Ultramicroscopy 188 (2018) 85–89

Contents lists available at ScienceDirect

Ultramicroscopy
journal homepage: www.elsevier.com/locate/ultramic

High quality ultrafast transmission electron microscopy using resonant

microwave cavities
W. Verhoeven a, J.F.M. van Rens a, E.R. Kieft b, P.H.A. Mutsaers a, O.J. Luiten a,∗
a
Department of Applied Physics, Coherence and Quantum Technology Group, Eindhoven University of Technology, P.O. Box 513, 5600, MB Eindhoven,
Netherlands
b
Thermo Fisher Scientific, Achtseweg Noord 5, 5651, GG Eindhoven, Netherlands

a r t i c l e i n f o a b s t r a c t

Article history: Ultrashort, low-emittance electron pulses can be created at a high repetition rate by using a TM110 deflec-
Received 1 August 2017 tion cavity to sweep a continuous beam across an aperture. These pulses can be used for time-resolved
Revised 7 March 2018
electron microscopy with atomic spatial and temporal resolution at relatively large average currents. In
Accepted 9 March 2018
order to demonstrate this, a cavity has been inserted in a transmission electron microscope, and pi-
Available online 10 March 2018
cosecond pulses have been created. No significant increase of either emittance or energy spread has
Keywords: been measured for these pulses. At a peak current of 814 ± 2 pA, the root-mean-square transverse nor-
Ultrafast transmission electron microscopy malized emittance of the electron pulses is εn,x = (2.7 ± 0.1 ) · 10−12 m rad in the direction parallel to
Pump–probe the streak of the cavity, and εn,y = (2.5 ± 0.1 ) · 10−12 m rad in the perpendicular direction for pulses
Microwave cavities with a pulse length of 1.1–1.3 ps. Under the same conditions, the emittance of the continuous beam
Coherent ultrashort electron pulses is εn,x = εn,y = (2.5 ± 0.1 ) · 10−12 m rad. Furthermore, for both the pulsed and the continuous beam a full
width at half maximum energy spread of 0.95 ± 0.05 eV has been measured.
© 2018 Elsevier B.V. All rights reserved.

1. Introduction continuous sources, including recent developments that promise a

higher brightness in the future [9]. Furthermore, any possible in-
Ultrashort high quality electron pulses at energies ranging from stabilities in electron emission due to the intrinsic pointing stabil-
30 to 200 keV have become a useful and powerful tool to inves- ity of a drive laser are circumvented.
tigate dynamical systems on sub-picosecond timescales through Recently, it has been shown that pulsing a beam can be done
diffraction [1], imaging [2] or spectroscopy [3], offering a vast using a microwave cavity oscillating in the TM110 mode while
amount of new information. Typically, inside an ultrafast trans- maintaining the low emittance of a continuous source [10–12]. This
mission electron microscope (UTEM) electron pulses are extracted can be accomplished using a conjugate blanking scheme, where
from a photocathode using an intense pulsed laser. Accurately the electron beam is focused at the center of the cavity, allowing
timed with a clocking laser pulse, dynamic processes can then be for 100 fs pulses to be created with a high beam quality. Since the
investigated with pump–probe measurements. Using photoemis- power in the cavity can easily be adjusted, the pulse length can
sion, a very large operational parameter-space can be spanned [4]. also be changed without influencing the electron emission process.
Furthermore, by using sideways illumination of a Schottky emitter, In Fig. 1(a) this chopping principle is shown.
the emission characteristics of the source are maintained, allowing In order to perform pump–probe experiments, the phase of
for high quality electron pulses to be created [5]. these microwave cavities can be accurately synchronized to a
Although photoemission is commonly used in UTEM systems, pump laser pulse. Using state-of-the-art synchronization schemes,
there is an interesting alternative to use a blanking method, where timing jitter between the electron pulses and the laser pulses can
a continuous beam is periodically swept across a slit or aper- be suppressed to levels well below 100 fs [13,14].
ture [6–8]. Creating pulses in this way has the advantages that Alternatively, it has been proposed to use a microwave signal
amplified laser systems are no longer required, and that no intru- as a pump pulse to drive electronic or semiconductor devices for
sive alterations to the source have to be made. Instead, the system laser-free stroboscopic imaging with repetition rates in the GHz
benefits from the vast amount of research done on state-of-the-art regime [15]. This is an interesting aspect of using microwave cav-
ities, as they can provide a higher repetition rate and therefore a
∗
higher average current for samples with a fast relaxation time.
Corresponding author.
E-mail address: [email protected] (O.J. Luiten).

