Open Access
Add to BibSonomy
Abstract
Laser heating is well-established to impart optical functionality into glass by local modification and crystallization. In this paper, we demonstrate electron beam heating in scanning and transmission microscopes as an alternative for the local crystallization of Sb2S3 in Sb-S-I glasses as a model system. Governed by different absorption physics, the electron beam expands morphological control of crystal cross section relative to laser, producing nanoscale (∼50 nm) single crystal architectures. We also report the effects of accelerating voltage and probe current and characterize the curved lattice of crystals formed in glass with electron diffraction techniques.
© 2022 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Laser heating of glass has proven itself as a versatile tool for imparting optical functionality into glass. Local modifications [1,2] can be introduced to fabricate waveguides [3], nanogratings [4], and fiber Bragg gratings [5]. Further optimization of glass compositions and laser heating can lead to space-selective crystallization [6–10] for the patterning of higher index polycrystalline and single crystal [11] waveguides, the formation of noncentrosymmetric crystals for second harmonic generation [6,12], and possibly electro-optic modulators. In general, every glass has its own absorption profile and heating characteristics, which strongly depend on laser wavelength. Changing the laser wavelength into the linear absorption regime above a material’s bandgap is nontrivial requiring different laser sources for different materials; in many cases no suitable laser source exists. Alternatively, glasses can be doped with absorbing transition or rare earth metal ions, which has been demonstrated with oxide glasses to increase linear absorption [13]. Where these dopants become undesirable, glasses can be heated by nonlinear absorption using ultrafast pulse lasers [14]. In both of the latter cases, architectures can even be fabricated in three dimensions.
Electron heating offers an alternative to laser heating which relies on an entirely different absorption mechanism than light [15]. Electron beam heating of glasses has been used in welding metallic glasses [16–18], selective patterning of nanostructures [19,20], and crystallization of thin films [21–23], specifically phase change materials [24–28]. Electron beam heating is less material dependent and circumvents the complications and shortcomings of laser heating mentioned above. Moreover, the smaller focal size and readily controllable electronic interaction volume by choice of accelerating voltage can expand the morphological control of crystal cross section for electron beam-fabricated crystal architectures relative to conventional laser-fabrication. To explore this hypothesis, in this work we contrast the morphology of crystals fabricated by both types of radiation. The resulting higher resolution crystal architecture fabricated by electron beam irradiation would be particularly useful for nanophotonics and waveguiding applications.
Laser-fabricated crystals in glass often exhibit curved lattice structures [29] with potential optical properties beyond those expected of perfect single crystals. The laser-fabrication of so-called rotating lattice single (RLS) crystals of Sb2S3 in Sb-S-I glass [30] is one such heavily studied example of these lattice engineered crystals. It has emerged as a model material for understanding the engineering of lattice of crystals grown in glass, as well as for its potential applications in phase change memory [31], solar cells [32], photocatalysis [33], or as an anode material in Li/Na ion batteries [34]. A few examples of crystals with curved lattices have also been reported in thin films heated by static electron beam irradiation [21–23]. Here, we have investigated the intentional and controlled formation of curved lattice in both bulk and thin film samples by scanned electron beam irradiation, which makes these crystals more suitable for device applications.
2. Experimental procedures
2.1 Preparation and characterization of chalcogenide glasses
Chalcogenide glasses were prepared in the xSbI3 – (100-x)Sb2S3 system from reagent grade precursors using the ampule quench method described previously [35]. For x = 4-8, glasses were prepared in quartz tubes with 7 mm ID and for x ≥ 12, in 11 mm ID, both with 1 mm wall thickness. The lower glass forming ability of pure Sb2S3 required higher cooling rates achieved with 1 mm ID quartz tubes with 10 µm wall thickness following [30]. Each glass, up to x = 24, was within about 1 mol.% of nominal elemental compositions according to wavelength dispersive spectroscopy measurements (JEOL JXA-8900R Superprobe). Glasses were characterized with differential scanning calorimetry (404 Pegasus F3 Netzsch) using a 20 K/min heating rate under flowing nitrogen gas. At lower values of x, these glasses readily crystallize into the biaxial orthorhombic Sb2S3 stibnite phase (a = 11.314 Å, b = 3.837 Å, c = 11.234 Å) [36, 37]. The two compositions most studied in this work were x = 0 and 4, which had glass transition temperatures of 224°C and 215°C with peak crystallization temperatures of 263°C and 293°C, respectively. In general, SbI3 additions increase glass stability. Before laser or electron beam crystallization experiments, all glasses were polished with progressively finer abrasives to an optical finish with 50 nm colloidal silica. Laser crystallization was achieved under flowing nitrogen gas using a 639 nm continuous wave diode laser (LP637-SF70, ThorLabs) focused through a 50X, 0.75 NA objective lens.
