A versatile setup for hydrogen isotope permeation studies

Hydrogen isotope (HI) migration and retention in plasma-facing components and structural materials affect the safety and tritium self-sufficiency of future fusion power plants as well as the tritium doubling time after which an additional reactor can be started with the tritium surplus. In order to reliably predict HI transport phenomena in plasma-facing materials, a detailed understanding of potentially relevant processes, such as interstitial and grain boundary diffusion, trapping and de-trapping from defects, and surface or interface transfer rates, is necessary. A suitable way of studying these effects is to perform permeation experiments.

We present here a new, versatile permeation experiment dubbed TAPAS (Testbed for Analysis of Permeation of Atoms in Samples), which is attached as a second beam line to the mass-filtered, monoenergetic ion beam setup SIESTA.¹ The main focus of TAPAS lies in performing ion-driven permeation experiments under as controlled as possible conditions, while at the same time providing ion flux densities and particle energies comparable to the first wall of a fusion reactor.^2,3

In addition to delivering a fusion-relevant particle spectrum, ion-driven permeation also offers some unique possibilities complementing the information that can be gained from gas-driven permeation due to the different boundary conditions for hydrogen uptake. If interstitial hydrogen diffusion in the membrane is the rate-limiting process for permeation, in gas-driven experiments the surfaces are in thermodynamic equilibrium with the surrounding gas phase, and the upstream hydrogen solute concentration depends on temperature, gas pressure, and the membrane’s hydrogen solubility following Sieverts’ law. In ion-driven permeation, on the other hand, the upstream concentration is proportional to the flux of implanted particles and independent of the membrane’s hydrogen solubility. Thus, if diffusion-limited boundary conditions can be obtained on both surfaces, the steady-state permeation flux is temperature-independent⁴ (see also Sec. VI A). This allows measuring permeation transients also at low temperatures, where the permeation signal might become immeasurably low in gas-driven permeation. In addition, e.g., in bi-layer samples, the transition of HI from one material to the other can be studied in both directions independently, whereas gas-driven permeation would result in a linear combination of permeation resistance regardless of the permeation direction.^5,6

In addition to its main feature of ion-driven permeation, TAPAS offers some capability also for gas-driven permeation at low pressure. This additional feature cannot fully replace a dedicated gas-driven permeation setup but allows, e.g., to accurately determine gas-driven offsets to the ion-driven signal and also to work in the neutral gas pressure range that will be present near plasma-facing components in a fusion reactor. Furthermore, combining both permeation methods in a single device allows us to directly compare their results for the same sample in the same environment and, thus, helps to avoid reproducibility issues.

With its high-intensity ion beam paired with a highly sensitive quadrupole mass spectrometer, TAPAS is also particularly suitable for investigating strong barriers for permeation, such as permeation barrier layers or phase boundaries with large differences in HI solubility.

In this article, we present a detailed description of the TAPAS setup as well as performance optimization of the SIESTA ion source and the TAPAS ion optics. We describe calibration procedures for the particle flux to the sample, including high-energy charge-exchange neutrals originating from the beam line, as well as for the deuterium gas pressure in front of the sample. Finally, as a benchmark experiment, we present ion-driven deuterium permeation measurements through recrystallized tungsten foil, comparing them with forward simulations and discussing implications for the actual boundary conditions present in our experiment.

The ion-driven permeation experiment “TAPAS” is an extension to the previously described ion-beam experiment for sputter erosion studies “SIESTA”¹ and is connected to its mass-filtered dipole magnet with a deflection angle of 44° at the port labeled ”PERMEX” in Ref. 1. A birds eye view, as well as a more detailed description of mutually used machine parts (ion-source, differential pumping stage, and dipole magnet for mass filtering) can be found there. The ion-source is based on the duopigatron concept shown in Refs. 7 and 8. A set of operation parameters for a deuterium beam for erosion studies as well as details about the limitation due to space charge induced divergence of the beam can be found in Ref. 1. The working gas, typically deuterium, is fed into the ionization volume of the ion source, and a hot filament emits thermionic electrons that are accelerated into the same volume in order to form a plasma by collisions with the neutral gas particles. From the ionization volume, which is held at high potential (typically +5.2 kV), multiple species of deuterium ions (D⁺, D₂⁺, and D₃⁺) get extracted via grids toward ground potential before the beam passes through a molybdenum aperture with 16 mm diameter. In a large expansion volume in front of the aperture (differential pumping stage), most of the working gas gets pumped at typically 2.6 × 10⁻⁵ mbar. The beam finally enters the dipole magnet, which directs it into one of two beam lines and provides mass filtering. In front of the dipole magnet, a retractable beam stopper can be used to interrupt the ion beam without switching the ion source.

Setting the dipole mass filter to transmit D₃⁺, we deliver three deuterons per ion onto the sample. Thus, for a given ion current and deceleration bias at the sample, we reach the highest possible deuteron flux (3× the ion current) at the lowest possible kinetic energy per deuteron (1/3 the D₃⁺ ion energy). This conforms best to our goal to reach irradiation conditions similar to the first wall of a fusion reactor. If we used other D ion species extracted from the source, i.e., D⁺ or D₂⁺, we would need a lower extraction voltage and/or higher deceleration voltage at the sample to reach the same deuteron energy, which would lead to a penalty in the ion current arriving at the sample. In addition, these ion species deliver less raw ion current (i.e., measured without deceleration) in the first place, with also fewer deuterons per ion, which makes D₃⁺ by far the most efficient ion species for our purpose. However, for dedicated applications (e.g., high-energy ion implantation) and after appropriate tuning of the ion-optical system in TAPAS, the implantation of D⁺ or D₂⁺ can be realized as well. Mass spectra of the ion beam before mass filtering for different working gasses (H₂, D₂, and H₂ + D₂) of the ion source are shown in the supplementary material.

After mass separation, the monoenergetic D₃⁺ ion beam enters the TAPAS loading chamber, which is represented in Fig. 1. The beam passes the beam focusing system consisting of an electrostatic einzel lens (inner ø33 mm, axial gap 3 mm) as well as two vertical and two horizontal sweeper plates with a gap of ∼25 mm in each direction. The beam then is decelerated by a set of electrodes (“deceleration system”) consisting of a tungsten deceleration aperture (ø16 mm) about 84 mm in front of the target and a final focus lens. The final focus lens, which consists of a 48 mm steel tube terminated by two tungsten apertures (upstream ø17 mm, downstream ø15 mm), is located between the deceleration aperture and the sample holder. The axial gap between the final lens and the sample holder is about 7 mm. By decelerating the particles only shortly before the target, beam divergence due to space charge is minimized. All electrodes can be biased up to +6.5 kV with a maximum current of 1 mA (FUG, HCE 7-6500). To avoid the accumulation of charge on the electrodes above the intended potential due to impinging beam ions, all electrodes are connected to ground potential via electrical resistors of 10 MΩ.

