Experimental evidence on photo-assisted O − ion production from Al 2 O 3 cathode in cesium sputter negative ion source
O. Tarvainen, R. Kronholm, M. Laitinen, M. Reponen, J. Julin, V. Toivanen, M. Napari, M. Marttinen, D. Faircloth, H. Koivisto, T. Sajavaara
PPhoto-assisted negative ion production
Experimental evidence on photo-assisted O − ion production from Al O cathode in cesium sputter negative ion source O. Tarvainen, a) R. Kronholm, M. Laitinen, M. Reponen, J. Julin, V. Toivanen, M. Napari, M. Marttinen, D. Faircloth, H. Koivisto, and T. Sajavaara STFC ISIS Pulsed Spallation Neutron and Muon Facility, Rutherford Appleton Laboratory, Harwell, OX11 0QX, UK University of Jyväskylä, 40500 Jyväskylä, Finland University of Southampton, Southampton SO17 1BJ, UK (Dated: 2 July 2020)
The production of negative ions in cesium sputter ion sources is generally considered to be a pure surface process. It hasbeen recently proposed that ion pair production could explain the higher-than-expected beam currents extracted fromthese ion sources, therefore opening the door for laser-assisted enhancement of the negative ion yield. We have testedthis hypothesis by measuring the effect of various pulsed diode lasers on the O − beam current produced from Al O cathode of a cesium sputter ion source. It is expected that the ion pair production of O − requires populating the 5delectronic states of neutral cesium, thus implying that the process should be provoked only with specific wavelengths.Our experimental results provide evidence for the existence of a wavelength-dependent photo-assisted effect but castdoubt on its alleged resonant nature as the prompt enhancement of beam current can be observed with laser wavelengthsexceeding a threshold photon energy. The beam current transients observed during the laser pulses suggest that themagnitude and longevity of the beam current enhancement depends on the cesium balance on the cathode surface. Weconclude that the photo-assisted negative ion production could be of practical importance as it can more than doublethe extracted beam current under certain operational settings of the ion source. I. INTRODUCTION
The negative ion production in cesium sputter ion sources istraditionally attributed to surface ionization by resonant tun-neling of an electron from the conduction band or interstitialsites of a metal or compound material covered with a layer ofcesium to the affinity level of a neutral atom and subsequentejection of the negative ion from the surface. A general reviewsummarizing the relevant steps of the process can be found inthe literature (and original references therein). The ioniza-tion efficiency depends strongly on the work function of thesurface as first noted by Yu , the sputtering yield and the es-cape velocity of the negative ions . Thus, cesium atoms andions play a dual role in the process; the alkali metal coveragereduces the surface work function while the heavy ion bom-bardment releases negative ions from the surface at sufficientenergy to overcome the force exerted by the induced imagecharge within the material causing detachment of the anions .The surface production of negative ions by resonant tun-neling is unquestionable as witnessed also in plasma ionsources serving as H − /D − injectors for large-scale acceler-ator facilities and neutral beam injectors for thermonuclearfusion . However, it is plausible that there are other mech-anisms contributing to the negative ion yield. This claim issupported by the discrepancy between the yields of negativeions deduced with reasonable estimates of the electron affini-ties and work functions involved, and the measured negativeion currents from cesium sputter ion sources, first acknowl-edged by Krohn . Despite of the obvious disagreement be-tween the theory and observations, there are no systematicexperiments attempting to solve the mystery, which is proba- a) Electronic mail: [email protected] bly due to the lack of a testable hypothesis on the mechanismgoverning the negative ion production through an alternativepath. Thus, the majority of the notions in the literature areanecdotal, a famous example being the assertion of the strongbelief of Middleton that "the ionization occurs primarily in theblue-glowing plasma of Cs created in sputter-induced pits orin purposefully recessed samples" . The glow is presumablysustained by electron impact excitation of neutral Cs to the 7pstates by secondary electrons emitted from the cathode and thesubsequent de-excitation back to the 6s ground state emittingblue light.This work has been inspired by a recent publication byVogel , which introduces a physical mechanism that wouldnot only explain the enhanced yield of negative ions but canalso be probed in a controllable experiment. This hypothesisis based on resonant ion-pair production first noted by Leeet al. in thermal alkali vapors . The ion pair-productionis described by the (chemical) reaction equation A ∗ + B → A + + B − , which depicts the interaction between a neutral atomon an excited state (A ∗ ) and a ground state neutral atom (B)with positive electron affinity resulting to the formation ofpositive (A + ) and negative (B − ) ion pair. According to so-called Landau-Zener-Stückelberg (LZS) formalism the prob-ability, i.e. cross section, of the ion-pair production dependsstrongly on the energy difference of the electron donor ioniza-tion potential from the excited state and the electron affinityof the electron acceptor . In practice this means that exci-tations to specific electronic states of the donor atoms can en-hance the negative ion production of those anions with match-ing electron affinities (see Section II B for an example illustra-tion). These excitations can occur as a result of inelastic col-lisions between neutral Cs atoms and electrons emitted fromthe cathode or they can be facilitated externally by photon ab-sorption.The above-said paper describes a proof-of-concept exper- a r X i v : . [ phy s i c s . i n s - d e t ] J u l hoto-assisted negative ion production 2 E vac EACBE g VB Affinitylevel e - e - Distance atom/ion - surface E vac
EACBE g VB Distance atom/ion - surface a) b)
