Development of a variable-energy, high-intensity, pulsed-mode ion source for low-energy nuclear astrophysics studies

Knowledge of astrophysically relevant nuclear cross sections at low energies (ranging from tens to hundreds of keV) is crucial for defining nuclear physics constraints on stellar evolution and nucleosynthesis models. This information allows us to calculate the corresponding thermonuclear reaction rates at stellar burning temperatures. However, Coulomb repulsion between reacting nuclei results in exceedingly small cross sections (often ≲10⁻⁹ b) at stellar energies. In order to detect such low-statistics events directly, the Laboratory for Experimental Nuclear Astrophysics (LENA) at the Triangle Universities Nuclear Laboratory (TUNL) uses intense proton beams and multiple background suppression techniques to enhance the experimental signal-to-background ratio. A useful statistical measure of experimental feasibility (⁠$F$⁠) is the ratio of net signal counts to the statistical uncertainty of the total detected events in the spectral region of interest. When the signal strength is small, and this region is dominated by background, this reduces to

We use two accelerators to cover the energy range of 20-1000 keV: a modified 1 MV Van de Graaff electrostatic accelerator¹ and an electron cyclotron resonance ion source (ECRIS). As displayed in Fig. 1, the accelerator beam lines are joined via an analyzing magnet to a beam transport section that terminates at the experimental target. The ECRIS and its recent high-current upgrades are the topic of this article.

The primary task of the ECRIS at LENA is the production of intense, low-energy (≲230 keV) proton beams. Key features of its design can be traced to the system described by Wills et al. and used at Chalk River Laboratories.² The source and its acceleration column (Fig. 1), plus its necessary electronics, sit on an air-insulated, high-voltage table. This table is electrically isolated from earth ground during operation and can be raised to a desired potential using a +200 kV power supply.³ A dense electron-cyclotron-resonance-heated plasma is created using 2.45 GHz microwave radiation and an 87.5 mT solenoidal magnetic field, supplied by a permanent magnet array with a remotely adjustable axial position. Proton beams from this source were used over several years for experiments,^4–7 including periods of up to 12 months without significant source maintenance. Increases in target beam current to ∼2 mA over this period were acheived with operator experience and minor modifications that improved ECR source reliability. For further information concerning details of source design and operation, see Cesaratto et al.⁸ 

The following text describes source and acceleration column upgrades to the ECRIS at LENA for the production of high-current, pulsed proton beams. Section II discusses a high-power magnetron upgrade that enabled pulsed beams, yielded higher extracted current, and dramatically improved our experimental feasibility, $F$⁠. Section III begins by describing how acceleration of such higher extracted currents caused serious, irreversible damage to the prior acceleration column used to transport the beam from the ECRIS to ground. The remainder of Sec. III concerns the successful effort to design, construct, and install a more robust acceleration column, with features to enable transport of at least 2 kW of beam power. Descriptions of necessary high-current upgrades to beam monitoring, focusing, and control systems are discussed in Sec. IV. Experience with operating these new systems for early experiments is described in Sec. V. Finally, we summarize and suggest future upgrade directions in Sec. VI.

Equation (1) implies that our experimental priority should be to increase our signal count (i.e., beam current) alone, but in practice the experimental situation is more complex. Our water-cooled targets tend to sputter or degrade during high-beam-power deposition. Non-destructive average beam power can be maintained, while the beam current is raised if the beam can be pulsed. However, to maintain pulsed-beam data collection rates comparable to DC beam rates, peak-pulsed beam currents must be increased by a factor corresponding to the inverse of the pulse duty cycle. If pulsed operation can be achieved, then data can be collected and stored separately during beam-on and beam-off intervals. Pulsed-signal data collected would then, on average, contain lower background counts arising from continuous cosmic ray or environmental radiation sources, thereby increasing $F$⁠.

To verify whether this goal could be achieved, the microwave system for the ECRIS was upgraded with the goal of increasing the output beam intensity. The new system⁹ features pulsing capability with a maximum power output of 1.2 kW. Our data acquisition system was also modified to allow simultaneous storage of both beam-on and beam-off spectra.

