1. Introduction

Bulk lasers based on Cr²⁺:ZnSe and Fe²⁺:ZnSe have become well known, commercial mid-IR lasers because of their broad tunability throughout the 2 to 3 micron and 3.7 to 5 micron wavelength regions, respectively [1,2]. However, these bulk lasers have a fundamental limitation to their maximum power output because of thermal effects caused by the large thermo-optic coefficient of ZnSe (dn/dT = 7.0×10$^{-6}$ K⁻¹) [3,4]. Current strategies to overcome this issue include ultra-fast laser inscribed waveguides [5], planar waveguides [6], and rotating disk lasers [7]. These strategies have shown to be capable of power scaling Cr²⁺:ZnSe lasers to some degree; but thermal lensing is still problematic at high powers, or these solutions are not suitable for portable environments. Another strategy to overcome this issue has been to fabricate transition metal doped ZnSe (TM²⁺:ZnSe) material into an optical fiber, which due to its intrinsically high aspect ratio, offers superior thermal dissipation to a bulk crystal. In early attempts at this, Cr²⁺:ZnSe nanoparticles were embedded in index matched chalcoginide glass fibers [8]. Lasing was observed in these fibers, but the scattering losses were high. Most recently, Cr²⁺:ZnSe and Fe²⁺:ZnSe optical fiber lasers grown via high pressure chemical vapor deposition (HPCVD) have been demonstrated, producing mid-IR light via direct laser emission in completely crystalline optical fibers [9,10].

HPCVD grown transition metal doped ZnSe (TM²⁺:ZnSe) optical gain media are unique compared to commercial bulk TM²⁺:ZnSe gain media because they are doped in situ during the deposition, while the latter are diffusion doped [11]. Diffusion doping is a slow process (days to weeks) that is only capable of producing a uniform concentration of dopant on millimeter length scales [12]. Conversely, HPCVD is a simple, quick, and scalable technique that is able to produce centimeter to meter long fibers of semiconductor material such as Si, Ge, and ZnSe via templated high pressure chemical reactions [13–16]. The HPCVD of TM²⁺:ZnSe, which is a high pressure modification of ZnSe metalorganic chemical vapor deposition (MOCVD) [17], utilizes alkyl zinc and selenium precursors in flowing high pressure H₂ with the addition of a chromocene or ferrocene derivative to act as the source of transition metal dopant. Typically, 1 cm long fibers are produced within 2 days. The deposition is believed to occur through the following series of, simplified, gas phase reactions. Here, chromocene is used as an example reactant for Cr²⁺:ZnSe.

However, little is understood about the micro- and nanoscale dopant distribution within the HPCVD grown fiber structures. Currently, HPCVD grown TM²⁺:ZnSe fiber lasers have only demonstrated small output powers on the order of microwatts; nanoscale inhomogeneous doping could be a contributing factor. It is well known that crystalline defects, such as twinning and high concentrations ($>$ 2×10$^{19}$ ions/cm$^{3}$) of dopant lead to multiphonon quenching, inhomogeneous spectral broadening of the laser output, and decreased laser performance [18–20]. Additionally, such defects also increase scattering and absorption losses as well as provide nonradiative decay pathways which can limit lasing efficiency.

Analyzing the dopant distribution in our HPCVD grown structures presents an analytical challenge because of small dopant concentrations coupled with the size of our fiber cores which can range from 15 to 75 µm in diameter. Here we utilize synchrotron micro/nano X-ray fluorescence mapping and micro X-ray near edge absorption spectroscopy (XANES) techniques [21] to study the dopant distribution and local oxidation state in our fibers. These techniques allow us to map large areas of our structures (10s to 100s of square microns) with a fine spatial resolution (as small as 30 nm × 30 nm) as well as extract valuable chemical information about the local environment of the dopant ions. Furthermore, we analyze commercially available bulk laser crystals to see how our HPCVD grown material compares. These experiments indicate that dopant aggregation at grain boundaries exists in the HPCVD grown laser material which can ultimately limit lasing performance.

