Hawkes P.W., Spence J.C.H. (Eds.) Science of Microscopy. V.1 and 2

Подождите немного. Документ загружается.

Chapter 4 Analytical Electron Microscopy 405

Van Cappellen, E. and Doukhan, J.C. (1994). Ultramicroscopy 53, 343.

Varela, M., Findlay, S.D., Lupini, A.R., Christen, H.M., Borisevich, A.Y., Dellby,

N., Krivanek, O.L., Nellist, P.D., Oxley, M.P., Allen, L.J. and Pennycook, S.J.

(2004). Phys. Rev. Lett. 92, 095502.

Verbeeck, J. and Van Aert, S. (2004). Ultramicroscopy 101, 207.

Verbeeck, J., Van Dyck, D. and Van Tandeloo, G. (2004). Spectrochim. Acta B 59,

1529.

von Barth, U. and Grossman, G. (1982). Phys. Rev. B 25(8), 5, 50

Voyles, P. and Muller, D.A. (2004). Microsc. Microanal. 10, 291.

Vvedensky, D.D. (1992). In Theory of X-Ray Absorption Fine Structure, Unoccupied

Electronic States (J.C. Fuggle and J.E. Inglesﬁ eld, Eds.), p. 138 (Springer-

Verlag, Berlin).

Watanabe, M. and Williams, D.B. (1999). Microsc. Microanal. 5, 88.

Watanabe, M. and Williams, D.B. (2005a). Microscopy and Microanalysis 2005,

Vol. 11, Suppl. 2, edited by R. Price, P. Kotula, M. Marko, J.H. Scott, G.F. Van

der Voort, M. Mah Lee Ng, K. Smith, B. Grifﬁ n, P. Smith and S. McKernan,

p. 1362. Honolulu, HI: Cambridge University Press.

Watanabe, M. and Williams, D.B. (2005b). Scanning 27, 94.

Watanabe, M., Horita, Z. and Nemoto, M. (1996). Ultramicroscopy 65, 187–198.

Watanabe, M., Ackland, D.W., Burrows, A., Kiely, C.J., Williams, D.B., Kanno,

M. and Hynes, R. (2005). In Microscopy and Microanalysis 2005 (R. Price, P.

Kotula, M. Marko, J.H. Scott, G.F. Van der Voort, M. Mah Lee Ng, K. Smith,

B. Grifﬁ n, P. Smith and S. McKernan, Eds.), Vol. 11, Suppl. 2, p. 2132

(Cambridge University Press, Honolulu, HI).

Williams, D.B. and Carter, C.B. (1996). Transmission Electron Microscopy, A Text-

Williams, D.B., Watanabe, M., Papworth, A.J. and Li, J.C. (2003). Thin Solid Films

424, 50.

Woldseth, R. (1973). X-Ray Energy Spectrometry (Kevex Instruments).

Wong, K. and Egerton, R.F. (1995). J. Microsc. 178, 198.

Wooten, F. (1972). Optical properties of solids. (Academic Press, New York).

Zaluzec, N.J. (1982). Ultramicroscopy 9, 319.

Zaluzec, N.J. (1984). J. Phys. 45 (Suppl. 2), C2-429.

Zaluzec, N. (1985). Ultramicroscopy 18, 185.

Ziebold, T.O. (1967). Anal. Chem. 39, 858.

imaging of nonspherical silicon nanoparticles embedded in silicon oxide by

Tencé, M. (2002). Private communication.

Terauchi, M., Tanaka, M., Tsuno, K. and Ishida, M. (1999). J. Microsc. 194, 203.

Thomas, P.J. and Midgley, P.A. (2001a). Ultramicroscopy 88, 187.

Thomas, P.J. and Midgley, P.A. (2001b). Ultramicroscopy 88, 179.

Tiemeijer, P.C., van Lin, J.H.A. and de Jong, A.F. (2001). Microsc. Microanal. 7,

1130.

Titchmarsh, J.M. and Dumbill, S. (1996). J. Microsc. 184, 195.

Trebbia, P. (1988). Ultramicroscopy 24, 399.

Turowski, M.A. and Kelly, T.F. (1992). Ultramicroscopy 41, 41.

Uhlemann, S. and Haider, M. (2002). In Microscopy and Microanalysis 2002 (E.

Voelkl, D. Piston, R. Gauvin, A.J. Lockley, G.W. Bailey and S. McKernan,

Eds.), p. 584CD (Cambridge University Press, Cambridge).

