WIR SCHAFFEN WISSEN – HEUTE FÜR MORGEN Rasmus Ischebeck Accelerator Research EPFL, 2018-12-20 Accelerators at PSI

Rasmus Ischebeck > EPFL > Accelerator Research 2 25 Nobel Prizes in that had direct contribution from accelerators

Year Name Accelerator-Science Contribution to - 1980 James W. Cronin and Cronin and Fitch concluded in 1964 that CP (charge- Winning Research Val L. Fitch parity) symmetry is violated in the decay of neutral K 1939 Ernest O. Lawrence Lawrence invented the cyclotron at the University of mesons based upon their experiments using the Californian at Berkeley in 1929 [12]. Brookhaven Alternating Gradient Synchrotron [28]. 1951 John D. Cockcroft and Cockcroft and Walton invented their eponymous linear 1981 Kai M. Siegbahn Siegbahn invented a weak-focusing principle for Ernest T.S. Walton positive-ion accelerator at the Cavendish Laboratory in betatrons in 1944 with which he made significant Cambridge, England, in 1932 [13]. improvements in high-resolution electron spectroscopy 1952 Bloch used a cyclotron at the Crocker Radiation [29]. Laboratory at the University of California at Berkeley 1983 William A. Fowler Fowler collaborated on and analyzed accelerator-based in his discovery of the magnetic moment of the neutron experiments in 1958 [30], which he used to support his in 1940 [14]. hypothesis on stellar-fusion processes in 1957 [31]. 1957 Tsung-Dao Lee and Chen Ning Lee and Yang analyzed data on K mesons (θ and τ) 1984 and Rubbia led a team of who observed the Yang from Bevatron experiments at the Lawrence Radiation Laboratory in 1955 [15], which supported their idea in intermediate vector W and Z in 1983 using 1956 that parity is not conserved in weak interactions CERN’s proton-antiproton collider [32], and van der [16]. Meer developed much of the instrumentation needed 1959 Emilio G. Segrè and Segrè and Chamberlain discovered the antiproton in for these experiments [33]. 1955 using the Bevatron at the Lawrence Radiation 1986 Ruska built the first electron microscope in 1933 based Laboratory [17]. upon a magnetic optical system that provided large 1960 Donald A. Glaser Glaser tested his first experimental six-inch bubble magnification [34]. chamber in 1955 with high-energy protons produced by 1988 Leon M. Lederman, Lederman, Schwartz, and Steinberger discovered the the Brookhaven Cosmotron [18]. , and muon neutrino in 1962 using Brookhaven’s Alternating 1961 Hofstadter carried out electron-scattering experiments Gradient Synchrotron [35]. on carbon-12 and oxygen-16 in 1959 using the SLAC 1989 Paul’s idea in the early 1950s of building ion traps linac and thereby made discoveries on the structure of grew out of accelerator physics [36]. nucleons [19]. 1990 Jerome I. Friedman, Friedman, Kendall, and Taylor’s experiments in 1974 1963 Goeppert Mayer analyzed experiments using neutron Henry W. Kendall, and on deep inelastic scattering of electrons on protons and beams produced by the Richard E. Taylor bound neutrons used the SLAC linac [37]. cyclotron in 1947 to measure the nuclear binding 1992 Charpak’s development of multiwire proportional energies of krypton and xenon [20], which led to her chambers in 1970 were made possible by accelerator- discoveries on high magic numbers in 1948 [21]. based testing at CERN [38]. 1967 Hans A. Bethe Bethe analyzed nuclear reactions involving accelerated 1995 Martin L. Perl Perl discovered the tau lepton in 1975 using Stanford’s protons and other nuclei whereby he discovered in SPEAR collider [39]. 1939 how energy is produced in stars [22]. 2004 David J. Gross, , Gross, Wilczek, and Politzer discovered asymptotic 1968 Luis W. Alvarez Alvarez discovered a large number of resonance states using his fifteen-inch hydrogen bubble chamber and and freedom in the theory of strong interactions in 1973 high-energy proton beams from the Bevatron at the H. David Politzer based upon results from the SLAC linac on electron- Lawrence Radiation Laboratory [23]. proton scattering [40]. 1976 and Richter discovered the J/Ψ particle in 1974 using the 2008 Makoto Kobayashi and Kobayashi and Maskawa’s theory of mixing in Samuel C.C. Ting SPEAR collider at Stanford [24], and Ting discovered 1973 was confirmed by results from the KEKB accelerator at KEK (High Energy Accelerator Research the J/Ψ particle independently in 1974 using the and Yoichro Nambu Brookhaven Alternating Gradient Synchrotron [25]. Organization) in Tsukuba, Ibaraki Prefecture, , 1979 Sheldon L. Glashow, Glashow, Salam, and Weinberg cited experiments on and the PEP II (Positron Electron Project II) at SLAC , and the bombardment of nuclei with neutrinos at CERN in [41], which showed that quark mixing in the six-quark 1973 [26] as confirmation of their prediction of weak model is the dominant source of broken symmetry [42]. neutral currents [27]. 2013: François Englert and Peter W. Higgs "for the theoretical discovery of a mechanism that contributes to our understanding of the origin of of subatomic particles, and which recently was confirmed through the discovery of the predicted fundamental particle, by the ATLAS and CMS experiments at CERN's Large Hadron Collider" Nobel Prizes and Accelerators, L. Rivkin, PSI & EPFL Rasmus Ischebeck More Accelerators…

Rasmus Ischebeck > EPFL > Accelerator Research 4 How to Accelerate Charged Particles

electron

+ 1 MV

Rasmus Ischebeck > EPFL > Accelerator Research 5 What is the Total Energy of an Electron that was Initially at Rest, after Acceleration by a DC Potential Difference of –1 MV?

Easy: E(null) =1MeV

E = 1 MeV + 511 keV/c2 c2 =1.511 MeV

Easy: (null) ·

Easy: 2 2 E(null) = 1 +0.511 MeV = 1.123 MeV p

This depends on the trajectory of the electron

Rasmus Ischebeck > EPFL > Accelerator Research 6 Enhancement

Rasmus Ischebeck > EPFL > Accelerator Research 7 How to Accelerate Charged Particles How to Accelerate Charged Particles

Assume: • an ultrarelativistic particle of charge e k • moving along the z axis • accelerated by a plane electromagnetic wave that propagates at an angle ϑ to the z axis ϑ

e-

λ

Rasmus Ischebeck – Accelerators Beyond LHC and ILC. Uni Wuppertal, 2009-04-30

Rasmus Ischebeck > EPFL > Accelerator Research 8 How to Accelerate Charged Particles How to Accelerate Charged Particles

Then: • Position of the electron k

ϑ

- • Electric field e

λ

• Energy gradient

Rasmus Ischebeck – Accelerators Beyond LHC and ILC. Uni Wuppertal, 2009-04-30

Rasmus Ischebeck > EPFL > Accelerator Research 9 Lawson Woodward Theorem Lawson Woodward Theorem

• Every wave in far field can be written as a superposition of plane waves • The Lawson-Woodward Theorem states: • the total acceleration • of ultrarelativistic particles • by far-field electromagnetic waves • is zero ⇒ Need near-field structures

electron

Woodward, J. IEE 93 (1947) Lawson, IEEE Trans. Nucl. Sci. 26 (1979) electromagnetic Palmer, Part. Accel. 11 (1980) wave

Rasmus Ischebeck – Accelerators Beyond LHC and ILC. Uni Wuppertal, 2009-04-30

Rasmus Ischebeck > EPFL > Accelerator Research 10 What do I mean by “Near-Field Structure”?

A structure consisting of a perfect conductor

A structure that encloses the electron from all sides

A structure that is less than a few wavelengths away

A structure with a nearly-flat surface

Rasmus Ischebeck > EPFL > Accelerator Research 11 Rasmus Ischebeck > EPFL > Accelerator ResearchRasmus Ischebeck Accelerating Structures

Metallic Superconducting

Dielectric Plasma

13 Laser-Driven Accelerators

Rasmus Ischebeck > EPFL > Accelerator ResearchRasmus Ischebeck Shortly after lasers were invented it was suggested to use them to accelerate particles.

Koichi Shimoda, Applied Optics 1 (1), 33 (1961) Laser-Based Accelerators

Shimoda Appl. Opt. 1 (1), 33 (1961)

Rasmus Ischebeck > EPFL > Accelerator Research 15 Rasmus Ischebeck > EPFL > Accelerator ResearchRasmus Ischebeck Rasmus Ischebeck > EPFL > Accelerator ResearchRasmus Ischebeck Rasmus Ischebeck > EPFL > Accelerator Research Requirements on an Accelerating Structure

• Longitudinal electric field

• Phase velocity = particle velocity

• Uniform fields across the particle bunch • Low energy spread • Good emittance

• Sustain large fields • Fields on the surface should be comparable to the fields seen by the beam

Rasmus Ischebeck > EPFL > Accelerator Research 19 Rectangular Waveguide

• Starting with Maxwell’s Equations for vacuum: E~ =0 r · H~ =0 r · @H~ E~ = µ r⇥ 0 @t @E~ H~ = " r⇥ 0 @t • Ansatz for the electromagnetic fields: b i(!t kz) E~ (x, y, z, t)=E˜(x, y) e y z · a i(!t kz) H~ (x, y, z, t)=H˜ (x, y) e x · • Decomposing into longitudinal and transverse components

E˜(x, y)=ET (x, y)+~ezEz(x, y)

H˜ (x, y)=HT (x, y)+~ezHz(x, y)

Rasmus Ischebeck > EPFL > Accelerator Research 20 Rectangular Waveguide

• Solving the wave equation • Take into account the boundary conditions

• Transverse electric modes ⇡m ⇡n H (x, y)=H sin x sin y z 0 a b ⇣ ⌘ ⇣ ⌘ Ez(x, y)=0 m, n N 2 • Transverse magnetic modes E-field phase animation inside a rectangular waveguide. ⇡m ⇡n Source: cst.com E (x, y)=E sin x sin y z 0 a b ⇣ ⌘ ⇣ ⌘ Hz(x, y)=0

Rasmus Ischebeck > EPFL > Accelerator Research 21 Rectangular Waveguide

• A few important points: • Transverse fields can be calculated from this

• There is a cutoff frequency • Waves with lower frequency cannot propagate in the waveguide

1 m 2 n 2 fc = + 2p"µ a b r⇣ ⌘ ⇣ ⌘

• Phase velocity of the mode ! c v = = >c p 2 k 1 fc f 2 q More details: see Jackson, Electrodynamics

Rasmus Ischebeck > EPFL > Accelerator Research 22 ETH Zurich & PSI Villigen Chapter 3. Electromagnetic field in circular waveguide

Expressed in polar coordinates equation (3.10) is a Bessel di↵erential equation whose solutions are the Bessel functions of the first kind Jm(r).Circular Applying the Waveguides boundary condition gives = umn/b,

im mn(r, )=Jm(umnr/b)e ,m=0, 1, 2,..., n=1, 2,... (3.11)

Here, u denotes the nth root of the mth Bessel function. mn • Similar results for circular waveguides • Field distribution is given by the Bessel functions The electric transverseETH Zurich component & PSI as Villigen well as the magneticChapter field are now 3. Electromagnetic given by field in circular waveguide ik Et = (3.12) 2 r? µr✏r! Expressed in polarB coordinatest = ez equationE (3.10) is a Bessel di↵erential(3.13) equation whose solutions are ck ⇥ ? the Bessel functions of the first kind Jm(r). Applying the boundary condition gives = umn/b, so the field is fully described. im mn(r, )=Jm(umnr/b)e ,m=0, 1, 2,..., n=1, 2,... (3.11) For synchronicity with a particle beam the dispersion relation !(k) is relevant. The general expression for a z-symmetricHere, umn geometrydenotes the is givennth root implicitly of the inmth equationBessel (3.9). function. In case of a hollow cylinder it is • Again, the phase velocity is always larger than the particle velocity c The electric transverse component2 as2 well as the magnetic field are now given by !(kv)=>c k +(umn/b) (3.14) p p✏r p Considering a hollow waveguide with dielectric loading, a = 0, this dispersionik is clearly di↵erent. Et = (3.12) But in the limit of vanishing inner radius it has to satisfy6 this dispersion.2 Afterr? describing the field in such a loaded waveguide, the change in dispersion whenµar✏decreasesr! is investigated in Bt = ez E (3.13) sec. 6.1. ck ⇥ ? so the field is fully described. 3.1 Waveguide with dielectric loading ForRasmus Ischebeck > EPFL > Accelerator Research synchronicity with a particle beam the dispersion relation !(k) is relevant. The general 23 The relative permittivityexpression is di for↵erent a z-symmetric along the radial geometry direction is given implicitly in equation (3.9). In case of a hollow cylinder it is 1, 0 r a c ✏r(r)=  !(k)= k2 +(u /b)2 (3.15) (3.14) ✏ ,a r b mn ( r   p✏r p Considering a hollow waveguide with dielectric loading, a = 0, this dispersion is clearly di↵erent. which splits equation (3.10) into a part for 0 r a with 1 and a part for a 6 r b with 2, where But in the limit of vanishing  inner radius it has to satisfy this dispersion. After describing the field in such a loaded waveguide, the change in dispersion when a decreases is investigated in !2 !2 sec. 6.1. 2 = k2, 2 = ✏ k2 (3.16) 1 c2 2 r c2 It is clear that for a given frequency ! there is exactly one corresponding longitudinal wave vector k(!) resp.3.1 for a givenWaveguide wave vector k exactly with one frequency dielectric! corresponds loading to it, since the wave propagates through the structure everywhere with the same properties. This allows us to express 1 with 2Theand relative vice versa. permittivity Either depending is di↵erent on the along frequency the radial or the direction wave vector. !2 2 1, 0 r a 1(!)= 2 (1 ✏r)+2 (3.17) c ✏r(r)=   (3.15) ✏r,a r b 1 1 2 ( 1(k)= 2 + 1 k   (3.18) ✏r ✏r which splits equation (3.10) into✓ a part◆ for 0 r a with 1 and a part for a r b with 2, This will be usedwhere when solving the condition for the dispersion relation.    !2 !2 2 = k2, 2 = ✏ k2 (3.16) 1 c2 2 r c2 4 It is clear that for a given frequency ! there is exactly one corresponding longitudinal wave vector k(!) resp. for a given wave vector k exactly one frequency ! corresponds to it, since the wave propagates through the structure everywhere with the same properties. This allows us to express 1 with 2 and vice versa. Either depending on the frequency or the wave vector. !2 (!)= (1 ✏ )+2 (3.17) 1 c2 r 2 1 1 (k)= + 1 k2 (3.18) 1 ✏ 2 ✏ r ✓ r ◆ This will be used when solving the condition for the dispersion relation.

