25 W of average power at 172 nm in the vacuum ultraviolet from flat, efficient lamps driven by interlaced arrays of microcavity plasmas

Few sources of radiation having an average power $≥1$ W currently exist in the vacuum ultraviolet (VUV) and deep-UV regions of the spectrum (100–250 nm). Despite more than five decades of effort since the first UV laser (N₂: 337 nm) was reported by Heard,¹ the only lasers capable of generating 1−50 W are the F₂, ArF, and KrF systems (157 nm, 193 nm, and 248 nm, respectively). The latter two quite literally transformed the deep-UV spectral region but are physically large (⁠$>0.7$ m³ for a 30 W, 193 nm system), expensive, typically limited to a pulse repetition frequency (PRF) of 100 Hz−5 kHz, and offer efficiencies of no more than a few percent.^2,3 A similar situation exists with regard to incoherent optical sources. Although the Xe₂ molecular lamp (⁠$λ=172$ nm) is the most efficient emitter available in the VUV, intensities generated by commercially-available lamps are generally restricted to 30–50 mW cm⁻², and their cylindrical geometry is not amenable to the spatially uniform irradiance of a surface.^4–8 To briefly summarize the status of source development in the 100–250 nm region, the dearth of inexpensive and compact, but powerful, lamps and lasers has impeded the progress of VUV photonics and its applications (such as photochemistry, photopolymerization, water purification, and surface physics).

We report here the development, structure, and characteristics of flat, thin lamps generating more than 25 W of average power, and ≥600 W of peak power, at 172 nm from an active area of 80 cm² (for a 10 × 10 cm² lamp). Driven by at least two interlaced arrays of microcavity plasmas⁹ produced on an interior surface, the lamp has an electrical-to-optical conversion efficiency >20%, a thickness and volume of <6 mm and $6×10−5$ m³, respectively, and a duty cycle of 3% when the PRF of the bipolar, driving voltage waveforms is 135 kHz. The latter is four orders of magnitude larger than that for an ArF or F₂ laser operating with a PRF of 100 Hz. Producing Xe microplasmas in at least two arrays of microcavities of distinctly different dimensions and/or geometry has the effect of utilizing the power pulse (V⋅I) more efficiently than is possible with a single array of microcavities of given dimensions. The result is the realization of 172 nm intensities beyond 350 mW cm⁻² and the generation, in every cycle of the voltage waveform, of four 70 ± 10 ns FWHM, 40–60 $μ$J pulses having a peak power of 600–850 W.

These lamps are readily tiled to cover large surface areas (>1 m²), and the expected power output with current lamp designs is $>2.5$ kW m⁻². The power output observed to date corresponds to ∼0.3 W of average power at 172 nm per cm³ of lamp volume. Extensive testing shows the lifetime of lamps having an internal phosphor to be >20 000 h.

Figure 1 is a diagram in cross section of one design of a microplasma-driven VUV/UV lamp. At least two arrays of microcavities are fabricated into an interior face of a thin (1 mm in thickness) plate of fused silica which also serves as the rear window of the lamp. Because the wall profile and the refractive index of the fused silica substrate determine the spatial variation of the electric field strength in each microcavity during lamp operation, the geometry of the cavities in each of the two (or more) arrays can be chosen so as to ignite microplasmas at different points in the driving voltage waveform. As discussed later, this aspect of the lamp design is largely responsible for the factor of 5–10 increase in the output intensity of these lamps, relative to conventional dielectric barrier discharge (DBD) sources. A variety of microcavity dimensions and geometries have been tested to date, but most designs have employed hemispherical or parabolic cavities (in cross section) with diameters in the 200 μm−2 mm range. Also, the microcavities can be fabricated by any of a number of processes, including laser or micropowder ablation¹⁰ and chemical etching. In all cases, the array topology is defined by photolithography.

A second fused-silica plate in Fig. 1 serves as the output (front) window and the lamp is hard-sealed with custom, lead-free frit designed to function stably at temperatures ≤1000 °C. A time-varying voltage is applied to the lamp through electrode arrays applied to the exterior face of the two fused-silica windows. Several electrode configurations have been studied, including mesh comprising 15 μm wide Ni lines with a pitch of 1 mm that was defined on each window face by photolithography and deposited by the evaporation of a sequence of CrO_x, Cr, and Ni films. The transmission of this electrode pattern was calculated to be 96%-98%. For most of the results presented here, the voltage waveform was that of a bipolar pulse sequence in which a positive voltage pulse 1.5 μs in width was followed, after a 1.75 μs delay, by a negative pulse, also 1.5 μs in duration. The rise time of each pulse was $<75$ ns and, between the pulses, the lamp was grounded to remove surface charge built-up on the interior face associated with the cathode during the preceding voltage half-cycle.

