betaFeedback

WorksPhysics and Astronomy

The Coming of Age of the Ultracold Electron Source: A Review

Abstract

The ultracold electron source is a unique approach to the generation of high-brightness electron beams. We give an overview of its development over the past 20 years, including the underlying physical principles, technical details and recent experiments, and give a flavor of the exciting prospects that the future may hold.

Introduction

Two decades ago, a radically different approach to the generation of high-brightness, low-emittance electron beams was born: based on the photo-ionization of laser-cooled atoms [1]. Dubbed the UltraCold Electron Source or ’UCES’, it is characterized by an extremely low (order 10 K) transverse electron temperature, promising for the generation of high-brightness electron beams. Such beams find applications in e.g. ultrafast electron diffraction (UED) [2, 3, 4], compact X-ray sources based on inverse Compton scattering [5, 6], or even as injectors for free-electron lasers (FEL) [7, 8]. Now, twenty years later, the time is ripe to give a complete overview of how the UCES-technology developed, and what promises the future may still hold. While a previous review ([9]) gave an overview of the field in 2016, the present work provides much greater technical detail and includes the developments of the past decade.

We restrict this review to the generation of electron bunches through photoionization of neutral atoms in a magneto-optical trap (MOT). The same principle can be used to produce ultracold ion bunches - this is treated elsewhere ([10–12]). For the closely related technique of creating electron beams through (photo- or field) ionization of neutral atom beams, see e.g. [13, 14, 15].

The paper is organized as follows. In Section II we first recall the theory on electron bunch quality and show why it is attractive to go to low beam temperatures. Then we describe the “Birth” of the ultracold electron source in Section III: the idea, the principle and its first realizations. In section IV we will see how the application of femtosecond pulses in the ionization process lead to lower bunch temperatures than initially expected, sparking a series of studies to achieve a proper understanding of the electron emission process. Dubbed “Youth: play and learn”, this section also covers performed experiments on phase-space shaping and the first electron diffraction experiments. § 7 covers the more recent developments towards usable instruments, thus titled “Adolescence”. We end with Conclusions (§ 8) and an overview of exciting future research directions (§ 9).

Electron Bunch Quality

Brightness and emittance

The ideal electron bunch carries a large amount of charge with a narrow energy spread, in a short time, and can be focused down to a tiny spot. The metric that quantifies this is the bunch brightness:

B^n≡1mcNe(2π)3ϵxϵyϵz,(1)\hat{B}_n \equiv\frac{1}{mc}\frac{Ne}{(2\pi)^3\epsilon_x\epsilon_y\epsilon_z}, \tag*{(1)}

where NN is the number of electrons in a bunch, ee and mm the electron charge and mass respectively. ϵx,y,z\epsilon_{x,y,z} are the emittances in the corresponding directions, and represent the phase-space volume occupied by the bunch. Here, the longitudinal emittance (typically ϵz\epsilon_z) roughly represents the product of bunch length and energy spread, while the transverse emittances (ϵx\epsilon_x and ϵy\epsilon_y) correspond to the focusability of the beam, analogous to the étendue or beam parameter product (BPP) in light optics.

More formally, the (normalized) emittance is defined as

ϵx=1mc⟨x2⟩⟨px2⟩−⟨xpx⟩2,(2)\epsilon_x = \frac{1}{mc}\sqrt{\langle x^2\rangle\langle p_x^2\rangle-\langle xp_x\rangle^2}, \tag*{(2)}

where xx and pxp_x are the position and momentum of each electron (and equivalently for yy and zz), and the brackets denote averaging over the entire bunch. In other words, the emittance is the determinant of the covariance matrix of the beam’s phase space coordinates, and thus a measure of the volume that the beam occupies in this phase space. Following Liouville’s theorem, the emittance is conserved as the electron bunch propagates through a system, although in practice it will increase monotonically. In absence of correlations between position and momentum - as is the case for the transverse dimensions at the source or at a focus - the expression for the emittance reduces to the simple product between beam size and momentum spread:

ϵx=1mc⟨x2⟩⟨px2⟩(at source or focus)=σxσpxmc.(3)\epsilon_x = \frac{1}{mc}\sqrt{\langle x^2\rangle\langle p_x^2\rangle}\quad\text{(at source or focus)} = \frac{\sigma_x\sigma_{p_x}}{mc}. \tag*{(3)}

It is instructive to rewrite the longitudinal emittance in units of time and energy using σz≈cσt\sigma_z \approx c\sigma_t and σpz/mc=σU/mc2\sigma_{p_z}/mc = \sigma_U/mc^2, such that:

ϵz=σtσUmc(4)\epsilon_z = \frac{\sigma_t\sigma_U}{mc} \tag*{(4)}

with σU\sigma_U the rms energy spread of the beam. Often (for many cases in UED for example) this energy spread is deemed less important, and the 5-D or transverse brightness is used instead:

B⊥≡I^(2π)2ϵxϵy=Q(2π)3/2ϵxϵyσt.(assuming Gaussian temporal profile)(5)B_\perp\equiv\frac{\hat{I}}{(2\pi)^2\epsilon_x\epsilon_y} = \frac{Q}{(2\pi)^{3/2}\epsilon_x\epsilon_y\sigma_t}. \quad\text{(assuming Gaussian temporal profile)} \tag*{(5)}

σpx\sigma_{p_x}: Temperature, MTE and coherence length

The momentum spread σpx\sigma_{p_x} appears in different forms in different domains of science. From a statistical physics point of view, σpx\sigma_{p_x} describes the random motion of the particles, and can therefore be intuitively expressed as a beam temperature. Indeed, for a thermal, Gaussian distribution:

σpx2=σpy2=σpz2=mkBT.(6)\sigma_{p_x}^2 = \sigma_{p_y}^2 = \sigma_{p_z}^2 = m k_B T. \tag*{(6)}

This means that the emittance can also be expressed as

ϵx=σxkBTmc2.(7)\epsilon_x = \sigma_x\sqrt{\frac{k_B T}{mc^2}}. \tag*{(7)}

In photocathode development, the transverse momentum spread is commonly expressed as an energy, known as the Mean Transverse Energy:

MTE=⟨px2⟩2m=σpx22m,(8)MTE = \frac{\langle p_x^2\rangle}{2m} = \frac{\sigma_{p_x}^2}{2m}, \tag*{(8)}

which equals kBT/2k_B T/2. And finally, to those seeking to apply a beam to UED experiments for example, it is more intuitive to talk about the transverse coherence length:

L⊥=ℏσpx,(9)L_\perp= \frac{\hbar}{\sigma_{p_x}}, \tag*{(9)}

which needs to be larger than the lattice constant such that constructive interference can occur and a diffraction pattern can be formed. It is intuitive to look at the ratio of the coherence length to the transverse beam size as a definition of the relative coherence of a beam:

C⊥=L⊥σx,=ℏϵxmc=λCϵx,(10)C_{\perp} = \frac{L_{\perp}}{\sigma_x}, = \frac{\hbar}{\epsilon_x mc} = \frac{\lambda_C}{\epsilon_x}, \tag*{(10)}

which, as the second line shows, is inversely proportional to the emittance. In the last step the Compton wavelength λC=ℏ/mc≈0.4\lambda_C = \hbar/mc \approx0.4 pm is introduced.

Routes towards high-brightness beams

In the quest for brighter beams, the most obvious approach may seem to try to extract a high charge from a small source size, to maximize current density at the source. It is indeed a common approach with conventional photocathodes to minimize σx\sigma_x in order to produce low-emittance, high-brightness beams, and techniques like laser-triggered emission from nanotips successfully take this approach to the extreme [16, 17]. However, this approach runs into two limitations. First, the extraction of a charge QQ from a surface gives rise to an image charge field of Eimage=Q/(2πϵ0σx,source2)E_{\mathrm{image}} = Q/(2\pi\epsilon_0\sigma_{x,\mathrm{source}}^2). This image charge field has to stay small compared to the accelerating field (by at least an order of magnitude[18]). For typical accelerating fields of tens of MV/m that means Q/σsource2<100Q/\sigma_{\mathrm{source}}^2 < 100 attoCoulomb/μm2\mu\mathrm{m}^2.