4 How Could We Slow Down the Phase Velocity?

There is nothing we can do about it

Give the waveguide a diameter that varies along the direction of motion

Partially fill the waveguide with a dielectric

Both of the above ( and )

Rasmus Ischebeck > EPFL > Accelerator Research 24 ETH Zurich & PSI Villigen Chapter 3. Electromagnetic field in circular waveguide

Expressed in polar coordinates equation (3.10) is a Bessel di↵erential equation whose solutions are the Bessel functions of the first kind Jm(r). Applying the boundary condition gives = umn/b,

im mn(r, )=Jm(umnr/b)e ,m=0, 1, 2,..., n=1, 2,... (3.11)

Here, umn denotes the nth root of the mth Bessel function.

The electric transverse component as well as the magnetic field are now given by ik Et = (3.12) 2 r? µr✏r! Bt = ez E (3.13) ck ⇥ ? so the field is fully described.

For synchronicity with a particle beam the dispersion relation !(k) is relevant. The general expression for a z-symmetric geometry is given implicitly in equation (3.9). In case of a hollow cylinder it is c 2 2 !(k)= k +(umn/b) (3.14) p✏r p Considering a hollow waveguide with dielectric loading, a = 0, this dispersion is clearly di↵erent. But in the limit of vanishing inner radius it has to satisfy6 this dispersion. After describing the field in such a loaded waveguide, the change in dispersion when a decreases is investigated in sec. 6.1. Dielectric Lined Waveguide 3.1 Waveguide with dielectric loading ETH Zurich & PSI Villigen Chapter 3. Electromagnetic field in circular waveguide The relative permittivity is di↵erent• Assuming a geometry with: along the radial direction

1, 0 r a To fully describe the boundary✏r(r)= value problem  the transition at the interface(3.15) is required. The (✏r,a r b longitudinal electric field component E is continuous  as well as the transverse magnetic field B z electron whichat r = splitsa. But equation in transverse (3.10) into a direction part for 0 ther electrica with field1 and is a not part continuous for a r b sincewith the2, displacement     beam wherefield D = ✏E is instead. • Continuity of the electromagnetic fields at the boundaries: !2 !2 E 2 = k2E, z(r2 == b✏) = 0 k2 (3.16) (3.19) 1 c2 2 r c2 Ez(r+ = a)=Ez(r = a) (3.20) It is clear that for a given frequency ! there is exactly one corresponding longitudinal wave vector k(!) resp. for a given wave vector kBexactly(r+ = onea)= frequencyB(r !=correspondsa) to it, since the (3.21) wave propagates through the structure everywhere with the same properties. This allows us to ✏0✏rEr(r+ = a)=✏0Er(r = a) (3.22) metal express 1 with 2 and vice versa. Either depending on the frequency or the wave vector. In the region of the dielectric, r>a, the solution of eq. (3.10) has to be extended with the Bessel dielectric 2 d a a functions of second kind, Y ( r). The! coecients2 of the superposition are already chosen such m1(2!)= 2 (1 ✏r)+2 (3.17) b that the condition of vanishing longitudinalc electric field is already satisfied. 1 1 2 Due to time constraints of this1( thesisk)= we2 + will limit1 usk to the first mode from now(3.18) on. It will have ✏r ✏r to be discussed elsewhere how to couple an✓ external◆ laser beam to this mode. ThisThe will solution be used for when the solving field is the condition for the dispersion relation.

Rasmus Ischebeck > EPFL > Accelerator Researchi(kz !t) Max Kellermeier 25 E1J0(1r)e ,r

Ez = 8 4 (3.23) > J0(2b) i(kz !t) >E J ( r) Y ( r) e ,a :> k i(kz !t) i E J 0 ( r)e ,r (3.24) > > k J0(2b) i(kz !t) > : ! i(kz !t) i 2 E1J00 (1r)e ,r > ! J0(2b) i(kz !t) where the parameters:> 1, 2, as well as the amplitudes E1 and E2, are determined by eq. (3.16) and the boundary conditions. By applying the boundary condition (3.20) of a continuous longitudinal electric field the ampli- tudes are related by J ( a)Y ( b) E = E 0 1 0 2 (3.26) 2 1 J ( a)Y ( b) J ( b)Y ( a) 0 2 0 2 0 2 0 2 Including this result in the boundary condition (3.22) requires 1 J ( a) ✏ F ( a) 00 1 r 00 2 = 0 (3.27) 1 J0(1a) 2 F0(2a) where

J0(2b) F0(2a)=J0(2a) Y0(2a) (3.28) Y0(2b)

J0(2b) F00(2a)=J00 (2a) Y00(2a) (3.29) Y0(2b)

5 ETH Zurich & PSI Villigen Chapter 3. Electromagnetic field in circular waveguide ETH Zurich & PSI Villigen Chapter 3. Electromagnetic field in circular waveguide

Only two unknown parameters remain. Together with the relations in eq. (3.16) the transcen- dental equation (3.27) can be solved numerically for a given wave vector k or a given frequency To fully describe the boundary value problem the transition at the interface is required. The 2 !. One has to pay attention to the square root in the definition of 1.While2 will always be longitudinal electric field component Ez is continuous2 as well as the transverse magnetic field B positive, 1 can become negative and therefore 1 is imaginary. From eq. 3.16 we know that 2 at r = a. But in transverse direction the electricdecreases field while is increasing not continuousk.Sincethecoe since thecient displacement of the second term in 1(2,k) is negative, eq. field D = ✏E is instead. 3.17, it will cancel the first one when k reaches a certain point. The transition from 1 being real to imaginary occurs at the of the dispersion with the Ez(r = b) = 0 (3.19) speed of dispersion, meaning ! = ck. At this point 1 vanishes. For increasing wavevector 2 2 2 Ez(r+ != /ca)=kEzis(r negative.= a) (3.20) In this region ra, the solutionstarts to of drop eq. (3.10) until it has vanishes to be at extendedr = b. with the Bessel Dielectric LinedFigure Waveguide 3.1(a) the resulting dispersion is shown, together with the speed of light dispersion and the functions of second kind, Ym(2r). The coeonecients of a fully of the loaded superposition waveguide, a are= 0. already Additionally chosen the such point of operation to match synchronic- that the condition of vanishing longitudinality electric is marked field as is the already crossing satisfied. with !(k)=ck. For this example the radial field dependence is Due to time constraints of this thesis we willplotted limit in us figure to the 3.1(b). first mode from now on. It will have to be discussed• Fields can be obtained analytically elsewhere how to couple an external laser beam to this mode. The solution for the field is i(kz !t) E1J0(1r)e ,r

Ez = 8 (3.23) > J0(2b) i(kz !t) >E J ( r) Y ( r) e ,a :> k i(kz !t) i E J 0 ( r)e ,r (3.24) > > k J0(2b) i(kz !t) > : ! i(kz !t) i 2 E1J00 (1r)e ,r > ! J0(2b) i(kz !t) corresponding maximum field distribution with E0 = 100 MeV. Since the transverse components where the parameters:> 1, 2, as well as the amplitudes E1 and E2, are determined by eq. (3.16) vanish when Ez is maximal and vice versa, Er and B are shown at a shifted period of ⇡/2. andRasmus Ischebeck > EPFL > Accelerator Research the boundary conditions. Max Kellermeier 26 By applying the boundary condition (3.20) of a continuous longitudinal electric field the ampli- tudes are related by J ( a)Y ( b) E = E 0 1 0 2 (3.26) 2 1 J ( a)Y ( b) J ( b)Y ( a) 0 2 0 2 0 2 0 2 Including this result in the boundary condition (3.22) requires 1 J ( a) ✏ F ( a) 00 1 r 00 2 = 0 (3.27) 1 J0(1a) 2 F0(2a) where 6 J0(2b) F0(2a)=J0(2a) Y0(2a) (3.28) Y0(2b)

J0(2b) F00(2a)=J00 (2a) Y00(2a) (3.29) Y0(2b)

5 T uih&PIVilligen PSI & Zurich ETH hpe .Dvlpeto iuainprogram simulation a of Development 5. Chapter

Rasmus Ischebeck Max Kellermeier

Figure 5.2: Screenshot showing the complete user interface with initial settings including the input section, the output section and the figures. 12 PHASE STABLE NET ACCELERATION OF ELECTRONS ... Phys. Rev. ST Accel. Beams 11, 101301 (2008)

TABLE I. Experimental parameters for the net acceleration large interferometer, with the electron beam serving to experiment. All width values are fwhm. communicate the phase between the IFEL and the second Parameter Value stage accelerator. To find the phase between the IFEL and ITR paths, a second laser path is installed that closely Ebeam energy 60 MeV follows the 800 nm drive laser. Prior to entering the main Energy spread 45 keV (typical) chamber containing the accelerators, the monitor laser is Ebeam pulse length 0.5 ps (typical) brought out and recombined with the other arm of the Ebeam spot size 100 "m (nominal) Laser pulse length 0.75 ps (nominal) interferometer. The interference pattern is observed with Laser energy 1 mJ=pls a CCD and used to interpret the phase of the drive laser in IFEL/ITR laser split ratio 35=65 the interferometer. IFEL laser spot size 500 "m Since the phase of the actual drive laser is not directly ITR laser spot size 200 "m measured during the experiments, it is important to verify Undulator period 1.8 cm that the phase monitor system tracks the actual phase of the Number of periods 3 drive laser. To do this, prior to beam experiments, another Undulator strength (aw) 0.46 CCD was temporarily placed inside the experiment cham- Chicane R56 0.06–0.22 mm ber to observe interference fringes from the IFEL and ITR ITR crossing angle 8 mrad lasers. Comparisons were made between the phase of the 800 nm light to that of the phase monitor. Each of the two interferometers were found to have a therefore necessary to restrict the transverse size of the fast uncorrelated jitter of around 13. When the phase was electron beam to less than half a fringe. This is done using a actively scanned using the piezo mirror [Fig. 3(a)], the tungsten slit. Electrons that do not pass through the slit are phase between the two monitors tracked well. Over very scattered on the tungsten and do not reach the energy long time scales, however, the phase between the two spectrometer. interferometers would wander [Fig. 3(b)]. In this case the The net acceleration experiment builds on the hardware slow drift was more than 2% during the course of an hour. for the microbunching experiment [6]. Figure 2 shows the Fortunately, this is much longer than the typical data set layout for the experiment. The laser light is split to drive which are acquired in a few minutes. In addition, it was the two stages with each path having separate focusing found during postanalysis that the data itself can be used to optics and remote controllable steering mirrors. In the arm account for the slow drift. driving the second stage there is also a delay arm to obtain Prior to net acceleration, the IFEL and ITR interactions temporal coincidence in the two laser-electron interactions are first observed individually using the cross-correlation and a piezo mounted mirror for scanning the timing on the technique described in [6]. The coarse delay between the optical scale. Inside the vacuum chamber are the undulator IFEL and ITR optical paths is adjusted so that the two for the IFEL interaction, the chicane, and the second stage interactions occur with the same laser timing. Both the ITR accelerator. IFEL and ITR interactions are then used together with a Arguably the key technical challenge in this experiment fixed laser to electron delay to obtain net acceleration. The is the measure and control of the optical phase between the optical phase offset between the two stages is scanned laser-electron interaction stages. The layout is essentially a First Laser Acceleration Experimentsusing the at SLAC piezo driven mirror producing phase correlated

halfwave piezo optical IFEL laser path plate delay mirror ITR laser path coarse delay steering mirror

vertical slit

polarizing Electron beamsplitter Spectrometer

e- Phase Monitor camera IFEL Chicane ITR

Sears et al., FIG. 2. (Color) LayoutPRST-AB of 11 the, 101301 net (2008) acceleration experiment. The laser is split at a polarizing beamsplitting cube to drive both the ITR acceleration and theRasmus Ischebeck > EPFL > Accelerator Research IFEL. The ITR path includes a piezo driven mirror for varying the optical phase. Each armChris Sears is 28 5min path length. Interferometric noise measurements indicate the optical paths are stable to within 20 nm or 2 parts per billion of the total path. 

