1. Introduction

Tunable optical fiber lasers that emit in the technologically important 3 to 5 micron region remain elusive, despite many recent advances in optical fiber technology [1]. This mid-infrared (IR) region is of much interest due to the presence of vibrational absorption lines of several chemically important functional groups, such as carbonyl and alkyl, and the presence of an atmospheric transmission window, which allows for uses such as remote sensing and lidar [2–5]. Silica optical fiber lasers have become extremely prevalent due to their minimal alignment requirements, mechanical robustness, and extremely low loss. Fundamental advantages such as distributed heat dissipation due to 360$^{\circ}$ radial symmetry and a large surface-area-to-volume ratio which enables small cross-sections (tens of microns) to be coupled with long lengths of active fiber material (centimeters to meters) have led to average power outputs in the thousands of watts in both the pulsed and continuous wave (CW) regime [6–10]. These properties make a fiber laser an ideal and robust alternative to existing bulk laser systems. While optical fiber lasers based on doped silica glasses have shown great success in emitting high-power continuous wave and ultrafast pulsed light in the near-IR region, silica’s high absorption in the mid-IR region and amorphous structure make it unsuitable to host IR emitting transition metal dopants whose d-orbitals are more sensitive to their coordination environment than the f-orbitals of rare-earth elements that are often doped into glasses [1]. To move beyond the near infrared, new mid-IR fiber optic materials are needed.

To date, several silicate and ZBLAN doped glass fiber mid-IR lasers have been fabricated, but direct CW mid-IR emission is limited to the 3.95 µm light provided by Ho$^{3+}$:ZBLAN fiber [1,11]. Several other strategies have currently been employed to expand the laser emission range of the optical fiber into the mid-IR, including rare-earth element doped fluoride optical fiber lasers [12], gas-filled photonic crystal fiber lasers [13], and supercontinuum generation [14]. Recently, broadband 4 µm emission was reported from Dy:InF₃ optical fiber, but no lasing in this region was reported [15]. Multiphoton quenching and nonradiative decay pathways can severely limit high power operation in glassy dopant hosts. Additionally, the indirect methods of laser light generation mentioned above typically aren’t as efficient and lack the spatial and temporal coherence of direct laser generation. Thus, to achieve powers of hundreds of watts or greater, crystalline materials are preferable due to their thermal conductivity, mid-IR transmission, advantageous crystal field splitting, and vibronic coupling, which allows them to be tuned over large wavelength ranges.

Transition metal doped crystalline II-VI semiconductors, such as iron doped zinc selenide (Fe²⁺:ZnSe) are well known mid-IR gain media used in bulk lasers [16,17]. Fe²⁺:ZnSe in particular has a useful emission range that spans from 3.7 to 5 microns [16,18,19]. However, bulk Fe²⁺:ZnSe and related Cr²⁺:ZnSe lasers are fundamentally limited in the maximum power they can produce because of thermal effects caused by the large thermo-optic coefficient of ZnSe (dn/dT = 7.0$\times$10$^{-6}\;$K)[20,21]. Bulk, freespace Fe²⁺:ZnSe lasers have achieved CW powers of 9.2 W and pulsed powers up to 35 W [22,23]. Current strategies to overcome this limitation include ultrafast laser inscribed and planar waveguides, which can help mitigate thermal lensing by confining the pump light to a waveguide situated near a cooling source [24–27]. However, these waveguide structures lack 360$^{\circ}$ radial symmetry to dissipate heat, leading to asymmetric thermal gradients, which cause instability at higher powers. Other competing approaches, such as a recently demonstrated 140 W Cr²⁺:ZnSe rotating disk laser, are impressive for their power output but less practical for high stability or portable environments where mid-IR lasers and sensors are likely to operate [28].

Another strategy to overcome this fundamental issue could be through the fabrication of an optical fiber with a Fe²⁺:ZnSe core which would benefit from the aforementioned optical fiber advantages. However, the high vapor pressure of ZnSe makes it incompatible with traditional fiber drawing techniques. High pressure chemical vapor deposition (HPCVD) has emerged as a promising technique to deposit semiconductor materials, including ZnSe, into the optical fiber waveguide geometry [29–32]. HPCVD employs high pressure (tens of MPa) gas phase reactions to deposit layers of semiconductor material in microstructured templates. Most recently, the first Cr²⁺:ZnSe optical fiber lasers, deposited via HPCVD, that operate in the gain switched and CW regime have been reported [33,34].

