Multi-Watt cavity for 266 nm light in vacuum

Intense coherent ultraviolet radiation is gaining increasing importance in advanced quantum technologies—from optical clocks and quantum computers to matter-wave interferometry—as well as in photochemistry, life sciences, semiconductor industry, and space applications. Since the preparation of multi-Watt light sources is still an open challenge for many ultraviolet wavelengths, resonant enhancement in a cavity is an attractive alternative. However, many experiments with atoms, molecules or nanoparticles require isolation in high vacuum where UV optics often show fast degradation. Here, we present stable performance of a cavity for 266 nm light with several Watt of intra-cavity power in high vacuum despite the presence of hydrocarbons. Comparing two sets of cavity mirrors indicates that this feat is connected to the micro-chemical environment at the topmost coating layer. Our study emphasizes the need for further developments in this direction to facilitate robust, compact, and high-performing devices employing UV radiation.

An integral part of this process is the chemical micro-environment at the topmost mirror layer. In particular, substrate coating technology has a strong influence on the binding of adducts. Both the porosity and the availability of possible reaction sites on the surface are important [23,24]. Even catalytic properties of defects have been discussed [26], in line with the observation that coatings with high internal strain are more strongly contaminated than uncoated substrates [6,27,28].
While the covalent bonds between the mirror and the contamination prevent thermal cleaning [7], the deposited carbon can be removed by ozone [9,26,29], oxygen plasmas [3,30] or oxygen at a partial pressure above 20 Pa [4,9,12,31], preferentially in the presence of ultraviolet laser radiation. Constantly flushing the optics with oxygen prevents degradation entirely [2,18] and high power UV-cavities in a segmented vacuum chamber have been demonstrated [15]. Alternatively, removing organic contamination entirely is a very successful route to ensure long-term stability [11,32,33].
These methods require either a very clean system or a highly elaborate and large setup to prevent or undo the degradation of the optics. This becomes an issue as soon as the size is restricted as in space missions or as organic components are to be investigated since they are precursors to contamination on the substrate. For wavelengths around 390 nm, it has been argued that sealing the surface with a layer of SiO 2 can prevent degradation of the Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
optics [19]. However, this finding has been contested by others who reported contamination even for optics with a thick SiO 2 top layer [20].
The discussion is further complicated by the fact that the experiments were performed under a large variety of boundary conditions. These encompass the particulars of the coating process, the pressure regime used, the chemical nature of the hydrocarbons, and so on. Here we compare the performance of two sets of cavity mirrors, both coated for 266 nm, in the same vacuum setup. The mirrors were manufactured by two different companies according to the same specifications and both systems feature a SiO 2 layer on top. Still, their properties vary dramatically. While mirror set (A) degrades on the time scale of seconds as soon as the cavity is operated in prevacuum, mirror set (B) retains its properties under all tested conditions (5 × 10 −10 − 0.1 mbar). Set (B) thus allows us to run a multi-Watt cavity at 266 nm in vacuum in the presence of hydrocarbons without degradation. This strongly suggests that the chemical micro-environment on the mirror surface determines its performance under the tested conditions.

