Wake Fields Excited in a Micron-Scale Dielectric Rectangular Structure by a Train of Femtosecond Bunches T. C. Marshall1, J-M. Fang1, J. L. Hirshfield2'3, Changbiao Wang2, V. P. Tarakanov4, and S. Y. Park5 Applied Physics Department, Columbia University, New York, New York 10027, USA 2 Physics Department, Yale University, New Haven, CT 06520 USA f Omega-P, Inc., New Haven, CT 06520 USA 4 ITES, RAS, Moscow, Russia 5 Postech, Korea Abstract. We study the longitudinal wake field components which are induced in a rectangular, dielectric-lined structure having micron-scale dimensions by the passage of one or more charge bunches having femtosecond duration. The bunches would be obtained from a 500 MeV LACARA "chopper" which uses a TW optical wave from a CO2 laser [1]; the bunches are chopped from a macrobunch having duration ~1 psec obtained from a high brightness 500 MeV rf linac. The high intensity laser wave accomplishes the chopping of the macrobunch into slices which are roughly 10% of the 10.6 um radiation wavelength. These microbunches can be shaped into a rectangular cross section, approximately 10 um x 150 um in dimension, and will excite wake fields when injected into a rectangular dielectric wake field accelerating structure. We compute sample 3D wake fields, using the PIC code KARAT, as well as by means of an analytic method. The passage of just one pC bunch will set up a longitudinal wake field ~ 40 MeV/m, and a train of ten properly-timed such bunches can produce a cumulative wake field ~ 600 MeV/m. The choice of dimensions causes the wave solutions to approximate a single-mode excited by an infinitely-tall bunch in a 2D structure; a highly uniform longitudinal wake field in the crosssectional plane of the structure results, suitable for accelerating a correctly positioned "test bunch". KARAT includes the effect of interference between the Cerenkov radiation of the bunch with the transition radiation emitted as the bunch enters the structure. The wake field structure is several cm in length, and is both rigid and capable of microfabrication accuracy; it could accordingly be a reproducible module in a staged array. The stability of the bunches and the analytic formulation are dealt with in a companion paper [2]. INTRODUCTION It is possible that narrow, femtosecond duration, sheet-like bunches can be created and injected into an optical-scale dielectric-slab accelerator structure, which will allow generation therein of a very strong longitudinal accelerating electric field (~ 1 GV/m) [1]. In this paper we describe the wake field set up in a planar dielectric slab structure energized by a train of high energy microbunches spaced by 10.6 jim. The bunches are to be approximately 3.5 fsec in duration (-1 jim) and form a train of up to 30 bunches, each containing -1 pC. This dielectric wake field accelerator structure is a vacuum device that will pass a sheet beam having energy -500 MeV, approximately 10 jim x CP647, Advanced Accelerator Concepts: Tenth Workshop, edited by C. E. Clayton and P. Muggli © 2002 American Institute of Physics 0-7354-0102-0/02/$19.00 361 150 jim in transverse dimensions. The slab dielectric which lines the parallel planar walls of the structure would be a prepared microstructure a few microns in thickness. The small transverse dimension of the device permits a buildup of very high accelerating fields, limited only by the vacuum breakdown of dielectrics when exposed to ultra-short (fsec) pulses of electromagnetic energy. A slab structure has the advantage that it can pass more beam charge and has better stability to transverse beam deflection than a cylindrical structure of comparable dimension [2, 3], particularly if the ratio of the height to width is > 10. A schematic of a rectangular wake field accelerator structure is shown in Fig. 1. The dielectric-lined waveguide supports a longitudinal electric wake field induced by the passage of an electron bunch (the "drive bunch"). Phase velocities for the modes of a dielectric-lined waveguide can be less than the speed of light, so that Cerenkov radiation occurs, manifesting itself as a wake field that reflects periodically from the conducting wall and fills the waveguide behind the drive bunch. If a "test bunch" of lesser charge is injected at a suitable interval following the drive bunch, it can move in synchronism with the wake field and experience net acceleration. Thus no external radiation is required for acceleration. In a study of a high energy accelerator based on these principles [1], femtosecond bunches obtained from a SOOMeV rf linac followed by a LACARA chopper were used to excite the dielectric structure with drive bunches and provide the energy for accelerating test bunches. The LACARA system requires a TW-level CO2 laser, but uses it only for chopping the bunch train, not for acceleration. A schematic of this concept is shown in