Here we report the fabrication of the first continuous wave Fe²⁺:ZnSe optical fiber laser deposited using HPCVD. Micro-X-ray fluorescence measurements indicate that these deposited structures have iron dopant along centimeters of length and radially throughout the fiber core. Additionally, optical pumping experiments performed on these Fe²⁺:ZnSe optical fibers indicate laser threshold behavior and CW operation. These fibers are a first significant step towards a tunable crystalline semiconductor optical fiber capable of generating direct laser emission in the 3 to 5 micron region.

2. Fabrication

Fabrication of the Fe²⁺:ZnSe optical fiber was performed using a modified method of the HPCVD synthesis reported for ZnSe optical fibers [31,35]. In this synthesis we utilize organometallic precursors, dimethyl zinc and dimethyl selenide (ratio roughly 1:2), pressurized in UHP hydrogen as carrier/reactant gas to 70 MPa. This gas mixture flows through a silica microcapillary template with an inner diameter that can range from microns to tens of microns. Heating the template to 450$^{\circ}$C results in atomically smooth, conformal deposition of polycrystalline cubic ZnSe along the capillary wall that nearly fills the template. Once a pore on the order of 100s of nanometers remains, the reaction arrests due to the limited mass transport of reaction byproducts out of the reactor. The choice of methylated organometallic precursors ensures that reaction only occurs in the heated region of the template [36].

To deposit Fe²⁺:ZnSe, we utilize n-butylferrocene as our iron precursor in the reaction scheme. n-bultylferrocene has been used in the metalorganic CVD of iron(III) oxide films and has a vapor pressure and decomposition temperature that is compatible with the HPCVD process [37]. Using commercially available 360 µm outer diameter (OD) silica capillaries with various inner diameters (ID), we constructed a microfluidic reaction vessel for the HPCVD reaction. Figure 1 shows a drawing of the silica microfluidic reactor. Optical micrographs of the reactor are provided in the supporting information, see Supplement 1. Liquid n-butlyferrocene was loaded into a 1 cm long, 6 µm ID, 125 µm OD capillary, via capillary action, which was then placed inside a larger 150 µm ID silica capillary. The larger capillary was fused to a 50 µm ID diameter capillary in which the deposition occurred. The reactor was configured to allow dimethyl zinc and dimethyl selenide with hydrogen carrier gas at 70 MPa to flow through it. A two zone furnace was used to independently control the vapor pressure of the n-butylferrocene and the growth rate of the Fe²⁺:ZnSe in the 50 µm ID capillary. Heating the n-butylferrocene to 180 $^{\circ}$C and deposition at 450$^{\circ}$C experimentally gave the highest concentration of and best dopant uniformity of Fe²⁺ and the deposition typically completed within 2 days, producing a usable fiber section that is approximately 1 cm long. Currently, this limit is largely due to the differences in the expansion coefficient between ZnSe and silica causing periodic cracking in the core. Nonetheless, the resulting deposition leaves an almost fully filled capillary, leaving only a very small central pore (Fig. 2). For comparison, it takes 5 to 14 days to dope a 3 cm thick ZnSe wafer to the appropriate iron concentration needed for lasing via diffusion doping [38]. Our method is comparatively faster and more convenient by producing centimeter long optical waveguides in approximately 2 days.

 

Fig. 1. The HPCVD reactor is fabricated by fusion splicing silica capillaries together to form a flowing reactor. The larger 150 µm diameter capillary is used to hold a small 6 µm capillary containing n-butylferrocene, which is held at 180$^{\circ}$C. The reactant flow carries the reactants into the 50 µm diameter capillary where the reaction occurs to deposit Fe²⁺:ZnSe and 450$^{\circ}$C.

Download Full Size | PDF

 

Fig. 2. Optical micrograph of Fe²⁺:ZnSe optical fiber cross-section (a) and fiber core in silica template/cladding (b). A small central void, $\sim {}$700 nm, remains in the center of the ZnSe core due to the buildup of reaction products in the reactor. Due to cylindrical lensing, the pore appears much larger than it physically is.