2.1 Electron microscopy
Electron beam crystallization on bulk samples as prepared above was performed inside a scanning electron microscope (SEM) and on thin samples in a transmission electron microscope (TEM) at the Materials Characterization Facility of the Institute for Functional Materials and Devices, Lehigh University. The former was performed using a FEI Scios FIB (focused ion beam) instrument, which is a dual beam system equipped with both electron and ion beams. To prevent sample charging in the SEM, glasses were sputter coated with about 1 nm of Ir and electrically grounded. Also, no charging of the sample was observed in TEM. Electron beam crystallization in the SEM was performed at 70° sample tilt at 15 mm working distance unless otherwise noted. This is a standard setup for electron backscatter diffraction (EBSD) analysis. EBSD patterns were acquired with a Hikari camera using EDAX TEAM EBSD data acquisition software. Pattern orientations were indexed using an Sb2S3 crystal reference and characterized by several parameters such as the image quality, fit, and confidence index. To only include information from the crystal, all inverse pole figure (IPF) colored maps were generated using patterns with the top 60% of image quality and confidence index greater than zero. Imaging of SEM electron radiation damage was performed using lower beam currents in a Hitachi 4300 SE. Within the FEI Scios FIB instrument, electron transparent samples were prepared for the TEM following the same FIB milling procedure as [38]. TEM analysis and crystallization of thin samples was performed using a JEOL JEM2100 operated in both TEM and nanobeam diffraction (NBD) modes with an accelerating voltage of 200 kV.
3. Crystallization in SEM
3.1 Exploration of parameter space
The most relevant parameters for electron beam heating using a focused beam are the accelerating voltage and probe current. The product of the electron energy, controlled by the accelerating voltage, and the electron flux incident on the sample, controlled by the probe current, is the overall beam power. The electron beam power is often limited to a few mW, well within the range to crystallize the chalcogenide glasses used in this work. In general, increasing probe currents will increase the size of the electron probe according to
where ${\textrm{d}_\textrm{p}}$ is the probe diameter, ${\textrm{i}_\textrm{p}}$ is the probe current, ${\mathrm{\alpha }_\textrm{p}}$ is the probe convergence angle, and $\mathrm{\beta }(\textrm{V})$ is the brightness of the electron source for a given accelerating voltage [39]. For the SEM experiments in this work, the focused probe should never exceed 100 nm. However, the region with a sufficient temperature rise to induce crystallization is assumed to be significantly larger than the focused probe size.Figure 1(a) shows the effect of electron beam irradiation on the glass surface. With insufficient heating, the glass structure is modified without crystallization, seen as a slight swelling of the glass surface. Increasing the beam current leads to crystallization and a slight depression presumably due to ensuing densification. Diffraction patterns were recorded to determine the formation of crystal by the presence or lack of Kikuchi lines shown in Fig. 1(b). Using this setup, a section of accelerating voltage and probe current parameter space was explored for both the x = 0 and 4 glass compositions in Fig. 1(c). For each set of beam parameters, five regions of the glass were irradiated for 30 s. The presence of Kikuchi lines after the hold was used to indicate crystal formations and the fraction of the five replicates is indicated by the color in Fig. 1(c). ‘X’ indicates where the beam current was high enough to ablate the sample beyond which crystallinity could not be determined. The x = 0 glass crystallized more readily across beam parameters than the x = 4 glass, matching the glass stability behavior expected from differential scanning calorimetry measurements. Although the beam parameter space was explored in detail for the x = 0 and 4 compositions, up to x = 12 glasses could be crystallized with the electron beam. This is contrasted when using a similarly powered laser, where glasses up to x = 37 could be readily crystallized [40,41].
Fig. 1. Comparison of modification and crystallization by electron beam. (a) SEM images comparing x = 0 glass modification with 20 kV beam at 16 nA and crystallization at 32 nA probe current. (b) 30 kV EBSD patterns used to distinguish modification and crystallization. (c) Exploration of accelerating voltage and probe current parameters space for stoichiometric Sb2S3 and x = 4 glasses. ‘X’ indicates significant beam damage beyond which crystallinity could not be determined.
The shape of the crystallized regions in Fig. 1(c) can be understood with consideration toward the electron interaction volume. Increased penetrating power at higher accelerating voltages increases the electron interaction volume, subsequently creating a more diffuse energy input into the sample. Conversely, at lower accelerating voltages, the electron interaction volume becomes smaller than the volume with a sufficient temperature rise to induce crystallization to the extent that a point source heating model may become valid. Overall, crystallization at higher accelerating voltages is limited by the power density of the electron interaction volume, but by the beam power alone at lower accelerating voltages. The crossover between 15 and 20 kV is likely where the electron interaction volume matches the region with a sufficient temperature rise to induce crystallization. Although there are methods to approximate temperature distribution under an electron beam [15,42], a more comprehensive approach would require Monte Carlo simulations of electron interactions combined with finite element modeling of the 3D temperature distribution. For well-grounded SEM or sufficiently thin TEM samples of the present investigations showing no charging, electron irradiation effects such as electron emission and X-ray generation would lead to some proportion of the beam power not absorbed by the sample. Although these losses might be considered like reflection losses during laser irradiation, their unique interactions may produce additional structural changes that impact the crystallization process.
