Kinetics of alkali-based photocathode degradation Open Access

The pace of high brightness electron source development has increased in response to design requirements for future x-ray free-electron lasers (XFEL) and x-ray energy-recovery linacs (XERL).¹ The majority of work still involves conventional photocathode materials, both metallic and semiconductor, but emphasizes improved beam brightness, quantum efficiency (QE), and effective lifetime.² Strategies for improving each of these vary, but in all cases the focus is on understanding and influencing the relevant cathode surface physics.³ Recent strategies for improving photocathode lifetimes have been proposed including the use of ultra-thin protective films.⁴ Thus, quantitative metrics are needed for meaningful characterization of these techniques as targeted in this Letter.

While alkali-based photocathodes with high QE in the visible range⁵ are attractive candidates as sources of high brightness electron beams,^6–8 they are known to suffer from fast degradation under typical UHV conditions.^7,9–11 Several distinct processes lead to their QE degradation, such as surface “poisoning,” deviation from stoichiometry, ion back bombardment, electrolysis, and thermally activated decomposition.⁵ Of these, poisoning due to surface reactions with residual gases such as O₂, CO₂, CO, and H₂O that persists even in the absence of illumination and current extraction is arguably the most important for RF-based photoinjectors and photocathode storage systems in service today.

Despite the fact that degradation due to poisoning by reactive gases (all of which contain oxygen) has been known for decades, a complete description of the phenomenon is complicated because it occurs among other competing mechanisms.^8,12 From an experimental standpoint, the main reason for this lack of knowledge is that the extreme reactivity of alkali-based photocathodes almost completely precludes ex situ diagnostics. In general, in situ surface diagnostics techniques cannot deliver sensitivities that exceed, or are at least comparable to, that of the photoelectric effect itself. It is well known that as little as a monolayer coverage on a semiconductor photocathode surface can enhance or inhibit QE, at a given wavelength, by many orders of magnitude due to induced changes in the electron affinity.^5,13,14

Lifetimes of photocathodes are typically characterized by indicating how much the QE decreases over a given time interval at a certain base pressure in the UHV system.^7,10,12 Many authors quote 1/e lifetimes without determining whether the degradation curve is exponential.^8,11,12,15 Some attempts have been made to introduce a more practical and universal metric for the lifetime by integrating pressure over time to obtain exposure expressed in Langmuir units.^6,10,11 Only one study conducted for (NEA) GaAs photocathodes provides reliable data that correlates QE with the time and the pressure of individual gases.¹¹ These materials owe their high QE under visible illumination to a monolayer of Cs-based dipoles produced during the surface activation process.¹³ Besides investigating this previous report, in this Letter we present QE degradation data on a representative sample of another class of photocathodes, the alkali antimonides that we fabricated and measured. These complementary data sets were used to validate the kinetic model presented below and to obtain values for the rate coefficients that characterize QE degradation.

We used an established procedure^16,17 to fabricate Cs₃Sb photocathodes on Si substrates in a UHV chamber with a base pressure of 8×10^-10 Torr. QE degradation curves were obtained by monitoring the current of electrons collected by a ≈25-mm diameter anode tube positioned 2-3 mm away from the cathode plane. The cathode was illuminated at 405 nm by a few-mW laser diode at near normal incidence. The corresponding photon energy (∼3.06 eV) lies in a region where the spectral response curve is relatively flat.⁵ A low voltage (18 V) was applied between anode and cathode to minimize residual gas ionization. Our measured QE’s at 405 nm were typically slightly lower than 10%, indicative of deviation from ideal Cs₃Sb stoichiometry attributable to less than ideal vacuum quality during photocathode fabrication: the pressure would increase to the high 10^-9 Torr range due to gases released by heating the Sb evaporation cell and Cs chromate sources (SAES). For controlled degradation, pure O₂ was introduced by means of a leak valve. All gas partial pressures were measured using a residual gas analyzer (SRS RGA200 with CEM).

