1. Introduction

The integration of electronic and optoelectronic architectures within thin and flexible fibers is bringing novel opportunities for smart sensors [1,2], biological probes [3], energy harvesting [4,5] and advanced textile [6]. Such multi-material fibers can be fabricated using high pressure chemical vapor deposition of semiconductors in the internal channels of glass capillaries [7]. In this method, the as-deposited semiconductors, such as silicon or germanium, exhibit an amorphous structure post-deposition [8,9]. An annealing step is therefore required to induce crystallization and comply with the microstructure requirements of the targeted functionality [8–12]. Another strategy relies on the well-established thermal drawing of macroscopic preforms that already integrate the different materials at prescribed positions [1]. This approach is highly scalable and allows for the integration of complex structures uniformly distributed along kilometers of fiber length. It has been shown that the same high melting point semiconducting materials such as Si or Ge can be co-drawn within a silica matrix [13,14]. The core of the fiber, during the drawing, crystallizes into a polycrystalline microstructure with undesired grain sizes, crystallographic orientation and segregation [15,16]. The use of a high energy light source is then demonstrated to modify the microstructure of the semiconductor core [17–20]. Another class of semiconducting materials, i.e. selenium rich chalcogenide glasses, has been drawn within polymer fibers. Compared to their silica counterparts, polymer fibers allow for low temperature processing, robust mechanical properties, and the ability to simply interface semiconductors with electrodes in a variety of complex architectures [21, 22]. Moreover, the pure chalcogen element selenium (Se) exhibits excellent photosensitive properties [23] exploited in xerography, ultrasensitive imaging tubes [24], chemical sensors [25], as well as photodiodes [4]. Thus far however, the optical, electronic and optoelectronic performances of these fibers have been limited due to the amorphous state of the chalcogenide semiconductor in the as-drawn fiber. One strategy for improving the performance of the device is to control the geometry of the active material in the fiber. It was in particular shown that going from a bulk glass solid core to a thin film configuration led to a significant improvement in photosensitivity of the device [26]. Another approach relies on the post-drawing annealing of the amorphous semiconductor. So far however, the microstructure has been poorly controlled in terms of phase, grain size, orientation and crystallization depth [27]. From an optical point of view, a grain boundary accumulates impurities and is generally detrimental for light transmission [20]. From an electronic point of view, grain boundaries serve as charge carrier recombination centers which generally impair charge transport and limit optoelectronic properties [28]. If the microstructure of the semiconducting element can be controlled to exhibit large grain sizes and ideally single crystal domains between the connecting electrodes, a drastic increase of performance is expected.

In this study, we investigate the post-drawing furnace-based and laser-based crystallization schemes of an electrically addressed Se core within a multi-material fiber. All the crystallization treatments were performed at the fiber tip. It facilitates the characterization of the resulting microstructure, compared to crystallization along the fiber through the cladding. Note that the investigation and conclusions we draw would very well apply to crystallization procedures along the fiber length. Moreover, thin, flexible and long fibers with a functional tip are promising for many applications, for example, in remote optical sensing [29], fluorescence imaging or biosensing [23, 29], and nanophotonics [29]. We demonstrate that the laser-based approach enables a combination of significantly larger grain size and a better control over the crystallization depth, which allows for orders of magnitude better optoelectronic properties, compared to the furnace-based approach. Thanks to the built-in electrodes, we highlight the key role of grain size and crystallization depth via a systematic study of the interplay between the microstructure and optoelectronic properties. In particular, we established a method to extract the crystallization depth via a nondestructive approach. We then demonstrate that laser treatment allowed us to fabricate a fiber device with larger grains and hence less grain boundaries in the direction of the photocurrent, resulting in significantly better optoelectronic properties. We also show that the laser-based treatment enables not only the precise location of crystallized domains on the fiber cross-section, or potentially also along its axis, but also to control the crystallization depth inside the material. This depth can be engineered to be of the order of magnitude of the photon penetration depth, hence optimizing light collection and charge extraction, while minimizing dark current, resulting in high sensitivity.

