Electron Beam Generation for Transmission Electron Microscope

Abstract

In one aspect, the present invention provides a method of generating an electron beam in a transmission electron microscopy device. The method includes: generating an electron pulse [306] by a pulsed electron source [300], accelerating the electron pulse in a first resonant microwave cavity [302], passing the accelerated electron pulse through a drift space [314], and correcting the energy spread of the accelerated electron pulse in a second resonant microwave cavity [304] by operating it out of phase by 90 degrees from the first resonant cavity [302].

Claims

1. A method of generating an electron beam in a transmission electron microscopy device, the method comprising: generating an electron pulse by a pulsed electron source; accelerating the electron pulse in a first resonant microwave cavity; passing the accelerated electron pulse through a drift space; and correcting an energy spread of the accelerated electron pulse in a second resonant microwave cavity by operating the second resonant microwave cavity out of phase by 90 degrees from the first resonant cavity.

2. The method of claim 1 wherein accelerating the electron pulse in the first resonant microwave cavity comprises passing the electron pulse through a booster cavity followed by a decompression cavity operating out of phase by 90 degrees from the first resonant cavity and out of phase by 180 degrees from the second resonant cavity.

3. The method of claim 1 wherein generating the electron pulse by the pulsed electron source comprises delivering the pulse in phase with an oscillation of the fields in the first resonant microwave cavity.

4. The method of claim 1 wherein accelerating the electron pulse in the first resonant microwave cavity comprises passing the electron pulse through the first resonant microwave cavity operating in TM.sub.010 mode.

5. The method of claim 1 wherein correcting the energy spread of the accelerated electron pulse in the second resonant microwave cavity comprises passing the electron pulse through the second resonant microwave cavity operating in TM.sub.010 mode.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0016] FIG. 1 shows the evolution of the longitudinal phase space distribution of a freely propagating pulse at three subsequent positions along the beam path, according to an embodiment of the invention.

[0017] FIG. 2 is a graph of on-axis electric field strength vs time, showing both the electric field in case of regular single-frequency operation and the electric field resulting if the second harmonic is added, according to an embodiment of the invention.

[0018] FIG. 3 is a schematic drawing of an energy-spread-corrected electron microscope, according to an embodiment of the invention.

[0019] FIG. 4 is a schematic drawing of an energy-spread-corrected electron microscope including a decompression cavity, according to an embodiment of the invention.

DETAILED DESCRIPTION

[0020] The invention provides techniques for producing electron beams with very high energy stability using resonant RF cavities. In one aspect, two RF cavities oscillating 90 degrees out of phase are used, where the second cavity reduces the energy spread of the beam accelerated by the first cavity. The electron beam is pulsed at a time scale much less than the cavity oscillation period. The technique allows for the realization of transmission electron microscopy, which requires sub-eV energy spread, without the need for conventional high-voltage electrostatic acceleration. Such a combination of RF cavities may be realized in an electron microscope.

[0021] Embodiments of the invention include a combination of (at least) two separate resonant RF cavities. In the present description, each cavity can also be implemented as a set of (RF) cavities. Thus, without loss of generality, the following description will refer to first and second cavities, with the understanding that this is equivalent to first and second sets of cavities.

[0022] In a preferred embodiment, first RF cavity accelerates electrons to the desired beam energy, in which are subsequently sent into the second cavity. In the second cavity, the electrons experience RF fields oscillating 90 degrees out of phase with the fields experienced in the first cavity. In the second cavity, deviations from the targeted beam energy due to variations in the RF power are corrected by decelerating electrons which have too much energy and accelerating electrons which have too little energy.

[0023] A pulsed electron source is used with a pulse duration much shorter than the RF period, which delivers pulses in phase with the oscillation of the RF fields in the cavities. The oscillatory fields in the RF cavities are phase-locked, which can be realized by driving them all with a single RF source. The pulsed electron source can be realized by either pulsed (sub-)ps laser photoemission, pulsed (sub-)ps laser photoionization, or by pulsed field emission with RF fields. By using mode-locked lasers, the laser pulses can be synchronized to the phase of the RF fields at the appropriate level of accuracy.

[0024] Embodiments of the present invention make it possible to use RF acceleration while retaining low energy spread and thus realize low-energy spread, radiofrequency electron microscopy. For example, a few 100 keV TEM can be realized without the need of high-voltage electrostatic acceleration, while retaining sub-eV energy spread.