Experimental setup
We generate continuous high-power radiation at λ = 266 nm by frequency-doubling a high-power, single-line DPSS laser beam (Coherent Verdi V10, P 10 W, λ = 532 nm, TEM 00 , FWHM line width Δν = 5 MHz) in an external cavity frequency doubler (Sirah WaveTrain2). We reach an output power of P 266 = 1 W for a fundamental input power of P 532 = 3.5 W and even P 266 = 3 W when pumping with P 532 = 8 W. However, we usually restrict the output to P 266 0.5 W to increase the lifetime of the UV optics. To facilitate the coupling to the cavity in vacuum, the UV beam is shaped using a cylindrical and a spherical telescope as shown in figure 1. The cylindrical telescope consists of the lenses c1 ( f = 75 mm) and c2 ( f = 50 mm) while the spherical telescope is composed of the lenses s1 ( f = 200 mm) and s2 ( f = 100 mm). These are adjusted to realize a Gaussian beam with a 1/e 2 diameter of 640 ± 20 μ m at the position of the plan-convex coupling lens CL ( f = 250 mm). Coupling into the cavity is optimized using two bending mirrors (not shown) behind the coupling lens. The light reflected by the cavity is registered with a suitable photodetector Det1 (Thorlabs DET25K/M).
The cavity consists of two spherical mirrors produced by ion beam sputtering. They have a radius of curvature of r = 100 mm and a nominal reflectivity of R = (98.75 ± 0.25)%. The high-reflective mirror surfaces are separated by L c = 3.34 mm, resulting in a free-spectral range of ν FSR = 44.9 GHz and a cavity waist of w 0 = 33 μ m in the center and only 0.3 μ m larger on the mirror [34,35]. Given a finesse of F = 250, the line width amounts to Δν 1/2 = 180 MHz (FWHM), which is more than one order of magnitude larger than the one of the pump laser. The cavity length can be adjusted using a ring piezo (Piezomechanik HPCh 150/10-5/3), which carries the cavity mirror M2. The impedance matching and coupling efficiency of the light to the cavity are estimated by scanning over the cavity resonance and monitoring the amount of the reflected light. To record the cavity spectrum, 10% of the transmitted light is guided by the beam splitter BS to the photo-detector Det2 (Thorlabs PDA10A2). The reading from this detector is also used to lock the cavity to the incident beam via a side-of-fringe locking scheme. Here, we typically set the lock to 50 − 75% of the peak intensity of the TEM 00 mode. The intensity inside the cavity is calculated from the power detected with the power meter PM (Coherent PowerMax USB PM 150-50C) behind the cavity, the enhancement factor F/π ≈ 80, and the power reduction due to the beam splitter (10%).
The cavity is mounted in a simple aluminum frame and placed in a UHV-compatible vacuum chamber. A turbo molecular pump (700 l/s) backed by a rotary vane pump together with a cold baffle (77 K) can reduce the pressure at room temperature to below 5 × 10 −10 mbar. The main sources of contamination originate from the rotary vane pump and several Viton rings. The latter are used to seal the two entrance windows of the chamber, damp vibrations of the cavity, and preload the ring piezo.

Performance of set (A)
We start with characterizing the cavity mirror set (A). At ambient pressure, these mirrors show the desired behavior: Once optimized, the cavity spectrum remains stable and does not degrade over time. This allows for locking the cavity with several Watt of intra-cavity power for more than an hour in air, as shown in the Supplementary Material. Comparing the cavity spectrum before and after locking shows no signs of degradation either.
The situation changes as soon as we reduce the pressure p. Below a certain threshold, the cavity spectrum decays, leading to a time-dependent decrease in the intensity of the TEM 00 mode (Supplementary Material). For the current data set, this point is reached around 1-2 mbar. To follow this process, we optimized the mode coupling at a pressure where no degradation takes place, that is, at 10 mbar. Then, we blocked the incoming laser and evacuated the cavity to the target pressure. As soon as we removed the beam dump and exposed the cavity to the UV light, its spectrum started to decay as shown for p = 0.1 mbar in figure 2(a)). During the whole procedure, we continuously scanned the cavity with 100 Hz over a large portion of the spectrum (about 40 GHz). Fitting the time-dependent peak value of the TEM 00 with a single-exponential decay yields the amplitude A, the offset y 0 , and the mean decay lifetime τ dec , which amounts to 7.7 ± 0.2 s in this run. The degradation can be reversed completely by leaking in laboratory air to increase the pressure above 2 mbar while pumping the cavity with UV light, as shown in figure 2(b). Venting the vacuum chamber with pure nitrogen had no beneficial effect. The time-dependent increase in transmitted power is reproduced well by a logistic function which allows us to extract the mean recovery time τ rec and the midpoint t c . Repeating the cycle of decay and recovery in the pressure range between 0.1 − 22 mbar shows that these processes approach a threshold at about 1 − 2 mbar as mentioned before (see figure 3). At this pressure p thresh , neither a complete decay nor complete recovery is observed.  The decay lifetime strongly depends on the cleanliness of the vacuum chamber. After baking and keeping the test setup at 10 −9 mbar for a few weeks, we repeated the measurement at 0.1 mbar. This led to an increase of τ dec from 7.7 ± 0.2 s to 195 ± 3 s, suggesting that p thresh is shifted to lower pressures. To see whether this allows for stable operation of the cavity in UHV, we performed the decay experiment at p = 5 × 10 −10 mbar. Although the mean lifetime τ dec was greatly enhanced to 68 ± 3 minutes, we still observed an exponential decay even under these conditions (Supplementary Material).
In the experiments shown in figure 2 the pump power coupled in was held constant at 40 mW.