Fig. 2. A solenoidal magnetic field of 1.7 T is used to set up a gyro-resonant interaction between the particles and the laser wave. The laser fields cause the bunch of electrons to spiral around the longitudinal axis, so that the electrons fall onto an annular pattern at the beamstop. The electrons intercepting the top half of the beamstop in one laser period are shown in Fig. 3. Thus a small hole in the beamstop (also shown) will transmit a pulse of charge having duration of a few fsec, repeated every laser period (35.3 fsec). We estimate the quantity of charge in this pulse to be approximately 1 pC, derived from a InC macrobunch; approximately thirty microbunches are generated from each macrobunch. A microbunch transmitted through the hole in the beamstop is distorted into a rectangular cross section shape ~35 cm downstream by a quadrupole (Fig. 4). This rectangular cross section profile is maintained for several cm of axial travel [1] and determines the location of the dielectric wake field structure. The longitudinal spreading of the fsec bunch due to space charge or finite emittance is not expected to broaden the bunch for a distance of at least 1m from the beamstop. A high acceleration field (~1 GeV/m) can be built up in this dielectric structure by superimposing the wake field radiation of several bunches; such a large field in a dielectric structure is thought to be possible [4] because the dielectric is exposed to an intense field for only ~ 0.3 nsec. It is interesting to comment that similar acceleration gradients are expected [5] in a practical plasma acceleration scheme that could achieve GeV energy, yet the method we describe will enjoy the higher efficiency typical of the rf linac as compared with a laser power source. 362 FIGURE 1. Schematic of slab bunch within a planar optical wake field structure. Gaussian envelope X quadrupole m r beam envelope ^_-I-----I n typical trajectory ——— 5 m ——— superconducting magnet beam stop wakefield element FIGURE 2. Schematic of a LACARA-type accelerator used as a chopper [1] for bunches obtained from a 500 MeV rf linac. The magnetic field is 1.7 T, and it uses a 5 TW circularly polarized CO2 laser in a Gaussian beam with 266.7 cm Rayleigh range. The superconducting magnet lies between z = 60 cm to 560 cm; the beam stop extends from z = 620 cm to 625 cm; the quadrupole is located between z = 630 cm to 635 cm. 0.03 0.02 _ transmitted electrons aperture E o 0.01 -0.015 0 0.015 x (cm) FIGURE 3. Distribution of electrons hitting the LACARA beamstop (upper half plane shown). The aperture will transmit a pulse of current. 0.002 at z = 666 cm E o at z = 671 cm -0.002 -0.01 -0.005 0 _L 0.005 0.01 x (cm) FIGURE 4. Sheet-like transverse profile of a bunch ~35 cm downstream from the quadrupole. Two views are shown at different locations (note different scales for the x and y axes). The dielectric wake field accelerator module is to be located here. The quadrupole has a pole-face field of 1.05 T and a pole radius of 1 cm. Wake field structures are attractive because no external source of energy is used in the structure itself. The breakdown limit of the wake field structure is not determined by the slow filling time of the structure by electromagnetic energy, but rather by the much shorter time of the passing field pulses set up by the short microbunches. The proposed vacuum-based wake field structure is rigid geometrically and therefore should be a more reproducible and controllable element in a staged system than would be an array of pulsed plasma elements. Also, the structure is capable of microfabrication accuracy, an important consideration when staging a large number of modules. The use of rectangular cross section bunches and structures may result in a tolerably-weak transverse wake field instability which would overcome one of the chief obstacles to realization of a practical dielectric wake field accelerator [3]. A laser wake field accelerator such as we have described may combine the high efficiency of a microwave rf system (which produces the drive bunches) with the advantages of short bunch operation (large wake fields) permitted by the high power laser. Use of several holes in the beamstop will result in several beams of drive bunches, improving the utilization of energy provided by the rf linac. Because the fields travel at the same speed as the highly relativistic charges, slippage is not an important factor within the spatial scale of the wake field structure. We emphasize the difference between chopping and bunching. Chopping, as we have it here, consists of chopping out a slice of current from the bunch, leaving zero current on either side. The axial definition of the chopped bunch remains crisp for a considerable distance of travel. Bunching, such as is done in an optical IFEL (e.g. STELLA at ATF [6]), is a spatially-transient affair that occurs as a result of an energy modulation imposed upon the particles. Spatial bunching forms at a certain downstream position and is transient; thus it may be suitable for use in a single wake field module, but perhaps not an extended interaction such as staged wake field acceleration. 