[0025] Following is a description of the theoretical principles underlying embodiments of the present invention. Consider an RF pillbox cavity with electromagnetic (EM) fields resonantly oscillating in the TM.sub.010 mode, with the symmetry axis (z-axis) of the cylindrical cavity coinciding with the electron beam. In this case the electric field only has a field component in the z-direction:

(x,y,z,t)=E(x,y,z)sin(t+.sub.0)(1)

where is the RF oscillation frequency and .sub.0 is the RF phase at time t=0. Close to the z-axis the field in a pillbox cavity can be approximated by:

(x,y,z,t)E.sub.0 sin(t+.sub.0)(2)

[0026] This approximation simplifies calculations but is not essential for the invention. Other resonant modes can be used as well. The momentum p.sub.z that is gained by a charged particle with charge q during passage through the acceleration cavity is given by:

[00001]$\begin{array}{cc}{p}_z={}_{-t_0}^{t_0}.Math.{\mathrm{qE}}_0.Math.\sin \left(.Math..Math.t+{}_0\right).Math.\mathrm{dt}=\frac{{\mathrm{qE}}_0}{}.Math.\left(\cos \left(.Math..Math.t_0+{}_0\right)-\cos \left({}_0-.Math..Math.t_0\right)\right)=-\frac{2.Math.{\mathrm{qE}}_0}{}.Math.\sin .Math..Math.{}_0.Math.\sin .Math..Math..Math..Math.t_0& \left(3\right)\end{array}$

[0027] For .sub.0=/2 and t.sub.0=/2 the maximum momentum gain is realized:

[00002]$\begin{array}{cc}{p}_{z,\max }=-\frac{2.Math.{\mathrm{qE}}_0}{}& \left(4\right)\end{array}$

[0028] Assume a particle with mass m is injected into the cavity at t=t.sub.0=/2 with a velocity much smaller than the final velocity. This assumption is not essential but simplifies the calculations. Furthermore we neglect relativistic effects, which are not essential to explain the invention. The kinetic energy after acceleration is then given by:

[00003]$\begin{array}{cc}U\frac{2.Math.q^2.Math.E_0^2}{m.Math..Math.{}^{.Math.2}}& \left(5\right)\end{array}$

[0029] As an example, consider a pillbox RF cavity, oscillating in TM.sub.010 mode at a frequency of /2=3 GHz with an electric field amplitude E.sub.0=15 MV/m. When an electron is injected with zero velocity at the proper RF phase for maximum acceleration it will be accelerated in this field to U200 keV, comparable to beam energies typical for state-of-the-art TEMs. This can be achieved in a RF cavity z=qE.sub.0/m.sup.2=2 cm in length with an RF power supply of a approximately 10 kW. By accelerating the electrons in several coupled cavities in a row the same final energy can be achieved with much less RF power.

[0030] Equation (3) shows how the final particle momentum after acceleration depends on .sub.0 and t.sub.0, i.e. on the synchronization of the moment of injection of the electron and the RF oscillation. For t.sub.0=/2 and .sub.0<<1:

[00004]$\begin{array}{cc}{p}_z\frac{2.Math.{\mathrm{qE}}_0}{}.Math.\left(1-\frac{{}_0^2}{2}\right)& \left(6\right)\end{array}$

[0031] RF phase instability .sub.0 will therefore translate in relative momentum and energy spread:

[00005]$\begin{array}{cc}.Math.\frac{.Math..Math.U}{U}.Math.=2.Math..Math.\frac{.Math..Math.p_z}{p_z}.Math.=.Math..Math.{}_0^2& \left(7\right)\end{array}$

[0032] For state-of-the-art TEMs, an energy stability |U/U|10.sup.6 is desired, and therefore an RF HI phase stability ||10.sup.3. For an RF frequency /2=3 GHz this corresponds to 50 fs electron injection stability. This is within present-day technical possibilities.

[0033] The energy stability of RF acceleration is generally limited by the stability of the high-power RF amplifier: |(U/U).sub.amp|=|E.sub.0/E.sub.0|10.sup.4, which is the main reason RF amplification is generally not used in Transmission Electron Microscopy.

[0034] Since the final energy after RF acceleration depends strongly on the RF phase, the resulting beam is pulsed, with a pulse length much shorter that the RF period: <<2/, and a maximum pulse repetition rate f.sub.rep,max=/2. In FIG. 1 the evolution of the longitudinal phase space distribution of a freely propagating pulse is illustrated by plotting the distribution of particles in longitudinal phase space at three subsequent spatial observation positions in the beam path. In the three plots the energy of the particles with respect to the average particle energy is plotted along the vertical axis; along the horizontal axis the arrival time of the particles at the observation point is plotted with t=0 corresponding to the arrival time of the central particle. The distributions 100, 102, 104 are schematically indicated by uniformly filled ellipses.