Cse - Δ Se - e - En e r g y En e r g y Δ S FIG. 1. A sketch of the electron capture process of atoms released by heavy ion bombardment from (a) insulator and (b) cesium coveredinsulator. The numbers indicate selected positions of the outbound atom/ion relative to the material surface depicted by the quantum mechanicalpotential well and explained in the bulk text. iment where the C − beam current extracted from a cesiumsputter ion source was enhanced by approximately 10% byshining a 450 nm / 5 W laser beam radially into the recessedcathode . It is argued that approximately 20 µW of the powerwas absorbed by the excitation of the Cs(7p)-states, which aresaid to be in resonance with the effective ionization potential I p , eff = E A + ∆ E in the reaction Cs ( p )+ C → Cs + + C − . Here E A is the electron affinity of the anion and ∆ E the endother-micity of the reaction. A direct quote from the paper reads:"The quiescent 40 µA current of C − immediately jumped to45 µA when the laser passed across the front of the samplewithin the 3 sec time resolution of data collection. The sourceheld the higher current as long as the laser passed behind theimmersion lens, but no long term data was taken."Given the temporal resolution of the experiment, lack ofactual data and staggering nearly 70% efficiency of the pair-production (see section V for further discussion) reported inthe original publication , we decided to subject the allegedmechanism to further scrutiny attempting to compare it to al-ternative explanations of enchanced negative ion production.All experimental data described hereafter was obtained withO − ion beams extracted from an Al O cathode of a cesiumsputter ion source. The choice of O − and its pair produc-tion mechanism in interaction with electronically excited Csatoms are explained in Section II B. The experimental setupincluding the SNICS ion source, the adjacent beamline andthe diode lasers employed for the experiment are described inSection III. Finally, the experimental results and conclusionsare presented in Sections IV and V. II. NEGATIVE ION FORMATION
Interpretation of the experimental data requires un-derstanding the principles of negative ion surface pro-duction via cesium sputtering as well as the possibleCs(5d) + O → Cs + + O − ion pair production mechanism.A qualitative description of each relevant process is therefore given below. These include electron transfer to the electronaffinity level of the anion either directly from the material orvia photoelectron emission, bond-breaking of chemical com-pounds and the possible ion pair production mechanism. A. Surface production of negative ions by cesium sputtering
The negative ion production by cesium sputtering is de-scribed schematically in Fig. 1 for (a) bare insulating material(e.g. Al O ) and (b) Cs covered insulating material. The ma-terial surface is exposed to bombardment by Cs + ions eject-ing atoms from the bulk material. The electron affinity levelof these atoms depends on the distance from the material sur-face and the material properties affecting the Debye screen-ing radius . At close proximity of the surface, denoted with(1) in Fig. 1, the affinity level of the outbound atom overlaps(in energy) with the valence band (VB) of the insulating ma-terial allowing electron tunneling through the potential bar-rier and formation of the anion. As the anion moves awayfrom the surface (2), the reverse process i.e. electron tunnel-ing back into the material is prohibited as the affinity levelnow overlaps with the band gap ( E g ) between the valence andconduction (CB) bands. Far from the surface (3) the elec-tron affinity level reaches the value of a free atom (ion), e.g.1 .
46 eV below the vacuum level for oxygen. The negative ionyield of the process can be greatly enhanced by deposition ofCs atoms onto the material surface as depicted in Fig. 1(b).This is because Cs modifies the band structure lowering theeffective work function of the surface. In this case (4) thequantum-mechanically allowed interaction distance of elec-tron tunneling to the atoms sputtered from the bulk materialis much longer. That is because the required downshift of theelectron affinity ∆ S resulting in overlap with electronic statesof the surface layer is reduced in comparison to bare insula-tor. For both cases shown in the figure ∆ S is referenced to thevalue of the electron affinity far from the surface (5). Detailedaccounts and surface material / ion-specific variations of thehoto-assisted negative ion production 3 D D S P ○ P ○ F ○ S Ionization (a) P ○ P ○ F ○ D D (0.909)(1.282) % . % . % . % . % . % . % . % % . % . % % . % . % (0.158)(0.133)(0.048) (0.034)(0.029) . % . % C s e xc i t a t i on ene r g y ( e V ) σ (arb.) 0.00.51.01.52.02.53.03.54.0 (b) S D D P ○ P ○ FIG. 2. A partial Grotrian diagram of neutral cesium. The diagram (a) shows the putative excitations corresponding to the neutral cesiumelectronic energy levels and the wavelengths of the diode lasers used in our experiment. The red arrows indicate the optically allowedtransitions to 7p states and the orange arrows the energetically possible but very unlikely transitions to 5d and 4f. The photon energies of eachlaser are marked with horizontal lines with the line colors matching with the corresponding wavelenghts and the line widths to the FWHMs.The radiative decays populating the 7s, 5d and 6p -states that overlap in energy with the normalized O − ion pair production cross section σ (b) are indicated by black arrows. The most likely excited state contributing to the O − pair production is the metastable 5d (populated bythe transitions marked with solid downward arrows) almost matching the peak of the cross section as indicated by the projected black barscorresponding to the fine structure of the electronic state(s). The spontaneous lifetime of each excited state in µs is shown in parentheses. Thebranching ratios of the de-excitations calculated from the reported oscillator strengths are displayed next to each downward transitions. Twobranching ratios associated with a single arrow correspond to the transition to/from different fine structure levels of the excited states. described process for metals and insulators can be found fromthe literature .For the purpose of this paper it is important to acknowl-edge that negative ions can be produced directly from com-pound materials, e.g. Al O , by heavy ion bombardment in-duced bond-breaking of the molecular solids into cation-anionpairs . The relative importance of such direct negative ionproduction and the electron capture by neutral sputtered parti-cles as explained above is unknown as the probability of eachprocess depends on the material, Cs coverage and cathodebias. Finally, it could be argued that photon absorption mightincrease the negative ion yield by allowing photoelectrons toovercome the surface potential barrier and become bound tothe affinity level of the ejected atom, thus supplementing the described electron capture by tunneling. To our knowledgethe possible role of photon absorption in the electron captureprocess has not hitherto been studied whereas Blahins et al.have recently used a laser to study photodetachment with acesium sputter source . B. Resonant ion pair production
The ion pair production cross section depends on the energylevels of the electron donor and the electron affinity of the an-ion as summarized by Vogel . The cross section peaks whenthe afore-mentioned resonance condition I p , eff = E A + ∆ E issatisfied. In our experiments we focused on O − for whichhoto-assisted negative ion production 4 E A = .
