Mid-infrared hollow core fiber drawn from a 3D printed chalcogenide glass preform

: We report the fabrication of a microstructured optical fiber drawn from a soft glass 3D printed preform. For this proof of concept, a chalcogenide glass that is well known for its capability to be shaped at low temperature and its mid-infrared transmission was selected: Te 20 As 30 Se 50 . The obtained negative curvature hollow core fiber shows several transmission bands in the 2-12 µm range that are reproduced numerically using finite element-based simulations and coupled mode theory.


Introduction
Additive manufacturing, or 3D printing has proven to be a powerful elaboration method in materials science. Predominantly used for polymers [1][2][3] additive manufacturing has been extended to metals [4], to ceramics [5], and quite recently to glasses [6][7][8][9][10][11][12]. In this study, the fabrication of a chalcogenide microstructured optical fiber (MOF) drawn from a 3D printed preform is reported for the first time. Transmission of infrared light through the fiber core is demonstrated, and the recorded main optical transmission bands of the as-prepared fiber are successfully compared to simulated ones. For this proof of concept, a soft glass such as chalcogenide glass has been chosen because of its well-known capability to be shaped at low temperature (T<600 °C) and its mid-IR transparency. Chalcogenide glass fibers have been implemented with success in innovative mid-IR systems for versatile mid-IR fiber transmission [13,14], supercontinuum generation [15][16][17][18], and sensing [19]. The objective of this work is to realize a chalcogenide hollow core (HC) MOF for transmission in the mid-IR region (2-12 μm) [20,21], by using an original additive manufacturing process. Actually, with HC MOF structure, it is possible to obtain transmission beyond the transmission limits of a solid core fiber and very low attenuation [22]. Laser power delivery would be greatly improved by using such fibers as compared to solid core fibers. It would then be possible to propagate a laser beam in the infrared, like CO 2 laser at 10.6 μm, for industrial cutting or for surgery. In this context the Te 20 As 30 Se 50 (TAS) chalcogenide glass showing good mid-IR transmission and a low glass transition temperature (T g ) suitable for 3D printing by thermal filamentation has been chosen. The glass transition temperature of TAS glass is 137°C. Thanks to this low T g , a viscosity of 10 3 -10 2 Pa.s is obtained around 280°C, which is perfectly well adapted for fused filamentation fabrication (FFF) [23]. Such additive manufacturing approach allows the fabrication of preforms with complex designs in a single step within a couple of hours, with a high degree of repeatability and accuracy of the geometry. One can note that the classical way for making MOF preforms is the stack and draw technique [24], which is a time-consuming method and presents limited possible geometries. One can note, that for soft glasses like chalcogenide glasses preform extrusion is also an alternative method to the stack and draw technique [20]. The 3D printing way for obtaining optical fibers is A c c e p t e d M a n u s c r i p t utilized currently with polymers for which various geometries have been obtained including MOF geometries. In addition, it has been shown that a silica fiber can be drawn from 3D printed preforms made by stereolithography with UV-curable composite resin containing silica particles [11]. However, no MOF obtained from a 3D-printed inorganic glass preform, including chalcogenide glass, has ever been reported. The closest related work deals with the demonstration of 3D printed bulk As 2 S 3 glass obtained by FFF [12].
In this paper, a commercial 3D printer, commonly used with plastic filaments, is modified to accept chalcogenide glass rods for the purpose of building structured fiber optic preforms. Chalcogenide glasses are fabricated using methods described elsewhere [25,26] in order to produce suitable feedstocks. A solid glass cylindrical preform was first produced as a vehicle to optimize the 3D printing technique and then drawn into a single-index fiber. This allowed to identify optical losses introduced by the 3D printing technique in excess of those from traditional fiber drawn from solid melt-cast preforms. The technique is then validated by printing a structured preform and drawing that into a hollow-core micro-structured optical fiber (HC-MOF). Fiber transmission was measured in the mid-IR and the resulting optical performance was then simulated using the finite element method (FEM).

