Liquid-Phase Electron Microscopy with Controllable Liquid Thickness

Jun 19, 2019 - Liquid-phase electron microscopy (LPEM) is capable of imaging nanostructures and processes in a liquid environment. The spatial resolut...
0 downloads 0 Views 2MB Size
Download PDF
Letter Cite This: Nano Lett. 2019, 19, 4608−4613

pubs.acs.org/NanoLett

Liquid-Phase Electron Microscopy with Controllable Liquid Thickness Sercan Keskin,† Peter Kunnas,† and Niels de Jonge*,†,‡ †

INM − Leibniz Institute for New Materials, D-66123 Saarbrücken, Germany Department of Physics, Saarland University, D-66123 Saarbrücken, Germany

‡

Downloaded via UNIV OF SOUTHERN INDIANA on July 23, 2019 at 04:27:29 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

S Supporting Information *

ABSTRACT: Liquid-phase electron microscopy (LPEM) is capable of imaging nanostructures and processes in a liquid environment. The spatial resolution achieved with LPEM critically depends on the thickness of the liquid layer surrounding the object of interest. An excessively thick liquid results in broadening of the electron beam and a high background signal that decreases the resolution and contrast of the object in an image. The liquid thickness in a standard liquid cell, consisting of two liquid enclosing membranes separated by spacers, is mainly defined by the deformation of the SiN membrane windows toward the vacuum side, and the effective thickness may differ from the spacer height. Here, we present a method involving a pressure controller setup to balance the pressure difference over the membrane windows, thus manipulating the shape profiles of the used silicon nitride membrane windows. Electron energy loss spectroscopy (EELS) measurements to determine the liquid thickness showed that it is possible to control the thickness precisely during an LPEM experiment by regulating the interior pressure of the liquid cell. We demonstrated atomic resolution on gold nanoparticles and the phase contrast using silica nanoparticles in liquid with controlled thickness. KEYWORDS: Liquid phase electron microscopy, liquid cell, EELS, pressure controller, atomic resolution, phase contrast

L

liquid enclosing membranes6,9,19−21 It is then practically impossible to achieve atomic resolution on metal nanoparticles or phase contrast on low atomic number materials unless a gas bubble is created, below which a thin liquid layer resides that is possibly too thin for the system under investigation and is, moreover, rather difficult to control and maintain constant.22 A monolithic liquid cell constructed with pillars to bind the top and bottom SiN membranes to minimize bulging was reported to achieve atomic resolution.23 But such a system is difficult to load with a sample and is not available to a broad audience. Another option to construct a liquid cell is via a graphene enclosure providing the highest possible resolution.10,24 But for this design also, the liquid thickness varies with the position in the sample, and this type of liquid cell is less versatile for experiments compared to a liquid cell using microchips.10,25,26 Various other types of liquid cells have been reported as well but are less common, such as liquid cells with silicon oxide, carbon or polymer windows.27−29 Here, we introduce a pressure controller setup capable of adjusting the liquid thickness in an LPEM experiment. The schematic illustration of the setup is shown in Figure 1a. The liquid cell is positioned in the tip of a TEM holder and filled

iquid-phase electron microscopy (LPEM) is a rapidly developing microscopy technique that is used to obtain morphological information about materials at the atomic scale in liquid, acquire time-resolved data from dynamic processes, and observe biological specimens such as proteins and eukaryotic cells in their natural liquid environment.1,2 In an LPEM experiment, both the spatial and chromatic broadenings of the electron beam in the liquid layer surrounding the object of interest limit the attainable spatial resolution, and so a thin liquid is desirable.3,4 A thin sample is also needed to allow imaging at a low electron dose, considering that the liquid specimens and also processes under examination are typically sensitive to electron beam irradiation.5−8 However, an infinitely thin liquid layer does not necessarily represent a realistic environment of the sample under investigation. A thickness in the range 0.1−0.3 μm strikes a balance between a representative liquid thickness and high-resolution TEM of a numerous relevant samples,9 such as a liquid containing nanocrystals for studying their growth,10,11 an in situ battery study,12,13 and biological samples including viruses and protein complexes.14−17 A standard liquid cell consists of a stack of two silicon microchips supporting thin, electron transparent silicon nitride (SiN) windows between which the liquid sample is enclosed. In theory, the liquid thickness can be set by means of a spacer between the microchips.18 But the liquid thickness is often found to vary by over a factor of 5 due to bulging of the © 2019 American Chemical Society

