Chemical and Bonding Analysis of Liquids Using Liquid Cell Electron Microscopy

formatting your Liquid cell transmission electron microscopy (TEM) has become an essential tool for studying the structure and properties of both hard and soft condensed matter samples, as well as liquids themselves. Liquid cell sample holders, consisting often of two thin window layers separating the liquid sample from the high-vacuum of the microscope column, have been designed to control in-situ conditions, including temperature, voltage/current or flow through the window region. While high-resolution and time-resolved TEM imaging probes the structure, shape and dynamics of liquid cell samples, information about the chemical composition and spatially-resolved bonding is often difficult to obtain due to the liquid thickness, the window layers, the holder configuration or beam induced radiolysis. In this article, we provide an overview of different approaches to quantitative liquid cell electron microscopy, including recent developments to perform energy dispersive X-ray (EDS) and electron energy-loss spectroscopy (EELS) experiments on samples in a liquid environment or the liquid itself. We will cover graphene liquid cells and other ultra-thin window layer holders. aberration corrected research focuses on atomic resolution electron tomography, scanning nanodiffraction (4D-STEM), in situ liquid cell electron microscopy and 2D/3D image analysis. on ultra-high energy resolution monochromated EELS analysis of infrared phonons, plasmons, polaritons, and molecular vibrations in a high-spatial resolution aberration-corrected STEM. on in-situ characterization of materials using aberration-corrected scanning transmission electron microscopy and electron spectroscopies. recent years, Klie developed novel approaches of electron energy-loss spectroscopy (EELS) to study materials, including nano-scale thermometry using low-loss EELS, or 2-dimensional layer liquid cell to characterize water, biological systems and solid-liquid interfaces.


Introduction
Liquid cell transmission electron microscopy (TEM) has a long history in quantitative materials science and life-science microscopy and has evolved significantly since the early work by prominent pioneers of the field, such as Ruska and Marton, in the early 1930s [1] and 1940s [2][3][4]. Early liquid cell TEM required dedicated instruments, where the liquid sample was either directly exposed to the vacuum (open-cell) or separated by a thick window layer (usually 100's of thick metal foil, such as Al, or thick plastic layers) capable of withstanding the pressure differential (closed-cell). Over the last 20 years, these approaches have been significantly improved and were used to successfully characterize the growth of Si nanowires from a liquid phase [5] and identify the dynamics of Cu plating on Au. [6] However, both open-and closed-cell approaches severely limit sample choice; either to low-vapor pressure liquids that can withstand the vacuum in the TEM, or high-contrast samples with sufficient signal to overcome the thick liquid and window layers. Resolution was also limited by multiple scattering.
Furthermore, conventional analytical approaches, such as energy-dispersive X-ray spectroscopy (EDS) or electron energy-loss spectroscopy (EELS), are also precluded in dedicated liquid cell instruments due to the small differential pumping apertures in an open-cell microscope and the thick window layers blocking most of the photons or inelastically scattered electrons.
The development of Si/SiNx in-situ heating/biasing holders in the early 2000s [7] (see timeline of selected liquid cell development in Figure 1) spurred the implementation of liquid cells using modular side-entry stages in conventional and Ercius, Hachtel, Klie/Sept. 2020 3 aberration-corrected TEMs. The resolution of these in-situ liquid cell experiments, starting in 2009, reached 1 nm as the result of significant reduction in the SiNx window layer thickness (~25-50 nm) and the dramatic reduction in the liquid layer thickness. [8] In addition to higher resolution imaging of colloidal nanoparticles in solution, the ability to flow and mix solvents also allowed for the observation of growth dynamics and imaging of whole cells in a liquid environment.
The need for more analytical capabilities was met by redesigning the window shape to support ultra-thin SiNx layers. The thinner window layers now also allowed for selective EELS quantification of samples, including energy-filtered TEM, for materials which avoid the silicon (Si) and nitrogen (N) core-loss excitations (~100 eV and ~400 eV respectively). When combined with the ability to apply a bias and using continuous liquid flow, the plating and stripping behaviors of anodes in rechargeable batteries became a focus of in-situ electrochemical characterization in a TEM (see Ref. [9] as one of many examples). Yet, the window layer thickness and sensitivity of the liquids to the electron beam did not allow for atomicresolution imaging and spectroscopy.
The development of ultra-thin windows layers, including graphene and BN, significantly reduce liquid layer thickness and largely eliminated the scattering by the window-layer resulting in both higher resolution imaging and spectroscopy of in-situ imaging. In addition, ultrathin graphene seems to increase the tolerance of the encapsulated materials to the high-energy electron beam and reduce the formation of bubbles from radiolysis. [10,11] The ability to perform electron energy-loss spectroscopy (EELS) or EDS during in-situ experiments has the Ercius, Hachtel, Klie/Sept. 2020 4 potential to overcome many of the current limitations for quantitative in situ TEM, providing information about beam-sample interactions and local reactions during imaging. In this article, we will review recent developments in quantitative and analytical liquid-cell TEM with a focus on quantitative chemical and bonding analysis enabled by innovations in holder and window layer designs, as well as new detector and electron source technologies. We will also present our vision for next generation analytical liquid phase microscopy exploiting parallel electron microscopy breakthroughs in tomography, direct electron detection, and monochromation to probe new aspects and dimensions of novel materials in liquid cell samples.

EDS and EELS in SiN-based liquid cells
While high-resolution and dynamic imaging of nanoparticles and whole cells flourished using liquid cell TEM, chemical and bonding analysis in SiNx liquid cells faced several challenges. EELS was primarily used to measure the relative total sample thickness and the quantify the extent of window bowing.
[12] Examples of early quantitative measurements include low-loss valence-EELS, as well as coreloss spectroscopy of the O K-and Fe L-edges, to examine the de/intercalation of Li-ion battery cathodes in electro-chemical cells. [13] Until recently, the limitations to EDS in liquid cells were primarily due to geometric constraints of the Si chips used to support the SiNx window layers and the TEM holder. Specifically, the cutouts in the Si chips produced by KOH and HF etching produce a 52° take-off angle, blocking the majority of X-rays generated by the Ercius

2-D window-layer cells for high spatial/energy resolution
Since all material in the beam path interacts with the electron beam, often overwhelming the signal from the material of interest due to multiple scattering, a critical problem with in situ liquid-cells is the window thickness and the large volume of liquid trapped between them. However, thinner windows can also bulge out far larger than the expected thickness.  Fig. 2b). This effect has been attributed to the high electron mobility in graphene or a catalytic reaction at the graphene/bubble interface. [11] It is also observed that the low-dose bubbles in GLCs reabsorb into the liquid, while the high-dose/larger bubbles in both GLC and BNLC remain or grow (Fig. 2e), indicating the potential to directly control radiolysis in liquid cells using the electron beam to manipulate the chemistry or pH [15]. Lastly, one can avoid direct electron beam irradiation entirely [36] while still effectively probing the sample by placing the beam ~30 nm away from the sample and allowing the beam to couple to low-energy excitations in the sample without the high energy electrons ever directly interacting with the liquid in the cell (the so called 'aloof' configuration). [37,38] Spatial resolution is governed by delocalization of inelastic scattering where the interaction volume is inversely proportional to energy loss.
[39] Next-generation electron monochromators, which can possess energy resolutions as low as 4.2 meV and significantly reduced background in the infrared, [40] allow for high-sensitivity, high precision measurements of phonons and molecular vibrations of organic molecules and other beam-sensitive 8 samples. [41][42][43][44][45] In Fig. 2f, the difference between a non-monochromated (black) and monochromated (red) EEL spectrum is shown in the infrared energy range, demonstrating the improved energy resolution and reduced background. In Fig. 2g, the aloof monochromated EEL spectrum of liquid-water in a BNLC is shown demonstrating the ability to sample the O-H molecular vibration without direct irradiation.

New detectors and approaches
Many of these new experiments and results are facilitated by tangential breakthroughs in non-liquid-cell electron microscopy techniques. Here, we outline more recent advancements, and provide perspective on their influence in liquid phase TEM.

Detectors
In situ liquid cell experiments are often focused on the dynamics of the systems of interest, meaning that time resolution is an important metric. Previously, charged coupled detector (CCD) technology limited the capabilities of in situ TEM to at best 30 frames per second (fps), but much lower for HR-TEM in the 1 frame per second range. The main limitation was in the CCD readout rate, which also required the detector was "blanked" during readout reducing the duty cycle significantly.
Blanking was often done below the sample while the sample was still illuminated and potentially changing without adding to the image signal. We are now in a new era of direct electron detectors (DED) which can be fully readout on the millisecond or less time scale with no blanking. [46] The detector point spread function (PSF) has also been improved producing sharper images [47] and enhanced sensitivity for 9 spectrometers. [48] DEDs are revolutionizing several aspects of TEM, such as cryo-EM 3D bio-structural imaging, [49,50] 4D-STEM, [51] EELS, [52] and in situ TEM is no exception. Faster data (image or spectra) captured with higher quality improves traditional imaging experiments [53,54] while also allowing new capabilities such as tomography discussed later. The speed and size of DEDs is still increasing, growing data set sizes exponentially, where 2.5 TB of data can now be collected per minute. [55] However, even higher readout speed could achieve the ultimate in in-situ imaging and analytical analysis by utilizing electron counting [48] at the dose rates needed for in situ TEM to record only electron strikes and reduce the noise in TEM images and low-loss spectra to the ultimate limit of Poisson noise. This will simultaneously improve the resolution and contrast in in situ TEM movies while also compressing the data 100x.

Beam sensitive materials
TEMs are now capable of resolving most atomic spacings in materials due to aberration correctors. The limiting factor is now the electron dose applied to the sample in both materials science and biology. [56] New developments in both liquid-cell window layer materials and faster detectors (mentioned earlier) are providing new opportunities for imaging beam sensitive materials and biostructures in the native buffer solution with higher contrast and resolution. In situ TEM imaging of biological materials while hydrated in a native buffer solution has been achieved using SiN windows using SEM, TEM and STEM. [57] The biological structures studied range from whole cells to viruses. De Jonge, for example, used STEM and Au nanoparticles as high contrast tags to image whole cells at 4 nm 10 resolution while also providing dynamical information. [58] It has also been shown in several different ways that GLCs can reduce the effects of beam damage when imaging beam sensitive materials, [59] which is critically important as resolution in TEM is now more limited by applied dose than microscope resolution. [60,61] Chemical bonding information was obtain using EELS from beam sensitive materials, such as polyphosphate nanoparticles encapsulated in sterically stabilized liposomes from samples encapsulated in graphene liquid cells.  provides the ability to determine the NPs' properties by ab initio quantum mechanical simulations. Next steps utilizing next generation DEDs and spectrometers could expand this capability to include compositional as well as structural 3D information.

Aloof Monochromated EELS
Aloof EELS, as shown in Figure 2f and 2g, allows for virtually damage free spectroscopy of even the most beam sensitive materials. This is possible due to the fact that the aloof coupling strength is much higher for low-energy infrared (IR) excitations, such as phonons and molecular vibrations, than it is for damaging highenergy (UV) excitations, needed for core-loss EELS. This can be understood by considering the radial dependence of the aloof scattering probability, where is the radial distance from the probe, is the frequency of the energy loss being considered, and v is the velocity of the primary electron [63]. For > ⁄ the interaction strength exponentially decays, meaning the intensity of damaging UV excitations attenuates long before the damage-free IR excitations.
We can use this probability to estimate the maximum potential damage-free spatialresolution achievable with monochromated EELS.
Vibrational spectroscopy is capable of reaching atomic-resolution, the same as conventional core-loss EELS, by using experimental geometries that emphasize highly-localized impact scattering in the collected EELS signal [64][65][66], but these techniques all require direct irradiation and long acquisition times and, thus, are not suitable for beam-sensitive liquid-cell samples. For damage-free aloof EELS, we can follow the formalism put forth by Egerton [63], where we imagine the probe at some impact parameter, b, away from a truncated slab and calculate radial distance from which the majority of the aloof signal originates as a function of the aloof scattering probability. Figure 3 shows this calculation for two different energy losses at 0.42 eV (O-H stretch mode in water [30]) and at 4.2 eV (where absorption in organic molecules begins to break bonds [67]), and two different impact parameters (10 nm and 30 nm). We can see that at 10 nm, the majority of the IR signal comes from a nano-sized region surrounding the probe, but that there is significant UV excitation in this area as well. By moving the probe back to a 30 nm impact parameter, the spatial resolution for the IR excitation is not significantly reduced (< factor of 2), but the ratio of IR excitation to UV excitation is improved by an order of magnitude.

Nano-reactors
Static liquid cell configurations that are optimized for spatial resolution or Alternatively, the electron beam can be used to change the local composition (or concentration of OHin water) to initiate a reaction. The high dose-rate tolerance Ercius, Hachtel, Klie/Sept. 2020 14 of samples in GLCs, or the use of damage-free aloof EELS, will then allow these reactions to be observed without negatively affecting the reaction pathways.

Outlook
This