Van Cappellen, E. (1990). Microsc. Microanal. Microstruct. 1, 1.

A.J. Lockley, G.W. Bailey and S. McKernan, Eds.), p. 68 (Cambridge Univer-

sity Press, Cambridge).

Yang, Y.Y. and Egerton, R.F. (1995). Micron 26, 1.

plasmon tomography. Applied Physics Letters 89, 151920.

Westwood, A.D., Michael, J.R. and Notis, M.R. (1992). J. Microsc. 167, 287.

*Yurtserver, A., Weyland, M. and Muller, D.A. (2006). Three-dimensional

book for Materials Science (Plenum, New York).

*References added since the first printing.

406

High-Speed Electron Microscopy

Wayne E. King, Michael R. Armstrong, Oleg Bostanjoglo, and

Bryan W. Reed

1 What Is High-Speed Electron Microscopy?

Historically, transmission electron microscope (TEM) images have

been recorded on ﬁ lm using exposure times of a few seconds. The use

of the TEM, and especially the high-voltage electron microscope, for in

situ experiments stimulated interfacing of video cameras with the

microscope column. This enabled capture of frames with millisecond

exposure times and interframe times of milliseconds and proved to be

very useful for the study of dynamic phenomena with characteristic

times that were consistent with the time resolution of the detectors,

such as dislocation motion in deforming metals.

Many phenomena in chemistry, biology, and materials science

proceed at rates that are signiﬁ cantly faster than could be captured

with standard video methods. Figure 5–1 attempts to represent some

of those phenomena in terms of spatial resolution and time resolution.

Experimental techniques exist for probing most of this space, but the

quadrant at high space and time resolutions has proved to be the most

difﬁ cult to access.

The chemistry community, which studies phenomena with charac-

teristic time scales of the order of the atomic vibrational period, has

employed short pulse lasers and optical techniques to study ultrafast

phenomena. Likewise, X-rays are becoming an important tool for ultra-

fast studies. Recently, ultrafast electron diffraction (UED) has been

developed with time resolutions approaching that of optical and X-ray

methods.

1–9

Although optical and X-ray methods are powerful, the widespread

use of TEMs and the accompanying extensive literature demonstrate

that there are many scientiﬁ c investigations that beneﬁ t from high

spatial resolution imaging. In materials science, these include observa-

tion of defects in crystalline materials (e.g., dislocations, grain bound-

aries, and heterophase boundaries) in which imaging affords a more

direct interpretation of the defect structure than does diffraction.

One group has demonstrated the potential of high-speed TEM or

dynamic (DTEM).

10–16

This group has produced an imaging instrument

Chapter 5 High-Speed Electron Microscopy 407

with time resolution ∼3–10 ns

10,11

and has demonstrated pump/probe-

train “movie” or nanosecond multiframe operation.

17,18

In this chapter

we review the events leading to the development of the DTEM. We

then describe the technologies employed in DTEM. Finally, we discuss

the performance, limitations, and applications and conclude with a

look to the future of these instruments.

2 Technologies of DTEM

2.1 Anatomy of a DTEM

Figure 5–2 shows a photo of the DTEM that was developed at TU

Berlin. The instrument in Berlin is based on a Siemens Elmiskop 1A

constructed in 1961. Several modiﬁ cations, shown in Figure 5–3, were

required to convert the instrument from a conventional continuous-

wave (CW) mode to an instrument that could carry out pump-probe

Figure 5–1. Phenomena classiﬁ ed by spatial and temporal resolution. Spatial resolution, is deﬁ ned as

follows: (1) if the technique is an imaging method, the low end of the bounding box would be deﬁ ned

by the smallest resolvable feature and the high end by the typical ﬁ eld of view or (2) for a nonimaging

technique, the bounding box would be deﬁ ned by the range of probe or spot sizes. Time resolution is

deﬁ ned as that for a single-shot investigation of irreversible processes. So time resolution is deﬁ ned

as the single-shot exposure time to obtain data that demonstrate a particular spatial resolution. (See

color plate.)