2. Experimental methods

The fiber in this study was made by the thermal drawing technique, as described in [30]. In a polysulfone (PSU) slab we cut out two grooves where carbon-loaded polycarbonate (CPC) and Sn₉₁Zn₉ electrodes were positioned. Between these two PSU slabs we introduce a thinner one that encapsulates a Se plate. In this way, the Se is completely encapsulated between the PSU and the CPC, preventing any leakage of molten Se during drawing. Subsequently, the stack was consolidated in vacuum at 215 °C for 30 minutes. The composite structure was then drawn at 285 °C, a temperature above the glass transition temperature of PSU and CPC and the melting point of Se, into a fiber more than 100 m in length.

All the scanning electron microscopy (SEM) samples were coated with 10 nm carbon film. The SEM images were taken with a Zeiss Merlin field emission SEM (Zeiss, Göttingen, Germany) equipped with a GEMINI II column operating at 2.0 kV with a probe current 150 pA. The In-Lens annular detector allowed for high resolution secondary electrons imaging at all magnifications. Transmission electron microscope (TEM) samples were prepared by embedding them in epoxy resin followed by sectioning thin slices (60 nm) using ultramicrotomy (diamond knife) which were transferred on a carbon/Cu grid supports (300 Mesh). The TEM images and selected area electron diffraction (SAED) were taken using a Talos F200X operating at 200 kV. For the samples subject to differential scanning calorimetry (DSC) and X-ray diffraction characterizations, a ~0.9 mm part from the fiber tip was cut before the surrounding materials−PSU, metal and CPC were dissolved away by N, N-dimethylacetamide. The harvested bulk Se was then grinded into powders for the DSC and XRD characterization. In order to have enough signals, we collected the fully crystalline part from many fiber tips. All the DSC curves were taken using DSC 8000 from Perkin Elmer. The Electron BackScatter Diffraction (EBSD) maps were acquired at 20 kV on a FEG-SEM XLF30 (FEI) equipped with a Nordlys II camera; they were treated with the Aztec software (Oxford Instruments).

To only crystallize the fiber tip, the fiber was placed perpendicularly with respect to the heat source of the hot plate and the laser beam during the annealing processes. The electrical and photoresponse characteristics were measured using a Keithley 6517B under dark and illuminated conditions. A SuperK Extreme from NKT Photonics with tunable wavelength between 410 and 830 nm was used as light source. The optical laser power was adjusted using an afocal lens system and was measured using a laser power meter (Thorlabs PM 100D). The laser beam used to characterize the optoelectronic properties of the devices was incoming perpendicularly to the fiber cross-section (hence parallel to the fiber axis). The beam size was set to be the same as the size of the larger side of the rectangular Se domain in the fiber cross-section. All measurements were performed in air and at room temperature.

3. Results and discussions

A schematic of the preform to fiber architecture of the structure described in the experimental section is shown in Fig. 1(a). The Se core is in intimate contact with the CPC nanocomposite electrodes. The Sn₉₁Zn₉ eutectic alloy in contact with CPC ensures a good axial conductivity and a straightforward electrical contact to external circuits. The transparent PSU cladding ensures a good encapsulation and mechanical support of the whole assembly. Figure 1(b) illustrates the scalability of the process where hundreds of meters of flexible and functional light weight fibers are produced in a single draw. The optical photograph of the fiber cross-section shows that the rectangular Se domain is sandwiched between CPC electrodes, with shape and aspect ratio unchanged from those of the preform (Fig. 1(c)). The DSC and the diffuse ring diffraction pattern of the selected-area electron diffraction (SAED) from TEM characterization on the Se harvested from the as-drawn fiber indicate that Se is amorphous after drawing (Fig. 1(d)). The crystallization temperature is higher than that in the literature [31], which may be due to stress-induced enhancement of the thermal stability of amorphous Se [32].

 

Fig. 1 (a) Schematic of the fiber drawing process; (b) Photograph of the preform and of drawn fibers of tens of meters showing their robustness and flexibility; (c) Optical photograph of the cross-section of the as-drawn fiber; (d) DSC (heating rate of 10 K/min) curve and SAED of the as-drawn Se. The heat absorbed during the melting is nearly identical with the heat released during the crystallization .