https://doi.org/10.1016/j.ultramic.2018.03.012
0304-3991/© 2018 Elsevier B.V. All rights reserved.
86 W. Verhoeven et al. / Ultramicroscopy 188 (2018) 85–89

Since the differential reduced brightness varies throughout the

beam, its maximum on-axis value is often used, called the axial
or peak brightness. Within the typical working regime of a micro-
scope, a large portion of the emitted electrons is cut away at the
condenser aperture, leading to an approximately uniform current
distribution. After focusing the beam at semi-angle α , this then re-
sults in a uniform angular distribution and a Gaussian position dis-
tribution within the beam waist, so that the peak brightness can
be written as
1 I
Br =
V ∗ 2 π 2 α 2 σx σy
|qe | I
= , (2)
me c2 4π 2 εn,x εn,y
with σ x and σ y the root-mean-square (RMS) size of the beam
waist, qe the electron charge, me the electron mass, c the speed
of light, and ε n, x and ε n, y the RMS normalized emittance in the x
and y direction respectively, given by
1  2
εn,x = x  p2x  − xpx 2
me c
γ vz 
Fig. 1. (a) General principle of the creation of pulses using a TM110 deflection cavity, ≈ x2 x2  − xx 2 , (3)
where a continuous beam is deflected over a chopping aperture. Definition of the c
with vz the velocity, px the transverse momentum and x = vx /vz
parameters is discussed in Section 2.2. (b) A typical cavity, with (1) the entrance
aperture, (2) the antenna, (3) the tuning stub, (4) the dielectric material, and (5)
the lid used to close the cavity. Shown left is the side of the cavity, and right is the the angular distribution of the particles. In this equation, . . .  in-
bottom of the cavity with the lid removed. dicates the averaging over a distribution.

2.2. Beam chopping

For samples with slower relaxation times, it has been proposed
to use two perpendicular deflecting modes at different frequencies,
The main advantage of using a microwave cavity is that the low
which can be placed in a single cavity [10]. Electrons will then be
emittance of the continuous beam is maintained in pulsed mode.
created at the difference frequency of these modes, allowing for
This is only the case when using the cavity in a conjugate blank-
the repetition rate to be lowered to tens of MHz. If lower frequen-
ing scheme, in which all electrons deflected by the cavity originate
cies are desired, a fast beam blanker can be used to pick specific
from the same virtual image. For a regular beam blanker conjugate
pulses, which are now separated by tens of ns. In this way, mi-
blanking is achieved by placing a crossover in the pivot point of
crowave cavities can also provide lower repetition rates for sam-
the blanker.
ples with slow relaxation times, allowing for the repetition rate of
Inside a microwave cavity the fields vary rapidly compared to
the setup to be optimized for each experiment.
the transit time of the electrons, so that it is no longer possible to
In order to facilitate the implementation in a TEM column,
distinguish a single pivot point. However, it can be shown that it
these deflection cavities can be filled with a dielectric material,
is still possible to maintain the virtual image by proper placement
which allows for a reduction in both the size and power consump-
of a crossover [10,12]. For a beam chopped by an on-axis aperture,
tion [16]. Fig. 1(b) shows a typical dielectric filled cavity used for
the optimal longitudinal position of this crossover is at the center
chopping an electron beam. Shown to the left is the outside of the
of the cavity.
cavity, and to the right is a bottom view of the cavity with the lid
This is also shown in Fig. 1(a), where the beam is focused at the
removed.
center of the cavity. As a result, it arrives at the chopping aperture
In this paper, the implementation of a TM110 deflection cavity
with a certain width w. Sweeping this beam with a magnetic field
in a TEM is presented. Design considerations are discussed, and the
amplitude B0 and an angular frequency ω over an aperture with
performance of a cavity-based UTEM is demonstrated.
width s results in a full width at half maximum (FWHM) pulse
length of
2. Theory
γ me ( s + w )
τ=  , (4)
2.1. Brightness 4|qe |lB0 sin f π2

An important figure of merit for a charged particle beam is its where l is the distance to the chopping aperture, and f = Lcav /Lmax
current density per unit of solid angle, called the transverse bright- the fractional length of the cavity Lcav compared to the maximum
ness. As the solid angle subtended by the beam, and therefore the useful cavity length Lmax = vz π /ω for which electrons feel exactly
brightness, depends on the beam energy, the beam quality is often half the oscillation period. From this equation it can be seen that
expressed in terms of the reduced brightness, which can be defined in order to create short pulses, the focusing angle has to be small
in differential form as [17] to restrain w from becoming too large.
Besides deterioration of the brightness, increase of the energy
1 ∂ 2I spread is also an important effect that has to be considered. Unfor-
Br = , (1)
V ∗ ∂ A∂  tunately, electrons moving through a cavity will probe the off-axis
with I the current through an area A at a solid angle , and V ∗ = electric fields of the TM110 mode. This will not only cause the total
(1/2 + γ /2 )V the acceleration voltage V multiplied by a relativistic beam energy to change, but also the energy spread to increase. Fo-
correction term, with γ the Lorentz factor. The reduced brightness cusing the beam at the center of the cavity minimizes this increase
is a conserved quantity during acceleration of the electrons. in energy spread, but does not completely eliminate it.
 W. Verhoeven et al. / Ultramicroscopy 188 (2018) 85–89 87