The second limitation is given by the Coulomb repulsion of the electrons, or space charge. These space charge fields are of a magnitude comparable to EimageE_{\mathrm{image}} [19], but their effect on the emittance depends drastically on the profile of the bunch density. Only for uniformly filled ellipsoidal (or ‘waterbag’) bunches are the space charge forces strictly linear in space, avoiding emittance growth [20, 21]. All other density profiles lead to emittance growth, which can quickly become dominant even at moderate bunch densities. For example, ‘pancake’-shaped bunches common in UED, with a Gaussian transverse profile with typical values of σsource=100 μm\sigma_{\mathrm{source}} = 100\,\mu\mathrm{m} and σz=0.4 μm\sigma_z = 0.4\,\mu\mathrm{m}, already show emittance degradation due to transverse shock formation at a charge of just 1000 electrons [22], corresponding to a limit on Q/σsource2Q/\sigma_{\mathrm{source}}^2 that is 4 orders of magnitude lower than that posed by the image charge.

Instead of minimizing σx\sigma_x, the alternative approach towards the realization of a bright beam is to minimize σpx\sigma_{p_x} instead, i.e. to minimize the beam temperature at the source. In modern photocathode development, this is now known as ”intrinsic emittance reduction”. For continuous electron beams, this principle was already applied as early as 1991, in the production of cold electron beam for the stochastic cooling of ion beams at MPIK (see [23]). But to the best of our knowledge, the UltraCold Electron Source was the first pulsed electron source to take this approach, and as this review will show, it takes this approach to the extreme.

The birth of the ultracold electron source

The ultracold electron source was born out of the merger of two research groups at the TU Eindhoven: one focusing on accelerator physics and applications, and the other on atomic physics and quantum technology. Soon the idea rose to investigate the potential of using an ultracold plasma (UCP) [24] as a high-brightness source of electrons.

The principle

In the past 20 years of development, the main principle behind the ultracold electron source has stayed the same. The UCES seminal paper [1] described it as follows, with reference to figure 1:

(I): A cold (T<1 mKT < 1\,\mathrm{mK}) cloud of atoms is trapped in a magneto-optical trap (MOT) [26] in a volume of a few mm3\mathrm{mm}^{3}.

(II): Part of the cold atom cloud is excited to an intermediate state with a quasicontinuous μs\mu\mathrm{s} pulse.

(III): Then, a pulsed-laser beam propagating at right angles to the excitation laser ionizes the excited atoms only within the volume irradiated by both lasers. Here a UCP is formed.

(IV): The bunches are extracted by an electric field.

While this main principle has stayed the same, the experimental details on today’s realization of this source differ considerably from the original proposition. Before describing that evolution we would first like to note four important characteristics unique to the UCES, that were already realized from its inception.

First, the fact that the electrons are created over a volume (instead of a surface like with conventional photocathodes) means that electrons in the back of the bunch experience a larger acceleration potential difference than the electrons in the front of the bunch. Thus, the accelerated electron bunches go through a self-compression point (as indicated in step IV in Figure 1). The initial simulations already predicted sub-ps bunch lengths at this point, which was later confirmed in experiment [27] as we will see in § 7 B. If the bunches are accelerated over a distance daccd_{\mathrm{acc}}, the self-compression point lies at 2dacc2d_{\mathrm{acc}} after the accelerator.

Schematic of the four-step procedure to realize a pulsed UCP electron source

Figure 1. Schematic of the four-step procedure to realize a pulsed UCP electron source. From [25].

Second, as the electron bunch is created at the overlap between the excitation and ionization laser beams, the initial charge distribution can be controlled in 3D. This offers a direct way to create the uniformly filled ellipsoidal bunches mentioned above, suppressing emittance degradation caused by space charge.

Third, while most solid state electron emitters require conditioning and suffer from aging, with the UCES each electron bunch is essentially generated from a ‘fresh’ cathode.

Finally, it is actually not strictly necessary to use a UCP in order to achieve low electron temperatures [28]. The initial transverse momentum spread (or temperature, or MTE) is dominated by the excess energy of the ionizing photons. Even if an atomic gas at room temperature was used, the electrons are extracted well before thermal equilibrium between the electrons and ions is achieved. The higher temperature of the atoms before photoionization does cause a Doppler broadening of the photoionization threshold, which for room temperature atoms would be on the order of 10 K. In other words: as long as the electron extraction is sufficiently fast, a 10 K electron beam can in principle also be produced by near-threshold photo-ionization of a room temperature atomic gas. “Sufficiently fast” here means compared to the plasma period Tp=2πϵ0mne2T_p = 2\pi\sqrt{\frac{\epsilon_0 m}{n e^2}}, which is on the order of a nanosecond for our plasma densities. There are however two reasons to use a UCP, one practical and one fundamental. First, a MOT allows to confine a high density of neutral atoms in a high-vacuum environment, such that the accelerated electrons experience little scattering from background gas. Second, UCPs may allow electron temperatures substantially lower than 10 K, and are therefore a potential way towards reaching the fundamental Fermi-degenerate limit (see Outlook).

Initial design

The UCES’ seminal publication proposed using a “dark” MOT [29], which allows atom densities up to 1018 m−310^{18}\,\mathrm{m}^{-3} to extract a 100 pC electron bunch from a 1 mm31\,\mathrm{mm}^{3} volume, through near-threshold ionization with a nanosecond laser pulse. The narrow spectrum of the nanosecond laser pulse causes the electrons to have a very low initial momentum spread, associated with a bunch temperature on the order of 1 mK. Next, electron disorder-induced heating (DIH) [30] or the liberation of correlation energy [31] causes the temperature to rise to about 10 K on the timescale of the plasma period or 1 ns. The bunch is then to be extracted by an electric field of at least 100 MV/m, which should be switched on in ≲1\lesssim1 ns using laser-triggered spark gap technology [32]. Particle tracking simulations show that combining this (rather extreme) set of parameters with uniformly filled ellipsoidal bunches, either ‘pancake’ or ‘cigar’ shaped, leads to a record peak brightness of B⊥=5×1013 A rad−2 m−2B_\perp= 5 \times10^{13}\,\mathrm{A\,rad}^{-2}\,\mathrm{m}^{-2} or 5×1014 A rad−2 m−25 \times10^{14}\,\mathrm{A\,rad}^{-2}\,\mathrm{m}^{-2}, respectively. The projected repetition rate would be 100 Hz, limited by the MOT loading rate.

First version

Needless to say, the first version of the source did not immediately reach these parameters.

It did however produce record-breaking cold electron bunches. A relatively simple first try ([33]) showed it was possible to extract either electrons (∼1.2\sim1.2 pC in 200 ns rms) or ions (∼0.3\sim0.3 pC in 800 ns rms) from a UCP. The plasma was created as follows: Rb-85 atoms were trapped and cooled in a MOT; The trapping beams were left on, thus populating the 5p5p state; a 480 nm, ns dye laser excited electrons to the 44d44d Rydberg state; The Rydberg atoms spontaneously ionize on a µs timescale, forming a UCP (as described in [34]). An upper limit for the electron bunch temperature was estimated at 500 ±\pm 400 K. After this initial test, a first dedicated UCES was built. As is shown in Figure 2, the entire source was built on a single vacuum flange. The MOT was created with three pairs of counterpropagating laser beams, in combination with a set of water-cooled, in-vacuum current coils in anti-Helmholtz configuration. In this first version, the trapping laser was kept on continuously, and no separate excitation pulse was used (skipping step II from the previous section). Electrons were thus created along a line, with a bunch charge of 10 fC, using a 2 ns (rms) laser pulse from a dye laser with an adjustable wavelength near the ionization limit λ0=479.06\lambda_0 = 479.06 nm, and accelerated by a constant potential applied to the cathode, of down to -30 kV. The electron bunch length was 2 ns (rms), about equal to the laser pulse length.

Schematic and photograph of the first version of the UltraCold Electron Source

Figure 2. The first version of the UltraCold Electron Source. Top: schematic showing the cathode (tapered to reduce reflections in case of HV pulses with sub-ns rise times) and the MOT (orange) created by three pairs of counterpropagating laser beams (red) and a pair of coils in anti-Helmholtz configuration (not indicated). The ionization laser is indicated in blue. Bottom: Picture of the source, showing also the coils, and the MCP used to detect the electrons (right).

