Ligand-Driven and Full-Color-Tunable Fiber Source: Toward Next-Generation Clinic Fiber-Endoscope Tomography with Cellular Resolution.

In many biomedical applications, broad full-color emission is important, especially for wavelengths below 450 nm, which are difficult to cover via supercontinuum generation. Single-crystalline-core sapphires with defect-driven emissions have potential roles in the development of next-generation broadband light sources because their defect centers demonstrate multiple emission bands with tailored ligand fields. However, the inability to realize high quantum yields with high crystallinity by conventional methods hinders the applicability of ultra-broadband emissions. Here, we present how an effective one-step fiber-drawing process, followed by a simple and controllable thermal treatment, enables a low-loss, full-color, and crystal fiber-based generation with substantial color variability. The broad spectrum extends from 330 nm, which is over 50 nm further into the UV region than that in previously reported results. The predicted submicrometer spatial resolutions demonstrate that the defect-ligand fields are potentially beneficial for achieving in vivo cellular tomography. It is also noteworthy that the efficiency of the milliwatt-level full-color generation, with an optical-to-optical efficiency of nearly 5%, is the highest among that of the existing active waveguide schemes. In addition, direct evidence from high-resolution transmission electron microscopy together with electron energy loss spectroscopy and crystal-field ligands reveals an excellent crystalline core, atomically defined core/cladding interfacial roughness, and significant enhancements in new laser-induced electronic defect levels. Our work suggests an inexpensive, facile, and highly scalable route toward achieving cellular-resolution tomographic imaging and represents an important step in the development of endoscope-compatible diagnostic devices.

Introduction

The increasing utility of biophotonics has stimulated the development of clinical optical coherence tomography (OCT) to achieve cellular spatial resolution and, eventually, three-dimensional tissue imaging in vivo.[1−3] This development of a noninvasive label-free technique has resulted in increased demand for endoscope-compatible broadband light sources (BLSs) with efficient luminescence. This urgent work has included exploration of available broad emissions, especially in the full-color (UV to red) region because the axial resolution of OCT is primarily governed by a bandwidth of 3 dB and is inversely governed by the square of the center wavelength of the light source.

Over the past decade, several alternatives have been developed to fabricate bulk-type BLSs, including crystalline (or amorphous)-assisted multiple emission ions,[4−6] photonic crystal fiber generated supercontinuums,[7−9] and high-power superluminescent diodes (SLDs).[10−12] However, each of these approaches presents one or more drawback. For example, the design of multiple active ions in a crystalline host is highly challenging because of the deteriorated crystal quality and, consequently, weak luminescence, which is insufficient to meet practical requirements.[13] In addition, amorphous matrices suffer from poor chemical durability and require stringent processing to suppress cluster formation and crystallization.[14] Second, for emissions below 450 nm, obtaining the aforementioned supercontinuum is difficult because of the large detuning resulting from pulse laser excitation. Additionally, expensive and extremely sophisticated ultrafast lasers become unfavorable when integrating with compact scanning apparatuses, limiting their size and simplicity as well as increasing their cost.

Despite the success of high-quality InGaN growth, white SLDs with a blue InGaN chip combined with a yellow phosphor blended with organic resins have attracted intensive research interest because of their unique properties and potential applications.[15] However, this configuration presents a poor Commission Internationale de l’Eclairage (CIE) chromaticity coordinate because a red light component, typically above 600 nm, is lacking.[16] Moreover, the high working temperature of high-power SLDs would cause severe carbonization at the phosphor/resin interface, resulting in the degradation of illumination and inducing color shifting.[17]

Unlike the above bulk-type examples, the compact, low-cost, and highly heat dissipating fiber-based BLSs are highly suitable for long-range biomedical endoscopic applications and high-capacity telecommunications.[18] These materials feature directional high luminescence and better light collimation, leading to higher coupling efficiencies of light into optical fibers where no traditional light source is attainable. Thus, finding suitable materials with fiber-waveguide and BLS functionalities remains challenging because simple band-to-band emission in a single material host is difficult. To this end, a facile route for achieving fiber-based BLSs covering the full-color range (UV to red) for applications in an OCT system is highly desirable.

Herein, we demonstrated defect-driven full-color broadband emission by engineering a glass-clad sapphire crystalline-core fiber using the laser-heated pedestal growth (LHPG) technique to create new defect levels within the bandgap and conduction band.[19−21] Glass-clad sapphire fibers with and without thermal treatment presented tunable full-color emission when pumped with 325 nm excitation. Unlike heavy particle irradiation[22−26] and thermochemical reaction,[27−29] which are often accompanied by poor crystallinity and insufficient formation of F- and F2-type color centers, leading to a limited bandwidth and low-yield emission, our proposed facile approach can not only provide a range extending over 50 nm deeper into the UV region than that previously achieved and efficient red emission over 600 nm but also can preserve the material’s high crystallinity. The predicted submicrometer spatial resolutions demonstrate that the impact of defect–ligand fields on the relative strength and controlled emitters is potentially beneficial for full-color emission to achieve in vivo ultrahigh resolution OCT. Additionally, comparing the measured full-color generation obtained with defect ligands and examining the high-crystallinity sapphire core by electron energy loss spectroscopy (EELS) confirmed that these extra bandwidths constitute an effect of considerable oxygen and aluminum defect centers, which were effectively facilitated by the inclusion of small amounts of Ti and Cr ions. Equally importantly, we report for the first time the direct atomic observation of the impacts of annealing on the crystallinity of a sapphire crystal fiber. Excellent crystallinity with low propagation loss was observed on a large scale along the core/cladding interface of the annealed samples, benefiting from atomic-scale interfacial roughness, far behind the optical wavelength scale. These results unambiguously demonstrate the applicability of defect-driven emission as an inexpensive, facile, and highly fluorescent route.