Chalcogenide glass synthesis
For the additive manufacturing study, the TAS chalcogenide glass is prepared by the conventional melt-quenching method by, first, introducing high purity As (99.999%), Se (99.999%) and Te (99.999%) in a silica ampoule (12 mm inner diameter), which is then pumped under vacuum for 3 h. Further, the silica ampoule is sealed and placed in a rocking furnace at 850° C during 10 hours to homogenize the melt. The glass is then quenched by immersion of the ampoule in water, and annealed above T g to remove mechanical constraints. For obtaining the rods used as raw materials for the 3D printer, the 12-mm diameter rods were stretched into 3-mm canes within the drawing tower.
For obtaining a higher optical quality glass, the addition of TeCl 4 and aluminum (1000 weight ppm and 100 weight ppm, respectively) to the raw starting elements permits the capture at high temperature of remaining hydrogen, carbon and oxygen by forming HCl, CCl 4 , Al 2 O 3 . The mixture is then cooled down to 700 °C, quenched into water by immersion of the ampoule, and annealed above the glass transition temperature T g . Then, the chalcogenide glass rod, obtained after opening the ampoule, is placed into a distillation ampoule to be purified by several distillation steps. It is first distilled under dynamic vacuum to eliminate the low vapor pressure impurities such as carbon tetrachloride and HCl residues. In the second step, a distillation under static vacuum permits the elimination of refractory impurities, such as carbon, silica, alumina and other refractory oxides. When the distillation process is complete, the glass is further homogenized at 850 °C for 10 hours, cooled to 550 °C, quenched in water, and annealed slightly above T g . This purified glass rod is used for drawing a single-index fiber whose optical attenuation serves as a reference for the results obtained with fibers drawn from 3D-printed preforms.

3D printing process
The 3D-printing set-up is based on a customized commercial cartesian Arduino-based RepRap-style 3D printer upgraded for soft glasses and running Marlin firmware [27]. Especially, the feeding mechanism is customized to handle suitably brittle materials. This mechanism is supplied with 500 mm long glass rods of 3 mm diameters that are produced by the fiber drawing method. An extruder drives the raw material filament to a nozzle heated well above T g . The printer head moves along the X and Z axes while depositing the TAS glass as 100-µm thick layers and the bed plate moves along the Y axis. The temperature of the extruder head was fixed at 300 °C. The size of the nozzle, and consequently, the width of A c c e p t e d M a n u s c r i p t the printed filament was about 400 µm. The TAS glass was deposited on a sodalime silicate glass bed plate heated around 140 °C (i.e. slightly above T g ) in order to ensure a good adherence of the first printed layers. For optimum control of the extrusion and the nozzle displacement during printing, a simple G-code input file was written using circular motion commands. Let us note that for those proof-of-concept experiments, the 3D printing process is not yet done under controlled atmosphere but in ambient conditions. The 3D printer is installed under a fume hood in order to protect the room environment from any volatile gases that could be generated during printing. Figure 1 shows an example of a 3D-printed chalcogenide TAS glass cylinder (preform used for drawing the single index "printed" fiber, see Section 3.1) and an inner part of it (obtained by cleaving the cylinder), observed by an optical microscope and (inset) a scanning electron microscope. No interface is visible between the filaments in the XY-plane or layers along the z-axis of deposited glass, indicating a good intrinsic quality of the printing. However, several bubbles are observed due to imperfect filaments collapsing and/or gas bubbles formed by the degassing of volatile compounds such as arsenic oxides or selenide oxides [26]. The physical properties of the printed glass have been compared to the ones of the initial glass and reported in Table 1. Energy dispersive spectroscopy (EDS) and differential scanning calorimetry (DSC) measurements confirm that the composition and the glass transition temperature have not been altered during the printing process. Small bubbles present in the glass account for the lower density and result in an increased light scattering that can explain the higher optical losses, from 8 to 28 dB/m in the mid-IR (Table 1).

(a) (b) (c)
A c c e p t e d M a n u s c r i p t

Hollow core 3D printed preform
The selected geometry for investigating the realization of 3D printed preform is a simple 6capillary ring of an anti-resonant structure [28][29][30] as shown in Fig. 2. After 4 hours of printing, a 30-mm long hollow-core chalcogenide preform has been obtained. The outer diameter is 16 mm, the inner diameter 12 mm, and the 6 capillaries show a diameter of 3.4 mm with a thickness close to 0.4 mm (Fig. 2), which is actually the size of a single trace of the printing nozzle. Let us note that the cladding is made from 5 traces.

Drawing process
The printed preform was used to manufacture an anti-resonant hollow-core fiber by using a homemade drawing tower specifically designed for low-T g glasses. The preform is placed in a narrow drawing furnace that can permit to draw short preforms. When the furnace reaches the drawing temperature around 270 °C, a glass drop appears and falls down under gravity forming a glass fiber which is then reeled on a spool in rotary motion. At the same time, the preform is moved down, feeding the drawing furnace. For a given feeding speed of the preform, the diameter of the fiber is controlled by the spool speed. A helium gas flow of 2.5 L/min provides an inert atmosphere around the preform. Usually, during the drawing of MOFs, a differential pressure is applied in the core and in the different capillaries in order to control the geometry. However, for this proof of concept study, no differential pressure was applied. Consequently, the geometry of the capillaries has changed during drawing due to surface tension, more particularly the ratio between the inner and the outer diameter (Table  2). Nevertheless, it has been demonstrated that a preform obtained by additive manufacturing can be drawn into a fiber in which the initial shape is globally maintained (see Fig. 3). The final parameters of the fiber are summarized in Table 2. Although the printed hollow core preform is only 3 cm long, several meters of fibers have been obtained.
A c c e p t e d M a n u s c r i p t