Performance of mirror set (B)
Exchanging mirror set (A) with (B) has a pronounced effect on the behavior of the cavity at low pressures. In contrast to set (A), we do not observe any time-dependent degradation for mirror set (B), when scanning over the cavity spectrum in vacuum. This allows us to stabilize the cavity with several Watt of intra-cavity power, typically at p = 5 × 10 −10 mbar. An exemplary run is shown in figure 4: here we locked the cavity with an intracavity power of 4.4 W for more than an hour. The highest intra-cavity power we observed was 5.6 W, which remained stable for half an hour. In total, we have accumulated several hours of stable performance at this power without any signs of degradation over a period of two months. Between the runs, the cavity was always kept at a pressure below 10 −8 mbar and at 300 K.

Discussions
In the experiments, mirror set (A) shows typical signs of degradation due to contamination. First, we observe a rather sharp pressure threshold at which the decay of the cavity sets in [36]. Second, oxygen is required for the recovery of the cavity spectrum, as venting the chamber with pure nitrogen has no effect [29]. The partial pressure of oxygen at the threshold in figure 3 is around 20 Pa, as observed before [4,9,12,31]. After extensive cleaning of the setup, p thresh shifts to lower values. This is in agreement with the observation that p thresh is determined by the ratio of the partial pressures of oxygen and the residual hydrocarbons [36]. Thus, reducing the partial pressure of the hydrocarbons allows for reducing the partial pressure of oxygen as well. While this increases the longevity of the mirrors, it cannot prevent their degrading in the presence of UV radiation [4]: Even after baking and pumping to p = 5 × 10 −10 mbar, the cavity spectrum is degrading. To mend this, we would have to remove all hydrocarbons from the vacuum setup [11,32]. However, as the purpose of the cavity is to provide an intense laser grating for matter-wave diffraction of large organic molecules [37,38], this is not an option. Flushing the mirrors with oxygen has to be discarded as well, as the resulting increase in pressure would induce collisional decoherence of the matter-wave [39].
In our experiments, we observe comparatively short values of τ dec compared to the literature. On the one hand, this might be connected to the composition of the hydrocarbons. In most previous experiments, highly volatile substances such as small aromatics have been used to precisely control the respective partial pressures. As our main source of contamination is pump oil, we expect the hydrocarbons to be large, long-chained molecules. These molecules have been identified to pose a greater threat for contamination in EUV lithography than lighter ones [40]. On the other hand, also the wavelength might play a role. Light at 266 nm is resonant both to electronic transitions in aromatic hydrocarbons and non-bonding oxygen hole center-defects (NBOHC) in strained silica films [6,25,41]. Hence, we might excite both the contaminating particles and the possible reaction sites very effectively [6].
The most important question, however, is why the mirrors of set (B) are not degrading over time. As the top layer in both sets is SiO 2 , this cannot be explained by the material, alone. Instead, it seems to be related to the micro-chemical environment of the surface. Previous studies have highlighted the effect of Si-OH groups, strained silica bonds, and defects [8,23,24,41,42]. Especially the latter seems to play a critical role in the creation of reactive hydrocarbons on the silica surface. Doping silica with fluorine has been shown to reduce the number of strained three-and four-membered rings considerably [43,44]. This can reduce the defect concentration by about one order of magnitude compared to pure samples [45] and significantly increase the lifetime of mirrors in UV cavities [46].
Based on our observations, it seems that this feat can also be achieved without the use of a process gas, as demonstrated by mirror set (B). This insight also gives the recent discussion on protective SiO 2 layers for UV optics [19][20][21] a new direction where the micro-chemical environment of the coatings was largely ignored. So, the question is not whether SiO 2 helps or not, but which requirements the top layer has to meet to obtain contamination-resistant coatings. The present findings suggest that the answer to this question might be of vital importance to reliably manufacture contamination-resistant coatings for UV wavelengths in a vacuum environment.

Conclusion
We tested the performance of two sets of cavity mirrors for 266 nm in vacuum. While these were fabricated to the same specifications and featured SiO 2 on top, they behaved quite differently in the presence of hydrocarbons. Set (A) showed typical signs of contamination, which led to a decay of the cavity spectrum on the order of seconds while scanning over the cavity resonance. Set (B), however, did not show this behavior, which allowed locking the cavity with several Watts of intra-cavity power in vacuum. This required neither a completely clean setup nor flushing the mirrors with oxygen. We attribute this difference in performance to the micro-chemical environment of the topmost coating layer. Our study highlights the need to better characterize the deposited thin films to reliably produce contamination-resistant coatings. This is of high interest for the realization of compact and reliable setups employing intense radiation in the UV.