[0035] The initial distribution 100 is an uncorrelated time-energy distribution with an initial pulse duration .sub.i and a finite energy spread U.sub.i=v.sub.zp.sub.z, with v.sub.z the average velocity of the bunch of particles. The pulse duration will increase as the pulse propagates, with the more energetic, faster particles moving ahead and the less energetic, slower particles lagging behind, resulting in a distribution 102 in which particle energy U and time t are, to a good approximation, linearly correlated: a chirped beam. After a drift length L the most energetic particles will outrun the least energetic ones by a time difference =L v.sub.z/v.sub.z.sup.2. The energies of the individual particles do not change and the area of the elliptical distribution in longitudinal phase space is conserved: as the ellipse is tilted it is stretched in time, it becomes both longer and thinner.

[0036] The drift space gives rise to an energy-time or longitudinal-velocity-position correlation, which can be undone with the RF correction cavity to produce distribution 104, which has reduced energy spread.

[0037] The linear correlation can be undone by the time dependent field of a second RF cavity in TM.sub.010 mode, operated with a field amplitude E.sub.1 and a phase .sub.1, so that the oscillating field close to the z-axis is given by:

(x,y,z,t)E.sub.1 sin(t+.sub.1){circumflex over (z)}(8)

[0038] If the pulse enters the second cavity at time t=t.sub.1 and exits at time t=t.sub.1, the momentum change is given by

[00006]$\begin{array}{cc}.Math..Math.p_{z,1}={}_{-t_1}^{t_1}.Math.E_1.Math.\sin \left(.Math..Math.t+{}_1\right).Math.\mathrm{dt}=-\frac{2.Math.{\mathrm{qE}}_1}{}.Math.\sin .Math..Math.{}_1.Math.\sin .Math..Math..Math..Math.t_1& \left(9\right)\end{array}$

[0039] If the particles at the center of the pulse experience a phase .sub.1=0, no net momentum is gained after having passed through the cavity. The particles at the front of the pulse then experience a phase .sub.1=/2 and those in the back a phase .sub.1=/2. If the second cavity is at a distance L form the acceleration cavity we have =L v.sub.z/v.sub.z.sup.2<<1 so that

p.sub.z,1=qE.sub.1L.Math.sin t.sub.1v.sub.z/v.sub.z.sup.2(10)

[0040] To undo the linear energy-time correlation, the momentum change induced by the second cavity should be equal to:

p.sub.z,1=p.sub.z/2(11)

which then results in an expression for the electric field amplitude E.sub.1 for undoing the linear energy chirp acquired during drift over a distance L:

[00007]$\begin{array}{cc}{E}_1=\frac{-U}{\mathrm{qL}.Math..Math.\sin .Math..Math..Math..Math.t_1}& \left(12\right)\end{array}$

[0041] For a well designed correction cavity t.sub.1=/2 and therefore |E.sub.1,min|=U/qL. Example: for a beam energy U=400 keV and a distance L=0.5 m between accelerator cavity and correction cavity, a modest electric field amplitude E.sub.1=0.8 MV/m can is sufficient to undo the energy chirp and thus minimize the energy spread. Note that any energy spread U present at the beginning of the drift space (also the energy spread due to the limited stability of the RF amplifier) will be minimized as long as L v.sub.z/v.sub.z.sup.2<<1.

[0042] The ultimately achievable minimum energy spread can be estimated using the fact that the product of pulse duration and energy spread U is conserved for an unchirped beam, i.e., in absence of energy-time correlation (see FIG. 1):

.sub.i.Math.U.sub.i=.sub.f.Math.U.sub.f(13)

[0043] For example: a 100 fs pulse with 10.sup.4 relative energy spread can be converted into a 10 ps pulse with 10.sup.6 relative energy spread. The achievable energy spread is limited by the fact that the final pulse length should be much smaller than the RF period, so that the electron pulse experiences an electric field in the second cavity that changes sufficiently linearly in time.

[0044] In practice, the effectiveness of the correction cavity, and thus the achievable minimum energy spread, is also limited by (1) variations of the electric field amplitude E.sub.1 in the correction cavity; (2) variations in the arrival time of the pulse in the correction cavity; (3) nonlinearity of the time dependence of the electric field experienced by the electrons in the correction cavity. These will now be discussed in more detail.

[0045] The dependence of the final energy spread on variation of the electric field amplitude of the correction cavity is given by U.sub.f=U.sub.i (E.sub.1/E.sub.1), which is completely negligible for typical variations E.sub.1/E.sub.110.sup.4, caused by the limited stability of the RF amplifier.