Fig. 2. Schematic overview of the adapted microscope column. Below the C2 aper-
ture the column has been extended by 20.3 cm, in which a cavity, additional de-
flectors, and a chopping aperture have been inserted.

It can be shown that this additional energy spread can be de-

creased further by using a shorter cavity length with a higher field
amplitude in order to maintain the same pulse length [12]. The
tradeoff is that more power has to be dissipated by the cavity,
which brings along technical difficulties. For the results shown in
this paper, a shorter cavity length is chosen at the cost of pulse
length.

3. Methods

For the experiments, a 200 kV FEI Tecnai TEM has been elon-
gated with a 203 mm long vacuum chamber below the C2 aper-
ture. In this chamber a TM110 cavity has been mounted, and an Fig. 3. (a) Minimal focus size of the pulsed electron beam for varying focus position
additional aperture holder has been inserted at the bottom. Above within the cavity. C2 current is increased from left to right. (b) Angular distribution
the cavity an extra set of beam deflectors has been added. Fig. 2 of the pulsed beam, together with a known diffraction ring at 10.6 mrad used as
gives a schematic overview of the adapted column. The distance calibration. (c) Energy spread measured with a spectrometer for both the pulsed
and the continuous beam. (d) Emittance along the long and short axis for each spot
from the center of the cavity to the chopping aperture is l = in (a), plotted against the focus position with respect to the center of the cavity.
122.2 mm. Both apertures are 30 μm in diameter. Curves show the emittance expected from simulations.
In order to prevent an increase in emittance in pulsed mode,
a crossover is placed at the center of the cavity by fixing the
C2 lens current. The field-of-view is controlled with the minicon- the General Particle Tracer (GPT) code [18], in which realistic fields
denser (MC) lens. This means that its original functionality of al- inside the cavity are taken into account, including fringe fields.
tering the divergence at the objective lens is now lost. However,
the appropriate choice of apertures can mimic this functionality.
A water-cooled cavity has been designed with a resonant fre- 4. Results
quency ω/2π = 2.9985 GHz, and a length Lcav = 16.67 mm. The
cavity is loaded with ZrTiO4 , a dielectric material with a high per- Shown in Fig. 3(a) is the pulsed electron beam focused on the
mittivity and low loss tangent. The typical magnetic field ampli- detector for varying longitudinal positions of the crossover in the
tude in such a cavity is B0 = 1.2 ± 0.1 mT at an input power of cavity. Moving from left to right, the current through the C2 lens
10 W [11]. For the measurements shown in this paper, the input is increased, raising the focus position through the point of mini-
signal is amplified to 16 W. mal emittance growth. At either too low or too high currents spot
As the electron beam is swept back and forth by the cavity, is elliptical, with the long and short axes corresponding to the di-
pulses are created twice every oscillation period. However, as these rection parallel to and perpendicular to the sweeping direction of
leave the chopping aperture under different angles [10], half of the cavity, respectively.
these must be blocked. This is currently done with the SA aper- With increasing C2 current, the focused spots in Fig. 3(a) also
ture. rotate. This is due to the change in MC current to refocus the beam
Using a Faraday cup, the current of the beam is measured. The into the detector, which also rotates the beam.
energy spread of the beam is measured using a Gatan ENFINA Fig. 3(b) shows the angular distribution of the pulsed beam. An-
spectrometer, with a dispersion of 0.05 eV/ch. Furthermore, mea- gles were calibrated using a known diffraction ring at 10.6 mrad
surements will be compared to particle tracking simulations using from a typical cross grating sample [19], which is also shown
88 W. Verhoeven et al. / Ultramicroscopy 188 (2018) 85–89

Table 1
Measured parameters of the continuous beam and the
pulsed beam, compared to simulations with the same
continuous beam as input and a magnetic field of 1.45
mT.