3.2 Patterning of electron beam induced crystallization
In addition to fabrication of crystal dots, like shown in Fig. 1(a), the electron beam can by scanned and selectively rastered to form crystal lines and patterns. Scanning in the SEM is achieved by beam scanning in contrast to stage scanning in laser setups. Beam scanning introduces some variance in the angle of incidence of the electron beam across the scanned area, but this can be minimized with larger working distances (15 mm in this study). The quality of crystal lines fabricated by electron beam scanning is a function of the beam parameters explored in Fig. 1(c) and the effective scanning rate. This effective scanning rate is obtained by a series of static irradiations for some dwell time spaced some pitch apart, usually on the order of 2-30 ms and 25-40 nm, respectively. The formation of single crystal architectures by electron beam heating is governed by principles similar to those for laser heating in similar systems [30,40,43]. For a given beam power, the effective scanning rate must be slow enough to allow sufficient temperature rise for crystal growth and fast enough to prevent nucleation of additional undesirable crystal grains. This constraint limits electron beam scanning to < 10-20 µm/s for isolated crystal lines with beam currents below 100 nA. This range is contrasted with laser-fabricated single crystals which can be scanned at rates exceeding 200 µm/s for similar laser powers in the same glasses [44].
By allowing the beam to be scanned freely, continuous crystal curves can be fabricated like the cursive ‘Lehigh’ in Fig. 2(a). The continuous color in the surface normal inverse pole figure (IPF) map next to the SEM image indicates the ‘Lehigh” crystal has a highly oriented lattice with no indication of large-angle grain boundaries. The slight gradual variation of color from left to right indicates a lattice rotation of about 20° across the crystal length which is highlighted by the crystal orientation deviation (COD) map. This type of rotation is expected of surface crystallized architectures similar to laser-fabricated RLS crystals studied by Savtyskii et al. [30]. The electron beam can also be selectively rastered to form architectures like the Lehigh University logo in Fig. 2(b). Here, the fast scan direction at 62.5 µm/s was left to right, and slow scan direction top to bottom which could be as slow as 0.03 µm/s. The actual crystal growth speed is likely between these two values since the fast scan direction exceeds the 10-20 µm/s maximum scanning observed for isolated, straight crystal lines, and the grain boundaries form in a direction between the two scanning directions. Due to the simple rastering procedure and the complicated logo shape, many regions are irradiated without any preexisting crystal to serve as a seed for subsequent growth. This complicated configuration of architecture forces nucleation and growth of several independent crystals, resulting in a multi-crystalline pattern. For example, the heart shape in the middle starts at both top regions, hence two independent crystals were nucleated and allowed to grow throughout the region. These independent nucleation events can be avoided by more complicated scanning procedures to avoid irradiating any new regions beyond the initial starting location, but at the cost of a consistent crystal growth direction across the entire pattern.
Fig. 2. Examples of crystals patterned on the surface of x = 4 glass in SEM by scanning the electron beam. (a) cursive ‘Lehigh’ fabricated by continuously and freely scanning the beam (b) Lehigh University logo patterned by selectively rastering the electron beam. Both patterned crystals are shown as SEM images and inverse pole figure (IPF) colored maps, as well as a crystal orientation deviation (COD) map for (a) indicating a 20° rotation from right to left. IPF maps are colored relative to the surface normal according to the unit triangle shown and were acquired with 250 nm step size on a hexagonal sampling grid.
Care should be given when imaging or mapping and crystallizing using the same electron beam in SEM to avoid excessive surface modification or spurious crystallization. Figure 3(a) shows modification during imaging along electron beam raster lines over a previously fabricated crystal line. Like in Fig. 1(a), the glass surface experiences some swelling under electron irradiation – however, the crystal is resistant to this modification. At even greater electron dosages, heating during imaging can lead to spurious crystallization across the sample surface, especially around previously fabricated lines, which tend to act as crystal seeds. To avoid excessive surface modification and spurious crystallization, imaging should be limited to lower electron dosages and faster effective scanning rates achieved through lower beam currents, image resolution, pixel dwell times, or magnification. Often the beam dwells slightly longer at the edges of each raster line, leading to a greater potential for damage and spurious crystallization at the edges of the field of view (FOV) – shown in Fig. 3(b).
Fig. 3. Examples of undesirable electron beam modification during imaging in x = 0 glass. (a) glass modification from rastered electron beam during imaging around a previously fabricated crystal line. (b) exacerbated sample damage at the edge of the imaging field of view (FOV).
3.3 Comparison of laser and electron beam-fabricated crystal morphology
Laser and electron beam irradiation have already been contrasted in the previous sections for the propensity to crystallize different glasses and the maximum allowable effective scanning rates to maintain crystal growth. Further differences can be highlighted when considering the morphology of crystal width and transverse cross sections. For patterning, fine crystal widths are desired to achieve higher resolution architectures. Figure 4 compares crystal lines grown in x = 4 glass from a single seed line by both laser and electron beam in an SEM with parameters optimized for thinner crystal widths. Laser-fabricated crystal lines are limited to about 1500 nm when grown with 2 mW laser power at 20 µm/s – the very edge of their parameter space for sustained single crystal growth. In contrast, electron beam-fabricated crystal lines readily achieve submicron widths of 600 nm when grown using a 10 kV and 32 nA beam scanned at 1 µm/s. The electron beam focal size is orders of magnitude smaller than any visible laser, and when combined with the limited electron interaction volume at lower accelerating voltages, a very small volume of glass can be heated and transformed to single crystal. Even thinner crystal lines are likely possible with further optimization at even lower accelerating voltages.
Fig. 4. Comparison of crystal resolution possible between electron beam and laser-fabricated crystal lines. (a) SEM image for crystal lines fabricated with a 10 kV and 32 nA electron beam scanned at 1 µm/s and crystal lines fabricated by 639 nm laser with 2 mW power scanned at 20 µm/s. Laser focus size indicated. (b) IPF map of region indicated in (a) colored according to unit triangle shown referenced to the surface normal acquired with 50 nm step size on a hexagonal sampling grid.
Given the ability of laser to crystallize more stable, higher SbI3 content glasses and faster allowable scanning rates, laser heating appears more efficient for crystallization compared to electron beam, but direct comparison between the two radiation sources shows a more nuanced understanding is required. TEM samples were prepared with transverse cross sections of crystal lines fabricated by electron and laser irradiation with identical powers and scanning rates. Crystal lines were fabricated in x = 0 glass with 2 mW beam powers (30 kV and 64 nA for electron beam) and scanned at 10 µm/s. Both lines were fabricated with beams perpendicular to the sample surface. Figure 5(a) and (b) directly compare these two transverse crystal cross sections. The coordinate system we developed previously [29] is indicated, where the 1-axis is the scanning direction, and the 3-axis is opposite beam propagation. Figure 5(a) includes inserts of selected area electron diffraction (SAED) patterns distinguishing the crystal region, marked by bend contours and diffraction contrast from the remaining amorphous regions. Using this distinction, the electron-beam fabricated crystal has a much larger cross section at 3500 nm wide and 4700 nm deep compared to the laser-fabricated crystal in Fig. 5(b) at about 1700 nm wide and only 400 nm deep. This direct comparison implies electron beam heating is more efficient for transforming glass to crystal. One explanation for the difference in crystal cross section size would be higher temperatures under electron beam, but this cannot explain the electron beam’s slower maximum scanning rates and inability to crystallize x > 12 glass compositions. An alternative explanation is a potential time delay between initial electron interaction and full absorption as heat. The electrons could serve as a more efficient heating source, but would be less efficient in terms of time, resulting in larger crystallized volumes, but slower maximum scanning rates. Another plausible explanation is that the temperature profile under both radiation sources is largely the same and electron-material interactions (e.g., electron emission or X-ray generation) act to reduce the energy barrier of crystallization by some mechanism allowing the crystal to grow in lower temperature regions in the same amount time relative to laser heating. The exact cause for the difference in crystallized volume under identical energy inputs by electron and laser beams is currently unknown and requires a more detailed investigation.
Fig. 5. Cross sections of crystal lines fabricated in glass imaged in TEM. Transverse cross sections of (a) laser and (b) electron beam-fabricated crystal lines in stoichiometric Sb2S3 glass. Both lines were fabricated with 2 mW focused beams (30 kV and 64 nA for the electron beam) and scanned at 10 µm/s. Inserts in (a) are SAED patterns to distinguish between crystal and glass regions. Higher magnification image is included in (b). All scale bars correspond to 500 nm and coordinate system is indicated.
Sudden changes in diffraction contrast in TEM images are typically indicative of grain boundaries. The electron beam-fabricated crystal in Fig. 5(a) then appears to have some potential damage near the surface – likely from light polishing to eliminate surface roughness – and a polycrystalline section near the bottom. The laser-fabricated crystal in Fig. 5(b) has significant changes in diffraction contrast which may be caused by either polycrystalline regions or the large lattice curvatures developed at these particular laser parameters. At better optimized laser parameters for single crystal formation and lower rotation rates yield a more typical single crystal structure inside the TEM as seen in previous work [38].
4. Crystallization in TEM
4.1 Crystal nucleation and growth
Crystallization experiments in TEM were performed on thin samples (< 150 nm) extracted from bulk x = 0 glass by the FIB milling procedure. The electron beam was operated at 200 kV with probe currents between 2 and 50 nA and scanning rates between 20 and 75 nm/s, much slower than the µm/s scanning rates used with bulk crystallization. To first nucleate new crystals in thin samples, a focused beam was held over amorphous regions for some incubation time. Due to the stochasticity of crystal nucleation in glass and the small interaction volume of thin samples, the probability of nucleation was very low. Despite this rarity, independent nucleation did seldom occur, but exclusively at sample edges, corners, or the interface with leftover Pt from FIB milling. After the initial crystal nucleation, crystals could be readily grown or new grains nucleated from already formed crystal surfaces. In later experiments, the rarity of initial crystal nucleation was overcome by restricting experiments to only the crystal growth stage. Thin samples were prepared with cross sections of crystals already fabricated by laser or electron beam in the SEM (e.g., the sample prepared in Fig. 5). Figure 6(a) shows an example of using a laser-fabricated crystal as a seed for subsequent crystal growth inside TEM with 7 nA beam current scanned at 50 nm/s. Coordinate systems are indicated for both crystals to help orient the image.
Fig. 6. Example of crystal lines fabricated by 200 kV electron beam in TEM in stoichiometric Sb2S3 glass thin samples. (a) electron beam-fabricated crystal line grown from previously laser-fabricated crystal line. Coordinate systems for both lines are indicated. (b) 50 nm thin line fabricated by electron beam.
Growth of crystal lines in TEM was accomplished by focusing the electron beam and scanning the sample stage – in contrast with beam scanning in SEM, but similar to laser crystallization. Stage scanning axes are rotated ∼20° relative to the imaging axes in this particular TEM. Electron beam focal size was between 10-30 nm, following the same beam current relationship defined in Eqn. (1). The smaller interaction volume of thin samples limited the crystallized region to the nanoscale, but this region still extended well beyond the focal size. At lower beam currents, crystal lines were fabricated with widths down to 50 nm of indefinite lengths (see Fig. 6(b) fabricated with 2 nA beam also scanned at 50 nm/s). Crystal lengths were restricted in sample sizes prepared by FIB, but thin glass specimens can be prepared from bulk samples by other means [45]. A variety of thin film deposition methods could also be used to create much larger samples with more uniform thicknesses than FIB milling, but would require substrates for support [46]. These methods could be used to more rigorously study any thickness-dependent absorption or other phenomena similar to Sb3.6Te crystallization under static electron beam irradiation investigated by Kooi and De Hosson [22].
Like in SEM, care should be taken when imaging and crystallizing with the same electron beam in TEM. Although the beam can be focused to locally heat the sample, even parallel illumination can heat the sample leading to morphological changes. The results of excessive and uncontrolled heating are undesirable crystal grains that nucleate and grow from previously fabricated lines – see Fig. 7. This can occur as many smaller grains when the beam is too focused or can approach faceted crystal growth when the overall sample temperature becomes too high. If spurious crystallization becomes excessive, stresses from density changes upon crystallization can even lead to sample folding inside the microscope. Excessive sample heating can be avoided by lower electron dosages with smaller beam currents and diffuse illumination or actively with specialized cooling sample holders. Similar experiments could be performed in scanning transmission electron microscopes (STEM), where considerations similar to SEM should be taken to avoid excessive sample heating.
Fig. 7. Undesirable crystal growth from previously fabricated single crystal line in TEM by excessive sample heating. (a) crystal line as fabricated with 60 nm/s scanning rate. (b) crystal line after several minutes of illumination with high beam current (50 nA).
4.2 Lattice curvature in thin samples
Similar to the lattice curvature observed in crystal lines fabricated on the surface of bulk glasses, crystal lines fabricated in thin samples also exhibit lattice rotation along their length. Azimuthal lattice rotation was measured by collecting SAED patterns along the length of these lines, as shown in Fig. 8 for a crystal line fabricated with 50 nA beam scanned at 75 nm/s. Using the coordinate system introduced earlier, this line has 3-axis rotation along the crystal growth direction (i.e., 1-axis). The lattice curvature component, κij, used to describe this rotation about the i-axis along the j-direction is κ31 in this case. This κ31 lattice curvature component is 44 ± 5°/µm, but values less than 1°/µm have also been reported [29]. Electron beam-fabricated crystal lines in thin samples do not have any detectable dislocations in contrast to RLS crystal lines laser-fabricated in bulk samples where dislocations have been directly observed [38]. In bulk samples, dislocations are the geometrically necessary and energetically favorable manifestations of lattice rotation due to the larger crystal volume. The thickness of the film from Fig. 8 and other electron beam crystallization experiments in TEM in this work are apparently below the critical thickness where dislocation formation becomes energetically favorable. The lattice curvature is then entirely elastic with similar results reported in other electron beam induced crystals formed in thin films [21–23]. The κ31 lattice curvature in Fig. 8 is further differentiated from crystal lines fabricated on the surface of bulk glasses by either radiation source which exhibit rotation about the 2-axis along the 1-axis or a κ21 lattice curvature, typically with much lower magnitudes. This κ21 lattice curvature has also been reported in other crystals fabricated in thin films by static electron beam heating [21–23] with similar magnitudes to the present crystal lines. SAED patterns are easily acquired in TEM, but the distortions of the patterns in Fig. 8 indicate that the lattice rotation is not limited to only the 3-axis; unfortunately, the full 3D rotation is not easily measured by this method.
Fig. 8. Crystal line with lattice curvature fabricated in stoichiometric Sb2S3 glass by electron beam in TEM. Selected areas and their diffraction patterns indicated with major azimuthal angle measurement shown. Plot insert shows this azimuthal angle as a function of distance along crystal line with slope (κ31) indicated in °/µm and the 95% confidence interval.
Full 3D lattice rotations are more easily measured by rotation and translation of Kikuchi lines like in the EBSD pattern shown in Fig. 1(b). Kikuchi lines can also be generated in TEM by convergent beam electron diffraction (CBED), but this method requires focusing the electron beam on the sample leading to increased heating and potentially spurious crystallization around the diffraction volume. Figure 9(a) shows two crystal lines fabricated in TEM with 2 nA beam current with 45 nm/s scanning rates, but with different lattice orientations indicated by differences in the diffraction contrast. The line on the left was unmodified after fabrication, whereas the arrows labeled (b) and (c) indicate regions where CBED patters were acquired along the right line and spurious crystallization occurred. These two CBED patterns are shown in Fig. 9(b) and (c) and are misorientated by 17.2 mrad of azimuthal (3-axis) rotation and by 4.2 mrad of translation within the plane of the detector. The translation was then resolved about the remaining 1- and 2-axes. This particular section of crystal was indexed near the <102 > zone axis with a few Kikuchi bands highlighted. The <201 > zone axis may also be a valid solution due to the pseudosymmetry in stibnite, but does not change the misorientation or lattice curvature measurements. For this section of crystal line, the lattice curvature can be resolved about the three axes as κ11 = -0.34 ± 0.06°/µm, κ21 = -0.08 ± 0.01°/µm, and κ31 = 1.43 ± 0.26°/µm. The error is estimated by the uncertainty of the distance measurement, and is larger than the error associated with the rotation measurements. Significant rotation about all three sample axes in crystal lines fabricated by electron beam in thin samples is more complicated than in bulk and is not yet fully understood. Kooi and De Hosson showed that lattice curvature strongly depends on film thickness for Sb3.6Te crystallized under static irradiation [22], which may also be contributing to the results here. It should also be noted that CBED patterns can also yield information about lattice strain [47] and recent developments of 4D-STEM techniques can be used to map lattice orientation and curvature like EBSD in SEM [48].
Fig. 9. Analysis of lattice curvature by CBED patterns. (a) crystal line fabricated with 2 nA beam current scanned at 45 nm/s with enlarged crystal regions indicating areas where CBED patterns were acquired. (a) and (b) CBED patterns misoriented by some rotation. Some Kikuchi lines are highlighted and labeled with associated planes.
5. Conclusions
Electron beam heating has been demonstrated as a viable alternative to laser heating for fabricating single crystal architecture in the model system of Sb2S3 in xSbI3 – (100-x)Sb2S3 glasses. These crystals can be grown on the surface of bulk samples in SEM and in thin samples in TEM. Exploring the relevant beam parameters (accelerating voltage and probe current) revealed that the propensity of glass to crystallize under electron irradiation in bulk samples depends strongly on the choice of accelerating voltage, which subsequently controls the electron interaction volume. In thin samples, the smaller interaction volume significantly limits independent crystal nucleation, but growth can proceed unhindered.
After crystal formation, scanning the beam or the stage can be used to pattern arbitrary shapes with nanoscale precision in both microscopes (e.g., 50 nm width but indefinite length). These feature sizes are much smaller than the finest resolution achieved with conventional laser-fabrication method due to the smaller probe size and interaction volume of the electron beam. When directly comparing the two heat sources for the same energy input, electron beam irradiation produces much larger crystal volumes relative to laser. These differences indicate a different mechanism of heating or crystallization, but the exact cause of this distinction remains unknown. Sample modification and spurious crystallization can occur in both microscopes when imaging and crystallizing with the same electron beam, but these effects can be mitigated by reducing electron dosage through a variety of methods.
As with laser-fabricated crystals in glass, the growth of crystal lines by electron beam in bulk and thin samples imparts lattice curvature. Whereas the lattice of crystal lines fabricated in bulk samples rotate as expected of similarly laser-fabricated crystals, crystal lines fabricated in thin film morphology exhibit a more complicated 3D rotation. Lattice curvature can be measured by electron diffraction techniques such as by EBSD in bulk and by SAED or more completely by CBED in thin films.
Overall, electron beam offers an alternative localized heat source with expanded morphological control over crystal cross section, which has potential uses for fabricating lattice engineered single crystal architectures in glass for photonic structures or waveguiding applications.
Funding
U.S. Department of Energy (DE-SC0005010).
Acknowledgments
The authors would like to thank Robert Keyes, Bill Mushock, Masashi Watanabe, Joseph Cline, and Joshua Smeltzer for assistance with electron microscopes.
Disclosures
The authors declare no conflicts of interest.
Data availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
References
1. J. D. Musgraves, K. Richardson, and H. Jain, “Laser-induced structural modifications, its mechanisms, and applications in glassy optical materials,” Opt. Mater. Express 1(5), 921–935 (2011). [CrossRef]
2. T. T. Fernandez, M. Sakakura, S. M. Eaton, B. Sotillo, J. Siegel, J. Solis, Y. Shimotsuma, and K. Miura, “Bespoke photonic devices using ultrafast laser driven ion migration in glasses,” Prog. Mater. Sci. 94, 68–113 (2018). [CrossRef]
3. K. M. Davis, K. Miura, N. Sugimoto, and K. Hirao, “Writing waveguides in glass with a femtosecond laser,” Opt. Lett. 21(21), 1729–1731 (1996). [CrossRef]
4. Q. Sun, F. Liang, R. Vallée, and S. L. Chin, “Nanograting formation on the surface of silica glass by scanning focused femtosecond laser pulses,” Opt. Lett. 33(22), 2713–2715 (2008). [CrossRef]