46 eV and ∆ E = .
55 eV. Thus, the pair produc-tion of O − ions is believed to be in resonance with the 5dstates of neutral Cs with I p , eff of 2 .
08 eV (5d D / ) and2 .
10 eV (5d D / ) as shown in Fig. 2. The correspond-ing normalized cross section as a function of energy differ-ence between the donor and acceptor states for the reactionCs ∗ + O → Cs + + O − was taken from the literature and isprojected onto the vertical scale in Fig. 2. The ion pair pro-duction reaction between Cs and O has been studied earlierby Vora et al. reporting that Cs(6p) increases O − produc-tion over ground state Cs at lower collision energies, which isencouraging as the ion pair production cross section from 6pstates of Cs is smaller than the expected cross section from the5d states (see Fig. 2).Oxygen (O − ) was chosen for our experiments for a num-ber of practical reasons despite of the large endothermicity ofthe pair production reaction corresponding to relatively highcollision energies with an ideal donor compared to e.g. C − production . Firstly, oxygen has a high electron affinity and,therefore, O − beams are relatively easy to produce, which al-leviates the experimental effort. Secondly, the 5d donor statesof neutral Cs relevant for O − pair production have longer life-times than other excited states of Cs, which presumably max-imizes the interaction probability. Finally, the 5d states areaccessible from a number of upper electronic states, which al-lows probing the ion pair production hypothesis with multiplelaser wavelengths and population pathways. The energies cor-responding to each laser used in the experiments are marked inFig. 2 with horizontal lines (the line colors correspond to thewavelenghts of each laser and the line widths to the FWHM ofeach diode laser). Furthermore, the figure displays the closestexcited states of Cs accessible with each laser and the branch-ing ratios of the electronic de-excitations leading to 5d statesfrom these upper states. The relevant transitions were identi-fied using the National Institute of Standards and Technology(NIST) atomic spectra database . The relative probabilitiesof each energetically possible transition from the ground stateare discussed in Section IV where they are used for the inter-pretation of the experimental data. The branching ratios listedin Fig.2 were calculated from the oscillator strengths reportedin the literature taking into account the known degenera-cies of the excited states, and finally normalizing the sum ofthe resulting Einstein coefficients to unity. It is worth notingthat the resulting branching ratios differ from those used byVogel , which is probably due to discrepancies in reported os-cillator strengths of the relevant transitions as outlined in theliterature . Also, the spontaneous lifetimes of each state inµs units are marked in the figure. From the experimental view-point these are more important than the branching ratios indetermining the expected population densities of the excitedstates and, therefore, the interaction probabilities of donor-acceptor pairs.The conditions for efficient ion pair production are suffi-cient flux of photons causing excitation of Cs(5d) and ade-quate density of both Cs(5d) as well as oxygen atoms interact-ing with each other. If these conditions are met, the ion pairproduction can occur anywhere along the path of the oxygenatoms sputtered from the ion source cathode (see Section III) and interacting with the surface layer of Cs atoms and Cs va-por in the proximity of the cathode surface. It is underlinedthat the ion pair production is not strictly a surface processper se but involves two unbound atoms. III. EXPERIMENTAL SETUPA. SNICS ion source
The experimental data were taken on a Multi-CathodeSource of Negative Ions by Cesium Sputtering (MC-SNICS) by National Electrostatics Corporation (NEC). Fig-ure 3 shows a schematic drawing of the ion source with mostof the detail omitted for simplicity. FIG. 3. Schematic drawing of the SNICS ion source. (1) Cesiumoven and transfer line, (2) ionization chamber, (3) ionizer, (4) cath-ode with Al O powder, (5) focusing electrode (immersion lens) and(6) extraction channel and electrodes. Cesium (blue in Fig. 3) is evaporated from the oven into theionization chamber where some of it condenses on the surfaceof the cathode creating a thin, ideally sub-monolayer cover-age (greatly exaggerated in the figure) and, thus, lowers theeffective work function of the cathode material. Some of thecesium vapor is surface ionized on the hot surface of the ion-izer. The Cs + ions (orange) are accelerated toward the cath-ode by applying a kV order of magnitude negative potential,typically 4 kV with the specific source used in this work, andare focused on the front face of the cathode by the "cesiumfocus lens" or "immersion lens". Increasing the cathode biasenhances the negative ion current through the sputtering yieldhoto-assisted negative ion production 5 SNICS ion source 30 deg. bending magnet Faraday cupLaser viewport 1446 mm from the SNICS cathode
FIG. 4. The layout of the MC-SNICS beamline at JYFL Pelletron facility. The multicathode SNICS ion source is connected to the upperbranch of the beamline. Other ion sources (to the left) and the Pelletron accelerator (to the right) are not shown. at the expense of reduced cathode lifetime and beam currenttemporal stability. In our experiment the cathode was pre-pared by pressing Al O powder into a cylindrical 1 mm (indiameter) notch on its surface. The O − ions (red) liberatedfrom the cathode by cesium sputtering are self-extracted fromthe ion source through the ∼