No adjustments of the high voltages applied are required with this pulsing method. Thus, the plasma is fully extinguished between beam pulses and, subsequently, reignited during each pulse. This repeated cycle of extinguishing and striking a plasma presented a significant challenge for microwave impedance matching between the waveguide and plasma chamber. The waveguide impedance is adjusted via a 3-stub tuner and is typically optimized prior to striking a plasma by introducing microwave power and minimizing the power reflected back toward the magnetron. Once the plasma lights, its impedance changes drastically. During DC beam operation, this change is easily accommodated by re-optimizing the 3-stub tuner for maximum microwave power transmission into the plasma. However, with this optimization, if microwave power is switched off to extinguish the plasma, the impedance match changes such that the plasma will not light again when microwaves are reintroduced. Thus, to achieve stable pulsed-beam operation, we intentionally introduce a slight impedance mismatch between the waveguide and the plasma chamber for microwave power both on and off.

The beam pulses observed under these conditions are presented in Fig. 2. The blue and yellow traces depict DC and pulsed beams, respectively, showing that the beam intensity within the 1 Hz, 100 ms wide pulses (i.e., a 10% operational duty cycle) is equal to that of the DC beam and has a pulse rise time of ∼100 μs.

As a test of the effectiveness of such 10%-duty-cycle beam pulsing to reduce environmental backgrounds in our spectra, data were taken on the ¹⁸O(p,γ)¹⁹F resonance at $Ercm$ = 151 keV. Data were acquired during successive runs with DC and peak-pulsed beam currents on targets of 37 and 370 μA, respectively. No target degradation occurred between data sets. We sought this factor-of-10 difference between DC and peak-pulsed beam currents in order to observe and verify spectral improvement via suppression of continuous environmental backgrounds. Data obtained on this resonance using DC and pulsed ECRIS beams are presented in Fig. 3 as the red and black histograms, respectively. The pulsed spectrum contains only the data taken when the beam was on target. The peak shown at channel 918 is from the 3908 → 1554 keV decay in ¹⁹F initiated by proton capture onto ¹⁸O, while the peak labeled ²⁰⁸Tl is a common environmental background line. Note that, since the net beam charge accumulated on the target is nearly identical between DC and pulsed data sets, the resonance peak at channel 918 is nearly unchanged. However, the total time during which pulsed data were taken is $110$ that of the DC data set. Thus, the intensity of the environmental ²⁰⁸Tl background line decreased by the same factor in the pulsed data set. These tests clearly demonstrate that data collection with pulsed beams can drastically reduce environmental backgrounds. However, because the continuum displayed in Fig. 3 is a beam-induced background resulting from Compton events, including those from higher lying transitions, it is not reduced by pulsing.

The ECRIS and its beam-extraction system⁸ were originally mated with a 200 kV acceleration tube built by High Voltage Engineering Corporation (HVEC) and modified for this application. This consisted of 24 aluminum electrodes bonded to intermediate Pyrex glass insulators with polyvinyl acetate (PVA).

In this prior design, a large turbopump at ground potential and a low-conductance pumping path for residual gas from the ion source created pressures decreasing from ∼10⁻⁴ to ∼10⁻⁵ Torr in the beam extraction region and downstream acceleration tube. This advantageously facilitated beam space-charge neutralization through the production of electrons via gas ionization. However, this also resulted in excessive bremsstrahlung radiation levels caused when electrons were accelerated toward the high-voltage end of the tube and struck interior surfaces. After installation of the new microwave system, the higher extracted beam currents magnified this problem and also increased the loading by both electrons and beams on the resistor chain that established the voltage gradient. This overheated the resistors and changed their values which, in turn, perturbed the focal properties of the tube and further increased the beam loading. As a result, the tube overheated and the PVA seals weakened, creating vacuum leaks, which exacerbated the above problems. This led to internal electrical discharges that permanently damaged several glass insulators. Thus, the voltage gradient of the tube was significantly reduced, which required that it be replaced.

With this upgrade, we sought to improve the acceleration system for the ECRIS on three fronts: beam optics, electrostatic stability, and a reduction of beam-induced bremsstrahlung radiation. We were guided in our new design by important features of the Hyperion acceleration system developed by scientists now at Neutron Therapeutics.¹⁰ 

To ensure the transport of high-brightness beams to target, we needed to find a satisfactory electrode geometry for accelerating intense H⁺ beams over our required experimental energy range (20-230 keV). We wished to use the extraction electrode voltage and axial position to control the shape of the plasma meniscus from which the beam was extracted. The goal was then to let the beam diverge, thereby minimizing space-charge effects, and optically match it to the acceleration column entrance downstream. Experience with the prior, now-damaged HVEC acceleration column had shown that beam transport over this energy range was best enabled by a strong voltage gradient at the column entrance. This gradient was maintained by shorting an increasing number of downstream sections of the acceleration column to ground as the beam energy was lowered.