Mid-IR optical measurements
Attenuation of optical fibers were measured by means of a Fourier-transform infrared (FTIR) spectrometer (Bruker, Vector 22) equipped with a mercury-cadmium-tellurium (MCT) detector. Four types of fibers have been analyzed: a single-index fiber made from purified glass, a single-index fiber made from unpurified glass, a single-index fiber made from printed unpurified glass, and a hollow-core fiber made from printed unpurified glass. The intrinsic mid-IR window of the TAS glass fiber was determined by measuring the attenuation of a single-index fiber made from purified glass (see Fig. 4). The measurement was realized by the classical cut-back method applied to a fiber of 12-meter initial length. The attenuation curve in Fig. 4 corresponds to the average of 3 measurements, the final length of the fiber being equal to 1 meter. Considering that the transparency limit corresponds to a 10 dB/m absorption, one can conclude that a TAS fiber made of purified glass transmits light A c c e p t e d M a n u s c r i p t between 2 and 12 µm, with a minimum attenuation of 1 dB/m at 7.5 µm. The peak at 4.55 µm is due to Se-H absorption, which results from a contamination by residual hydrogen adsorbed on the silica ampoules utilized for TAS glass synthesis [25,26].

Bulk single-index fibers
Complementary measurements were carried out to better understand the effects of 3D printing on the glass optical properties. In this study, where the objective is to demonstrate the concept of making microstructured optical fibers from a printed preform, unpurified TAS glasses were implemented. The preparation of unpurified glasses is more straightforward than that of purified ones. Attenuation curves were obtained by using the cut-back method on 1-meter long fibers drawn from the initial glass preform, made from the same raw glass composition as the one used for feeding the 3D printer, and from a printed glass cylinder (8 mm diameter).
Results are shown in Fig. 5. Comparison of Figs. 4 and 5 indicates that single-index fibers possess the same overall Mid-IR transparency window, from 2 to 10 µm, whether the glass is initially purified or not. However, the minimum loss is significantly higher in the unpurified fiber with a value of nearly 8 dB/m at 7.5 µm. Also, two intense absorption peaks at 2.9 µm and 6.3 µm due to, respectively, OH chemicals bonds and molecular H 2 O chemicals bonds were observed. OH and H 2 O are classical pollutants in unpurified chalcogenide glasses [25]. Compared to these two intense absorptions, the Se-H peak at 4.55 µm is too weak to be observed in Fig. 5. Above 10 µm, the apparent decrease of attenuation is an artefact due to degraded signal intensity at the output of the fiber. The single-index fiber obtained from a printed glass shows an important increase of the optical losses, which reach 28 dB/m at 7.5 µm. This important increase can be explained by two main reasons: the presence of scattering bubbles (already observed in a bulk 3D printed TAS glass, Fig. 1) and some weak crystallization of the glass occurring during printing, which is done under ambient air for those proof of concept experiments. Moreover, in addition to the OH and molecular H 2 O signatures at 2.9 µm and 6.3 µm, numerous other IR absorption peaks are observed. The peak at 3.3 µm is due to C-H chemical bonds induced by an organic pollution, while Se-H bonds are responsible for the absorption at 4.55 µm. Carbon impurities, which are not present in fibers prepared by the classical method, are likely to originate from traces of hydrocarbons present in the 3D printer. The reappearance of hydrogen results from a chemical reaction between the glass and moisture in ambient atmosphere, during printing. The origin of bands at 5.8 µm, 6.8 µm, and 7.2 µm could not be established. Finally, the 7.9 µm and 9.1 µm absorption peaks could be attributed to metal-oxygen vibrations due to a weak oxidation of the chalcogenide glass during printing. A c c e p t e d M a n u s c r i p t