2. Experimental

2.1 Fabrication

Fabrication of the TM²⁺:ZnSe optical fibers was carried out via previously published methods [9,10]. Briefly, a high-pressure microfluidic reactor was constructed from commercially available 360 µm outer diameter silica capillaries through which the pressurized reactants flowed and deposition occurred in one step on the silica capillary walls. A 150 µm inner diameter (I.D.) capillary is used to contain the transition metal dopant, a 50 to 75 µm I.D. fiber is used as the deposition template, and a 15 µm I.D. capillary is used to limit the flow through the reactor. The 3 capillaries are spliced together using a fiber fusion splicer to construct the entire reactor. Figure 1(a) shows a diagram of the reactor construction.

 

Fig. 1. a) The HPCVD reactor constructed by splicing together silica capillaries and allows for the reactants to flow through it, driven by high pressure carrier gas . A two zone furnace allows the vapor pressure of the dopant source and the deposition kinetics to be controlled independently. b) Optical micrograph of deposited ZnSe fiber and fiber facet. Due to the nature of the deposition, a small central pore remains. Here the pore’s size is magnified by the curvature of the glass capillary.

Download Full Size | PDF

Electronic grade dimethyl zinc and dimethyl selenide vapor (1:2 vapor pressure ratio) were pressurized in 40 MPa UHP H₂ gas and configured to flow through the reactor. Liquid transition metal precursor was contained within a smaller 1 cm long, 5 µm I.D. capillary within the larger 150 µm I.D. silica capillary. The reactor is placed inside of a custom made two zone furnace which allows for independent control of the vapor pressure of the dopant precursor and the deposition kinetics. The first zone of the furnace is 10 cm long and the second zone is 25 cm long. The reactor is held within a 1 mm I.D. alumina tube in the center of the furnace. Using this furnace, the dopant is heated to 200 °C to produce the appropriate vapor pressure necessary to deposit either chromium or iron doped zinc selenide. For Cr²⁺:ZnSe, bis(ethylcyclopentadienyl)chromium(II) is used as the source of Cr, for Fe²⁺:ZnSe, t-butylferrocene is used. The deposition zone of the reactor was heated to 450 °C and deposition completed within 2 days, indicated by no flow coming through the reactor. The deposition proceeds inward toward the center of the silica capillary, starting at the walls, and stops once a plug forms in the material. This invariably leaves a small, central pore on the order of 200 nm to 700 nm that spans the entire fiber core. In the case of the Cr²⁺:ZnSe fiber used in this study, heating of the dopant precursor was delayed to deposit an intrinsic layer of ZnSe around the doped core. Figure 1(b) shows an optical micrograph of a ZnSe optical fiber imaged from the side through the silica template as well as a polished cross section of the fiber as well as a diagram of the HPCVD reactor.

2.2 X-ray sample preparation

Important considerations are necessary to produce good samples for X-ray characterization. The sample should be thick enough to produce enough signal to detect X-ray emission but also to avoid significant self-absorption. Additionally, the sample thickness will dictate the ultimate spatial resolution we are able to achieve with mapping since the produced maps are a 2D projection of a 3D sample. Ideally, sample thickness should be on the same order as beam size.

To prepare HPCVD samples for the X-ray studies, first the TM²⁺:ZnSe fiber cores are etched from the silica deposition template/reactor using a 49$\%$ hydrofluoric acid solution. Hydrofluoric acid readily etches the silica, but leaves the ZnSe core intact. Next, the cross-section of the exposed core was polished using an Ar⁺ ion mill (Hitachi IM4000plus). Lastly, a thin cross-section, roughly 1 µm thick was milled from the etched core using a FEI Helios 660 focused ion beam (FIB) and mounted to a molybdenum FIB grid using the FIB’s micromanipulator and carbon deposition. Because of the large diameters of the fiber cores, 50 µm to 75 µm, only sectors of the cross-sections could be milled out. Cross-sections of Cr²⁺:ZnSe and Fe²⁺:ZnSe optical fibers were produced. Figure 2 shows an SEM image of the Cr²⁺:ZnSe fiber cross-section.

 

Fig. 2. SEM image of FIB Cr²⁺:ZnSe fiber cross-section. The cross-section is attached to a Mo FIB grib with carbon deposition. Beam damage from the FIB process is visible on the sample surface. The feature in the center is the fiber void. The cross-section had to be squared off due to it’s large 75 µm diameter.