The ceramic insulators of the new column would be protected against high-voltage breakdown by overlapping interior shield rings, inter-electrode spark gaps, and a voltage gradient entirely supported by chilled, deionized (DI) water resistors. Bremsstrahlung radiation levels would be reduced by employing an open, well-pumped electrode geometry to extract, focus, and transport the beam successfully from the ECRIS discharge chamber into the acceleration column, thereby reducing beam ionization of background gas. Any secondary electrons generated by the beam would be suppressed via alternating, transverse magnetic fields along the acceleration column, in conjunction with an electron suppression electrode at its grounded end.

Electrodes and insulators of the new column would be held together under compression with O-rings. The benefits of this design are that we avoid volatile adhesives to obtain a high-vacuum seal and reduce the electrostatic issues associated with triple junctions.¹¹ In addition, the column could be disassembled should repairs become necessary.

Simulations of electrostatic effects by an acceleration column geometry similar to that of the Hyperion system¹⁰ were carried out using an early version of the software package kassiopeia,¹² supplied to us for testing purposes by the KATRIN (KArlsruhe TRItium Neutrino experiment) group at the University of North Carolina (UNC). kassiopeia is a charged-particle propagation and tracking software suite divided into two main subpackages.

One is an electrostatic field calculator^13,14 that solves Laplace’s equation within charge-free regions using two different methods: the indirect boundary element method (BEM) and, in axially symmetric regions far from charge distributions, a zonal harmonic expansion. With the aid of boundary conditions, zonal harmonic coefficients of a quickly converging expansion are solved in terms of charge densities provided by the initial BEM calculations. This output is then passed to the second subpackage, a particle tracking code,¹⁵ which calculates a complete description of the kinematics of charged particles using an eighth-order Runge-Kutta integrator in the user-defined region. kassiopeia does not account for space-charge effects, so it cannot reliably simulate trajectories within our extraction region where they are most significant. Also, any particles that were generated on collisional trajectories with our electrodes were removed from the simulation.

These simulations led to the geometry and associated voltage configurations illustrated in Figs. 4(a)–4(c). After initial extraction from the ECR plasma chamber, the beam is allowed to expand radially in two conical electrodes, minimizing space-charge effects before further acceleration. The last conical section can be biased positively to retard the energy of the beam before it enters ten successive accelerating electrode gaps of up to 20 kV each (depending on final beam energy) that step the voltage to ground. The initial internal electrode spacing was specifically chosen to be small for strong initial focusing, followed by eight larger electrode gaps that propagate the beam to ground. This configuration yields a nearly parallel beam upon column exit [see the right side of Figs. 4(a) and 4(c)] that can be propagated effectively by solenoid lenses further down the beam line. Raising the last conical electrode voltage increases the focusing strength of the subsequent small gaps, as seen in Fig. 4(b).

To accomplish effective focusing over the desired beam energy range of 110–230 keV, we chose to limit the maximum energy of the beam extracted from the ECR plasma to ∼30 keV. We assumed in the design phase that our ECR plasma provided a maximum current density of ∼200 mA/cm², comparable to that available at Los Alamos National Laboratory (LANL)¹⁶ and Neutron Therapeutics.¹⁰ Our 200 kV, 36 mA power supply³ used for acceleration provided a maximum current limit and, therefore, determined the plasma aperture size from which we extracted the beam. Extraction and focusing of this beam is optimized by varying the gap between the plasma aperture and the extraction electrode. A suitable tuning range for extraction gap sizes (see Table I) was determined using the Child-Langmuir Law^17,18 and optical estimations from Alton and Bilheux.¹⁹ 