Received: April 16, 2019 Revised: May 25, 2019 Published: June 19, 2019 4608

DOI: 10.1021/acs.nanolett.9b01576 Nano Lett. 2019, 19, 4608−4613

Letter

Nano Letters

windows were assembled with 0.5 μL sample solution between them and then connected to the pressure controller. The liquid was pure water with dissolved citrate functionalized gold nanoparticles (AuNPs) of 20 nm diameter. Electron energy loss spectroscopy (EELS) was used to measure the liquid thickness in scanning transmission electron microscopy (STEM) mode at a pressure range of 0.05−1.0 bar. Figure 2a shows a STEM image of the liquid cell; the edge of

Figure 2. Controlling the liquid thickness in LPEM. (a) Lowmagnification scanning transmission electron microscopy (STEM) image of a region of the window with gold nanoparticles (AuNPs). This region was used to measure electron energy loss (EEL) spectra. EEL spectra were collected along the yellow line, from 0 to 9 μm toward the window center. The silicon edge of the SiN window was marked with a dashed line. (b) Measured liquid thickness at a pressure range of 0.05−1.0 bar using EELS log-ratio technique as a function of the distance from the edge of the SiN window.

Figure 1. Pressure controller setup for liquid-phase electron microscopy (LPEM). (a) Schematic representation of the vacuum system to control the internal pressure of the liquid cell (LC) in the tip of a liquid flow specimen holder. The system comprises a pressure gauge (P), a needle valve (NV), a plug valve (PV), and a roughing pump (RP). Arrows are pointing the direction to the vacuum pump. (b) Schematic of the cross-sectional view of a liquid cell showing the flat and bulged configurations of the silicon nitride (SiN) membranes with straight and dashed lines, respectively (not drawn to scale). The electron beam passes through the liquid cell. (c) Photograph of the pressure controller setup with the in situ liquid TEM holder.

the SiN window is seen at the bottom, and several AuNPs are distributed over the field of view. We collected EEL spectra over a line perpendicular to the edge of the window and repeated the experiment for internal pressures ranging from 1.0 to 0.05 bar. The thickness of the water layer enclosed by SiN windows was calculated from the EEL spectra using the intensity ratio between zero-loss peak and inelastically scattered electrons (see Supporting Information) from the measured ratio of the thickness and the inelastic mean free path length λ. We determined the measurement error of the EELS measurements at three different pressures of 0.05, 0.5, and 1.0 bar, resulting in a liquid thickness range of 0.23−072 μm. It was estimated by others that the accuracy of the method is 20%, representing a systematic error in the calculation of the thickness from the measured ratio of the thickness over λ due to inaccuracy originating from the mathematical model and the used value for λ since factors such as multiple scattering are not considered.31 The relative error between measurements at the same thickness, however, was less than 5% (Figure S1). Comparing relative thicknesses for the same sample for which one value of λ is used can thus be accomplished with higher accuracy than 20%. As seen in Figure 2b, the membranes are considerably bulged at 1.0 bar, resulting 0.65 ± 0.13 μm deflection additional to the spacer thickness of 0.2 at 9 μm far from the membrane edge. The liquid thickness was gradually reduced by decreasing internal pressure down to 0.05 bar, where the measured thickness of liquid (0.26 ± 0.05 μm) close to window center almost matched the spacer thickness (0.2 μm), suggesting a near-flat window geometry with only (0.26 − 0.20)/2 = 0.03 μm deflection for each window. Besides reducing the bulging significantly, this situation also reflects a more uniform liquid thickness throughout the imaging area

with liquid, and the liquid chamber in the tip is connected via tubing in the holder to a gas handling system outside of the microscope maintaining a certain set pressure. The degree of bulging of the SiN membranes is influenced by the tension in the membrane pulling it flat, and the pressure on the interior liquid pushing the membrane outward. The liquid thickness is thus controlled via the pressure applied to the tubing segment outside the microscope. The system was tested for a spacer thickness of 0.2 μm between the silicon microchips. Highresolution TEM was demonstrated for both gold and silica (SiO2) nanoparticles.