10 mm (in diameter) extractionchannel by the cathode potential and finally accelerated to thedesired energy further downstream. The ion beam current isadjusted by the user through cesium oven temperature, ion-izer temperature, cathode potential and focusing lens voltage.These affect the neutral cesium flux into the ionization cham-ber and cathode surface, Cs + ionization probability and fluxonto the cathode, sputtering yield and negative ion escape ve-locity, and beam optics, respectively.The extracted negative ion beam was then focused with anEinzel lens into a 30 degree dipole magnet. The transported,mass-analysed, O − beam current was then measured with aFaraday cup located downstream from the magnet. The lowenergy MC-SNICS beamline of the JYFL Pelletron facility isshown in Fig. 4. The distance between the laser viewport andthe SNICS cathode is indicated in the figure. Typical beamcurrents detected from the Faraday cup range from a few nA toseveral µA. Hence, the current was measured with a low noisecurrent amplifier (Stanford Research SR570), which affectsthe temporal resolution of the experiment. The implicationsof the inevitable RC -constant are discussed in Section IV. B. Setup for photo-assisted negative ion production
The photo-assisted production of O − was probed with sev-eral diode lasers listed in Table I with their nominal wave-lengths ranging from 405 nm to 638 nm and maximum pow-ers from 0 . TABLE I. The models of the Lasertack GmbH diode lasers, theirnominal wavelenghts and maximum powers.Diode laser Wavelength [nm] Power [W]LDM-405-1000 405 1.0LE-445-6000 445 6.0LDM-450-1600 450 1.6LDM-520-1000-A 520 1.0LDM-638-700 638 0.7
The laser beam was focused straight onto the ion sourcecathode through a viewport of the bending magnet. Thisarrangement differs from the one used in the earlierexperiments where the laser irradiated the volume adjacentto the cathode surface radially. The focal point of our lasersetup, shown in operation in Fig. 5, was first adjusted off-lineto match the distance to the SNICS cathode, then rotating thelaser optics to center the beam spot with the viewport and to il-hoto-assisted negative ion production 6luminate the cathode. The lasers were changed during the ex-periment using the viewport and two optical apertures shownin Fig. 5 as alignment fixtures. The accuracy of this procedurewas assessed by measuring the laser power at the SNICS cath-ode surface by replacing the cathode with a quartz windowand using a 1 mm diameter collimator in front of the win-dow, mimicking the cathode cross section. The output powerof the LDM-450-1600 diode and the power delivered to thecathode were then measured with a Thorlabs S415C thermalpower sensor. It was observed that less than 10% of the outputpower (i.e. 0 . .
09 W out of 1 W) reaches the power sen-sor when the above alignment procedure was applied repeat-edly. The maximum power delivered to the cathode in thisconfiguration was less than 20% (0 .
18 W out of 1 W) whenthe alignment was adjusted while observing the power read-ing "on-line". Altogether, this translates to estimated 5–20%of the laser power being delivered to the cathode. The poweris limited by the mismatch between the beam spot size andthe geometrical apertures i.e. extraction channel and samplediameter.
LaserViewport MirrorMirror
FIG. 5. The optical table and components used for focusing the laserbeam onto the ion source cathode at approximately 1 . The effect of the laser exposure on the O − beam currentwas probed by modulating the output of the lasers by puls-ing their drive current (on/off) at various frequencies as wellas adjusting the drive signal amplitude to control the laserpower. The emission spectrum of the diode lasers dependson their operating temperature and, therefore, the lasers weretemperature stabilized to 20 ◦ C by a Peltier cooling elementequipped with a thermocouple feedback control. Neverthe-less, it was observed that the emission spectrum shifts slightlywith the output power as well as in the beginning of each laserpulse, the emission intensity increasing rapidly with time by ∼
10% in the latter case. Both these effects are presumablyrelated to the exact temperature (and its transient) of the light-emitting element, and have implications on the interpretationof the data. Tables II and III serve to quantify these effectsfor the 450 nm laser (used here as a representative example) at 20 ◦ C temperature. The emission spectra of each laser weremeasured with Ocean Optics USB 2000+ survey spectrome-ter. Their full width half maximum (FWHM) was measuredto be 1 . . TABLE II. The peak wavelength of the 450 nm blue laser with dif-ferent output powers at 20 ◦ C set temperature.Output power [mW] Peak wavelength [nm]320 446.3640 447.0960 448.11280 449.21600 449.5TABLE III. The peak wavelength and normalized intensity of the450 nm blue laser with 1 . It was later confirmed that the results achieved by pulsingthe laser could be reproduced by pulsing the light signal witha mechanical chopper. Pulsing the laser was preferred as itallows studying both, prompt and long-term effects withoutthe added complication of a data acquisition system based ona lock-in-amplifier as described elsewhere . IV. EXPERIMENTAL RESULTS
Several experimental campaigns were carried out todemonstrate photo-assisted production of O − ions and to scru-tinize the ion pair production -hypothesis. The experimentalresults presented hereafter have been organized chronologi-cally in order to allow the reader to follow the reasoning be-tween each step. A. Experiments with the 450 nm laser
The experiments were started following the footsteps ofVogel i.e. with the 450 nm laser. Absorption of the 2 .