The bunch temperature was fully determined by the excess photon energy of the ionization laser, as is nicely shown in Figure 3. This excess photon energy is given by

Source temperature as a function of ionization laser wavelength

Figure 3. Source temperature as a function of ionization laser wavelength λ\lambda. The solid line represents the theoretical temperature due to the excess photon energy in absence of any electric field. The dots represent measured bunch temperatures and the 55 K offset is (partially) attributed to the accelerating field, effectively reducing the ionization threshold. From [35].

Eλ=hc(λ−1−λ0−1),(11)E_\lambda= hc\left(\lambda^{-1} - \lambda_0^{-1}\right), \tag*{(11)}

with λ\lambda the wavelength of the ionization laser and λ0\lambda_0 the field-free ionization limit given above, hh the Planck constant and cc the speed of light. As the ionization laser wavelength is increased, the excess photon energy is reduced. Down to about 50 K, the slope equals dTdλ=2hc3kλ2\frac{\mathrm{d}T}{\mathrm{d}\lambda} = \frac{2hc}{3k\lambda^2} as expected assuming equal distribution of the excess energy over the 3 degrees of freedom. Below about 100 K the temperature drops asymptotically, at the cost of reduced bunch charge, with the lowest measured bunch temperature being 15 K. Note that the value for λ0\lambda_0 stated above is for the field-free case, while in this experiment the applied accelerating field effectively reduces the ionization threshold, qualitatively causing the 55 K offset indicated in the figure. This is known as a Stark shift, with the expression for the

Stark shift energy given by

EF=2EHFF0,(Stark shift energy)(12)E_F = 2E_H \sqrt{\frac{F}{F_0}}, \qquad\text{(Stark shift energy)} \tag*{(12)}

with EH=27.2 eVE_H = 27.2\,\mathrm{eV} the Hartree energy, FF the accelerating field strength and F0=5.14×1011 V/mF_0 = 5.14 \times10^{11}\,\mathrm{V/m} the atomic unit of field strength. We can now write the total excess energy as:

Eexc=Eλ+EF=hc(λ−1−λ0−1)+2EHFF0.(13)\begin{aligned} E_{\mathrm{exc}} = E_\lambda+ E_F \\ &= hc\left(\lambda^{-1} - \lambda_0^{-1}\right) + 2E_H \sqrt{\frac{F}{F_0}}. \tag*{(13)} \end{aligned}

In the same experiment, an alternative approach was also explored: here, the atoms were excited to Rydberg states and subsequently ionized by applying a pulsed field to the cathode (3 kV, 17 ns rise time), creating sub-ns electron bunches. This produced even colder electron bunches, with temperatures down to 9 K.

In order to approach the brightness from the seminal paper, significant efforts were made to develop a pulsed HV source that would be able to switch 1 MV with a sub-ns rise time, employing laser-triggered spark gap technology. Despite substantial progress [36] and previous experience with such switches [37, 38], these goals proved technically too challenging, and the idea was abandoned.

Melbourne initiative

The exciting results described above proved contagious: the group of Rob Scholten at the University of Melbourne built their own ultracold electron source using a different geometry, shown in Figure 4. Here, the Rb atoms are trapped and cooled in a MOT between two parallel plates which form the DC accelerator. The cathode is optically transparent, allowing the trapping laser beams to pass through, while the anode doubles as a mirror for the trapping laser beams (a design borrowed from [39]). Excitation happens through a separate excitation beam, so ionization does not take place on a line, but in a volume defined by the overlap of an excitation laser (CW, switched on for about a µs) and the ionization laser (5 ns), as explained in § 4 A. Their pioneering use of a Spatial Light Modulator (SLM) in the excitation beam enabled novel phase-space shaping experiments, as will be treated in section

The setup used in Melbourne

Figure 4. The setup used in Melbourne: Rb-85 atoms are trapped and cooled in a MOT in a DC accelerator, which accelerates the electrons produced through photoionization in the overlap volume between the ionization and excitation laser beam. The latter can be shaped with a spatial light modulator. From [40].

IV D. Consistent with the experiments in Eindhoven, bunch temperatures down to 10±510 \pm5 K were reported, with a bunch length determined by the length of the ionization laser pulse, 5 ns.

In order to bring this pulse length down and achieve ultrafast, cold electron bunches, the next step for both Eindhoven and Melbourne was to consider ionization by a femtosecond laser pulse.

YOUTH: PLAY AND LEARN

Femtosecond ionization

There is a strong argument against femtosecond ionization: when creating an ultracold plasma, the initial electron temperature (so on timescales shorter than the plasma period TpT_p) is determined by the spectral bandwidth of the ionization laser pulse, on the order of Eexc=hΔfE_{\mathrm{exc}} = h \Delta f.

For a Fourier-limited, Gaussian pulse, (Δt)FWHM(Δf)FWHM=2ln⁡2/π(\Delta t)_{\mathrm{FWHM}}(\Delta f)_{\mathrm{FWHM}} = 2 \ln2 / \pi, so the electrons would then acquire a thermal distribution with kBT=πln⁡2ℏΔtk_{\mathrm{B}}T = \frac{\sqrt{\pi\ln2}\hbar}{\Delta t}. For a 50 fs pulse this corresponds to 225 K, compared to 2 mK for the ns pulses used in the experiment above. This would mean that the curve traced out by the measured source temperature in Figure 3 would plateau at 225 K, severely limiting the electron bunch coherence.

Luckily, the experiment showed that this picture is too simple. In [41], two important changes were made with respect to the first Eindhoven experiment. First, a 50 fs ionization laser pulse was used, from an optical parametric amplifier (OPA) pumped by a Ti:Sa laser system. Second, ionization only took place in the overlap volume of the excitation laser and the ionization laser. The electron bunch temperature was then determined through a waist scan. The blue squares in Figure 5 show the obtained temperatures for different excess energies, with a minimum at 18±418 \pm4 K. Later, these measurements were repeated in [42] with a higher-resolution detector and a higher accelerating field (F=0.813 MV/mF = 0.813\,\mathrm{MV/m}) leading to a minimum bunch temperature of 10 K. These values are well below the limit expected from the spectral bandwidth of the ionization pulse (see [43] for a detailed experimental comparison between fs and ns ionization). To understand how this is possible, we need to take a closer look at the electron emission process, as we will in the next section. Around the same time, the Melbourne group combined femtosecond excitation with nanosecond ionization [40] to produce short pulses. While they initially estimated the pulse length to be determined by energy spread at a minimum of 150 ps, this was not supported by the measurements, and rectified by a later publication [44]. Their ability to shape the excitation laser beam allowed formation of multiple beamlets, from which the emittance could be determined like in a pepper-pot measurement, for which a value of ϵ=538±26\epsilon= 538 \pm26 nm was found.

Femtosecond ionization: experimentally measured bunch temperatures and analytical temperature models

Figure 5. Femtosecond ionization: Experimentally measured bunch temperatures (blue squares) and the temperature expected from equipartition (green dash-dotted line). Also indicated are the curves for the analytical temperature model with and without taking the ionization laser spectral bandwidth into account (blue solid and red dashed resp.). From [45].