Download Full Size | PDF

For purposes of comparison, bulk Cr²⁺:ZnSe and Fe²⁺:ZnSe laser crystals were purchased from IPG Photonics (Oxford, MA). The Cr²⁺:ZnSe crystal was reported to have a concentration of 1.01 × 10$^{19}$ ions/cm$^{3}$ and the Fe²⁺:ZnSe crystal was reported to have a concentration of 2.23 × 10$^{18}$ ions/cm$^{3}$. 1 µm thick 10 µm × 10 µm FIB slices were taken from these samples and subjected to the same measurements and analysis as our HPCVD grown material.

2.3 X-ray characterization

2.3.1 $\mu$XRF and nano-XRF

$\mu$X-ray fluorescence measurements were performed at beamline 2-ID-D at the Advanced Photon Source at Argonne National Lab. An X-ray excitation energy of 9.60 keV was used to probe the transition metal ions in the samples. An x-ray zone plate provided a spot size of 200 nm × 200 nm and a 3-axis stage allowed the sample to be moved so that concentration could be spatially mapped with a resolution of 200 nm × 200 nm. The sample compartment was continuously flushed with He to reduce the air background signal. The k$_{\alpha }$ emission of Cr (5.415 keV) and Fe (6.405 keV) were used to create elemental distribution maps in their respective samples.

The nano-XRF measurements were performed at beamline 26-ID-C at the Advanced Photon Source at Argonne National Lab. Here, an x-ray excitation energy of 9.50 keV was used to probe the transition metal ions in the samples. An x-ray zone plate provided a spot size of 30 nm × 30 nm and a 3-axis stage allowed the sample to be moved so that concentration could be spatially mapped with a resolution of 20 nm × 20 nm to create an over-sampled image. The sample compartment was evacuated with a turbo pump to 1×10$^{-5}$ torr to reduce the air background. Again the k$_{\alpha }$ emissions of Cr and Fe were used to produce the maps.

The data from both beamlines were quantified using a NIST XRF standard, however the reported concentrations can only be considered estimates of the true concentrations due to matrix differences between the samples and the standards. The data was analyzed using the MAPS software package [22].

2.3.2 $\mu$XANES

$\mu$XANES measurements were performed at beamline 2-ID-D at the Advanced Photon Source at Argonne National Lab. A similar setup to the $\mu$XRF experiments was used. XANES was collected by scanning x-ray excitation energies close to the element’s absorption edge and measuring the k$_{\alpha }$ fluorescence intensity at each wavelength. A fluorescence spectrum was collected for 5 s for each excitation wavelength. Each scan was performed 3 times and the results were averaged. The fluorescence intensity was then normalized to the incident x-ray intensity to calculate the absorption coefficient, $\mu$(E), for each excitation wavelength. For Cr containing samples, excitation wavelengths were scanned from 5980.0 eV to 6040.0 eV in 0.5 eV increments. For Fe containing samples, excitation wavelengths were scanned from 7100.0 eV to 7160.0 eV in 0.5 eV increments. Using the x-ray zone plate, we were able to collect XANES from 200 nm × 200 nm spots along the sample. Cr and Fe standards were used to compare the absorption edge energy of the samples to compounds of known oxidation state. For Cr containing samples; Cr metal, CrSe, and Cr₂O₃ were used as standards. For Fe containing samples; Fe metal, FeSe, FeO, and Fe₂O₃ were used as standards. Data reduction was performed using Athena [23].

3. Results and discussion

3.1 XRF

By utilizing beamlines with x-ray micro- and nanoprobe capabilities, we can create both long and short range images of the dopant distribution in our structures. We consider these maps representative samples of HPCVD grown Cr²⁺:ZnSe and Fe²⁺:ZnSe fiber lasers as both samples were sourced from fibers that demonstrated lasing. Figure 3 shows the $\mu$XRF maps collected from a quarter of the FIB cross section of the Cr²⁺(a) and Fe²⁺(b) fibers. The center of fibers are oriented to the bottom and top left hand corners respectively. These maps have a spatial resolution of 200 nm. In the Cr²⁺ map (Fig. 3(a)), there are several distinctive regions. The outermost region, labeled "1" in the figure shows the boundary layer between the initially deposited intrinsic ZnSe cladding layer and the beginning of the deposition of the Cr²⁺:ZnSe core. Once the doping process begins in region 2, the amount of Cr incorporated into the material appears to increase. The doped material seems to eventually aggregate into highly doped lobes, as seen in the XRF image. Once the Cr²⁺ concentration reaches a maximum average of 0.135 µg cm⁻² (1.6×10$^{19}$ ions/cm$^{3}$), it immediately begins to sharply drop (region 3) to a concentration of 0.03 µg cm⁻² (3.5×10$^{18}$ ions/cm$^{3}$).

