Thulium pumped mid-infrared 0.9–9μm

: We theoretically demonstrate a novel approach for generating Mid-InfraRed SuperContinuum (MIR SC) by using concatenated ﬂuoride and chalcogenide glass ﬁbers pumped with a standard pulsed Thulium (Tm) laser (T FWHM =3.5ps, P 0 =20kW, ν R =30MHz, and P avg =2W). The ﬂuoride ﬁber SC is generated in 10m of ZBLAN spanning the 0.9-4.1 μ m SC at the -30dB level. The ZBLAN ﬁber SC is then coupled into 10cm of As 2 Se 3 chalcogenide Microstructured Optical Fiber (MOF) designed to have a zero-dispersion wavelength ( λ ZDW ) signiﬁcantly below the 4.1 μ m InfraRed (IR) edge of the ZBLAN ﬁber SC, here 3.55 μ m. This allows the MIR solitons in the ZBLAN ﬁber SC to couple into anomalous dispersion in the chalcogenide ﬁber and further redshift out to the ﬁber loss edge at around 9 μ m. The ﬁnal 0.9-9 μ m SC covers over 3 octaves in the MIR with around


Introduction
The Mid-InfraRed (MIR) part of the spectrum is a very exciting frequency region as absorption bands of many organic compounds reside herein, such as sugars, lipids, and proteins, which can be probed to differentiate malignant from benign tissue in detection of skin cancer [1,2]. Absorption bands of atmospheric molecules, such as CO 2 and NO x , likewise reside within this part of the spectrum, as well as explosive materials such as TNT, which also can be probed for detection of air pollution [3] or in stand-off ranged detection of hazardous materials [4], respectively.
Traditionally, Fourier Transform InfraRed (FTIR) spectrometers are used to obtain transmission or reflection spectra of samples with weak globar type sources (emitting in the 1-20μm range). With the advent of Quantum Cascade Lasers (QCLs) spectroscopy has improved as the traditional broadband light sources like the globar in the FTIR could be replaced with a spatially coherent laser source yielding a high power density [5]. While tunable MIR QCLs operating at several wavelengths have been developed [6], a single QCL is still a laser operating at a narrow spectral range, so in order to have the spectral range offered by the globar, multiple expensive QCLs operating at different wavelengths need to be combined together. Supercontinuum sources combine all features of a broadband and spatially coherent source having a high intensity [7]. Mid-IR SC laser sources are currently based primarily on ZBLAN fibers covering the 1-4.75μm spectral range [8][9][10][11][12] and tellurite fibers covering the 1-5μm range [13,14]. Mid-IR SC sources have already been successfully applied in FTIR spectroscopy [15] and in hyperspectral imaging [16]. A further advantage of fiber based laser sources is that the average power can be scaled up by increasing the pump power [9,17], or by increasing the pump repetition rate [8,9], with the current record being a 10.5W average power ZBLAN SC demonstrated by Xia et al. [9].
In order to extend the SC spectrum further into the MIR region, materials other than fluoride and tellurite are needed. For this the chalcogenide glasses (CHALCs) are good candidates as some compositions have been shown to be able to transmit light out to 25μm [18]. Apart from the broadband MIR transmission windows, CHALCs posses very large linear and nonlinear  [20]. Gattas et al. show experimentally that a 1.9-4.8μm MIR SC is obtainable when pumping As-S fibers with Tm amplified picosecond pulses coming from an Er source [21]. Yuan has numerically shown how formation of SC in a As 2 Se 3 Microstructure Optical Fiber (MOF), having an idealised loss without extrinsic impurities, otherwise typically present in a fabricated fiber, can give a 2-10μm output when pumped with femtosecond pulses having 10kW peak power at 4.1μm [22]. Wei et al. likewise numerically obtained a broadband 2-12μm SC in As 2 Se 3 MOF, but in this case pumping at 2.78μm with 1kW peak power femtosecond pulses from a mode-locked Er:ZBLAN laser [23]. In a low loss LiInS 2 planar waveguide Bache et al. showed numerically that it is possible to obtain a 1-8μm SC in just 15mm of waveguide when pumping with femtosecond pulses having a peak power of 120MW launched at 3μm from a tunable Optical Parametric Amplifier (OPA) [24]. Experimentally Yu et al. used an OPA generating femtosecond pulses having 20MW peak power at 5.3μm, where they were able to demonstrate a 2.5-7.5μm SC by pumping a bulk Ge 11.5 As 24 Se 64.5 sample [25]. Waveguides of the same glass composition have also been successfully utilised in MIR chemical sensing [26].