Understanding the electron emission process

The naive idea that the excess energy (Eq. (13)) is distributed equally across all three degrees of freedom ignores the fact that the Stark shift is directional. For high photon energies this is irrelevant, but close to the ionization threshold Eλ<EFE_{\lambda} < E_{F} and so electron emission will be confined to a cone opening in the direction of the accelerating field. The Stark shifted atomic potential has a saddle point, with a minimum saddle point energy of −EF-E_{F}. This is illustrated in the top panel of Figure 6, for an accelerating field FF pointing in the negative zz-direction. From this top panel it is already clear that electron emission will be towards the positive zz-direction. To gain quantitative understanding an analytical model is developed in [46], based on a previously developed classical and non-relativistic model for ionization of a hydrogen atom [47, 48]. With this model, closed analytical expressions are derived for the electron trajectories in a combined Coulomb-Stark potential UCS=−1/r+FzU_{CS}=-1/r+Fz (atomic units), as a function of EλE_{\lambda}, FF and β\beta, where β\beta is the starting angle with respect to the zz-axis. Only trajectories with a starting angle sufficiently close to the zz-axis can escape the potential well, i.e. β<βc\beta< \beta_c, with βc=cos⁡−1(−Eλ/EF)\beta_c=\cos^{-1}(-E_{\lambda}/E_F). In Figure 6 these trajectories are drawn for different excess energies. From these trajectories the (asymptotic) transverse velocity spread σvx\sigma_{v_x} and thus the transverse temperature can be calculated, assuming a probability distribution for β\beta. Here, a uniform probability would correspond to an unpolarized laser pulse. For a linearly polarized pulse electrons are preferentially ejected along the laser polarization axis ([49]), following a cos⁡2α\cos^2\alpha distribution, with α\alpha the angle between the initial electron velocity and the laser polarization. This assumes that the electron orbits of the state from which the atom is ionized (5p5p in our case) do not have a preferential direction. This assumption is valid at the center of the MOT, where the magnetic field is zero and the magnetic quantization axis is therefore undefined.

Near-threshold photo-emission from a single atom in an external electric field

Figure 6. Modelling the near-threshold photo-emission from a single atom in an external electric field. Top: The field-free Coulomb potential (green dash-dot), the Stark potential (red dash) and the Combined Coulomb-Stark potential UCSU_{CS} (blue, solid), for Eexc=5 meVE_{\mathrm{exc}} = 5\,\mathrm{meV} and F=−0.155 MV/mF = -0.155\,\mathrm{MV/m}. Bottom: UCSU_{CS} potential landscape in xzxz-plane (FF in zz-direction), with electron trajectories for Eexc=5E_{\mathrm{exc}} = 5, 15 and 25 meV and starting angles 0≤β<βc0 \le\beta< \beta_c, where the color of the trajectory indicates β\beta. From [45].

The model uses a simple Coulomb potential, which is only correct for the hydrogen atom. For a more realistic Rb potential ([46]) the electron trajectories can no longer be described analytically and are calculated instead using the General Particle Tracer (GPT) code [50].

In this potential no closed orbits exist, so all electrons (with Eexc>0E_{exc} > 0) can escape the ion. Electrons with starting angle β<βc\beta< \beta_c will escape the potential immediately as in the analytical model, while electrons with β≥βc\beta\ge\beta_c will make recursions in the potential until they scatter on the ion core. The electron will escape if the ‘new’ β\beta after scattering is smaller than βc\beta_c. The resulting transverse temperature is therefore very close to that of the analytical model.

We can now use this model to quantitatively explain the transverse temperature from the femtosecond ionization experiment, as is drawn in red (dashed) in Figure 5. An important correction is however still missing: the finite spectral bandwidth of the ionization laser pulse. This can be taken into account by convoluting the model curve with a Gaussian distribution corresponding to the spectral bandwidth, cut off at Eexc=0E_{exc} = 0 (as no electrons are created for negative excess energies). This leads to the blue solid curve in Figure 5.

Note that the analytical model is for a single-atom emitter, while in the experiments the transverse bunch temperature is measured from a collection of such emitters. For these measurements to be representative of the single-emitter dynamics, care was taken to avoid space charge effects by keeping the bunch charge low (typically a few hundred electrons).

Longitudinal quality

Now that the effect of the emission dynamics on the transverse beam quality is understood, it is time to consider the longitudinal quality, i.e. the pulse length σt\sigma_t and energy spread σU\sigma_U. First considering σU\sigma_U, two contributions can be distinguished: the (perfectly uncorrelated) energy spread due to the finite longitudinal temperature; and the (perfectly correlated) energy spread due to the length of the ionization volume in the direction of the accelerating field. From the analytical model above, we expect the longitudinal temperature to be much higher than the transverse one, but it will be limited by the total excess energy as

Eexc=k2(Tx+Ty+Tz),(14)E_{exc} = \frac{k}{2}(T_x + T_y + T_z), \tag*{(14)}

which for our cold electrons is typically 1-10 meV. The correlated energy spread due to the length of the ionization volume is given by

σU=σioneF,(15)\sigma_U = \sigma_{ion} e F, \tag*{(15)}

which is typically 0.1-1 eV, so at least an order of magnitude larger than the thermal contribution (this was also confirmed experimentally in [51]). As mentioned in § 4 A, because of this correlated energy spread, the electron bunch of the UCES passes through a self-compression point at 2dacc2d_{\mathrm{acc}} behind the DC accelerator. This is schematically depicted in Figure 7). At this longitudinal focus, pzp_z and zz are uncorrelated, such that the pulse length consists of three contributions: the ionization laser pulse length τlaser\tau_{\mathrm{laser}}, the spread in ionization delays τion\tau_{\mathrm{ion}}, and a thermal contribution τth\tau_{\mathrm{th}} from the uncorrelated energy spread due to the finite longitudinal temperature. So

Schematic of the UCES showing excitation and ionization laser beams and longitudinal phase space diagrams

Figure 7. Schematic of the UCES, showing the excitation (780 nm) and ionization (480 nm) laser beams, and the electrons created in their overlap volume at a distance daccd_{\mathrm{acc}} from the ground electrode. The size of the overlap volume in the direction of the accelerator potential (determined by σion\sigma_{\mathrm{ion}}) causes the electrons to self-compress at a distance 2dacc2d_{\mathrm{acc}} behind the ground electrode. The diagrams below show the longitudinal phase space at different instances. From [52].

τsc=τlaser2+τion2+τth2(16)\tau_{\mathrm{sc}} = \sqrt{\tau_{\mathrm{laser}}^{2} + \tau_{\mathrm{ion}}^{2} + \tau_{\mathrm{th}}^{2}} \tag*{(16)}

and the longitudinal emittance (equation (4)) is then given by:

ϵz=σUmcτlaser2+τion2+τth2(17)\epsilon_z = \frac{\sigma_U}{mc}\sqrt{\tau_{laser}^{2}+\tau_{ion}^{2}+\tau_{th}^{2}} \tag*{(17)}

The first, τlaser\tau_{laser}, is typically tens of fs (rms). The second, τion2\tau_{ion}^{2} can be estimated from the trajectories in the analytical model, and is typically a few picoseconds (depending on the FF, λ\lambda and the polarization of the ionization laser). The third is given by

τth=mkBTzeF.(18)\tau_{th} = \frac{\sqrt{m k_B T_z}}{eF}. \tag*{(18)}

In the first experimental characterization of the longitudinal quality of the electron beam, an RF cavity was used to measure the electron pulse length [42]. Due to geometric constraints, it was not possible to measure the pulse length in the self-compression point (but see § 7 B for the follow-up measurement). Instead, the RF cavity had to be placed further downstream, such that the pulse length measurement was dominated by the energy spread σU\sigma_U. Still, from the measured value of 25 ps we can conclude that it is possible to produce electron bunches that are both ultracold and ultrafast. In addition, a transverse temperature of 10 K was measured, at an excess energy of Eexc=10 meVE_{exc} = 10\,\mathrm{meV}, such that we can estimate TzT_z using equation (14) to be about 225 K, and τth=0.4 ps\tau_{th} = 0.4\,\mathrm{ps}. We can thus safely conclude that the longitudinal emittance is dominated by the spread in ionization delays τion\tau_{ion}. Taking a value of τion=2.5±1.0 ps\tau_{ion} = 2.5 \pm1.0\,\mathrm{ps} (as obtained from the analytical model for typical experimental values), the longitudinal emittance is ϵz=36±14 nm rad\epsilon_z = 36 \pm14\,\mathrm{nm\,rad} or ϵzmc=62±24 ps eV\epsilon_zmc = 62 \pm24\,\mathrm{ps\,eV}.

