Optically transparent Pelco^® membranes from pure and amorphous, thermally deposited silicon dioxide (50 μm × 50 μm windows in a silicon nitride frame) are purchased from Plano GmbH (Germany), which closely resembles the structure of silica glass (see section Impact of Ion- and e-beam Irradiation on the Structure of Silica Glass). The site-specific preparation of silica glass tensile specimens was carried out using a Helios NanoLab 660 DualBeam SEM/FIB system from FEI Company (Netherlands). The SEM/FIB is equipped with an ultra high-resolution Elstar™ e-beam column with a monochromator system and a resolution of ~1 nm at 1–30 kV, a high-resolution Tomahawk™ ion beam column, a gas injection system for C deposition, an EasyLift™ micromanipulator, and an Oxford X-Max^N detector for energy dispersive X-ray (EDX) spectroscopy. More details on the preparation procedure of tensile specimens on push-to-pull (PTP) devices for in situ tensile experiments inside the TEM are given in Sections “Workflow for In Situ Tensile Testing of Silica Glass Membranes in the TEM” and “Site-Specific Preparation of Silica Glass Tensile Specimens.”

Transmission electron microscopy was performed by using a Titan³ Themis 300 (FEI Company, Netherlands) operated at 300 keV acceleration voltage. The microscope is equipped with a high-brightness field-emission gun (X-FEG), a monochromator system, two C_s-correctors (probe and image side) from Corrected Electron Optical Systems (CEOS GmbH, Germany), a Super-X detector for EDX spectroscopy, a Gatan Imaging Filter Quantum for electron energy loss spectroscopy (EELS), a high-angle annular dark-field (HAADF) detector for atomic number contrast imaging, and a 4K CMOS (complementary metal-oxide semiconductor) camera. In situ tensile testing was performed with the PI95 TEM Picoindenter^® (Hysitron, Inc., USA) and PTP devices inside of the Titan³ Themis 300. Comparable experimental setups have been recently used for in situ tensile testing of individual Ag nanowires (Kobler et al., 2015) and in situ tensile experiments of Au and alloyed Cu thin (Liebig et al., 2016). In the present study, the PI95 holder was operated under displacement control, while experiments are carried out with and without e-beam irradiation. In our previous studies on nanoscale silica spheres, we noticed a strong impact of e-beam irradiation on their mechanical behavior inside the TEM (Mačković et al., 2014, 2016) and SEM (Romeis et al., 2012). Similarly, in the present work, the impact of e-beam irradiation was investigated in case of silica glass. Quantitative nanomechanical data obtained from the tensile experiments are then directly correlated with in situ movie data recorded with the CMOS camera of the TEM.

Raman spectroscopy was performed using a LabRAM HR—Evolution (Horiba Scientific, USA) micro Raman setup. The excitation wavelength was 532 nm (frequency-doubled Nd-YAG laser source). Grating, slit, and confocal hole were 1,800 l mm⁻¹, 100 μm, and 100 μm, respectively. The spectral resolution was ~0.5 cm⁻¹. The laser power in plane of the samples was 20 mW.

The membrane thickness was measured with an atomic force microscope (AFM, NanoWizard3, JPK Instruments, Berlin, Germany) in the intermittent contact mode using a NSC15/AlBS cantilever tip (MikroMasch^®, Bulgaria; resonance frequency 325 kHz, spring constant 46 N m⁻¹, nominal tip radius <10 nm). Measurements were performed with the fast scanning direction perpendicular to the membrane edge for three different membranes that have been prepared on PTP devices. For image processing and thickness determination, the software JPKSPM Data Processing was used. After background subtraction and inclination correction, a histogram-based approach was used for thickness evaluation.