Results and Discussion

Impact of Thermal Treatments on the Nanostructure

Figure a shows the schematics of the procedure by which borosilicate glass-clad sapphire crystalline-core fibers were grown in a reducing environment that favors Ti3+ over Ti4+ ions.[30]Figure b shows the scanning electron microscopy (SEM) side-view image of an LHPG-grown 40 μm diameter sapphire crystalline core with a uniform diameter and smooth surface along the fiber. A smooth waveguide surface ensures low-loss propagation because scattering loss stems from random fluctuations of the core diameter and could become the dominant attenuation factor.[31−33] The bright-field end view of an as-grown glass-clad sapphire crystalline-core fiber is shown in Figure c. It should be noted that dealing with this type of “heterostructure” material is an issue in fiber fabrication. The hardness of the crystalline core and that of the glass cladding are quite different. We have developed a face preparation process for the hybrid crystalline-core glass-clad fibers that comprises grinding (using SiC with grain sizes ranging from 10 to 3 μm) followed by fine polishing (using diamond with grain sizes ranging from 3 to 0.1 μm). When using this preparation procedure, the surface processing between the core and cladding can be controlled, with a depth variation of less than 15 nm. Figure d presents the corresponding compositional distributions of Al and Si. Energy-dispersive X-ray spectroscopy (EDX) results indicate that the 40 μm sapphire core bonds closely to the surrounding borosilicate glass cladding, ensuring effective optical confinement. In addition, the EDX mappings clearly delineate hexagon-like core signatures, corresponding to c axis hexagonal close-packed (HCP) structures, as illustrated in Figure e.

Figure 1

Fabrication of borosilicate glass-clad sapphire crystalline-core fibers. (a) Schematic of the growth of a glass-clad sapphire crystalline-core fiber. (b) Side-view image of an LHPG-grown 40 μm sapphire crystalline core. (c) End-face images of a glass-clad sapphire crystalline-core fiber. (d) EDX mappings for Al (green) and Si (red), taken from the square region in (c). (e) Crystallographic orientation of the c axis of sapphire.

To directly examine the impact of thermal treatments on the interfacial nanostructure and luminescent efficiency, we investigated the nanostructure at the core/cladding interface. A high-resolution transmission electron microscopy (HRTEM) image (Figure ) of a representative sample without annealing shows atomically abrupt interfaces, arranged in a close-packed manner and in the [21̅1̅0] orientation. The interface plane is nearly parallel to (011̅0). The lattice image (Figure a) from the local area of this sample presents further dark tweed contrast and planar defects, as indicated by the two-dimensional (2D) Fourier transform of the [21̅1̅0] zone axis (inset of Figure a), which is streaky along [0002]. The inverse Fourier transform, shown in Figure b, confirms that abundant nanometer-sized faults occur in the {0002} planes and propagate in the ⟨011̅0⟩ direction, as denoted by the red lines. Figure c shows a Fourier-filtered image of a thin region within the sapphire core in which the misfit dislocations are marked by ⊥. In fact, the rapid heating and cooling that occur during such an LHPG process could cause considerable residual stress. The employed CO2 laser heating may result in a temperature of over 2000 °C, as evidenced by the change of the sapphire crystal into a melt, observed using yttrium–aluminum–garnet (YAG) crystal fibers.[34] Thus, defect formation under the influence of laser heating during LHPG growth, as occurs during pulsed laser ablation, can result in faults, dislocations, and twins.[35,36] Nevertheless, because of the transformed sapphire crystalline core, which has a defective structure and ionic vacancies, these defect centers could act as color emitters for broadband generation, as has been discussed later. Recently, dislocations in the c plane of GaN grown on sapphire were suggested to be significantly reduced from an initially random dispersed state during high-temperature annealing.[37−39] High-temperature treatments have been shown to be effective in recovering the corundum-type structural misorientation introduced during crystal growth, which is relevant in the present case of III-nitrides.[40,41] In addition, surface roughness has been shown to significantly influence key high-brightness light-emitting diode (LED) wafering. Although annealing treatment has been correlated to effective charge transfer (Ti4+ to Ti3+) in bulk sapphire for efficient laser actions,[42,43] the effects of annealing on glass-clad sapphire crystal fibers have not been studied yet. In this work, the impacts of high-temperature treatment were investigated to understand how the nanostructure of the core/cladding interface is influenced by annealing and to elucidate the process of defect reduction with respect to the fluorescent properties for potential OCT applications.

Figure 2

(a) HRTEM lattice image of a sapphire crystalline-core/borosilicate glass-clad interface before annealing in the [21̅1̅0] zone axis, in which the planar defects are clearly observed. The double diffractions are denoted by D in the inset. (b) 2D Fourier transforms of the square region in (a). (c) Inverse Fourier transforms of (b) showing the misfit dislocation (marked by ⊥), with extra half-planes parallel to the (011̅0) edge-on and abundant faults along (0002). (d) Lattice image of the crystalline-core/glass-clad interface showing a γ-Al2O3 nanocrystallite with a local crystallographic relationship to the sapphire core. Insets: magnified image with labeled d-spacings and a 2D Fourier transform of the square region in (d), showing oriented attachment in the zone axis [1̅11]γ∥[011̅2]α. (e) 2D Fourier transforms of the square region in (d). (f) Inverse Fourier transforms of (e) showing the orientation relationship (01̅1̅)γ∥(202̅1̅)α and (1̅01̅)γ∥(011̅1̅)α, with misfit dislocations at the γ-Al2O3/α-Al2O3 interface.