Transverse magnetic fields are an effective means to suppress any secondary electrons created by beam interactions with residual gas in our new column. They steer any secondary electrons into column electrodes before they gain appreciable kinetic energy, thereby limiting the maximum intensity of bremsstrahlung X-rays produced upon collision, and offering a solution to the radiation problems of our prior system. Furthermore, the optical effects of magnetic suppression upon the beam are not dependent upon its velocity,²⁰ so the suppression system is electrostatically decoupled from the column electrodes, along with all of their vulnerabilities to electron and beam loading effects.¹¹ This means that, for any proton injection energy, the magnetic fields will always have the net effect of preserving the dynamics of the beam, unlike inclined electric field suppression systems which can produce divergent beam profiles.²⁰ 

We used a simplified kinematic model to estimate the resultant transverse displacement of our proton beam centroid as it passes through regions of transverse magnetic fields in the acceleration column. The Lorentz forces that act on an axial proton on its journey down the acceleration column were considered [illustrated in Fig. 4(d)]. Since the injection energy of the proton, electrode axial spacings, and voltage differences across each gap were known, the axial acceleration and transit time of the proton across any region of the column were calculable. By reversing the direction of the magnetic field sections appropriately along the length of the column, an initially axial proton experiences transverse impulses at each section until attaining a final off-axis displacement of ∼0.5 mm [see Fig. 4(d)].

An overview of the upgraded ECR ion source and acceleration column, designed to achieve the objectives described above, is presented in Fig. 5. A more detailed view is given in Fig. 6(a), where key elements discussed in the text below are defined and labeled. Figure 5 displays in the foreground the (yellow) ECR source permanent magnet array and (gray) plasma chamber described by Cesaratto et al.⁸ Structures to the right extract, focus, and accelerate the beam to ground through the sequence of (gray) annuli whose electrostatic potentials are established by twin, coaxial, DI-water-resistor channels in the upper left and bottom corners of the triangular column electrodes.

The overall structural and high-vacuum integrity of the new acceleration column comes from a self-aligning, compression design facilitated by two sets of three G10 fiberglass rods that are screwed into opposing sides of the 0-200 kV high-voltage-table electrode and upon which all triangular electrodes and ceramic insulators are supported. As hex nuts on opposite ends of the column are tightened, all of the electrodes and ceramics are squeezed together and the O-rings, which provide high-vacuum seals at their interfaces, are compressed.

Our triangular electrode design is based on one used for the Hyperion accelerator.²¹ Two principal electrode geometries compose the main part of the column: the shield ring and magnet ring electrodes. These electrodes have cylindrical ends that extend axially in both directions and, together, block any line-of-sight trajectory for stray beam charge to reach the ceramic insulators. Referring to Fig. 6(a), the final magnet ring, which is biased to ∼−3 kV, acts as a column electron suppressor electrode to retard any backstreaming electrons that might enter from beyond. Twenty-eight NdFeB magnets²² are mounted along the outside circumference of the magnet ring electrodes and are not exposed to beams or secondary electrons. All electrodes were machined to accommodate these features for beam extraction and acceleration by the UNC Physics and Astronomy Instrument Shop.

The plasma and ringless electrodes of Fig. 6(a), and the adjacent ceramic insulators on the left side of the high-voltage-table electrode, support the ECR plasma chamber and enclose the beam extraction system. Microwaves enter from the left into the water-cooled, aluminum plasma chamber through a 3.2-mm-thick Al₂O₃ window that seals vacuum-tight against an O-ring. To protect this window from thermal stress caused by fast electrons, it is completely covered inside by a thin boron-nitride (BN) disk with a 1-cm-diameter, 1-cm-long cylindrical axial stub.

The plasma chamber and plasma electrode are biased up to ∼+30 kV with respect to the (green) high-voltage-table electrode [see Fig. 6(a)]. Positive hydrogen ions created by the ECR discharge are formed into a beam as they emerge from the plasma aperture [see Fig. 6(b)] toward the adjacent (green) electrodes, which are at the same potential as the high-voltage-table electrode. The (yellow) suppressor electrode is biased to ∼−1 kV and impedes backstreaming electrons from reaching the plasma chamber. The beam then enters the 27.4-cm-long expansion cone electrode where it drifts and expands radially until it reaches the 9.4-cm-long conical 0 to +10 kV focus electrode [indicated in red in Fig. 6(a)].

The expanded view of this beam extraction system in Fig. 6(b) shows a modified version of that used at LANL¹⁶ and in the Hyperion accelerator.²¹ The molybdenum beam defining apertures of the plasma, extraction, source suppressor, and expansion cone electrodes are all removable for ease of maintenance. The present beam defining aperture diameters are compared in Table I with those used at LANL, Neutron Therapeutics, and in our earlier LENA extraction system.