Continuous Pulsed GPT

ε n, (pm rad) 2.5 ± 0.1 2.7 ± 0.1 2.62

ε n, ⊥ (pm rad) 2.5 ± 0.1 2.5 ± 0.1 2.51
Iavg (pA) 814 ± 2 2.8 ± 0.3 3.38
Efwhm (eV) 0.95 ± 0.05 0.95 ± 0.05 1.01
τfwhm (ps) 1.31

ity as the temperature changes. In either case, a pulse length of

1.33 ± 0.06 ps is deduced from the fitted curve, whereas a pulse
length of 1.1 ± 0.1 ps is deduced from the current measurement.
Shown in Table 1 are the emittance, current and energy spread
measured for both the continuous and the pulsed beam. Also
shown in this table are simulation results starting with a continu-
Fig. 4. Measured duty
√ cycle as a function of the cavity input power, fitted with the ous beam with the same parameters, and a magnetic field strength
expected behavior ∝ P .
of 1.45 mT inside the cavity. Good agreement is found between
simulations and measurements.

in Fig. 3(b). From this, the focusing angle is determined to be 5. Conclusions and outlook
8.74 mrad.
In Fig. 3(c) the energy spread measured with the spectrometer To summarize, it has been experimentally verified that TM110
for both the pulsed and the continuous beam is shown. The mea- cavities can be used to create a pulsed electron beam in a 200 keV
surement with the pulsed beam seems to give a slightly lower en- TEM without a significant increase in emittance. For pulse lengths
ergy spread. However, the difference is well below the resolution of 1.1–1.3 ps, no measurable increase in energy spread or deterio-
of the spectrometer, and is more likely to be due to small mis- ration in performance of the microscope is found. This makes an
alignments. These can easily arise between the two measurements RF-based UTEM a viable alternative to photocathodes.
as different settings have to be used for a continuous beam to pre- As a next step, cavities will be developed further, allowing for
vent the spectrometer from saturating. synchronization to a clocking laser pulse at a frequency of 75 MHz.
Fig. 3(d) shows the corresponding emittance plotted against the Furthermore, higher input powers will be tested, and smaller chop-
difference in focus position. Also shown in this figure is the emit- ping apertures will be used. With increasing field strength in the
tance found in simulations. These show good agreement. Devia- cavity care must be taken to prevent an increase in energy spread.
tions are attributed to the error in estimating the focus position Reference [12] explains in more detail how this can be achieved. In
from the lens current. The emittance of the continuous beam un- this way, the pulse length can be reduced towards 100 fs, allowing
der the same conditions has been determined to be εn,x = 2.5 ± for pump–probe experiments with both a high temporal resolution
0.1 pm rad. and a high transverse coherence.
From this, it can be seen that the beam quality is unaffected by
the cavity in the direction perpendicular to the streak; parallel to Acknowledgement
the streak the growth of emittance can be minimized by correct
placement of the crossover. At the minimum, both simulations and This work is part of an Industrial Partnership Programme of the
measurement give a negligible increase in emittance. Foundation for Fundamental Research on Matter (FOM), which is
The minimum RMS spotsize found from these measurements part of the Netherlands Organisation for Scientific Research (NWO).
is 0.61 by 0.56 nm, at a focusing angle of 8.74 mrad and a peak The authors would like to thank E.H. Rietman, I. Koole, H.A. van
current of 814 ± 2 pA. From Eq. (2), we find a peak brightness of Doorn, and A.H. Kemper for their invaluable technical support.
6.6 · 106 A/(m2 sr V). As a comparison, the RMS spotsize of the
continuous beam has been measured to be 0.55 nm, resulting in a References
brightness of 7.5 · 106 A/(m2 sr V). However, the actual brightness is
presumably larger, since the measured spotsize also includes con- [1] G. Sciaini, R. Miller, Femtosecond electron diffraction: heralding the era of
atomically resolved dynamics, Rep. Prog. Phys. 74 (2011) 096101, doi:10.1088/
tributions from abberations. 0034-4885/74/9/096101.
In order to estimate the field strength in the cavity and the as- [2] D. Flannigan, A. Zewail, 4D electron microscopy: principles and applications,
sociated pulse length, the currents of both the continuous beam Acc. Chem. Res. 45 (2012) 1828–1839, doi:10.1021/ar3001684.
[3] R. van der Veen, T. Penfold, A. Zewail, Ultrafast core-loss spectroscopy in four-
and the pulsed beam have been measured with a Faraday cup dimensional electron microscopy, Struct. Dyn. 2 (2015) 024302, doi:10.1063/1.
at different cavity input powers. Fig. 4 shows the measured cur- 4916897.
rent in pulsed mode divided by the continuous current on the [4] D. Plemmons, D. Flannigan, Ultrafast electron microscopy: instrument response
from the single-electron to high bunch-charge regimes, Chem. Phys. Lett. 683
left y-axis as a function of input power. The right axis shows the
(2017) 186, doi:10.1016/j.cplett.2017.01.055.
corresponding pulse length acquired in the simulations with the [5] A. Feist, N. Bach, N. da Silva, T. Danz, M. Möller, K. Priebe, T. Domröse, J. Gatz-
same
√ ratio. The solid line shows a fit with the expected behavior
mann, S. Rost, J. Schauss, S. Strauch, R. Bormann, M. Sivis, S. Schäfer, C. Ropers,
Ultrafast transmission electron micrroscopy using a laser-driven field emitter:
∝ P [16]. At an input power of 15.3 W, a magnetic field strength
femtosecond resolution with a high coherence electron beam, Ultramicroscopy
of 1.45 ± 0.06 mT is expected from the fit, in good correspondence 175 (2017) 63, doi:10.1016/j.ultramic.2016.12.005.
with values determined before on similar cavities [11]. [6] L. Oldfield, A rotationally symmetric electron beam chopper for picosecond
At higher input powers, the measured current is smaller than pulses, J. Phys. E 9 (1976) 6, doi:10.1088/0022-3735/9/6/011.
[7] T. Hosokawa, H. Fujioka, K. Ura, Generation and measurement of subpicosec-
expected. This could be due to measuring errors or instabilities in ond electron beam pulses, Rev. Sci. Instrum. 49 (1978) 624, doi:10.1063/1.
the electron beam, or by a change in the quality factor of the cav- 1135464.
 W. Verhoeven et al. / Ultramicroscopy 188 (2018) 85–89 89