Download Full Size | PDF

The electrical and optoelectronic properties of amorphous materials suffer from short carrier lifetime, low carrier mobility and short carrier diffusion length induced by the disordered structure. Even though the carrier density in amorphous Se is increased when the material is under illumination, the abundant defects acting as recombination centers still severely impair charge carrier drift, resulting in a small photocurrent. In order to improve the performance of the device, we first subjected the amorphous Se embedded in the fiber to an annealing treatment. To facilitate the characterization of the resulting microstructure, we induced and investigate dthe crystallization at the fiber tip rather than along the fiber through the cladding. We established a protocol where we heat up the tip of the fiber by placing it vertically on a hot plate at 150 °C, above the crystallization temperature, for 5 minutes. The DSC, SAED as well as X-ray diffraction characterizations on the Se in the annealed fiber prove that Se becomes polycrystalline, as indicated in Fig. 2(a) and 2(b). The SAED and X-ray diffraction can be indexed as the trigonal phase. We then characterized the optoelectronic properties of the annealed fiber by measuring the photocurrent of the device at a bias between −10 and 10 V, under different illumination powers. The linear form of the I-V curve in Fig. 2(c) reveals a good Ohmic contact between the crystalline Se and the CPC electrodes. The photocurrent at a given voltage is raised by increasing the power of the incident light, as it increases the number of photo-generated carriers. The photoresponsivity of the annealed fiber device is therefore two to three orders of magnitude higher than the as-drawn fiber in which Se is amorphous, for a wide range of incident powers (Fig. 2(d)). However, compared to the dark current, the photocurrent of the in-fiber device remains rather small, as shown in Fig. 2(c). Such a large dark current is due to a large crystallization depth along the fiber axis which increases the area over which the current density is integrated.

 

Fig. 2 (a) DSC (heating rate of 10 K/min) curve and the insert SAED of Se heat treated on a hot plate; (b) X-ray diffraction of the annealed Se; (c) I-V curves of the device versus light power at λ = 532 nm; (d) Photoresponsivity of the device versus light power at λ = 532 nm and the bias of 10 V.

Download Full Size | PDF

To assess the region over which crystallization has occured, we developed a non-destructive method that relies on the local photo-response of the in-fiber device. As the fiber tip is placed on the hot plate, heat diffuses and a temperature gradient is generated along the fiber axis. The Se domain at the fiber tip was heated at a higher temperature (~150 °C) compared to the Se domain away from the tip. We assume Se became fully crystalline at the fiber tip and then partially crystalline in the region away from the hot plate. The different photoconductivity of the amorphous and annealed Se domains allows us to extract the crystallization depth of Se along the fiber axis, as we scan it with a laser beam this time along the fiber length. Indeed, thanks to the transparency of PSU cladding, photons can reach the Se core when the fiber is illuminated from its side between the CPC electrodes. The photocurrent generated was then recorded as the laser beam of 3 mm in diameter (CPS532 Laser Diode, Thorlabs [33] The wavelength is 532 nm and the power is 4.81 mW) was scanning the fiber along its axis with steps of 100 micrometers. Figure 3(a) shows a schematic of the process, and in Fig. 3(b) we plotted the current versus the position of the right edge of the laser beam (i.e. at 0 mm the beam does not cover the fiber). We can then divide the current measurement in different stages: in stage I, the current increases linearly as the beam illuminates an increasing length of the fully crystallized Se that has the same photo-conductivity. At around 0.9 mm the slope changes because the beam now also illuminates a partially crystallized region, with a lower photoconductivity. As the beam progresses further, the crystalline volume fraction of the partially crystalline part becomes smaller when the position is furtheraway from the fiber tip, leading to a reduced increase rate of current in stage II. Once all the crystalline part is illuminated, the current reaches a maximum. As we continue scanning, the current remains roughly stable in stage III when the amorphous part is being illuminated, as it contributes a very small photocurrent. Subsequently, the current decreases linearly because the fully crystalline part starts to move outside of the beam, as shown in stage IV. The current then reduces with a rate similar to stage I until all the partially crystalline part moves out of the beam (stage V), corresponding to a similar width of fully crystalline part of around 0.9 mm. Finally, the current keeps the same low value when only an amorphous domain intercepts with the beam (stage VI). We can use this curve to extract the length of fully and partially crystalline parts. They correspond to 0.9 mm and 1.3 mm, respectively, for the particular heat treatment performed. The axial length of the amorphous domain in the beam in stage III can be estimated to be 0.7 mm. The total length of the fiber in the beam at the end of stage III is therefore 2.9 mm that matches with the effective beam size of the laser.