Download Full Size | PDF

3. Results and discussion

3.1 Fiber Fe²⁺ distribution

To gain a better understanding of the HPCVD deposition and doping process, we carried out synchrotron micro-X-ray fluorescence (XRF) mapping experiments to quantitatively map the distribution of transition metal dopants in HPCVD produced Fe²⁺:ZnSe optical fibers. Micro-XRF is a useful technique because of its high sensitivity and sub-micron spatial resolution coupled with the ability to map large regions up to centimeters in dimension. These chemical maps give a direct observation of how the distribution and concentration of Fe changes during the deposition process. Homogeneous doping is crucial towards supporting high power lasing. Low concentrations of Fe²⁺ will limit laser performance through inefficient absorption of the pump light while excessive Fe²⁺ will lead to fluorescence quenching [39]. In bulk Fe²⁺:ZnSe crystals, diffusion doping of Fe is the most widely used technique to produce the required concentration needed for lasing [40]. However, as previously discussed, diffusion doping only produces uniform concentration gradients over millimeters of length in a bulk crystal and is comparatively slower than direct deposition of the material.

Figure 3(a) shows the micro-XRF map of a focused ion beam milled fiber cross-section and that the iron dopant is distributed throughout the fiber radially. This indicates that the reaction decomposition kinetics of the n-butylferrocene is well matched with the decomposition kinetics of dimethyl selenide and dimethyl zinc since there is no global concentration gradient present throughout the structure. The average radial concentration is 0.116$\pm$0.024 µg/cm$^{2}$. The large standard deviation suggests that there may be locally high concentrations of Fe, but higher resolution chemical maps would be needed to investigate this further. We also mapped the Fe concentration along the fiber length by polishing a Fe²⁺:ZnSe in half along the fiber length and analyzing it with the x-ray microprobe (Fig. 3(b)). Along the fiber length iron concentrations were again found to be principally homogeneous although some clustering on orders of hundreds of nm may be present. The hot spots in the image and in the profile appear to be particles on the sample surface leftover from polishing the cross-section. This is evidenced by some of the hot spots appearing on the glass cladding of the fiber core. We excluded the most obvious points from the plotted line profile. The average concentration along the fiber length was 0.57$\pm$0.06 µg/cm$^{2}$. It’s important to note that the XRF concentrations are not an absolute measure of concentration. Matrix effects such as self-absorption of the emitted X-rays will significantly affect the reported concentration. However, the concentrations are useful for relative comparisons, such as concentration gradients within the structure.

 

Fig. 3. micro-XRF map of Fe²⁺:ZnSe fiber cross-section (a) and map of Fe distribution along 0.80 cm length (b). Globally, along the fiber cross-section and fiber length, concentration remains uniform.

Download Full Size | PDF

To measure the absolute Fe²⁺ concentration in the fiber, a FT-IR microscope was used to collect a room temperature absorption spectrum from the fiber.The FT-IR spectrum is shown in Fig. 4 and matches well with the expected tetrahedral crystal field splitting of Fe²⁺ doped into cubic ZnSe. We found that the average concentration in the 0.8 cm fiber was 2.2$\times$10$^{18}\;$ ions/cm$^{3}$. The absorption cross-section reported by Adams, 6.5$\times$10$^{-19}\;$cm$^{2}$ at 2.698 µm, was used to calculate the concentration [18].

 

Fig. 4. Room temperature FT-IR absorbance spectrum of Fe²⁺:ZnSe optical fiber. The absorbance is normalized to the length of the fiber.