The sample holder is a modified commercial DN16 CF straight spacer tube, where the knife edge of one CF flange has been removed on one side. This allows for clamping the sample membrane using a ø19.6 mm spring-loaded HELICOFLEX^® (Technetics Group) metal seal, which compensates for thermal expansion during temperature cycling. The main jacket of the seal is about 200 μm of Inconel. The downstream side of the jacket, which includes the sealing surfaces, is plated with 35 μm of silver by the manufacturer to reduce hydrogen permeability and provide a soft material for sealing with minimal force. The circular sample membrane has a diameter of 21 mm and is gently pressed onto the seal by a flat stainless steel ring (modified CF16 flange) held down with 6 × M4 screws, which are tightened at a defined torque. For fragile samples, such as recrystallized W, a torque of ∼2 Nm turned out to be optimal. On the beam-facing side, the steel ring is coated with 3 μm of tungsten to minimize erosion by impinging beam particles. The sample holder tube is connected to the analyzing chamber via an additional CF16 spacer tube. A ceramic insulator separates the sample holder from ground potential in order to measure the ion current and to apply a bias voltage to the target, i.e., define the energy of the impinging beam particles. A variable bias voltage in the range of ±6.5 kV can be applied by a bipolar power supply (FUG, HCB 7-6500) at <1 mA. Details about the calibration of the ion current measurement can be found in Sec. V B. To reduce the deuterium permeability of the sample holder, the CF16 parts have been coated on the upstream side alternatingly with two alumina and two yttria layers, each to a total oxide thickness of about 200 nm, whereas yttria is the topmost layer.

The sample heater cassette holds several windings of molybdenum resistor wire wound around six ceramic supports. Tungsten sheets around the winding pack serve as heat shields to protect the chamber walls and homogenize the temperature field inside the cassette. The sample membrane and the clamping flanges are centered in the heater cassette such that, in thermal equilibrium, a flat temperature profile can be assumed. Sample temperatures from room temperature to 1000 K can be realized. A K-type thermocouple is clamped to the modified CF16 tube next to the membrane to accurately measure the sample temperature for feedback control. Fast temperature changes of the membrane, e.g., when ion implantation starts or stops, are precisely monitored using a pyrometer working at a wavelength of 2.3 μm. These temperature excursions are not counter-regulated by the feedback control since they would lead to oscillations of the sample temperature, but they can later be considered in the data evaluation by numerical modeling. The pyrometer is mounted outside of the analyzing chamber and views the downstream surface of the sample membrane through a window and the sample holder tube, as shown in Fig. 1. Since the emissivity of the membrane’s surface depends on temperature and may be influenced by surface contaminations, the pyrometer is calibrated using the thermocouple reading each time the thermal equilibrium of a new temperature set-point is reached. The variation of the sample temperature during an experiment at each temperature set-point is small (<5 K, e.g., due to heating from the impinging beam particles); hence, no re-calibration of the pyrometer during the experiment is necessary.

The neutral gas pressure in the loading chamber is measured using a gas-type independent capacitance manometer with a resolution of ±1 × 10⁻⁴ mbar (Pfeiffer, CMR 364) in the high-pressure range and a full-range pressure gauge (Pfeiffer, PKR 261) in the low-pressure range. To minimize possible influences of thermal radiation from the heater or charged beam particles on the pressure measurement, the gauges are located in a small recessed volume, which is connected to the loading chamber without a direct line of sight to the beam or heater (Fig. 1). In the same recessed volume, a fine dosing needle valve allows for an additional neutral gas inlet from a minican (oxygen, deuterium, etc.). By this, gas-driven permeation experiments with upstream pressures between 1 × 10⁻⁶ and 1 mbar as well as surface oxidation studies can be realized. The loading chamber is pumped by a turbomolecular pump (TMP) type “Leybold TURBOVAC 350i” with a pumping speed of 350 l/s for hydrogen and a base pressure in the order of 1 × 10⁻⁷ mbar is achieved. The open design of the heater cassette and the tube shaped sample holder facilitates the pumping of recycled beam particles that desorb from the target as neutral gas. Accumulation of neutral gas in front of the target would increase the number of collisions with beam particles and potentially deteriorate the beam performance. Details about the measurement and distribution of neutral gas pressure during ion loading will be given in Secs. IV and V D.

Implanted particles permeate through the sample membrane and desorb into the analyzing chamber, where they are detected using the quadrupole mass spectrometer (QMS) system “QMG 422” from Pfeiffer. Details about the operation, calibration, and linearity of the QMS will be shown in Sec. III. To facilitate the pump-down process after a sample exchange, the analyzing chamber is divided into three sections by valves. The sample holder tube connects to the first section, which can be connected to the loading chamber via a bypass. During sample changes, only the loading chamber and the first section of the analyzing chamber need to be vented. Since the sample membranes are fragile, it is crucial to maintain equal pressure on both sides during pump-down. To achieve this, the bypass is opened to pump the downstream side of the sample membrane through the loading chamber. Once the pressure on the downstream side of the sample has reached about 1 × 10⁻⁴ mbar, the bypass is closed, and the valve between the first section and the ultra-high vacuum of the second section is opened. The second section is then pumped by a TMP type “Leybold TURBOVAC 90i” until it reaches a base pressure of the order of 1 × 10⁻⁹ mbar. Then, the valve to the third section with the QMS and a TMP type “Agilent TwisTorr 304 FS” is opened. The base pressure in the third section reaches <1 × 10⁻¹⁰ mbar. The chamber walls of all three sections of the analyzing chamber can be baked up to 450 K by external heating bands to reduce the pumping time. A typical pumping time after exchanging the sample membrane is about 72 h. For venting the sample into the atmosphere, the TMP in the loading chamber, the valve between the dipole magnet and loading chamber, as well as the valve between the first and second sections of the analyzing chamber, are closed. Subsequently, the bypass is opened to vent both sides of the sample with nitrogen.

To facilitate monitoring and control of long term experiments over days and weeks, all relevant components are remotely accessible via a LabVIEW integration.