 

Fig. 3 (a) The schematic of the fiber advancing into the laser beam. The green dot represents the laser beam while the fiber is depicted in red. The dark red is the fully crystallized part (left). The lighter red in the middle represents the partially crystallized section, and the light pink part is the area that remained amorphous; (b) Photocurrent versus the length of the fiber irradiated by the laser beam, as the beam scans the fiber from left to right. The voltage was set to 6 V.

Download Full Size | PDF

To validate this indirect and non-destructive approach, the fiber was cut in the longitudinal direction by ultramicrotomy. The optical photograph of the longitudinal section of the fiber is shown in Fig. 4(a). The left and right edges of the sample are not smooth because the force applied on the diamond knife is only uniform in the center of the blade but not on the sides. It is apparent that the crystalline region consists of fully and partially crystalline regions. The SEM image of region 1 (see Fig. 4(a)) reveals the Se is fully crystalline with the typical spherulite microstructure of crystalline Se (Fig. 4(b)-1). The insert bright field TEM image further demonstrates the crystallinity and indicates the grain size of Se in this region is around 100 nm. Figure 4(b)-2 and (b)-3 show the SEM images of region 2 made of both fully and partially crystalline parts, and region 3−partially crystalline part, respectively. The insert bright field TEM image in Fig. 4(b)-3 of the Se in this region indicates Se is partially crystalline and the grain size is tens of nanometers. Region 4 is far away from the fiber tip and the SEM image (Fig. 4(b)-4) shows that Se on the right in this region remains amorphous. The grain size of Se in this work agrees well with that of crystalline Se fabricated by a similar approach [34,35]. From this SEM analysis we can extract the lengths of the fully and partially crystalline Se to be 0.89 and 1.43 mm, respectively. This matches well with the values of 0.9 mm and 1.3 mm obtained from the previously described non-destructive method.

 

Fig. 4 (a) Optical photograph of the longitudinal section of the fiber; (b) SEM micrographs of the regions 1 to 4 in (a). The inserts in Fig. 4(b)-1 and Fig. 4(b)-3 are right field TEM images.

Download Full Size | PDF

The nanoscale grain size of the polycrystalline Se after annealing corresponds to a large number of grain boundaries. These can act as charge carrier recombination centers which is detrimental to the electronic and optoelectronic properties of the material [28,36]. To improve the responsivity of the device, a better control over the microstructure with larger grains and potentially single crystal domains between the electrodes is required. For a good device sensitivity, it is also important that the dark current and associated noise be minimal. Ideally, the crystallization depth would be just enough to absorb as much light as possible, that is the order of the penetration depth of the wavelength considered. Beyond this depth, crystalline domains would only participate to noise, which is detrimental to the sensitivity. In other words, the photocurrent remains the same while the dark current can be significantly reduced if the crystallization depth can be controlled to be of the order of the photon penetration depth, giving rise to a high responsivity and sensitivity [26]. Regardless of the temperature and time of annealing, we have observed that the resulting crystallization domains have a microstructure with rather small grains, and a depth hard to control and of the order of several tens of micrometers. We therefore turned to a laser-based approach to induce crystallization, used for Si or Ge materials, but never for Se-rich semiconductors.

