A protocol for preparation of graphene-supported microwell liquid cells for in situ electron microscopy of gold nanocrystals from HAuCl4 precursor solution is presented. Furthermore, an analysis routine is presented to quantify observed etching and growth dynamics.
The fabrication and preparation of graphene-supported microwell liquid cells (GSMLCs) for in situ electron microscopy is presented in a stepwise protocol. The versatility of the GSMLCs is demonstrated in the context of a study about etching and growth dynamics of gold nanostructures from a HAuCl4 precursor solution. GSMLCs combine the advantages of conventional silicon- and graphene-based liquid cells by offering reproducible well depths together with facile cell manufacturing and handling of the specimen under investigation. The GSMLCs are fabricated on a single silicon substrate which drastically reduces the complexity of the manufacturing process compared to two-wafer-based liquid cell designs. Here, no bonding or alignment process steps are required. Furthermore, the enclosed liquid volume can be tailored to the respective experimental requirements by simply adjusting the thickness of a silicon nitride layer. This enables a significant reduction of window bulging in the electron microscope vacuum. Finally, a state-of-the-art quantitative evaluation of single particle tracking and dendrite formation in liquid cell experiments using only open source software is presented.
Modern materials science, chemistry and cell biology require a deep understanding of underlying dynamic processes and effects at the sub-micron scale. Despite the power of advanced optical microscopy techniques such as stimulated-emission-depletion fluorescence microscopy1, direct imaging techniques to access detailed morphologies require electron microscopy. In particular, in situ (scanning) transmission electron microscopy (S)TEM has been shown to illuminate valuable insights into process dynamics by encapsulating liquids in dedicated, vacuum-tight cells2. Various experiments such as quantitative investigations of nanostructure formation kinetics and thermodynamics3,4,5,6, imaging of biological specimens7,8,9,10 and studies of energy storage-related mechanisms11,12 along with comprehensive studies of corrosion process dynamics13 or nanobubble physics14,15,16 have unraveled many phenomena using (S)TEM that were not accessible using standard microscopy techniques.
During the last decade, two major approaches to realize in situ liquid cell TEM (LCTEM) have been established. In the first approach, the liquid is encapsulated in a cavity between two Si3N4 membranes produced via Si process technology17, whereas in the second, small liquid pockets are formed between two graphene or graphene oxide sheets10,18. The handling of both silicon-based liquid cells (SiLCs) and graphene-based liquid cells (GLCs) has been demonstrated19,20,21. Although both approaches have undergone significant improvements22,23,24,25, they still lack in the combination of the respective advantages. In general, a tradeoff exists between encapsulating the sample in often undefined graphene pockets with a small liquid volume that enables high-resolution imaging18, and well defined cell volumes resulting in thicker membranes and liquid layers, which provide an environment closer to the natural situation in bulk liquid26 at the expense of resolution2. Furthermore, some experiments depend on a liquid flow26,27 which has only been realized in SiLC architectures and requires a dedicated TEM holder28.
Here, we present the fabrication and handling of a liquid cell approach for high-performance in situ LCTEM via static graphene-supported microwell liquid cells (GSMLCs) for TEM analyses. A sketch of the GSMLC is presented in Figure 1. GSMLCs have proven to be capable of enabling in situ high-resolution transmission electron microscopy (HRTEM) results6 and are also feasible for in situ scanning electron microscopy29. Their Si technology-based frame allows for mass production of reproducibly shaped cells with tailored liquid thickness and extra-thin membranes from a single wafer. The graphene membrane covering these cells also mitigates electron beam-induced perturbations8,30,31 since the electron beam passes through the top graphene membrane first. The flat topography of the cells allow for complementary analysis methods such as energy-dispersive X-ray spectroscopy (EDXS)6 without any shadowing effects arising from the liquid cell itself, enabling a variety of high-quality in situ liquid cell electron microscopy experiments.
1. Fabrication of microwell-based liquid cell templates
2. Transfer of graphene onto TEM grids
3. Specimen preparation
4. GSMLC loading
5. TEM Imaging and video analysis
After loading of the cell, a successful graphene transfer is indicated by a differently shaded appearance on the wells under an optical microscope. This is visible, for example, in the right membrane of Figure 3c. As mentioned, it is crucial to carefully remove the TEM-grid in order to not break the thin Si3N4 layer. In case of a broken membrane, lucent and curved residuals are clearly visible in the optical microscope, as shown in the left two membranes of Figure 3c. Due to the multiple viewing areas in the utilized GSMLC design, the cell can be used as long as at least one membrane is intact. Broken membranes can be used for TEM alignment without exposing the specimen to the electron beam.
A successful encapsulation of the specimen solution can be verified during electron microscopy. Figure 5 presents individual micrographs of Supplementary Video 1, where the dissolution of an ensemble of nanoparticles and the growth of a dendritic structure is statistically evaluated in a GSMLC. Besides the drift-induced movement of the image, minor individual orthogonal particle movements are visible, indicating that particles in solution are present. Furthermore, the prevalence of particle dissolution proves that a wet-chemical reaction is present which would not be possible without a successful liquid enclosing. Other typical indications for enclosed liquids are beam-induced bubble formation19 or particle motion. The presence of Au particles in graphene-featured cells alone does not conclusively indicate a liquid environment, since the particles could also stem from the graphene-induced reduction of HAuCl440. A quantification of the oxygen peaks of the enclosed liquid via electron energy loss spectroscopy (EELS) can also be performed to verify a liquid environment41.
In order to gain insights into particle growth and dissolution kinetics, it is important to investigate each particle individually rather than to analyze the development of average parameters42. It is also crucial to exclude particles at the frame edges that are only partially captured by the camera because drift effect-related position changes of such particles might be mistaken as growth or dissolution processes. Etching is believed to be caused by oxidative species generated by electron beam-induced radiolysis43. In order to yield sufficient statistics, computational single particle tracking is required. By estimating the growth exponent α of the equivalent radius variation of individual particles over time, information of the underlying reaction kinetics can be obtained. To do so, it is possible to introduce an equivalent radius based on the projected particle area, even if not all particles are completely spherical6,44. Figure 5b shows the tracking of equivalent radii over time for six representative particles which are highlighted in Figure 5a. Figure 5c shows the distribution of α based on 73 dissolving particles from the present study. Only particles where an allometric model explains the radius decline to at least 50% (adjusted coefficient of determination) are regarded.
Furthermore, a dendrite structure emerges rapidly after about 42 s in the same well depicted in Figure 6a. Dendrite formation is another typical, well documented process in liquid cells45,46. To quantify dendrite growth, the structural outlines (see inset in Figure 6a) are analyzed. The evolution of tip radius and velocity over time (see Figure 6b,c) reveals the expected hyperbolic relationship47 (Figure 6d). Dendrite growth is caused by local supersaturation of Au-ions due to the aforementioned particle etching. In Figure 5a, it is clearly visible that particles are still dissolving whilst the oversaturated system relaxes into dendrite growth. This may be caused by local concentration variations in both the Au-ions and the oxidative species as a result of the high viscosity of the liquid in the GSMLC which has been observed before6. A detailed discussion of this phenomenon, however, is beyond the scope of this work.
Figure 1: Sketch of a GSMLC: Schematics of the structure of a graphene-supported microwell liquid cell. Reprinted from https://pubs.acs.org/doi/abs/10.1021/acs.nanolett.8b03388 6. Further permissions are to be directed to the American Chemical Society (ACS). Please click here to view a larger version of this figure.
Figure 2: Fabrication of GSMLC-frames. The fabrication process of GSMLC-frames is schematically sketched. (a) Oxidation of Si wafer after cleaning. (b) LPCVD of Si3N4. (c) Front side Si3N4 patterning by photolithography and RIE to define the cell volume. (d) Deposition of Si3N4 to form the bottom cell window. (e) Back-side lithography and RIE. (f) Bulk micromachining with KOH to create a freestanding Si3N4 membrane containing microwells. Please click here to view a larger version of this figure.
Figure 3: Transfer of few-layer CVD-graphene with PMMA protection layer onto a TEM-grid. The transfer of few-layer CVD-graphene on PMMA onto the top of a holey carbon-coated TEM-grid is displayed. (a) Immersion of the few-layer CVD graphene on PMMA in a Petri dish filled with DI water. (b) Transferred graphene/PMMA stack on a filter paper is cut into pieces suitable to cover the GSMLC-frames. (c) Re-immersion of a cut graphene/PMMA piece. (d) Transfer of the graphene/PMMA layer onto a holey carbon-coated TEM grid (e) Graphene/PMMA stack after a successful transfer. Please click here to view a larger version of this figure.
Figure 4: Removal of the top TEM grid. The drying process of a loaded GSMLC is documented with the help of an optical microscope. (a) A graphene-coated TEM grid is placed on top of the GSMLC directly after loading. The graphene layer is visible as turquoise rectangle covering all three viewing areas. Its outlines are roughly sketched by the black rectangle. (b) An almost completely adhered membrane is visible by the contrast change between the wet (dark, compare with (a)) and the adhered area (turquoise) after approximately two minutes. (c) A GSMLC after the lift-off of the TEM grid is shown, revealing two broken membranes (left and middle), and one membrane with successfully loaded and sealed microwells (right). Please click here to view a larger version of this figure.