Download Full Size | PDF

3.2 Lasing characterization

To analyze the lasing characteristics of the HPCVD grown Fe²⁺:ZnSe optical fibers, we utilized an optical setup where the Fe²⁺:ZnSe was pumped with a 2940 nm CW Er:YAG laser. The full details are provided in the Experimental Section. The Fe²⁺:ZnSe optical fiber was mounted inside a cryostat on a machined copper cold finger, which allowed lasing experiments to be performed at 77 K with liquid nitrogen cooling. Fe²⁺:ZnSe has too short of a radiative lifetime at room temperature to support lasing, necessitating cryogenic cooling of the crystal in order to achieve CW operation [41]. For these experiments, no external cavity optics were used since they were found to be ineffective, however 17$\%$ Fresnel reflection at each fiber facet provided some optical feedback. We observed a lasing threshold at 16 mW of pump power by an increase in output intensity and the appearance of sharp peaks in the emission spectrum (Fig. 5). Below the threshold at 12 mW of pump power, the typical fluorescence spectrum of Fe²⁺:ZnSe is observed. At the threshold of 16 mW of pump power, the fluorescence spectrum narrows and a peak is observed around 4100 nm, which is indicative of lasing behavior. At 20 mW of pump power, more laser peaks appear out of the fluorescence background. The laser threshold plot is shown in Fig. 6. This lasing threshold is the lowest reported for any Fe²⁺:ZnSe laser to our knowledge. Currently, we attribute this low threshold to the waveguide confinement effect. Additionally, our polycrystalline fibers have low loss ($\sim {}$1.5 dB/cm @ 2600 nm), where the loss mechanism is largely due to Rayleigh scattering at grain boundaries [35]. The low optical loss of our optical fibers combined with the optical confinement provided by the waveguide and the long radiative lifetime of cryogenically cooled Fe²⁺ should significantly lower the threshold as has been observed in Cr²⁺:ZnSe laser inscribed waveguides [27].

 

Fig. 5. Emission spectrum of Fe²⁺:ZnSe optical fiber near the lasing threshold taken at 77 K.

Download Full Size | PDF

 

Fig. 6. Laser threshold plot showing low threshold behavior of 16 mW for Fe²⁺:ZnSe optical fiber laser.

Download Full Size | PDF

Figure 7 shows the lasing spectrum of a Fe²⁺:ZnSe optical fiber laser at 600 mW of pump power. At powers above 600 mW the pump laser beam quality became unstable, and so the pump light was not able to be coupled well into the optical fiber. The lasing modes are centered around 4120 nm with a spectral width of 100 nm, which is in agreement with previous studies of inhomogeneously broadened, free-running Fe²⁺:ZnSe lasers [41]. Many modes are present due to the extreme multimodal nature of the waveguide and because this sample was not in a selective optical cavity. The effect of chopping the pump with different duty cycles was also explored and we found that the laser operation was largely unaffected by varying the duty cycle, besides a slight red shift in the lasing spectrum which has been previously reported in the literature [41].

 

Fig. 7. Lasing spectrum of free running Fe²⁺:ZnSe optical fiber at full pump power (600 mW) with cyrogenic cooling. The laser emission is centered at 4.12 µm.

Download Full Size | PDF

The slope efficiency of the laser was characterized by measuring the output of the fiber laser at one end at increasing pump powers. From the efficiency plot (Fig. 8), the slope efficiency is 0.1$\%$. This low efficiency can be attributed to several factors. First, due to the large refractive index contrast between silica (n = 1.38) and ZnSe (n = 2.42) and the large core diameter, our optical fiber is capable of supporting 1000s of modes. Additionally, the central pore will affect the mode structure such that a Gaussian shaped mode structure cannot be supported. The loss in these higher order modes will be greater than if the fiber was single mode and will allow for some modes to leak into the silica cladding which the Fe²⁺:ZnSe is deposited into. At 4.1 µm, silica will be 90$\%$ to 95$\%$ absorptive depending on the formulation, meaning the effective gain of the fiber is expected to be decreased due to evanescent field interaction with the cladding. To combat this a cladding layer of intrinsic ZnSe or other chalcogenide such as ZnS could be deposited to form a fiber with an undoped cladding and doped core to create a refractive index contrast that can support a single mode.

 

Fig. 8. Efficiency plot for Fe²⁺:ZnSe fiber laser. The efficiency is 0.1$\%$ with a total power output of 0.40 mW.

Download Full Size | PDF

Another source of loss in our fiber could be through localized heating of the fiber core. At elevated temperatures above 180 K, the lifetime of the laser transition will collapse and nonradiative pathways will dominate [18]. Experiments at higher duty cycles (25% and 50%) showed a red shifting of the emission spectrum, which may indicate local heating at the input face of the fiber [41]. We are currently investigating alternative thermally conductive cladding materials such as diamond [42,43]. Despite the low efficiency, this result demonstrates the first Fe²⁺:ZnSe fiber laser and also provides clear direction on how to improve the laser design such that power scaling should be achievable.