101301-3 First Laser Acceleration Experiments at SLAC

• Acceleration by inverse transition radiation on a surface

Sears et al., PRST-AB 11, 101301 (2008)

Rasmus Ischebeck > EPFL > Accelerator Research 29 First Laser Acceleration Experiments at SLAC

Rasmus Ischebeck > EPFL > Accelerator Research 30 First Laser Acceleration Experiments at SLAC

CHRISTOPHER M. S. SEARS et al. Phys. Rev. ST Accel. Beams 11, 101301 (2008) • Net energy gain observed 1.5 105 0.15 y = 1.2 cos (θ + 5.2) y = 6.9 cos (θ + 0.1) + y = 0.12 cos (θ + 1.3) 89.5 1 100 0.1 0.5 0.05 95 0 0 90 −0.5 Asymmetry −0.05

Centroid Shift (keV) −1 85 −0.1

−1.5 Energy Spread (fwhm; keV) 80 246 246 246 Corrected Phase Corrected Phase Corrected Phase

FIG. 6. (Color) Correlation of electron spectrum centroid energy (left), energy spread (center), and asymmetry (right) with the corrected phase subtracting slow drift. The error bars are the deviations of the means for the binned data. Also shown are sinusoidal fits Sears et al., PRST-ABto each 11 set., 101301 (2008)

Rasmus Ischebeck > EPFL > Accelerator Research Future experiments with a stronger second stage inter- [3] C. M. S. Sears, E. R. Colby, B. M. Cowan, R. H. Siemann,Chris Sears31 action will be needed to further explore issues such as J. E. Spencer, R. L. Byer, and T. Plettner, Phys. Rev. Lett. charge capture. This might be provided by using photonic 95 , 194801 (2005). band gap structures [12] to build optical scale dielectric [4] P. Musumeci et al., Phys. Rev. Lett. 94 , 154801 (2005). waveguides for guiding the laser [13]. Such a device could [5] W. D. Kimura et al., Phys. Rev. Lett. 86 , 4041 (2001). [6] C. M. S. Sears et al., Phys. Rev. ST Accel. Beams 11, sustain large accelerating gradients approaching 1 GeV=m 061301 (2008). while at the same time requiring less pulse energy due to [7] S. Baccaro, F. DeMartini, and A. Ghigo, Opt. Lett. 7, 174 the small transverse leading to good coupling (1982). efficiency [14]. The first experiments with such devices are [8] A. Luccio, G. Matone, L. Miceli, and G. Giordano, Laser currently being planned. Part. Beams 8, 383 (1990). [9] C. McGuinness, R. Byer, E. Colby, R. Ischebeck, R. ACKNOWLEDGMENTS Noble, T. Plettner, C. M. S. Sears, R. Siemann, J. Spencer, and D. Walz, in Proceedings of the 2007 The authors wish to acknowledge the efforts and con- Particle Accelerator Conference, Albuquerque, New tributions of the NLCTA Operations group: Justin May, Mexico, 2007 (IEEE, Albuquerque, New Mexico, 2007), Doug McCormick, Tonee Smith, Richard Swent, and Keith p. 4195. Jobe. We would also like to thank Walt Zacherl and Bruce [10] T. Plettner, R. L. Byer, E. Colby, B. Cowan, C. M. S. Sears, Rohrbough for their work on laser diagnostics. This work J. E. Spencer, and R. H. Siemann, Phys. Rev. Lett. 95 , is supported by Department of Energy Contracts No. DE- 134801 (2005). AC02-76SF00515 and No. DE-FG02-03ER41276. [11] T. Plettner, R. L. Byer, E. Colby, B. Cowan, C. M. S. Sears, J. E. Spencer, and R. H. Siemann, Phys. Rev. ST Accel. Beams 8, 121301 (2005). [12] J. D. Joannopoulos, R. D. Meade, and J. N. Winn, Photonic [1] M. J. Feldman and R. Y. Chiao, Phys. Rev. A 4 , 352 Crystals: Molding the Flow of Light ( (1971). Press, Princeton, New Jersey, 1995), p. 08540. [2] W. D. Kimura, G. H. Kim, R. D. Romea, L. C. Steinhauer, [13] X. E. Lin, Phys. Rev. ST Accel. Beams 4 , 051301 (2001). I. V. Pogorelsky, K. P. Kusche, R. C. Fernow, X. Wang, and [14] Y.C. N. Na, R. H. Siemann, and R. L. Byer, Phys. Rev. ST Y. Liu, Phys. Rev. Lett. 74 , 546 (1995). Accel. Beams 8, 031301 (2005).

101301-6 Double Grating Struture

plane wave

Rasmus Ischebeck > EPFL > Accelerator Research Yelong Wei 32 RESEARCH LETTER

Acceleration Experiments with a Dielectric Structure ab Laser pulse (λ = 800 nm) Spectrometer magnet Magnetic Cylindrical lens lenses Electron DLA device RESEARCH LETTER beam

ab B Bunch length:Laser pulse (λ = 800 nm) Spectrometer magnet Magnetic10 ps Cylindrical lens lenses Scattered Energy Electron DLA device beam electrons Electrons B Transmitted Scattered Energy electrons electrons Electrons Transmitted electrons

2 μm Lanex screen Intensifed CCD 2 μm camera Figure 1 | DLA structure and experimental set-up. a, Scanning electron field polarization direction and the effective periodic phase reset, depicted as Lanex screen microscope image of thePeralta longitudinal et al., cross-section of a DLA structure alternating red (acceleration) and black (deceleration) arrows. A snapshot of Intensifed CCD fabricated as depictedNature in Extended 503 Data, Fig.91 1a.(2013) Scale bar, 2 mm. the simulated fields in the structure shows the corresponding spatial b, Experimental set-up. Inset, a diagram of the DLA structure indicating the modulation in the vacuum channel. See text for details. camera than an 800-nmRasmus Ischebeck > EPFL > Accelerator Research gap structure but requires tighter tolerances on the across the electron beam, forming a cross-correlation signal. A sample 33 Figure 1 | DLA structure and experimentalelectron beam. set-up. a, Scanningmeasurement electron at a laser pulse energyfield of 91.8 polarization6 1.3 mJ over a laser delay direction and the effective periodic phase reset, depicted as The NIR pulses, 1.24 6 0.12 ps long, from a regeneratively amplified of 6 ps is shown Fig. 3a. The orange circles (laser-off data) show no Ti:sapphire mode-locked laser are focused to an r.m.s. spot size of variation correlated with laser delay, as expected, and have an r.m.s. microscope image of the longitudinal30 mm 3 300 m cross-sectionm at the interaction point. We of use a a DLA motorized structurefour- deviation of 4.5 keV, which is takenalternating as the noise floor level red of the mea- (acceleration) and black (deceleration) arrows. A snapshot of axis stage for precise alignment of the structure with the electron beam. surement. The blue circles (laser-on data) show the expected sech2 dis- fabricated as depicted in ExtendedOnce aligned, Data the electron Fig. beam 1a. leaving Scale the structure bar, goes 2 throughmm. a tribution with a full-width at half-maximumthe simulated (FWHM) of 1.89 6 0.09 fields ps, in the structure shows the corresponding spatial point-to-point focusing spectrometer magnet, which disperses the b, Experimental set-up. Inset, aoutgoing diagram electron beam of in the energy DLA onto a Kodak structure Lanex phosphor indicating the modulation in the vacuum channel. See text for details. screen that is imaged by an intensified CCD (charge-coupled device) Charge density (arbitrary units) camera. According to particle tracking simulations, 2.2% of the 60- 0 0.2 0.4 0.6 0.8 1 MeV beam is transmitted through the vacuum channel of the 400-nm a gap structure (see Methods). A segment of the spectrometer screen 15 Laser off than an 800-nm gap structurefocusing but on this transmitted requires distribution tighter is shown in tolerancesFig. 2a. The on the across the electron beam, forming a cross-correlation signal. A sample horizontal axis represents beam energy, and the entire image spans 12 electron beam. 240 keV. The central pixel location of the 60-MeV beam is taken as the 6 m reference point, corresponding to zero energy deviation (DE). measurementEnergy gain at a laser pulse energy of 91.8 1.3 J over a laser delay 9 The spectrometer image in Fig. 2a is a median filtered average of a b The NIR pulses, 1.24 6 0.12dozen ps shots. long, The least-squares from fit to the a distribution regeneratively of electrons scat- amplified15 of 6 ps is shownLaser on Fig. 3a. The orange circles (laser-off data) show no tered by the fused silica substrate and the grating teeth has been removed Ti:sapphire mode-locked laserfrom this image are to emphasize focused the transmitted to distribution an r.m.s. (see Methods). spot12 size of variation correlated with laser delay, as expected, and have an r.m.s. A similarly averaged set of laser-on spectrometer images within 0.5 ps

of the optimal timing overlap for laser pulses with energy 93 mJper Position (mm) Position (mm) 30 mm 3 300 mm at the interactionpulse is shown in Fig. point. 2b. The white We contour in use both Fig. a 2a and motorized Fig. 2b 9 four- deviation of 4.5 keV, which is taken as the noise floor level of the mea- denotes the location where the spectral charge density is 4.5% of the c 2 axis stage for precise alignmentmaximum of density the (at structure the peak of the scattered with distribution). the In electron the 0.2 beam. surement.Laser The off blue circles (laser-on data) show the expected sech dis- Spectrum ft presence of a laser field, there exists a higher charge density on either Laser on Once aligned, the electron beamside of the original leaving peak at DE 5 the0. The white structure contour shows a sizablegoes through0.15 a tributionModel with a full-width at half-maximum (FWHM) of 1.89 6 0.09 ps, fraction of electrons with maximum energy that is ,60 keV higher Simulation than in the laser-off case. 0.1 point-to-point focusing spectrometerThe laser-induced energy modulation magnet, is readily apparent which in the energy disperses the spectra (Fig. 2c). Using the fits to these spectra, a maximum energy outgoing electron beam inshift energy of 53.1 keV is calculated onto from the a abscissa Kodak of the half-width Lanex at half- phosphor0.05 Accelerated maximum (HWHM) point in the high-energy tail. We use an analy- electrons Charge density (arbitrary units) screen that is imaged by antical intensified interaction model (see CCD Methods and (charge-coupled Extended Data Fig. 3) to 0 device) Charge density (arbitrary units) calculate an accelerating gradient from this measurement. Figure 2c –100 –50 0 50 100 shows the input electron beam distribution used in the model (blue Energy deviation, ΔE (keV) 0 0.2 0.4 0.6 0.8 1 camera. According to particlecurve), which tracking is a fit to the measured simulations, spectrum in the absence 2.2% of a laser of the 60- field (light blue crosses). The calculated energy modulation (red curve) Figure 2 | Demonstration of energy modulation. a, Image of the transmitted agrees with our measurement (pink crosses), and gives a correspond- electron beam on the spectrometer screen,a with the laser off. b,Asa but when MeV beam is transmitted through the vacuum channel of thethe laser 400-nm field is present. c, Energy spectra from a and b showing energy ing accelerating gradient of 151.2 MeV m21 for this example. Particle modulation. A fit (blue curve) to the measured laser-off15 spectrum (light blue Laser off gap structure (see Methods).tracking A simulations segment (black dots; see of Methods) the at this spectrometer gradient level give crosses) is used screen as input for the simulations. The calculated energy modulation an independent confirmation of the observed modulated spectrum. (red curve) and particle tracking simulations (black dots) agree with our To determine the maximum gradient at a given laser power level, we measured spectrum (pink crosses). Images of the entire spectrometer screen are focusing on this transmittedmeasure distribution the energy modulation as the is laser pulse shown is temporally in scanned Fig.shown in 2a. Extended The Data Fig. 2. horizontal axis represents beam92 | NATURE energy, | VOL 503 | 7 NOVEMBER and the 2013 entire image spans 12 ©2013 Macmillan Publishers Limited. All rights reserved 240 keV. The central pixel location of the 60-MeV beam is taken as the reference point, corresponding to zero energy deviation (DE). Energy gain 9 The spectrometer image in Fig. 2a is a median filtered average of a b dozen shots. The least-squares fit to the distribution of electrons scat- 15 Laser on tered by the fused silica substrate and the grating teeth has been removed from this image to emphasize the transmitted distribution (see Methods). 12 A similarly averaged set of laser-on spectrometer images within 0.5 ps

of the optimal timing overlap for laser pulses with energy 93 mJper Position (mm) Position (mm) pulse is shown in Fig. 2b. The white contour in both Fig. 2a and Fig. 2b 9 denotes the location where the spectral charge density is 4.5% of the c maximum density (at the peak of the scattered distribution). In the 0.2 Laser off Spectrum ft presence of a laser field, there exists a higher charge density on either Laser on side of the original peak at DE 5 0. The white contour shows a sizable 0.15 Model fraction of electrons with maximum energy that is ,60 keV higher Simulation than in the laser-off case. 0.1 The laser-induced energy modulation is readily apparent in the energy spectra (Fig. 2c). Using the fits to these spectra, a maximum energy shift of 53.1 keV is calculated from the abscissa of the half-width at half- 0.05 Accelerated maximum (HWHM) point in the high-energy tail. We use an analy- electrons Charge density (arbitrary units) tical interaction model (see Methods and Extended Data Fig. 3) to 0 calculate an accelerating gradient from this measurement. Figure 2c –100 –50 0 50 100 shows the input electron beam distribution used in the model (blue Energy deviation, ΔE (keV) curve), which is a fit to the measured spectrum in the absence of a laser field (light blue crosses). The calculated energy modulation (red curve) Figure 2 | Demonstration of energy modulation. a, Image of the transmitted agrees with our measurement (pink crosses), and gives a correspond- electron beam on the spectrometer screen, with the laser off. b,Asa but when 21 the laser field is present. c, Energy spectra from a and b showing energy ing accelerating gradient of 151.2 MeV m for this example. Particle modulation. A fit (blue curve) to the measured laser-off spectrum (light blue tracking simulations (black dots; see Methods) at this gradient level give crosses) is used as input for the simulations. The calculated energy modulation an independent confirmation of the observed modulated spectrum. (red curve) and particle tracking simulations (black dots) agree with our To determine the maximum gradient at a given laser power level, we measured spectrum (pink crosses). Images of the entire spectrometer screen are measure the energy modulation as the laser pulse is temporally scanned shown in Extended Data Fig. 2.