 

Fig. 3. a) $\mu$XRF map of Cr²⁺:ZnSe optical fiber cross-section. Region 1 is the intrinsic ZnSe cladding layer. Region 2 is defined by lobes of doped regions. Region 3 shows how Cr concentration quickly drops off as the deposition completes. b) $\mu$XRF map of Fe²⁺:ZnSe optical fiber cross section. The bright border around the sample originates from a thin layer of glass on the sample. c) $\mu$XRF map of IPG Cr²⁺:ZnSe sample. d) $\mu$XRF map of IPG Fe²⁺:ZnSe sample.

Download Full Size | PDF

In the Fe²⁺:ZnSe sample (Fig. 3(b)), the overall Fe concentration is more uniform, however locally high Fe concentrations appear inside individual grains or aggregation in grain boundaries. The bright ring that surrounds the fiber cross-section is due to a thin layer of glass that remained on the fiber. In this sample, concentration ranges from 0.120 µg cm⁻² (1.3×10$^{19}$ ions/cm$^{3}$) to 0.062 µg cm⁻² (6.7×10$^{18}$ ions/cm$^{3}$) .The map reveals the grain structure of the sample as well as the columnar growth pattern that was reported previously [14]. Figures 3(c) and 3(d) show $\mu$XRF maps of the commercial samples. At this length scale these samples do not show any inhomogeneity.

Figures 4(a) and 4(b) shows the nano-XRF maps from the Cr²⁺:ZnSe (4(a)) and Fe²⁺:ZnSe (4(b)) fibers. In the case of the Cr²⁺:ZnSe, the nano-XRF image was taken at the bottom of the center lobe in Fig. 3(a). Here the drastic change in nanoscale dopant concentration can be more clearly observed, with values ranging from 0.14 to 0.08 µg cm⁻². The nano-XRF image of the Fe²⁺:ZnSe fiber has similar features, with the high dopant regions forming a web like structure in the image. Dopant aggregation in the grain boundaries is suspected to cause this.

 

Fig. 4. nano-XRF maps of a) Cr²⁺:ZnSe fiber, b) HPCVD Fe²⁺:ZnSe fiber, c) Cr²⁺:ZnSe IPG crystal, d) HPCVD Fe²⁺:ZnSe IPG crystal.

Download Full Size | PDF

Lastly, Figs. 4(c) and 4(d) show the nano-XRF maps collected from commercially available Cr²⁺:ZnSe and Fe²⁺:ZnSe diffusion doped gain media. The most striking difference between the HPCVD grown samples and the commercial samples is that the commercial samples do not have any observable nanoscale dopant structure. The commercial samples appear to have a random distribution of the dopant. Here it is important to note a key difference between the grain size of HPCVD and diffusion doped commercial samples. Bulk diffusion doped samples have grain sizes on the order of 100’s of microns, while HPCVD grown ZnSe is known to have grain sizes on the order of 100’s of nanometers. Thus, the nano-XRF map of the HPCVD grown fibers contain many more crystalline grains than the maps of the commercial samples.

From the XRF maps, it is clear that inhomogeneous doping is present in the HPCVD grown fiber material. The nXRF maps indicate that the difference between the areas of high and low concentration differ by approximately 20%. Additionally, the concentration of dopant reaches concentrations near 2×10$^{19}$ ions/cm$^{3}$ to 3×10$^{19}$ ions/cm$^{3}$ which is where concentration quenching in both Cr²⁺:ZnSe and Fe²⁺:ZnSe begins to affect lasing through decreases in fluorescence lifetime and other nonradiative processes, such as Cr-Cr or Fe-Fe energy transfer [20,24]. Ideal concentrations of both Cr and Fe fall within the 3×10$^{18}$ ions/cm$^{3}$ to 1×10$^{19}$ ions/cm$^{3}$ for diffusion doped material. Furthermore, nanoscale changes in dopant concentration lead to nanoscale changes in refractive index, potentially increasing the scattering losses in the fiber. Losses reported in Cr²⁺:ZnSe gain-switched fiber lasers have been reported to be as high as 3.0 dB cm⁻¹ [9] which is higher than the 1.0 dB cm⁻¹ loss reported for intrinsic ZnSe optical fibers [14]. Nanoscale refractive index changes in the doped material could explain the high loss values in the doped fibers.