Around the same time, the Melbourne group also characterized their bunch lengths, using a pair of streaking electrodes. As the Melbourne setup employs fs excitation followed by ns ionisation, the possible ionization pathways differ from those in Eindhoven. This is reproduced in Figure 8. Four ionization pathways can be identified as follows. Sequential excitation (SE), as employed by the setup in Eindhoven, produces cold electrons but the pulse length is determined by the ionisation laser pulse length, as the lifetime of the 5p5p state is 26 ns. With ns ionization, this thus produces ns electron bunches. Multiphoton excitation (MPE) and Resonance-enhanced multiphoton excitation (REMPE) occur at higher laser intensities, and produce hot electrons. Two-color multiphoton excitation (TCMPE) is the only possible route of generating electrons bunches that are both ultracold and ultrashort with this setup. Unfortunately, since this process is non-resonant, the ionization cross-section is three orders of magnitude lower than that of SE.

Ionization pathways and measured transverse momentum distributions

Figure 8. Top: The four ionization pathways that can occur in the Melbourne setup. Bottom: Measured transverse momentum distributions for the different pathways, showing that only SE and TCMPE generate cold bunches. From [44].

Phase-space shaping

The fact that the electrons are created in the overlap of two perpendicular laser beams enables unique opportunities to control the initial shape of the electron bunch. The relatively simple ‘cigar’ or ‘pancake’-shaped bunches - trading longitudinal for transverse coherence or vice versa - were already described in section III B. But from the very start [53], the Melbourne group went much further: the incorporation of a spatial light modulator (SLM) in their excitation beam enabled generation of electron bunches with arbitrary transverse shapes. This is illustrated by the image in Figure 9, which is the measured electron density of an electron bunch generated with an excitation beam in the shape of an atom. Thanks to the low bunch temperature, this shape is largely maintained during propagation down the beamline. This is not just a fun trick, as the generation of initial bunches with sharp edges [53] or consisting of multiple beamlets [40] allowed direct determination of the transverse emittance.

Image not extracted

Figure 9. An electron bunch transversely shaped like an atom, measured with the Melbourne setup. From [53].

In a different experiment [54], the Melbourne group used their shaped excitation beam to track space-charge dynamics. This was done on the ion bunches instead of the electrons, as this is much easier: their lower initial temperature (mK vs ∼10\sim10 K) and their higher mass make that the initial bunch structure is preserved for much longer, and that space charge dynamics are slowed down, approximately by a factor mion/me≈355\sqrt{m_{\mathrm{ion}}/m_e} \approx355. This means that the dynamics of their ion bunches of a few ns should be analogous to the dynamics of electron bunches of about 10 ps. As shown in [54], at the source nine closely spaced bunches were formed, and their Coulomb interactions give rise to compression layers between the bunches and phase-space wave-breaking at the boundary of the full beam, already at moderate bunch charges of a few fC. Building on this result, and still employing ions, [55] compared bunch generation with different transverse profiles, showing how space-charge driven bunch expansion is self-similar for bunches with half-spherical or flat-top profiles, allowing a reduction of space-charge-driven emittance growth by up to 50 % compared to a standard Gaussian profile.

In Eindhoven, efforts on shaping of the initial bunch distribution focused rather on the longitudinal phase space: in [42] the ionisation laser was transversely shaped to create two foci, thus creating two electron bunches initially 90 µm apart in the longitudinal direction. This was further developed in [56], by shaping the excitation beam using an SLM. These results open up the possibility of creating pulse trains, or microbunches, with applications in burst-mode UED or superradiant X-ray generation [57].

The first diffraction experiments

With the peculiarities of the ultracold electron source now well understood, we can discuss the proof-of-principle diffraction experiments that both the Eindhoven and Melbourne groups performed around the same time. In ([58]), the ultracold electron beam was made to diffract of a graphite single crystal. Panel (a) in Figure 10 shows the first-order diffraction spots with the electron beam focused on the detector. In panel (c) and (e) the beam was focused to a tight, micron-sized spot on the sample instead, such that the size of the diffraction spots on the detector was dominated by the transverse coherence of the beam. By changing the ionization laser wavelength from 478 to 498 nm, the transverse temperature at the source is changed from 250 K (c) to 10 K (e), as measured using a waist scan (like in Figure 3 and Figure 5). Together with a source size of σx,source=32 μm\sigma_{x,\mathrm{source}} = 32\,\mu\mathrm{m} these values correspond to a relative coherence of C⊥,source=5.9×10−5C_{\perp,\mathrm{source}} = 5.9 \times10^{-5} and $2.9 \times 10^{-4}$ or, equivalently, an emittance of 6.6 and 1.3 nm rad. The increase in transverse coherence at the sample is evident from the increased brightness and reduced size of the diffraction spots. From this diffraction spot size, the transverse coherence length can be calculated as 0.30 and 0.55 nm resp., which together with the beam size σx,sample\sigma_{x,\mathrm{sample}} (8.7 and 3.3 µm) correspond to $C_{\perp,\mathrm{sample}} = 3.4 \times 10^{-5} and 1.7×10−41.7 \times10^{-4} or, equivalently, an emittance of 11 and 2.3 nm rad. Comparing the transverse coherence at the sample with that at the source (which are measured independently), we can see the effect of apparent emittance growth: a loss of coherence roughly on the order of a factor of two. This illustrates an important challenge in low-emittance beams: creating a low-emittance beam is only the first challenge, the second challenge is to preserve that beam quality down to the sample and detector. The two main causes for such emittance growth are space charge and imperfect electron optics. In these experiments, bunch charges were kept low (a few hundred electrons) to avoid space-charge effects, such that the emittance growth can be attributed to imperfect electron optics. Indeed, a simulation incorporating these imperfections ([43]) reproduces the coherence at the sample to reasonable accuracy (Figure 4 in [58]).

Image not extracted

Figure 10. A first diffraction experiment with the UCES, on mono-crystalline graphite. (a): 13.2 keV electrons are focused on the detector. Indicated are first order spots (1); a second order spot (2); a beam block (3); and the detector edge (4). (c): Diffraction pattern with the beam focused on the sample and a source temperature of 250 K. (e): Idem for a source temperature of 10 K. (b, d and f): One of the diffraction spots (inset) and a lineout along the minor axis of that spot, for each the patterns above. From [58].

Soon after, the Melbourne group also performed a diffraction experiment, with a significant improvement [59]. Their higher emittance and longer pulse length allow for a higher bunch charge before space-charge becomes problematic. Leveraging a bunch charge of 5×1055 \times10^{5} electrons, they performed single-shot electron diffraction of a gold crystal, as shown in the left panel of Figure 11. These single-shot patterns contain enough signal to perform registration of successive patterns, thus allowing compensation for beam jitter and drift. A registered average of 2000 patterns is shown in the right panel of Figure 11, with Bragg reflections up to (6ˉ60)(\bar{6}60).

Image not extracted

Figure 11. Left: Electron diffraction pattern of a gold crystal, from a single cold electron bunch. Right: Diffraction pattern obtained after registration and averaging of 2000 single-shot patterns. Grayscales of the main images are logarithmic, of the insets are linear. From [59].

The Manchester AC-MOT

We will end this section by mentioning the initiative in Manchester to build an ultracold electron source as well [60]. Uniquely, this source was to be based on the AC-MOT developed previously in Manchester [61]. In an AC-MOT, an alternating current runs through the coils that generate the BB-field required for atom trapping and cooling, such that the average BB-field is zero. As a result, no net current is induced in the rest of the setup, allowing the BB-field to be switched of in just tens of μ\mus, as opposed to several ms for a standard DC MOT. Indeed, in both the Eindhoven and Melbourne setups, the BB-field of the MOT is left on continuously, creating a focus (or even multiple foci) in the electron beam downstream. Additionally, keeping the BB-field on also causes a hyperfine splitting of the energy levels of the cold atoms which increases with distance from the trap center. This splitting causes a (small) contributes to the spread in ionization energies, which can be eliminated by wwitching the B-field off before ionization. With the AC-MOT, the B-field can be switched off before each ionization pulse, and still operate at a 5-10 kHz repetition rate. According to the design the source was to generate 22 keV, 1 pC electron bunches at an emittance of 0.35 µm rad and to be the basis of an ultracold electron diffraction facility. This unfortunately never materialized.

Adolescence: towards usable instruments

In this section we will see how the Eindhoven setup is transitioning from an experiment in itself towards a usable instrument. The activities in Melbourne unfortunately ceased after the results described in the previous section.