92 | NATURE | VOL 503 | 7 NOVEMBER 2013 ©2013 Macmillan Publishers Limited. All rights reserved RESEARCH LETTER ab Laser pulse (λ = 800 nm) Spectrometer magnet Magnetic Cylindrical lens lenses Electron DLA device beam

B RESEARCH LETTER Scattered Energy electrons Electrons Transmitted electrons ab 2 μm Laser pulse (λ = 800 nm) Lanex screen Spectrometer Intensimagnetfed CCD Magnetic Cylindrical lens camera lenses Figure 1 | DLA structure and experimental set-up. a, ScanningElectron electron field polarization directionDLA and device the effective periodic phase reset, depicted as microscope image of the longitudinal cross-section of a DLA structurebeam alternating red (acceleration) and black (deceleration) arrows. A snapshot of fabricated as depicted in Extended Data Fig. 1a. Scale bar, 2 mm. the simulated fields in the structure shows the correspondingB spatial b, Experimental set-up. Inset, a diagram of the DLA structure indicating the modulation in the vacuum channel. See text for details. Scattered Energy electrons than an 800-nm gap structure but requires tighter tolerancesElectrons on the across the electron beam, forming a cross-correlationTransmitted signal. A sample electron beam. measurement at a laser pulse energy of 91.8electrons6 1.3 mJ over a laser delay The NIR pulses, 1.24 6 0.12 ps long, from a regeneratively amplified of 6 ps is shown Fig. 3a. The orange circles (laser-off data) show no Ti:sapphire mode-locked laser are focused to2 an μm r.m.s. spot size of variation correlated with laser delay, as expected, and have an r.m.s. Lanex screen 30 mm 3 300 mm at the interaction point. We use a motorized four- deviation of 4.5 keV, which is taken asIntensi the noisefed CCD floor level of the mea- axis stage for precise alignment of the structure with the electron beam. surement. The blue circles (laser-on data)camera show the expected sech2 dis- OnceFigure aligned, 1 | DLA the structure electron and beam experimental leaving set-up. the structure a, Scanning goes electron through a tributionfield polarization with a full-width direction and at half-maximum the effective periodic (FWHM) phase of reset, 1.89 depicted6 0.09 as ps, point-to-pointmicroscope image focusing of the longitudinal spectrometer cross-section magnet, of whicha DLA structure disperses the alternating red (acceleration) and black (deceleration) arrows. A snapshot of outgoingfabricated as electron depicted beam in Extended in energy Data Fig. onto 1a. a Scale Kodak bar, 2Lanexmm. phosphor the simulated fields in the structure shows the corresponding spatial screenb, Experimental that is imaged set-up. Inset, by an a intensified diagram of the CCD DLA (charge-coupled structure indicating device) the modulation in the vacuum channel. SeeCharge text for density details. (arbitrary units) camera. According to particle tracking simulations, 2.2% of the 60- 0 0.2 0.4 0.6 0.8 1 MeV beam is transmitted through the vacuum channel of the 400-nm a than an 800-nm gap structure but requires tighter tolerances on the across15 the electron beam, forming a cross-correlation signal. A sample gapelectron structure beam. (see Methods). A segment of the spectrometer screen measurement at a laser pulse energy of 91.8 6 1.3 mJ over aLaser laser off delay focusingThe NIR on pulses, this transmitted 1.24 6 0.12 distribution ps long, from is a regeneratively shown in Fig. amplified 2a. The of 6 ps is shown Fig. 3a. The orange circles (laser-off data) show no horizontalTi:sapphire axis mode-locked represents laser beam are energy, focused and to the an entire r.m.s. image spot size spans of variation12 correlated with laser delay, as expected, and have an r.m.s. 24030 m keV.m 3 The300 centralmm at the pixel interaction location of point. the 60-MeV We use beam a motorized is taken as four- the deviation of 4.5 keV, which is taken as the noise floor level of the mea- reference point, corresponding to zero energy deviation (DE). Energy gain 2 axis stage for precise alignment of the structure with the electron beam. surement. 9 The blue circles (laser-on data) show the expected sech dis- OnceThe aligned, spectrometer the electron image beam in Fig. leaving 2a is a the median structure filtered goes average through of a a tributionb with a full-width at half-maximum (FWHM) of 1.89 6 0.09 ps, dozenpoint-to-point shots. The focusing least-squares spectrometer fit to the magnet, distribution which of electrons disperses scat- the 15 Laser on teredoutgoing by the electron fused silica beam substrate in energy and the onto grating a Kodak teeth has Lanex been phosphor removedObserved Energy Gain fromscreen this that image is imaged to emphasize by an the intensified transmitted CCD distribution (charge-coupled (see Methods). device) 12 Charge density (arbitrary units) Acamera. similarly According averaged to set particle of laser-on tracking spectrometer simulations, images 2.2% within of the 0.5 60- ps 0 0.2 0.4 0.6 0.8 1

ofMeV the beam optimal is transmitted timing overlap through for the laser vacuum pulses channel with energy of the 93 400-nmmJper aPosition (mm) Position (mm) pulsegap structure is shown (see in Fig. Methods). 2b. The white A segment contour of in the both spectrometer Fig. 2a and Fig. screen 2b 15 9 Laser off denotesfocusing the on location this transmitted where the distribution spectral charge is shown density in is Fig. 4.5% 2a. of The the c Laser off maximumhorizontal density axis represents (at the peak beam of energy, the scattered and the distribution). entire image In spans the 0.212 Spectrum ft presence240 keV. of The a laser central field, pixel there location exists of a the higher 60-MeV charge beam density is taken on either as the Laser on sidereference of the point, original corresponding peak at DE 5 to0. zero The energy white contour deviation shows (DE). a sizable 0.15 Energy Modelgain 9 Simulation fractionThe spectrometer of electrons image with maximum in Fig. 2a is energy a median that filtered is ,60 average keV higher of a b thandozen in shots. the laser-off The least-squares case. fit to the distribution of electrons scat- 15 Laser on 0.1 teredThe by laser-induced the fused silica energy substratemodulation and the is grating readily teeth apparent has been in the removed energy spectrafrom this (Fig. image 2c). to Using emphasize the fits the transmitted to these spectra, distribution a maximum (see Methods). energy 12 shiftA similarly of 53.1 keV averaged is calculated set of laser-on from the spectrometer abscissa of the images half-width within at 0.5 half- ps 0.05 Accelerated maximum (HWHM) point in the high-energy tail. We use an analy- electrons of the optimal timing overlap for laser pulses with energy 93 mJper Position (mm) Position (mm)

Charge density (arbitrary units) ticalpulse interaction is shown in model Fig. 2b. (see The Methods white contour and Extended in both Fig. Data 2a and Fig. Fig. 3) 2b to 09 c calculatedenotes the an locationaccelerating where gradient the spectral from this charge measurement. densityPeralta et al., is 4.5% Figure of theEnergy deviation, 2c ΔE Nature 503, 91 (2013) –100 –50 0 50 100 showsmaximum the input density electron (at the beam peak distribution of the scattered used distribution). in the model In (blue the 0.2 Laser off Energy deviation, ΔE (keV) Spectrum ft curve),presence which of a is laser a fit field, to the there measured exists spectrum a higher charge inRasmus Ischebeck > EPFL > Accelerator Research the absence density on of a either laser Edgar Peralta34 Figure 2 | Demonstration of energy modulation. a, ImageLaser of the on transmitted fieldside (light of the blue original crosses). peak The at D calculatedE 5 0. The energy white contour modulation shows (red a sizable curve) Model electron0.15 beam on the spectrometer screen, with the laser off. b,Asa but when agreesfraction with of our electrons measurement with maximum (pink crosses), energy and that gives is ,60 a correspond- keV higher Simulation the laser field is present. c, Energy spectra from a and b showing energy ingthan accelerating in the laser-off gradient case. of 151.2 MeV m21 for this example. Particle modulation.0.1 A fit (blue curve) to the measured laser-off spectrum (light blue trackingThe laser-induced simulations (black energy dots;modulation see Methods) is readily at this apparent gradient in the level energy give crosses) is used as input for the simulations. The calculated energy modulation anspectra independent (Fig. 2c). confirmation Using the fits of to the these observed spectra, modulated a maximum spectrum. energy (red curve) and particle tracking simulations (black dots) agree with our shiftTo of determine 53.1 keV the is calculated maximum from gradient the abscissa at a given of the laser half-width power level, at half- we measured0.05 spectrum (pink crosses). Images of the entireAccelerated spectrometer screen are measuremaximum the (HWHM) energy modulation point in the as the high-energy laser pulse tail. is temporally We use an scanned analy- shown in Extended Data Fig. 2. electrons Charge density (arbitrary units) tical interaction model (see Methods and Extended Data Fig. 3) to 0 92 | NATURE | VOL 503 | 7 NOVEMBER 2013 calculate an accelerating gradient from this measurement. Figure 2c –100 –50 0 50 100 ©2013 Macmillan Publishers Limited. All rights reserved shows the input electron beam distribution used in the model (blue Energy deviation, ΔE (keV) curve), which is a fit to the measured spectrum in the absence of a laser field (light blue crosses). The calculated energy modulation (red curve) Figure 2 | Demonstration of energy modulation. a, Image of the transmitted agrees with our measurement (pink crosses), and gives a correspond- electron beam on the spectrometer screen, with the laser off. b,Asa but when 21 the laser field is present. c, Energy spectra from a and b showing energy ing accelerating gradient of 151.2 MeV m for this example. Particle modulation. A fit (blue curve) to the measured laser-off spectrum (light blue tracking simulations (black dots; see Methods) at this gradient level give crosses) is used as input for the simulations. The calculated energy modulation an independent confirmation of the observed modulated spectrum. (red curve) and particle tracking simulations (black dots) agree with our To determine the maximum gradient at a given laser power level, we measured spectrum (pink crosses). Images of the entire spectrometer screen are measure the energy modulation as the laser pulse is temporally scanned shown in Extended Data Fig. 2.

92 | NATURE | VOL 503 | 7 NOVEMBER 2013 ©2013 Macmillan Publishers Limited. All rights reserved RESEARCH LETTER ab Laser pulse (λ = 800 nm) Spectrometer magnet Magnetic Cylindrical lens lenses Electron DLA device beam

B

Scattered Energy electrons Electrons Transmitted electrons

2 μm Lanex screen Intensifed CCD camera Figure 1 | DLA structure and experimental set-up. a, Scanning electron field polarization direction and the effective periodic phase reset, depicted as microscope image of the longitudinal cross-section of a DLA structure alternating red (acceleration) and black (deceleration) arrows. A snapshot of fabricated as depicted in Extended Data Fig. 1a. Scale bar, 2 mm. the simulated fields in the structure shows the corresponding spatial b, Experimental set-up. Inset, a diagram of the DLA structure indicating the modulation in the vacuum channel. See text for details. than an 800-nm gap structure but requires tighter tolerances on the across the electron beam, forming a cross-correlation signal. A sample electron beam. measurement at a laser pulse energy of 91.8 6 1.3 mJ over a laser delay The NIR pulses, 1.24 6 0.12 ps long, from a regeneratively amplified of 6 ps is shown Fig. 3a. The orange circles (laser-off data) show no Ti:sapphire mode-locked laser are focused to an r.m.s. spot size of variation correlated with laser delay, as expected, and have an r.m.s. 30 mm 3 300 mm at the interaction point. We use a motorized four- deviation of 4.5 keV, which is taken as the noise floor level of the mea- axis stage for precise alignment of the structure with the electron beam. surement. The blue circles (laser-on data) show the expected sech2 dis- Once aligned, the electron beam leaving the structure goes through a tribution with a full-width at half-maximum (FWHM) of 1.89 6 0.09 ps, point-to-point focusing spectrometer magnet, which disperses the outgoing electron beam in energy onto a Kodak Lanex phosphor screen that is imaged by an intensified CCD (charge-coupled device) Charge density (arbitrary units) camera. According to particle tracking simulations, 2.2% of the 60- 0 0.2 0.4 0.6 0.8 1 MeV beam is transmitted through the vacuum channel of the 400-nm a gap structure (see Methods). A segment of the spectrometer screen 15 Laser off focusing on this transmitted distribution is shown in Fig. 2a. The horizontal axis represents beam energy, and the entire image spans 12 240 keV. The central pixel location of the 60-MeV beam is taken as the reference point, corresponding to zero energy deviation (DE). Energy gain 9 The spectrometer image in Fig. 2a is a median filtered average of a b dozen shots. The least-squares fit to the distribution of electrons scat- 15 Laser on tered by the fused silica substrate and the grating teeth has been removed from this image to emphasize the transmitted distribution (see Methods). 12 A similarly averaged set of laser-on spectrometer images within 0.5 ps Accelerated Electrons

of the optimal timing overlap for laser pulses with energy 93 mJper Position (mm) Position (mm) pulse is shown in Fig. 2b. The white contour in both Fig. 2a and Fig. 2b 9 denotes the location where the spectral charge density is 4.5% of the c maximum density (at the peak of the scattered distribution). In the 0.2 Laser off Spectrum ft presence of a laser field, there exists a higher charge density on either Laser on side of the original peak at DE 5 0. The white contour shows a sizable 0.15 Model fraction of electrons with maximum energy that is ,60 keV higher Simulation than in the laser-off case. 0.1 The laser-induced energy modulation is readily apparent in the energy spectra (Fig. 2c). Using the fits to these spectra, a maximum energy shift of 53.1 keV is calculated from the abscissa of the half-width at half- 0.05 Accelerated maximum (HWHM) point in the high-energy tail. We use an analy- electrons Charge density (arbitrary units) tical interaction model (see Methods and Extended Data Fig. 3) to 0 calculate an accelerating gradient from this measurement.Peralta Figure et al., 2c –100 –50 0 50 100 Nature 503, 91 (2013) shows the input electron beam distribution used in the model (blue Energy deviation, ΔE (keV) curve), which is a fit to the measured spectrum in the absenceRasmus Ischebeck > EPFL > Accelerator Research of a laser Edgar Peralta35 field (light blue crosses). The calculated energy modulation (red curve) Figure 2 | Demonstration of energy modulation. a, Image of the transmitted agrees with our measurement (pink crosses), and gives a correspond- electron beam on the spectrometer screen, with the laser off. b,Asa but when 21 the laser field is present. c, Energy spectra from a and b showing energy ing accelerating gradient of 151.2 MeV m for this example. Particle modulation. A fit (blue curve) to the measured laser-off spectrum (light blue tracking simulations (black dots; see Methods) at this gradient level give crosses) is used as input for the simulations. The calculated energy modulation an independent confirmation of the observed modulated spectrum. (red curve) and particle tracking simulations (black dots) agree with our To determine the maximum gradient at a given laser power level, we measured spectrum (pink crosses). Images of the entire spectrometer screen are measure the energy modulation as the laser pulse is temporally scanned shown in Extended Data Fig. 2.