For several generations of lamps, additional silica plates were affixed to the outside faces of both windows of the lamp for physical and electrical isolation of the electrodes. Despite the addition of the two external plates, the lamp thickness remains <6 mm. A small tube installed through the rear window of the lamp allows for the lamp interior to be evacuated by a turbomolecular pump to ∼10⁻⁷ Torr. All lamps are filled with a mixture of research grade Xe and He or Ne, and degassing is facilitated by operation of the lamp, followed by evacuation and refill. Figure 2 is a photograph (in plan view) of a 10 × 10 cm² (4″ × 4″) Xe₂ lamp operating in a 70% Xe/Ne mixture at a pressure of 550 Torr (300 K pressure). The inset at upper right is a time-averaged and magnified view of a portion of the rear interior surface and, specifically, two interlaced microcavity plasma arrays. In an effort to show more clearly the microplasma arrays, a composite of two optical micrographs for one lamp design having the fine-line electrode mesh described earlier and hemispherical microcavities is presented in Fig. 3. As exhibited by the inset (upper left) in Fig. 3, visualization of the microplasmas is enhanced by introducing a few ppm of air to the lamp, thereby forming the XeO molecule¹¹ responsible for the observed green emission (⁠$λ≃$ 550 nm). Recorded for a 20 kHz sinusoidal voltage exciting the lamps, the two images of Fig. 3 show microplasma produced in both arrays of cavities but the intensification of emission near the axis of the larger cavities, in particular. The diameters of the larger hemispherical cavities for most lamps (such as those presented in Figs. 2 and 3) are normally in the 800 μm−2 mm interval, and the blue circles are silica spacers.

False color images of the spatial intensity maps for the microplasma arrays, recorded at two specific times during the bipolar voltage waveform, are shown in Fig. 4. Panel (a) of the figure was acquired by an intensified charge-coupled device (ICCD) camera in a 10 ns window at t = 80 ns, where t = 0 is defined as the onset of the positive pulse in the waveform (i.e., the microcavity arrays are in proximity to the mesh serving as a cathode). This image captures the emission intensity spatial distribution during the period of maximum instantaneous power production by the lamp−i.e., the peak of the first of four pulses illustrated in Fig. 4(c). Although the color scale is nearly saturated, only a small portion of the lamp surface is presented here for clarity. Specifically, three periods of the dual microcavity array topology are shown, and it is evident that plasma fully occupies the smaller cavities at this point in time but, for the large cavities, emission emanates primarily from the cavity perimeter. However, when t = 1.59 μs, the spatial distribution of 172 nm emission in the larger cavities reverses. The image of panel (b) in Fig. 4 was recorded 90 ns after the positive voltage pulse in the bipolar waveform was terminated, thereby generating the second $Xe2*$ fluorescence pulse in the four pulse train of Fig. 4(c). In this situation, the central portion of the larger microcavities is emitting uniformly but the fluorescence produced near the cavity wall is weakened. We interpret this behavior as the result of charging of the fused-silica interior surface. Early in the positive pulse (t ≲ 100 ns), the plasma sheath is not well-developed¹² and emission occurs primarily near the wall because of the role of electron recombination at the cavity surface. Later in the driving pulse waveform, the sheath and the bulk plasma are fully established, and the latter is critically dependent upon ionization driven by electrons accelerated in the sheath. Similar dynamics are at work during the negative voltage pulse which gives rise to the third and fourth fluorescence pulses of Fig. 4(c). All four of the emission pulses generated during each cycle of the bipolar voltage waveform coincide with either the fast rising or falling portions of the waveform and have a duration of 70 ± 10 ns FWHM. It must also be emphasized that the images of Fig. 4, representative of those recorded for other lamps, demonstrate that interweaving at least two arrays of microcavities is effective in improving the utilization of the electrical power (V⋅I) pulse driving the plasma.

Lamp emission spectra were recorded with a 0.2 m scanning VUV spectrometer and a photomultiplier, and Fig. 5 illustrates the spectral profiles observed for 70% Xe/He gas mixtures for which the total pressure is raised in 50 Torr increments from 100 Torr to 700 Torr. For this range in pressure, no significant change to the spectrum is observed, although the wavelength-integrated emission rises rapidly with increasing pressure because the transient Xe₂ molecule is known to be formed by teratomic recombination.¹³ Peak emission occurs at ∼173 nm, and the breadth of the spectrum is ∼9 nm FWHM which is ∼35% smaller than that reported for previous Xe₂ lamps (14 nm¹⁴). The reduction in bandwidth relative to existing sources is presumed to be linked to the increased operating pressure afforded by the microcavities, thereby raising the vibrational-translational (V-T) relaxation rate that is responsible for thermalizing the population of the $AΣu+1$ excited state (⁠$AOu+$ in the Hund’s case (c) representation) and compressing the breadth of emission spectrum.

