Spatiotemporal oscillation of an ion beam extracted from a potential-oscillating plasma source

A radiofrequency oscillation is successfully superimposed on a plasma potential of a filamented source plasma in an ion beam source while maintaining a constant plasma density, in order to investigate effects of oscillating source plasma potential on an extracted ion beam. The experiment is preliminarily performed with a positive argon ion beam source. A class-D amplifier operational over a wide range of a frequency from a few tens of kHz to several MHz is installed; leading the oscillation of the plasma potential in the plasma source for the frequency range being tested. The beam current profile downstream of the extraction grids shows an oscillation of the beam current at the peripheral region of the ion beam; implying that the oscillation of a beam halo is induced by the potential oscillation of the source plasma.


Introduction
A neutral beam injection (NBI) heating is one of the powerful and promising techniques for plasma heating in nuclear fusion researches [1][2][3]. In typical hydrogen/deuterium negative ion beam sources, the negative ions produced in a plasma source (called extraction region) are electrostatically extracted and accelerated by a grid system. The negative ion beam is neutralized via a gas neutralizer system, where the energetic ion beam is converted into the neutral beam via a charge exchange process and reaches the fusion plasma core surrounded by a strong magnetic field [4]. The extraction energy and current density of the negative ion beam required in International Thermonuclear Experimental Reactor (ITER) are about 1 MeV and 280 A m −2 for deuterium [2,5]; the conversion efficiency from the positive ion beam to the neutral beam via a charge exchange process is significantly lowered for such a high energy range due to the decrease in the reaction cross section, while the higher conversion efficiency close to 60% is still maintained for the negative ion beam case [4]. Therefore development of a hydrogen/deuterium negative ion beam source is crucial to establish the fusion reactor, where the ITER-NBI requires the divergence angle in the range of ∼0.17°-0.4°(3-7 mrad) [5].
The negative ion sources are typically divided into two types of a filamented-arc source and a radiofrequency (rf) source. In the former, the plasma is produced by the filamented arc discharge; the high energy electrons destroying the negative ions are filtered by a magnetic filter located in the vicinity of the first grid (called a plasma grid) and cooled down to the temperature less than 1eV [6][7][8]. The negative ions are mainly produced via a surface reaction process at the plasma grid and extracted electrostatically, where the grid system has to be designed to minimize the ion loss to the grids and the beam divergence, e.g. a beam divergence of about 4mrad has been obtained in [9]. The plasma source is replaced by an rf plasma source operated at several hundreds of kHz to a MHz typically for the latter case for extending the lifetime of the plasma source [10,11]. Although the plasma is mainly sustained by an inductively coupled discharge, a part of the rf power is coupled via a capacitive mode, which often enhances the rf oscillation of the plasma sheath or the plasma potential [12]. A Faraday shield has been installed inside the plasma source of the rf negative ion source to minimize the thermal load to the insulator source tube [2]; simultaneously it can suppress the capacitive coupling between the rf antenna and the plasma as demonstrated in the studies on the inductively coupled plasma [13]. The driving frequency is known to affect many aspect of the plasma device performance, e.g. the electron energy distribution function, the rf potential oscillation, the power coupling between the antenna and the plasma, and the plasma production, as investigated in the field of the plasma processing, e.g. in [14]. A recent experiment on the rf helicon hydrogen negative ion source has also shown the standing helicon wave characteristics in the bent magnetic field structure and the resultant electron heating process [15]. The well-collimated negative ion beams have been obtained in the negative ion source using the filamented-arc source as described above, while the collimation of negative ion beam in the rf ion source is still a challenging issue toward the ITER-NBI development since a great effort has been devoted to the investigation of the attainable current density rather than the beam optics; the divergence of the negative ion beam extracted from the rf negative ion source is recently reduced to about 1.5° [16]. For the case of the rf hydrogen negative ion beam source, the rf effect on the potential structure and particle dynamics near the beam extraction grid might be one of the additional key issues to improve the performance, since both the static and dynamic potential structures will affect the transport of the negative ions and the extracted beam optics. Furthermore, themodulation of the ion density by the rf power [17] might cause the fluctuated ion beam current. A previous study on the beam characteristics in a focused ion beam for industrial applications has implied that the energy of the focused ion beam spreads due to the potential oscillation of the plasma source attached to the acceleration grids [18]. More recently, a fluctuation of the negative ion beam current has been detected in the rf ion beam source [19].