92 | NATURE | VOL 503 | 7 NOVEMBER 2013 ©2013 Macmillan Publishers Limited. All rights reserved Why is the Energy Spread of the Electrons Increased?

The electrons scatter in the material

The laser energy is fluctuating

The structure is misaligned

The electron bunch length is longer than the wavelength of the laser

Rasmus Ischebeck > EPFL > Accelerator Research 36 Gradients have been observed that are 10 times higher Gradients have been observed that are 10 times higher than thanthe mainthe main SLAC SLAC linac linac…… Peak gradient as a function of Laser Field 300300 MV/m MV/m

• Joel England

Peralta et al., Nature 503, 91 (2013)

Rasmus Ischebeck > EPFL > Accelerator Research Joel England37 Acceleration of Relativistic Electrons Kerr Saturation

Cesar et al., Communications Physics 1, 46 (2018)

Rasmus Ischebeck > EPFL > Accelerator Research David Cesar, Xinglai Shen, Kent Wootton, Joel England, Pietro Musumeci BREUER et al. Phys. Rev. ST Accel. Beams 17, 021301 (2014)

radiation source [21,22] at a metal grating of 250 μm characterized by n oscillations per grating period λ [n 3 p ¼ period, but acceleration gradients were too small (keV=m) in Figs. 1(a–c)]. Therefore, it propagates along the grating to compete with rf accelerators. Our proof-of-concept surface with a phase velocity v fλ =n cλ = nλ . ph ¼ p ¼ p ð Þ experiment exploits this effect in the vicinity of a dielectric Hence, particles with the velocity v βc vph can surf grating structure. The difference to the originally proposed on and continuously interact with this¼ synchronous¼ mode. inverse Smith-Purcell effect is that here the grating modes This yields the synchronicity condition [25] are excited in transmission as opposed to reflection [20]. The dielectric grating used in our experiment is directly λ β p : (1) compatible with the double grating structures proposed by ¼ nλ Plettner et al. [9,23–26]. Recently, dielectric laser accel- eration of relativistic electrons has been observed at SLAC By solving the wave equation c2 2 − 2= t2 E⃗ r;⃗ t 0, [27], in parallel with our demonstration of nonrelativistic it can be shown that the field strengthð ∇ falls∂ off∂ exponentiallyÞ ð Þ¼ electron acceleration at a similar structure [28]. Because of with increasing distance from the grating surface with a the intercompatibility of the two experiments, an all-optical decay constant laser-driven accelerator, including nonrelativistic and rela- tivistic sections, seems now feasible. In this paper we βγλ Γ ; (2) present a detailed overview of our experiment, including ¼ 2π setup, simulations, and results. with γ 1= 1 − β2 [26]. Because of the field geometry of ¼ II. THEORY the synchronous mode the particle can experience accel- eration, deceleration,pffiffiffiffiffiffiffiffiffiffiffiffiffi or deflection, depending on its relative Particle acceleration at a grating is based on the position inside the field [Figs. 1(a–c)]. The effect of all other diffraction of laserSingle light Grating oscillating with a frequency asynchronous spatial harmonics averages to zero over time. f c=λ, with c the speed of light and λ the laser wave- ¼ As nonrelativistic electrons significantly change speed length. The diffracted light field is associated with evan- during the acceleration they may run out of phase and escent modes, also known as spatial harmonics, in close eventually become decelerated again. Therefore, in future vicinity• Grating transforms plane wave into accelerating field of the grating surface. The nth spatial harmonic is experiments the grating period needs to be adaptively

Breuer et al., PRST-AB 17, 021301 (2014)

Rasmus Ischebeck > EPFL > Accelerator Research 39 FIG. 1 (color online). (a–c) Three consecutive snapshots in time (a quarter optical period apart) of the electric field distribution of the third spatial harmonic (blue arrows) above a grating (light blue structure). This surface wave, excited by a linearly polarized laser from below (red arrows), propagates synchronously with the charged particles (numbered circles) along the grating surface. Here the charged particles are assumed to be positrons. Depending on the relative position of the positron inside the field the force can lead to either acceleration (1), deceleration (2), or deflection (3,4), as indicated by the blue arrows and the color shading. (d) Electron microscope image of the fused silica grating that is located on top of a mesa, 20 μm above the substrate. The closeup shows the grating with a grating period λ 750 nm, a trench width of 325 nm, and a depth of 280 nm. (e) Top view of the mesa with a width of 25 μm. p ¼

021301-2 BREUER et al. Phys. Rev. ST Accel. Beams 17, 021301 (2014) radiation source [21,22] at a metal grating of 250 μm characterized by n oscillations per grating period λ [n 3 p ¼ period, but acceleration gradients were too small (keV=m) in Figs. 1(a–c)]. Therefore, it propagates along the grating to compete with rf accelerators. Our proof-of-concept surface with a phase velocity v fλ =n cλ = nλ . ph ¼ p ¼ p ð Þ experiment exploits this effect in the vicinity of a dielectric Hence, particles with the velocity v βc vph can surf grating structure. The difference to the originally proposed on and continuously interact with this¼ synchronous¼ mode. inverse Smith-Purcell effect is that here the grating modes This yields the synchronicity condition [25] are excited in transmission as opposed to reflection [20]. The dielectric grating used in our experiment is directly λ β p : (1) compatible with the double grating structures proposed by ¼ nλ Plettner et al. [9,23–26]. Recently, dielectric laser accel- eration of relativistic electrons has been observed at SLAC By solving the wave equation c2 2 − 2= t2 E⃗ r;⃗ t 0, [27], in parallel with our demonstration of nonrelativistic it can be shown that the field strengthð ∇ falls∂ off∂ exponentiallyÞ ð Þ¼ electron acceleration at a similar structure [28]. Because of with increasing distance from the grating surface with a the intercompatibility of the two experiments, an all-optical decay constant laser-driven accelerator, including nonrelativistic and rela- tivistic sections, seems now feasible. In this paper we βγλ DIELECTRIC LASER ACCELERATION OF … Phys. Rev. ST Accel. Beams 17, 021301 (2014) Γ ; (2) present a detailed overview of our experiment, including ¼ 2π setup, simulations, and results. Experiments with Non-Relativistic Electrons with γ 1= 1 − β2 [26]. Because of the field geometry of ¼ II. THEORY the synchronous mode the particle can experience accel- eration, deceleration,pffiffiffiffiffiffiffiffiffiffiffiffiffi or deflection, depending on its relative Particle acceleration at a grating is based on the position inside the field [Figs. 1(a–c)]. The effect of all other diffraction of laser light oscillating with a frequency asynchronous spatial harmonics averages to zero over time. f c=λ, with c the speed of light and λ the laser wave- ¼ As nonrelativistic electrons significantly change speed length. The diffracted light field is associated with evan- during the acceleration they may run out of phase and escent modes, also known as spatial harmonics, in close eventually become decelerated again. Therefore, in future vicinity of the grating surface. The nth spatial harmonic is experiments the grating period needs to be adaptively

FIG. 1 (color online). (a–c) Three consecutive snapshots in time (a quarter optical period apart) of the electric field distribution of the third spatial harmonic (blue arrows) above a grating (light blueBreuer structure). et al., This surface wave, excited by a linearly polarized laser from below (red arrows), propagates synchronously with the chargedPRST-AB particles (numbered17, 021301 circles) (2014) along the grating surface. Here the charged particles are assumed to be positrons. Depending on the relative position of the positron inside the field the force can lead to either acceleration (1), deceleration (2), or deflection (3,4), as indicated by the blue arrows and the color shading. (d) Electron microscope image of the fused silica grating that is located on top of aRasmus Ischebeck > EPFL > Accelerator Research mesa, 20 μm above the substrate. The closeup shows the grating with a grating John Breuer40 period λ 750 nm, a trench width of 325 nm, and a depth of 280 nm. (e) Top view of the mesa with a width of 25 μm. p ¼ 021301-2 FIG. 2 (color online). (a) Conceptual picture of the electron acceleration detection. Electrons (blue trajectory) are emitted from an electron column (left) and interact with the laser pulses (red line), derived from a Titanium:sapphire oscillator, at the fused silica grating. A microscope objective is used to monitor the position of the laser focus. The electrons pass through a retarding field spectrometer, which blocks all unaccelerated electrons (counter voltage UG), and are detected with a MCP. The trajectory entering the spectrometer is drawn as slightly offcenter to illustrate the effect of the retarding field. A magnetic field B⃗ deflects the electrons around an x-ray beam stop to separate them from high energy photons originating inside the electron column. A time-to-digital converter (TDC) is used to measure the time delay Δt between a detector event and the following laser pulse. This way, a signal of accelerated electrons appears at a fixed Δt in a histogram while background counts are distributed equally over all delays (inset). (b) Technical drawing of the experimental setup with zoom-in of the grating and interaction region. The experiment is placed inside a vacuum chamber (base flange size DN 350 CF) evacuated to ∼5 × 10−7 mbar. The microscope objective (xyz-degrees of freedom), the spectrometer (yz-degrees of freedom) and the grating mount (xyz-degrees of freedom) can be positioned relatively to the electron beam with motorized translation stages. Additionally, the grating can be fine positioned with stick-slip piezoelectric actuators (translation in z-direction and rotation in xz-plane). The achromatic lens which is used to focus the laser pulses can be manually positioned with a compact dovetail linear stage.

increased to stay in phase with the accelerating electrons. λ 787 nm, a pulse duration τp 110 fs, and a pulse Otherwise, without changing the grating period, dephasing energy¼ of 160 nJ (450 mW average¼ output power) as the limits the distance over which electrons can be accelerated. laser source [30]. A typical spectrum is shown in Fig. 3. From our calculations we infer that this acceleration The laser pulses are focused with an achromatic spherical distance is about 25 μm for acceleration gradients below lens with focal length of 30 mm onto the grating with a 25 MeV=m for 30 keV electrons (β 0.33). Extensive focal waist radius wl of 9.0 0.4 μm. A microscope simulation results and implications are¼ reported in [29]. objective is used to monitorð theÆ laserÞ focal spot size and position relative to the grating and electron beam and is also placed inside the vacuum chamber. III. EXPERIMENTAL SETUP A. Overview B. Grating Two schematic overviews of the experimental setup are The fused silica grating [Figs. 1(d) and 1(e)] has been shown in Fig. 2. The electron beam passes through the manufactured using electron beam and laser lithography in interaction region where it interacts with the evanescent combination with reactive ion etching. With the laser field excited at the grating surface by laser pulses imping- wavelength of 787 nm, a grating period of 250 nm is ing from behind. The electrons enter a retarding field needed to accelerate E kin 27:9 keV electrons (β 0.32) spectrometer. Accelerated electrons that pass the spectrom- using the first spatial harmonic.¼ Because of a lower¼ bound eter are deflected by a magnetic field and are detected with of λp 600 nm set by the manufacturer [31], we chose a a microchannel plate detector (MCP). The beam is grating≳ period of λ 750 nm and hence use the third p ¼ deflected to reduce the background signal at the detector, spatial harmonic, as depicted in Figs. 1(a–c). The grating is as will be discussed below. A camera, placed outside the located on top of a mesa (size: 2 mm × 35 μm × 20 μm) vacuum chamber, can be used to observe the MCP counts. that sits on top of the substrate (size: 3 mm× We use a Titanium:sapphire long-cavity oscillator with a 20 mm × 1 mm). This mesa structure allows spatial access repetition rate f 2.7 MHz, a center wavelength to the grating and therefore the electron beam focus can rep ¼

021301-3 DIELECTRIC LASER ACCELERATION OF … Phys. Rev. ST Accel. Beams 17, 021301 (2014) Experiments with Non-Relativistic Electrons this to be the experimental limit for the minimum distance • Energy gain for two different laser peak electric fields between the electrons and the grating. We assume that • E = 2.85 GV/m (orange circles), beam clipping together with residual surface charging 1 • E = 2.36 GV/m (blue squares) prevents a closer approach in the current setup. Only the component of the laser electric field that is parallel to the electrons’ velocity can excite the accelerating 0.1 spatial harmonic. Hence, the accelerated fraction depends on the laser polarization angle φ between the incident laser electric field and the electrons’ trajectory like cos ϕ (Fig. 11). This measurement proves the direct accelerationð Þ 0.01 with the laser field and rules out intensity-dependent but Accelerated fraction polarization-independent ponderomotive acceleration [45]. Its effect we estimate to a maximum of Gpond 2 2 2 ¼ -3 e E p exp −0.5 = 2mω wl 12 keV=m, with the elec- 1x10 tron massð m andÞ ð the laserÞ¼ angular frequency ω 2πf. Breuer et al., 0 100 200 300 ¼ PRST-AB 17, 021301 (2014) Energy gain (eV) The sinusoidal fit in Fig. 11 gives a limit on a possible

Rasmus Ischebeck > EPFL > Accelerator Research 41 angular misalignment of the grating with respect to the FIG. 10 (color online). Measurement of the accelerated fraction electron beam of φ0 3.2 5.3 °. This angular mis- as a function of energy gain (bottom axis) and acceleration alignment implies an¼ð offset betweenÆ Þ the experimentally gradient (top axis) for two different laser peak electric fields measured electron energy and the energy used in the [E 2.85 GV=m (orange circles), E 2.36 GV=m (blue p ¼ p ¼ simulations, where we set ϕ0 0, in the following way. squares)]. We measure a maximum energy gain of 280 eV The component of the electron¼ velocity that is parallel to corresponding to a maximum acceleration gradient of the grating vector has to satisfy the synchronicity con- 25 MeV=m. The curves represent simulation results, which were dition and therefore β cos φ0 λp= nλ . However, in obtained according to Eq. (8) for z0 120 nm and we 77 nm. ð Þ¼ ð Þ The overall amplitude of the simulated¼ values has been¼ scaled to the experiment we measure the total kinetic energy related fit the experimental data. to an electron velocity β,whichisafactorof1= cos ϕ0 larger than the design velocity λp= nλ . We calculate the resulting shift of the measuredð kineticÞ energy to 0.03eE πw 100 pp int be −100þ eV. zmax Γ ln : (9) −570 ¼ E The measurement of the accelerated fraction as a  Δ ffiffiffi  function of the relative distance z0 between the electron For example, assuming E 2.85 GV=m, electrons have p ¼ to pass the grating within zmax 94 nm to gain more than 100 eV in energy. ¼

V. EXPERIMENTAL RESULTS 0.4 In this section we discuss the experimental results published in [28] in more detail. 0.3 In Fig. 10 we compare measurements of the accelerated fraction as a function of the acceleration gradient for two 0.2 different laser peak electric fields of E 2.85 GV=m p ¼ (average power P 450 mW, peak intensity Ip 12 2 ¼ 2 ¼ 0.1 2.2 × 10 W=cm , peak fluence F p 0.13 J=cm ) and Accelerated fraction ¼ 12 2 E p 2.36 GV=m(P 300 mW, Ip 1.5 × 10 W=cm , F ¼ 0.09 J=cm2) with¼ simulation results.¼ We observe a 0.0 p ¼ maximum measured energy gain of ΔE 280 eV. It corresponds to a maximum acceleration¼ gradient of -100 -50 0 50 100 Polarization angle (degrees) Gmax 25 MeV=m, according to Eq. (6). This is already comparable¼ with state-of-the-art rf linacs. The simulated FIG. 11 (color online). Accelerated fraction of electrons as a curves of the accelerated fraction assume a distance of the function of the laser polarization angle ϕ relative to the electrons’ electron beam center from the grating surface of z0 trajectory. ϕ 0 means that the laser polarization is parallel to 120 10 nm and an electron beam waist of w 77 nm.¼ ¼ ∘ ð Æ Þ e ¼ the electrons’ momentum, ϕ 90∘ that it is perpendicular to it We deduce from our simulations that the maximum (see inset). The data agree well¼ with the expected cosine behavior acceleration occurs for electrons that pass the grating at (orange fit curve) and proves that the electrons are directly a distance of ∼50 nm due to the finite beam width. We infer accelerated by the light field.