4. Conclusion and outlook

In this work, we demonstrate the first Fe²⁺:ZnSe optical fiber laser and to our knowledge, the first solid state optical fiber laser that operates in the CW regime via direct laser generation extending past 4 µm. To achieve higher power Fe²⁺:ZnSe optical fibers, new cladding materials such as ZnS or ZnS_xSe_1-x are crucial to shield the gain in the doped core from interacting with the absorptive silica cladding. We have already demonstrated the ability to deposit ZnS and ZnS_xSe_1-x using HPCVD and extending this to deposit core-cladding structures is currently under investigation [31]. Additionally, the silica cladding can be removed and a more thermally conductive outer cladding such as diamond can be deposited or adhered, which will make it much easier to cool these fiber structures to avoid thermal lensing and luminescence lifetime collapse at high power operation. Sapphire capillaries could also be used as deposition templates, but currently there are no commercial options available that are compatible with HPCVD. Removal of the central void will also be necessary to achieve low mode order fibers. Techniques such as laser and thermal annealing are under investigation, and may be able to close the void and improve the material quality of the fibers [44,45]. We have also developed novel glass compositions to allow us to anneal and deposit our fibers in expansion matched templates [46]. Laser inscribed Fe²⁺:ZnSe waveguide lasers have exhibited slope efficiencies as high as 58$\%$ under liquid nitrogen cooling suggesting these HPCVD grown optical fibers should be capable of high efficiency operation once the cladding and thermal issues are resolved [27]. Furthermore, appropriate optical coatings on the fiber facets will make the optical cavity more efficient by enhancing feedback and lead the way towards direct high power optical fiber laser emission in the mid-IR.

5. Experimental methods

5.1 micro-X-Ray fluorescence

Micro-X-ray fluorescence measurements were performed at beamline 2-ID-D at the Advanced Photon Source at Argonne National Lab. Fiber cross-section samples were prepared by first etching the silica cladding of a Fe²⁺:ZnSe fiber with 10$\%$ hydrofluoric acid. The etched fiber core was then milled using a focus ion beam (FIB) instrument (FEI) to form a 1-micron thick disk of the cross-section (See Supplement 1). To study the longitudinal Fe concentration, fiber samples were polished in half biaxially to reveal the fiber core. These samples was attached to a Mo FIB grid which was mounted in the X-ray microprobe instrument at the beamline. An X-ray excitation energy of 9.6 keV was used to excite the Fe ions in the sample. An X-ray zone plate provided a spot size of 200$\times$200 nm and a 3-axis stage allowed the sample to be moved so that concentration could be spatially mapped with a resolution of 1000$\times$1000 nm. The third dimension is used to move the sample into the focus of the X-ray beam. The data were quantized using a NIST XRF standard. The data was analyzed using the MAPS software package [47].

5.2 Spectroscopy and laser characterization

A room temperature IR absorption spectrum was used to calculate the concentration of Fe$^{2+}$ in the fiber. The spectrum was collected using a Brucker Hyperion 3000 FT-IR microscope. 20x reflective microscope objectives were used to couple light into and out of the fiber. The spectrum was collected using a cryogenically cooled MCT detector.

Continuous wave laser pumping of Fe²⁺:ZnSe was carried out at using a Er:YAG laser operating at 2940 nm (Sheaumann) chopped with a 10$\%$ duty cycle. The sample was mounted in an evacuated Dewar (0.13 Pa) with anti-reflective (AR) coated CaF₂ windows and kept at cryogenic temperatures using liquid nitrogen (77 K). Fe²⁺:ZnSe suffers from multiphoton quenching at room temperature, necessitating cryogenic cooling of the crystal in order to achieve CW operation [41]. An AR coated CaF₂ lens with a focal length of 40 mm was used to couple the pump light into the sample in an end pumped geometry. Fluorescence transmitted through the sample was collected by another 40 mm AR CaF₂ lens and directed into a 150 mm monochromator (Acton Research Corporation) with a 300 g/mm grating blazed at 3000 nm. The signal was collected using a lock-in amplifier (Stanford Research Systems) and cryogenically cooled InSb detector (InfraRed). Three LP-3000 long pass filters (Spectrogon) were used to filter pump light from the signal transmitted through the fiber.