Towards a compact, stable setup

One of the main challenges in operating the setup of Figure 2, was the careful alignment of the six laser beams for trapping and cooling, in addition to the excitation and ionisation beams, in a configuration that also leaves room for the electron beam to exit the accelerator and ideally even for a camera to image the MOT. The next generation of the source considerable alleviated this challenge through the use of a grating MOT, a technique developed at the University of Strathclyde [62]. This works as follows (see Figure 12). A single trapping laser beam is used, which is incident upon a patterned chip (c). This patterned chip consists of three identical linear gratings, lying in a plane with 120° relative angles (a and b). The incoming laser beam is diffracted by each grating, and these first-order diffracted beams partially overlap with the incoming beam to create a diamond-shaped overlap volume. The force balance between the incoming and diffracted beams can be used to trap and cool atoms [63], similar to the conventional setup with 3 pairs of counterpropagating beams [64].

Image not extracted

Figure 12. Left: Principle of the grating MOT. (a): Orientation of the three parts of the grating chip. (b): SEM image of part of the grating chip. (c): Simple setup of the grating MOT requiring only a single trapping beam. (d): Creation of the overlap volume between incoming and diffracted laser beams. (e): Geometry of the overlap volume. Right: The grating-MOT-based UCES. Adapted from [52] and [56].

By embedding the grating chip in the ground electrode of a parallel-plate DC accelerator (right panel in Figure 12), the principle of the grating MOT can be used to construct an UCES that is more compact and easier to align [65]. It also allows easier optical access for the excitation and ionization beams as well as fluorescence imaging of the trapped atoms. This configuration additionally has the negative electrode of the DC accelerator outside vacuum, avoiding the need for a HV feedthrough. The electrons leave the accelerator through a laser-drilled hole in the center of the grating chip. The whole module is mounted on a single reentrant flange, while the MOT coils (not shown) are outside vacuum. Commissioning of this ‘GMOT-UCES’ is described in [65], which reports measurements of beam energies up to 10 keV at an accelerating field of F=0.37F = 0.37 MV/m. The normalized emittance was measured at 1.9−1.5+0.91.9^{+0.9}_{-1.5} nm rad which, with a source size of σx=30\sigma_x = 30 µm, corresponds to a transverse temperature of 25±2525 \pm25 K. Although seemingly compact and convenient, this setup initially suffered from a major drawback: the Rb+^+ ions created at every shot impinge on the vacuum window of the accelerator. Because of the Rb background pressure, that window surface is covered with a thin layer of Rb atoms, which act as a cesiated cathode surface. As the accelerated Rb+^+ ions hit this surface, a bunch of secondary, hot electrons is generated and accelerated down the beamline, often disturbing the measurement. Additionally, the (dielectric) window surface becomes gradually positively charged, thus gradually reducing the accelerator potential at the MOT position. This changes the electron beam energy, which in turn changes the beam’s focal position and beam pointing. This was solved by implementing a fast HV-switch that quickly inverts the accelerator potential once the electron bunch has

left the accelerator, slowing down the ions before they hit the vacuum window [56]. More recently, carefully designed mu-metal shielding was inserted around the electron beam, between the ground electrode and the sample plane. This greatly reduces the focusing effect of the MOT coils, avoiding the unwanted foci mentioned in section IV F.

Ultracold and ultrashort: sub-picosecond UCES

The improved ease of optical access offered by the new setup finally enabled a measurement that was long overdue: the pulse length at the self-compression point. In [27] this pulse length was measured using a ponderomotive scattering technique, with a setup depicted in Figure 13 (right). By mounting two lenses on the back of the ground electrode of the DC accelerator (at a distance 2dacc2d_{\mathrm{acc}} behind the front of the electrode) it was possible to focus a femtosecond laser pulse (‘interaction’ laser) down to a μm\mu\mathrm{m}-sized spot, reaching a peak intensity of about 1.1×1016 W/cm21.1 \times10^{16}\ \mathrm{W}/\mathrm{cm}^{2}. When this interaction laser overlaps with the electron beam, part of the electrons are ponderomotively scattered in the transverse direction, which is visible on the electron detector downstream. The graph in Figure 13 (left) shows the number of scattered electrons as a function of delay between the interaction laser and the electron bunch. This yields an rms pulse length of τsc=735±7 fs\tau_{\mathrm{sc}} = 735 \pm7\ \mathrm{fs}. The measurement was performed for polarizations parallel and perpendicular to the accelerating field (resp. blue and red[66]), and for six different wavelengths of the ionization laser, i.e. for different excess energies. Consistent with the electron emission model described above, the shortest τsc\tau_{\mathrm{sc}} (Figure 13) corresponds the case where the entire spectrum of the ionisation laser is above hc/λ0hc/\lambda_{0}, such that only electrons with an excess energy larger than the Stark shift are produced, which can directly leave the atom potential. Here, the departure from equipartition due to the directional Stark shift explained in the model of section IV B no longer applies, so these electrons are hot. Indeed, the central wavelength of the measurement of Figure 13 is 474.4 nm, well below the Stark-shifted ionization threshold of 495 nm. This corresponds to an excess energy of 111 meV or an (equipartitioned) temperature of 857 K (while [27] erroneously uses an equipartitioned temperature of 25 K in simulations and the calculation of ϵz\epsilon_{z}). As the spectrum is gradually shifted to lower photon energies, to where most electrons have negative excess energy, τsc\tau_{\mathrm{sc}} broadens to 1 ps with the addition of a few-ps tail [27, 56]. This confirms the estimates from Sec. IV C and [42], that the longitudinal quality of the ultracold electron bunches is determined by the spread in ionization times, which is on a few-ps timescale. This means that there is room to minimize the electron excess energies by reducing the bandwidth of the ionization laser pulse by at least an order of magnitude, thus matching τlaser\tau_{\mathrm{laser}} to τion\tau_{\mathrm{ion}}, at the cost of a minor increase in longitudinal emittance.

Image not extracted

Figure 13. Left: The result of ponderomotive measurement of the pulse length at the self-compression point. In (a) the number of scattered electrons is plotted as a function of delay between the interaction laser and the electron bunch. The ionization laser was polarized either perpendicular (red) or parallel (blue) to the accelerating field. The background signal (no ionization laser) is plotted in black. The insets show typical electron beams on the detector with (c) and without (b) interaction laser. Electrons are counted as ponderomotively scattered if they are detected in the area outlined in red. Right: A picture of the reentrant flange with the DC accelerator. For the ponderomotive measurement two lenses are mounted on the back of the ground electrode, to focus the interaction laser (red). Adapted from [56].

Increasing the electron energy

Other than the stability and ease of use addressed by the new setup, another point requires attention in the transition from an experiment to a useful instrument: the electron energy. While the 1-10 keV energies reported above can be used for experiments in reflection (e.g. SEM or RHEED [59]), experiments in transmission geometry require much higher energies for all but the thinnest (few nm) samples. Unfortunately, the required presence of a MOT limits the accelerating field that can be applied, due to the induced Stark shift of the atomic transitions used for trapping and cooling. The GMOT UCES uses the highest accelerating field so far (1.2 MV/m).

In [67] an RF-cavity was used to post-accelerate the electrons produced by the GMOT-UCES to up to 35 keV, while the transverse quality of the electron bunches was derived from the diffraction pattern of a single gold crystal. The transverse emittances at the sample were found to be between 3-10 nm rad, mainly attributed to imperfect beam dynamics in the RF-cavity, which was originally designed as a compression cavity for 100 keV electrons [3].

A more promising attempt is a dedicated setup that combines DC and RF acceleration. A complete design for such a setup was recently proposed ([68]). After further improvements, the ultracold DCRF source is currently in production. Figure 14 (left) shows the final design, illustrating the complexity that results from the competition between optical, RF, vacuum, high-voltage, thermal and mechanical constraints. A set of preliminary particle tracking simulations (using GPT) shows that a beam energy of 100 keV is reached with limited emittance growth.

Image not extracted

Figure 14. Left: CAD model of the ultracold DCRF source, illustrating the complexity of the setup. Right: Preliminary particle tracking simulations. (Work by A. Patwardhan, subject of a future publication by A. Patwardhan et al.)