021301-9 Circular Structures

• Structure geometry:

electron beam E

metal d a dielectric

Nanni et al. Proceedings of IPAC 2014

Rasmus Ischebeck > EPFL > Accelerator Research Franz Kärtner 42 WEOAB03 Experiments Proceedings of IPAC2014, Dresden, Germany

1 Measured Measured Modeled Modeled 0.8

0.6

0.4 Counts (Arb.) 0.2

0 45 50 55 60 65 70 45 50 55 60 65 70 75 Energy (keV) Energy (keV)

Nanni et al. (a) (b) Proceedings of IPAC 2014 (a) Franz Kärtner Rasmus Ischebeck > EPFL > Accelerator Research Figure43 4: Measured (black) and modeled (red) energy spec- −1.5 1 1 trum with THz (a) off and (b) on at a gun voltage of 59 kV. −1 0.8 0.8 is limited to 3 mm due to the low initial energy of the elec- −0.5 0.6 0.6 trons which results in the rapid onset of a phase-velocity 0 mismatch between the electron bunch and the THz pulse 0.4 0.4 0.5 once the electrons have been accelerated by the THz pulse. Distance (mm) 1 0.2 0.2 The energy spectrum from the electron bunch with and 1.5 0 Counts (Arb. Units.) 0 −1−0.5 0 0.5 1 1.5 −1 −0.5 0 0.5 1 1.5 without THz is shown in Fig. 4 for an initial mean energy Distance (mm) Distance (mm) of 59 keV. The electron bunch length after the pin-hole, (b) (c) σz = 45 µm, is long with respect to the wavelength of = Figure 3: (a) THz LINAC and source with the THz acceler- the THz pulse in the waveguide, λg 315 µm, resulting ation chamber and accompanying power supplies, chillers in both the acceleration and deceleration of particles. The and pumps fit on a portable optical cart. (b) Image of THz waveguide is sensitive to the initial energy of the elec- the electron beam from an MCP at 50 kV. (c) Compari- tron bunch, due to the rapidly varying velocity of the elec- son between simulated (black) and measured (red) electron trons. If the initial energy is too low, acceleration is not ob- served. With the available THz pulse energy, a peak energy bunch at MCP, with 25 fC per bunch, a σ⊥ = 513µmand ∆E/E = 1.25 keV. After the pin-hole the transverse emit- gain of 7 keV was observed by optimizing the electron beam tance is 25 nm-rad and the longitudinal emittance is 5.5 nm- voltage and timing of the THz pulse. The modeled curve in rad. Fig. 4(b) was fit with on on-axis gradient of 4.9 MeV/m, indicating some loss of THz pulse energy due to misalign- THz LINAC ment. At the exit of the LINAC, the modeled transverse and longitudinal emittance are 240 nm-rad and 370 nm-rad, A60kVDCphoto-emissionelectrongunwasusedasthe respectively. This increase in emittance is due to the long injector for the THz-driven LINAC. Dielectric-loaded cir- electron bunch length compared to the THz wavelength and cular waveguides, described in the previous section, were can be easily remedied with a shorter UV pulse length. optimized for non-relativistic electron beams and used as CONCLUSION the acceleration structures. The accelerating waveguide is 10 mm in length, including a single tapered horn for cou- THz pulses generated via optical rectification of a 1µm pling the THz into the waveguide. The electron beam pro- laser were used to accelerate electrons in a simple and prac- duced by the DC electron gun, shown in Fig. 3(a), operates tical THz traveling-wave accelerating structure. A gradient with 25 fC per bunch at a repetition rate of 1 kHz. The of ∼10 MeV/m with 2 µJwasachievedduringtransmission photo-emission laser is a 350 fs UV pulse produced by 4th testing. A energy gain of 7 keV was achieved over a 3 mm in- harmonic generation from the 1 µmlaser.Fig.3(b)shows teraction length. Performance of these structures improves an image of the electron beam produced by the MCP camera. with an increase in electron energy and gradient making Afocusingsolenoidisusedtocollimatethebeamafterthe them attractive for compact accelerator applications. With THz LINAC. The electron beam energy is determined via upgrades to pump laser energy and technological improve- energy-dependent magnetic steering with a dipole located ments to THz sources, GeV/m gradients are achievable in after the accelerator. PARMELA simulations were used to dielectric-loaded circular waveguides. The available THz model the DC gun, Fig. 3(c), and the THz LINAC. Align- pulse energy scales with IR pump energy, with a recently re- ment between the THz waveguide and the DC gun is pro- ported result of 0.4 mJ and ∼1% conversion efficiency [11]. vided by a pin-hole aperture in a metal plate with a diameter Multiple stages of THz acceleration can be used to achieve of 100 µmthatabutsthewaveguide.TheTHzpulseiscou- higher energy gain with additional IR pump lasers for sub- pled into the waveguide downstream of the accelerator and sequent stages. it propagates the full length of the waveguide before being reflected by the pin-hole aperture, which acts as a short at REFERENCES THz frequencies. After being reflected the THz pulse co- [1] Dolgashev, Valery, et al., “Geometric dependence of radio- propagates with the electron bunch. The interaction length frequency breakdown in normal conducting accelerating ISBN 978-3-95450-132-8 03 Particle Sources and Alternative Acceleration Techniques

Copyright ©1898 2014 CC-BY-3.0 and by the respective authors A16 Advanced Concepts Circular Structures: Measurements WEOAB03 Proceedings of IPAC2014, Dresden, Germany

1 Measured Measured Modeled Modeled 0.8

0.6

0.4 Counts (Arb.) 0.2

0 45 50 55 60 65 70 45 50 55 60 65 70 75 Energy (keV) Energy (keV) (a) (b) (a) Figure 4: Measured (black) and modeled (red) energy spec- −1.5 1 1 trum with THz (a) off and (b) on at a gun voltage of 59 kV.

−1 0.8 0.8 Nanni et al. Proceedings of IPAC 2014is limited to 3 mm due to the low initial energy of the elec- −0.5 0.6 0.6 Rasmus Ischebeck > EPFL > Accelerator Researchtrons which results in the rapid onset of a phase-velocity 44 0 mismatch between the electron bunch and the THz pulse 0.4 0.4 0.5 once the electrons have been accelerated by the THz pulse. Distance (mm) 1 0.2 0.2 The energy spectrum from the electron bunch with and 1.5 0 Counts (Arb. Units.) 0 −1−0.5 0 0.5 1 1.5 −1 −0.5 0 0.5 1 1.5 without THz is shown in Fig. 4 for an initial mean energy Distance (mm) Distance (mm) of 59 keV. The electron bunch length after the pin-hole, (b) (c) σz = 45 µm, is long with respect to the wavelength of = Figure 3: (a) THz LINAC and source with the THz acceler- the THz pulse in the waveguide, λg 315 µm, resulting ation chamber and accompanying power supplies, chillers in both the acceleration and deceleration of particles. The and pumps fit on a portable optical cart. (b) Image of THz waveguide is sensitive to the initial energy of the elec- the electron beam from an MCP at 50 kV. (c) Compari- tron bunch, due to the rapidly varying velocity of the elec- son between simulated (black) and measured (red) electron trons. If the initial energy is too low, acceleration is not ob- served. With the available THz pulse energy, a peak energy bunch at MCP, with 25 fC per bunch, a σ⊥ = 513µmand ∆E/E = 1.25 keV. After the pin-hole the transverse emit- gain of 7 keV was observed by optimizing the electron beam tance is 25 nm-rad and the longitudinal emittance is 5.5 nm- voltage and timing of the THz pulse. The modeled curve in rad. Fig. 4(b) was fit with on on-axis gradient of 4.9 MeV/m, indicating some loss of THz pulse energy due to misalign- THz LINAC ment. At the exit of the LINAC, the modeled transverse and longitudinal emittance are 240 nm-rad and 370 nm-rad, A60kVDCphoto-emissionelectrongunwasusedasthe respectively. This increase in emittance is due to the long injector for the THz-driven LINAC. Dielectric-loaded cir- electron bunch length compared to the THz wavelength and cular waveguides, described in the previous section, were can be easily remedied with a shorter UV pulse length. optimized for non-relativistic electron beams and used as CONCLUSION the acceleration structures. The accelerating waveguide is 10 mm in length, including a single tapered horn for cou- THz pulses generated via optical rectification of a 1µm pling the THz into the waveguide. The electron beam pro- laser were used to accelerate electrons in a simple and prac- duced by the DC electron gun, shown in Fig. 3(a), operates tical THz traveling-wave accelerating structure. A gradient with 25 fC per bunch at a repetition rate of 1 kHz. The of ∼10 MeV/m with 2 µJwasachievedduringtransmission photo-emission laser is a 350 fs UV pulse produced by 4th testing. A energy gain of 7 keV was achieved over a 3 mm in- harmonic generation from the 1 µmlaser.Fig.3(b)shows teraction length. Performance of these structures improves an image of the electron beam produced by the MCP camera. with an increase in electron energy and gradient making Afocusingsolenoidisusedtocollimatethebeamafterthe them attractive for compact accelerator applications. With THz LINAC. The electron beam energy is determined via upgrades to pump laser energy and technological improve- energy-dependent magnetic steering with a dipole located ments to THz sources, GeV/m gradients are achievable in after the accelerator. PARMELA simulations were used to dielectric-loaded circular waveguides. The available THz model the DC gun, Fig. 3(c), and the THz LINAC. Align- pulse energy scales with IR pump energy, with a recently re- ment between the THz waveguide and the DC gun is pro- ported result of 0.4 mJ and ∼1% conversion efficiency [11]. vided by a pin-hole aperture in a metal plate with a diameter Multiple stages of THz acceleration can be used to achieve of 100 µmthatabutsthewaveguide.TheTHzpulseiscou- higher energy gain with additional IR pump lasers for sub- pled into the waveguide downstream of the accelerator and sequent stages. it propagates the full length of the waveguide before being reflected by the pin-hole aperture, which acts as a short at REFERENCES THz frequencies. After being reflected the THz pulse co- [1] Dolgashev, Valery, et al., “Geometric dependence of radio- propagates with the electron bunch. The interaction length frequency breakdown in normal conducting accelerating ISBN 978-3-95450-132-8 03 Particle Sources and Alternative Acceleration Techniques

Copyright ©1898 2014 CC-BY-3.0 and by the respective authors A16 Advanced Concepts Acceleration and Deflection with THz Pulses DESY

Zhang et al., Nature Photonics (2018). doi:10.1038/s41566-018-0138-z Rasmus Ischebeck > EPFL > Accelerator Research 45 ARTICLES NATURE PHOTONICS

Electric mode Magnetic mode abe Max 0 1 qν × B 60 10 55 Min

Energy (keV) 50 5

B field cancelled qE E field cancelled qν × B –qE qE 0 1234 5 0 1234 5 0 c d 10 Deflection (mrad) –5

0 –10 –10 Deflection (mrad) –qν × B B field cancelled qE E field cancelled qν × B 0 1234 5 0 1234 5 50 55 60 Time (ps) Time (ps) Energy (keV)

Fig. 2 | Concept and implementation. a, Measured energy modulation of e pulse as a function of electron–terahertz delay for constructive interference of the E fields entering the device and cancellation of the B fields. b, Corresponding beam deflection measured for constructive interference of the B fields, that is, E field cancellation scenario. c,d, Time-dependent deflection diagrams measured by varying the electron–terahertz delay in the B field (c) and E field (d) cancellation scenarios. e, Measured shape of e beam on MCP detector for maximum acceleration, deceleration, and right and left deflection points plotted in one image. IntensityAcceleration was normalized and with image THz contrast Pulses was tuned to show the relative positions more clearly. The red dashed line represents the predicted locus of beam positions corresponding to a sweep of the relative phases of the two THz waveforms. This demonstration was performed using a Yb-doped potassium yttrium tungstate (Yb:KYW) laser with ~2!× !0.5!µ J of terahertz radiation coupled into the device and a bunch charge of ~1!fC.

ab 90 1.0 With terahertz >30 keV Without terahertz

80 intensity (keV) 0.5 70 Energy Normalized 60

0.0 0 200 400 600 40 50 60 70 80 90 100 110 –1 Energy (keV) ETHz (kV cm ) cd 1323 1 23 1 2 1,000 90 Zhang et al., Nature Photonics (2018). doi:10.1038/s41566-018-0138-z Rasmus Ischebeck > EPFL > Accelerator Research 46 Layer 1 Layer 2 500 80 E z (kV cm (keV ) Layer3 0 70 –1 Energy )

–500 60

0.00.5 1.0 –1,000 0 ps 1.6 ps 3.3 ps Propagationdistance(mm)

Fig. 3 | Terahertz acceleration. a, Measured electron energy spectra for initial input beam (blue curve) and accelerated beam (red curve) that shows an energy gain of more than 30!keV. An increased energy spread is observed due to the long length of the initial electron bunch, as well as the slippage between the terahertz pulse and the electron bunch. b, Relative energy versus input terahertz field strength with the red circle indicating the energy spectra plotted in a. The linear relationship supports a direct, field-driven interaction. c, Temporal evolution of the electric field inside each layer with the red arrow indicating the electron propagating. d, Calculated acceleration along the electron propagation direction with ~2!×! 6!µJ terahertz radiation and beam diameter of 3!mm. This illustration was performed using the Yb:YLF laser with ~2!× !6 µ J terahertz radiation coupled into the device and a bunch charge of ~5!fC.