The QMS is equipped with a secondary electron multiplier (SEM) and operates in ion counting mode at a typical working pressure of 1 × 10⁻¹⁰ mbar. For absolute calibration of the deuterium sensitivity, a small calibrated deuterium leak of 6.9 × 10¹⁰ D₂/s is fed into the first section of the analyzing chamber, i.e., close to the sample holder. To increase the ionization probability, the cross-beam-type ionization unit was modified with an electron collimator magnet available from the manufacturer. The achieved sensitivity of the measurement setup is 9.1 × 10⁷ D₂/count. The background signal of the QMS with the valve to Sec. III of the analyzing chamber closed is <0.01 counts per second (cps). The background of the inside of the sample holder without upstream deuterium loading is dependent on temperature and dominates the background signal with typical values between 1 and 10 cps. For a typical integration time of 20 s, the background noise has a standard deviation of 0.7 cps. Considering this background noise as the sensitivity limit, permeation fluxes >6.4 × 10⁷ D₂/s can be detected. Considering the permeating area of 33 mm² in ion-driven permeation experiments (see Sec. V C), this sensitivity limit is reached at about 4 × 10¹² D₂/m²;s, corresponding to about 4 × 10⁻⁷ monolayers per second. As will be shown in Sec. VI B, typical ion-driven permeation signals are about ×2700 higher than the sensitivity limit of the QMS.

In the low signal regime, linearity between deuterium partial pressure in the ionization unit of the QMS and ion count rate in the SEM is assumed. However, at higher partial pressures, non-linear behavior due to space charge effects in the ionization unit, dead time of the ion counting unit, or warming of the secondary electron-emitting surfaces in the SEM must be excluded by explicit measurements. For this purpose, the QMS signals resulting from deuterium gas flux from the loading chamber to the analyzing chamber via a thin capillary sample have been measured for various upstream pressures (Fig. 2). Using a focused ion beam, the capillary with a diameter of ∼10 μm was sputter eroded at an angle of 45° into a polycrystalline tungsten foil of 25 μm thickness, leading to a capillary length of about 35 μm. The capillary sample was mounted temporarily onto the sample holder in the same fashion as the permeation samples. For upstream pressures of 1 mbar, the mean free path of the deuterium molecules is about 100 μm. Since all dimensions of the capillary are much smaller than this, there is on average at most a single deuterium molecule inside the capillary at any given time, so that molecular flow through the capillary can be assumed at pressures below <1 mbar. The upstream pressure was applied with closed TMP in the loading chamber and monitored using the capacitance manometer. The red line in Fig. 2 is a linear fit through the origin, and its slope, the effective conductance, is proportional to the conductance of the capillary and the sensitivity of the QMS. The scatter for upstream pressures <1 × 10⁻³ mbar is due to the limited resolution of the capacitance manometer. The QMS signal shows excellent linearity with the upstream deuterium pressure even for high count rates.

The capacitance manometer is not sensitive to pressures below 1 × 10⁻⁴ mbar, while, during ion-driven permeation or low-pressure gas-driven permeation experiments, the pressure in the loading chamber can be as low as 1 × 10⁻⁷ mbar. For this pressure range, the full-range pressure gauge needs to be used. However, the readings of the full-range gauge, which combines a Pirani gauge and a cold cathode ionization gauge, are gas-type dependent. Furthermore, during low-pressure experiments, the TMP in the loading chamber near the gauge is typically open and may cause an inhomogeneous pressure profile between the sample and the gauge. Therefore, calibration is required not only to accurately account for the influence of the gas type but also to compensate for potential pressure gradients in the loading chamber.

For the calibration, the deuterium flux through the same capillary sample as described in Sec. III was used. Assuming the linear behavior of the QMS in the low signal regime, the real deuterium pressure in close proximity to the sample can be calculated using the experimentally determined effective conductance (Fig. 2) of that capillary. In Fig. 3, the real upstream pressure in close proximity to the sample, which is derived from the mass four QMS signal and the effective conductance of the capillary, is plotted vs the corresponding full-range gauge readings for a wide range of D₂ pressures in the loading chamber. Different pressures were achieved by varying the deuterium gas flow and the pumping scheme (TMP open or closed). The internal transition of the full-range pressure gauge from heat conductivity (Pirani) to ion current measurement (cold cathode) is clearly visible in the pressure range of around 1 × 10⁻³ mbar, as specified by the manufacturer. When the TMP is closed, the loading chamber is pumped only via the dipole magnet chamber and the differential pumping stage, and a homogeneous pressure profile within the loading chamber is expected. As shown in Fig. 3 comparing the red and black data points, the relation between capillary flux (“real pressure”) and gauge reading is identical for both pumping schemes. This indicates that an active TMP does not create a significant pressure gradient between the sample and the pressure gauges. The real deuterium pressure in close proximity to the sample is about 0.8× the uncalibrated reading in the high-pressure regime (heat conductivity measurement) and 3.4× the uncalibrated reading in the low-pressure regime (cold cathode). Those factors lay well within the specifications of the manufacturer for the influence of the gas type but reflect a more accurate correlation. Applying the calibration of the full-range gauge, as shown in Fig. 3, the upstream deuterium pressure in close proximity to the sample can be accurately determined for pressures between 1 × 10⁻⁷ and 1 mbar.

The original machine settings of the ion source for erosion studies as shown in Ref. 1 have been adjusted for the usage of the D₃⁺-beam in the TAPAS beam line for permeation studies. A detailed description of the adjustments made to the original ion source settings as well as the procedure for optimizing the parameter space of the electrodes in the loading chamber is provided in the supplementary material. The material of the hot filament of the ion source has been changed to increase lifetime, beam stability, and reproducibility. The voltages of the electrodes of the einzel lens, sweeper plates, and deceleration system were optimized to satisfy the following criteria in a well-balanced compromise:

• Low ion energy.

• High ion flux.

• Homogeneous beam footprint.

• Low number of charge exchange neutrals (CXNs).

• Low number of secondary electrons (SEs).

The chosen set of parameters for a 170 eV/D D₃⁺-beam for permeation studies in TAPAS is given in the supplementary material. In this section, a detailed characterization of the ion beam is presented.

The U–I characteristic of the selected parameter set was determined by varying the bias voltage of the sample holder from 0 to 5.3 kV while measuring the required drain current during beam operation, as illustrated in Fig. 4. The target consists of a quartz plate covered with a nickel mesh to prevent electrostatic charging by the ion beam. The ionoluminescence of the quartz was used for qualitative real-time observation of the beam footprint using a CCD camera with a line of sight through the sample holder tube and a window in the downstream chamber (this system replaced the pyrometer during beam optimization; see also the supplementary material). A positive current indicates the flow of electrons from the power supply to the sample holder, compensating for the impinging D₃⁺ ions to maintain the sample bias voltage set-point. The energy per deuteron is calculated as the difference between the beam potential (5.2 kV) and the sample bias, divided by 3, the number of deuterons in the D₃⁺ molecular ion. The inset provides detailed information about the behavior for bias voltages close to the beam potential.