In some of the pump lasers used the peak power is very high of several megawatts and coming from a tunable laser source such as in [24,25]. While these are good approaches for a first theoretical and experimental realisation of a CHALC MIR SC source, such lasers are impractical to introduce in commercial SC products without making them complex and very expensive. The MIR SC designs using a single wavelength pump laser with peak power around 1-10kW, such as 2.5μm in [20], 2.78μm in [23], and 4.1μm in [22], are potentially more commercially feasible. However, these laser sources are not yet commercially available, and thus the proposed SC sources are currently only of academic interest, such as in particular the 4.1μm pump proposed to be generated by down-converting a thulium laser [22].
We circumvent these two issues by approaching the generation of MIR SC by concatenating fluoride and CHALC fibers as this is an easier setup to implement in practice with all the required components being readily available, which makes it also a more commercially friendly approach. The SC concatenation approach relies on the fact that a typical SC spectrum is composed of many solitons in the long wavelength part of the spectrum with anomalous dispersion, pushing the light to longer wavelengths [7]. The solitons are usually femtosecond long pulses having high peak powers, so when concatenating fibers the solitons from the first fiber can continue to redshift in the second fiber, which allows the spectral broadening to proceed, provided the effective soliton number of each individual soliton in the second fiber is larger than 0.5 [27]. In our case the second fiber is a highly nonlinear CHALC fiber and so the transferred solitons will have soliton numbers significantly above 1.5 and therefore undergo soliton fission, which leads to extended redshift and spectral broadening [27].
Conceptually the SC concatenation approach was used before to generate a ZBLAN SC to about 4μm by using an erbium laser pumping a piece of silica fiber to generate a silica SC out to 2.2μm and then coupling this into a ZBLAN fiber [8,9,28]. Later this concatenated approach was improved by amplifying the silica SC using a section of core-pumped thulium amplifier, before coupling it into a ZBLAN [29,11,30] or a chalcogenide fiber [30]. Here we present the first demonstration of how the cascading approach, without intermediate amplifiers, can be used to generate an SC all the way to 9μm.

Concatenated fluoride and chalcogenide glass fiber supercontinuum generation
We extended the ZBLAN fluoride fiber MIR SC by concatenating it with a CHALC fiber that allowed the formation of SC to proceed further into the MIR part of the spectrum, as shown in Fig. 1(a). .5ps, P 0 =20kW, ν r =30MHz, P avg =2W) was coupled to a ZBLAN step-index fiber with a coupling loss of 3dB, where it generated a 0.9-4.1μm SC (-30dB level). The output SC was further coupled into a CHALC As 2 Se 3 MOF with a coupling loss of 6dB, in which it generated a 0.9-9μm SC (-30dB level). The spectrum above shows how a single pulse undergoes spectral broadening through the setup. (b) Dispersion for a ZBLAN fiber with core diameter of 5.7μm and NA=0.30 based on material dispersion by F. Gan [32] and measured loss provided by FiberLabs Inc., Japan [33]. (c) Optical properties of the chalcogenide grapefruit MOF with core diameter of 5 (solid) and 20μm (dashed) based on As 2 Se 3 material dispersion by Amorphous Materials Inc. [34] and loss for the fiber of the same material and design provided by Perfos [35].
A Tm fiber laser operating at 2μm emitting Gaussian shaped pulses having a temporal duration T FW HM =3.5ps, peak power P 0 =20kW, repetition rate ν R =30MHz, and average power P avg =2W [31] drives the entire SC formation process by pumping first the ZBLAN fluoride Step-Index Fiber (SIF). We assumed a coupling loss of α c =3dB when coupling the laser beam into the Fundamental Mode (FM) of the ZBLAN SIF, which occurred due to mode mismatch and Fresnel reflection at the interface of the ZBLAN fiber (n=1.49 at 2μm), so effectively P 0 =10kW drives the broadening process. The ZBLAN fluoride fiber SC is then transferred to an As 2 Se 3 CHALC MOF. Due to the broad bandwidth of the first SC we assume a typical increased coupling loss of α c =6dB due to a large frequency dependent mode mismatch between the two different fiber geometries [36] as well as an enhanced Fresnel reflection at the CHALC fiber interface (n=2.77 at 3.55μm).