364 NUMERICAL SIMULATION USING KARAT Because of the good beam quality obtained using a 500 MeV bunch [1], as well as the fact that higher bunch energy is more appropriate for a series of wake field accelerator modules (fewer modules required), the theory of the wake field acceleration in a rectangular dielectric-lined module is carried out at 500 MeV energy. However, insofar as wake field acceleration itself is concerned, the choice is based upon beam quality. In Figure 1, we standardize the following parameters: H (height of structure and bunch) = 150 jim; 2a = 15 jim; 2b = 18.8 jim; dielectric constant = 3.0; the bunch width is 10 jim, length = 1 jim; bunch charge = 1 pC and energy = 500 MeV. The parameters are such that a micron-scale wake field period will be set up, but not necessarily at 10.6 jim (this can be obtained by readjusting the dielectric thickness). KARAT is a versatile, relativistic, fully electromagnetic ID, 2D, or 3D code based on the PIC method, developed by Vladimir Tarakanov [7]. Although the code permits a full computation of all the fields set up by the passing bunch, we shall show only a few examples involving the axial ( E z ) component. The first of these (Fig. 5) shows the wake field trailing a single 3D rectangular bunch charge, showing a powerful wakefield ~40 MV/m set up behind the traveling bunch; the period of this wake field is about 20 jim. (The color scale indicates the magnitude and sign of the Ez field; the wake field reverses sign [color] periodically.) The wake field period depends on the dimensions of the structure and the dielectric constant of the material. The fields are shown when the bunch has reached the right-hand side of the figure at z = 60 jim; the field is displayed on the mid-plane, and the two dielectric slabs are indicated by the boundaries at the top and bottom of the figure. Figure 6 shows the Ez wake field at z = 51 jim in the x-y plane; the field is relatively uniform in the cross section of the structure. We have found that smaller channel width is more desirable in terms of improved wake field amplitude and definition. The wake field patterns obtained here do not have the sharp localization found in a previous study [8]. The reason for this has to do with two conditions: first, the bunch length chosen here results from chopping and therefore is ~10% of the laser period; secondly, the structure contains a large channel bounded by thin slabs of dielectric. As a result, only the lower spatial modes are excited and there is no dispersion effect to complicate the timing of multiple bunches. Another advantage is that a test bunch undergoing acceleration in the wake field of a drive bunch can have a length ~ 1 fsec without experiencing appreciable longitudinal energy spread. The test bunch need not have a rectangular cross section; indeed it may be desirable that it is centered in the structure and have a width of the order of its height [9]. Stability issues will affect bunch transport if the injected bunch is not centered accurately [2]. An overall view of the 3D wake fields is shown in Figure 7 for two timed bunches, located at z = 60 and z = 40 jim. The wake field is displayed as a color pattern on 2D x-y sheets at selected axial locations. Qualitatively, from this figure it can be seen that the additional bunch causes approximately a doubling of the wake field intensity, showing that superposition of the separate bunch fields occurs. As one might expect, the wake field is rather uniform along the long (y) dimension, and appears to be of 365 satisfactory quality. quality. Thus, Thus,aatall tallrectangular rectangularstructure structurecan canbebewell-approximated well-approximatedbybya a satisfactory 2Dmodel, model,asasone onemay mayexpect. expect.InIncontrast, contrast,a astructure structure3030µm jimtall tallwas wasfound foundtotoproduce produce 2D wake field field that thatwas wasless lessuniform uniformininthe thex-y x-yplane, plane,and andtherefore thereforeisisless lessdesirable. desirable. aa wake Otherexamples exampleswere wererun run(e.g. (e.g.varying varyingbunch bunchlength, length,dielectric dielectriccoefficient), coefficient),but butgiven given Other limited space here, we believe this should clarify the understanding of the wake fields limited space here, we believe this should clarify the understanding of the wake fields in the 3D rectangular geometry. in the 3D rectangular geometry. Accordingtotothe thesimulation, simulation,the thewake wakefield fieldstructure structureappears appearsdiffuse diffuseininthe theregion region According near z = 0 (z < 30 jim). This is caused by the