One of the possible approaches to investigate the above-mentioned effects of the temporally varying potential is detailed diagnoses of a plasma source, a potential structure near the grids, and a beam extracted by the grids in a fundamental laboratory experiment. However, it would be difficult to control only the potential oscillation in the rf negative ion source in the laboratory, since some types of the rf coupling processes (capacitive, inductive, and wave couplings) are actually superimposed in the device and simultaneously affect both the plasma-density and plasma-potential oscillations. Hence superimposition of a controlled rf voltage to the plasma potential of the filamented ion source will be one of options to understand the physics underlying the plasma dynamics near the extraction grids and the rf effect on the beam performance, e.g. such an experiment might give optimization of the driving frequency from the view point of the beam extraction.
Here a plasma device yielding the superimposition of the rf potential oscillation on filamented plasma source is developed, where the rf voltage source utilizing a class-D switching circuit operational over a wide range of the frequency up to a few MHz is installed. Only the potential oscillation can be superimposed on the plasma source while having no density modulation. The preliminarily experiment is performed with the positive argon ion beam and the spatial profile of the beam current downstream of the grids is investigated by using a retarding field energy analyzer. The results implies the current oscillation at the peripheral region of the ion beam, i.e. the oscillation of the beam halo. Figure 1 shows the schematic diagram of the experimental setup. The beam extraction and acceleration grids consisting of a plasma grid (PG), an extraction grid (EG), and a grounded grid (GG) are attached to a 280 mm diameter and 600 mm long diffusion chamber, which is evacuated by a turbomolecular pumping system to a base pressure less than 10 −3 Pa. The grids have 9 mm diameter single center holes for the beam extraction and electrically isolated by insulator vacuum flanges. A plasma source consisting of a spiral tungsten filament and a 158mm diameter and 145 mm long discharge chamber is further attached to the upstream side (left side of figure 1) of the grids via an insulator structure, where z=0 is defined as the axial location of the upstream flange of the discharge chamber and then the axial location of the PG is z=150 mm. To inhibit the plasma loss to the discharge chamber, permanent magnets forming cusp magnetic fields are set around the chamber wall. Argon gas is introduced into the discharge chamber via a mass flow controller; the pressure measured in the diffusion chamber is maintained at 0.5 Pa. A dc current of about I heater ∼40 A is supplied to the tungsten filament located at z=15 mm to emit thermionic electrons, which are accelerated by applying a discharge voltage V dis between the filament and the discharge chamber via a resistor of 10Ω limiting the discharge current and ionize the argon neutrals via an electron impact ionization process. After turning on the dc power supply for the heater current, the discharge voltage can be pulsed by using an insulated gate bipolar transistor (IGBT) inserted between the dc power supply and the resistor to minimize thermal load. The plasma density measured by a Langmuir probe located at z=70 mm (not shown in figure 1) is about 10 m 17 3

Experimental setup
in the present experiment. The whole structure of the plasma source can be biased by inserting the bipolar power supply (V bias ) between the discharge chamber and the plasma grid for the control of the plasma potential with respect to the PG, being similar to [20,21]. The positive ion beam can be extracted and accelerated from the plasma source by applying extraction and acceleration voltages V ext and V acc between the EG and PG and between the GG and EG, respectively, where V ext =1.5 kV and V acc =8 kV are chosen in the present experiment. It should be mentioned that the fast ions in the diffusion chamber may cause the ion-neutral ionization process there; providing the electrons and slow ions, i.e. a beam-oriented plasmas.