The extraction, source suppressor, and expansion cone electrodes are screwed tightly together into a self-aligned, rigid unit, with the source suppressor electrically isolated by using Vespel®²³ washers. Electrode parts and washers are machined so they seal together with O-rings to provide a vacuum-tight, interior DI-water-cooling channel.

This rigid beam extraction system is firmly clamped to the upper of two small platforms located below it inside the high-voltage-table electrode [see Fig. 6(a)]. Vertical, horizontal, pitch, and yaw adjustment of the upper platform with respect to the lower one enables alignment of the rigid beam extraction system. Also, the plasma chamber exit aperture position is horizontally and vertically adjustable. All apertures can then be aligned with the downstream acceleration column axis to within ±0.1 mm using a precise telescopic transit²⁴ and removable optical targets.

The lower platform can be positioned axially on roller bearings by using an external, gear-coupled stepper motor. During operation, this allows remote adjustment of the axial extraction gap length, s_PE. Our average gap size and tuning range are given in Table I. Other axial electrode spacings, s_ESS and s_SSEC, and thicknesses t_E and t_SS, are, 2.7, 2.6, 3.2, and 7.9 mm, respectively.

Our high-density Al₂O₃ column ceramic insulators²⁵ were chosen to have high-vacuum properties and 50.8 mm axial thicknesses to withstand voltages greater than 20 kV per section. Undulations visible in Fig. 6(c) on the interior, high-vacuum side extend the surface spark length; the exterior, atmospheric side is glazed for cleanliness. The shoulder of this glazed surface can be seen in Fig. 6(d). All ceramic insulator ends are aligned by 1.5-mm-deep annular shoulders in adjacent electrodes. Both of their end surfaces are ground smooth and, to minimize chipping, have chamfered corners as seen in Figs. 6(d) and 6(e). We rely on 12 mm spark gaps between electrodes, calibrated to discharge at 26 kV in atmosphere, to prevent ceramic damage.

The voltage gradient of the acceleration column and its allowed leakage current are supported by the parallel resistance of the twin, chilled DI water channels like the one visible at the bottom of Fig. 6(a). Each consists of successive sections of short “female” stainless steel and “male” fiberglass-impregnated Delrin²³ tubes. Each stainless steel tube is secured both mechanically and electrically to an acceleration column electrode by using a set screw, and joints between the stainless and Delrin tubes are sealed with O-rings. We chose the cross-sectional area of the tubes to provide up to 25 mA of leakage current for the maximum gradient of 20 kV per accelerating electrode gap. We also control both the conductivity and the temperature of the DI water supplied to these channels to adjust our leakage current to specific needs dictated by beam loading or power limits imposed by the 0-200 kV supply.³ 

As indicated in Figs. 4(a) and 4(c), the beam emerging from the acceleration column is large (5-7 cm) in diameter and either parallel or slightly diverging, so a lens is required to focus the beam through the 2.5-cm-vertical gap of the bending magnet located 2.0 m downstream. Hence, we removed the existing 5-cm-aperture magnetic quadrupole lens previously located between the ECR source and the bending magnet and replaced it with a large-aperture magnetic solenoid lens at the end of the acceleration column. A second, nearly identical magnetic lens was added after the bending magnet (see Fig. 1).

Each of our solenoidal lenses is comprised of eight, separate, resin-encased coils of hollow copper tubing with a cross-sectional area of 36 mm². Each coil has an inner diameter of 25 cm and an outer diameter of 50 cm and is 19-mm-thick. Current up to 250 A flows in series through the eight coils, but each coil is cooled independently by chilled water. For each solenoid, sets of four coils are held firmly and aligned coaxially inside large, machined, square blocks of aluminum. Two such aluminum blocks are then encased in a flux-return box made of 19-mm-thick cold-rolled steel plates. The pair of flux-return end plates for the solenoid lens nearest the ECRIS and nearest the target have inside diameters of 19.7 and 16.5 cm, respectively.