Conclusion: Lessons Learned

In summary, the ultracold electron source has evolved into a stable, reliable setup, with a well-characterized electron beam. The fact that the electron beam is generated in a 3-dimensional ionization volume (rather than at a two-dimensional cathode surface) allows for a unique level of control over the initial phase space distribution of the electron beam. A relatively simple, classical model of the ionization process explains the observed low transverse temperatures. In order to fully exploit the potential offered by the ultracold electron source, several research directions are still waiting to be explored.

Outlook

The development that is nearest in the future is that of the DCRF accelerator, expected to bring the electron energy up to 100 keV. Near-to-medium term applications that are envisioned are for example the application to UED and even protein crystallography experiments, injection into a dielectric laser accelerator[69], or as injector for inverse Compton scattering [6]. Highly relevant to such applications is the potential benefit offered by emittance compensation approaches [3]. More exotic would be the generation of spin-polarized electron beams, or beating disorder-induced heating through the use of optical lattices, Rydberg crystals, or even through photodetachment from an anion Coulomb crystal [70]. Clearly, the story of the ultracold electron source has only just begun.

Acknowledgments

J.V. Huijts is supported by the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No. 101066850, and by a Branco Weiss Fellowship - Society in Science, administered by the ETH Zürich.

References

  1. [2]B. J. Siwick, J. R. Dwyer, R. E. Jordan, and R. D. Miller, An atomic-level view of melting using femtosecond electron diffraction, Science 302, 1382 (2003).pubmed.ncbi.nlm.nih.gov/14631036
  2. [3]T. Van Oudheusden, P. Pasmans, S. Van Der Geer, M. De Loos, M. Van Der Wiel, and O. Luiten, Compression of subrelativistic space-charge-dominated electron bunches for single-shot femtosecond electron diffraction, Physical review letters 105, 264801 (2010).
  3. [4]D. Filippetto, P. Musumeci, R. Li, B. J. Siwick, M. Otto, M. Centurion, and J. Nunes, Ultrafast electron diffraction: Visualizing dynamic states of matter, Reviews of modern physics 94, 045004 (2022).DOI
  4. [5]W. Graves, J. Bessuille, P. Brown, S. Carbajo, V. Dolgashev, K.-H. Hong, E. Ihloff, B. Khaykovich, H. Lin, K. Murari, et al., Compact x-ray source based on burst-mode inverse compton scattering at 100 khz, Physical Review Special Topics-Accelerators and Beams 17, 120701 (2014).DOI
  5. [6]I. J. M. van Elk, C. W. Sweers, D. F. J. Nijhof, R. G. W. van den Berg, T. G. Lucas, X. F. D. Stragier, P. Tack, M. N. Boone, O. J. Luiten, and P. H. A. Mutsaers, First x-rays from a compact and tunable linac-based compton scattering source (2025), arXiv:2504.11897 [physics.acc-ph].arxiv.org/abs/2504.11897
  6. [7]R. Akre, D. Dowell, P. Emma, J. Frisch, S. Gilevich, G. Hays, P. Hering, R. Iverson, C. Limborg-Deprey, H. Loos, et al., Commissioning the linac coherent light source injector, Physical Review Special Topics—Accelerators and Beams 11, 030703 (2008).DOI
  7. [8]S. van der Geer, E. Vredenbregt, O. Luiten, and M. de Loos, An ultracold electron source as an injector for a compact sase-fel, Journal of Physics B: Atomic, Molecular and Optical Physics 47, 234009 (2014).DOI
  8. [13]A. Gallagher and G. York, A photoionization source of monoenergetic electrons, Review of Scientific Instruments 45, 662 (1974).DOI
  9. [14]M. Kurokawa, M. Kitajima, K. Toyoshima, T. Odagiri, H. Kato, H. Kawahara, M. Hoshino, H. Tanaka, and K. Ito, Threshold photoelectron source for the study of low-energy electron scattering: Total cross section for electron scattering from krypton in the energy range from 14 mev to 20 ev, Physical Review A—Atomic, Molecular, and Optical Physics 82, 062707 (2010).
  10. [15]A. McCulloch, R. Speirs, J. Grimmel, B. Sparkes, D. Comparat, and R. Scholten, Field ionization of rydberg atoms for high-brightness electron and ion beams, Physical Review A 95, 063845 (2017).DOI
  11. [16]P. Hommelhoff, Y. Sortais, A. Aghajani-Talesh, and M. A. Kasevich, Field emission tip as a nanometer source of free electron femtosecond pulses, Physical review letters 96, 077401 (2006).pubmed.ncbi.nlm.nih.gov/16606139
  12. [17]C. Ropers, D. Solli, C. Schulz, C. Lienau, and T. Elsaesser, Localized multiphoton emission of femtosecond electron pulses from metal nanotips, Physical review letters 98, 043907 (2007).pubmed.ncbi.nlm.nih.gov/17358773
  13. [18]T. Van Oudheusden, E. De Jong, S. Van der Geer, W. t Root, O. Luiten, and B. Siwick, Electron source concept for single-shot sub-100 fs electron diffraction in the 100 kev range, Journal of Applied Physics 102 (2007).
  14. [19]T. van Oudheusden, Electron source for sub-relativistic single-shot femtosecond diffraction, Ph.D. thesis, Eindhoven University of Technology (2010).
  15. [20]I. Kapchinskii and V. Vladimirskii, Proceedings of the international conference on high energy accelerators, cern, geneva (1959).
  16. [21]O. Luiten, S. Van der Geer, M. De Loos, F. Kiewiet, and M. Van Der Wiel, How to realize uniform three-dimensional ellipsoidal electron bunches, Physical review letters 93, 094802 (2004).pubmed.ncbi.nlm.nih.gov/15447108
  17. [22]B. Zerbe, X. Xiang, C.-Y. Ruan, S. Lund, and P. Duxbury, Dynamical bunching and density peaks in expanding coulomb clouds, Physical Review Accelerators and Beams 21, 064201 (2018).DOI
  18. [23]S. Zwickler, D. Habs, P. Krause, R. Neumann, D. Schwalm, and A. Wolf, Photocathode studies for an ultracold electron beam device, in High Energy Spin Physics: Volume 2: Workshops (Springer, 1991) pp. 20–24.DOI
  19. [24]T. Killian, S. Kulin, S. Bergeson, L. A. Orozco, C. Orzel, and S. Rolston, Creation of an ultracold neutral plasma, Physical Review Letters 83, 4776 (1999).DOI
  20. [25]B. J. Claessens, Dynamics and applications of excited cold atoms, Ph.D. thesis, Eindhoven University of Technology (2006).DOI
  21. [26]E. L. Raab, M. Prentiss, A. Cable, S. Chu, and D. E. Pritchard, Trapping of neutral sodium atoms with radiation pressure, Physical review letters 59, 2631 (1987).pubmed.ncbi.nlm.nih.gov/10035608
  22. [27]T. De Raadt, J. Franssen, and O. Luiten, Subpicosecond ultracold electron source, Physical Review Letters 130, 205001 (2023).DOI
  23. [28]O. Luiten, B. Claessens, S. Van Der Geer, M. Reijnders, G. Taban, and E. Vredenbregt, Ultracold electron sources, International Journal of Modern Physics A 22, 3882 (2007).
  24. [29]W. Ketterle, K. B. Davis, M. A. Joffe, A. Martin, and D. E. Pritchard, High densities of cold atoms in a dark spontaneous-force optical trap, Physical review letters 70, 2253 (1993).pubmed.ncbi.nlm.nih.gov/10053514
  25. [30]J. Maxson, I. Bazarov, W. Wan, H. Padmore, and C. Coleman-Smith, Fundamental photoemission brightness limit from disorder induced heating, New Journal of Physics 15, 103024 (2013).DOI
  26. [31]S. Kuzmin and T. O’Neil, Numerical simulation of ultracold plasmas: How rapid intrinsic heating limits the development of correlation, Physical review letters 88, 065003 (2002).pubmed.ncbi.nlm.nih.gov/11863816