To superimpose an rf voltage between the PG and the plasma source, a class-D switching amplifier shown in figure 2(a) is connected to the circuit as shown in figure 1 via an inductance of about 400μH. The amplifier includes two field-effect transistors switched alternatively. The output voltage from the amplifier is transferred to the load via a transformer and a LC resonance circuit. Since the impedance of the inductance inserted into the discharge circuit in figure 1 is large, the resonance conditions is mainly determined by the values of (L, C 1 , C 2 ) in figure 2(a); the frequency giving the LC resonance can be adjusted by changing the two capacitors of C 1 and C 2 . The amplifier is confirmed to be operational from a few tens of kHz to 2MHz by changing the frequency of the gate signals, where the amplitude of the output voltage V rf is changed by both the frequency and the dc voltage in figure 2(a).
A cylindrical Langmuir probe (LP) is inserted at z=125 mm from the sideport of the discharge chamber. The first derivative of the current-voltage characteristic of the LP give a steady-state local plasma potential V p , where the probe voltage is swept for about 50ms during the discharge pulse and the first derivative can be obtained by an analogue differentiation technique [22,23]. The LP is also used to identify the oscillation of the plasma potential in the plasma source by measuring the floating potential via a high-impedance voltage probe. A retarding field energy analyzer (RFEA) is located at z=275 mm to measure a spatial profile of the ion beam extracted by the grids, where a detailed structure of the RFEA is shown in figure 2(b) and consisting of a 6 mm diameter entrance ceramic orifice, three grids (G 1 , G 2 , G 3 ), and a collector electrode. By supplying a negative voltage to the second grid (G 2 ), the electrons are reflected there and only the positive charge can reach to the third grid (G 3 ) and the collector, where G 3 and the collector are electrically connected to minimize the effect of the secondary electron emission from the collector surface. Since the secondary electrons emitted from G 3 and the collector would be reflected by the negatively biased G 2 , the effect of the secondary electrons would be small in the present configuration. Furthermore, only the high energy positive beam ions can reach the collector and G 3 by biasing those at the positive potential higher than the plasma potential V p as sketched in figure 2(b). The collector current signal I c is converted into a voltage signal via a resistor inserted between the collector and the dc power supply, where the voltage of the resistor is measured by a digital oscilloscope via a precise isolation amplifier operational up to 1MHz. The signal of the applied rf voltage can be taken by connecting a high voltage ceramic capacitor (V rf ∼V ref in figure 1), which can remove the dc high voltage. The rf amplitude I c | |of the collector current and its phase difference Δf with respect to V ref can be obtained from the amplitude spectrum and cross spectrum analyses. The RFEA is mounted on a L-shaped shaft passing through a vacuum port at the downstream flange; approximately radial measurement can be performed with rotating the shaft by using a stepping motor.  Figure 3 shows the steady-state local plasma potential V p measured at z=125 mm as a function of the bias voltage V bias between the PG and the plasma source, where the extraction and acceleration voltages (V ext and V acc ) and the rf voltage V rf are set to zero. It is clearly observed that the plasma potential in the upstream plasma source is proportional to the bias voltage V bias for V bias >0, while no clear change can be observed for V bias <0. To force to oscillate the upstream plasma potential by the rf voltage V rf (t), the source should be operated with V bias +V rf (t)>0.   figure 4(a). No noticeable change by applying the rf voltage is observed in I is , while the oscillation of the floating potential is observed during the discharge pulse between t=0 and t∼105 ms as in figure 4(b). Since the ion saturation current is proportional to be the plasma density and the square root of the electron temperature, the stable signal of I is implies that both of them are not oscillated by the rf voltage. Hence only the plasma potential is forced to be oscillated while maintaining the constant plasma density and electron temperature.

Results
By taking the amplitude spectrum of V f from the data for t=50-100ms via a Fast Fourier Transform, the amplitude at the frequency corresponding to the applied rf voltage can be obtained. Figure 5 shows the amplitude of V f normalized by the applied rf voltage as a function of the frequency, which are taken at z=125 mm for V ext =V acc =0. It is found that the potential oscillation induced by applying the rf voltage reduces with an increase in the frequency. However the present setup will be able to provide the investigation for the wide range of the frequency close to a few MHz with a single rf power supply.