In the present work, silica glass was chosen as a model system, in order to get more insights on the intrinsic strength, deformation behavior, and elastic properties of amorphous materials. Furthermore, nanoscale specimens and nanomechanical testing approaches are one way to avoid/minimize the general problem of surface defects, which control the strength of brittle materials and usually prevent the exploration of intrinsic effects. Hence, the aim of the present work was to develop a workflow for the site-specific preparation of thin, nanoscale silica glass specimens suitable for quantitative in situ tensile testing inside the TEM and get insights on the mechanical properties occurring on micro-/nanoscale. Figure 1 schematically demonstrates this workflow. In the first step, a piece of a thin silica glass membrane is extracted from a Pelco^® membrane (also see section Sample Preparation) by using the FIB. In the following step, the extracted silica glass membrane is transferred onto a PTP device. Next, a tensile specimen with dog bone geometry is prepared by careful FIB cutting. The PTP device together with the FIB-prepared silica glass tensile specimen is then mounted into the Hysitron PI95 TEM Picoindenter^®, and in situ tensile testing is performed inside the TEM. As a result, in situ TEM movie data of the deformation event and nanomechanical data are obtained simultaneously and can be directly correlated. In the following chapters, the single steps of the workflow are described in more detail, whereby characteristics and challenges are emphasized.

As introduced above, structural changes of amorphous silica (Primak and Kampwirth, 1968; Snoeks et al., 1994) are reported in literature upon irradiation with heavy ions. Moreover, structural changes and changes in the mechanical properties of vitreous silica (Primak and Kampwirth, 1968; Dellin et al., 1977; Storm et al., 2005) and silica glass (Yamaguchi et al., 2004; Iwata and Nishi, 2008) are reported upon irradiation with electrons on the macroscopic and microscopic scale. To account for this, we carried out a systematic analysis of the impact of Ga⁺ irradiation (during FIB preparation) and e-beam irradiation (during experiments in the TEM) on the structure of silica glass membranes. Targeted irradiation experiments are carried out with electrons (parallel illumination in the TEM) according to our previous approach (Mačković et al., 2014) and Ga⁺ (in scanning mode in FIB) with beam current densities of ~10⁻³ and ~10⁻⁶ A cm⁻² in the TEM and FIB, respectively. Identical beam current densities are also used later on during the preparation routine of silica glass membranes in the FIB and in situ tensile testing in the TEM. Raman spectroscopy is used to evaluate the structural integrity of the glass network in the silica glass membranes and to investigate possible damage induced by Ga⁺ and/or electron irradiation.