  27. [32]S. Brussaard and D. Vyuga, A 2.5-mv subnanosecond pulser with laser-triggered spark gap for the generation of high-brightness electron bunches, IEEE transactions on plasma science 32, 1993 (2004).DOI
  28. [34]M. Robinson, B. L. Tolra, M. W. Noel, T. Gallagher, and P. Pillet, Spontaneous evolution of rydberg atoms into an ultracold plasma, Physical review letters 85, 4466 (2000).pubmed.ncbi.nlm.nih.gov/11082572
  29. [35]G. Taban, A cold atom electron source, Ph.D. thesis, Eindhoven University of Technology (2009).DOI
  30. [36]G. Taban, M. Reijnders, B. Fleskens, S. Van der Geer, O. Luiten, and E. Vredenbregt, Ultracold electron source for single-shot diffraction studies, Europhysics Letters 91, 46004 (2010).DOI
  31. [37]D. Vyuga, Subnanosecond Pulsed-DC ultra-high gradient photogun for bright relativistic electron bunches, Ph.D. thesis, Eindhoven University of Technology (2006).DOI
  32. [38]J. Hendriks, The physics of photoconductive spark gap switching: Pushing the frontiers, Ph.D. thesis, Eindhoven University of Technology (2006).DOI
  33. [39]B. Knuffman, A. V. Steele, J. Orloff, and J. J. McClelland, Nanoscale focused ion beam from laser-cooled lithium atoms, New Journal of Physics 13, 103035 (2011).DOI
  34. [40]A. McCulloch, D. Sheludko, M. Junker, and R. Scholten, High-coherence picosecond electron bunches from cold atoms, Nature communications 4, 1692 (2013).DOI
  35. [41]W. Engelen, M. Van Der Heijden, D. Bakker, E. Vredenbregt, and O. Luiten, High-coherence electron bunches produced by femtosecond photoionization, Nature communications 4, 1693 (2013).DOI
  36. [42]J. Franssen, T. Frankort, E. Vredenbregt, and O. Luiten, Pulse length of ultracold electron bunches extracted from a laser cooled gas, Structural Dynamics 4 (2017).DOI
  37. [43]W. Engelen, E. Smakman, D. Bakker, O. Luiten, and E. Vredenbregt, Effective temperature of an ultracold electron source based on near-threshold photoionization, Ultramicroscopy 136, 73 (2014).DOI
  38. [44]R. W. Speirs, A. J. McCulloch, B. M. Sparkes, and R. E. Scholten, Identification of competing ionization processes in the generation of ultrafast electron bunches from cold-atom electron sources, Physical Review A 95, 053408 (2017).DOI
  39. [45]W. Engelen, Coherent electron bunches from laser-cooled gases, Ph.D. thesis, Eindhoven University of Technology (2013).DOI
  40. [46]W. Engelen, E. Vredenbregt, and O. Luiten, Analytical model of an isolated single-atom electron source, Ultramicroscopy 147, 61 (2014).DOI
  41. [47]V. Kondratovich and V. Ostrovsky, Resonance and interference phenomena in the photoionisation of a hydrogen atom in a uniform electric field. i. resonances below and above the potential barrier, Journal of physics B: Atomic and molecular physics 17, 1981 (1984).DOI
  42. [48]C. Bordas, Classical motion of a photoelectron interacting with its ionic core: Slow photoelectron imaging, Physical review A 58, 400 (1998).DOI
  43. [49]W. Engelen, D. Bakker, O. Luiten, and E. Vredenbregt, Polarization effects on the effective temperature of an ultracold electron source, New Journal of Physics 15, 123015 (2013).DOI
  44. [50]S. van der Geer and M. de Loos, General particle tracer, www.pulsar.nl/gpt (2011).
  45. [51]J. Franssen, J. Kromwijk, E. Vredenbregt, and O. Luiten, Energy spread of ultracold electron bunches extracted from a laser cooled gas, Journal of Physics B: Atomic, Molecular and Optical Physics 51, 035007 (2018).
  46. [52]J. Franssen, An ultracold and ultrafast electron source, Ph.D. thesis, Eindhoven University of Technology (2024).
  47. [53]A. McCulloch, D. Sheludko, S. Saliba, S. Bell, M. Junker, K. Nugent, and R. Scholten, Arbitrarily shaped high-coherence electron bunches from cold atoms, Nature Physics 7, 785 (2011).DOI
  48. [54]D. Murphy, R. Speirs, D. Sheludko, C. Putkunz, A. McCulloch, B. Sparkes, and R. Scholten, Detailed observation of space-charge dynamics using ultracold ion bunches, Nature communications 5, 4489 (2014).DOI
  49. [55]D. Thompson, D. Murphy, R. Speirs, R. Van Bijnen, A. McCulloch, R. Scholten, and B. Sparkes, Suppression of emittance growth using a shaped cold atom electron and ion source, Physical review letters 117, 193202 (2016).pubmed.ncbi.nlm.nih.gov/27858456
  50. [56]T. C. H. de Raadt, The ultracold electron source: from experiment to instrument, Ph.D. thesis, Eindhoven University of Technology (2024).
  51. [57]B. H. Schaap, T. De Vos, P. Smorenburg, and O. J. Luiten, Photon yield of superradiant inverse compton scattering from microbunched electrons, New Journal of Physics 24, 033040 (2022).DOI
  52. [58]M. Van Mourik, W. Engelen, E. Vredenbregt, and O. Luiten, Ultrafast electron diffraction using an ultracold source, Structural Dynamics 1 (2014).
  53. [59]R. W. Speirs, C. T. Putkunz, A. J. McCulloch, K. A. Nugent, B. M. Sparkes, and R. E. Scholten, Single-shot electron diffraction using a cold atom electron source, Journal of Physics B: Atomic, Molecular and Optical Physics 48, 214002 (2015).DOI
  54. [60]G. Xia, M. Harvey, A. Murray, L. Bellan, W. Bertsche, R. Appleby, O. Mete, and S. Chat topadhyay, An ultracold low emittance electron source, Journal of Instrumentation 9 (06), P06011.
  55. [61]M. Harvey and A. J. Murray, Cold atom trap with zero residual magnetic field: The ac magneto-optical trap, Physical review letters 101, 173201 (2008).pubmed.ncbi.nlm.nih.gov/18999746
  56. [62]C. Nshii, M. Vangelyen, J. P. Cotter, P. F. Griffin, E. Hinds, C. N. Ironside, P. See, A. Sinclair, E. Riis, and A. S. Arnold, A surface-patterned chip as a strong source of ultracold atoms for quantum technologies, Nature nanotechnology 8, 321 (2013).DOI
  57. [63]J. P. McGilligan, P. F. Griffin, E. Riis, and A. S. Arnold, Phase-space properties of magneto-optical traps utilising micro-fabricated gratings., Optics express 23, 8948 (2015).
  58. [64]The force balance is not identical: while for a conventional six-beam MOT the force balance is isotropic, for the grating MOT the axial force balance and radial force balance are typically different, except for a diffraction angle of arccos(1/3).
  59. [65]J. Franssen, T. De Raadt, M. Van Ninhuis, and O. Luiten, Compact ultracold electron source based on a grating magneto-optical trap, Physical Review Accelerators and Beams 22, 023401 (2019).DOI
  60. [66]Parallel and perpendicular were erroneously swapped in [27].
  61. [67]D. Nijhof, T. de Raadt, J. Huijts, J. Franssen, P. Mutsaers, and O. Luiten, Rf acceleration of ultracold electron bunches, Structural Dynamics 10 (2023).arxiv.org/abs/2306.05022
  62. [68]D. F. Nijhof, P. H. Mutsaers, and O. Luiten, Design and optimization of a 100 kev dc/rf ultracold electron source, Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment 1055, 168469 (2023).DOI
  63. [69]R. J. England, R. J. Noble, K. Bane, D. H. Dowell, C.-K. Ng, J. E. Spencer, S. Tantawi, Z. Wu, R. L. Byer, E. Peralta, et al., Dielectric laser accelerators, Reviews of Modern Physics 86, 1337 (2014).DOI
  64. [70]D. Comparat and H. Lignier, Anions get cold, nature physics 19, 1225 (2023).DOI

Paper details

vlnc:W_vv6mcp5m
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents