, N, O, F, Ne and Ar can be (and have been) measured with the core CXRS systems when these impurities are present in the machine. Typical B CXRS spectra from the COR diagnostic are shown in Fig. 10. Here, a spectrum from a LFS LOS around mid-radius, a LOS near the magnetic axis, and a HFS LOS also around mis-radius are shown. The standard fit recipe for the boron spectrum is comprised of 5 Gaussians, an active and passive line for both the n=7-6 and the n=11-8 BV transitions plus a single passive line for the edge BII emission line at 494.038nm. Here, active refers to emission that results from charge exchange reactions with the injected neutrals in the 17 Extension of the core CXRS in AUG Intensity (1017 Ph/m2/str/nm/s) 1.5 1.0 0.5 0.0 0.0 0.4 0.2 0.8 0.2 0.6 0.4 0.0 Intensity (1017 Ph/m2/str/nm/s)Intensity (1017 Ph/m2/str/nm/s) 2.0 493.5 494.0 494.5 495.0 495.5 496.0 496.5 Wavelength (nm) 0.1 0.5 0.3 Discharge 32866, 4.045s COR LOS 09 R=2.045m LFS COR LOS 30 R=1.744m ~Mag. Axis COR LOS 54 R=1.417m HFS BII 494.038nm BV (n=7-6) 494.467nm Active Passive BV (n=11-8) 495.01nm Active Passive Wavelength range used for fit FIG. 10. Measured B spectra from the COR diagnostic taken in discharge 32866 at 4.045s. The top panel contains a spectrum from a LFS LOS around mid-radius, the middle panel a spectrum from near the magnetic axis, and the bottom panel a HFS spectrum from around mid-radius. main plasma volume. Passive refers to emission from the plasma edge that originate due to electron impact excitation and other processes. The number of free parameters in this fit is reduced due to the fact that the temperatures and rotations derived from the n=7-6 and n=11-8 come from the same population of boron ions and, therefore, must be the same. In addition, the positions and widths of the passive contributions are constrained to reasonable temperature and rotation ranges consistent with the ionization stage of the impurity and its location in the plasma. These constraints help to correctly separate out the active and passive emissions. In all cases, the measured CXRS spectra are fit to a convolution of the Gaussians defined in the fit recipe and a Gaussian approximation to the instrument function (see previous section) for the corresponding spectrometer channel. 18 Extension of the core CXRS in AUG 0 1 2 3#32866 4.00−4.10 s LFS HFS 0 1 2 3 Ti[keV] CER COR CUR CMR 1.2 1.4 1.6 1.8 2.0 2.2 R[m] 0 40 80 120 vφ[km/s] CER COR CUR CMR 0.0 0.2 0.4 0.6 0.8 1.0 ρtor 0 20 40 60 ω[krad/s] LFS HFS Ti[keV] FIG. 11. Impurity ion temperature and toroidal rotation measured using the new and pre-existing CXRS diagnostics in discharges 32866 at 4.045s as a function of major radius (left) and mapped onto the normalized toroidal flux coordinate ρtor (right). The rotation shown on the left as a function of major radius is the rotation velocity in km/s, while on the right the toroidal rotation frequency ω=vφ/R is shown. The full ion temperature and toroidal rotation profiles corresponding to these spectra are shown in Fig. 11. On the left hand side the profiles are shown as a function of major radius and the data from two pre-existing toroidal CXRS diagnostics from NBI 3 (CER in black squares and CMR in blue stars) are provided for comparison. The agreement between the diagnostics on the LFS is extremely good and one can see that the ion temperature profiles significantly constrain the position of the magnetic axis. On the right hand side, the same profiles are shown again as a function of the normalized toroidal flux coordinate, ρtor = p(ψ0−ψ)/(ψ0−ψsep). Here, the individual diagnostics are not distinguished. Rather, only whether the data was measured on the LFS (red squares) or on the HFS (blue diamonds) is indicated. The HFS and LFS boron temperatures overlay extremely well. In this case, a pressure constrained equilibrium was used for the mapping. Without the pressure constraint, radial shifts of up 3 cm are required to align the HFS and LFS profiles, which is outside of the 19 Extension of the core CXRS in AUG error bars on the measurement. The relative alignment of the HFS and LFS Tiprofiles, therefore, provides additional information on the magnetic flux surfaces, which can be used to check the quality of the equilibrium reconstruction. Note that the diagnostic LOS are not purely in the toroidal direction, see Fig. 1, and, therefore, do not measure purely toroidal rotation. The vertical tilt angle of the LOS varies between 7 and 13 degrees with the largest angles for the LFS LOS (bottom head in Fig. 2) and the smallest for the HFS LOS. This means that if there is poloidal velocity in the plasma, these LOS pick up some projection of this flow. However, this projection is very small. Typical core poloidal velocities are on the order of 1km/s and the projection of this onto the LOS would result in a correction to the measured LOS velocities of 0.1-0.2 km/s. Core poloidal velocities of 10km/s and upwards would be needed before the projection of this velocity onto the LOS reached the uncertainty threshold of the measurement. Therefore, within the error bars, the LOS measured velocities represent the projection of the toroidal rotation into the LOS direction. Once the spectra are fit for the LOS velocities, the real plasma equilibrium is used to calculate the projection back into the purely toroidal direction for every LOS and every time frame. Note that the rotation shown on the left-hand side of Fig. 11 is the toroidal rotation velocity (vtor) in km/s, while on the right hand side, the toroidal rotation frequency ω=vtor/R in kHz is shown. If the poloidal rotation in the plasma is zero, this quantity should be symmetric on a flux surface and the total plasma flow can be well described by simple solid body rotation. If the poloidal rotation is non-zero, however, then this will cause a poloidal asymmetry in the measured toroidal rotation frequency4–6 from the LFS to the HFS. In the example shown here, one can see that the measured HFS frequency is slightly lower than the LFS frequency, indicating a small poloidal flow. Detailed calculations of the poloidal flow in the core of AUG plasmas using this method and comparisons to neoclassical theory will be the subject of an upcoming publications, and is beyond the scope of the work presented here. The measured B+5 intensity profiles and corresponding impurity density profiles from the new and old CXRS diagnostics are shown in Fig. 12. Here, one can see that the intensities measured by the new (COR/CUR) and old (CER) diagnostics agree well. This is expected because, as previously mentioned, the difference in the charge exchange crosssections for the 60 and 93KeV beams is compensated by the different number of particles 20 Extension of the core CXRS in AUG 1.2 1.4 1.6 1.8 2.0 2.2 R[m] 0 2 4 6 I[Ph m−2sr−1nm −1s−1] ×1016 #32866 4.00−4.10 s CER COR CUR 0.0 0.2 0.4 0.6 0.8 1.0 ρtor 0 1 2 3 4 5 nB5+ [m−3] ×1017 LFS HFS FIG. 12. (Top) Measured boron line intensities from the core CXRS diagnostics. (Bottom) Impurity density profiles calculated from both the CER (NBI 3, 60KeV) and the COR/CUR (NBI 8, 93keV) diagnostics. injected by the different sources. The agreement in intensity between the two different sets of measurements provides an additional check on the intensity calibrations of the diagnostics and has been useful for identifying problems that occur during campaigns. The attenuation into the plasma for the two different beams, however, is different and this is reflected in the steeper attenuation observed for the CER (60KeV beam), which only just becomes visible near the plasma center. The boron densities measured by the new and old diagnostics are shown in the bottom panel of Fig. 12. There is an asymmetry outside of ρφ=0.6, but inside of this radius the LFS and HFS density profiles overlay well. The good agreement between the calculated boron densities from the old and new CXRS diagnostics (all red squares) indicate that the intensity calibrations, the beam geometries for the two NBI boxes, the beam energy composition and 21 Extension of the core CXRS in AUG excited state calculations, and the neutral density attenuation calculations have all been implemented correctly. Outside of ρtor = 0.6 the calculated HFS densities are larger than on the LFS. The reason for this is not yet known. However, it is a direct result of the slight increase in measured intensities that can be seen on the innermost points of Fig. 12. This asymmetry is not the result of errors in the intensity calibrations. Nor can it be explained by an incorrect beam geometry or beam attenuation. One possible explanation is that an additional unknown and unaccounted for emission in the measured spectrum could lead to an erroneous increased emission in this region. However, careful examination of the fitted spectra and the residuals show no evidence of this and, moreover, the asymmetry has been observed while measuring both boron and nitrogen. This makes an additional emission, which effects both spectral lines more or less identically in this region, very unlikely. Alternatively, the increased emission could be explained by an increased population of neutrals from the high field side SOL being transported inward and undergoing charge exchange reactions. This possibility will be tested in future experiments by using beam modulation to isolate the NBI-related active CXRS signal. V. ATOMIC EFFECTS ON THE CXRS SPECTRA The LOS geometries of the new systems were optimised to maximise the radial resolution of the diagnostics (tangency to flux surfaces). This lead, however, to large angles between the LOS and the injection trajectory of the neutral beam varying from 76◦on the LFS to 45◦on the HFS. This can be seen, for example, in Fig. 1. Due to these angles and the energy dependence of the charge exchange cross-sections, the cross-section effects described in von Hellermann et al13 need to be evaluated for each LOS and, when applicable, included in the analysis of the data. To evaluate the effect of cross-section effects on the measured spectra an approach similar to that described in von Hellermann et al was followed. The results of these calculations for the boron n=7-6 CX emission line are shown in Fig. 13. In the top plot the percent error between the true and the apparent impurity ion temperature (%∆Ti= (Ttrue −Tapp)/Ttrue ×100) is shown as a function of the true temperature for the COR LOS. The bottom curve (black) corresponds to the outermost LOS on the LFS and the top curve (red) to the innermost LOS on the HFS. Typically, the edge temperatures at 22 Extension of the core CXRS in AUG AUG are less than 1 keV, while core temperatures are typically less than 5 keV. However, for unusual cases core ion temperatures in the vicinity of 10 keV can be achieved. In all cases, the error on the ion temperature due to this effect remains less than 5% and has, therefore, not been included in the standard analysis. For the rotation the cross-section effects can become important under certain circumstances. In the middle and bottom panels of Fig. 13 the percent error on the toroidal rotation as a function of ion temperature is shown for two different rotation velocities. At low rotation, significant corrections can become necessary at high ion temperatures. However, as the effect scales much more strongly with temperature than it does with the rotation velocity, the error on the measurement drops quickly with increased rotation. In the experiment, toroidal rotation and ion temperature are strongly correlated, meaning low rotation velocities typically only occur together with low ion temperatures and high velocities with high temperatures. Even for very high ion temperatures (∼10 keV) in the plasma core, the effect remains under 10%. However, for the special circumstance of low rotation and high ion temperature, particularly on the HFS, these corrections should be considered. These calculations are done considering only the first energy component of the beam, as this is the dominant contribution to the measured intensities. However, even small contributions from the lower beam energy components can potentially cause disproportionately large apparent velocities due to the increased slope of the charge exchange cross-section at lower energies for the standard CX transitions. For boron, the slope of the reaction rate is maximised around ∼25 keV/amu, which is quite close to the second energy population for NBI box II (∼23 keV/amu). For the LFS LOS, the second energy component of the beam can be responsible for up to 25% of the total signal, while on the HFS, it is less than 10%. The cross-section effects on the measured CXRS spectra were also evaluated taking into account the contributions from the other neutral populations. To do this the charge exchange spectra, including cross-section effects, were calculated separately for each energy component as a function of true ion temperature and velocity and then weighted according to their contribution to the total CXRS signal and summed. The final resultant spectra were then fitted with single Maxwellians to determine the “total” cross-section effects on the measured spectra. While this procedure yields slightly larger differences between true and apparent temperature and rotations than if only the contribution from the first energy component is considered, the effects remain negligible for the typical range of AUG operation. Therefore, 23 Extension of the core CXRS in AUG 0 2 4 6 8 10 % Error vφ% Error Ti True Ti (keV) 0 2 4 6 8 10 LFS HFS AUG standard Ti range 0 2 4 6 8 vφ=100km/s 0 5 10 15 20 0 5 10 15 20 vφ25km/s % Error vφ FIG. 13. Effect of the energy dependent charge exchange emission rates on the ion temperatures and toroidal rotations measured along the LOS of the new COR diagnostic. The apparent parameters here are calculated following the equations presented in von Hellermann et al13 and taking into account only the first energy component of the beam. (Top) Percent error on the ion temperature and (Middle/Bottom) percent error on the toroidal rotation for two rotation values as a function of the true ion temperature. In all panels the bottom curve of each data set (black) corresponds to the outermost LOS on the LFS and the top curve (red) corresponds to the innermost LOS on the HFS. 24 Extension of the core CXRS in AUG these effects are not included in the routine data analysis. It should be noted, however, that the cross-section effects on both Tiand vφare systematically higher for the HFS LOS than for the LFS LOS and this could potentially lead to systematic errors when using the LFS-HFS Tidata to constrain the magnetic equilibrium or using the LFS-HFS vφdata to infer the core poloidal rotation. Therefore, cross-section effects are included in a post-processing analysis whenever the data is used for these purposes. In all cases considered to date, the cross-section induced differences in the measured LFS-HFS impurity ion temperatures have been found to be well within the error bars of the measurements and have not made a substantial impact on the reconstructed equilibrium. For extremely high temperature plasmas, however, this may not be the case. The reconstruction of the poloidal velocity relies on the absolute magnitude of the difference between the HFS and LFS toroidal rotation frequency, ω=vφ/R. This means the differences in the toroidal rotation velocity induced by the cross-section effects is amplified by the difference in the major radii of the measurements on a given flux surface. For the evaluation of the poloidal rotation, the systematic error introduced by the cross-section effects has been found to be small, but not negligible, and is routinely included in all analyses of the poloidal rotation. Gyro-orbit effects14 on the measured spectra have been neglected as all of the LOS of this diagnostic are almost parallel to magnetic field lines within the neutral beam volume. However, it is necessary to correct for the Zeeman effect. The Zeeman effect causes a splitting of the spectral line when the atom is in the presence of a magnetic field. While this effect is typically small on AUG due to the relatively high ion temperatures (causing Doppler broadening to dominate over Zeeman splitting) and low magnetic fields, it is important for low temperature measurements, particularly on the HFS where the magnetic field can be up to 4T. The measurements from all AUG CXRS diagnostics are corrected for Zeeman splitting and fine structure by means of correction curves. The full splitting (including Doppler) has been calculated for each measured impurity transition (B n=7-6, C n=8-7, ...) for a dense grid of magnetic fields, ion temperatures, and angles of the LOS with respect to the magnetic field. These spectra are then fit by a single Gaussian to determine the apparent ion temperature. In all cases, the final Zeeman broadened spectral line is found to fit well to a single Gaussian. After the experimental data is fit for the apparent temperature, the 25