The fiber cross-section was illuminated by a laser beam (λ = 490 nm) with a power of 1.5 mW and a diameter of 160 μm for up to 3 hours. We chose these parameters as they result in a drastically better control over microstructure and performance compared to the furnace-based crystallization scheme described above, highlighting the comparison between the two methods. An in-depth report of the influence of these various parameters on the resulting microstructure and optoelectronic properties is beyond the scope of the current study. The cross sectional optical photograph in Fig. 5(a) (left) shows that the Se region illuminated by the laser beam is crystalline with many grains, while the region outside the beam remains amorphous. The microstructure was further investigated by EBSD. The corresponding EBSD map of the crystalline Se is shown in Fig. 5(a) (middle), displaying a polycrystalline structure consistent with the optical microscope imaging. The histogram in Fig. 5(a) (right) indicates an average grain size of 4.5 μm in the cross-section that is orders of magnitude larger than the grain size of Se crystallized with the furnace-based approach described above. In order to study the crystallization depth and microstructure along the fiber axis, we used a combination of TEM and electron diffraction technique. We prepared TEM lamella cut out of the fiber in the longitudinal direction with ultramicrotomy, as previously described. The bright field TEM in Fig. 5(b)-1 reveals that the crystallization depth is 450 nm at the edge and 700 nm in the center of the crystalline region, which can be explained by the Gaussian-like wave-front of the laser beam. The SAED pattern (Fig. 5(b)-2) taken from region A of Fig. 5(b)-1 clearly shows a polycrystalline structure with different orientations and the clustered diffraction spots suggest that the grain boundaries are mostly low-angle type; Region B on the other hand only contains a few grains, as indicated by the bright field TEM image in Fig. 5(b)-3. Indeed, the SAED pattern taken from region C of Fig. 5b-3 shows a large single crystal (Fig. 5(b)-4). Using an objective aperture in the back focal plane to collect the diffracted electrons that contribute to the spots of the SAED pattern, as shown by the squares in Fig. 5(b)-2, dark field TEM images, formed by diffracted beam 1 to 4, are taken in Fig. 5(b)-5-5(b)-8 which allow us to study the grain size and morphology. It can be seen that the grain size is a few hundreds of nanometers in the fiber axis direction, and the morphology of each grain is irregular. We hence could achieve a microstructure with grains that are large in the direction perpendicular to the electrodes, which reduces the number of grain boundaries, and small in the fiber axis direction, reducing the dark current.

 

Fig. 5 (a) Left: cross sectional optical photograph of the fiber heated with laser; Middle: corresponding EBSD map on the crystalline Se in (a) (Here we have chosen a IPFz representation which encodes the angle between the crystallographic c-axis and sample Z-axis (the green and blue colors mean that the c-axis is at 90° with respect to the Z axis, the grains orientated at 0° would be in red, but there are no such grains in the sample)); Right: the histogram of grain size distribution; (b) TEM micrographs and SAED patterns of the crystalline Se in the longitudinal section; (c) Left: comparison of photoresponsivity of laser-induced crystallization fiber and hot plate-induced crystallization fiber. Right: ratio of I_ph/I_dark versus power for the same fibers.

Download Full Size | PDF

Such an optimized microstructure is expected to exhibit improved performances. We then characterized the optoelectronic properties of the device by measuring the current under different powers and at fixed wavelength (λ = 532 nm), as previously done. In Fig. 5(c) we compare the photoresponsivity and the ratio I_ph/I_dark measured at a bias of 10 V as a function of power, for the two different devices. As expected, the fiber crystallized with the laser-based approach exhibits a much higher photoresponsivity in a wide range of powers. The ratio I_ph/I_dark of the fiber crystallized with the laser beam is several orders of magnitude higher, compared with the fiber crystallized with hot plate. These performances are excellent and compare favorably with reported trigonal Se microtube planar photodetector [37].

4. Conclusions

The in-depth understanding of the interplay between the materials microstructure and device performance could open new opportunities for increasingly sophisticated functionalities in fiber devices. Two annealing approaches were employed to crystallize the amorphous semiconducting core in a multi-material fiber. Regardless of temperature and time, the regular annealing-based method induces a microstructure with nanoscale grains and a large and hard to control crystallization depth. In contrast, the local laser-based approach induces crystallization in a controlled depth and a microstructure with microscale grains. The optoelectronic properties are drastically enhanced, featuring high photoresponsivity and photosensitivity. Further work is underway for the systematic study of the effect of laser power, beam size and exposure time on the microstructure formation. Optoelectronic fibers with a functionalized tip can find several applications in highly sensitive remote detection and sensing, optoelectronic probes, minimally invasive in situ and in vivo bio-compatible probing and imaging of biological tissues. The small cross-section and large aspect ratio of the high-performance fiber indeed allows access to remote and confined environments where a rigid and planar point photodetector is unable to reach. The laser annealing approach demonstrated here also has the potential to be applied in the longitudinal direction of the fiber. This would enable the ability to write, at any location along kilometer-long fibers, devices of controlled microstructure and performance. Possible applications can be envisioned in large area, flexible optoelectronics, energy harvesting systems, and advanced fibers and textiles.