The measured current exhibits a stable plateau of ∼150 μA for a wide range of ion energies in the high-energy regime, followed by a relatively sharp drop observed as the sample bias approaches the beam potential. Within the plateau, the shape of the footprint, as observed by the camera on the quartz, showed only little dependence on the deceleration voltage. However, when decelerating the ions to energies below 200 eV/D, the footprint begins to widen with increasing deceleration voltage, possibly due to a change in direction of the electrical field lines between the sample holder and the final focus lens or due to stronger beam divergence (space charge) at lower ion velocities. Based on the U–I characteristic and qualitative shape of the footprint from the quartz image, an optimum working point for low ion energy has been identified at 170 eV/D, which is well below the sputtering threshold energy for deuterium in tungsten.⁹ 

While the ion currents for lower ion energies would still be acceptable in terms of resulting permeation signal intensities, reliability would be reduced due to increasing fractions of the beam particles impinging on the sample holder and not the sample as a result of the stronger divergence. In addition, at lower ion energies, the beam appears to be less stable, resulting in fluctuations in the measured ion current.

As shown in the inset in Fig. 4, the zero crossing of the U–I curve of the sample holder occurs at a bias voltage very similar to the source potential of 5.2 kV. The plasma potential in the ion source, where ions are generated from collisions with 200 eV electrons, could explain the slight difference (≈+25 V) between the potential of the zero crossing and the nominal beam potential. The uncertainties of the voltages of the high voltage power supply for the extraction grids as well as of the high voltage power supply for the sample bias as stated by the manufacturer are in the range of ±10 V.

When biasing the sample holder above the beam potential, a small current with the opposite polarity of about −15 μA is established. This is assumed to be due to SE that are created from beam ions (E = 5.2 kV) impinging on the final focus lens (4.74 kV), which are then accelerated onto the sample holder due to its higher potential. However, since the sample holder has a 50 V lower potential at the working point established above (4.69 kV), those electrons are expected to be redirected back to the final focus lens and not to significantly influence the ion current measurement on the sample holder (see also Sec. V B).

The wide range of ion energies for which the ion current is independent of the deceleration potential and the sharp drop of the ion current only at low ion energies show the functionality of the chosen geometry of the deceleration system.

The electrical current on the sample holder is a measure of the incident ion flux. Electrons from upstream sections of the beam line are being filtered out due to the high potential of the final focus lens, as described above. However, SE emitted from the sample upon ion irradiation might influence the ion current measurement, i.e., over-estimate the real number of impinging ions due to secondary electrons escaping to the final focus lens. Due to space restrictions within the heated zone, ion current measurement using a retractable Faraday cup or a Faraday cage around the sample is not feasible. To correlate the measured current on the sample holder with the real number of impinging ions, a calibration of the measured sample current based on sample erosion by physical sputtering was performed. A bulk copper sample of about 1.9 mm thickness was polished to mirror finish and ultrasonically cleaned in ethanol. Markers were scratched onto the surface with a fine diamond tip. To remove possible swarfs from marker scratches, the final polishing step was done after marker scratching. The sample and a reference sample were weighted with a precision balance (“Sartorius MC21S,” typically achievable an accuracy of ±1 μg) multiple times over several days after cleaning. The sample was then eroded for about 48 h using the beam configuration described above (170 eV/D). The average measured current of 153.3 μA during copper erosion was almost identical to the current measured on the quartz at the working point (see Fig. 4). To determine the weight loss due to physical sputtering, the sample and the reference sample were weighed directly after breaking the vacuum as well as after several days of exposure to the atmosphere to estimate the influence of possible oxidation. The drift of the balance due to changing atmospheric humidity and pressure was estimated using the weight measurements of the reference sample and found to be negligible (<1 ppm). The influence of dust particles and/or surface adsorbates was estimated by weighing multiple times before and after cleaning with ethanol. The potential loss of material due to clamping in the sample holder and handling was estimated by repeating the procedure without erosion. All variations were in the range of the typical reproducibility of the measurement method (±10 μg). From initially ∼5 g, a total of 2.153 mg of copper was eroded. Due to the large amount of eroded copper, the total influence of the uncertainties discussed here on the mass loss measurement is well below 1%.

As will be shown in Sec. VI C, the fraction of CXN in the ion beam is about 1% and has an energy of 1733 eV/D. The sputter yields for 170 eV (ions) and 1733 eV (CXN) deuterium on copper were interpolated using the empirical fitting function proposed by Bodhansky¹⁰ on measured sputtering data from Eckstein et al.⁹ The yields are (3.5 ± 0.4) ×10⁻² Cu/D and (8.2 ± 0.3) ×10⁻² Cu/D, respectively. The small fraction of CXN had to be considered due to the larger physical sputtering yield compared to the ions. The resulting calibration function for the ion current is I_ions = 1.14 × I_measured. The proportionality between impinging deuterium flux and measured sample holder current is, therefore, 2.13 × 10¹⁹ D/sA.

To clarify the influence of SE from the sample holder or the final focus lens on the ion current measurement, the procedure has been repeated for a beam configuration with an identical sample holder bias but a lower potential of the final focusing lens, i.e., at 4.52 kV instead of 4.74 kV. This results in a reversed polarity of the electrical field between sample holder and final focus lens and, if significant amounts of SE are created at the target, should affect the measured sample holder current. The measured calibration factor for the reversed polarity case was 1.17 and is very similar to that of the previous case (1.14). Hence, it is assumed that the influence of SE on the ion current measurement is small, at most about 3%. In addition, for identical beam configurations, the measured ion current for the different target materials was very similar. In particular, the quartz plate covered by a nickel mesh measured 151 μA (see Fig. 4), while the copper erosion sample measured 153.3 μA (discussed in this section). This further supports the assumption of a small influence of SE on the ion current measurement.