In Figure d, the HRTEM observations also indicate that the core/cladding interface consists primarily of a corundum phase with some nanocrystallites of spinel-like γ-Al2O3, presumably because of glass cladding-induced interdiffusion. The γ-to-α transformation involves a change in the face-centered cubic (FCC)–HCP oxygen framework and is commonly thermally activated in alumina at temperatures above 500 °C.[44−46] In our case, however, the transformation of the α-form to the metastable γ-Al2O3 occurs via a different mechanism, that is, sapphire/silica interdiffusion. Therefore, glass cladding by laser irradiation in the LHPG system would introduce nonstoichiometric numbers of Al, Si, and O atoms. More Si and O atoms will be displaced than Al atoms because of the weak bonding of SiO2. Further, the displacement energies of Si and O atoms in silica are estimated to be 20 and 10 eV, respectively, assuming an Si–O bond energy of ∼5 eV.[47] In contrast, the displacement energies of Al and O atoms in sapphire are 20 and 65 eV,[48] respectively. As a result, during irradiation, α-to-γ transformation occurs via inward diffusion of excess Si and O atoms from the borosilicate glass cladding. In fact, all of the Al3+ in α-Al2O3 are in octahedral sites, whereas γ-Al2O3 contains a random distribution of Al3+ and vacancies in the tetrahedral and octahedral sites in the spinel structure. In contrast, a small amount of Si signal, detected by point-count EDX, was collected at a depth of 10 nm across the interface by HRTEM; this signal is considered to be nearly inevitable when fabricating such glass-clad/crystalline-core fiber waveguides. According to the HRTEM observations, γ-Al2O3 does not form distinguishable clusters but rather appears to exhibit oriented attachment to the sapphire core, as shown in Figure d. The electron-diffraction pattern and magnified lattice fringe in the inset of Figure d allow the measurement of interplanar spacings of 0.251 nm for (202)γ, 0.219 nm for (02̅2̅)γ, 0.342 nm for (011̅1̅)α, and 0.218 nm for (202̅1̅)α. (Note: γ-Al2O3 and α-Al2O3 [i.e., sapphire] are denoted by γ and α, respectively; the diffractions of γ and α are circled in cyan and red, respectively.) Figure e shows the identified nanocrystallites with close attachment to γ-Al2O3 and α-Al2O3 (denoted as γ and α, respectively), following a crystallographic relationship (01̅1̅)γ∥(202̅1̅)α and (1̅01̅)γ∥(011̅1̅)α, which were also identified by the high-resolution lattice image and 2D and inverse Fourier transforms in the [1̅11]γ∥[011̅2]α zone axis. Data from the local regions of the γ-Al2O3 nanocrystallites processed in Figure f indicate the existence of numerous misfit dislocations (marked by ⊥), with the half-plane parallel to the (1̅01̅)γ edge-on along the γ-Al2O3/α-Al2O3 interface. The interfacial misfit dislocations occur periodically and are evenly spaced in nearly all 2(011̅1̅)α planes. Although spinel-like structures are present at the core/cladding interface, their abundance is negligible because the size of these nanocrystallites is far behind the wavelength of light when the signal passes through the crystalline core of the fiber. Specifically, the scattering loss of nanocrystallites shows an a6 dependence (where a is the radius of the nanocrystallite).[49]

Figure a,b shows the micrographs and corresponding 2D Fourier-filtered images of the processed sample after annealing recorded along the sapphire [011̅1] zone axis, respectively. The interfacial selected-area electron-diffraction (SAED) pattern is also shown in the inset of Figure a, which reveals a high crystallinity and the absence of planar defects. Significantly, excellent large-scale crystalline quality of the sapphire core was achieved along the interface between the annealed samples, suggesting that half-plane dislocation and planar defects disappeared. The defect-free sapphire crystalline core was obtained from prolonged thermal treatment at 1650 °C for 3 h. Careful examination of around 30 TEM images showed that the defect densities of the fibers with and without annealing are estimated to be ∼10–6 #/nm2 and ∼10–2 #/nm2, respectively. On the basis of these data (Figures S1 and S2 in the Supporting Information (SI)), we can conclude that the effectiveness of high-temperature annealing is related to the reduction of the interfacial root-mean-square roughness (0.5350 vs 0.1933 nm) and correlation length (2.4938 vs 1.0660 nm). Both small interfacial roughness and correlation length favor achieving very low propagation losses within the glass-clad/crystalline-core fibers. This is because the interplay of light with roughness increases as the interfacial variation increases, and scattering increases when the correlation length is comparable to the wavelength. In addition, scattering loss α can be deduced on the basis of a theoretical model in terms of interfacial roughness σ and correlation length Lc, developed by Payne and Lacey[50]where k0, d, and n1 are the free-space wave vector, core radius, and core index, respectively. Function g is primarily determined by the waveguide geometry, whereas f is determined by σ and Lc. The extracted scattering losses for our sapphire crystalline-core fiber with interfacial roughness values of 0.5350 (nonannealed) and 0.1933 nm (annealed) are calculated to be 1.236 and 0.069 dB/cm, respectively. This nearly atomically smooth core/cladding interface represents an order of magnitude improvement in roughness and hence a low propagation loss compared to that in the traditional ultrafast-laser-inscribed waveguide, which exhibits typical values from 1 dB/cm to a few dB/cm.[31] These results highlight the clear superiority of our LHPG-based fiber-drawing technique over other techniques, such as ultrafast-laser inscription. It is also noteworthy that unlike that in most ceramic-prepared sapphire fibers that are polycrystalline and rather void after annealing, and[51−53] in some cases contain intragranular amorphous phases at grain boundaries if incomplete crystallization occurs, we find no evidence of an amorphous inclusion nor any impurity precipitation within the core region. By contrast, a single-crystalline-core fiber exhibits a relatively small scattering loss because of the lack of grain boundaries and pores that reduce optical quality.

Figure 3

(a) HRTEM image of the interface between the sapphire crystalline core and borosilicate glass cladding after annealing, which shows a defect-free interior of the sapphire core and the (21̅1̅0) interface viewed edge-on in the [011̅1] zone axis. (b) The corresponding Fourier-filtered image confirms the effectiveness of the dislocation reduction during annealing and hence (c) sapphire cores with high crystallinity are obtained.

Ligand-Driven Full-Color Evolution

The impact of annealing treatments on the full-color evolution from the starting material to the as-grown glass-clad sapphire core fibers is shown in Figure a. Strong, sharp peaks superimposed on the broadband emission are distinctly observed near the excitation wavelength. These peaks can be assigned to the spontaneous Raman scattering modes of the sapphire crystalline core, suggesting the effectiveness of waveguide confinement with high crystallinity and low propagation loss. Both the fluorescent and Raman spectra were probed by a 325 nm laser in the center of the core region at room temperature. Figure a shows the spectra of step 2, denoted by pink (nonannealed) and blue (annealed) curves, with three broad fluorescent bands at 330–400, 400–600, and 600–650 nm. These differ significantly from those of the starting material (the orange curve in Figure a) and the as-grown 40 μm sapphire core (the cyan curve in Figure a), which exhibit commonly observed green-to-red emissions with weak blue and UV–visible parts in the sapphire crystals[27,54,55] because of insufficient oxygen monovacancies, as discussed in detail below. The chromaticity coordinates of these four studied samples are depicted in the 1931 CIE diagram (Figure b), indicating that the emission colors from the starting material to the as-grown glass-clad sapphire crystalline-core fiber can be tuned from orange to yellow-green by step 1 and then to white or green by step 2 (annealing). The corresponding CIE coordinates are (0.413, 0.402), (0.316, 0.424), (0.287, 0.333), and (0.286, 0.406), respectively. Notably, the color coordinate of the nonannealed sample (nearly white emission) is located in the vicinity of the pure white point of (0.333, 0.333), whereas the annealed sample shows intense green light and could be useful in improving the radiant efficiency of the typical colored LEDs that lack green emitters.[56] The above results clearly demonstrate the versatile applicability of this method as a facile route and indicate a different physical color center-formation mechanism during the LHPG-based fiber-drawing processes.

Figure 4

(a) Representative 325 nm excited full-color generations of the starting material (orange), LHPG-grown 40 μm sapphire core (cyan), and glass-clad sapphire crystalline-core fiber with (blue) and without (pink) annealing, with insets showing the luminescent image from the fiber end faces. (b) Corresponding color evolutions of (a) during the growth and thermal treatment of the glass-clad sapphire crystalline-core fibers. The predicted axial resolutions in air of (c) the nonannealed and (d) Ar-annealed samples estimated from (a) are approximately 0.97 and 1.03 μm, respectively; the power envelop spectra are shown by pink curves. This submicrometer spatial resolution demonstrates enormous potential for achieving in vivo tomographic imaging with ultrahigh resolution. (e) Full-color emission as a function of excitation, showing a maximum output power up to milliwatt order. This result is comparable to that obtained using III–V and III-nitride light-emitting diodes. (f) Picture of (e) coupled to an SMA optical fiber and dispersed by a reflective holographic grating.

More importantly, full-color emission with a 3 dB bandwidth covering the whole UV–visible range has potential as a cellular-resolution OCT BLS, as shown in Figure a. To test this applicability, Figure c,d presents simulated interferometric signals of the nonannealed and Ar-annealed samples estimated from the broadband spectrum in Figure a, respectively. The normalized power envelope, 20 log[H(z)], is calculated via the Hilbert transform, H(z), of the simulated interferometric signal, as shown by the pink curves in Figure c,d. The predicted axial resolutions in air, estimated from the full-color bandwidth, are approximately 0.97 and 1.03 μm, respectively. The predicted submicrometer-scale axial resolution of the ligand-driven full-color fiber source in air is smaller than the corresponding resolutions of OCT images based on conventional broadband gain media (2.2 μm for Ti3+:sapphire[57] and 1.45 μm for Ce3+:YAG[58]) and white-light sources (1.6 μm[59]). The broadband emissions from representative Ti3+:sapphire and Ce3+:YAG crystals are among those used as high axial resolution OCT light sources in the near infrared and visible wavelength ranges, respectively. A variety of imaging techniques, including magnetic resonance imaging (MRI), photoacoustic tomography (PT), and confocal and multiphoton microscopy, have been used for in vivo and ex vivo physiology investigations.[60,61] However, these techniques have inherent limitations, such as poor imaging resolution (MRI and PT) and shallow penetration depth (confocal and multiphoton microscopy). In this regard, OCT finds a niche among these techniques. The imaging resolution of OCT is typically limited to a few micrometers, which is better than that of MRI and PT; however, its imaging depth is greater than that of confocal and multiphoton microscopy. The predicted submicrometer-scale axial resolution of our proposed fiber source is smaller than the corresponding resolution of OCT based on conventional broadband gain media. This ability has been recognized to be important for understanding the dynamic interactions of cellular behaviors at different stages. Probing the response at a cellular level may provide further pharmacokinetic and pharmacodynamic information. Ideally, it is hoped that the desired OCT configuration, consisting of a full-color fiber source with a fast scanning rate, can offer a truly viable architecture for high-resolution tomography. Investigations are currently underway toward the realization of this ultimate goal. Overall, these submicrometer spatial resolutions demonstrate that the impact of defect–ligand fields on the relative strength and controlled emitters is potentially beneficial for full-color emission to achieve in vivo ultrahigh resolution OCT.