Extensive measurements of the lamp intensity and emission uniformity have been made over the last two years on scores of lamps with a calibrated Hamamatsu H9535-172 VUV power detector. This sensor has an active area of 28 mm² (6 mm in diameter) and is specifically designed for power measurements in the 172 ± 10 nm wavelength interval. Representative intensity measurements are given in Fig. 6 for a 10 × 10 cm² lamp when the PRF of the bipolar voltage waveform is varied between 18 kHz and 140 kHz. The standard deviation for each point in Fig. 6 is smaller than the diameter of the symbols, and the linearity of the lamp intensity over the entire range in PRF is evident. Intensities above 170 mW cm⁻² are observed reproducibly at 137 kHz, and we hasten to mention that the intensity values shown are those for emission through the front window only. In the absence of a rear reflector, an average of 42% of the total power radiated by the lamp exits through the rear window. That is, the intensity radiated through the rear window is $∼72%$ of that generated from the front window. Tests in which an aluminum reflector was installed behind the rear window of a 100 cm² lamp yielded intensities of ∼250 mW cm⁻², as measured at the front window of the lamp. However, it should be mentioned that the total intensity produced by the lamp (i.e., emerging from both front and rear windows) of Fig. 6 is $>290$ mW cm⁻². With regard to the total emitted power, the radiating area is calculated to be 80.5 cm² (excluding the spacers and the region outside the seal of Fig. 1) and, thus, a conservative value for the maximum power generated by the Fig. 6 lamp is 23.3 W. With further optimization of the lamp design, generating higher power levels is straightforward. For example, the most powerful lamp examined to date generates 228 ± 2 mW cm⁻² from the front window and 164 ± 2 mW cm⁻² through the rear window which corresponds to 31.5 W of 172 nm power (assuming A = 80.5 cm²) for 145 W of electrical power delivered to the lamp. Therefore, the efficiency of the lamp itself is $>20%$⁠. As discussed briefly in the introduction, the $λ∼$ 172 nm intensities reported here are a factor of 5–10 larger than those produced by commercially available lamps, despite little effort to date devoted to optimizing the microcavity and array geometries. Further development focusing on the back reflector and increasing the PRF of the driving electronics is expected to yield intensities above 400 mW cm⁻². It should also be noted that the duty cycle of these Xe₂ lamps at present is ∼3% when the PRF is 135 kHz, a value four orders of magnitude larger than that for an excimer laser operating at 100 Hz.

Tests conducted with lamps that are virtually identical to those described above, except for the absence of the microcavity arrays, show that the microplasmas are responsible for the increase in output intensity reported here. It is reasonable to attribute this significant performance enhancement to an increase in the plasma electron temperature (T_e) brought about by the intensification of the local electric field strength (E) within the cavities. Since T_e drives the production rates for both Xe 6s[3/2]₂ (the precursor to $Xe2*$(A)) and electrons by inelastic electron-impact processes, the cavities have a profound impact on lamp performance.

Intensity maps were also obtained for several distances (d) from the lamp surface, and the Xe₂ emission was found to be spatially uniform to within ±2.5% over the entire surface when d ≥ 10 mm, at which the magnitude of the intensity has fallen by 15% from its value at d = 5 mm. At closer distances, the silica spacers have the effect of modulating the spatially-resolved intensity profiles by ±15%.

Lifetime testing of Xe₂ lamps with an internal phosphor coating and a borosilicate glass structure has been in progress since 2011. Lamp lifetimes $>2×104$ h are standard, and most of the lamps in a lifetime trial begun in the Fall of 2011 remain in continuous operation to this day. Testing of monolithic fused-silica lamps is at an earlier stage but the results to date demonstrate lifetimes (to 70% of initial output) beyond 2000 h with the present manufacturing process. Lamps based on the structure of Fig. 1 but emitting at other wavelengths (222 nm, 225–270 nm, 254 nm, and 308 nm) have also been demonstrated but the results will be presented elsewhere.

In summary, $10×10$ cm² flat lamps driven by two interleaved arrays of microcavity plasmas have produced more than 25 W of average power at 172 nm at a PRF of 137 kHz. When excited by a bipolar voltage pulse format, power is emitted by the lamp in a train of 40–60 μJ, 70 ns FWHM pulses. Having a conversion efficiency >20% and a 3% duty cycle, these lamps are capable of generating unprecedented power levels in the VUV spectral region. Tiling 100 cm² lamps, for example, is expected to yield 1 kW of average power when 40–50 lamps are assembled into a common frame. Other wavelengths in the VUV/UV region have also been generated by downconverting Xe₂ radiation with a phosphor (225-270 nm) or by producing other transient molecules, such as KrCl or XeCl, in the lamp. The thin, flat form factor of microplasma lamps, combined with the power levels now available at 172 nm (in particular), is expected to impact VUV photonics significantly and open the door to commercial applications of VUV photochemistry¹⁵ and surface modification.

The support of the early development of this technology by the U.S. Air Force Office of Scientific Research (H. Schlossberg and J. Luginsland) under Grant Nos. FA9550-14-1-0002 and FA9550-13-1-0006 is gratefully acknowledged.