Figure 5: Representative development of nanoparticle radii. The radius development of 183 individual particles has been tracked. (a) Image sequence taken from Supplementary Video 1. Six representative particles are highlighted. The colored circles correspond to the obtained equivalent radius. (b) Logarithmic plot of the particle radii. (c) The histogram of 73 particles where a negative allometric exponent α has been determined using an automated routine. Please click here to view a larger version of this figure.
Figure 6: Dendrite dynamics: The tip radius of five dendrite branches is analyzed. Error bars account for the respective standard deviation. (a) Image sequence taken from Supplementary Video 1 showing the emerging dendrite, which is visible after about 42 s. The inset in the right image shows the evolving dendrite contours. Here, the pink outlines correspond to 42.09 s, red to 42.7 s, and purple to 43.3 s. (b) Development of the (averaged) tip radius over time. (c) The mean tip velocity plotted over time. (d) The averaged tip radius logarithmically plotted against the averaged tip velocity, revealing a hyperbolic dependency (orange curve). Please click here to view a larger version of this figure.
Figure 7: SEM Image of a loaded GSMLC: A representative SEM image acquired in HAADF STEM mode in an SEM of a loaded GSMLC at low acceleration voltage (29 kV) is displayed. Besides the prominent 5 µm wide microwells, two partially overlapping circular holey carbon grids (2 µm diameter) stemming from the graphene transfer elucidated above is visible. The first carbon grid stems from an unsuccessful graphene transfer. It is clearly visible that the membrane shading stays mostly constant over the well region, but slightly darkens towards the well center. This accounts for weak, negative bulging. Please click here to view a larger version of this figure.
Supplementary Video 1: In situ video showing representative results of a liquid cell bright field TEM study of etching of Au nanoparticles and subsequent growth of a dendrite structure caused by supersaturation of the surrounding specimen solution. Please click here to download this file.
In contrast to commercially available liquid cells, custom-made GSMLCs have the advantage that they can be designed to fit into readily available TEM holders and do not require an expensive, dedicated liquid cell TEM holder.
The GSMLC architecture demonstrated here combines aspects of SiLCs and GLCs that could potentially lead to unique advantages. On the one hand, SiLCs allow for a precise determination of cell position and shape, but require relatively thick Si3N4 membranes to reduce bulging effects while ultimately reducing the achievable resolution. GLCs, on the other hand, exhibit exceptionally thin membrane walls consisting of graphene, yet suffer from random pocket sizes and positions. By combining these two membrane approaches via GSMLCs, the resolution limitation caused by the cell boundaries35 can be bypassed. As the well structure is fabricated directly into the Si3N4 layer, the actual Si3N4 membrane can be constructed even smaller than in SiLCs, simplifying HRTEM analyses which has already been demonstrated in GSMLCs6. Still, it should be noted that HRTEM in general is possible with SiLCs as well48. Moreover, large viewing areas can be realized without severe window bulging due to the small membrane areas of the individual specimen chambers. Thereby, bulging-related thickness increase35 can be ruled out to a large extent, as shown by Dukes et al.49. This is demonstrated in Figure 7, where a representative high-angle annular dark field (HAADF) STEM image of al loaded GSMLC is displayed. This image was acquired using a Dual-beam system. Since the image brightness acquired in this setup is directly related to the specimen thickness, it is clearly visible that the sealed microwells exhibit only small negative bulging. Kelly et al.24 have demonstrated that the negative bulging and partial well drying visible in Figure 7 depends on the well diameter. Reducing the well diameter is therefore a feasible approach to homogenize the liquid thickness even further.
Due to the equilibrium pocket shape of GLCs, the liquid thickness is also strongly site-dependent35. SiLCs follow the design of two membranes stemming from different Si wafers. By replacing the top Si3N4 membrane with graphene, liquid-cell fabrication is simplified. This means that possible delamination of two bonded Si-wafers during the subsequent wet etching steps can be avoided and the alignment of two wafer pieces during the cell loading is omitted. The flat surface on one side of this cell architecture enables complementary in situ analysis methods such as EDXS analysis of the specimen6, which is restricted in conventional SiLC architectures by shadowing effects at steep Si edges50.