Figure e shows the measured and fitted output powers as a function of the 325 nm incident pump power. Figure f shows the corresponding picture of the generated full-color emission coupled to a commercial SMA fiber and dispersed by a reflective holographic grating. The maximum output power of the full-color emission was limited by the available pump power. It is noteworthy that as much as ∼1.2 mW (white light) and ∼1.6 mW (green light) of the output power were achieved, corresponding to optical-to-optical efficiencies of 4.7 and 6.6%, respectively. To our knowledge, these high conversion efficiencies are the highest among conventional glass fibers.[62−64] The high optical-to-optical conversion efficiency can be ascribed to the large emission cross-sections of the color centers produced by the odd-parity phonon vibrations.[65] The emission cross-sections of the F- and F2-type color emitters in the sapphire crystal are 1–2 orders of magnitude higher than those found in typical transition-metal-doped sapphire[66−69] and YAG[21,34,70] wideband gain media (10–23 m2 vs 10–21–10–22 m2). The resulting high conversion efficiency is also the highest reported to date among the existing waveguide-type white light-production techniques, including that obtained using supercontinuum lasers. It is also noteworthy that an increase in the fiber length generally leads to higher losses, as introduced by absorption and propagation losses. However, a longer fiber length is necessary in practice to extract the maximum output power from the sapphire crystalline core, because the gain saturation increases with increasing emission cross-sections for the F- and F2-type color emitters. The nature of these larger emission cross-sections of the color emitters in the sapphire crystalline core ensures a higher output power. Moreover, on the basis of the calculated numerical aperture of 0.89 and solid angle of 2.89 in a 40 μm diameter sapphire core fiber, one can evaluate luminous fluxes of ∼0.8 lm (white light) and ∼1.1 lm (green light) from the milliwatt-order full-color generations. These results lead to a luminance of ∼108 cd/m2 and are comparable to those obtained using III–V and III-nitride light-emitting diodes.[71,72]

Conventional X-ray diffraction (XRD) can give direct information about the crystalline quality and size as well as the presence of defects in the crystal. This is especially true for the large single crystals or polycrystalline materials in bulk forms. Although XRD is a promising technique for the characterization of the overall degree of crystallinity of bulk samples, it is rather difficult to measure the fiber quality precisely between the core and the cladding with micrometer-scale spatial resolution when using this method. This is because XRD is limited by the beam diameter of the X-ray source, which is typically in the range of a few to a few tens of millimeters. The spatial resolution of XRD is often poor and is further exacerbated when directly examining a micrometer-sized crystal fiber core and cladding. Several previous reports have addressed stress field and crystal quality determination, with submicrometer spatial resolution,[73−77] in bulk samples and waveguide structures, in which it is not possible to use XRD techniques. The spatially resolved nature of Raman microscopy allows direct and quantitative analysis of any type of fiber-waveguide structure. Of course, both Raman spectroscopy and XRD techniques are complementary on different orders of scale. Additionally, in our specific cases, direct probing of the micrometer-sized crystal fiber core by XRD remains challenging because of the presence of the amorphous glass cladding that surrounds the core. Considering the volume difference ratio (1/63) between the crystalline core (radius = 20 μm) and the glass cladding (radius = 160 μm), we expect that the XRD data will have a strong broad background (from the amorphous glass cladding) and remarkably weak XRD peaks (from the crystalline core). Unfortunately, the etching approach[78] would not be practical either because the stress-dependent XRD results obtained after removal of the cladding would be different from those for the original glass-clad case. As an alternative, a relatively reliable and rapid optical testing technique of confocal Raman microscopy[79−81] has been used to investigate both the fiber quality and the residual strain with high spatial resolution, as shown in Figure a,b.

Figure 5

(a) Excited Raman spectra (325 nm) of the starting material (orange), LHPG-grown 40 μm sapphire core (cyan), and glass-clad sapphire crystalline-core fiber with (red) and without (blue) annealing, respectively. Typical peaks are observed, including two A1g peaks associated with five Eg Raman active modes. (b) Close-up view of (a) in the range of 700–800 cm–1, showing that the sapphire crystalline core was subject to tensile strain, as indicated by a red-shifted A1g peak of ∼−1.256 cm–1 in the inset. The full-color luminescence spectra of the (c) nonannealed and (d) Ar-annealed samples fitted by Gaussian showing a laser-induced interconversion between various defect centers.

The details of the Raman signals in Figure a are magnified in Figure b. Seven unambiguous characteristic c axis sapphire peaks can be identified on the basis of group theory, showing the typical two A1g and five Eg Raman active modes.[77] Another important conclusion from Figure a is that the frequency shift of the second Raman peak presents a strain-dependent behavior, as shown in an enlarged plot of the Raman spectra from 700 to 800 cm–1 in Figure b. By fitting of the Raman spectra shown in Figure b, the peak wavenumber of the annealed fiber was determined to be 754.374 cm–1, with a narrowest 3 dB linewidth of 12.32 cm–1. The extracted linewidths for the nonannealed fiber and the bulk source rod are 13.11 cm–1 (754.470 cm–1) and 12.74 cm–1 (755.630 cm–1), respectively. This ∼6.0% reduction in the 3 dB Raman linewidth following subsequent preferential annealing agrees well with the HRTEM results shown in Figures and 3 and also corroborates the reported improvement in fiber core quality.