76 eVphotons presumably populates the 7p-states of neutral Cs,which then populate the long-lived 5d-states that are expectedto contribute to O − pair production (see Fig. 2). Figures 6(a)-(c) show the extracted O − beam current as a function of timewhen the Al O -cathode was exposed to long (250 s) laserpulses with different powers. Three plots are shown to accountfor the inherent temporal variation of the beam current char-acteristic to the SNICS ion source, causing the initial beamhoto-assisted negative ion production 7 O - b e a m c u rr e n t [ n A ] Time [s]320 mW640 mW960 mW1280 mW 0 50 100 150 200 250 300 350 0 50 100 150 200 250 0 10 20 30 40 50 60 70 0 50 100 150 200 250Laser ON(b) C h a n g e o f t h e O - b e a m c u rr e n t [ n A ] Time [s] 0 10 20 30 40 50 60 70 0 50 100 150 200 250 0 5 10 15 20 25 30 0 50 100 150 200 250Laser ON(c) O - b e a m c u rr e n t g a i n [ % ] Time [s] 0 5 10 15 20 25 30 0 50 100 150 200 250
FIG. 6. The effect of the 450 nm laser power on the O − beam current. The plots show (a) the absolute current, (b) the change of the current innA and (c) the achieved gain in % units. current to vary between data sets. Figure 6(a) shows the effectof the laser on the O − beam current at different power levels,Fig. 6(b) the corresponding change of the beam current in nAand Fig. 6(c) the gain achieved with the laser normalized tothe O − beam current just before the leading edge of the laserpulse. The contribution of the laser on the O − beam currentis evident with the magnitude of the effect increasing with thelaser power (photon flux) despite of the several nm shift ofthe emission spectrum with power (see Table II). The beamcurrent, which was constant before the laser was switched on,increases for the whole duration of the 250 s laser pulse im-plying that the expected prompt effect, namely populating therelevant excited states of Cs followed by ion pair production,cannot alone explain the observed gain in O − yield.The data recorded at varying laser pulse lengths and shownin Figs. 7(a)-(c) reveal three different time scales in the O − beam current response to the 450 nm / 1 W laser exposure.Three different time scales can be clearly distinguished: (a)long-term linear increase lasting for several minutes until theend of the long laser pulses, (b) a logarithmic rise in 3–5 sand (c) a prompt effect when the laser pulse is applied. Thefirst two trends are mirrored (qualitatively) between the laserpulses and the last one at the trailing edge of the pulse.It was confirmed that the observed prompt effect is indeedinstantaneous by measuring the signal rise times at differenttransimpedance amplifier gain settings and comparing them tothe theoretical rise times of a forced step change in beam cur-rent. The rise time is determined by the amplifier impedancetogether with the Faraday cup (approx. 70 pF) and cable (ap-prox. 170 pF at high frequency) capacitances. Figure 8 showsan example of the O − beam current response at the leadingedge of the laser pulse. The expected signal rise time of 60 µscorresponding to the calculated time constant of the measure-ment setup is marked in the figure. It matches with the ob-served rise time implying that the laser-induced contributionof O − yield is truly a prompt one. It is important to note thatwe are not claiming this to be evidence for the ion pair pro-duction mechanism but instead argue that the observation con-firms the existence of a photo-assisted negative ion production channel (of yet unknown origin).The other two time scales cannot be explained easily. Theshape of the 3–5 s pulse response is typical to thermal tran-sients that could affect the Cs coverage of the cathode surfaceby changing the equilibrium density of Cs atoms determinedby deposition, evaporation and sputtering. The long-term in-crease of the O − beam current could be related to the photonabsorption affecting the surface properties, most importantlyits work function via a gradual change of the Cs density on thecathode surface. It is worth noting that the ionizer temperature(power) was kept lower than its nominal operational value,which results in modest extracted current. This was done inorder to observe the photo-assisted effect superimposed on thecontinuous beam current signal. The ionizer temperature af-fects the Cs coverage of the cathode surface by limiting theincident Cs + flux. Altogether this implies that at constantCs oven temperature the duration of the long-term transient issensitive to the ionizer temperature and it varies with the ex-tracted beam current as demonstrated in subsequent sections. B. The effect of laser wavelength - experiments with 450,520 and 638 nm lasers
In order to establish whether the laser-induced increase ofthe extracted O − current is wavelength specific, we first irradi-ated the Al O cathode with 450, 520 and 638 nm (blue, greenand red) lasers. Figures 9(a)-(b) show the response of the O − beam current to the above said blue, green and red laser expo-sures using the maximum power of each diode (see Table I).The long-term effect is observed with all of the three lasersirrespective of their wavelength whereas the prompt effect isinduced only by the 450 nm laser, not with the 520 and 638 nmones. The relative magnitude of the long-term increase of theO − beam current matches the difference in total powers be-tween the above said laser diodes, which indicates a thermalorigin although the exact mechanism acting on the cathodesurface and affecting the beam current remains elusive.The excited states of Cs that are accessible and most likelyhoto-assisted negative ion production 8
85 90 95 100 105 110 115 120 -400 -200 0 200 400Laser ON Laser ON(a) O - b e a m c u rr e n t [ n A ] Time [s] 85 90 95 100 105 110 115 120 -400 -200 0 200 400 90 95 100 105 110 -10 -5 0 5 10Laser ON Laser ON(b) O - b e a m c u rr e n t [ n A ] Time [s] 90 95 100 105 110 -10 -5 0 5 10 90 95 100 105 110-0.6 -0.4 -0.2 0 0.2 0.4 0.6ON ON ON ON ON(c) O - b e a m c u rr e n t [ n A ] Time [s] 90 95 100 105 110-0.6 -0.4 -0.2 0 0.2 0.4 0.6
FIG. 7. The effect of the 450 nm laser on the extracted O − beamcurrent at 1 W laser power with pulse lengths of (a) 250 s, (b) 5 s and(c) 100 ms. to be populated from the 6s ground state of neutral Cs withthe 520 nm laser do not populate the 5d states, which are mostrelevant for the ion pair production mechanism as describedabove. The 6p states that overlap with the tail of the crosssection curve are accessible from the 7s state. However, the6s →
7s transition is optically forbidden, which implies thataccessing the 7s state from the ground state would involve atwo-photon excitation. Hence, the probability of populatingthe 7s state with the 520 nm laser can be considered negligi-ble. The 638 nm laser could presumably promote electrons
90 92 94 96 98 100-300 -200 -100 0 100 200 300Laser ON60 µ s O - b e a m c u rr e n t [ n A ] Time [ µ s] 90 92 94 96 98 100-300 -200 -100 0 100 200 300 FIG. 8. The measured O − beam current response (raw and smootheddata) at the leading edge of the 450 nm laser pulse. The expectedsignal rise time corresponding to a step change of current and takinginto account the time constant of the measurement setup is 60 µs.
