1. Introduction

While searching for high average power ($\def\uplambda{\unicode[Times]{x03BB}}\geq$ kW) CW laser sources, many different methods have been discovered and pursued. Lasers with kilowatt power levels have been demonstrated using different types of gain media and configurations such as fibers [1,2], semiconductors [3], thin disks [4,5] and atomic and molecular gases [6–12]. For many of these lasers, the output power of a single source is not sufficient for a high power ($>$ kW) system. Instead, many of them are combined into a system using one of many different ways such as coherent, spectral or polarization beam combining, where spectral beam combining is the often the preferred method. Problems that often arise when attempting to integrate one or more of these lasers into high-power systems are thermal management and intensity limits of the beam combiner. In addition, most applications require the laser to be single-mode, ruling out several designs such as multi-mode fiber lasers. For these reasons, and others such as system size and complexity, single-aperture high-power sources are desired. For single-mode, single-source systems, the issues of thermal management and intensity limits of the gain medium must also be addressed. Gas lasers, and especially atomic gas lasers, have seen much interest for these applications due to the inability to damage the gain media and the fact that they can be circulated through a system enabling efficient cooling of the gain media at very high average powers. For example, gasdynamic lasers [6,13,14] have demonstrated CW average powers up to 100’s of kWs to MWs [15], but several limiting factors, such as safety, limit the feasibility of using such a system in real-world applications. Atomic gases have seen increased interest as potential single-source systems due to the ability to achieve high gain without resorting to a gasdynamic arrangement. Much effort has gone into diode-pumped alkali laser (DPAL) [11,16–19], and more recently, optically pumped rare gas laser (OPRGL) [20–30]. Due to the reactivity of alkalis with chamber seals and windows, dimer and trimer formation, and the current necessity of using simple hydrocarbons (to induce rapid spin-orbit relaxation) that also react with the alkalis, DPALs have had many challenges to overcome. As an alternative gain media to the alkalis used in DPALs, metastable state noble gases can be used with analogous lasing gain results, but with none of the reactivity issues of the gain media. Han and Heaven [20] first proposed that metastable rare gas atoms (Rg$^*$) will have similar level structures, lifetimes and coupling strengths as alkalis. In that work, near-IR lasing was achieved in Kr$^*$, Ar$^*$, and Xe$^*$ with pulsed dye laser pumping. For CW operation, however, a diode-pumped laser is the only practical source for creating an efficient system since high-power diodes are currently the most efficient high-power sources. Since atomic lines are very narrow, an ideal pump source will also need to have maximal bandwidth overlap within the broadened atomic absorption region. A CW diode-pumped rare gas laser (DPRGL) was demonstrated in Ar$^*$ by using a volume Bragg grating (VBG) narrowed diode laser pump [22]. Xenon, however, has some advantages over other noble gas species. For instance, due to the large number of naturally occurring Xe isotopes (9 total with 7 comprising >99%), the absorption linewidth is quite broad (many GHz) even at modest pressures (100 mTorr). At 750 Torr, the absorption bandwidth increases substantially ($>$20 GHz), allowing the use of broad-band pump sources.

In the past, xenon received significant interest as a mid-IR laser source [31–52]. In most of those schemes xenon was excited either by nuclear reactor radiation, electron gun or electric discharge. With such pumping schemes, lasing efficiencies of 2–3% were achieved at wavelengths in the 2–3 $\mu$m range. However, for most applications, nuclear pumping is not feasible. In addition, other methods were invented to generate very high-average-power CW lasers in the IR to mid-IR, such as the gasdynamic laser, which also achieved modest efficiencies. Interest in xenon as a high power CW source did not garner much attention until tunable, high-power pump diodes would become commercially available.

In this work, we demonstrate the first quasi-CW diode-pumped metastable Xe laser. We first discuss the energy level structure for the various Xe isotopes, and then describe the setup for our proof-of-principle experiment. Finally, we discuss the results of the experiment and discuss improvements for better efficiency and power scaling.