At maximum power, the proton beam produced by the ECRIS and acceleration column is potentially ∼5 kW, so all sensitive beam line hardware is protected by tungsten apertures. To accommodate this, we designed new chilled-water-cooled beam stops comprised of a sandwich of two copper plates. We machined the beam-facing plate with an interior 6 mm wide, 10 mm deep spiral channel, as displayed in Fig. 7(a). Chilled water from the opposite second plate enters this spiral at the center and exits at its outer radius. The spiral-channel plate was braised vacuum-tight in a hydrogen furnace to the facing second plate and covered with a 0.5-mm-thick tungsten sheet.

This water-cooled sandwich is attached via a bellows to an air-operated piston that allows it to be inserted into its beam line housing when needed. Surrounding the housing is a frame containing eight NdFeB magnets²² attached to the moving piston-bellows assembly [Fig. 7(b)]. When this assembly is inserted, the transverse B-field produced on-axis by these magnets is sufficient to suppress secondary electrons produced by the beam.

We placed a high-power beam stop followed by a 7.5-cm-aperture beam profile monitor²⁶ just before analyzing the magnet. After a 15° deflection at the magnet, the diverging proton beam must be refocused to the target 4.4 m downstream (see Fig. 1). To accomplish this, we added a second high-power beam stop and beam profile monitor before the second solenoidal lens. A third high-power beam stop was installed just before the target. The average areal target beam power density is reduced by rastering coil currents in the final magnetic steerer to sweep the beam horizontally and vertically ∼±1 cm at the target position.

Because the beam line between the bending magnet and the target must also provide efficient transport of the higher energy proton and ⁴He beams from the 1 MV accelerator,¹ the original 5-cm-aperture quadrupole lens was kept in place. The beam line following the bending magnet also contains new variable-aperture, water-cooled slits used primarily for precise momentum analysis of the higher energy JN accelerator beams.

With the new acceleration column, the ECRIS at LENA has produced stable, higher intensity beams. The intense bremsstrahlung X-ray levels produced with the old system are now comparable to environmental background and sparking between electrodes has also been mitigated. Both results point to reduced beam loading during normal operation and more efficient transport of the extracted beam to ground. Critical ECRIS tuning parameters for a 30 keV extracted proton beam, ranked in descending order of importance, are the gap spacing, s_PE, between the plasma and extraction electrodes; axial position of the (yellow) permanent magnet array surrounding the plasma chamber (see Fig. 5); source gas pressure; and injected microwave power. Experience shows also that achieving high target beam current is quite sensitive to focus electrode voltage because it improves optical matching with the downstream solenoidal lenses.

During normal operation, extracted beam currents of 4–9 mA are measured from the leakage current on the plasma chamber power supply. Typical transmission of this current to target ranges from 10% at 100 keV to 80% at 200 keV. Target proton currents in excess of 3.5 mA have been achieved, which now often exceed the power our chilled-water-cooled targets can sustain. Thus, beam pulsing as described in Sec. II has become routine, both to limit the average target current and reduce the experimental background. Pulse repetition rates between 1 and 20 Hz, and pulse duty cycles between 100% (DC) and 10% have been used, depending on experimental requirements. Optimizing tuning parameters to attain highest average currents and square-top pulses like those found in Fig. 2 becomes harder at higher repetition rates and is best achieved using simultaneous oscilloscope and analog meter beam current displays.

The extraction region and acceleration column pressures tend to rise during pulsed-beam operation. On average, shorter pulse widths result in more neutral gas and fewer ions emerging from the ECR plasma chamber. Since beam current and column gas pressure control beam-induced gas ionization, pulsing can increase electron loading, associated electrostatic instabilities, and the possibility for column sparking, especially at higher beam energies. Because this system is evacuated by a single, 2200 l/s turbopump at the ground end of the column, experience shows that these situations require larger leakage current in the DI water channels.

Though operated typically to provide proton beams, the ECRIS at LENA has provided beams of heavier species for ion implantation into host materials that are later used as targets for experiments. Usable beams of deuterium, neon, and argon have been accelerated and momentum analyzed before implantation. Milliampere-levels of 20–100 keV^20,22 Ne⁺ have been obtained at the 45°-left exit port of the magnet (see Fig. 1), while smaller currents of ²⁰Ne²⁺ and ²⁰Ne³⁺ are available. Also, several microamperes of 240 keV ⁴⁰Ar⁴⁺ have been obtained.

The tuning conditions of both the ECRIS and beam-transport system are generally reproducible from day to day, and operation for periods of up to a year has been achieved without opening the ECRIS for maintenance. When maintenance is needed, the most frequent repair is replacement of the sputtered extraction aperture and the BN window inside the plasma chamber.