120 121 122 123 124 125 126 0 0.2 0.4 0.6 0.8 1ON ON ON ON ON(b) O - b e a m c u rr e n t [ n A ] Time [s] 120 121 122 123 124 125 126 0 0.2 0.4 0.6 0.8 1 112 114 116 118 120 122 124 126 128 130 0 20 40 60 80 100 120 140 160ON ON(a) O - b e a m c u rr e n t [ n A ] Time [s]450 nm520 nm638 nm 112 114 116 118 120 122 124 126 128 130 0 20 40 60 80 100 120 140 160
FIG. 9. The effect of (a) 50 s and (b) 100 ms laser pulses at 450 nm /1 . . . − beam current directly to the 5d states but that can be argued to be unlikelyas the quadrupole transition strength from the ground state tothe 5d states is extremely low . Furthermore, the > . . .
81 eV excitation energy and 1 .
94 eVhoto-assisted negative ion production 9photon energy is too large for a resonant excitation mecha-nism. These arguments are in line with the experimental re-sult, namely the the lack of the prompt effect with the 520 nmand 638 nm lasers.Similarly, the energy difference between 450 nm laser andthe 6s →
7p transition could be argued to question the reso-nant nature of the prompt effect observed with the blue laserpresumably populating the 5d states of neutral Cs via excita-tion to 7p states. This is because less than a 10 − th fractionof the 1 . . → / transition of neutral Cs. This discrepancy motivated us tocontinue the experiments with laser wavelengths shorter than450 nm as explained hereafter. C. Experiments with the 405 nm laser
The experiments were continued with the 405 nm laser,which in principle allows accessing the 4f states of neutral Csfrom the 6s ground state and further populating the 5d states(see Fig. 2). However, the probability of the 6s →
4f excita-tion is extremely low due to corresponding change in orbitalangular momentum being high ( ∆ l = + − current with 1 Wlaser power is shown in Fig. 10. The absolute effect (in nA)of the 405 nm laser is virtually identical to the effect of the450 nm laser at corresponding power and O − beam current.This is strong evidence against a resonant ion pair produc-tion explaining the observed prompt effect. This is due to thedifference in expected excitation probabilities between the 7pand 4f states from the 6s ground state and subsequent branch-ing to the metastable 5d state altogether suggesting that the7p excitation should be more efficient catalyst of resonant ionpair production.
130 132 134 136 138 140-0.2 -0.1 0 0.1 0.2 0.3ON ON ON ON ON O - b e a m c u rr e n t [ n A ] Time [s] 130 132 134 136 138 140-0.2 -0.1 0 0.1 0.2 0.3
FIG. 10. The effect of the 405 nm / 1 W laser exposure on the O − current from the Al O -cathode at 10 Hz laser pulse repetition rate. Following the discovery of the prompt effect with the405 nm laser it was confirmed that the photo-assisted contri-bution on negative ion production can be observed irrespec-tive of the beam current – an important step in assessing the practicality of the method. The beam current was adjusted byvarying the ionizer temperature, i.e. the flux of Cs + ions im-pinging on the cathode surface. Figure 11 shows the O − cur-rent response exhibiting both, the prompt effect and a slow,few second transient.
860 880 900 920 940 960 980 -10 -8 -6 -4 -2 0 2 4 6Laser ON Laser ON O - b e a m c u rr e n t [ n A ] Time [s] 860 880 900 920 940 960 980 -10 -8 -6 -4 -2 0 2 4 6
FIG. 11. The effect of the 405 nm laser exposure on the O − currentfrom the Al O -cathode at elevated beam current. D. Experiments with the high power 445 nm laser
The data presented in Sections IV A-IV C motivated us toconduct further experiments with a high-power, i.e. 6 W,445 nm laser. The purpose of this campaign was to study theeffect of the laser power and wavelength at elevated beam cur-rents as well as observe long-term transients with significantlocalised power deposition presumably having a prononuncedeffect on the cathode surface Cs balance. Fig. 12 shows thebest result obtained with the 6 W laser at elevated O − cur-rent. Both, the prompt effect inducing a 50–100% step andthe subsequent long-term increase (up to another 35–40%) ofthe beam current were observed. O - b e a m c u rr e n t [ n A ] Time [s] 0 500 1000 1500 2000 2500 3000 -100 -50 0 50 100 150
FIG. 12. The best recorded example of the effect of the 445 nm, 6 Wlaser on the O − current. hoto-assisted negative ion production 10The fact that the 445 nm laser, which should not be effi-cient in exciting those states of neutral Cs that populate the5d states, casts doubt onto the ion pair production hypothe-sis. Nevertheless, the observed photo-assisted negative ionproduction effect is encouraging as especially the high powerlaser offers a route to boost the performance of the SNICSsource. However, the magnitude and persistence of the ef-fect was observed to depend on the ion source settings. Un-der some operating conditions the effect fades away i.e. thebeam current starts to decrease gradually following the initialincrease (prompt effect and gradual rise). It is believed thatsuch long-term trends are due to evolving Cs-coverage of thecathode surface. This view is supported by the data shownin Fig. 13 demonstrating that the magnitude of the prompteffect and the time constant of the gradual decrease of thebeam current depend strongly on the time between the laserpulses. When the laser is off Cs presumably accumulates onthe cathode surface and is then removed by ablation or evapo-ration when the laser pulse is applied. In the particular case ofFig. 13 the saturation O − beam current at the end of the laserpulses is lower than without the laser unlike in the examplein Fig. 12. The obvious implication is that the neutral Cs fluxfrom the oven should be adjusted for each set of ion sourceparameters and laser power. O - b e a m c u rr e n t [ n A ] Time [s] 0 500 1000 1500 2000 2500 3000 50 100 150 200 250 300 350 400
FIG. 13. An example of the effect of the 445 nm, 6 W laser on theO − current with randomly varied laser pulse duration and repetitionrate. V. DISCUSSION
The experiments described above have confirmed the exis-tence of a photo-assisted enhancement of negative ion produc-tion in cesium sputter ion sources. In the case of O − ions pro-duced from Al O -cathode the effect consists of two compo-nents; a prompt effect that appears to be insensitive to the laserwavelength above a certain threshold photon energy achievedsomewhere between 2 .
38 eV and 2 .
76 eV (corresponding tothe 520 nm and 450 nm lasers), and a long-term effect whichis presumably driven by evolving Cs coverage of the cathodesurface. The experimental setup and the obtained results do not allow quantifying the possible contribution of photode-tachment (threshold energy of 1 .
46 eV for a free anion) by thelaser-emitted photons on the extracted O − current.The insensitivity of the prompt effect on the laser wave-length disputes the hypothesis of resonant ion pair produc-tion in interaction between excited states of neutral Cs atoms(donors) and oxygen atoms (acceptors). Taking into accountthe discrepancy between the laser emission spectrum and ex-citation wavelength as well as the modest 3 s time resolutionin the earlier experiment with C − ions it is possible thatthe authors of that paper were observing the long-term ef-fect instead of a prompt increase of the beam current. Fur-thermore, the photon-to-anion conversion efficiency deducedfrom the data given in Ref. , namely 20 µW power (fractionof 10 − of the total 2 W power) at 455 . η = N ion N photon = I ion P laser E photon of 68% of the 7p excitations po-tentially resulting in negative ion formation. Here, I ion is thebeam current increase, P photon the laser power at the excitationenergy and E photon the laser photon energy in eV. Taking intoaccount the branching ratio from the 7p state to the 7s state,which is considered to be the relevant one for C − ion pairproduction, would increase the efficiency above unity, thusviolating the conservation of energy even without consideringthe spontaneous lifetime of the excited state and the interac-tion probability of the two atoms. We therefore suggest thatthe effect observed in the cited work is probably due to laser-induced variation of the Cs density (on the cathode surfaceand in the volume in front of the cathode), which could affectthe negative ion production in non-linear manner and wouldbe insensitive to the laser wavelength hence making the 10 − absorption factor irrelevant.It is emphasized that our experiments do not exclude thepossibility of secondary electrons promoting neutral Cs to rel-evant excited states in front of the cathode and thus contribut-ing to the negative ion yield through the pair production mech-anism. Instead we retrospectively question the use of diodelasers for studying the putative mechanism as the observedphoto-assisted negative ion production appears to be insen-sitive to the laser wavelength above a certain threshold en-ergy. It is concluded that the contribution of ion pair produc-tion on the negative ion currents extracted from cesium sput-ter ion sources should be confirmed or disputed with an ad-justable wavelength laser scanning across the relevant wave-lengths corresponding to excitations of neutral Cs.It is possible that secondary electrons emitted from the cath-ode surface promote Cs atoms to the excited states relevant forthe ion pair production as suggested in the literature . In thiscase the laser-induced prompt effect would be best explainedby photoelectrons ejected from the low work function sur-face ( φ < .
38 eV), and contributing to the negative ion yieldby direct attachment or by promoting the ion pair productionthrough enhanced electron impact excitation to the relevantexcited states of neutral Cs in the close proximity of the cath-ode surface. Alternatively, the absorption of photons by theelectrons within the cathode material band structure could in-crease their tunneling probability through the surface potentialhoto-assisted negative ion production 11barrier to the affinity state of the anion . The observed thresh-old behavior together with the photo-assisted gain of the O − beam current apparently depending on the Cs balance, andtherefore the work function, of the cathode surface are consis-tent with the photoelectron hypothesis.Yet another possible mechanism that could explain the ob-servation is photoionization from the metastable 5d states. Inthis case the laser would promote neutral Cs to the metastablestate via the excitation to upper states followed by ionization,i.e. Cs + h ν → Cs ∗ / Cs ∗ + h ν → Cs + . Alternatively the firststep of the process could be facilitated by electron impact ex-citation, i.e. Cs + e → Cs ∗ + e. Such scheme would explain thefact that the photo-assisted effect was observed in the earlierexperiment where the laser was not irradiating the cathodesurface and thereby releasing secondary electrons but insteadexposing the Cs vapor in front of the cathode. The enhancednegative ion yield in this case would be due to increased Cs + flux to the cathode and the corresponding change of the neg-ative ion sputtering rate and Cs balance on the cathode sur-face. Although such effect cannot be excluded, we argue thatpopulating the 5d state via photon absorption and subsequentcascading to the metastable state should be a resonant effectinvolving the same initial step as the putative ion pair pro-duction mechanism. This does not apply if the metastable Cspopulation was produced through electron impact excitationinstead. In that case the enhancement of the Cs + ion flux andthe negative ion yield would be a threshold process with a2 .