Spatial Resolution (m

-1

)

Time Resolution (s

-1

)

Diffraction of

Phase

Transformations

structural

changes in

biology

nucleation

and

growth

of damage

Imaging of

Phase

Transformations

Making

and

Breaking

of Bonds

Dislocation

Dynamics at

Conventional

Strain Rates

magnetic

switching

melting

and

resolidification

408 W.E. King et al.

experiments. These included the development of a photoelectron gun

to replace the conventional thermionic tungsten ﬁ lament. This was

realized by interfacing a laser to the electron-optical column via a

vacuum feed-through and directing that laser to the ﬁ lament using a

turning mirror. The photocathode is driven by a Q-switched frequency-

quadrupled Nd:YAG laser (wavelength 266 nm). Using the photoelec-

tric effect, a short laser pulse stimulates the emission of a low-energy

electron bunch that is accelerated by the high-voltage stage of the elec-

tron microscope. In CW mode, the self-biasing Wehnelt brings the

electrons emitted at the ﬁ lament to a focus, thus deﬁ ning the object

plane for the condenser lens system. In the pulsed mode, circuit time

constants inhibit the self-biasing effect of the Wehnelt, no crossover is

formed, and the optics are a bit different than CW operation. The pho-

tocathode in the Berlin instrument has been designed so that it can

operate in either CW or pulsed modes. This facilitates easy electron-

optical alignment of the instrument. Since the electrons are emitted

from slightly different regions of the cathode in thermionic operation

compared with pulsed operation, it is essential to be able to align the

gun in pulsed mode. This is achieved by pulsing the photocathode

periodically at a rate of several pulses per second, observing the image

on the digital image capture system (discussed later), and carrying out

a gun alignment.

Accelerators for TEMs are designed to give smooth electric ﬁ eld

gradients and stable potentials. The TU Berlin instrument typically

operates at 80 kV with an accelerating gap of 1 cm. This corresponds to

Figure 5–2. The DTEM at TU Berlin. Cathode and sample drive lasers are at left. (See color plate.)

Chapter 5 High-Speed Electron Microscopy 409

Figure 5–3. TU Berlin dynamic transmission electron microscope. (See color plate.)

a ﬁ eld gradient of 8 × 10

V/cm. When electrons are photoemitted at

low energy from the cathode, they are most vulnerable to space charge

effects. It is therefore advantageous to optimize the extraction potential

at the cathode and the ﬁ eld gradient to minimize space charge effects.

This may be achieved by using an extraction electrode similar to those

used in ﬁ eld emission electron guns. The steady-state current limit due

to space charge in a parallel-plate accelerating gap is dictated by the

Child–Langmuir law,

and varies as V

3/2

, with V the voltage and d

the distance between the cathode and the extraction electrode. Thus it

is most effective to have both a large electric ﬁ eld (possibly enhanced

by surface curvature of the cathode) and a short gap.

410 W.E. King et al.

Since the photoemitted electrons will be incoherent, it is not essential

to have a very small source size. Therefore, it may be possible to spread

the laser beam across a relatively large area to ensure that sufﬁ cient

electrons can be generated without damaging the cathode with the

laser. Of course, the size of the souce will control the size of the smallest

electron beam spot on the sample. Therefore, there will be a trade-off

between the region of interest and source size. The source size should

be adjusted to achieve the desired viewing area and no larger (to mini-

mize space charge effects).

Apertures will play an important role in limiting space charge effects

in DTEM. The condenser aperture should be removed to maximize the

electron ﬂ uence on the sample. The objective aperture can be used to

produce contrast by blocking scattered electrons as is routinely done

in diffraction contrast imaging. This will decrease the number of elec-

trons in the beam as the beam passes through the projector lens system.

Alternatively, it is possible to block the main beam and allow the scat-

tered electrons to pass.

In a double condensor lens system common to most electron micro-

scopes, the ﬁ rst condensor will likely be opperated at low or zero excita-

tion to allow as many electrons as possible to pass to the second

condensor lens. This lens can then be used in conjunction with the

source size to optimize the illumination area on the sample.

Because of the energy spread in pulsed beams, chromatic aberration

will limit the obtainable resolution. Thus constraints on the design of

the objective lens can be relaxed until a need for high resolution arises.

At that point, an aberration correction may be possible.

The projector lens has the potential to introduce distortion in the

image if space charge and trajectory displacement effects are signiﬁ -

cant at the numerous crossovers in the projector lens system. For suf-

ﬁ ciently short pulses, a lateral space charge effect could become

signiﬁ cant at the crossovers.

The time resolution of the TU Berlin instrument is currently 3–10 ns.