3.2 XANES

X-ray absorption spectroscopy is a useful technique for probing the oxidation state and coordination chemistry of a molecule. Here we will be concerned with the x-ray absorption edge which gives information about the oxidation state of the atom being probed and the region immediately adjacent it, and the XANES region which can give information about the atom’s local chemical environment. The X-ray absorption edge energy is extremely sensitive to oxidation state because of the different electronic configuration each state has [25]. This can be seen in the absorption spectra of our standard compounds in the supporting information (See Supplement 1). By comparing the edge energy of our samples to standard compounds of known oxidation state, we can determine the oxidation states of the probed ion in our sample. If mixed oxidation states are present, the edge energy would be shifted to an intermediate energy between two oxidation states. Using the beamline instrumentation, we were able to collect XANES spectra at several points on the sample. For the Fe²⁺:ZnSe fiber XANES spectra were collected in 1 micron increments starting from the sample edge; for the Cr²⁺:ZnSe fiber spectra were collected in 5 micron increments.

Figure 5 summarizes the major results from the XANES study, the full dataset is shown in the supporting information (Supplement 1). By comparing the collected spectra for all samples (fibers and commercial) to the standards, it is clear that all samples contain the doped transition metal in the 2+ oxidation state. The absorption edge for Cr²⁺ was found to be 5995 eV and the edge for Fe²⁺ was found to be 7119 eV. Figures 5(a) and 5(b) shows CrSe and FeSe x-ray absorption spectra overlayed over a representative spectrum from the corresponding transition metal doped fiber and commercial sample. The edge energies are closely matched, indicating the prevalence of Cr²⁺ or Fe²⁺ in the samples. However, there are differences in the XANES region between the selenide standards and the sample crystals. However, this is expected since the coordination environments and crystal structure of pure FeSe/CrSe are very different that the corresponding transition metal doped ZnSe. The structure in the XANES region of the commercial samples closely resembles absorption spectra collected in previous studies [26,27]. The absorption spectra from our HPCVD doped fibers closely matches the spectra collected from the commercial samples, which indicates that the chemical environment of the dopant in the fiber is the same as diffusion doping.

 

Fig. 5. a) XANES spectra of CrSe standard, Cr²⁺:ZnSe crystal purchased from IPG, and HPCVD Cr²⁺:ZnSe fiber. b) XANES spectra of FeSe standard, Fe²⁺:ZnSe crystal purchased from IPG, and HPCVD Fe²⁺:ZnSe fiber.

Download Full Size | PDF

4. Conclusion

In this work we’ve demonstrated how synchrotron XRF mapping and XANES can be used to study the distribution and oxidation state of dopants in optical fibers. These synchrotron techniques can be especially useful to study new fiber fabrication techniques, such as HPCVD, and give insight on how to solve the unique material challenges that arise. Here, HPCVD grown TM²⁺:ZnSe optical fibers offer a promising path toward realizing direct mid-IR fiber laser emission. This work shows that HPCVD is capable of depositing material chemically similar to bulk TM²⁺:ZnSe produced via diffusion doping. However the nanoscale dopant distribution is not homogeneous, which if ameliorated provides the likely path to further improving the HPCVD TM²⁺:ZnSe fiber laser efficiencies. We believe this issue can be addressed via either choosing different dopant precursors to have more favorable dopant chemistry, or through unique post-processing techniques which were developed for this unique material system. Laser annealing [28] and micro-chemical vapor transport [29] are two recent techniques that have been shown to be capable of recrystallizing HPCVD grown semiconductor material. Both of these recrystallization techniques should homogenize the doping profile in these fibers and also lower the optical loss through grain growth, potentially leading to power scaling in HPCVD Cr²⁺:ZnSe and Fe²⁺:ZnSe optical fiber lasers.