It is assumed that the calibration factor differs from 1, mainly due to the limited accuracy of the sputter yield measurements in Ref. 9. It is also plausible that the samples investigated in Ref. 9 had a slightly different microstructure, i.e., crystallinity and distribution of grain orientations, than the copper samples used here. As shown in Ref. 11, the crystal orientation of the target can significantly influence the sputter yield. To estimate the potential influence of the microstructure of the copper samples on the calibration of the ion current measurement, we calculated the sputter yields for 170 eV deuterium on copper with SDTrimSP 7.00.¹² This version of the software is able to simulate both amorphous and crystalline targets. We calculated two cases: The first case was an amorphous target, where all projectile–target collisions are randomly chosen in a Monte Carlo process and resulted in a sputtering yield of 4.2 × 10⁻² Cu/D. The second case was a polycrystalline target, where first the distribution of the various grain orientations was measured via electron backscatter diffraction (EBSD) on the copper erosion sample used here, and subsequently, the mean value of sputter yields from a representative subset of grain orientations was calculated. The resulting average sputtering yield was 3.0 × 10⁻² Cu/D. The difference between the calculated sputter yields for amorphous and crystalline copper exceeds the uncertainty of the measured data from Ref. 9, for which no information about the target’s microstructure is available. Therefore, as a more conservative estimate, the range of the two calculated sputter yields is taken as the uncertainty for the ion current calibration.

The rise and decay transients at the beginning and end of a permeation experiment give insight into the transport of HI through the material. If a substantial number of traps are present, the transients do not only depend on the interstitial diffusivity but also on the density and properties of the traps as well as on the ion flux density. For high ion flux density, defects will be saturated faster than for low ion flux density, leading to a faster break-through time. If the impinging ion beam is inhomogeneous, which in practice is often the case, the transients become a superposition of regions with different ion flux densities. Furthermore, for temperatures where dynamic trapping and de-trapping are happening simultaneously during steady-state permeation, the local fill level of trapping sites in the sample depends on the local solute concentration, i.e., local implanted ion flux density. In permeation experiments with samples where sputtering cannot be neglected, the useful lifetime of the sample is limited by erosion in the high-flux region, i.e., where critical fluence is reached first.

A horizontal line profile, representing the vertical average between the two white lines through the center of the beam spot, is shown in the top plot. As shown in the line profile, the beam profile can be described as peaked. The granular appearance within the beam spot is due to the grain-orientation-dependent sputter yield¹¹ in the polycrystalline copper sample and does not indicate a lateral variation in the ion flux density. The periodic vertical stripes in the center of the beam spot are an effect of the laterally periodic beamlets emitted from the extraction grid of the ion source and do reflect a lateral variation of the ion flux density. Since the dipole magnet focuses only horizontally, the vertical periodicity of the grid is de-focused at the sample. The dimension of lateral ion flux density variation (≈1 mm) is much larger than the typical sample thickness (50 μm). Hence, significant smoothing of the lateral deuterium solute concentrations toward the downstream surface by lateral diffusion is not expected. Suppression of such features and smoothing of the footprint, e.g., for studying flux- and fluence-dependent surface modification of samples, could be achieved by de-focusing the ion beam, which would also reduce the total ion current, or by wobbling the beam position, which would locally lead to periodic ion flux variations.

The number of particles that are cutoff by the deceleration system and, consequently, the creation of SE and CXN is avoided as far as possible by minimizing the ion currents onto those electrodes during beam optimization. The circumference of the footprint is, therefore, slightly irregular and does not resemble the circular shape of the elements of the deceleration system. The outer edge of the crater shows rather large areas of low erosion from halo trajectories of the beam [red areas in Fig. 5(a)]. Suppression of such halo regions could be achieved again at the expense of total ion current if necessary.

We do not observe the footprint of the CXN on the height data in Fig. 5. CXN does not get focused and should create a footprint that is limited by the area of the last aperture (ø15 mm). However, assuming that the 1% CXN (see Sec. VI C) impinges on the sample homogeneously within an area limited by the last aperture, the flux density of the CXN yields 2 × 10¹⁷ CXN/m²s and is much lower than the flux density of the ions (Fig. 6). Therefore, despite the larger sputter yield of the CXN as compared to ions, no significant features of erosion by CXN are expected in Fig. 5.

The total measured volume of the crater was 15% smaller than the expected volume loss considering the measured mass loss and the atomic density of copper. The influence of CXN on the mass loss has been considered. This slight disagreement is assumed to be due to the subtraction of the background from the height data, the limited resolution of the CLSM measurement, as well as errors due to automated lateral stitching of adjacent measurement areas. However, this does not significantly influence the shape of the histogram but mainly causes a slight offset on the x-axis in Fig. 6. Therefore, the x-axis of the histogram has been converted to ion flux density such that the integral of the histogram corresponds to the average ion flux density as determined by the mass loss measurement (see Sec. V B) normalized by the total ion beam area.

Figure 7 shows the mean ion flux density for different fractions of the beam ions in a percentile plot. The mean ion flux density (black curve) was determined using the sputter yield from Ref. 9, as described in Sec. V B. The uncertainty range reflects the limiting cases based on sputter yields calculated using SDTrimSP 7.00¹² for amorphous and polycrystalline copper targets as described in Sec. V B. As shown in Fig. 7, 95% of the beam particles impact regions of an average flux density larger than 0.9 × 10²⁰ D/m²s and are localized within 50% of the total ion beam area, i.e., an area of 33 mm². From here on, this 95% percentile will be used as a practical representation of the beam, e.g., to calculate the nominal sample fluence for a given beam time. Parts of the beam, however, show much higher ion flux density. For instance, as indicated by the left dotted line in Fig. 7, 10% of the beam particles are wetting only 1.7% of the total beam area, which results in an average ion flux density of 2.9 × 10²⁰ D/m²s in that region.

To illustrate the influence of the distribution of ion flux densities shown here on the transients in ion-driven permeation experiments, simulations for a 50 μm thick recrystallized tungsten sample between 600 and 900 K were performed using the diffusion trapping code TESSIM-X.¹³ A trap site density of 1.5 × 10⁻⁵ per tungsten atom and an activation energy for de-trapping of 1.5 eV as determined in Ref. 14 were used. An attempt frequency for de-trapping of 10¹³ Hz was assumed. As computational time increases proportionally to the number of bins describing the histogram, a reduced histogram (100 bins) was used for the calculations. It was found that the much faster modeling approach using the nominal mean implantation flux density describes the transient permeation in recrystallized tungsten sufficiently accurately. However, in samples with larger trap site densities (1.4 × 10⁻² as typical, e.g., for self-damaged tungsten¹⁵), the calculated transient permeation flux using the simplified nominal mean implantation flux density deviates significantly from the more detailed case using the histogram of implanted flux densities. Accordingly, for such materials, the histogram of ion flux densities should be included in the model, or a more homogeneous beam configuration needs to be chosen to minimize the effect of lateral ion flux density distribution on the measurement of transient permeation.