On the other hand, the blueshift of the second Raman peak in the as-grown glass-clad forms of sapphire relative to the peaks of the starting material and the as-grown 40 μm sapphire core shows that the sapphire matrix experiences considerable strain.[77] The presence of this residual strain is attributed to the thermal expansion coefficient (TEC) mismatch between the sapphire crystalline core and borosilicate glass cladding, which develops when the temperature changes during the LHPG process and remains within the crystal fiber as it cools to room temperature. Considering the average reported value of the strain-induced Raman shift of ∼1.700 cm–1/GPa in bulk sapphire,[77] this difference in the Raman shift of ∼−1.256 cm–1 (from 755.630 to 754.374 cm–1) corresponds to a tensile stress change of ∼−0.74 GPa, as found for typical luminescent Nd:YVO4, Nd:YAG, and Cr:YAG analogs.[34,82,83] In contrast, one can reasonably estimate the thermally induced residual stress, σ, in the sapphire core in terms of Young’s modulus ν, Poisson’s ratio E, core-to-clad volume fraction f, TEC α, and temperature change ΔT using the following relation[84]whereAssuming that the physical properties of the micrometer-sized fiber are similar to those of the millimeter-sized bulk sample, the values of ν, E, and α for the c axis sapphire core are 0.23, 400 GPa, and 4.5 × 10–6 °C–1,[85,86] respectively. Furthermore, f = (20 μm)2/(160 μm)2 = 1/64, and typically, ΔT = 525–25 °C = 500 K (i.e., the transition temperature for borosilicate glass is 525 °C). Additionally, ν, E, and α for the borosilicate cladding are 0.2, 64 GPa, and 3.8 × 10–6 °C–1,[87,88] respectively. On the basis of this theoretical model, the residual stress was calculated to be ∼−0.80 GPa, which compares favorably with the experimentally determined value of ∼−0.74 GPa. Thus, the rapid heating and cooling in this glass cladding process caused significant residual stress between the crystalline core and the glass cladding, which exhibit considerably different TECs. Therefore, the thermally induced tensile stress, reaching values as high as the GPa range, and not the γ-Al2O3-to-sapphire lattice mismatch, induced strain at the core/clad interface. Additionally, Pezzotti et al. reported the piezospectroscopic effect of c plane sapphire to account for the observed strain dependence of F+-type-center emission under application of biaxial stress.[89] This is similar to the present case of glass-clad c axis sapphire crystalline-core fiber with TEC mismatch between the crystalline core and glass cladding. The room-temperature TECs of the sapphire crystalline core and the borosilicate glass cladding are 4.5 × 10–6 and 3.8 × 10–6 °C–1, respectively. When compared with the unclad bulk sapphire rod, this TEC mismatch generates a tensile strain field radially within the core region during the fiber-drawing process. For an isolated Al3+ ion in this structure, the local structural arrangement comprises an Al3+ ion surrounded by six oxygen ions in an octahedral site. This octahedral site may then be distorted by the aforementioned tension along the C3 symmetry axis, which is analogous to the case produced under a controlled strain field in ref (89). (Note that there may be some of planar defects that are not studied in ref (89).) This may illustrate the origin of the strain-tailored optical properties of hybrid glass-clad/crystalline-core fiber devices.

Ligand-Engineered Full-Color Tuning

As demonstrated in Figure , achieving full-color light emission from a single fiber with high crystallinity by engineering the electronic structure using the LHPG technique represents a significant breakthrough. Thus, further investigating the origin of color center formation using different fiber growth conditions is important. It should be noted that the scattering of photons in biological tissue is less pronounced in the near infrared region than that in the UV region. Short-wavelength emission in the UVA (320–400 nm) range typically penetrates tissue to only a few hundred micrometers. However, many biomedical diagnostic methods for conditions, such as colonic dysplasia and polyps, use UVA emission to discriminating adenomas from normal tissues in vitro.[90,91] UVA can penetrate the skin more deeply than UVB (290–320 nm) and is thought to play a key role in skin aging and wrinkling. It is believed that UVA does not cause major damage to the epidermis; however, prolonged exposure to UVA radiation can induce tissue-specific damage and cellular mutagenicity and produce reactive oxygen species that cause oxidative DNA damage.[92,93]

Referring to step 2 in Figure , the luminescence spectra of the sapphire crystalline-core/glass-clad fibers both with and without annealing were found to primarily comprise nine emission bands according to Gaussian deconvolutions, as shown in Figure c,d. These principal luminescent bands exhibit peaks at approximately 352, 378, 400, 406, 455, 475, 530, 587, and 636 nm and agree well with the known F- and F2-type centers, Ti3+/Ti4+ ions, Ti4+-facilitated aluminum vacancies (Ti4+–VAl‴), and Cr3+ ions with peaks at 330, 380, 410, 420, 460, 480, 510, 590, and 650 nm,[42,94−97] as listed in Table . Alternatively, the luminescent band at ∼530 nm can be assigned to donor emission centers, such as interstitial aluminum ions (AlAl×).[98] Note that the center wavelengths are nearly identical and that the corresponding discrepancies are all within 5%, except for those of the F+ center, which are shifted by 6.7% from 330 to 352 nm because of the longer excitation wavelength (260 vs 325 nm). This red-shifted Stokes shift shows that the typical F-type center emission originates from a trapped charge coupled to lattice vibrations by analogy to the phonon-assisted alkali halides.[99]