[0046] Jitter of the cavity phase .sub.1 with respect to the arrival time of the electron pulse will translate in arrival time jitter .sub.f=.sub.1/, which amounts to less than 100 fs using presently available synchronization electronics. As a result .sub.f/.sub.f<<1, which makes this contribution negligible as well.

[0047] Around a zero crossing the electric field in the correction cavity can be approximated by E.sub.z(t)=E.sub.1(t(t).sup.3/6+(t).sup.5/120+ . . . ) where t is the time with respect to the zero crossing. For proper operation of the correction cavity the field experienced by the electrons should have a linear time dependence. Using |t|=/2 for the electrons in the front or the back of the pulse this implies (.sub.f).sup.2/24<<1. Assuming, for example, that we allow 1% nonlinearity, we find that .sub.f0.5.sup.125 ps for a 3 GHz correction cavity.

[0048] Ideally the correction cavity should have an exactly linear time dependence, i.e. a saw tooth function, in which case .sub.f150 ps for a 3 GHz cavity.

[0049] The linearity of the electric field around the zero crossing can be improved, however, by allowing higher harmonics in the cavity as well. For example, adding a field at the second harmonic with the proper amplitude:

E.sub.z(t)=E.sub.1 sin(t)(E.sub.1/8)sin(2t)(14)

will eliminate third order nonlinearities at the zero crossing. The field given by Eq. (14) is illustrated in FIG. 2, in which the on-axis electric field strength is plotted as a function of time for two cases. One curve 200 shows the electric field in case of regular single-frequency operation, i.e. only the first term on the righthand side of Eq. (14); the other curve 202 shows the electric field resulting if the second harmonic is added with an amplitude as given in Eq. (14) Since the third order nonlinearity at the zero crossing has been removed, the nonlinearity criterion becomes (.sub.f/2).sup.4/120<<1. Allowing 1% nonlinearity now results in .sub.f2.sup.1100 ps and therefore substantial further reduction of the energy spread.

[0050] For a RF TEM according to embodiments of the present invention, a pulsed electron source is used, which produces ultrashort electron pulses that are synchronized to the RF phase of the cavities with a phase jitter .sub.i. For an initial pulse length .sub.i=100 fs the RF phase jitter should be 210.sup.3. The phase jitter can be accomplished by either creating the electron pulses (1) directly in phase with the oscillating RF field or by generating electron pulses by (2) photoemission or (3) photoionization with a mode-locked femtosecond laser that is synchronized to the RF oscillation. The choice of the electron source is important, as it determines the initial pulse duration, the beam current, and the beam emittance.

[0051] Using an RF cavity in TM.sub.110 mode as an ultrafast beam blanker, the continuous beam emitted by a standard field emission gun can be turned into a pulsed beam with conservation of the beam emittance and energy spread. The pulse train will have a repetition rate equal to the RF frequency and pulse durations of .sub.i100 fs can be achieved straightforwardly. The current may be boosted by applying an RF accelerating field to the field emitting tip so that electrons will preferentially be emitted when the oscillating RF field is maximal. These pulses are not necessarily ultrashort, but an RF beam blanker may be used to cut out the central part of the pulse, where the current peaks.

[0052] By femtosecond laser photoemission from a flat metal photocathode 100 fs electron pulses are readily created, but with a relatively poor emittance due to the large emission area. The resulting beam quality is insufficient for high-resolution TEM. Femtosecond pulses may also be generated by illuminating a sharp field emission tip from the side by a femtosecond laser pulse, with the linear polarization of the laser pulse pointing in the acceleration direction. Field enhancement of the optical field then leads to the emission of ultra-low-emittance femtosecond electron pulses. By femtosecond photoionization of an ultracold, laser-cooled atomic gas in a magneto-optical trap, low-emittance femtosecond electron pulses can be generated.

