Micro-engineered Nanowire Electron Source for Atomic Resolution Imaging

The size tunability and chemical versatility of nanostructures provide attractive engineering potential to realize an electron source of high brightness and spatial temporal coherence, which is a characteristic ever pursued by high resolution electron microscopy. (1–3) Regardless of the intensive research efforts, electron sources that have ever produced atomic resolution images are still limited to the conventional eld emitters based on a bulk W needle. It is due to the lack of fabrication precision for nanostructured sources, that is required to align a nanometric emission volume along a macroscopic emitter axis with sub-degree angular deviation. (4) In this work, we produced a LaB 6 nanowire electron source which was micro-engineered to ensure a highly collimated electron beam with perfect lateral and angular alignment. Such electron source was validated by installing in an aberration-corrected transmission electron microscope, where atomic resolution in both broad-beam and probe-forming modes were demonstrated at 60kV beam energy. The recorded un-monochromated 0.20eV electron energy loss spectroscopy (EELS) resolution, together with 20% probe forming eciency and 0.4% probe current peak-to-peak noise ratio under a wide vacuum range, presented the unique advantages of nanotechnology and promised high performance low-cost electron beam instruments.


Introduction
Electrons emitted through electric eld-induced tunneling are partially coherent in space and time. An electron source with high spatial temporal coherence is desired by applications of scanning electron microscope (SEM), transmission electron microscope (TEM) and scanning transmission electron microscope (STEM). To achieve high spatial coherence, electron emission volume is expected to be small; and to achieve high temporal coherence, the emission surface is expected with low work function. (5,6) Conventional electrochemical etching technique is effective to produce nanoscale sharp needles out of refractory metals. Though it ful lls the small emission volume requirement, few work function options are available among this category of materials. W single crystal oriented along [310] lattice direction was a result of compromise and has been used as the high brightness electron source for electron microscopes that pursue the highest resolving power. (7,8) This situation has not changed ever since the rst atomresolved STEM image was produced by Albert Crewe in 1970 using a W eld emitter. (9) The concept of nanostructured electron source separates the material of support body and the material to emit electrons.
While emitter support body could simply follow the conventional structure of a metal needle, the electron emitting unit could adopt nanostructures of various geometry and chemical composition. It became possible to tune independently both the emission volume and emission surface work function. Electron source-relevant nanostructures, to name a few, include: 0D atom clusters, 1D nanotube/nanowires (NW/NT), 2D nanosheets, high work function noble metal pyramids and negative electron a nity lms. (2,(10)(11)(12)(13)(14)(15) However, only few such nanostructured electron sources succeeded in producing SEM images of nanometer resolution and none has proven capable of atomic resolution imaging in TEM. (1,(16)(17)(18) Angular alignment is one of the most important procedure to achieve high resolution imaging in electron microscopy. In a modern TEM column, electrons from emitter surface are expected to travel over 2 meters in a straight line with nearly 0 o angular deviation. Even though sophisticated aligners were used in the column to steer electron beam, there is no angular adjustment mechanism available for the electron source in electron gun chamber, due to the di culty of implementation at high voltage above ground. Therefore, in the electron source fabrication stage, emission direction must be pre-aligned along the emitter support body axis with sub-degree precision. Such nanomanipulation capability was not available to date.
In this work, a micro structure was fabricated on a support needle tip by focused ion beam (FIB); and a single LaB 6 nanowire was installed and aligned through SEM nanomanipulation. (19,20) An angularly aligned emitter was validated in an aberration-corrected (S)TEM platform with a customized 60kV electron gun. Besides atomic imaging resolution, zero-loss peak (ZLP) energy resolution was also compared with that produced by a conventional W (310) electron gun using the same lens column and EELS spectrometer.