When the ECRIS and acceleration column have been exposed to atmosphere or not used for several days, both usually require conditioning to achieve optimal performance. Prior to use for an experiment it is necessary to outgas the source and remove impurities (e.g., water vapor) in the BN and on interior surfaces of the plasma chamber by striking a plasma discharge and extracting a low-energy beam for up to 8 h.

Conditioning of the acceleration column without the beam is also necessary to remove adsorbed gases. High voltage is applied incrementally to the air-insulated table while the pressure within the column and the leakage current to ground are monitored. This leakage current is composed both of charge carried by the DI water channels and ionized surface contaminants stripped from the electrodes and accelerated to ground. Several hours of this gradual conditioning may be required before the desired operating voltage and column current load are met. The temperature and conductivity of the deionized water supply are also adjusted during this process.

Beam emittance is useful for characterizing the optical quality and transport efficiency of charged-particle beams. Energy-normalized beam emittances allow for a quick optical efficiency comparison of various ion sources and acceleration systems. A simple emittance measurement method, using a scan of the beam via a profile monitor and a corresponding shadow signal on a downstream Faraday cup, which is too small to capture the entire beam, provides the necessary information to calculate emittance. Furthermore, this simple measurement²⁷ facilitates the study of the ECRIS beam emittance as a function of various ion source parameters.

The necessary instrumentation for these measurements includes a well-understood beam profile monitor (BPM)²⁶ and a 2.54-cm-diameter Faraday cup, placed downstream at a distance z (1.10 m). The current signal of the Faraday cup is read out using an oscilloscope triggered by the magnetic pickup of the rotating wire shaft of the BPM. A beam scan is taken by the BPM, while the beam current signal on the downstream Faraday cup is simultaneously recorded (see oscilloscope traces in Fig. 8). Gaussian fits to these measurements yield the beam radius at the BPM (r_b) and the radial extent of the beam reaching the Faraday cup that was briefly blocked by the rotating wire of the BPM. The difference between the Faraday cup radius and that of the occulted beam, divided by z, yields the angular divergence of our beam (r′). One must ensure that the beam is focused to a symmetric waist at the BPM to satisfy the assumption that the transverse position-momentum phase-space ellipse of the beam is vertically upright. In accordance with the definitions given by Reiser²⁸ for beam emittance and brightness, this measurement configuration gives

where the phase-space area is given by A_r = π · ϵ. From Eq. (2), the normalized beam emittance and brightness are

and

where the non-relativistic limit is assumed, β = $v$/c, and I is taken as the leakage current on the plasma chamber power supply.

Ideally, an infinitesimally thin scanner wire would be used to measure a true trace of the Gaussian beam profile. However, when a wire of finite thickness is employed, the measured signal is temporally broadened. Hence, these normalized emittance and brightness results represent upper and lower limits, respectively.

We report in Table II normalized emittance and brightness measurements at 30 keV and 100 keV, and at different beam extraction region pressures and extracted currents (P_extr and I_extr, respectively). The 30 keV tests only incorporated the extraction gap as a focusing and accelerating element, whereas the 100 keV test also included the first four acceleration gaps of the column and allowed us to quantify their optical effects. Horizontal scans were used to measure the beam and shadow radii since the beam was limited vertically by the aperture of the analyzing magnet. The Faraday cup was placed inside the 30°-right exit port of the analyzing magnet, thereby allowing for momentum analysis of the ECRIS beam by both the first focusing solenoid and the analyzing magnet (see Fig. 1). Beam profiles via the BPM in each of these tests showed very little contaminant ion species within our beam prior to the analyzing magnet. Uncertainty values for all measured quantities were calculated by propagation of instrumental errors.

Ions extracted from a solenoidal magnetic field undergo emittance growth when emerging through the axially symmetric fringe-field, so the beam quality produced by ECR ion sources is limited by Busch’s theorem.^29–32 Using the definition of ϵ in Eq. (2), the following lower limit for the ϵ_n of their extracted beams is

where Z is the ion charge state, A is the ion mass given in atomic mass units, B₀ is the axial magnetic field in the extraction zone in units of 10⁻⁴ T, and r is the plasma aperture radius in millimeters.