Table 1

Summary of the Luminescent Efficiency and Gain Properties of Ligand-Driven Emissions in LHPG-Grown Sapphire Fibers and Referenced Bulk Sapphire

	glass-clad fiber with nonannealed sapphire crystalline core		glass-clad fiber with annealed sapphire crystalline core		bulk sapphire for reference
color emitters	center wavelength (nm)	gain–bandwidth product × 103 (au)	center wavelength (nm)	gain–bandwidth product × 103 (au)	center wavelength (nm)
F+	352	358	352	169	330[94]
F2+	379	41	377	36	380[94]
F	400	332	402	181	410[94]
Ti4+	406	2	406	1	420[97]
Ti3+	446	43	452	162	460[97]
Ti4+–VAl‴	471	128	478	32	480[42]
F2/AlAl×	537	301	523	960	510[85,98]
F22+	588	43	586	176	590[95]
Cr3+	637	–	636	–	650[97]

To date, abundant reports on the formation of defect emitters in sapphire crystals have been published because of their impressive electronic transitions and thus a variety of luminescent properties over the entire full-color spectral range have been developed. However, to the best of our knowledge, no direct observations of the effects of the interplay between color center evolution and thermal treatments have been reported, especially in the fiber form. As summarized in Table , the first salient feature is that the gain–bandwidth product (σ)(Δν) (i.e., the area of the Gaussian-fitted peak) ratio of Ti3+/Ti4+ ions in the annealed sample is almost 8 times higher than that in the nonannealed sample, suggesting that heating under reducing conditions can generate a significant number of Ti3+ centers. In fact, the reduction of sapphire increases the ratio of Ti3+ to Ti4+ ions and thus this conclusion persists.[42,43] Therefore, it is common to anneal Ti:sapphire crystals in strongly reducing environments to optimize the laser efficiency in favor of the Ti3+ ions. By comparison, a nonannealed sample contains a rather large (σ)(Δν) ratio of Ti4+ emission band at 420 nm, implying that the formation of the Ti4+ ions must be accompanied by the simultaneous introduction of aluminum vacancies for charge compensation. This is because local charge neutrality requires that an aluminum vacancy be formed for every three Ti4+ ions. Thus, charge-compensating defect centers (OO× and VAl‴) associated with the Ti3+ and Ti4+ ions should occur via the following reaction in Kröger–Vink notation:Here, TiAl× and TiAl• represent the Ti ions with 3+ and 4+ charges, respectively, and OO× is the interstitial oxygen. Indeed, compared to that in the Ar-reduced sample (i.e., the annealed fiber) in Figure d, we found that the growth of the 40 μm sapphire core under oxidation results in a 4-fold enhancement in (σ)(Δν) at 471 nm, indicating the domination of Ti4+–VAl‴ in the nonannealed sample. In addition, Ti4+–VAl‴ also acts as a hole trap in the charge-exchange state, generally resulting in blue-green emissions in the 450–500 nm range. For the nonannealed sample, the negatively charged Ti4+–VAl‴ is likely to be compensated for by positive charges, namely, the F+ and F2+ centers, resulting in rather large F+ and F2+ concentrations compared to those in the Ar-annealed sample. Thus, the enhanced oxygen divacancy concentrations in Mg-doped sapphire single crystals result from such a heating process in air.[28,95] It should be noted that the total (σ)(Δν) intensity of the F+ and F2+ centers decreased by over 2-fold after annealing. Therefore, the reduction of the sapphire core decreases the concentrations of F+ and F2+ to a greater extent than that in the nonannealed sample, as denoted by the first two purple Gaussian deconvolutions in Figure c,d. This behavior arises because heating in an inert Ar atmosphere can efficiently suppress the occurrence of many more oxygen vacancies than heating in ambient air. Thus, we observe a remarkable color deviation between the nearly white light and green light produced by the absence of short-wavelength blue emissions, as evidenced by the chromaticity coordinates in Figure b. Analogously, the F+/F concentration ratio of the Ar-annealed sample was reduced by ∼13%, which is consistent with the values obtained by high-energy bombardment and thermochemical reduction in Ar.[22]

In addition to the aforementioned Ar-prepared sample in the wavelength range of <400 nm, it is also noteworthy that the (σ)(Δν) intensity of the F+ center is 2 times smaller than that of the sapphire fiber without annealing, as shown in Figure c,d. In contrast, in the same sample, the F22+ center at ∼590 nm shows an intense yellow emission with a four-fold enhancement. Thus, the interaction after thermal annealing is expected to significantly impact the tailoring of both the short (blue)- and long (orange)-wavelength emissions, given the previously observed thermal-induced interconversion between the F+ and F22+ centers. This phenomenon was recently proven experimentally by Ramírez et al.[95]where ϕ is the binding energy of the F22+ centers. For thermal treatment at 1650 °C, F+ centers aggregate more energetically and form a cluster. The resulting Ar-annealed sapphire crystalline core contains substantial amounts of F22+ centers, yielding strong emission at 586 nm, as shown in Figure d. Accordingly, this high-temperature heat treatment often leads to large aluminum interstitial concentrations and observable quantities of oxygen vacancies because of their high formation energy, as confirmed by the intense green light at 500–550 nm in Figure d. Alternatively, for the sapphire fiber without annealing, the F+ center’s mobility is too low for it to be an F22+ center, and relatively weak luminescence at ∼590 nm is observed. Figure shows the corresponding EELS spectra of both the Ar-annealed and nonannealed glass-clad sapphire crystal fibers. For comparison, these spectra were probed in regions with similar thicknesses. The intensity of the O-K edge of the Ar-annealed sample is higher than that of the nonannealed sample, demonstrating that the oxygen concentration in the Ar-annealed sample exceeds that in the nonannealed sample. However, as evidenced by the luminescence analyses in Figure c,d, heating the sample under an inert gas can substantially reduce the total amount of oxygen vacancies (F+, F2+, and F22+ centers) by ∼14%. Thus, we suggest that the nonannealed sample can include numerous misfit dislocations and stacking faults within its sapphire core, as indicated by HRTEM and EELS. Such dislocation formation may be responsible for the appearance of the considerable numbers of vacancies and interstitial centers in the luminescence spectrum.