5. A. Martinez, M. Dubov, I. Khrushcev, and I. Bennion, “Direct writing of fibre Bragg gratings by femtosecond laser,” Electron. Lett. 40(19), 1170–1172 (2004). [CrossRef]
6. Y. Yonesaki, K. Miura, R. Araki, K. Fujita, and K. Hirao, “Space-selective precipitation of non-linear optical crystals inside silicate glasses using near-infrared femtosecond laser,” J. Non-Cryst. Solids 351(10-11), 885–892 (2005). [CrossRef]
7. Y. Teng, J. Zhou, K. Sharafudeen, S. Zhou, K. Miura, and J. Qiu, “Space-selective crystallization of glass induced by femtosecond laser irradiation,” J. Non-Cryst. Sol. 383, 91–96 (2014). [CrossRef]
8. T. Komatsu and T. Honma, “Laser patterning and characterization of optical crystals in glasses,” J. Asian Ceram. Soc. 1(1), 9–16 (2013). [CrossRef]
9. T. Komatsu and T. Honma, “Laser patterning and growth mechanism of orientation designed crystals in oxide glasses: a review,” J. Sol. State Chem. 275, 210–222 (2019). [CrossRef]
10. J. Cao, M. Lancry, F. Brisset, L. Mazerolles, R. Saint-Martin, and B. Poumellec, “Femtosecond laser-induced crystallization in glasses: growth dynamics for orientable nanostructure and nanocrystallization,” Cryst. Growth Des. 19(4), 2189–2205 (2019). [CrossRef]
11. A. Stone, H. Jain, V. Dierolf, M. Sakakura, Y. Shimotsuma, K. Miura, K. Hirao, J. Lapointe, and R. Kashyap, “Direct laser-writing of ferroelectric single-crystal waveguide architectures in glass for 3D integrated optics,” Sci. Rep. 5(1), 10391 (2015). [CrossRef]
12. T. Komatsu, R. Ihara, T. Honma, Y. Benino, R. Sato, H. G. Kim, and T. Fujiwara, “Patterning of non-linear optical crystals in glass by laser-induced crystallization,” J. Am. Ceram. Soc. 90(3), 699–705 (2007). [CrossRef]
13. T. Honma, Y. Benino, T. Fujiwara, and T. Komatsu, “Transition metal atom heat processing for writing of crystal lines in glass,” Appl. Phys. Lett. 88(23), 231105 (2006). [CrossRef]
14. D. Tan, B. Zhang, and J. Qiu, “Ultrafast laser direct writing in glass: thermal accumulation engineering and applications,” Laser Photonics Rev. 15(9), 2000455 (2021). [CrossRef]
15. R. F. Egerton, P. Li, and M. Malac, “Radiation damage in the TEM and SEM,” Micron 35(6), 399–409 (2004). [CrossRef]
16. Y. Kawamura and Y. Ohno, “Successful electron-beam welding of bulk metallic glass,” Mater. Trans. 42(11), 2476–2478 (2001). [CrossRef]
17. S. Kagao, Y. Kawamura, and Y. Ohno, “Electron-beam welding of Zr-based bulk metallic glasses,” Mater. Sci. Eng., A 375-377, 312–316 (2004). [CrossRef]
18. G. Wang, Y. Huang, W. Cao, Z. Huang, M. Huttala, Y. Su, and C. Tan, “Microstructure and crystallization mechanism of Ti-based bulk metallic glass by electron beam welding,” J. Manuf. Process. 32, 93–99 (2018). [CrossRef]
19. M. Vlcek and H. Jain, “Nanostructuring of chalcogenide glasses using electron beam lithography,” J. Optoelectron. Adv. Mater. 8(6), 2108–2111 (2006).
20. V. Takats, F. Miller, H. Jain, C. Cserhati, and S. Kokenyesi, “Direct surface patterning of homogeneous and nanostructured chalcogenide layers,” Phys. Status Solidi C 6(S1), S83–S85 (2009). [CrossRef]
21. V. Yu Kolosov and A. R. Thölen, “Transmission electron microscopy studies of the specific structure of crystals formed by phase transition in iron oxide amorphous films,” Acta Mater. 48(8), 1829–1840 (2000). [CrossRef]
22. B. J. Kooi and J. T. M. De Hosson, “On the crystallization of thin films composed of Sb3.6Te with Ge for rewritable data storage,” J. Appl. Phys. 95(9), 4714–4721 (2004). [CrossRef]
23. A. G. Bagmut, S. N. Grigorov, V. Y. Kolosov, V. M. Kosevich, and G. P. Nikolaychuk, “The growth of Sb2S3 crystals with bend lattice during amorphous films annealing and condensation,” Func. Mater. 12(3), 461–466 (2005).