1.1 Theory

To optically pump Xe, it must first be excited into a metastable state (labeled as Xe$^*$) [53]. Figure 1 shows an energy level diagram for xenon with the ground state ($^1$S) and the excited metastable state (6s[3/2]$_2$). The 6s[3/2]$_2$ state is metastable because it is forbidden to decay to the ground state and relies on collisions or other excitation mechanisms to leave this state. Now the primary challenge associated with the DPRGLs is creating large ($\geq 10^{12}$ cm$^{-3}$) metastable population densities. While the ionization potential of group 18 elements is very high, it decreases as atomic number (and atomic radius) increases, making Xe the easiest stable noble gas to excite. A pulsed dc discharge is an efficient means to excite Xe and was used in our setup. The pulse excites the gas due to a rapid change in the electric field, leaving the gas either excited or ionized. By means of collisions and radiative decay, a portion of the excited xenon atoms will quickly decay into the 6s[3/2]$_1$ and 6s[3/2]$_2$ states. Through this method, a sufficient metastable population is created for lasing.

 

Fig. 1. Level diagram showing the transitions for the 979.9 nm lasing and 904.5 nm pump transitions for a three-level lasing scheme in Xe$^*$. $\textrm {k}_{\textrm {ij}}$ is the relaxation rate that results from collisions with the buffer gas.

Download Full Size | PDF

Once the Xe$^*$ population has been created, it can be optically pumped at 904.5 nm into the 6p[5/2]$_2$ state. As mentioned earlier, the large number of naturally occurring, stable isotopes of Xe results in an unusually wide Doppler and pressure broadened absorption profile compared to most atomic species. Figure 2 shows the optical absorption spectrum of a 10 mTorr, 1:1 Xe:He mixture at the 904.5 nm transition.

 

Fig. 2. Energy level structure of the 904.5 nm transition. The hyperfine structure of the odd isotopes in shown in the top of the figure, and the bottom shows the measured absorption of low pressure Xe. The roman numerals and Greek letters mark the energy levels in the upper figure that are responsible for the absorption dips in the absorption profile. The green line is the signal from the Fabry-Pérot and has a FSR of 1.5 GHz. The even isotopes all contribute to the strong absorption dip in the center (set as the 0 reference) that is broadened by isotope shifts.

Download Full Size | PDF

Most of the isotopes of Xe are even and contribute only to the central absorption profile in Fig. 2. They are shifted relative to each other only due to isotope shifts ($\mathcal {O}(100)$ MHz) [54]. The two odd isotopes of Xe that are stable, namely $^{129}$Xe and $^{131}$Xe, have nuclear spins of 1/2 and 3/2 respectively, and contribute to most of the broadening of the absorption spectrum. The calculated hyperfine splitting of these two odd isotopes is shown in the upper portion of Fig. 2, and their main contributions to the absorption profile are labeled and shown in the absorption spectrum. A Fabry-Pérot interferometer was used to calibrate the absorption spectrum and is shown as the solid green line in Fig. 2. For the Xe$^*$ laser, we operate at $\geq$1 atm, where this transition broadens to $>$ 20 GHz FWHM.

DPRGLs are often operated at atmospheric pressures, with most of the volume consisting of a buffer gas. The buffer gas is required to create a large collisional relaxation rate (k$_{ij}$ in Fig 1). Due to this collisional relaxation, population will be transferred from 6p[5/2]$_2$ state to create inversion in the 6p$[1/2]_1$ state. We measured the fluorescence lifetime of the 6p$[1/2]_1$ transition to be $\Gamma \simeq 10^{7}$ s$^{-1}$. This three-level system will lase at 979.9 nm from the 6p[1/2]$_1$ state to the 6s$[3/2]_2$ state (Fig. 1).

2. Experimental setup

The experimental setup is shown in Fig. 3(a). The regions in the dashed lines were used only for tunable diode laser absorption spectroscopy (TDLAS) and the gain measurements. The vacuum chamber (cube with 13.6 cm per inner dimension) was constructed with three windows; two windows along the optical path and one window on top of the chamber. This top window was used to collect light from the plasma for temporal and spectroscopic measurements. The chamber widows along the optical path were anti-reflection coated for the IR and were wedged with a 30 arcmin wedge to eliminate feedback into the gain region. A Xe:He mixture flowed into the chamber with 24 sccm and 1490 sccm flow rates of Xe and He respectively. The pressure was varied for pressure dependent measurements but maintained at about 760–780 Torr for laser operation. For thermal reasons, the gas is always flowing and is evacuated from the system with a vacuum pump after the pressure control valve.