The function of the device was thus selected by tuning the relative infrared pump beams. In focusing and streaking modes, the elec- delay of the two terahertz pulses and the electrons, all of which were tron beams were sent directly to a microchannel plate (MCP) detec- controlled by means of motorized stages acting on the respective tor. For acceleration measurements, an electromagnetic dipole was

338 NATURE PHOTONICS | VOL 12 | JUNE 2018 | 336–342 | www.nature.com/naturephotonics © 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved. Letter Vol. 41, No. 15 / August 1 2016 / Optics Letters 3437 Dielectric Structures for Attosecond Beam Characterization

3436 Vol. 41, No. 15 / August 1 2016 / Optics Letters Letter The theoretical accelerated current density distribution [Figs. 2(c) and 2(d)] describes qualitatively the reduction of nanostructureMeasured caused by the interaction with the electron beam wasaccelerated observed during the experiments. The silicon nano-grating the area for higher values of eU s − E k0 (minimum energy gain). electron [SEMcurrent detail in Fig. 1(b)] is fabricated by electron beam lithog- However, it slightly differs from the experimental data. This raphy (JEOL JBX-6300) and subsequent reactive ion etching. The grating period is chosen as 620 nm; the grating depth is can be explained by two effects. The first effect is the charging 450 nm; and the trench/tooth width ratio is 55% to optimize of our silicon structure. Even though the resistivity of the sub- the excitation efficiency of the acceleration mode by the inci- dent laser pulse. Structures with four different widths in the strate is 10–20 Ωcm, there is a charge accumulated on the sur- x-direction are fabricated, namely, w 250 (sample S1), face of the nanostructure which changes the accelerated electron x ˆ 500 (S2), 750 (S3), and 1000 nm (S4). The SEM image of path. The second effect is the electron scattering off the sides of the structures in the x-z plane is shown in Fig. 1(c). The accel- erating fields are excited by near-infrared femtosecond laser the nanostructure. Due to the nanostructure’s length in the pulsesSimulation generated by 1 MHz repetition rate thulium- and/or electron propagation direction (100 m), many of the electrons thulium-holmium-doped fiber laser systems delivering pulses μ at wavelengths of λ 1.93 and 2.05 μm, respectively. The are scattered at imperfections on the sidewalls. These imperfec- pulse durations are ˆτ 600 50 fs (thulium) and 390 tions are present due to the fabrication process. This is not the 30 fs (thulium-holmium).ˆ A laser beam with polarization along the electron propagation direction (z-axis) is focused onto the case for electrons above the structure, where the flatness is bet- nanostructure along the −y direction with a 1∕e2 radius of ter, as it is given by the flatness of the silicon wafer used for w 7 1 μm. The peak electric field is limited by the laser Kozák et al., ˆ  fabrication. (The RMS surface roughness is specified to 2 Å.) Optics LettersFig. 41, 1. 15 (2016)(a) Experimental setup used for characterization of the spatial damage threshold of silicon to 1.2 GV/m corresponding to the properties of the optical near-fields which are excited by the femtosecond peak intensity of 190 GW∕cm2. After interaction with the Rasmus Ischebeck > EPFL > Accelerator Research 47 We further characterize the transverse distribution of the laser pulse (red) at the silicon nanostructure (gray). The structure is scan- synchronized fields, electrons are spectrally filtered by a retard- ned by a precise 2D xy stage across the focused electron beam (black). accelerating fields as a function of the structure width w . The ing field spectrometer,Fig. where 2. a(a), retardation (b) Measured voltage U s acceleratedis ap- electron current as a function x After the interaction, electrons are filtered by a retarding field spectrom- − plied. Only electrons with an energy gain ΔE k > eUs Ek0 current of accelerated electrons with ΔEk > 30 eV is measured eter and detected by a MCP. The time delay between each MCP count are transmitted throughof nanostructure the spectrometerx andand detectedy coordinates by a mi-† with respect to the center of the and a fast photodiode signal is measured by a TDC. The accelerated crochannel plate detector (MCP). The temporal delay between along the line y 50 nm above the structure surface for all electron signal appears at a specific time delay. (b) SEM image of struc- electron beam for (a) ΔE k > 30 eV and (b) ΔE k > 500 eV. Nano- each MCP count and the next laser pulse detected by an ava- ˆ ture S2 (w 500 nm). (c) SEM images of all four structures S1-S4 samples. The results are shown in Fig. 3 and compared to theo- x lanche photodiodestructure is measured edges by a time-to-digital are represented convertor by black lines. (c), (d) Calculations of used in this Letter.ˆ Their widths are w 250, 500, 750, and 1000 nm. x ˆ unit (TDC) and plottedaccelerated in a histogram. electron Here current the accelerated corresponding to the measurements shown retical calculations. The measured widths of the distribution electron signal (whichin (a), has (b) to be (color temporally scale, synchronized see text for with details) with contour lines representing agree well with theoretical values. The minimum observed the laser pulse) appears as a peak (for details see [18]). The period. In the case of a simple nano-grating [12], the longi- acquisition time inthe the amplitude presented measurements of the longitudinal was limited accelerating field component. width equals 650 50 nm. tudinal field component of the synchronous spatial harmonics  by the DC electron source to 30 s per 1 pixel. However, with decays exponentially with the distance from the grating surface, ∼ Apart from the excellent spatial resolution, this method also a laser triggered electron source working at MHz repetition rate and it can be approximated in the electron’s rest frame as offers femtosecond temporal resolution. The accelerated elec- and a single electron per bunch, the single measurement acquis- 2 2 2 E z y Ez0 exp −Γy cos ωt0 × exp − ve t − t0 ∕w ition time can beresponse decreased below function 1 s. is wres wmeas − we , where wmeas is tron current reflects both the oscillating laser field at the driving †ˆ † † f ‰ † Š g In the proof-of-principle DLA experiment with sub-ˆ × exp −2 ln 2 t − t ∕τ 2 ; (1) FWHM of the measured distribution obtained from the 2D laser frequency and the laser pulse envelope function [see f ‰ 0† Š g relativistic electrons [12], a rectangular grating was used. Its pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi where E 0 is the accelerating mode amplitude at the grating sur- width in the x-directionGaussian (perpendicular fit. The to both resolutions laser and elec- in the x- and y-directions differ, Eq. (1)]. The results of temporal characterization measure- face, y is the transverse distance from the structure surface, ω is tron beams) was much larger than both the laser and electron namely, wresx 600 nm, wresy 490 nm at ΔE k > 30 eV ments of near-fields excited by the thulium laser at structure the laser angular frequency, t0 is electron arrival time with beam dimensions. Therefore, the acceleratingˆ fields were effec- ˆ respect to the laser pulse, w is laser spot size, τ is laser pulse tively independentand of x. Tow achieve high340 spatial nm resolution, w also 450 nm at ΔE > 500 eV. S2 are shown in Fig. 4. Here we use two spatially separated resx ˆ resy ˆ k duration, Γ βγλ∕ 2π is the transverse decay constant, and in the x-direction,The the width spatial of the resolution grating wx is is limited improved to by only detecting electrons laser pulses (with a separation distance in the z-direction of γ 1 − β2 ˆ−1∕2 is the Lorentz† factor (for details see [17]). In sub-wavelength dimensions (w 250–1000 nm) here. The ˆ † x ˆ our experiment with sub-relativistic electrons (E k0 28.4 keV) resulting acceleratingwith field larger is thus a energy rapidly decaying gains. function This is the consequence of the spatial d 18μm corresponding to an electron travel time λ ˆ ˆ and near-infrared laser pulses ( ≅ 2 μm), Γ ≅ 100 nm. with the distance fromdecay the of surface the of accelerating the nanostructure field. in both The highest energy gain is obtained ttr 190 fs) with an adjustable time delay. The first pulse in- Whether the electron is accelerated or decelerated depends the x- and y-directions. Transverse spatial resolution is cha- troducesˆ an energy modulation to the electron beam, while the on its arrival time t0 with respect to the phase of the laser field racterized by 2Donly scanning by with electrons the structure transmitted in the x- and in the closest proximity of the nano- [factor cos ωt0 in (1)]. The strength of the interaction is then y-directions with respectstructure. to the electron However, beam focus. by improving The laser the resolution, the accelerated second pulse probes this modulation in time. The accelerated given by the temporal† overlap of the transmitted electron with focal position remains fixed during the scan, but since the focal the laser pulse envelope. dimensions (widthcurrentw and Rayleigh decreases length dueof laser to beam) the are narrowing of the real- and phase- electron current is then measured as a function of time delay The experimental setup for characterization of the spatial much larger thanspace both the volume electron beam occupied focal width bywe detectedand electrons. between the two laser pulses (more details in [9]). Figure 4(a) resolution of the proposed technique is shown in Fig. 1(a).A the scanned range, the incoming laser field can be treated as a Hitachi S-Series scanning electron microscope column (SEM) spatially infinite planeOur wave in experimental the x-direction. results are compared to numerical results shows the phase-dependent oscillations of the measured current is used as a DC electron source with an energy tunable in the In Figs. 2(a) andshown2(b) we in show Figs. the comparison2(c) and of 2D2(d) scans. The accelerating fields are calcu- range of E k0 3–30 keV and a spectral width of 3 eV. The with the electron energy gains ΔE k > 30 eV and ΔEk > 500 eV. electron energyˆ is set to match the synchronicity condition in The measured acceleratedlated using electron a current finite-difference distribution corre- time-domain (FDTD) technique each experiment. An electron beam with a current of I DC sponds to the convolution of the device spatial response func- 2 ˆ with a commercial software [19]. The amplitude of the accel- 3 1 pA is focused to a transverse radius (1∕e ) of we 70 tion given by the shape of the accelerating fields with the initial 20 nm and traverses the nanostructure. No damageˆ of the transverse electronerating beam density mode distribution. is obtained The FWHM of from the the spatial Fourier transform of the field amplitude along the trajectory of each electron [visualized by contour lines in Figs. 2(c) and 2(d)]. The transverse electron beam density is described as n x;y n exp −4 ln 2 x − x 2 y − y 2 ∕w2 , where n is†ˆ the 0 f ‰ 0† ‡ 0† Š e g 0 peak density, and x0, y0 are coordinates of the beam center. d ⃗ ⃗ The equation of motion dt γm0v⃗ e q E v⃗ e × B is nu- merically integrated for all electron†ˆ arrival ‡ times to† obtain the final electron energies. Here E⃗ and B⃗ are the amplitudes of the electric and magnetic fields of the synchronous spatial harmonics, m0 is electron rest mass, q is electron charge, v⃗ e is 2 −1∕2 electron velocity, and γ 1 − v⃗ e:v⃗ e∕c is the Lorentz factor. We neglect the quiver motionˆ of electrons† at the laser frequency and use only the synchronous spatial Fourier component of the Fig. 3. Measured accelerated electron current (points) for energy gains field. This approximation is valid when the maximum electron higher than 30 eV along the line y 50 nm above the structure surface ˆ velocity change during the interaction is small compared to its for structures S1-S4 with four different widths wx . Data are compared to initial velocity (< 1% in our case). By integrating over all arrival theoretical calculations (solid curves). Data are vertically shifted for clarity. times and counting only electrons with final ΔE k above a certain Inset: FWHM of the measured accelerated current distribution (points) threshold, the accelerated electron current is obtained. compared to calculations (curve) as a function of structure width wx . Time Domain Simulations

Rasmus Ischebeck > EPFL > Accelerator Research Uwe Niedermayer 48 Transverse Focusing

• Transverse confinement crucial for long DLA • Alternating Phase Focusing (APF)

Analytical and numerical (rms) beam envelopes, scaled to identical initial beam size at ε = 100 pm.

U. Niedermayer, 2018 https://arxiv.org/pdf/1806.07287.pdf

Rasmus Ischebeck > EPFL > Accelerator Research 49 Plasma Wakefield Acceleration

Rasmus Ischebeck > EPFL > Accelerator Research Rasmus Ischebeck > EPFL > Accelerator Research Taylor Ratlif What is the Role of the Ions?