■

RESULTS AND DISCUSSION A commercial liquid flow TEM specimen holder was connected to a pressure controlling system (Figure 1). The liquid flow tubing to and from the liquid cell in the tip of the holder was dried of liquid and used as vacuum pump lines. The pressure was controlled manually by a needle valve with a precision of 20 mbar. A plug valve installed before the vacuum pump was used to shut off or to fully open the system. Others used a gas flow system presented elsewhere but then the space between the SiN membrane windows is not entirely filled with liquid, and so it is a challenge to adjust the exact liquid thickness.30 In our case, there is no gas flow, and the vacuum level in the tubing is directly used to control the liquid thickness in the specimen chamber. We investigated the effect of low pressure on the thickness of the liquid between two SiN windows. The microchips supporting the SiN membrane 4609

DOI: 10.1021/acs.nanolett.9b01576 Nano Lett. 2019, 19, 4608−4613

Letter

Nano Letters

Figure 3. Improvement of the resolution of LPEM of AuNPs due to usage of the pressure controller. (a) STEM image of AuNPs, overlapped with the EELS thickness map (20 × 20 EEL spectra represented by an interpolated two-dimensional contour map) of the imaged region at 0.05 bar. The electron dose was D = 50 ± 5 e−/Å2. Color coding represents liquid thickness in micrometers. (b, c) Line profiles obtained from an AuNP marked in the black box in a at 0.05 in (b), and 1.0 bar in (c), with a Gaussian fit and calculated full width at half-maximum, and the 25−75% edge width. Zoom-in images of the AuNP cluster are shown as insets. (d)−(f) Transmission electron microscopy (TEM) images of AuNPs at higher magnification to acquire atomic resolution showing the lattice fringes. The corresponding fast Fourier transform (FFT) of each image is shown as an inset. AuNPs were immobilized on the bottom window of the liquid cell. The liquid cell was imaged in dry state in (d), filled with 10% phosphate-buffered saline (PBS) in water with a measured liquid thickness (t in (e) and (f)) of 0.39 ± 0.08 μm at 0.05 bar in (e), and at 1.0 bar in (f) resulting in a liquid thickness of 0.74 ± 0.15 nm. D = (50 ± 5) × 102 e−/Å2 with an exposure time of 1 s for (d)−(f). The electron beam was blanked between each acquisition. The FFT in (e) shows a resolution of 2.4 Å obtained from AuNPs in 0.39 μm thick liquid.

overlapped with the experimental data but differed by an average of 22% for the investigated pressure range (Figure S 3d), suggesting that the pressure difference between the interior of the liquid cell and the microscope vacuum and material properties of SiN are the dominant factors determining the degree bulging. In order to evaluate the improvement in image quality with our pressure controller, we compared the obtained spatial resolution for AuNPs at 0.05 and 1.0 bar. A 0.5 μL droplet of citrate functionalized AuNP (20 nm in diameter) solution was drop-casted onto one of the microchips, and the microchips were immediately assembled in the in situ TEM holder. Figure 3a shows a STEM image of AuNPs at 0.05 bar, overlapped with the thickness map (20 × 20 spectra) of the same region obtained by EELS. This region was close to the window center so the minimum and maximum of the measured water thickness at 0.05 bar were 0.21 ± 0.04 and 0.41 ± 0.08 μm, respectively. At 1 bar, the water thickness ranged from 0.60 ± 0.12 μm up to 0.76 ± 0.15 μm in the same region. Parts b and c of Figure 3 show the image intensity line profiles over a particle from the box in (a), obtained at 0.05 and 1 bar, respectively. At 1 bar (without pressure adjustment), the AuNP imaged in 0.76 μm thick liquid appears wider than it should be, 32 nm instead of the known diameter of 20 nm, while the image shows “blurry” edges (Figure 3c, inset). At reduced pressure, on the contrary, the measured width of 21 nm is consistent with the AuNP size, and the edges appear sharper (Figure 3b, inset).