At this resolution, the electron pulse length is controlled primarily by

the laser pulse length and by the number of electrons in an electron

bunch required to obtain an image. By making a few simpliﬁ cations

and assumptions, the number of electrons needed to form an image

can be estimated. The current state of the art in low-dose TEM applica-

tions in the biological sciences requires an average of about 100 elec-

trons per pixel.

For a 1 k % 1 k pixel image, this requirement leads to

–10

total electrons at the image plane (see Figure 5–4). This number

of electrons per pulse is currently achieved by the nanosecond electron

microscope at TU Berlin.

However, many electrons are lost in the TEM

condenser lenses. Thus, the number of electrons that is required at the

cathode may be substantially higher. Other factors that will inﬂ uence

the number of electrons arriving at the image plane include specimen

thickness, diffracting conditions chosen for image formation, and the

size and placement of the objective aperture.

The number of electrons needed to form a diffraction pattern can be

substantially less than that required for an image. As few as 10

elec-

trons may be needed for the qualitative assessment of a diffraction

Chapter 5 High-Speed Electron Microscopy 411

pattern.

However, for quantitative assessment, more (as many as 10

)

could be required for improved signal-to-noise ratio (SNR).

It is desirable to have the ability to induce physical changes in the

sample that can be easily synchronized with the probe electron pulse.

This has been accomplished at TU Berlin by interfacing a pump laser

with the DTEM instrument. The pump beam is fed into the column at

a position above the objective lens and directed to the sample by a

turning mirror. Using this conﬁ guration, it is possible to bring an ∼20-

µm-diameter pump beam into coincidence with the probe electron

beam. In the TU Berlin instrument, two different lasers are used for

pump and probe. This is reasonable for systems with time resolutions

on a nanosecond scale. At higher time resolutions, unacceptable jitter

between the lasers will require the use of a single seed-laser with beam

split into pump and probe beams.

Because of the limited number of electrons per pulse, it is essential

to have a highly efﬁ cient detector, the so-called “every-electron camera.”

Space-charge limits on the number of electrons in a short pulse put a

premium on detection efﬁ ciency. However, because of the large spec-

Figure 5–4. Simulation demonstrating the number of electrons required for a given image quality.

The original 512% 512 8-bit grayscale image was normalized and taken as the probability distribution

for the detected position of each independent electron. Binning to a smaller number of pixels would

reduce the required number of electrons, but at the expense of resolution. Various methods for esti-

mating this (including the one shown here) consistently indicate that 10

–10

electrons are needed at

the detector, depending on the number of pixels and the required contrast. Crystalline diffraction

patterns require ∼10–100 times fewer electrons because of the concentration of intensity into relatively

few pixels.

412 W.E. King et al.

trum of desired electron energies (keV–MeV) and the disparate require-

ments of imaging and diffraction, it is clear that no single camera is

ideal for every application. At high energies (>100 keV), phosphor scin-

tillators are preferred (Figure 5–5). Phosphor thickness can be opti-

mized to achieve an energy conversion efﬁ ciency of ∼0.12 (with

intensiﬁ cation not required).

Light-scattering effects within the phos-

phor and the various optical interfaces limit system resolution.

The

point spread function (PSF) for a detector system composed of phos-

phor on a ﬁ beroptic faceplate is determined by electron and photon

scattering as well as the numerical aperture (NA) of the ﬁ beroptic plate.

The next step is to couple the phosphor to the charge-couple device

(CCD). This is typically achieved using a ﬁ beroptic bundle or lens

coupling. Fiberoptic coupling is preferred for energies below 400 keV

because of the high-photon-collection solid angle from the phosphor.

Because the PSF of the phosphor is greater than the pixel size of typical

CCDs, taper-ﬁ ber bundles are often employed to improve the PSF/pixel

size match. However, tapered bundles can exhibit spatial distortions of

∼2–3% that need to be corrected before quantitative measurements can

be made from images or diffraction patterns.

With the availability of

large CCD chips, 1 : 1 coupling becomes practical. Also, binning of adja-

cent pixels is used to boost the detective quantum efﬁ ciency (DQE).

Finally, to achieve images that follow the evolution of dynamic pro-

cesses, a multiframe movie mode is desirable. Interframe times less

than a nanosecond are desirable to observe dynamic processes. Cur-

rently, no technology exists that will enable such rapid ﬁ lling and

readout of CCD cameras. At TU Berlin, movie mode is achieved through

the use of an electrostatic beam shifter (Figure 5–3). An image-shifting

capacitor with low inductance leads is used to produce successive

frames by displacing an image patch across the detector between the

Electron Microscope

Base plate

Insertable

gate valve

Phosphor

Fibre optics

Preamplifiers

CCD

Peltier cooling element

Water jacket

Vacuum

system

Cooling

water

CCD drive

pulses

Figure 5–5. Directly coupled CCD camera system.