Modelin
The guiding p homemade m The simulatio experimentall fiber cross-se problem [33]. (FEM) [34]. propagation c the gmsh/getd in Ref. [31]. I fibers includ microscopy im hollow core fib ical losses of th it to achieve an ssion of a 20-cm investigated ( ometer mention of the FTIR w 1mm 2 ), part of ore fiber. To through the ho the surface of field measurem ength range (F the fiber, reco propagates in n the core of th rted in Fig. 6(a cteristic of the s therein for a 6. Mid-infrared tra ass preform: (a) Tr output profile r ng and comp properties of th ode solver bas ons were perfor ly and the TAS ection is invar . This eigenval We use a ve constants (eigen dp free open-so It has already ing ones whe mages [37,38]. ber he fiber (proba ny cut back me m long hollow- (Fig. 6). The t ned above (Fig  as injected into f the light was make sure tha ollow core of the fiber. Inde ments were car FLIR A655sc). orded with an i n the hollow he fiber was re a). The numero propagation b comprehensive nsmission spectru ransmission spectr recorder with a mi arison with e he HC-MOF co ed on the finite rmed from the S material disp riant along the lue problem is ector FEM ba nvalues) and th ource software been used suc ere the fiber It is worth me ably close to 10 easurement and -core fiber draw transmission cu g. 6(a)). The M o the 20-cm lo injected not o at the transmi the fiber, an a eed, this coatin rried out with . Figure 6(b) s infrared camer core of the corded by usin ous transmissi behavior of a n e review). m of the hollow co rum of a 20 cm-lon crobolometer cam experimenta overed by the a e element meth HC-MOF cro ersion and loss e propagation solved numeri ased on the G he associated m s [35,36]  As depicted i defined here bands observe The same con Fig. 7(b). The not present in bands, the usu must be consi considered as function of th this core-loca loss increase.  Fig. 7(a), th as the core of ed experimenta nclusion can be e transmission b n the simulated ual guidance m idered [29]. In invariant alon he wavelength, alized mode co . Usually, the ween the main c But, in our c the transmissio h the initially t modes (associ ue to the fabric uently, couple In this frame e fiber with its ding the lossy fundamental m  Fig. 6(a) or m ( Fig. 7(b)).In uch HC-MOF, case, it is well e losses of the fu oss regions are regions of the ction is optim modes and the mple loss analy ast for three r n, (II) it is not olarization axi the fiber is not y [33,41,42] i ly the fundam ating must be e partially loca hese modes are hollow core fib that it has not uggests that th thermore, these n spectrum. 6 not been erized yet ed by the ot explain and 7(d)), nsmission Fig. 6(a)). shown in ssion) are nce of the ure MOF, ber can be udied as a ons where ble for the e smallest at specific xplain the er profile y the two from each opagation observed he crossgher order g but still oefficients A c c e p t e d M a n u s c r i p t between them fundamental m localized mod fiber core wi Nevertheless, modes that st transmission [33,41,42], in ones) between formula used the ones give field and not i The sum of fundamental provided in pa Practically, fo to ensure that the transverse the first 14 co these transver longitudinal longitudinal o m and the funda mode and ther de couples onl ith effective l if the fundam trongly couple will vanish at n the coupled m n the modes m to compute th en in [41] with in the permittiv the modulus mode (m = 0 anel Fig. 7(a)  c c e p t e d M a n u s c r i p t imaginary parts of their effective indices are above 10 -3 (while the imaginary part of the effective index of the fundamental mode is typically in the [10 -6 -10 -5 ] range). It is worth to mention that only the results for the transverse coefficient sum for the modes are shown in Fig. 8(a), because like for most fibers, the longitudinal coefficients are typically one order of magnitude smaller than the transverse ones. As illustrated Figs. 8(a) and 8(b), the transmission peaks correspond to wavelength ranges with a smaller sum of these modulus coefficients between the fundamental core mode and other modes (highlighted spectrum part in Fig. 8) while low transmission regions originate from a large sum of the coupling coefficients to higher-order modes. Since the longitudinal variations of the fiber cross-section are not available, the coupled mode theory-based analysis cannot be taken further. Thus, this limitation can explain the small differences between the central wavelength positions of the computed transmission bands and the measured ones. Nevertheless, the negative curvature fiber guidance mechanism with the localization of its transmission bands is demonstrated.

Conclusion
In conclusion, the proof of concept of 3D printing of chalcogenide glass preforms has been demonstrated and such preforms have been drawn into guiding optical fibers (single index fiber and hollow core fiber). This first ever chalcogenide "printed" hollow core fiber shows significant light transmission bands in the 4-12 µm window. The spectral positions of the most intense bands are confirmed by numerical simulations, considering the recorded fiber cross-section. The important optical losses should strongly be improved by using high optical quality raw glasses, by printing the preforms under controlled atmosphere, and by a better control of the geometrical parameters during fiber drawing. Those results open a new way for the elaboration of chalcogenide and other soft-glass based components such as optical fibers, especially hollow-core MOFs, and also types of other waveguides or lenses.

Funding
This work was funded in part by the European Union through the European Regional Development Fund (ERDF), the Ministry of Higher Education and Research, the French region of Brittany, Rennes Métropole, the French National Research agency and DGA (grant ANR ASTRID DGA FOM-IR-2-20).