 

Fig. 3. (a) Experimental setup for the TDLAS, gain measurements and the Xe$^*$ laser. The area enclosed by dashed lines was only used for the TDLAS and gain measurements. The cavity mirrors were removed for TDLAS and gain measurements, however the pump was used for the gain measurement and lasing only. (b) Schematic of the electrodes and their cross section. BB – Beam Block, DM – Dichroic Mirror, F – Filters, FM – Flip Mirror, FP – Fabry-Pérot interferometer, HR – High Reflector, MFC – Mass Flow Controller, OC – Output Coupler, PCV – Pressure Control Valve, PD – Photodiode, PM – Power Meter, SL – Spherical Lens, TDL – Tunable Diode Laser

Download Full Size | PDF

The electrodes were made of titanium and had a cross-sectional area A$_\textrm {e}$ = $y_{\textrm {e}} \times z_{\textrm {e}}=6.5\times 11.5$ mm$^2$ (see Fig. 3(b)), and the spacing between the electrodes was S$_\textrm {x}\simeq$ 2.5 mm, giving a gain region volume of A$_\textrm {e}\times \textrm {S}_\textrm {x}\simeq 187$ mm$^3$. The electrodes were placed on several sheets of mica to insulate them from the bottom support platform inside the chamber. A high voltage generator (Gamma High Voltage Research; RR1.5-500P/115) and pulsing unit (DEI; PVX-4150) were used to deliver 60 ns pulses at 1.2 kV peak amplitude. The discharge was run with a pulse repetition rate of 240 kHz, which resulted in an 80 mA average current draw through the discharge. Therefore, the power required to sustain the discharge was about 96 W.

For the 904.5 nm absorption measurement, the TDL2 was used (Fig. 3). This probe laser was a tunable diode laser (Toptica Photonics; DL Pro, $\uplambda _0=905$ nm) that was tunable from 880 to 920 nm with a linewidth of $\approx$100 kHz. The output power sent into the plasma was attenuated to a few milliwatts or less to prevent power broadening effects and was frequency scanned over the absorption region. Part of the output of the diode laser was split off and sent to a wavemeter (High Finesse; WS-6) to verify the absolute frequency before the scan, and then to a 1.5 GHz free spectral range (FSR) Fabry-Pérot interferometer (Thorlabs; SA200-8B) to provide a frequency scale reference during the scan. The probe beam was reduced in size with a telescope and collimated before passing through the chamber. Balanced detection was used to subtract out the variations in laser power over the scan range. The signals from the photodiodes were collected by an analog-to-digital converter (National Instruments; cDAQ-9188 chassis, NI-9222 DAQ card) and processed by a custom written computer code.

The TDL1 (see Fig. 3(a)) was used for the gain measurements and was a tunable diode laser (Toptica Photonics; DL 100, $\uplambda _0=980$ nm) that was tunable from 920 to 990 nm with a linewidth of $\approx$100 kHz. The Xe$^*$ population was optically pumped by focusing the pump laser ($\uplambda =904.5$ nm) into the plasma to populate the 6p[1/2]$_1$ state. The pump laser was a spectrally narrowed diode laser (OptiGrate Shark; $\uplambda _0=905$ nm) that was tunable around 904.5 nm with an output of 90 W CW with 17.8 GHz linewidth. The pump laser is frequency controlled by adjusting the temperature (through software) of the volume Bragg grating (VBG). The VBG temperature was adjusted to tune to the maximum absorption of the pump laser. The probe beam was combined (but not focused) with the pump beam using a dichroic filter (Thorlabs; DMLP950L) and passed through the gain region with 35 mW of power while tuned to 979.9 nm. The signal was recorded with a photodiode (Thorlabs; DET10A) and an oscilloscope (Tektronix; DPO 70404) with and without the pump laser.