Chapter 5 High-Speed Electron Microscopy 413

electron pulses. For multiframe imaging of the specimen, the image is

conﬁ ned by an adjustable slit aperture in the second-intermediate

image plane, exploiting the space between the intermediate and projec-

tor lenses. For multiframe diffraction, the diffraction pattern in the

back focal plane of the objective lens is conﬁ ned with a suitable aper-

ture, and this pattern is as usual transferred with the intermediate and

projector lenses to the detector.

The beam shifter is a low-capacitance rectangular plate capacitor

mounted on a side-entry retractable holder. It was designed to ﬁ t into

the space between the intermediate and projector lenses. Two blades

glued onto the plate electrodes at the beam entrance serve as the image

ﬁ eld-conﬁ ning aperture. The plates of the capacitor are tilted to prevent

the deﬂ ected electrons from hitting the attracting electrode. The capac-

itor can be axially moved and the distance between the plates varied,

allowing alignment and ﬁ eld selection. The deﬂ ecting unit is inserted

at such a location that the second intermediate image can be positioned

near the ﬁ eld-conﬁ ning slit aperture.

Three-frame imaging is achieved by applying a two-step voltage

waveform across the capacitor (Figure 5–6). Because the deﬂ ecting ﬁ eld

is short due to limited space and since the deﬂ ection occurs near an

image plane, high voltages of several hundred volts are required to

shift 100-kV electron images. The shifting steps are produced by short-

ing voltages with fast transistor-based switches (Figure 5–7). When

both switches are open the image is on the left-hand side (Figure 5–6).

As switch S1 closes, the image is moved to the center. Finally, switch

S2 closes transferring a positive voltage step via the 6.8-nF capacitor to

the right electrode, thereby displacing the image to the right side.

A high-voltage radio frequency n-channel metal oxide semiconduc-

tor ﬁ eld-effect transistor and cascaded bipolar transistors (Motorola

2N5551) driven in the avalanche mode were used as switches. The

voltage fall times were 2–3 ns. Overshoot and ringing were minimized

by careful grounding, shielding, and placing of leads avoiding stray

capacitance, parasitic inductance, and closed loops. In addition, ferrite-

damping beads were suitably distributed along the leads. The switches

were mounted directly on the holder of the shifting capacitor to avoid

voltage distortions caused by reﬂ ections and to minimize electromag-

netic interference.

Electromagnetic interference easily was kept below a deleterious

level using the MOSFET-based switch. But it is rather inconvenient to

Figure 5–6. Generation

of voltage steps across

the image shifting

capacitor for three-

frames imaging.

414 W.E. King et al.

supply the required large gate currents if a high negative voltage is to

be switched, since gate and source are then at a high potential. In con-

trast, the voltage-triggered avalanche transistor cascades can readily be

used as switches for voltages of either polarity. Unfortunately, electro-

magnetic interference is pronounced, and it proved problematic to

prevent two cascades from ﬁ ring simultaneously. For these reasons

switch 1, handling a positive voltage, is based on the fast MOSFET, and

the avalanche transistor cascade is used to switch the negative voltage.

2.2 Photoelectron Guns

DTEM requires an extremely bright pulsed electron source. A 512 %

512 pixel image on the border of usefulness requires ∼10

electrons at

the detector, while a good image requires ∼10

(Figure 5–4). To obtain

a usable image in a single shot, this number of electrons must pass

through the sample region of interest in ∼1–10 ns, which implies a

Figure 5–7. Fast high-voltage switches. (a) High-power radiofrequency n-

channel MOSFET for shutting down a positive voltage below 500 V. DE275 is

supplied by Directed Energy Inc. (b) Cascaded avalanche npn transistors for

switching off a negative voltage of twice the collector-base breakdown voltage

(about −700 V). The transformer is a high-frequency ferrite ring (40–400 MHz)

with ﬁ ve windings of strand wire for each coil. The Schottky diodes and the

100-Ω resistor protect the trigger circuit against voltage steps passed by the

gain-multiplied collector-base capacitance during switching. The 200-pF

capacitor holds off the DC high voltage in case of isolation failure.