Main Text
Compared to other form of nanostructures, 1D NT/NW have been used the most extensively in electron source assemblies. (21) Its axial symmetry is compatible with an electron microscope column and its extended side surfaces provide sites for attaching to emitter support needle. (22) However, because of this side attachment con guration, axial symmetry of electric eld distribution around the NT/NW tip could vanish. Such effect becomes more pronounced when short suspension length of NT/NW is used to reduce mechanical vibration. As a result, even though the NT/NW body is perfectly centered laterally and aligned angularly, the emitted electron beam will still be out of axis. To solve this problem, we used FIB to reshape the support needle tip into a collimator structure, as shown in the SEM image of gure 1a. The yellow pseudo color represents a collimator structure which is used to form axially symmetric electric eld around the emission tip of a nanowire; Blue pseudo color represents a carbon pad that is used to stop chemical reaction between nanowire and metal support body during lament heating; Purple pseudo color represents the LaB 6 nanowire position. Height of the carbon pad is adjusted by FIB milling to ensure that the nanowire is positioned at the collimator center. Figure 1a inset is an optical microscope image of a TEM emitter showing an arch-shaped lament and the position of the fabricated microstructure at a Ta needle tip. The SEM snapshots of crucial steps of installing a LaB 6 nanowire into this microstructure are presented in gure 1b-d. In the rst step, one LaB 6 nanowire was attached on a nanomanipulator hand through Van Der Waals force. Such nanowire was then transported through the collimator hole by driving the manipulator hand. (Figure 1b) After the nanowire got into full contact with the carbon pad, electron beam induced carbon deposition was applied to the bonding region to x the nanowire. (Figure 1c) At the last step, manipulator hand was retracted and detached from the nanowire. It was realized because the Van Der Waals force is much weaker compared to carbon bonding force. (Figure 1d). Figure 1e and f showed a procedure where two manipulator hands are placed onto support needle at positions marked by two solid arrow heads. When the two manipulator hands moved in the direction of arrow heads, microscale bending of the support needle tip was realized. The bending amount was pre-determined by measuring off-axis angle using eld emission pattern, as shown in gure 1e inset. (1) In this case, even though the LaB 6 nanowire body (marked by a hollow arrow head) was already aligned with support needle axis, emission pattern still showed a 2.5-degree angle out of axis. Figure 1f is an SEM image of the emitter tip after 3-degree micro-bending performed by the two manipulator hands. Inset shows a perfectly aligned eld emission spot on the phosphor screen center hole after such a set of manipulations.
In order to validate the above described LaB 6 NW emitter in a commercial TEM, a 60kV electron gun and high voltage tank were customized to adapt the changes in extraction voltage and total emission current. The extraction voltage and emission current for the LaB 6 nanowire emitter are ~400V and ~40nA, which are much lower than ~4000V and ~15uA, as used by conventional W (310) emitter. A TEM equipped with such a LaB 6 NW electron gun is shown in gure 2a. Components marked along beam path from top to bottom are the NW emitter, condenser lens aperture (CLA), specimen, high angle annular dark eld (HAADF) detector, Faraday cage for probe current measurement, image screen, and EELS spectrometer. Figure 2b is a defocused TEM image about the CLA, taken after emitter centering with respect to the aperture using a gun motor capable of lateral movement. (23) Transmitted electrons and those scattered by the aperture edge form Fresnel diffraction fringes in the near eld. The visibility of such fringes on the TEM screen is determined by an amplitude damping envelop formed due to lens aberration. Therefore, axially symmetric fringe pattern is observed only when the electron source is placed on the axis passing through the CLA center and thereby forming the same angles to the entire aperture edge circumference. Only a small portion of the total emitted electrons could go through the apertures