The geometry of the sample holder, the heater cassette, including heat shields, and the deceleration system were designed to allow for efficient pumping of the recycled deuterium that desorbs from the upstream sample surface during ion beam operation. This is to reduce collisions of the beam particles with neutral deuterium gas in front of the sample, which could reduce the beam performance by creating CXN, reducing the number of implanted ions, and increasing the beam’s divergence. Furthermore, particularly at high temperatures, even the build-up of a low deuterium neutral gas pressure in front of the sample during ion-driven permeation experiments can lead to gas-driven permeation that may significantly contribute to the measured total permeation flux. Therefore, the neutral gas pressure in direct proximity to the sample during ion beam operation has been measured using the same capillary sample as described in Sec. III. The earlier determined conductance is again used to correlate the QMS signal with neutral deuterium gas pressure in front of the capillary. Due to the 45° angle between the capillary axis and the beam axis and its sufficiently large aspect ratio, only neutral deuterium gas can directly reach the analyzing chamber through the capillary, while deuterium ions, or CXN, are implanted into the tungsten membrane or reflected. The opening area of the capillary is much smaller (a factor of 2 × 10⁻⁵) than the beam area and positioned about 4 mm below the center of the membrane, i.e., slightly outside of the beam’s footprint. The measurement was performed at room temperature, and the estimated break through time of the ion-driven permeation flux due to interstitial diffusion of implanted deuterium as determined using transport properties from Ref. 16 is several days.

In Fig. 8, the neutral deuterium gas pressure at two different locations in the loading chamber is shown. The red curve shows the pressure in front of the sample as determined by the capillary flux into the QMS. The black curve shows the pressure from the re-calibrated full-range gauge (see Sec. IV) in the recessed volume. Two implantation scenarios are compared. The machine parameters for the 170 eV/D D₃⁺-ion beam (“Ions”) are given in the supplementary material. In the other case (“CXN”), the same machine parameters are used, but the sample holder is biased above the extraction potential of the ion source such that the ions are repelled and do not reach the target.

At both measuring locations, the neutral deuterium gas pressure differs only marginally between the different implantation scenarios. This indicates that the pumping efficiency is not sensitive to the exact beam configuration. The pressure in the recessed volume (black curve) and the pressure directly in front of the sample (red curve) are very similar in both cases. This indicates an almost constant pressure profile in the loading chamber during beam operation and that the open design, in fact, allows for sufficiently fast pumping of the recycled deuterium. The re-calibrated full-range gauge in the recessed volume can, therefore, be used to reliably estimate the neutral gas pressure in direct proximity of the sample, even during beam operation. The beam causes the neutral gas pressure in the loading chamber to increase by about a factor of 7. This indicates that most background gas in the loading chamber during beam operation is deuterium, which is transported into the loading chamber by the ion beam. Considering the measured operating pressure of about 2 × 10⁻⁶ mbar, the mean free path of gas and beam particles in the loading chamber during beam operation, as determined using kinetic gas theory, is about 100 m. This is much longer than the distance between the mass-filtering dipole magnet and the sample (≈1 m) or the length of the deceleration system (≈0.1 m). It can, therefore, be assumed that most ions impinge on the sample surface without previous collisions with neutral background gas. In particular, no charge exchange processes are to be expected within the deceleration system since there is no pressure build-up in that region.

For a 50 μm thick tungsten foil, assuming diffusion-limited boundary conditions and transport parameters from Ref. 16 as well as hydrogen solubility from Ref. 17, the expected gas-driven permeation from an upstream neutral background pressure of 2 × 10⁻⁶ mbar at a temperature of 900 K via the entire sealed membrane area is 9.4 × 10⁸ D/s. As will be discussed in detail in Sec. VI, the steady-state ion-driven permeation flux from the D₃⁺-ion beam (170 eV/D) described above is 3.5 × 10¹¹ D/s at an ion current of 160 μA. The gas-driven permeation flux, therefore, accounts for only 0.3% of the ion-driven permeation flux. However, for membranes with high deuterium solubility, particularly at high temperatures, the measured permeation flux must be seen as a superposition of ion- and gas-driven permeation. In the case of much more highly permeable “Eurofer 97” martensitic steel,¹⁸ the gas-driven permeation flux through the entire sealed area of a 50 μm membrane as a consequence of 2 × 10⁻⁶ mbar upstream neutral deuterium pressure at 900 K would be 4.7 × 10¹² D/s, i.e., exceed the ion-driven permeation flux by a factor of more than 10.

Depending on boundary conditions such as temperature, implantation flux, ion energy, surface contamination (e.g., oxides), and of course membrane material, various diffusion regimes for ion-driven permeation experiments are discussed in the literature (see, e.g., Refs. 19 and 4).

If surface contaminations such as oxides reduce the surface desorption rate to an extent where the solute deuterium accumulates beneath this surface (i.e., subsurface solute concentration ≠ zero), the probability for an implanted particle to desorb from either side of the membrane will additionally depend on the thermally activated recombination process on the upstream and/or downstream surfaces.²¹ In such a regime, the measured steady-state ion-driven permeation flux will, unlike Eq. (1), show a temperature dependence. In the extreme case of a strongly reduced recombination rate on both surfaces, implanted particles have a high probability of not desorbing when reaching a surface and diffusively migrate much further than a single membrane thickness before desorbing either upstream or downstream. In this case, the probability of desorption in steady-state is 50% of the implanted particle flux for both surfaces and does not depend on the sample thickness, the implantation depth, or the temperature.

The energy of a particle impinging on the membrane is reduced due to electronic and nuclear stopping until the particle is thermalized. The mean depth at which this implantation happens can be simulated with Monte-Carlo-based binary collision approximation codes, such as SDTrimSP 7.00,¹² and depends mainly on the initial energy and the species of projectile and target. However, the orientation of the target’s crystal structure with respect to the projectile’s angle of incidence can influence or even dominate the mean implantation depth due to channeling.^22,23 Along channeling directions, the mean implantation depth can be much larger than for random directions. Since the channeling particles penetrate deeper into the crystal before significant nuclear interaction takes place, the reflection probability is also reduced.

The reflection probability and the mean implantation depth, as well as the resulting permeation probability [Eqs. (1) and (2)], were calculated for deuterium ions (170 eV/D) and CXN (1733 eV/D), and the results are shown in Table I.