The Xe$^*$ laser was demonstrated by placing the cavity mirrors outside the gas chamber along the pump diode axis. The high reflector (Advanced thin films) was a flat mirror that is transparent at 904.5 nm and has $>99.9$% reflectivity at 979.9 nm. The output coupler (Advanced thin films) was a concave mirror (1.5 m radius of curvature, therefore 0.75 m focal length) with 90% reflectivity at 979.9 nm and $> 99.9$% reflectivity at 904.5 nm. A focusing lens before the back cavity mirror was used to focus the pump beam into the gain region. This focusing lens was mounted on a translation stage so that the focus could be moved along the optical axis (z direction) through the gain region and optimized to achieve maximum pumping efficiency. The residual pump light is diverging after the output coupler and was filtered by an iris and two 925 nm long-pass filters (Semrock; FF925-Di01-25x36) which provided $>$5 OD of attenuation at 905 nm while allowing $\approx$88% transmission of the probe beam. The output power and frequency spectrum were measured with a power meter and an optical spectrum analyzer (Yokogowa; AQ6370C).

3. Results

Figure 4(a) shows an image (the scale bar is 5 mm) of the XeHe plasma region where the Xe$^*$ population will be located. We used a TDLAS setup to measure the metastable population in the plasma and obtain the absorption linewidth. Figure 4(b) shows the FWHM of the 904.5 nm transition versus total gas pressure. The strait line fit in Fig. 4(b) gives a broadening coefficient of $\approx$30 MHz$\cdot$Torr$^{-1}$. The measured FWHM at 750 Torr was $>$20 GHz. To find the metastable density, we use an integration method employed in TDLAS [55] to sum up the area under the absorption curve relative to the reference. The equation is derived by setting the ratio of upper and lower state populations from the Boltzmann distribution of states equal to the same ratio calculated by detailed balance of a two-level system. The resulting equation is

 

Fig. 4. Experimental results from the TDLAS measurements and the streak camera. (a) A picture of the plasma at 760 Torr (scale bar is 5 mm). (b) The full width at half maximum of the 904.5 nm absorption profile versus total pressure in the cell. (c) The calculated metastable Xe densities in the plasma versus total pressure in the cell. (d) Data obtained from a streak image showing the rise in fluorescence versus time from the 979.9 nm transition using impulsive excitation at 904.5 nm.

Download Full Size | PDF

Population densities $\geq 10^{13}$ cm$^{-3}$ were measured at various pressures show in Fig. 4(c). To obtain an estimate of the collisional transfer rate from the 6p[5/2]$_2$ state to the 6p[1/2]$_1$ state, we first pumped the Xe$^*$ atoms with a 30 ps laser (Ekspla; PL2231-50-PRETRIG-PLD208-PG401) pulse at 904.5 nm. The fluorescence was then filtered with a monochromator set to 979.9 nm and the signal was recorded with a streak camera (Hamamatsu Photonics; C1587). The timing window was set to include the background a little before the rise in fluorescence. The resulting data is plotted in Fig. 4(d). From this data we get an estimate of about 0.9 GHz relaxation rate.

For single-pass gain measurements, we assume a small signal gain that is exponential

 

Fig. 5. Experimental data showing the gain and laser output power. (a) Measurement of the laser gain coefficient (black line) through the plasma discharge cycle. The green line is the probe with the pump off. (b) The 979.9 nm Xe$^*$ laser output power versus input pump power. (c) The observed Xe$^*$ laser linewidth measured relative to the transition frequency ($\nu _0$).

Download Full Size | PDF

Figure 5(b) shows the output power as a function of input optical power. With our setup we achieved a slope efficiency of $\approx$2.3% averaged over the gain cycle (Fig. 5(a)). The output spectrum of the Xe$^*$ laser is shown in Fig. 5(c). The plot shows the detuning in GHz from the central wavelength ($\nu _0=979.9$ nm). From this the FWHM of the laser linewidth was calculated to be 13.4 GHz with a Voigt fit. The pump laser linewidth was measured to be 17.8 GHz FWHM.

4. Discussion

A theoretical treatment of the XeHe laser by Demyanov et al. [56] predicts a maximum total efficiency of about 35% with a number density $2.3 \times 10^{13}$ cm$^{-3}$ (assuming 1 cm gain length), 0.65 atm pressure with a xenon fraction of 0.004. Our pressure was around 1 atm and our xenon fraction was about 0.02. We further investigated the use of ternary mixtures with Ar or Ne as the third gas in the mixture to investigate whether an improvement could be made to the metastable density and lasing efficiency. We did not observe the addition of Ar to help increase the metastable population densities, rather, we observed a deleterious effect. The discrepancies in performance could be our quoted efficiency and physical implementation and is discussed below. It should be noted the the binary XeNe mixture was tested as well and lasing was achieved, however lower efficiency ($<$1%) was observed compared to the XeHe mixture.