[8] I. Weppelman, R. Moerland, J. Hoogenbeem, P. Kruit, Concept and design of a 3 GHz radio frequency electric field inside a resonant cavity, Appl. Phys. Lett.
a beam blanker with integrated photoconductive switch for ultrafast electron 103 (2013) 141105, doi:10.1063/1.4823590.
microscopy, Ultramicroscopy 184 (2018) 8, doi:10.1016/j.ultramic.2017.10.002. [14] M. Walbran, A. Gliserin, K. Jung, J. Kim, P. Baum, 5-Femtosecond laser-
[9] H. Zhang, J. Tang, J. Yan, Y. Yamauchi, T. Suzuki, N. Shinya, K.akajima, L. Qin, electron synchronization for pump-probe crystallography and diffraction,
An ultrabright and monochromatic electron point source made of a LaB6 Phys. Rev. Appl. 4 (2015) 044013, doi:10.1103/PhysRevApplied.4.044013.
nanowire, Nat. Nanotechnol. 11 (2016) 273, doi:10.1038/nnano.2015.276. [15] J. Qui, G. Ha, C. Jing, S. Baryshev, B. Reed, J. Lau, Y. Zhu, GHZ laser-free time-
[10] A. Lassise, Miniaturized RF Technology for Femtosecond Electron Microscopy, resolved transmission electron microscopy: a stroboscopic high-duty-cycle
Eindhoven University of Technology, 2012 Ph.D. Thesis. method, Ultramicroscopy 161 (2015) 130–136, doi:10.1016/j.ultramic.2015.11.
[11] W. Verhoeven, J. van Rens, M. van Ninhuijs, W. Toonen, E. Kieft, P. Mutsaers, 006.
O. Luiten, Time-of-flight electron energy loss spectroscopy using TM110 deflec- [16] A. Lassise, P. Mutsaers, O. Luiten, Compact, low power radio frequency cav-
tion cavities, Struct. Dyn. 3 (2016) 054303, doi:10.1063/1.4962698. ity for femtosecond electron micsoscopy, Rev. Sci. Instrum. 83 (2012) 043705,
[12] J. van Rens, W. Verhoeven, J. Franssen, A. Lassise, X. Stragier, E. Kieft, P. Mut- doi:10.1063/1.3703314.
saers, O. Luiten, Theory and particle tracking simulations of a resonant ra- [17] P. Hawkes, E. Kasper, Principles of Electron Optics II: Applied Geometrical Op-
diofrequency deflection cavity in TM110 mode for ultrafast electron microscopy, tics, Academic Press, 1989.
Ultramicroscopy 184 (2018) 77, doi:10.1016/j.ultramic.2017.10.004. [18] (See http://www.pulsar.nl/gpt for more information on the software).
[13] G. Brussaard, A. Lassise, P. Pasmans, P. Mutsaers, M. van der Wiel, O. Luiten, [19] (Agar Scientific, Cross Grating S106. For more information, see http://www.
Direct measurement of synchronization between femtosecond laser pulses and agarscientific.com).