and reach the image screen if emitter angular pre-alignment is poor. In this case of perfect pre-alignment, more than 10nA beam current could reach the image screen for low magni cation TEM observation with large eld of view. Figure 2c is a 30kx low-magni cation TEM image showing a LaB 6 nanowire as synthesized by chemical-vapor-deposition method. Amplitude contrast is the main TEM image forming mechanism at this magni cation. To demonstrate phase-contrast imaging, a Si [110] specimen was used to take high resolution TEM (HRTEM) images with magni cations from 400kx to 1000kx. A region from a 1000kx HRTEM image is displayed in gure 2d to show clear Si (111) planes. Figure 2e is a fast Fourier transform (FFT) diffractogram of the HRTEM image. (S gure1 in supplementary information) The highest observable spatial frequency corresponds to 90pm spacing of Si (442) planes. This spot is likely to be produced by secondary diffraction from two diffracted beams, because (442) is a forbidden re ection in primary diffraction pattern of diamond structure crystals. (24) The second smallest observable feature corresponds to 96pm of Si (440) planes. Two vectors were drawn in the diffractogram connecting the center transmitted spot to one 90pm spot and its neighboring 96pm spot respectively. The 96pm vector passes through Si (220) primary re ection and the 90pm vector passes through no primary re ections. Based on the principle of transmission cross-coe cient, non-linear phase contrast transfer function in the reciprocal space reaches maximum along primary re ection directions. (25,26) High intensity damping occurs for secondary re ections that are out of those primary re ection directions, especially for high index re ections. If 96pm spot is also a secondary re ection, it would therefore be much brighter than 90pm spot as a result of higher phase contrast transfer. However, the fact that 96pm spot is weaker in intensity compared to 90pm spot suggests that 96pm spot is a primary re ection which intensity is damped by lens aberration envelop. Therefore, an upper bound of information limit is estimated to be 96pm. Information limit of this TEM using the conventional W (310) emitter is 110pm~120pm. (27) The ner resolvable features by the LaB 6 NW electron source is a proof of higher spatial temporal coherence, which determines the damping envelop in a linear phase contrast transfer process. (28, 29) It requires detailed knowledge of imaging condition and simulation to evaluate resolution quantitatively for an HRTEM image with spherical aberration. (30) STEM is a probe-forming mode of TEM. A demagni ed image of the electron source forms an electron probe which is used to interact with specimen atom columns in raster scan manner. Image resolution can be directly measured by nding the smallest feature resolved in a STEM image. Figure 3a is a STEM image taken with HAADF detector about the Si [110] specimen, which is the same as that used for the TEM image of gure 2d. Each white dot in gure 2d is now resolved to be a dumbbell structure with 136pm atomic separation. The inset is a diffractogram of the image. The re ection with the highest observable spatial frequency is Si (440) with 96pm lattice spacing. Four Si atom columns marked by a red rectangle were magni ed in gure 3b and their line pro le is displayed below the image. The specimen thickness was determined to be 50nm by EELS method. (31) A multi-slice simulation at 60kV was then carried out by QSTEM to t the image line pro le with the assumption that the electron probe shape is a convolution between aberration-related broadening and a Gaussian shaped source image. (32, 33) The best t was found when source image FWMH was set to be 60pm and it resulted in a probe shape with FWMH of 88pm. (S gure2 in supplementary information) Single atom imaging capability was demonstrated by STEM imaging of a single layer graphene specimen, which is presented in gure 3c. Three pairs of graphite (200) re ections with 107pm spacing were clearly shown in gure 3d, which is a diffractogram of gure 3c. Four neighboring C atoms in gure 3c as selected by a red rectangle were used to make a line pro le shown in gure 3e. The same electron probe with 88pm FWHM was again used to simulate a graphene image. The well-matched line pro le thereby veri ed our previous probe size estimation.