In addition and for direct comparison with the irradiated material, Raman spectra are obtained from pristine (non-irradiated) silica glass membranes. Figure 2 exemplarily shows the results from Raman spectroscopy, while in general good data reproducibility was given. The non-irradiated silica glass membranes exhibit typical Raman spectra with well-known characteristics of vitreous silica on micro- (Perriot et al., 2006) and nanoscale (Paul et al., 2015; Romeis et al., 2015). The intense band at 440 cm⁻¹ occurs due to stretching of bridging oxygen atoms between two Si atoms. As observed by Sharma et al. (1981) for vitreous silica, this intense band can also be observed at 437 cm⁻¹. The two defect lines D₁ and D₂, which are located at 495 and 606 cm⁻¹, are attributed to breathing modes of the four- and three-membered Si–O rings in vitreous silica (Galeener, 1982). The broad band at 795 cm⁻¹ is a ω3 mode with little transverse optical-linear optical splitting (Galeener, 1979). The band at 1055 cm⁻¹ can be attributed either to the high frequency antisymmetric Si-O stretching mode (McMillan et al., 1984) or to a transverse optical mode (Galeener, 1979) occurring in vitreous silica. In case of all three Raman spectra, a relatively weak signal is detected at ~1,060 cm⁻¹, which might arise from sp³-type vibrations of hydrogen-free carbons (Ferrari and Robertson, 2001). The Raman spectrum of the e-beam irradiated silica glass membrane is comparable to the one of the non-irradiated membrane, which indicates that no major structural changes occur upon e-beam irradiation. In our previous work (Paul et al., 2015), targeted e-beam irradiation experiments inside the TEM and subsequent Raman spectroscopy was performed on 200-nm diameter vitreous silica spheres (heat-treated SFB-type silica spheres (Romeis et al., 2014; Paul et al., 2015)). It was found that e-beam irradiation does not significantly alter the structure of vitreous silica spheres [only minor densification ≤2 vol.% (Paul et al., 2015) was observed], even though higher beam current densities of ~10⁻² A cm⁻² were used during irradiation, compared to beam current densities of ~10⁻³ A cm⁻² used for irradiation of the glass membranes in the present study. In this context, Luo et al. (2016) irradiated silica glass fibers with electrons inside the TEM by using a beam current density of 2.5 × 10⁻³ A cm⁻² for 1 min, and subsequently performed EELS measurements on the irradiated material. No differences in the chemical composition and electronic structure (or bonding characteristics) are noticed upon irradiation with the e-beam, in comparison to the non-irradiated silica glass fibers. These results confirm that moderate e-beam irradiation in the TEM does not significantly change the structure and chemistry of silica glass. In contrast, in the case of Ga⁺ irradiated membranes, we observed that the intense band, which in the case of pristine silica glass is usually located at 440 cm⁻¹, shifts to higher wavenumbers (see Figure 2). This phenomenon can be attributed to a densification of the silica glass network structure. Similar features are observed in the Raman spectra by Perriot et al. (2006) in indentation-densified amorphous silica and by Romeis et al. (2015) in compression-densified vitreous silica spheres. Previously, compression-induced densification of silica glass was observed by Tomozawa et al. (1998) in uniaxial compression experiments and infrared spectroscopy measurements and by Mochizuki and Kawai (1972) in pressure-densified vitreous silica. It has been explained to be accommodated by a decrease of the bond angles of inter-tetrahedral Si-O-Si bonds of the silica glass network structure (Mochizuki and Kawai, 1972; Tomozawa et al., 1998). In an earlier work, densification induced by heavy Xe⁺ irradiation was reported for 1.85 μm thick amorphous silica films (Snoeks et al., 1994). However, Ga⁺ irradiation-induced densification of nanoscale silica glass, which we observe here by Raman spectroscopy, has so far not been reported in literature.

In the next step, dog bone-shaped tensile specimens are prepared directly on the silica glass film of the Pelco^® membrane (see Figures 3A,B). During preparation of the silica glass membrane with the Ga⁺ beam redeposition of Ga in form of particles on the silica glass membrane occurred, which is evident from scanning TEM (STEM)-EDX measurements shown in Figures 3C–F. Furthermore, during FIB preparation, the silica glass membranes are cut “blindly,” meaning that direct irradiation of the dog bone-shaped tensile specimen was prevented during the full preparation procedure, which avoids/minimizes FIB-induced preparation artifacts. By using STEM-EDX analysis, we evaluated the Ga content after a typical FIB preparation of the glass membranes. The results are shown in Figures 3C–F. Measurements are performed in the middle and edge regions of the membranes, as exemplarily shown in Figures 3C,D and Figures 3E,F, respectively. In general, in the middle region of the membranes, a higher Ga content is observed. Furthermore, tilting experiments of the membranes by ±45° in the TEM showed that the Ga-enriched particles are mainly located on the bottom side of the membranes (see Figures 3G–J).

The specimen thickness is one of the crucial dimensions needed for the evaluation of the mechanical properties from in situ tensile experiments. In order to determine the membrane thickness, TEM and complementary (tapping mode) AFM measurements are performed. For the TEM measurement, silica glass membranes are transferred from a Pelco^® grid onto a silicon wafer, a Pt protective layer is deposited and cross-sections are prepared by FIB lift-out (see Figure 4A).