24. M. Kaiser, L. Van Pieterson, and M. A. Verheijen, “In situ transmission electron microscopy analysis of electron beam induced crystallization of amorphous marks in phase-change materials,” J. Appl. Phys. 96(6), 3193–3198 (2004). [CrossRef]
25. R. Pandian, B. J. Kooi, J. T. M. De Hosson, and A. Pauza, “Influence of electron beam exposure on crystallization of phase-change materials,” J. Appl. Phys. 101(5), 053529 (2007). [CrossRef]
26. D. Zhou, L. Wu, L. Wen, L. Ma, X. Zhang, Y. Li, Q. Guo, and Z. Song, “Electron-beam-irradiation-induced crystallization of amorphous solid phase change materials,” Jpn. J. Appl. Phys. 57(4), 041401 (2018). [CrossRef]
27. T. T. Jiang, J. J. Wang, L. Lu, C. S. Ma, D. L. Zhang, F. Rao, C. L. Jia, and W. Zhang, “Progressive amorphization of GeSbTe phase-change material under electron beam irradiation,” APL Mater. 7(8), 081121 (2019). [CrossRef]
28. M. K. Singh, C. Ghosh, B. Miller, and C. B. Carter, “Direct visualization of the earliest stages of crystallization,” Microsc. Microanal. 27(4), 659–665 (2021). [CrossRef]
29. E. J. Musterman, V. Dierolf, and H. Jain, “Curved lattices of crystals formed in glass,” Int. J. Appl. Glass. Sci. 13(3), 402–419 (2022), special issue on International Year of Glass. [CrossRef]
30. D. Savytskii, H. Jain, N. Tamura, and V. Dierolf, “Rotating lattice single crystal architecture on the surface of glass,” Sci. Rep. 6(1), 36449 (2016). [CrossRef]
31. M. Delaney, I. Zeimpekis, D. Lawson, D. W. Hewak, and O. L. Muskens, “A new family of ultralow loss phase-change materials for photonic integrated circuits: Sb2S3 and Sb2Se3,” Adv. Funct. Mater. 30(36), 2002447 (2020). [CrossRef]
32. R. Kondrotas, C. Chen, and J. Tang, “Sb2S3 solar cells,” Joule 2(5), 857–878 (2018). [CrossRef]
33. H. Wang, X. Yuan, H. Wang, X. Chen, Z. Wu, L. Jiang, W. Xiong, and G. Zeng, “Facile synthesis of Sb2S3/ultrathin g-C3N4 sheets heterostructures ebedded with g-C3N4 quantum dots with enhance NIR-light photocatalytic performance,” Appl. Catal., B 193, 36–46 (2016). [CrossRef]
34. X. Xiong, G. Wang, Y. Lin, Y. Wang, X. Ou, F. Zheng, C. Yang, J. H. Wang, and M. Liu, “Enhancing sodium ion battery performance by strongly binding nanostructured Sb2S3 on sulfur-doped graphene sheets,” ACS Nano 10(12), 10953–10959 (2016). [CrossRef]
35. P. Gupta, A. Stone, N. Woodward, V. Dierolf, and H. Jain, “Laser fabrication of semiconducting ferroelectric single crystal SbSI features on chalcohalide glass,” Opt. Mater. Express 1(4), 652–657 (2011). [CrossRef]
36. A. Kyono, M. Kimata, M. Matsuhisa, Y. Miyashita, and K. Okamoto, “Low-temperature crystal structures of stibnite implying orbital overlap of Sb 5s2 inert pair electrons,” Phys. Chem. Miner. 29(4), 254–260 (2002). [CrossRef]
37. D. Savytskii, K. Atwater, V. Dierolf, and H. Jain, “Formation of ferroelectric phases in Sb-S-I glasses,” J. Am. Ceram. Soc. 97(11), 3458–3462 (2014). [CrossRef]
38. E. J. Musterman, D. Savytskii, V. Dierolf, and H. Jain, “The source of lattice rotation in rotating lattice single (RLS) crystals,” Scr. Mater. 193, 22–26 (2021). [CrossRef]
39. M. S. Bronsgeest, J. E. Barth, L. W. Swanson, and P. Kruit, “Probe current, probe size, and the practical brightness for probe forming systems,” J. Vac. Sci. Technol. B 26(3), 949–955 (2008). [CrossRef]
40. D. Savytskii, B. Knorr, V. Dierolf, and H. Jain, “Demonstration of single crystal growth via solid-solid transformation of a glass,” Sci. Rep. 6(1), 23324 (2016). [CrossRef]
41. C. Au-Yeung, D. Savytskii, K. Veenhuizen, S. McAnany, H. Jain, and V. Dierolf, “Polarization and surface effects on the seed orientation of laser-induced Sb2S3 crystals on Sb-S-I glass,” Cryst. Growth Des. 21(8), 4276–4284 (2021). [CrossRef]
42. M. Liu, L. Xu, and X. Lin, “Heating effect of electron beam bombardment,” Scanning 16(1), 1–5 (1994). [CrossRef]
43. D. Savytskii, V. Dierolf, N. Tamura, and H. Jain, “Fabrication of single crystal architecture in Sb-S-I glass: transition from dot to line,” J. Non-Cryst. Solids 501, 43–48 (2018). [CrossRef]
44. D. Savytskii, E. Musterman, V. Dierolf, and H. Jain, “Influence of the laser scanning rate on the structure of rotating lattice single crystal lines,” Cryst. Growth Des. 19(11), 6324–6330 (2019). [CrossRef]
45. B. J. Kestel, “Preparation of damage-free glass TEM specimens,” Ultramicroscopy 83(1-2), 61–66 (2000). [CrossRef]
46. J. Orava, T. Kohoutek, and T. Wagner, “Deposition techniques for chalcogenide thin films,” in Chalcogenide Glasses (Woodhead Publishing, 2014).
47. A. Béché, J. L. Rouvière, J. P. Barnes, and D. Cooper, “Strain measurement at the nanoscale: comparison between convergent beam electron diffraction, nano-beam electron diffraction, high resolution imaging and dark field electron holography,” Ultramicroscopy 131, 10–23 (2013). [CrossRef]
48. C. Ophus, “Four-dimensional scanning transmission electron microscopy (4D-STEM): from scanning nanodiffraction to ptychography and beyond,” Microsc. Microanal. 25(3), 563–582 (2019). [CrossRef]