We based the ZBLAN fiber design on the material dispersion by Gan [32] and material loss by FiberLabs Inc., Japan [33]. The particular SIF design considered here was in [37] shown to be able to generate a 1-4.1μm SC directly from a 1550nm Er laser. The fiber has a core diameter of 5.7μm and Numerical Aperture (NA) of 0.30, which gave the FM a very low but anomalous dispersion with λ ZDW =1.59μm as seen in Fig. 1(b). The low anomalous dispersion allows the generated solitons to rapidly shift to longer wavelengths due to an enhanced Soliton Self-Frequency Shift (SSFS) yielding an efficient MIR SC broadening [37].
For the SC broadening to continue successfully when the ZBLAN SC was coupled into the CHALC fiber it was required that the fiber had; i) a low propagation loss, and ii) anomalous dispersion across most of the transmission window so that the ZBLAN fiber solitons were allowed to further broaden the spectrum. A CHALC MOF having a 20μm core diameter and a grapefruit design has been fabricated by Troles et al. with low loss out to 8μm [38] and a similar CHALC MOF was fabricated by Adam et al. with low loss out to 8.5μm [39], where in the latter case the long wavelength transmission was set by the As 2 Se 3 material loss. Using the grapefruit design we modeled the dispersion for the FM in the 1-9μm range for core diameters of 5 and 20μm based on the As 2 Se 3 material dispersion obtained from Amorphous Materials Inc. [34], which is seen in Fig. 1(c). In the large 20μm core the FM was primarily affected by the bulk material dispersion with λ ZDW =6.05μm, which is close to the material λ ZDW =7.55μm. Decreasing the core size enhanced the waveguide contribution to the total fiber dispersion increasing its numerical value thereby shifting the λ ZDW to shorter wavelengths [40]. Decreasing the core diameter down to 5μm shifts the λ ZDW to 3.55μm, which is below the long wavelength transmission edge of the ZBLAN fiber at 4.1μm. This allowed all solitons in the 3.55-4.1μm part of the ZBLAN SC to couple into anomalous dispersion in the CHALC MOF.
Decreasing the size of the fiber core to shift the λ ZDW to shorter wavelengths has the drawback that it increases the fiber losses due to micro-deformations [41][42][43]. Troles et al. measured the loss in a 20μm grapefruit fiber and in a suspended core fiber having a core diameter of 4.5μm at 1.55μm, where they observed an increase of 0.4dB/m larger loss in the smaller core fiber [38]. This is well-known to be due to micro-bending losses [43]. In order to take into account the increased loss due to the micro-deformations in the CHALC MOF having the 5μm core diameter we used the measured loss of a 20μm CHALC grapefruit fiber provided by Perfos [35] and to it add an additional loss of 0.4dB/m at all wavelengths as a worst case scenario, as seen in Fig. 1(c).
Using the dispersion and losses of the ZBLAN and CHALC fibers, as well as the nonlinear material response of the ZBLAN fiber given in [37], and of the As 2 Se 3 CHALC composition given by Ung et al. [44], the concatenated MIR SC generation was modeled based on the Generalised Nonlinear Schrödinger Equation (GNLSE). We detail on the use of the equation in modeling ZBLAN fiber SC in [37,45]. The generated MIR SC at the end of each fiber is shown in Fig. 2.