interference of the Cerenkov wake field near z = 0 ( z < 30 µm). This is caused by the interference of the Cerenkov wake field radiationwith withthe thetransition transitionradiation radiationthat thatisisemitted emittedwhen whenthe thebunch bunchemerges emergesatatz z= = radiation 00 [10]. ItItisisnot nottotobe befound foundwhen whenthe themathematical mathematicalproblem problemincludes includesthe theentire entirez-axis, z-axis, [10]. i.e., when whenthere thereisisno nopoint pointofofentry entryorordeparture departurefor forthe thebunch. bunch. This Thiseffect effectwas wasalso also i.e., observed earlier earlier inin another another KARAT KARATsimulation, simulation,for fora awake wakefield fieldstructure structurehaving having observed macroscopicdimensions dimensions[11]. [11]. (In (InRefs. Refs.[10] [10]and and[11], [11],the thecylindrical cylindricalcase caseisisstudied.) studied.) macroscopic m ), moving to to thethe right, shown in in thethe FIGURE FIGURE5.5. Wake Wakefield fieldfrom froma asingle singlebunch bunch(located (locatedatatz =z =6060 \i m), moving right, shown x-z areare x-zplane planeatatthe they-plane j-planelocated locatedhalfway halfwaybetween betweenthe thetop topand andthe thebottom. bottom.The Thedielectric dielectricboundaries boundaries the m sign from thehorizontal horizontallines linesatatxx==1.9 1.9and and16.9 16.9 \imm. .The Thewake wakefield fieldatatz =51 z =51 n mhas hasthetheopposite opposite sign fromthethe wake is is proportional to to thethe field wakefields fieldson oneither eitherside sideofofit,it,with withfield fieldnulls nullsininbetween. between.The Thegray grayscale scale proportional field intensity, m toto thethe intensity,and andthe themaximum maximumisis40 40MV/m MV/mininmagnitude. magnitude.Note Notethe thediffuse diffuseregion regionforforz <z <3030 \i mdue due interference interferenceofofthe thewake wakefield fieldand andtransition transitionradiation. radiation. 366 Wake field field in in the the x-y plane section section at at zzz = 51\im m... The The moving moving bunch bunch is is located x-y plane FIGURE 6. 6. Wake FIGURE section at == 51 51 m The moving bunch is located at at zz == Wake field in the plane x-y FIGURE 6. 60 m . Note the uniformity of the field. The dielectric boundaries are shown as vertical lines on left 60 n m . Note the uniformity of the field. The dielectric boundaries are shown as vertical lines on left 60 m . Note the uniformity of the field. The dielectric boundaries are shown as vertical lines on left and right. and right. and right. Sections located located at at various various z,z, showing showing the the wake wake fields fields in in the the x-y planeresulting resulting from from two two x-y plane FIGURE 7. 7. Sections FIGURE wake fields in the plane resulting from two FIGURE 7. bunches,moving movingto tothe theright. right. The The bunches bunches are are located located at at zzz == 60 60\im mand and zzz === 40 40\im m... Note Notethe theincrease increase bunches, m and 40 m Note the increase bunches, moving to the right. ingray grayscale scaleat atthe thelocation location of of the the second second bunch bunch (~ (~ 40 40\im m)) compared compared with with the theleading leadingbunch bunch(first (firstatat atthe the in m) compared with the leading bunch (first the in gray scale at the location right). Moving Moving from from right right to to left, left, relative relative to to the the first first panel, panel, the the second second wake wake field field isis is in in the the opposite opposite right). Moving from right left, right). second wake field in the opposite direction; the the fourth fourth panel panel has has the the same same wake wake field field direction direction as as the the first first one, one, and and the the sixth sixth has has the the direction; the fourth panel has direction; one, and the sixth has the wakefieldin inthe thesame samedirection directionas asthe thesecond secondpanel. panel. wakefield in the same direction the second wakefield 367 RESULTS FROM AN ANALYTIC TREATMENT In addition to finding analytic solutions for the wake fields themselves, analytic formulas have been derived for the forces on a test particle moving in the wake of other particles [2]. These solutions are needed to explore questions of stability. The mathematical development is presented in the companion paper [2], which examines stability issues. In this paper, we quote a few computational results which may be compared with KARAT. Although it is appealing to describe fields in dielectric-lined rectangular waveguides in terms of well-known LSE and LSM modes [12], their use for calculation of wake fields has been found to be restricted [2]. This is because the LSE and LSM families of modes each possess an axial electric field component Ez, and have their symmetry axis normal to the dielectric surface. Moreover, both families are