The standard description of the reactivity of gas molecules with surfaces dates back to the Langmuir’s seminal paper of 1918.¹⁸ The Langmuir model describes the surface as a collection of binding sites where the gas molecules reversibly react. The rate at which the surface is covered by gas molecules is a competition between the adsorption rate (k_ads) and the rate of desorption (k_ads). The number of surface unreacted sites (N), decreases exponentially from the initial number of available sites (N₀) with a rate $λ=kads+kdes$ as given by:

At long times, the number of available sites reaches an equilibrium population, N_eq = N₀ × k_des/(k_ads+k_des). A corollary of this model is that removing the gas pressure, i.e. setting k_ads=0, results in the surface recovering at the rate of desorption until it is completely healed.

In this work, we studied the QE degradation of alkali-based photocathodes upon exposure to three gases, O₂, CO₂, and CO. We focused on a macroscopic description of the degradation process, whereas a microscopic study will be the topic of future work. Our observation of QE degradation indicates that it follows a double exponential decay mode and exhibits only partial recovery upon improving vacuum conditions. This type of behavior cannot be accurately represented by the Langmuir adsorption model, but suggests a two-step mechanism wherein the gas molecules can adsorb and desorb, or irreversibly react with the surface as described by the following process:

where we have labeled a site in a pristine, photoemissive state before reaction as C₁, and interpret the two other states as a physisorbed intermediate state, C₂, and an irreversibly damaged chemisorbed state, C₃ (see Fig. 1). In this model we assume that the QE is proportional to the number of unreacted C₁ sites.

The rate k_ads is the same as the rate of adsorption in Langmuir’s model and is proportional to the gas pressure and the sticking coefficient of the gas molecules onto the cathode. The rate k_des corresponds to the rate of desorption of gas molecules from the physisorbed state C₂ back into the gas phase. Finally, the rate k_irr represents the rate at which physisorbed molecules irreversibly react with the surface to permanently change it via chemisorption.

The kinetics equations that govern the number of surface sites in each state are:

where N_i is the number of sites in state C_i. The analytical solutions of equations (2) thru (4) are given by:

where $λ±=12×[(kads+kdes+kirr)+α]$⁠, $α=(kads+kdes+kirr)2−4×kads×kirr$⁠, and N_i(0) is the initial number of sites in state C_i. With the appropriate reaction rates, the behavior of N₁ follows the experimentally observed degradation of photocathodes QE as discussed below.

The solid lines in Fig. 2 show the changes in QE for a NEA photocathode upon exposure to O₂, CO₂, and CO¹¹ and for a Cs₃Sb photocathode exposed to O₂. The QE degradation curves show up to three distinct regions: (I) under UHV conditions extremely little to no QE degradation occurs; (II) the surfaces are exposed to various gases beginning at t₀ and show significant decreases in QE; and (III) the gas is removed leading to some QE recovery (e.g., see Fig 2 top, NEA-O₂).

The model described above was numerically fit to region II. The rates obtained for the four cathode-gas systems are listed in Table I, and correspond to single degradation measurements. Variations in sample preparation yield dispersion in the fitted parameters of a factor of 2 to 4. However, the model consistently describes the experimental data over a wide range of sample conditions.

The dashed lines in Fig. 2 were obtained by integrating the kinetic equations (2) thru (4) using the values of k_des and k_irr indicated in Table I, since these rates represent the reactivity of the surface with gas molecules at a given temperature. However, k_ads changes in proportion to the gas pressure. Validation of the model and the parameter set is ensured by observing that the recovery steps for NEA-O₂ and NEA-CO₂ are accurately described, even though this data was not part of the input into the parameterization. We did observe similar recovery steps for the Cs₃Sb-O₂ system, but relatively high partial base pressure levels of the damaging gases complicate quantitative analysis.