08 eV minimum energy corresponding to the ionization po-tential of the Cs(5d) atoms and could, therefore, be driven bythe 520 nm laser which was not observed in the experiment.Altogether, our data showing that the prompt effect is not sen-sitive to the laser photon energy above a certain threshold, notachieved with the 520 nm laser, supports the above explana-tion based on photoelectron emission affecting the negativeion yield through an unknown mechanism.The role of Cs coverage is best illustrated by the experi-ments with the high power laser using varying pulse repeti-tion rates and pulse lengths, demonstrating that with sufficientlaser power and inappropriate Cs coverage the long-term ef-fect of the laser can be adverse. It has been observed that op-erating the SNICS source in pulsed mode can sometimes (un-der certain operating conditions) lead to enhanced beam cur-rents although systematic trends covering various ion specieswere not found. It is plausible that, similar to photo-assistednegative ion production with the laser exposure, the perfor-mance of the SNICS ion source in pulsed mode is sensitive tovariations of the cathode Cs coverage.The photo-assisted negative ion production could be ofpractical importance for the operation of cesium sputter ionsources as demonstrated by the factor of > − current achieved with the 6 W laser. Alternatively themethod could be applied for reducing the erosion rate of thecathode and, thus, increasing its lifetime by enabling to reachthe same beam current (as without the laser) at reduced Cs + flux. A complete assessment of the method’s potential re-quires experiments with other negative ions and cathode ma-terials, i.e. metals and compounds, especially those that typ-ically have low negative ion yields. The role of Cs could be best studied with cathode materials made of Cs compounds,such as CsCl typically used for the production of Cl − ions.The transient effects of extracted beam current, presumablycaused by the fluctuation of the cathode Cs coverage, couldbe suppressed in the case of cathode materials with intrinsicCs content.It is expected that in cesiated plasma ion sources the bene-fits of exposing the negative ion production surface to a photonflux from an external source are limited as plasmas naturallyradiate up to several tens of percent of the discharge power inUV/VUV-range resulting in significant photoelectron emis-sion from cesiated surfaces . However, if the follow-up ex-periments on cesium sputter ion sources with an adjustablewavelength laser were to reveal a significant contribution bythe resonant pair production effect, experiments on laser-assisted negative ion production in the discharge volume ofcesiated plasma ion sources would be justified. ACKNOWLEDGMENTS
MN acknowledges financial support from Her Majesty’sCustoms & Revenues (HMCR) Coronavirus Job RetentionScheme. G. D. Alton, "Negative ion formation processes: A generalreview", Tsukuba Polarized Workshop 1990:0339-374, Re-port number: CONF-900289-4, (1980), available online athttps://digital.library.unt.edu/ark:/67531/metadc1191487/ M. L Yu, Phys. Lett. 40, (1978) 574. C. F. A. van Os, E. H. A. Granneman and P. W. van Amersfoort, J. Appl.Phys. 61 (11), 1987. B. Rasser, J. N. M. van Wunnik and J. Los, Surface Science 118 (1982)697-710. D. Faircloth and S. Lawrie, 2018 New J. Phys. 20 025007 R. S. Hemsworth, D. Boilson, P. Blatchford, M. Dalla Palma, G. Chitarin,H. P. L. de Esch, F. Geli, M. Dremel, J. Graceffa and D. Marcuzzi, 2017New J. Phys. 19 025005. V.E. Krohn, J. Appl. Phys., 33 (1962), pp. 3523-3525. R. Middleton, and J. Klein, Physical Review A. 60 (1999) 3786. J. S. Vogel, Nuclear Instruments and Methods in Physics Research SectionB: Beam Interactions with Materials and Atoms Volume 438, 1 January2019, Pages 89-95. Y.-T.-T. Lee, and B. H. Mahan, The Journal of Chemical Physics 42 (1965)2893-2896. R. Grice and D. R. R. Herschbach, Mole. Phys. 27 (1974) 159-175. E. Y. Buslov and B. A. Zon, Phys. Rev. A. 85 (2012). A. A. Narits, E. S. Mironchuk, and V. S. Lebedev, Jour. Phys B:. 47 (2013)015202. G. Cartry, D. Kogut, K. Achkasov, J.-M. Layet, T. Farley, A. Gicquel, J.Achard, O. Brinza, T. Bieber and H. Khemliche, 2017 New J. Phys. 19025010. P. Williams, Surf. Sci. 20. (1979) 588. G. Slodzian, Surf. Sci. 42 (1975) 161. J. Blahins, T. Leopold, A. Apsitis, U. Berzins, A. Ubelis, J. Rohlén, D. Luand D. Hanstorp, Rev. Sci. Instrum. 91, 023322 (2020). R.B. Vora, J.E. Turner, and R.N. Compton, Phys. Rev. A. 9 (1974) 2532. B. Warner, Monthly Notices of the Royal Astronomical Society 139 (1968),115–128. A. A. Vasilyev, I. M. Savukov, M. S. Safronova and H. G. Berry, Phys. Rev.A, 66, 020101, (2002). R. Middleton, Nuclear Instruments and Methods in Physics Research Vol-ume 214, Issues 2–3, 1 September 1983, Pages 139-150. hoto-assisted negative ion production 12 R. Kronholm, T. Kalvas, H. Koivisto and O. Tarvainen, Rev. Sci. Instrum.89, 043506 (2018). S. Pucher, P. Schneeweiss, A. Rauschenbeutel and A. Dareau, Phys. Rev. A101, 042510 (2020). J.S. Vogel, AIP Conference Proceedings. 1515 (2013) 89-98. J. Komppula and O. Tarvainen, Phys. Plasmas 22, 103516 (2015).27