excited by the motion of a charge bunch, so ten field components would then be invoked in the solution, although only six of these are independent. Therefore, a different basis has been used in the development of the new analytical theory, in which each family of modes has all six independent components. In general, four such mode families must be considered, corresponding to symmetric and anti-symmetric solutions in both transverse coordinates x and y\ all four families can be excited by nonsymmetric beams, where stability issues in particular need to be examined. Orthonormalization constants have been found for all four families, following procedures established when deriving wake fields in dielectric-lined cylindrical waveguides. The theory provides the amplitudes of all the modes, and compact formulas have been found for the longitudinal and transverse forces on test particles. Some examples of the calculated longitudinal force Fz = -qEz on a test electron are shown in Fig. 8. For these examples, one to ten 1-pC sheet bunches each with width, height and length of 10, 150, and 1 jim respectively are sequenced periodically and injected symmetrically into the planar dielectric structure. The bunch spacing is chosen to be equal to the fundamental wake field period of 20 jim, so that cumulative build-up of the field from successive bunches can occur. It is seen that the calculated peak wake field amplitude of 40 MV/m from a single bunch is in good agreement with the value found in the KARAT simulations. Uniform build-up of the maximum wake field to nearly 600 MV/m from ten bunches is seen in Fig. 8c. Also to be noted is a long-period modulation of the wake field, caused by interference with an adjacent mode having slightly higher frequency and about 1/3 the amplitude of the dominant mode. These results affirm the expectation that injection of a train of pC sheet bunches into a micron-scale dielectric wake field structure can lead to accelerating fields that approach 1 GV/m. (For example, a Ipsec macrobunch obtained from the linac can be chopped into ~30 microbunches, and the lack of dispersion in the rectangular structure will permit a nearly-linear superposition of the individual wake fields.) In the companion paper [2], it is found that the transverse forces are modest, and would permit an off-centered bunch to propagate without external guiding about 7cm before hitting the dielectric wall. The instability is much weaker than for a bunch moving along a cylindrical structure, for the same choice of narrow gap dimension. 368 100 200 300 400 500 600 700 800 z (microns) FIGURE 8a. Axial force behind a single bunch, located at z = 800 (im and moving to the right. 100 200 300 400 500 iOO 700 800 z (microns) FIGURE 8b. Same as Fig.Sa, but for two bunches spaced by 20 (im. 100 200 300 400 500 600 700 800 FIGURE 8c. Same as Fig. 8a,b, but for ten bunches spaced by 20 (im. Note expanded scale. 369 The wake field structure would be about 10 cm in length, and would be canted so that a bunch will travel along its axis without grazing the walls. It can be pumped along its open sides, and unlike the numerical case studied above, these would include the top and bottom surfaces. The structure may be prepared by micro-fabrication. The question of the dielectric material is still open: the dielectric constant we used is compatible with a choice of NaCl or BaF2, both of which have very low losses in the IR where the wake field modes are set up. Other materials such as silicon could prove more versatile; indeed this would permit doping with impurities which might absorb away any unwanted modes. ACKNOWLEDGMENT This work was supported by the DoE, High Energy Physics Division. REFERENCES 1. Marshall, T. C, Wang, C, and Hirshfield, J.L., Phys, Rev, ST- Accel and Beams 4, 121301 (2001). 2. Park, S-Y, Wang, C., and Hirshfield, J. L.: see this volume. 3. Tremaine, A., Rosenzweig, J., and Schoessow, P., Phys, Rev, E 56, 7204-7216 (1997). 4. Sprangle, P., Hafizi, B., and Hubbard, R. R, Phys, Rev. E 55, 5964-5975 (1997). 5. Sprangle. P., et al, Phys, Rev, E 63, 056405 (2001). 6. Liu, Y., et al., Phys, Rev, Lett. 80, 4418-4421 (1998). 7. Tarakanov, V. P., "Users' Manual for Code KARAT", BRA Inc., VA, USA (1992). 8. Zhang, T-B., Hirshfield, J.L., Marshall, T.C., and Hafizi, B., Phys. Rev. E 56, 4647-4655 (1997). 9. Xiao, L., Gai, Wei, and Sun, Xiang, Phys. Rev. E 65, 016505 (2002). 10. Onishchenko, I.N., Sidorenko, D. Yu, and Sotnikov, G. V., Phys, Rev. E 65, 066501 (2002). 11. Marshall, T.C., Fang, J-M., Hirshfield, J. L., and Park, S-Y, "Multi-Mode, Multi-Bunch Dielectric Wake Field Resonator Accelerator" in Advanced Accelerator Concepts: Ninth Workshop, edited by P.L. Colestock and S. Kelley, AIP Conference Proceedings 569, 2000, pp. 316-325. 12. Collin, R. E., Field Theory of Guided Waves, 2nd ed. (IEEE, New York, 1991). 370
© Copyright 2025 Paperzz