In Fig. 2(a) is shown a single pulse spectrogram of SC at the end of 10m of ZBLAN fiber with spectrum below. We see that the solitons above the λ ZDW at 1.59μm and the Dispersive Waves (DWs) below 1.59μm, nicely follow the total linear group delay β 1 L curve (solid black), as is well-known in formation of SC [7]. The time trace in Fig. 2(c) of the ZBLAN fiber SC is only for the 3.5-4.5μm part of the SC and clearly shows the many MIR solitons with peak power around 10kW, which were responsible for the further broadening in the CHALC MOF. An ensemble average of ten pulses is used to evaluate the IR edge (-30 dB level) seen in Fig. 2(e). It shows that SC develops over three meters of the ZBLAN fiber, after which the broadening began to stagnate. The accumulation of the 3-5μm IR power started to stagnate at around 10m, which we took as the optimum length, as here the highest amount of IR power was transferred into the CHALC MOF for further broadening. A longer piece of fiber decreases the IR power due to the fiber losses [37].  MOF having the 5μm core diameter with the spectrum given below. The IR edge of the SC was shifted to 9μm when evaluated at the -30 dB level, which was due to the solitons undergoing SSFS. The short wavelength edge remained at 0.9μm, which was the same wavelength as the ZBLAN fiber SC. The time trace of the 3.5-9μm part of the CHALC MOF SC is seen in Fig.  2(d). Comparing it with the time trace of the ZBLAN SC (see Fig. 2(c)) shows clear soliton fission of the ZBLAN fiber solitons when coupled into the CHALC MOF, as the time trace of the CHALC MOF SC became much denser [7]. The plotted total linear group delay in the single pulse spectrogram evaluated at the end of the CHALC MOF for both the ZBLAN and the CHALC fibers are shown as black solid and dashed lines, respectively. They reveal that the final SC was composed of the SC originating from both fibers, as expected, where the spectral part of the CHALC fiber SC in the 4.5-5μm range on the spectrogram has a clear signature of the long wavelength part of the ZBLAN fiber SC. In Fig. 2(f) is shown a ten pulse averaged IR edge of the CHALC MOF SC and shows that the SC developed in around 5cm of CHALC MOF due to a combination of the high peak power ZBLAN fiber solitons and the strong nonlinearity of the As 2 Se 3 composition. Just as the ZBLAN fiber SC is broadened as the solitons undergo SSFS, the solitons coupling into the anomalous dispersion of the CHALC MOF continued to undergo SSFS and move the IR edge to the long wavelength transmission edge of the CHALC MOF. The accumulated power in the 6-9μm range shows that around 15mW power could be converted over 10cm of CHALC MOF. Finally, we compared the influence of the dispersion of the CHALC MOF on the SC generation by comparing the obtained spectrum in the small 5μm and the large 20μm CHALC MOF, as seen in Fig. 3. When transferring the ZBLAN SC into the 20μm CHALC MOF with normal dispersion below λ ZDW =6.02μm (green dashed), the solitons just undergo Self Phase Modulation (SPM) leading to a minor broadening. On the other hand, when transferring the ZBLAN fluoride SC into the 5μm CHALC fiber, the long wavelength part of the SC overlaps with the CHALC anomalous dispersion regime beginning at λ ZDW =3.55μm (blue dashed line) thereby allowing the ZBLAN fiber solitons to continue to undergo SSFS and further push the IR edge to 9μm. Apart from the optimised dispersion, the 5μm core diameter fiber also had a 16 times smaller effective mode area resulting in a much stronger nonlinearity, which additionally enhances the broadening in the 5μm core MOF compared to the 20μm one.
The presented results show that when pumping properly designed concatenated ZBLAN and CHALC fibers with a standard pulsed Tm laser it is possible to obtain an SC spanning from the short wavelength edge of the ZBLAN SC at 0.9μm to the long wavelength edge of the CHALC MOF SC at 9μm. A broadening of more than three octaves has thus been obtained without the need of expensive and intricate pump laser sources.

Conclusion
In conclusion, we have demonstrated how a more than three octave broad -0.9-9μm (-30dB level) -MIR SC could be generated from a standard 2μm Tm mode-locked laser by concatenating (commercially available) ZBLAN step-index fibers with λ ZDW below 2μm with properly designed (commercially available) CHALC MOFs.
The ZBLAN SC extended to 4.1μm, and we showed how a CHALC MOF with λ ZDW sufficiently below 4.1μm was suitable, because it allowed enough of the high peak power solitons in the IR part of the first ZBLAN fiber SC to couple into anomalous dispersion and continue to redshift and broaden the spectrum.
The particular MOF we chose was a 5μm core diameter grapefruit MOF, but the concept applies to any CHALC fiber with a λ ZDW sufficiently below the IR edge of the applied first SC. With the chosen configuration we were able to generate an MIR spectrum with an average power of 15mW above 6μm from the 2W average power 30MHz Tm pump laser. This can be improved upon in several ways. The ZBLAN SC can be improved to have a higher IR power and extend further into the infrared to say 4.75μm, as was recently demonstrated [12]. A tellurite or indium fluoride (InF 3 ) [46] fiber with the loss edge at a longer wavelength could be used instead of the ZBLAN fiber. The two approaches each serve to push more power out into the infrared part of the first SC. Given a fixed first SC, the CHALC fiber could be improved to have a lower λ ZDW and longer wavelength loss edge [18,47]. Thus there is amble room for improvement of this first demonstration of a 0.9-9μm SC, in which we used only commercially available fibers and pump lasers.