The time evolution in the population of sites in states C₁, C₂, and C₃ provides insight into the dynamics presented by the model. As shown in Fig. 3, Region I corresponds to the pristine, as prepared, GaAs NEA photocathode. Upon O₂ exposure (start of Region II), the physisorbed sites, C₂, quickly populate at the same rate as the C₁ sites are consumed. The population of C₂ sites then peaks due to the competition between physisorption/desorption and the chemisorption processes. Defining the surface coverage as in the Langmuir model, $η=N2/(N1+N2)$ , an equilibrium degradation mode is then reached as evidenced by the plateau at . Upon removal of the gas (Region III) the C₂ sites disappear via desorption and irreversible reaction, leading eventually to zero coverage. In the standard Langmuir model, the population of C₁ sites would totally recover back to its initial population of N₁=1. However, in the present model due to the permanently damaged surface sites, only partial recovery is observed.

The equilibrium coverages, $ηeq$⁠, in the region of exposure and decay for the four cases studied are also listed in Table I and indicate the relative importance of the physisorbed state: an equilibrium coverage closer to unity shows a more prevalent physisorbed state during the degradation. However, even when a low gas pressure results in a low $ηeq$ value, as would be evidenced by a smaller deviation of the QE degradation curve from a single exponential decay, the effective degradation rate would still be determined by the competition between k_ads, k_des and k_irr. Thus an attempt to characterize QE degradation by a single rate will always result in the loss of important information.

The reaction rates k_ads, k_des and k_irr provide quantitative metrics for photocathode lifetimes since the QE degradation curves can then be predicted for any particular environment. The rates for a given cathode-gas system are determined by the physical and chemical properties of the surface. The data in Table I indicate that the reactivity of the Cs₃Sb surface is much lower than that of the NEA GaAs surface. That difference is largely determined by the rates of adsorption, k_ads. We consider this factor as the primary reason for the longer observed lifetimes of the alkali antimonide photocathodes as opposed to the suggested “bulk” nature of such materials.¹⁷ 

As discussed above, when normalized by both pressure and density of sites, the Langmuir adsorption model rate constant, k_ads, represents a physically significant parameter - the sticking coefficient. Even though the pressure and density of sites are not known with quantitative accuracy, valid relative comparisons can be drawn from the data in Table I: For NEA-activated surfaces, the sticking probability of an O₂ molecule is about three times higher as compared to a CO₂ and more than an order of magnitude higher as compared to a CO molecule.

We note that there is an additional uncertainty in the Cs₃Sb parameters quoted in Table I since limitations of our photocathode fabrication system do not guarantee perfect stoichiometry of Cs₃Sb or ideal purity of the dosing gas. Nevertheless, our QE degradation measurement results are applicable based on their reproducibility, and improvements in their accuracy will be reported in future work. The main purpose of this Letter is not to provide precise rate coefficients, but to illustrate and encourage the adoption of the more quantitative metrics for describing QE degradation presented herein.

In conclusion, a kinetic model is presented that accurately describes the QE degradation process in alkali-based photocathodes due to background gases in the vacuum chamber. The model accurately predicts the observed QE recovery upon removal of the damaging gas, validating the model and highlighting the importance of including a physisorbed intermediate state in the process of degradation into an irreversibly damaged surface. Our model departs from the standard Langmuir adsorption model. This his model describes degradation for both “bulk” Cs₃Sb and surface-activated GaAs NEA photocathodes providing strong evidence that the QE degradation in these materials is determined by surface processes.

We gratefully acknowledge the support of the U.S. Department of Energy through the LANL/LDRD Program. Los Alamos National Laboratory is operated by Los Alamos National Security, LLC, for the National Nuclear Security Administration of U.S. Department of Energy (contract DE-AC52-06NA25396). We thank J. Lewellen and V. Bermudez for critical reading of the manuscript, D. Lizon, A. Mohite, G. Gupta, and H. Yamaguchi for help with the sample fabrication, and S. Gerashchenko, and A. Malyzhenkov for valuable technical discussions.