Multiple STEM-EDX line scans are performed perpendicular to the Pt–SiO₂–Si interfaces along the length of the FIB cross-section, which show an average thickness of 107.15 ± 1.05 nm. Figure 4B exemplarily shows a silica glass membrane with a thickness of 108 nm. In the case of AFM measurements, the thickness is directly measured on silica glass membranes that were already prepared on PTP devices. Figure 4C shows exemplarily a height profile of a section of a silica glass membrane viewed from top. In order to evaluate the thickness, Gauss fits are applied to the original experimental data in the ranges indicated in the intensity—thickness diagram shown in Figure 4D. The thickness of the membrane shown in Figure 4D is 92.4 nm. For two other membranes thickness values of 99.3 and 103.6 nm are obtained by AFM. Based on AFM and FIB cross-section measurements, an average thickness value of 100.6 ± 6.35 nm is obtained and used for further calculations. For more accuracy in future experiments, it is essential to determine the membrane thickness of each individual membrane before the tensile experiments.

For in situ tensile experiments in the TEM, the silica glass membranes are prepared from Pelco^® membranes (see section Sample Preparation and SEM image in Figure 3A) on PTP devices. Figure 5 shows an overview SEM image of a PTP device. The region where the specimen can be prepared is localized between a mobile and fixed part of the PTP device. By pushing with the indenter on the mobile part of the PTP device, the springs are deflected and a tensile force is applied to the specimen. Figure 6 exemplarily shows the site-specific preparation procedure of a silica glass membrane on the PTP device for in situ tensile testing in the TEM. A piece of the silica glass membrane is first cut out of the Pelco^® membrane using the FIB and is further on transferred onto the PTP device with the micromanipulator.

Carbon is used to attach the membrane to the manipulator and pure adhesion is used to fix the membrane to the PTP device. After placing the membrane onto the PTP device, the Ga⁺ beam is used to separate the membrane from the manipulator. From the practical point of view, it should be noted that in some case the silica glass membrane lifted up when detached from the manipulator and the transfer failed. The deposited silica glass membrane is cut step-wise starting from the outer and going to the inner part of the membrane. Finally, by using a self-designed mask for cutting (8-bit gray scale image), the tensile specimen is milled in a single step to the final dog bone-shaped geometry.

As proof of principle, in situ tensile experiments of silica glass membranes are performed at beam-on and beam-off conditions inside the TEM. Figure 7 shows in situ movie frames of a tensile experiment performed interruptedly under moderate e-beam irradiation (beam current density ~10⁻³ A cm⁻²) in the TEM. The corresponding in situ Movies S1–S3 (Video S1–S3) in Supplementary Material show the first, second, and third cycle of tensile experiments, respectively, performed on the same silica glass membrane. As can be seen in Figure 7 and the in situ TEM movies, large strains up to 41% are achieved without fracture of the glass membrane. At the end of the third run, a crack was initiated in the upper edge region of the membrane (see left side in the necked region of the silica glass membrane in Movie S3 (Video S3) in Supplementary Material) and would most probably have caused the failure of the membrane upon further tension. Moreover, the glass membrane is deformed relatively uniformly along the gauge length indicated with the dashed rectangles, showing only minor necking (see left side of the membrane in Figures 7B–D). So far, no quantitative nanomechanical data are reported in literature for nanoscale silica glass pulled under e-beam irradiation inside the TEM. In the present work, the glass membrane (average diameter ~221 nm, thickness ~100 nm) that was pulled under moderate e-beam irradiation in the TEM, exhibits a yield stress (σ_y) of 1.9 GPa measured at 0.2% plastic strain. After the first pulling cycle (at a strain of 25%), a tensile stress of 6.8 GPa is observed. The Young’s modulus (E) of 49 GPa, which is evaluated from the linear fit in the low deformation regime of the engineering stress-strain curve, is clearly below values known for bulk fused silica (72–76 GPa) (Adachi and Sakka, 1990), most probably caused by the e-beam. Moreover, this E value of 49 GPa has to be interpreted with care, since e-beam irradiation in principle could induce plastic flow of silica glass already in the early stages of deformation. Hence, in future, loading-unloading experiments need to be exploited under e-beam irradiation in order to determine the elastic deformation regime. It is well known from literature (Zheng et al., 2010; Yue and Zheng, 2014) that tensile testing of amorphous silica fibers under e-beam irradiation in the TEM can lead to superplastic deformations (more than 200% (Zheng et al., 2010)). In our previous works (Mačković et al., 2014, 2016), we observed the E of as-synthesized SFB-type nanoscale silica spheres (sol-gel-derived silica) to be strongly dependent on the e-beam irradiation dose used before compression of the spheres inside the TEM. An E value of ~21 GPa is observed in case of low dose e-beam irradiation (beam current density of ~10⁻⁴ A cm⁻²), while at e-beam irradiation with higher dose (beam current density of ~10⁻² A cm⁻²) an E of ~ 45 GPa is observed (Mačković et al., 2014). Also in this case, the E values of 21 and 45 GPa are significantly lower than the value known for bulk fused silica (72–76 GPa) (Adachi and Sakka, 1990). However, they are typical for such SFB-type silica particles, as reported for sol-gel-derived hollow silica spheres and silica microcapsules by Zhang et al. (2009, 2010) and submicron-sized silica spheres by Romeis et al. (2012, 2014) and Paul et al. (2015), respectively. In difference to this, Romeis et al. (2014) and Paul et al. (2015) reported E values of 71 ± 1.4 GPa for heat-treated SFB-type silica spheres approaching the values known for bulk fused silica (72-76 GPa) (Adachi and Sakka, 1990). Similar values of the E are observed by Paul et al. (2015) from in situ compression experiments for 200 nm-sized vitreous silica spheres in the SEM and TEM with the e-beam blanked.