Sealing prepatterned microwells with graphene on both the bottom and top well site has been demonstrated before24,25. Applying two graphene membranes may enhance the achievable resolution. A twofold graphene transfer, however, would complicate the preparation process further; especially since this has proven to be the most sensitive preparation step (see below). Furthermore, the above discussed membrane bulging is expected to be even more critical in case of two graphene membranes, because graphene is much more flexible than a Si3N4 layer. In those architectures, the microwells were constructed using sequential focused ion beam (FIB) milling. While this approach has proven to yield high-quality results, FIB milling is complicated and expensive cell production technique. Utilizing massively parallel single-shot patterning techniques that are already standard in today’s semiconductor industry such as nanoimprint- or photolithography, however, has the major advantage of being fast, cheap and scalable for mass production.
It should be noted that the approach presented here does not allow for liquid flow operation, which is achievable by other designs28. Since the loading and liquid volume are comparable for GSMLCs and GLCs, a contamination of high vacuum due to rupture of the membrane can be avoided19. This eliminates the need for a cumbersome seal check. Though the advantages of SiLCs and GLCs have been combined, the disadvantages of both approaches are still present in GSMLCs. The fabrication of the cells requires a clean room infrastructure for silicon technology, which is not necessarily present in TEM laboratories. In addition, the liquid loading is not trivial. It requires a dedicated training, similar to graphene cells. This, however, is also true for commercially available systems. Here, the most sensitive preparation step is the TEM-grid removal after the graphene transfer because rash movements or jittering is likely to break the Si3N4 layer. The redundant membrane windows, however, enhance the chances of preserving at least one membrane area. As a consequence, the yield (amount of operable GSMLC chips) achieved by a trained experimenter is three out of four6, and thus exceeds the one achieved with graphene-based cells (one to two out of four)19.
As with GLCs, the liquid encapsulation in GSMLCs is based on van-der-Waals interactions18. Consequently, interface contamination could lower the success rate in processing of GSMLCs19. Furthermore, depending on the Hamaker constant of the to-be-encapsulated liquid phase, the wetting characteristics during the loading procedure (and thus the achievable yield) may differ51 and therefore the preparation can be complicated. Our experience shows that this is the case if, for example, amphiphilic species are present.
The GSMLC architecture enables flexible configuration of well-depths, allowing for adaptation to various experimental prerequisites. Moreover, the architecture is suitable for electron tomography investigations over a broad tilt-angle range of ±75 °, which would also allow for in situ electron tomography52. Therefore, in situ and post mortem tomography of specimen in liquid could also be established with GSMLCs.
The authors have nothing to disclose.
We thank Tilo Schmutzler for the preparation of the HAuCl4 solution. Furthermore, we thank R. Christian Martens for proof reading. Financial support by the German Research Foundation (DFG) via the Research Training Group GRK 1896 “In situ microscopy with electrons, X-rays and scanning probes” and through the Cluster of Excellence EXC 315/2 EAM “Engineering of Advanced Materials” is gratefully acknowledged.
Acetone | VWR Chemicals | 50488858 | VLSI |
Deionized water | own production | ||
Dumont Anti-Capillary tweezers | Carl Roth GmbH + Co. KG | LH72.1 | 0203-N5AC-PO Dumoxel alloyed |
Ethanol | VWR Chemicals | 85651.360 | VLSI |
FIJI Is Just ImageJ | FIJI.sc | Version 1.51 | |
Gold Quantifoil, Amorphous Carbon TEM Grids | Plano GmbH | S173-8 | R 2/2 Au 300 mesh |
HAuCl4 · 3 H2O crystal | Alfa Aesar | 36400.06 | 5 g |
Jupyter Notebook | Project Jupyter | Version 5.7.2 | |
Matplotlib-Package | John Hunter, Darren Dale, Eric Firing, Michael Droettboom and the Matplotlib development team | Version 3.0.2 | |
NumPy-Package | NumPy developers | Version 1.15.4 | |
Pandas-Package | AQR Capital Management, LLC, Lambda Foundry, Inc. and PyData Development Team | Version 0.23.4 | |
Python | Python Software Foundation | Version 3.7 | |
Scipy-Package | SciPy developers | Version 1.1.0 | |
Seaborn-Package | Michael Waskom | Version 0.9.0 | |
Si wafer | Siegert Wafer GmbH | Thin silicon (100) wafer 175 +/-5 µm, 4", p-type, boron doped (1-30 Ohm cm), double-sided polished | |
single tilt TEM holder | Philips | Ensure that cell fits | |
Transmission Electron Microscope | Philips | CM 30 (S)TEM | 300 kV |
Trivial Transfer Graphene | ACS Material | TTG60011 | PMMA-covered, 6 — 8 MLs |