[0053] FIG. 3 shows, according to one embodiment, a schematic drawing of a possible realization of an energy-spread-corrected electron microscope operating at 3 GHz and accelerating to 200 keV. The device includes three major parts: a 5 keV pulsed electron source 300, delivering (sub)picosecond electron pulses, synchronized to the phase of a radio frequency (RF) oscillator; the first resonant RF cavity 302, which acts as a 3 GHz RF accelerator, boosting the energy of the electron pulses 306 supplied by the pulsed electron source from 5 keV to 200 keV; the second RF resonant cavity 304, which has its the RF fields oscillating 90 degrees out of phase with the fields in the first cavity 302, thus acting as a 3 GHz RF energy spread corrector. In this embodiment, the pulsed source 300 is operated at 5 kV and the electron pulses 306 produced by the source travel 0.03 m before entering the first cavity 302. The electron beam path is indicated by the dashed line 308. The oscillating on-axis RF fields are indicated by thick arrows. The sizes of the RF cavities indicated are typical for pillbox cavities operating at 3 GHz in TM.sub.010 mode. A single 3 GHz RF power supply 310 feeds RF power into both the first cavity 302 and the second cavity 304, thus assuring a fixed phase relation between the RF fields in cavities 302 and 304. A RF power supply has a RF oscillator and a RF amplifier. The first cavity 302 and the second cavity 304 can be operated each with its own RF amplifier. The signal fed into the RF amplifiers, however, is derived from a single RF oscillator to assure a fixed phase relation. A phase shifter 312 allows adjustment of the RF phase difference between cavity 302 and 304. The cavities 302 and 304 are separated by a drift space 314 of length L. In this embodiment the drift space 314 has length L=0.4 m. More generally, the value of L is preferably sufficiently large that the final .sub.f is substantially larger than .sub.i. So, if .sub.i is really small, then L can be relatively small as well, in theory. In practice, L is preferably larger than the cavity length, otherwise there would be no well-defined drift space. This implies that, in practice, L is preferably at least 10 cm.

[0054] The drift space between booster 302 and corrector 304 results in a correlation between position and longitudinal velocity in the electron pulses arriving at the corrector 304 (see FIG. 1) with the most energetic electrons arriving first and the least energetic electrons arriving last. By operating the corrector 304 at the proper RF field amplitude, which depends on the average energy of the electron pulse and the length of the drift space, the correlation can be undone, resulting in a reduced energy spread. By increasing the length of the drift space 314, the energy spread can be reduced further, until the pulse length acquired during drift becomes equal to one third of the RF period, e.g., 100 ps pulse length for 3 GHz RF frequency.

[0055] In practice, the availability of RF equipment and the sizes of the cavities allow the use of RF frequencies in the range 1-10 GHz. Low frequencies in principle allow further reduction of the energy spread but require large cavities and relatively large RF power.

[0056] The schematic drawing in FIG. 3 shows the possible realization for an RF frequency of 3 GHz. The cylindrically symmetric TM.sub.010 pillbox cavities 302 and 304 are drawn to scale. In this realization, accelerator 302 is comprised of three stacked TM.sub.010 pillbox cavities, with the length of each cavity such that the electrons experience half an RF period during passage. The three coupled cavities in booster module 302 collectively oscillate in three different resonant modes at slightly different resonant frequencies. The resonant mode that is selected has neighboring cavities oscillating in counter phase. As a result, the electrons keep on being accelerated when they move from one cavity to the next. As the speed increases, the length of the cavities is adjusted accordingly.

[0057] In the example of FIG. 3, the source 300 creates electrons by pulsed photo-emission, photo-ionization or field emission, and they are injected with an energy of 5 keV into the RF booster module 302. In module 302 a moderate RF field amplitude of E.sub.0=5 MV/m is used to accelerate the electrons to 200 keV. In an optimized cavity design (not the pillbox cavities shown here) this corresponds to P.sub.0=3 kW of RF power. The length of the drift space 314 in this example is L=0.4 m and the correction cavity has a length of 0.01 m. The RF electric field amplitude used to undo the correlation is E.sub.1=2 MV/m. In an optimized cavity design this corresponds to P.sub.1=80 W.

[0058] For very small initial energy spreads, an impractically long drift space would be required to reduce the energy spread even further. By adding an additional decompression cavity, the length of the drift space can be reduced significantly. FIG. 4 is a schematic overview of a device according to an embodiment of the present invention, which is similar the embodiment of FIG. 3, except for the addition of a decompression cavity 404 and phase shifter 410. A pulsed source 400 generates electron pulses that follow a beam path 414. A single RF power supply 408 feeds power in the three RF cavities 402, 406 and 404. The phase shifters 410 and 412 allow adjustment of the relative phase between RF cavity 402 and 404 and the relative phase between cavity 402 and 406, respectively. Cavity 404 is identical in design to correction cavity 406 and is positioned in between booster cavity 402 and corrector cavity 406, oscillating 90 degrees out of phase with respect to the fields in 402, i.e. in counter phase (180 degrees) with respect to the field in corrector 406. Cavity 404 increases the correlated energy spread and thus minimizes the required length of drift space between 402 and 406.