As introduced above, the high energy e-beam generates structural and bonding defects in the amorphous lattice and facilitates the bond-switching mechanism, which favors the rotation and migration of atomic clusters to accommodate plastic flow in silica (Zheng et al., 2010). Hence, the e-beam irradiation causes a softening of the silica network structure both reducing its rigidity and promoting plastic flow. This could explain the lower E that we observed in case of in situ tensile testing of a silica glass membrane under e-beam irradiation, in comparison to the tensile test performed at beam-off conditions.

In contrast, in situ tensile testing of a silica glass membrane (diameter ~351 nm, thickness ~100 nm) at beam-off conditions inside the TEM (see Figure 8) results in a high amount of elastic deformation, followed by brittle fracture at a fracture strain of ~ 8%. To obtain this value, the initial slip of the membrane is subtracted, as indicated in Figure 8E. The slip off of the membrane is also confirmed by SEM analyses after the tensile experiment. The fracture surface of the membrane appears to be relatively smooth, as can be seen in Figures 8C,D. The particle and/or island-like dark contrasts visible in Figures 8C,D are Ga-enriched particles redeposited during FIB preparation. In contrast to the tensile experiment performed at beam-on conditions (see Figure 7), here the deformation took place almost over the whole length of the tensile specimen (see the red dashed lines in Figure 8B), which indicate the initial shape of the membrane. The σ_f of 5.4 GPa is slightly below values reported for silica glass fibers with comparable dimensions (Brambilla and Payne, 2009), but clearly exceeds tensile strength values reported for microscale silica fibers with E-glass (ranging from ~2.1 to 3.4 GPa) and basaltic glass compositions (ranging from ~1 to 4 GPa) (Lund and Yue, 2010). The E of 76 GPa is in the range of values known for bulk fused silica (Adachi and Sakka, 1990). Silva et al. (2006) performed bending experiments on silica nanowires with diameters ranging from 280 to 1950 nm and observed values of E similar to ours (measured at beam-off conditions). Just recently, Luo et al. (2016) performed in situ tensile experiments with nanoscale silica glass fibers (with diameters well below 50 nm) at beam-off conditions in the TEM and performed complementary molecular dynamics simulations. The authors noticed plasticity at length scales of 10 nm and a significant brittle-to-ductile transition upon reducing the fiber diameter below 18 nm (Luo et al., 2016).