The amplitude reduction of the potential oscillation for the high frequency range has been observed in the previous study [24] and has been briefly understood by resistive and capacitive rf sheath models for the low and high frequencies, respectively. Very briefly, the amplitude ratio of the rf sheath voltages V V ps gs | | | |at the powered and grounded electrodes in a capacitive discharge configuration can be given as where the powered and grounded electrodes correspond to the discharge chamber and the PG, respectively, in the present experiment, and the numerator and denominator of the left-hand side in equation (1) correspond to the effective contact area of the plasma with the grounded and powered electrodes, respectively; being given by the sheath edge densities (n gs and n ps ) and the surface areas (A gs and A ps ). The factor q is known to increase with an increase in the rf frequency, i.e. variation from the low-frequency resistive model to the high-frequency capacitive model [12,24]. The amplitude V f | | of the potential oscillation measured by the LP would be close to V gs | |in the present configuration if assuming a negligible voltage drop within the plasma. If the cusp magnetic field set around the plasma source reduces the effective area n A ps ps and provide the numerator larger than the denominator in the LHS of equation (1), the value of V V ps gs | | | |increases with the increase in the factor q, i.e. the increase in the frequency. This is equivalent to the decrease in V gs , i.e. the reduction of the rf amplitude of the floating potential V f in figure 5 can be qualitatively understood. The detailed calculation of the potential oscillation is out of scope of the present paper.
To investigate the ion beam extracted from the source by applying the two high voltages V ext and V acc , the basic I-V characteristic of the RFEA is firstly investigated with the positive-beam extraction. It is once again mentioned that the extracted ion beam might cause the ion-atom ionization collision since the mean free path for this collisional process is about 150mm being shorter than the diffusion chamber, where the previously measured cross section of 5×10 −20 m 2 [25,26] and the argon pressure of 0.5Pa are used for the calculation. Since the beam-oriented slow ions will appear around the plasma potential in the diffusion chamber, it is  important to know the plasma potential. The first grid G 1 exposed to the beam and beam-oriented plasma is used as a Langmuir probe and the current I g1 -voltage V g1 characteristic of the G 1 grid is shown in figure 6(a). The change of the current is observed at around V g1 ∼30-40V corresponding to the downstream plasma potential. It should be mentioned that the current I g1 is positive over the whole range of V g1 , which originates from the presence of the high energy positive ion beam, i.e. the current for V g1 <20 V is the sum of the ion beam current I ib and the saturation current I ip of the slow ions in the beam-oriented plasma, while that for V g1 >50 V corresponds to the sum of I ib and the electron saturation current I ep of the beam-oriented plasma.
Then the electrical circuit of the RFEA is set as that in figure 2(b) and the collector current I c -the collector voltage V c characteristic is taken as plotted by open circles in figure 6(b) where the electrons are reflected by the negative bias voltage of the repeller. The dashed line shows the linear line fitted to the data for V 40 V c  , implying the deviation of the measured current from the linear line for V 40 V c  . This deviation seems to be due to the presence of the slow ions around the plasma potential close to ∼40V, i.e. the net current corresponds to the ion beam current I ib for V 40 V c  and to the sum of I ib and the saturation current of the slow ions for V c <40 V. Therefore, only the beam ions is detectable by biasing the collector above the plasma potential. Based on the above-described characteristic of the RFEA, the collector voltage for the beam measurement described hereafter is chosen as 100V.
Radial measurement of the ion beam current is performed with the rf voltage for V bias =50 V, V ext =1.5 kV, V acc =8 kV, where the amplitude and frequency of the rf voltage are chosen as ∼7.5±2 V p 0and 550kHz, respectively. The dc and rf components of the collector current I c are obtained by taking the amplitude spectrum via a Fast Fourier Transform. Figure 7