To validate the permeation setup, a series of ion-driven permeation experiments were carried out on a 50 μm tungsten membrane recrystallized at 2000 K for 30 min prior to the permeation experiments. The sample temperature was kept constant at 650 K during the whole series of experiments. The additional sample heating from ion implantation as observed with the pyrometer was typically about 4 K.

The ion-driven permeation flux normalized for the respective impinging ion current for three permeation experiments on the same sample for different sample conditions is shown in Fig. 9. The two black dashed lines mark the expected steady-state ion-driven permeation flux for diffusion-limited permeation regime [Eq. (1)] using the mean implantation depth and reflection probability of D₃⁺ ions as calculated for amorphous and crystalline tungsten in Table I. The confidence bands reflect the uncertainty of the ion current calibration due to uncertainties in the sputter yield of 170 eV deuterium on copper as shown in Fig. 7.

When the sample is exposed to deuterium for the first time (blue curve in Fig. 9), a non-monotonic permeation flux is observed, where the initial permeation flux quickly reaches a maximum and then slightly decays during 5 h. Based on the nominal mean ion flux density of 0.9 × 10²⁰ D/m²s (see Sec. V C), this corresponds to a fluence of about 1.7 × 10²⁴ D/m². Subsequently, a slow increase to the final, saturated permeation flux is completed after about 1.3 × 10²⁵ D/m² (40 h). Similar initial maxima in permeation flux have been reported in the literature^19,24 and are attributed to a reduced upstream re-emission rate due to surface contamination. We assume that a natural tungsten oxide layer of a few nm thickness on the upstream side is eroded chemically due to physical sputtering by the beam ions during the first 5 h. Such behavior of the permeation flux has not been observed for experiments after the first one, indicating that the re-oxidation at the given upstream base pressure of 5 × 10⁻⁸ mbar at 650 K is slow and that the upstream surface can be considered sufficiently clean in subsequent runs, respectively, that the surface is cleaned by the ion beam much faster than the rise time of the permeation signal. The slow increase after 5 h can be explained by assuming an increasing downstream desorption rate, e.g., due to the reduction of the downstream natural oxide by chemical erosion by the permeating deuterium particles, similar to what has been observed in Ref. 25 for outgassing of deuterium-loaded tungsten through surface oxides. In parallel to the stabilization of the mass four permeation signal, a decay of the QMS signals for mass 17 and 18 (heavy water) has been observed, which supports this possible explanation. After the initial conditioning of the sample’s surfaces is finished, the permeation flux stabilizes and remains constant for more than a day. The effective permeation probability for ions following Eqs. (1) and (2) using the total measured implantation and permeation fluxes was calculated to 9.5 × 10⁻⁵. This permeation probability lies in between the expectations for amorphous and crystalline targets with diffusion-limited boundary conditions (see Table I).

An example of a permeation curve with well-conditioned sample surfaces (red curve in Fig. 9) is also shown. A more detailed representation of that permeation curve, including time traces of the ion current and sample temperature, is provided in the supplementary material. The steady-state permeation flux through the well-conditioned sample is reached after about 0.5 h and shows very good agreement with the flux observed after the initial conditioning. After initial sample conditioning, for multiple experiments with identical implantation conditions, we observed highly reproducible permeation curves with respect to both transients and steady-state fluxes. The reproducibility of the experiments is given even after the implantation had been stopped for several hours, indicating the preservation of the well-conditioned state during typical decay time scales at 650 K.

Aging of the sample, however, has been observed after the ion implantation had been stopped for 32 h. For about 11 h during this period, only CXN had been implanted (see Sec. VI C). As shown in Fig. 9, the rising transient for the aged sample saturates much slower than in the well-conditioned case, and the steady-state is reached only after about 10 h. We assume that this is due to re-oxidation of the downstream surface from ambient residual oxygen in the absence of significant deuterium permeation flux. It appears, however, that such newly formed oxide layers can again be chemically eroded similarly to the native oxide from the initial condition. The steady-state flux recovers almost completely to 96% as compared with the well-conditioned case. A possible explanation for this slight reduction of the steady-state permeation flux is a reduced permeation probability due to a reduced crystallinity on the implantation side as a consequence of bulk displacements of target atoms from the 11 h (1 × 10²² CXN/m²) of CXN implantation. The respective damage dose within the ion implantation range, as estimated using SDTrimSP 7.00,¹² is about 2 dpa.

The steady-state ion-driven permeation fluxes have also been measured at various temperatures up to 830 K in randomized order using nominally identical samples. The permeation fluxes scattered by ±15% but were on average identical to the well-conditioned case shown here, and no systematic temperature trend was observed. In addition, permeation has been measured using a sample membrane of only half the thickness (25 μm). The surface conditions and implantation conditions were closely similar to the 50 μm sample. The measured steady-state permeation flux through the 25 μm membrane was in good approximation two times the flux of the 50 μm samples.

Given the temperature independence of the permeation flux, the behavior for thickness variation, the theoretical permeation probability, and the observations shown in Fig. 9, we are confident that, after sufficient sample conditioning, diffusion-limited boundary conditions can be reached in TAPAS for ion-driven permeation through tungsten. It appears that, despite a finite oxygen partial pressure at the downstream surface of the sample (<1 × 10⁻⁹ mbar), the steady-state permeation flux density of 1 × 10¹⁶ D/m²s keeps the surface conditions sufficiently clean for diffusion limitation.

After passing the dipole magnet, collisions of the D₃⁺ beam ions with neutral background gas particles or surfaces of the ion optical system lead to the creation of CXN with energies up to 1733 eV/D. Since they are uncharged, CXN will not be decelerated by the deceleration system and, therefore, lead to much deeper implantation and lower reflection probability (i.e., stronger contribution to permeation per particle than ions), as well as displacement damage in the sample and physical sputtering. The fraction of CXN in the 170 eV/D D₃⁺-ion beam was estimated by comparing the steady-state permeation signal with the case where the sample holder is biased above the beam potential, such that ion implantation is suppressed and only CXN can contribute to permeation. The sample was again a 50 μm tungsten membrane recrystallized for 30 min at 2000 K. The temperature was 650 K (Fig. 10).

As discussed in the previous Sec. VI B, for ion-driven permeation, the upstream and downstream surfaces of the tungsten sample are apparently cleaned by the impinging and the permeating flux of deuterium, respectively. However, if the beam ions are filtered out and only CXN are implanted, the deuterium fluxes to both surfaces are much lower. Accordingly, the equilibrium surface conditions may differ and cause slow deuterium desorption, i.e., surface limitation. To address this concern for the quantification of CXN by measured permeation fluxes, the permeation of CXN was measured in two different ways.