In a third experiment (see Figure 9), a silica glass membrane was pulled under e-beam irradiation (beam current density of ~10⁻² A cm⁻²), while during elongation the e-beam is switched off at a strain of ~20%. This kind of experiment directly evidences the impact of e-beam irradiation on the deformation behavior of silica glass inside the TEM. The corresponding in situ TEM Movie S4 (Video S4) is available in the Supplementary Material. As can be seen in the Movie S4 (Video S4) in Supplementary Material and in Figure 9D, during the elongation under e-beam irradiation, the silica glass membrane exhibits an elastic/plastic deformation behavior. After an initial elastic deformation regime, it yields at a strain of ~6%, while reaching a σ_y of ~2.2 GPa. The following plastic deformation regime suddenly ends at the point where the e-beam is switched off (see Figure 9D), where a stress of 4.5 GPa is reached. The absence of the e-beam is directly correlated with an instantaneous change in the mechanical response of silica glass. Starting from the point where the e-beam is blanked, the stress (and the force signal in the Movie S4 (Video S4) in Supplementary Material) increases in a linear manner, indicating a purely elastic behavior, finally reaching a value of 13.8 GPa where the membrane fractures. After subtracting the fraction of the initial e-beam-assisted deformation regime, a σ_f ~9.4 GPa (at ~0.12 fracture strain) and an E value of 78 GPa are observed for the silica glass membrane at beam-off conditions. The σ_f is in the range of values of silica glass fibers with comparable dimensions reported by Brambilla and Payne (2009) and again clearly exceeds values reported for microscale silica fibers by Lund and Yue (2010). The E of 78 GPa is in the range of values known for bulk fused silica (Adachi and Sakka, 1990) and also in the range of values reported by Silva et al. (2006) for silica nanowires with comparable dimensions. Moreover, the E value also agrees well with the one observed for the silica glass membrane pulled at beam-off conditions in the TEM (see Figure 8 and text above). This experiment directly demonstrates in situ how the e-beam irradiation in the TEM can be used to control the mechanical behavior of silica glass. Just recently, we observed a similar behavior in case of in situ compression experiments of nanoscale SFB-type silica spheres in the TEM (Mačković et al., 2016). Upon switching off the e-beam during in situ compression in the TEM, the deformation behavior of silica spheres switched from elastic/plastic to completely elastic (Mačković et al., 2016), similar to what we observed here for silica glass membranes. It has to be noted that the σ_max value of 4.5 GPa obtained at ε_max ≈ 0.2 during the beam-on deformation regime of the tensile experiment shown in Figure 9 is slightly lower in comparison to the stress value of ~5.4 GPa observed at a comparable plastic strain in the beam-on tensile test shown in Figure 7. The slight differences in the stress values are possibly caused by the variation of the membrane thickness, as we observed by AFM measurements and TEM analysis of the FIB cross-section.

Even though in the present work, based on the experimental setup, e-beam irradiation is not expected to cause significant heating of the silica glass specimens, future work is planned to check this systematically as a case study. As shown just recently by Niekiel et al. (2016), parallel beam electron diffraction can be used in situ in the TEM to perform local temperature measurements with high precision. A similar experimental setup is currently being developed to assess quantitative data of a possible temperature increase upon irradiation with electrons in the TEM. Furthermore, it is expected that silica glass shows an e-beam dose-dependent mechanical response during tensile testing in the TEM, similarly to our observations discussed in the text above for nanoscale SFB-type silica spheres in the TEM (Mačković et al., 2014, 2016). To account for this, in situ tensile experiments carried out at different e-beam conditions in the TEM are planned in near future.

As a final remark, it should be noted that all in situ tensile experiments showed good reproducibility.