It is clear that if a diode-pumped Xe$^*$ laser is to be used for high power ($\geq$kW) applications, greater efficiency will have to be achieved. Our calculated efficiency (2.3%) is based off of input pump power, not absorbed pump power. Due to the circumstances of the experiment, we could not reliably determine the absorbed pump power. However, if we based our efficiency off of absorbed pump power, our slope efficiency would probably increase. From Fig. 5(a), we also see that this efficiency is based off the average power. The slope efficiency at peak gain is $>$4%.

There are several areas of improvement for our setup. The first is to improve the production of metastable Xe. The metastable density is very sensitive to the state of the plasma and its generation. With a larger power supply and using a dielectric barrier discharge, we would be able to generate a larger and more uniformly dense plasma. In the relaxation process, some of the electrons settle into the 6s[3/2]$_1$ state. The population in that state can be optically-pumped and increase the population into the 6s[3/2]$_2$ metastable state by an order of magnitude [57].

The cavity design for our setup was hemispherical and was chosen due to the mirrors available at the time of the experiment. The gain region was placed close to the flat cavity mirror where the mode size is the smallest. This yielded the closest mode matching between our pump laser and the cavity mode, better extraction of the gain region, and allowed the lasing mode to fit between the electrodes without clipping. A better cavity for this laser would be a confocal design which would be a more stable cavity, and would also allow a more uniform depletion of the gain region.

The pump was a multimode source, and will not focus as tightly as a TEM$_{00}$ beam and therefore cannot be well collimated into a small beam. This can be problematic because the outer region of the pump beam can hit the electrodes and cause heating. This heating of the electrodes will affect the electron emission from the electrodes and will cause instability in the plasma generation by driving it into an arc filament. The metastable population will then be high in a very small region around the filament, but will be orders of magnitude lower everywhere else. Therefore a large uniform distribution of plasma is ideal for an optimal Xe$^*$ gain region. Moving the electrodes further apart to avoid being illuminated by the pump beam was not an option available to us since moving the electrodes further apart will require a larger power supply to sustain the discharge with the same density. To mitigate the laser induced electrode heating, the beam must be as tightly focused as possible. However, since the pump laser is on resonance there is absorption heating of the gain region. This will cause a reduction in the metastable density by thermal blooming [58–61] (this effect is very visible in the plasma) at the focus of the pump laser in the plasma, therefore degrading the Xe$^*$ population. There is then a trade-off between minimal heating of the electrodes and minimal heating of the gain region. We achieved this by moving the position of the pump laser focus (along the propagation axis) in order to give an optimal metastable population. An improvement would be either the use of a single mode pump beam that could be well collimated, or to use a transverse excitation scheme and pump the whole gain region from the side. This would allow for a more uniform pumping of the gain region and lead to near optimal energy extraction of the pump laser. Lastly, through investigation of the fluorescence intensity and lifetimes of many levels in Xe, there are different energy level schemes that we will be investigating in the future that will potentially provide a more efficient lasing scheme. One such energy scheme would be to pump at 881.9 nm (6p[5/2]$_{3}$ in Fig. 1) and then lase at the 904.5 nm transition. The energy difference between these energy levels is only $\approx$ 173 cm$^{-1}$ and is strongly coupled, whereas our current level scheme has an energy difference of $\approx$ 856 cm$^{-1}$. Given the strong collisional coupling and lower quantum defect, this would be a good alternative lasing level scheme.

5. Conclusion

In conclusion, we have demonstrated a diode-pumped, quasi-CW metastable Xe gas laser at atmospheric pressure. We have discussed the level structure and the broad absorption bandwidth of Xe due to the abundance of natural isotopes. A detailed outline of our experiment was presented, and the results for metastable population, collisional transfer rates, gain, laser output power and linewidth were presented. With our experimental configuration, we obtained an average slope efficiency of about 2.3%. The results were then discussed and short-comings to the experiment were addressed and improvements were proposed.