They attract the plasma electrons back to the axis, which creates the high charge density behind the electron bunch

They have higher mass than the electrons, and therefore move only very slowly

The electrons scatter on the ions, which increases the emittance slightly

All of the above

Rasmus Ischebeck > EPFL > Accelerator Research 52 Plasma Wakes - Theory

– – – – + + + + + + + – • Unlike electromagnetic waves – + + – + – – – – – + + + – + – + in vacuum, plasma wakes can – + – – – + + – – – – + have a longitudinal electric field + + + – – – + + + + – – + – + – – – – – – + + – + + – – + + + + – + – + + + – + + – – + + – + – – + – + – + – – – + + + – + + + • Tajima & Dawson, – – + – + – + – – – + – + – + + PRL, 43, 267(1979) + – – + – – +

E E E E E E

• Linear plasma wake:

− Limit:

Rasmus Ischebeck > EPFL > Accelerator Research 53 Plasma Wakes - Theory

• Above this limit: non-linear wakes, “Blow-out regime” • Fields can be calculated only with numerical methods

• Typical wavelength: 50 µm • Accelerating fields up to 50 GV/m

Rasmus Ischebeck > EPFL > Accelerator Research Miaomiao Zhou 54 Plasma Wakes - Reality

Rasmus Ischebeck > EPFL > Accelerator Research 55 The Choice of the Driver for Plasma Wakefields

• The plasma wakefields can be excited by several means:

−Short high power laser pulses ! many places, soon DESY

−Short electron bunches ! SLAC, BNL, Frascati and soon DESY

−Short (and long) proton bunches ! AWAKE experiment at CERN

• Each method has its advantages and disadvantages.

• All must be explored to propose optimal solution for a given project (can also be combination of different technologies).

Rasmus Ischebeck > EPFL > Accelerator Research Ralph Aßmann !56 Energy Doubling

• Plasma length: 85 cm • Density: 2.7•1023 m−3 • Incoming energy: 42 GeV • Peak energy: 85±7 GeV

Rasmus Ischebeck > EPFL > Accelerator Research 57 Measurement of Electromagnetic Fields

Rasmus Ischebeck > EPFL > Accelerator Research Malte Kaluza 58 Measured Electromagnetic Fields

Rasmus Ischebeck > EPFL > Accelerator Research Malte Kaluza 59 4.25 GeV beams have been obtained from 9 cm plasma channel powered by 310 TW laser pulses (15 J)

Electron beam spectrum INF&RNO simulation* Angle (mrad) 1 2 3 4 5 Beam energy [GeV]

Exp. Sim. • Laser (E=15 J): Energy 4.25 GeV 4.5 GeV - Measured longitudinal profile (T0= 40 fs) ΔE/E 5% 3.2% - Measured far field mode (w0=53 μm) Charge ~20 pC 23 pC • Plasma: parabolic plasma channel (length 9 cm, 17 -3 Divergence 0.3 mrad 0.6 mrad n0~6x10 cm ) W.P. Leemans et al.,PRL 2014, in print Rasmus Ischebeck > EPFL > Accelerator Research Ralph Aßmann 60 1% Relative Energy Spread

Rasmus Ischebeck > EPFL > Accelerator Research V. Malka 61 Self-Modulation of the Proton Beam Direct Seeded Self-Modulation Result Laser 10m Rb plasma 1st IS nd p OTR/CTR 2 IS Proton IonizingSPS laser off: beam protons Laser Laser dump dump dump

In 2018: inject electrons here

First milestone reached! • Self-modulated proton bunches present over long time scale from seed point • Reproducibility of the self-modulated proton bunches process against bunch parameters variation • Phase stability essential for e- external injection

44 Rasmus Ischebeck > EPFL > Accelerator Research Edda Gschwendtner 62 RESEARCH LETTER Electron Acceleration

Electron source system Accelerated electrons on the scintillator screen

Laser beam 20-MeV radio- Radio-frequency gun frequency structure

Dipole Electron beam Dipole

10-m Rb plasma Proton beam Imaging station 1

OTR, CTR screens Rb fask Quadrupoles Dipole

Scintillator screen

Electron 8 8 spectrometer Electron bunch Long Proton microbunches 6 proton bunch 6 4 4 Laser 2 2 dump 0 0 Imaging –2 –2 x (mm) x (mm) station 2 –4 –4 Ionizing –6 laser pulse –6 Captured electrons –8 –8 –25 –20 –15 –10 –5 0 –25 –20 –15 –10 –5 0 (mm) (mm) Adli et al., Nature 561, 363 (2018). Rasmus Ischebeck > EPFL > Accelerator Research 63 Fig. 1 | Layout of the AWAKE experiment. The proton bunch and rubidium (pink) is supplied by two flasks at each end of the vapour source. laser pulse propagate from left to right across the image, through a The density is controlled by changing the temperature in these flasks and 10-m column of rubidium (Rb) vapour. This laser pulse (green, bottom a gradient may be introduced by changing their relative temperature. images) singly ionizes the rubidium to form a plasma (yellow), which Electrons (blue), generated using a radio-frequency source, propagate a then interacts with the proton bunch (red, bottom left image). This short distance behind the laser pulse and are injected into the wakefield by interaction modulates the long proton bunch into a series of microbunches crossing at an angle. Some of these electrons are captured in the wakefield (bottom right image), which drive a strong wakefield in the plasma. These and accelerated to high energies. The accelerated electron bunches are microbunches are millimetre-scale in the longitudinal direction (ξ) and focused and separated from the protons by the quadrupoles and dipole submillimetre-scale in the transverse (x) direction. The self-modulation magnet of the spectrometer (grey, right). These electrons interact with of the proton bunch is measured in imaging stations 1 and 2 and the a scintillating screen, creating a bright intensity spot (top right image), optical and coherent transition radiation (OTR, CTR) diagnostics. The allowing them to be imaged and their energy inferred from their position.

central propagation axis by transverse electric fields that are present approximately 1.4 T. A large triangular vacuum chamber sits in the cavity only when the proton bunch undergoes modulation in the plasma. of the dipole. This chamber is designed to keep accelerated electron Electron bunches with a charge of 656 ± 14 pC (where the uncer- bunches under vacuum while the magnetic field of the dipole induces tainty is the r.m.s.) are produced and accelerated to 18.84 ± 0.05 MeV an energy-dependent horizontal deflection in the bunch. Electrons (where the uncertainty is the standard error of the mean) in a radio- within a specific energy range then exit this vacuum chamber through frequency structure upstream of the vapour source32. These electrons a 2-mm-thick aluminium window and are incident on a 0.5-mm-thick are then transported along a beam line before being injected into gadolinium oxysulfide (Gd2O2S:Tb) scintillator screen (Fig. 1; blue, the vapour source. Magnets along the beam line are used to control right) attached to the exterior surface of the vacuum chamber. The the injection angle and focal point of the electrons. For the results proton bunch is not greatly affected by the spectrometer magnets, presented here, the electrons enter the plasma with a small vertical owing to its high momentum, and continues to the beam dump. The offset with respect to the proton bunch and a 200-ps delay with respect scintillating screen is 997 mm wide and 62 mm high with semi-circular to the ionizing laser pulse (Fig. 1, bottom left). The beams cross approx- ends. Light emitted from the scintillator screen is transported over a imately 2 m into the vapour source at a crossing angle of 1.2–2 mrad. distance of 17 m via three highly reflective optical-grade mirrors to Simulations show that electrons are captured in larger numbers and an intensified charge-coupled device (CCD) camera fitted with a lens accelerated to higher energies when injected off-axis rather than with a focal length of 400 mm. The camera and the final mirror of this collinearly with the proton bunch17. The normalized emittance of the optical line are housed in a dark room, which reduces ambient light witness electron beam at injection is approximately 11–14 mm mrad incident on the camera to negligible values. and its focal point is close to the entrance of the vapour source. The The energy of the accelerated electrons is inferred from their hori- delay of 200 ps corresponds to approximately 25 proton microbunches zontal position in the plane of the scintillator. The relationship between 14 −3 resonantly driving the wakefield at npe = 2 × 10 cm and 50 micro- this position and the energy of the electron is dependent on the strength 14 −3 bunches at npe = 7 × 10 cm . of the dipole, which can be varied from approximately 0.1 T to 1.4 T. A magnetic electron spectrometer (Fig. 1, right) enables measurement This position–energy relationship has been simulated using the Beam of the accelerated electron bunch33. Two quadrupole magnets are located Delivery Simulation (BDSIM) code34. The simulation tracks electrons 4.48 m and 4.98 m downstream of the exit iris of the vapour source of various energies through the spectrometer using measured and and focus the witness beam vertically and horizontally, respectively, simulated magnetic-field maps for the spectrometer dipole, as well to more easily identify a signal. These are followed by a 1-m-long as the relevant distances between components. The accuracy of the C-shaped electromagnetic dipole with a maximum magnetic field of magnetic-field maps, the precision of the distance measurements

364 | NAT U RE | VOL 561 | 20 SE P TEMBER 2018 © 2018 Springer Nature Limited. All rights reserved. Accelerator Research at EPFL/PSI

Rasmus Ischebeck > EPFL > Accelerator Research 64 Load Lock

• Allows sample change without breaking the accelerator vacuum

Rasmus Ischebeck > EPFL > Accelerator Research 65 Experiments with Ultra-Relativistic Beams

Laser Profile monitor

Quadrupole magnets

Electron beam from SwissFEL Accelerating Magnetic structure spectrometer

Paraboloid mirror

Rasmus Ischebeck > EPFL > Accelerator Research 66 Experiments with Ultra-Relativistic Beams

Rasmus Ischebeck > EPFL > Accelerator Research 67 Focusing of the Electron Beam

90 x 80 y 70

60

50

40 -function (m) -function

β 30

20

10

0 0 2 4 6 8 10 12 14 16 18 s [m]

0 2 4 6 8 10 12 14 16 18

Rasmus Ischebeck > EPFL > Accelerator Research EduardEduard Prat Prat 68 Design of Interaction Chamber

Rasmus Ischebeck > EPFL > Accelerator Research Adriano Zandonella, Goran Kotrle, Eugenio Ferrari 69 Fabrication of the Chamber

Rasmus Ischebeck > EPFL > Accelerator Research 70 Marie Siegler: In-Vacuum Microscope Optics Summer Internship

Microscope objective Tube lens (infinity corrected) +

MTF = image contrast / object contrast “Resolution of the detection system”

~ 5000 cycles/mm — ~ 200 nm

Rasmus Ischebeck > EPFL > Accelerator Research 71 Yelong Wei: Modeling of Collective Effects Internship Acceleration

100-period rod-grating with Bragg reflector walls

Electron bunch

Wavelength=150 μm

Gaussian THz pulse Wake fields

Rasmus Ischebeck > EPFL > Accelerator Research 72 Max Kellermeier: Modeling of Acceleration in Dielectric Lined Tube Semester Thesis

Rasmus Ischebeck > EPFL > Accelerator Research 73 Dominique Zehnder: Positioning of Quadrupole Magnets Individuelle Praktische Arbeit

Verwendeter Aufbau System Position X 206.2007 Y 145.1001 Anzahl Messungen 200

RMS X 9.99 µm

RMS Y 6.25 µm

Rasmus Ischebeck > EPFL > Accelerator Research 74 Max Kellermeier: Photonic Band Gap Structures Master’s Thesis

Rasmus Ischebeck > EPFL > Accelerator Research 75 Simona Borrelli: Measurement of Sub-µm Beam Size Master’s Thesis

• 300 nm wire made with lithographic methods

• 491 nm rms beam size

Borrelli et al., Communications Physics 1, 46 (2018)

Rasmus Ischebeck > EPFL > Accelerator Research 76 What Happens When a 300 MeV Electron Beam with a Total of 60 mJ Hits the Wire?

The electrons deposit so much energy that the wire melts

The electrons lose energy, but most of the secondary particles leave the wire. Only very little energy is deposited

The wire charges up and peels off the membrane

The electron beam is deflected by the wire

Rasmus Ischebeck > EPFL > Accelerator Research 77 Benedikt Hermann: Wake Field Acceleration PhD (ongoing)

• Dielectric structure, manufactured using a combined laser exposure – wet etching process

Rasmus Ischebeck > EPFL > Accelerator Research 78 Benedikt Hermann: Wake Field Acceleration PhD (ongoing)

Spectrometer setup TDC located before ACHIP chamber, not useable (streaked beam would be to large) ! Create streaking field by orbit-offset (2 mm) in two C-Bands

tail

head

Rasmus Ischebeck > EPFL > Accelerator Research 79 Benedikt Hermann: Wake Field Acceleration PhD (ongoing)

Energy difference vs “Time” (wakepotential):

CST Simulation

250 keV energy gain! (in 3 mm)

200 pC

Tim axis uncalibrated… • non-linear streak • t0, offset 6D beam dynamics simulation incl. C-band streak started

Rasmus Ischebeck > EPFL > Accelerator Research 80 Accelerator Research

• Laser-driven acceleration in microstructures • Plasma wakefield accelerators

• Need much more than accelerating structures! • Electron source • Beam focusing • Beam control • Instrumentation and feedbacks

• Possible additional components • Beam delivery and steering • X-Ray generation • Detectors…

Rasmus Ischebeck > EPFL > Accelerator Research Ainu Havukainen81 Questions?

Nick Veasey Thank You

• Thank you for Simulations, Measurements, Illustrations, Animations and Photos: • Download this talk in the file format • Ainu Havukainen • Martin Koźak of your choice: • Adriano Zandonella • Matthias Moser • https://ischebeck.net • Chris Sears • Max Kellermeier • Daniele Filippetto • Micha Dehler • More information on the Accelerator- • Dominique Zehnder • Nick Veasey • ebeam Technologies • Peter Hommelhoff on-a-Chip International Program at • Edda Gschwendtner • Paul Scherrer Institut PSI: • Edgar Peralta • Peyman Yousefi • http://achip.ch • Eduard Prat • Pietro Musumeci • Eugenio Ferrari • Ralph Aßmann • Franz Kärtner • Ronny Huang • Goran Kotrle • Simona Bettoni • Gordon and Betty Moore Foundation • Simona Borrelli • John Breuer • Stanford University • Joshua McNeur • Ulrich Lorenz • Ken Leedle • Uwe Niedermayer • Leonid Rivkin • Wikimedia Commons • Marcel Drabbels • Yelong Wei • Marco Calvi • Yves Bellouard • Marie Siegler

© 2018 Paul Scherrer Institut 83