compared to a liquid cell with 1 bar internal pressure. The measured thickness near the edge of the viewing area showed variation at the increased pressures probably due to a slight misalignment of two windows. If both SiN windows were to be perfectly aligned in the liquid cell assembly, then no variation would be expected near the window edge. However, this is rarely the case considering the manual alignment of microchip pairs. We did not observe any bubble formation at this pressure so that it is reasonable to assume that the liquid cell was completely filled with liquid at the viewing area. We observed that the liquid thickness remained constant for up to 35 min (Figure S2) by collecting EEL spectra in the same region with 5 min intervals at 0.05 bar for AuNP solution with an initial volume of 0.5 μL. Thereafter, the liquid thickness started to decrease presumably by evaporation of the liquid from the liquid cell. We did not observe a sudden retraction of the liquid sample from the viewing area of a liquid cell when the pressure in the tubing was reduced. Such an event may possibly happen due to the hydrophobic nature of SiN. For this reason, the microchips were exposed to Ar−O2 plasma prior to an experiment to increase their hydrophilicity, and thus increase their wettability for water (Supporting Information). Bulging of SiN membranes under differential pressure depends on the mechanical properties, thickness, and the short side length of the membrane but liquid vapor pressure or fluid-membrane interfacial energy might affect the bulging as well. The deflection of the window was calculated using a theoretical model, not including liquid effects and compared to the experimental results (Figure S3). At 0.05 bar, the model 4610

DOI: 10.1021/acs.nanolett.9b01576 Nano Lett. 2019, 19, 4608−4613

Letter

Nano Letters

scattering.4 We, therefore, investigated TEM phase contrast on silica NPs at this transition liquid thickness. To avoid radiation damage as much as possible, we acquired a series of low-dose (D = 10 ± 1 e−/Å2) TEM images from the same sample region containing silica NPs immobilized at the bottom window at a defocus of −10 μm. This dose fractionation was furthermore used to observe the effect of electron dose cumulatively on the obtained phase contrast (Figure S4). Ten of these images were then aligned and averaged. The resulting averaged TEM image (Figure 4a) shows silica NPs with surrounding Fresnel