To take advantage of the high current stability of the LaB 6 nanowire emitter in moderate UHV condition, our customized electron gun was only pumped by ion pumps to keep vacuum in the ~10 -8 Pa range, while the original W (310) electron gun vacuum was in the ~10 -10 Pa range enhanced by non-evaporable getter pumps. (34) As gure 4a shows, 21% of total emission current from a LaB 6 nanowire emitter could be used as probe current, which is measured using a Faraday cage near the image screen. Low noise level in the probe current is especially important for single atom HAADF imaging of low atomic weight elements because of their small electron scattering cross-section. In the case of graphene, signal-to-noise ratio is typically ~10%. (35) To allow 8-bit greyscale information, the noise level of STEM probe current is expected to be below 0.4%. Figure 4b is the probe current pro le measured for 1 minute, which is longer than the usual STEM frame acquisition time. A peak-to-peak noise ratio of 0.36% was demonstrated, even after a full day of use without emitter heat treatment. EELS is an indispensable spectroscopy method that has become standard to combine with STEM to realize atomic resolution chemical analysis. (36) While its spatial resolution depends on STEM probe size, its energy resolution depends on probe temporal coherence, which can be measured as the FWHM of ZLP. Electron probe energy spread originated from emitter energy spread and column related factors such as instability of high voltage tank, objective lens and EELS spectrometer, all contribute to the nal measured ZLP width. To compare the ZLP between LaB 6 NW electron gun and W (310) electron gun on equal ground, we acquired ZLPs from both electron guns using the same column, spectrometer and high voltage tanks with negligible noise compared to the expected energy spread range. Figure 4c includes ZLPs from LaB 6 NW emitter, W (310) emitter and Schottky emitter all acquired at 60kV acceleration voltage and ~15pA probe current. 0.22eV, 0.28eV and 0.8eV were measured as FWHMs for the three electron guns respectively. ZLP FWHMs with respect to probe current were also plotted for both the LaB 6 NW electron gun and W (310) electron gun in gure 4d.
At probe current ~5pA, both electron guns reached the narrowest ZLPs which are 0.2eV and 0.27eV respectively. (S gure3 in supplementary information) It is noted that the LaB 6 nanowire electron gun energy spread is consistent with our previous measurement using hemispherical analyzer in a low voltage eld emission chamber. (1) However, energy spread from W (310) emitter was much higher in our previous measurement. It is due to the >100-fold higher vacuum in the JEM-Arm200F TEM electron gun where W (310) could maintain a clean surface for a much longer time. On the other hand, the less vacuum stringency and consistent performance of LaB 6 nanowire emitter in a wide range of vacuum conditions make it suitable for applications in various electron microscopes of a wide cost range. (37, 38) As a conclusion, micro-fabrication and nano-manipulation were carried out to make perfectly angularly aligned LaB 6 NW electron source. Such an electron source was installed in an aberration-corrected (S)TEM where sub-angstrom features in both broad-beam mode and probe-forming mode were resolved at 60kV acceleration voltage. The higher spatial temporal coherence offered by the nanostructured electron source produced advantages over the state-of-the art W (310) electron source in both microscopic spatial resolution and spectroscopic energy resolution. With the additional merits of low vacuum stringency and probe-forming e ciency from the LaB 6 NW electron source, a natural next step would be applying such emitter in high voltage monochromated (S)TEM applications to bene t frontier studies such as atomic resolution vibrational spectroscopy. (39, 40) Another development direction would be a new generation low-cost electron beam imaging and chemical analysis instruments with not only high performance but also broader applicability and affordability.

Methods
The LaB 6 NW electron gun was rst installed on an experimental JEM-2200FS TEM to con rm emitter alignment, emission stability and low magni cation imaging. After that, it was transported and mounted on a JEM-Arm200F (S)TEM, which is equipped with a UHR objective pole piece and a probe-forming CEOS ASCOR spherical aberration corrector. LaB 6 nanowire electron gun was constructed with 2-anode structure. The rst anode A1 provides extraction eld to determine total emission current and the second anode A2 controls beam current that comes out of electron gun exit. Under imaging condition of this work, total emission current was kept at 40nA while electron gun demagni cation was adjusted by varying A2 voltage. Electromagnetic sensors were not equipped with this experimental electron gun. During STEM image acquisition, demagni cation ratio was adjusted to minimize image distortion caused by stray electromagnetic eld interference. For STEM imaging, CLA size, beam semi-convergence angle, inner and outer detector angles, probe current are 50um, 38.5mrad, 48mrad, 200mrad, 1.7pA respectively.
Images are formed by either using a short dwell time of 5us with 20 scans accumulation or a long dwell time of 20us with single scan. Exposure time of 0.5s was used for TEM image acquisition. GIF-quantum ER was used for EELS spectrum acquisition. CLA size, camera length, spectrometer dispersion, entrance aperture size, exposure time are 50um, 2.5cm, 0.01eV/Channel, 2.5um, 0.01s, respectively. Figure 1 Micro

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. Supplementary210421.docx