Considering the first 30 keV test case in Table II (i.e., I_extr = 7.3 mA, B₀ = 50.0 mT, r = 2.5 mm, A = 1.0078 u, and Z = 1) and using Eqs. (5) and (4), one can calculate minimum ϵ_n and maximum B_n values of 0.0159π mm mrad and 5.86 $A(π⋅mmmrad)2$⁠, respectively, and find that our smallest emittance measurement is less than three times the lower limit given by Busch’s theorem. However, this emittance measurement indicates a brightness increase of our LENA beam by at least an order of magnitude compared with the results of Cesaratto et al.⁸ who, with the prior acceleration system, measured normalized emittance and brightness values of 0.19π mm mrad and 0.06 $A(π⋅mmmrad)2$⁠, respectively. This brightness increase is a consequence of over a factor-of-4 lower beam emittance, not an increase in extracted beam current. Furthermore, comparing the results at 30 and 100 keV, the first four acceleration gaps of the column seem to cause little emittance degradation.

Performance highlights following the ECRIS upgrades described above include the successful implementation of routine pulsed-beam operation; the total elimination of harmful beam-induced bremsstrahlung radiation; long-term operation without high-voltage breakdown; and over 3.5 mA of target beam current (1.75 times the prior value). The normalized beam emittance of this system is smaller by at least a factor of 4, and the normalized brightness is larger by over an order of magnitude, in comparison to our prior system. These optical improvements led to a higher beam transport efficiency to target. We also found that this system can produce mA-range heavier ion beams that are useful for implantation applications. Below, we describe future beam tests and optical studies with this system.

Our extracted beams indicate a lower current density at the plasma aperture than exhibited at LANL¹⁶ or Neutron Therapeutics.¹⁰ Recent measurements of the permanent magnet solenoidal field over the ECR chamber indicate that its peak axial magnitude has dropped by 6 mT when compared to measurements by Cesaratto et al.⁸ Presumably, this is because of accidental over-heating of the NdFeB permanent magnet arrays in the solenoid. This decrease in field strength has led to ECR zones within the plasma chamber that no longer intersect its axis but are rather cylindrical regions around that axis. We plan to install a new magnet with an improved axial magnetic field profile and ECR zone geometry, with the goal of increasing extracted beam current and brightness.

The focal properties of the acceleration column were optimized for 30 keV injected H⁺ beam energy. Having gained extensive operational experience under these conditions, we wish to explore further the effectiveness of increasing the injected energy of the beam toward 40 keV to raise the beam current extracted from the plasma chamber. Injecting higher energy beams into the column will require somewhat higher focus electrode voltages to optimize target beam currents.

The measurements in Fig. 8 represent the convolution of the assumed Gaussian beam signal with the sum of two opposing Heaviside functions.³³ Their sum represents the time-width of the opaque, hard-edged scanner wire, and its convolution with a Gaussian is equivalent to the sum of two Gaussian error functions.³⁴ Proper analysis of these emittance data must account for this since the measurements presented in Fig. 8 were collected with a 0.75-mm-radius wire and, therefore, over-estimate the beam radius. In the future, we also wish to study the emittance of the ECRIS beam as a function of various source tuning parameters with an analysis that includes this wire-thickness correction.

The results in Table II indicate lower emittance for higher gas pressure, as measured with similar ion sources (e.g., Roychowdhury et al.³⁵). Pressure values were recorded with a vacuum gauge placed just past the beam extraction gap. We suspect that space-charge neutralization following beam extraction is occurring within the conical expansion region. Further investigation of the dependence of beam emittance on energy and other source parameters is planned.

We would like to acknowledge the contributions of S. Hunt for his help with assembly and beam line design advice; L. Downen for her assistance with beam testing and emittance measurements; G. Rich for helping with the installation of our high-power microwave system; C. Caldorado for his help with column assembly; J. Martin for his collaboration on extraction system alignment challenges; S. Brogan for providing various laboratory diagnostic tools during column construction; and C. Westerfeldt, T. Calisto, J. Dunham, and R. O’Quinn for their invaluable technical advice. We also recognize the Duke University Instrument Shop for always being available for the countless, small machining jobs in the final stages of this project. This work was supported in part by the U.S. Department of Energy under Contract Nos. DE-FG02-97ER41041 (UNC) and DE-FG02-97ER41033 (Duke).