We also observed the AuNP crystal lattice fringes at higher magnification using TEM for a new sample in which AuNPs were immobilized on the SiN membrane in order to minimize drifting of AuNPs during image acquisition. Positively charged, polyethylenimine (BPEI) coated AuNPs of 15 nm in diameter (Nanocomposix, CA, USA) from a liquid droplet were immobilized on the plasma-treated, negatively charged bottom window of the liquid cell, and the sample was dried. Lattice fringes were visible for AuNPs in a dry liquid cell (Figure 3d). The fast Fourier transform (FFT) of the image revealed a spatial resolution of 1.3 Å (Figure 3d, inset), whereby higher spatial frequencies appeared further away from the center of the FFT image. The liquid cell was then filled with 10% phosphate buffered saline (PBS) using a syringe pump (Harvard Apparatus, MA, USA). Then, the system was disconnected from the syringe pump and connected to the pressure controller. The pressure controller system was found capable of adjusting the pressure even though the tubing was filled with liquid in this case. A liquid thickness of 0.39 ± 0.08 μm was measured at 0.05 bar pressure in a region with two stationary AuNPs. A TEM image of the AuNPs under these conditions displayed lattice fringes and a spatial resolution of 2.4 Å (Figure 3e). This region with 0.39 μm liquid thickness at 0.05 bar was close to the window center. A region closer to the window edge might have resulted in a thinner water layer leading to a higher resolution, but it was not trivial to find a stable NP under the irradiation at this high magnification (×500k). Some of the particles were drifting during the acquisition with a high electron dose of D = (50 ± 5) × 102 e−/Å2, which we used to obtain atomic resolution in these sample conditions. The error margin of 10% in the calculation of electron dose was explained elsewhere.32 At a pressure of 1.0 bar, the liquid thickness increased to 0.74 ± 0.15 μm in the same region and the TEM image appeared “blurred” without any lattice information. This would be the normal situation of a liquid cell with 0.2 μm spacer without the pressure-control system. Note that it was impossible to make the same observation in STEM mode since the AuNPs were drifting too much in the direction of scanning beam during acquisition and this blurred their projection in the image (data not shown). A further question of interest was whether it was possible to observe the motion of particles despite the lower internal pressure in the liquid cell. A concentrated solution of AuNPs was placed in the liquid cell, and the pressure was adjusted to 0.05 bar resulting in a liquid thickness of 0.29 ± 0.06 μm averaged throughout the viewing area. As seen in the STEM movie S1, many clusters of AuNPs are present in the entire viewing area that exhibit dynamics in their position. The observation of streaks instead of round objects is caused by fast motion in direction of the scanning beam. Having demonstrated that the pressure controller can be used to achieve high resolution for AuNPs, it was also tested if TEM of a material of low atomic number (Z) would benefit from the reduced and uniform liquid thickness over the field of view. The key obstacle when imaging low-Z materials in membrane-based liquid cells is that the liquid thickness prevents using phase contrast so that mass−thickness contrast must be used. But since mass−thickness contrast is much less dose efficient than phase contrast, it is impossible to image dose sensitive materials, which many of the relevant low-Z samples are, with sufficient resolution.3 The exact transition between these two contrast mechanisms is around 200 nm for water determined by the mean free path length for electron

Figure 4. Phase contrast TEM obtained from amorphous SiO2 nanoparticles using the pressure controller. (a), (b) TEM images of SiO2 NPs in 0.33 ± 0.07 μm thick liquid in (a) and 0.20 ± 0.04 μm thick liquid in (b) at an internal pressure of 1.0 and 0.05 bar, respectively. The defocus amounted to −10 μm. Each image represents the average of 10 images acquired at an electron dose of 10 ± 1 e−/Å2 so the total dose for each image was (10 ± 1) × 10 e−/ Å2. (c) FFT of the TEM images in (a) (left) and (b) (right), demonstrating the Thon rings. (d) Radially averaged pixel intensities of the FFT images from the image center for both liquid thicknesses and the dry liquid cell. The inset shows the baseline-corrected curve for 0.20 μm thick liquid with the averaged noise level (red). In 0.20 μm thick liquid, the intensities of the Thon rings were above the noise (signal-to-noise ratio = 3.2) until the third ring at 0.33 nm−1.

fringes.33 The effect of the liquid thickness on the phase contrast is apparent by comparing Figure 4a,b acquired at liquid thicknesses of 0.33 ± 0.07 μm (at 1.0 bar) and 0.20 ± 0.04 μm (at 0.05 bar), respectively. At first sight, the images do not seem to differ much in resolution but rather in contrast. The FFT of the TEM image exhibits concentric so-called Thon rings (Figure 4c) consistent with the presence of phase contrast.34 For the thicker liquid, the Thon rings disappear at a higher spatial frequency (further out of the center), and a higher background signal is visible; both the achieved spatial resolution and contrast are thus lower. In order to quantify the frequency components in the images, we radially averaged the pixel intensities of the FFT images of both liquid thicknesses with respect to the image center (Figure 4d). For comparison, the FFT information also 4611

DOI: 10.1021/acs.nanolett.9b01576 Nano Lett. 2019, 19, 4608−4613

Letter

Nano Letters

nonpressure-regulated liquid cell. The membrane bulging at atmospheric pressure increased the liquid thickness in the tested uncontrolled system from 0.2 μm set by the spacer to >0.8 μm in the middle of the liquid cell, whereas the pressure controller limited the liquid thickness to