In the first case (red curve in Fig. 10), the permeation was started with the full 170 eV/D D₃⁺-ion beam, and the steady-state permeation flux was identical to that of the well-conditioned sample case in Fig. 9. While the beam was still running, the sample holder was then biased above the beam potential to repel the ions and implant only the CXN (at t = 1.4 h). With this procedure, no time is given for re-oxidation of the sample surfaces, which supposedly were clean during the preceding period of full ion beam implantation. The permeation flux, now driven only by CXN, drops by about a factor of 20. When the beam stopper was inserted (at t = 2.5 h), the implantation of CXN was stopped as well, and the typical diffusive decay of the permeation signal toward the background level was observed.

In the second case (black curve in Fig. 10), the sample was initially well-conditioned (not shown here), but was then exposed to the background chamber pressure at 650 K for about 6 h after the ion implantation for surface conditioning had been stopped. Permeation was then started directly with the over-biased sample holder, i.e., by implantation of only CXN (t = 0 h). The steady-state permeation is reached after about 1 h. This was expected because the fluence at which residual material defects are decorated such that mobile deuterium could reach the downstream surface was reached later for lower deuterium implantation fluxes. The permeation driven by the upstream background deuterium gas pressure during the experiment (<2 × 10⁻⁶ mbar) was estimated using transport data from Ref. 16 and is negligible (<3 × 10⁻³ cps).

The steady-state permeation flux from CXN is identical in the first and second cases, i.e., regardless of whether there was a break between sample conditioning and CXN permeation measurement or not. The permeation flux remains constant for several hours, indicating stable surface conditions. If surface contamination due to chemical reactions with the background oxygen would have significantly influenced the permeation flux of CXN, a difference between the two cases would be expected. Since no difference was observed, we believe that the permeation regime for CXN implantation is diffusion-limited as well.

The effect of the target’s microstructure on the permeation probability of impinging ions is shown in Fig. 9. In order to quantify the fraction of CXN in the ion beam from the measured steady-state permeation fluxes of ions and CXN (Fig. 10), the effective microstructure determining the permeation probability of CXN is assumed to be identical to the ion case. Accordingly, the same weighting of the amorphous and crystalline cases as for the ions was chosen, resulting in an effective permeation probability of 4.3 × 10⁻⁴ for the CXN (see Table I). Thus, the probability of permeating is about 4.3 times higher for CXN compared with ions. Considering the permeation probabilities of ions and CXN, the measured steady state permeation fluxes in Fig. 10 yield a fraction of CXN in the beam of 1.0%.

In this article, we described the newly constructed ion-driven permeation setup TAPAS. We discussed the optimization of machine parameters as well as the characterization of various experimental boundary conditions for deuterium permeation through membranes. The ion source, mass filter, ion focusing, and ion deceleration systems were optimized for a stable and reproducible D₃⁺ ion beam with an ion energy of 170 eV/D, which is below the sputtering threshold of tungsten.⁹ At the same time, we achieved a high ion flux density of 0.9 × 10²⁰ D/m²s averaged over the beam footprint containing 95% of all impinging particles, with a central ion flux density of 2.9 × 10²⁰ D/m²s (containing 10% of all impinging particles). This is comparable to the conditions at the first wall of a fusion reactor.² A detailed description of the calibration procedure for the ion current measurement and the mapping of the ion flux density distribution based on the erosion of copper by physical sputtering is given.

By maintaining a low pressure of neutral D₂ gas in the loading chamber of about 2 × 10⁻⁶ mbar during operation, we minimized the scattering of beam particles with background D₂ molecules and kept the generation of fast charge-exchange neutrals to about 1% of the total ion flux, as measured via the comparison of permeation signals for the 170 eV/D ion beam and for fully repelling the ions by positively biasing the sample. A method based on the molecular flow of neutral deuterium gas through a small capillary was used to show the linearity of the quadrupole mass spectrometer in the high signal regime and to calibrate the measurement of low upstream deuterium pressures in direct proximity to the sample. The neutral gas background in the loading chamber is evidently dominated by recycled beam particles during operation but did not show any significant pressure gradients.

We note that the measured operating background D₂ pressure does not lead to significant gas-driven permeation during ion-driven permeation experiments of low-HI-permeability materials such as tungsten. However, for materials with a high HI permeability, the deuterium uptake can be dominated by gas loading. We calculated that in the case of Eurofer 97¹⁸ at 900 K, the permeation flux that is driven by background D₂ gas (2 × 10⁻⁶ mbar) exceeds the permeation flux driven by 0.9 × 10²⁰ D/m²s ion-implantation by a factor of more than 10.

We found that in the present setup the lower detection limit of the quadrupole mass spectrometer corresponds to a total permeation flux of 6.4 × 10⁷ D₂/s. This is about a factor of 2700 lower than the ion-driven permeation flux through 50 μm of tungsten measured here and, thus, provides the possibility of measuring also strong permeation barriers.

In a benchmark experiment, it was demonstrated by performing temperature and thickness variations as well as model calculations that diffusion-limited boundary conditions are adequate for describing ion-driven permeation of D through recrystallized tungsten foils for temperatures between 650 and 830 K. Based on calculations for the reflection probability and mean implantation depth using SDTrimSP 7.00¹² for crystalline and amorphous targets, we show the significance of the microstructure and crystallinity of the tungsten foil on the permeation probability.

The supplementary material is made available online by the publisher and contains detailed information about the routine for optimizing the machine parameters of the ion source and the ion optical system, the machine parameters for the D3+ ion beam used in this contribution, time traces of ion current, QMS signal, and sample temperature from the shown benchmark ion-driven permeation experiment (well-conditioned sample), as well as mass spectra of the un-filtered ion beam for different working gases.

This work has been carried out within the framework of the EUROfusion Consortium, funded by the European Union via the Euratom Research and Training Program (Grant Agreement No. 101052200—EUROfusion). Views and opinions expressed are, however, those of the author(s) only and do not necessarily reflect those of the European Union or the European Commission. Neither the European Union nor the European Commission can be held responsible for them.

The authors have no conflicts to disclose.

P. Sand: Formal analysis (lead); Visualization (lead); Writing – original draft (lead); Writing – review & editing (equal). A. Manhard: Conceptualization (lead); Formal analysis (supporting); Supervision (lead); Writing – review & editing (equal). U. von Toussaint: Formal analysis (supporting); Investigation (supporting); Methodology (supporting); Software (equal); Supervision (supporting); Writing – review & editing (supporting).

The data that support the findings of this study are available from the corresponding author upon reasonable request.