Figure 6

EELS spectra of the O-K edge of glass-clad sapphire crystalline-core fibers with and without annealing treatment. These results indicate that the intensity of the first O-K peak in the sapphire core decreases as the oxygen vacancy concentration increases.

Experimental Section

Fabrications of Hybrid Crystalline-Core Glass-Clad Fibers

As shown in Figure a, a c axis sapphire rod with a rectangular cross-section of 0.5 × 0.5 mm2 was used as the starting material. The commercially available starting materials were prepared by the Czochralski method. To engineer wideband full-color generation and effectively facilitate the generation of F- and F2-type color centers, ca. 50–100 ppm Ti and Cr ions were doped into the 40 μm sapphire core. A 40 μm diameter sapphire crystalline core was first grown at a rate of 3 mm/min in an ambient atmosphere by the LHPG technique (see the SI and Figures S3 and S4 for details). The grown fiber can be as long as 50 cm, depending on the fiber-drawing system that is used. The 40 μm core and cladding maintain single-crystal and amorphous structures, respectively. Apart from the growth rate, the parameter that most strongly affects the fiber quality is the temperature used for the high-temperature treatment, as discussed in the Section . Thermal annealing treatment was performed at 1650 °C in a strongly reducing atmosphere of 5% H2 and 95% Ar for 3 h. In the second step, the as-grown, 40 μm core with/without thermal treatment was inserted into a borosilicate glass capillary with a 320 μm outer diameter and grown again using the co-drawing LHPG technique to form the borosilicate glass-clad sapphire crystalline-core fiber. The growth rate of the glass cladding is 6.5 mm/min. Note that the glass cladding waveguide structure in this step was prepared in a ∼10–3 Torr reducing atmosphere to prevent void formation.

Characterization of Nanostructures and Optical Properties

After the co-drawing LHPG process, the surface morphology of the as-grown fiber was examined, and EDX analysis was conducted by SEM (JXA-8900R; JEOL). The atomic structure at the glass-clad sapphire crystal-core interface and the SAED pattern were investigated by field-emission HRTEM (Tecnai G2 F20; FEI), equipped with an EELS operating at 200 kV. The HRTEM specimen was prepared with a dual-beam focused ion beam (FIB, SMI3050, Seiko), which can make precise cuts in specific areas and along selected crystallographic orientations (Figure S5). Atomic image processing and 2D Fourier filtering of the HRTEM images were performed using the Gatan DigitalMicrograph software. HRTEM images were compared with diffraction patterns using the processed image to confirm the local orientation relationship between the nanometer-sized crystallites and the sapphire core matrix.

Wavelength-tunable full-color and Raman experiments were conducted with a 325 nm He–Cd laser (IK3802R-G; Kimmon Koha) as the light source and a high-spectral-resolution spectrometer (LabRAM HR 800; JOBIN-YVON) with an 1800 mm–1 grating. A 40× objective lens with a numerical aperture of 0.50 (LMU-40X-NUV; OFR) was employed to achieve submicrometer spatial resolution. The bright-field end view of an as-grown glass-clad sapphire crystalline-core fiber was obtained using a metallographic microscope (LV100ND; Nikon). The corresponding refractive index profile is shown in Figure S6. To measure the full-color output power, the 325 nm excitation was coupled to an 18 mm long fiber by an objective (LMU-40X-NUV; OFR). The full-color output and the residual excitation were first collimated by an achromatic lens with a 10 mm focal length and further blocked by a long-wavelength-pass filter (BLP01-325R-25; Semrock) before detection by a photodetector (818-UV, Newport). The sensing range of the photodetector was 200–1100 nm. The filtered full-color emissions were collected by an optical fiber (M15L02; Thorlabs) and then dispersed by a reflective holographic grating (GH13-36U; Thorlabs) in the whole UV–visible range.

Conclusions

In summary, we presented an effective and facile method of tailoring color-tunable BLSs with ultra-broadband emissions by engineering the defect–ligand properties. To demonstrate this phenomenon, we designed and fabricated a sapphire crystalline core–borosilicate glass cladding fiber with intentionally laser-introduced defects and dopants as color emitters using the LHPG technique. Interestingly, the full-color luminescence could be finely tuned to cover the whole UV–visible region while retaining high crystallinity and low waveguide loss, which remains challenging in optics and materials. In addition, HRTEM and EELS indicated that the full-color broadband emission resulted from an assembly of abundant dopant-facilitated aluminum vacancies and oxygen monovacancies and divacancies, providing high axial resolution in air for OCT in the submicrometer range. On the basis of these results, this study expands upon the potential for using functional defect–ligand-driven ultra-broadband emission from crystalline fibers in next-generation clinic endoscope-compatible diagnostic devices.