Wakefield Calculation for Superconducting TM110 Cavity Without Azimuthally Symmetry
Leo Bellantoni Fermi National Accelerator Lab Batavia IL 60510 U.S.A.
Draft of April 26, 2018
The physical problem 1 The 3.9GHz TM110 mode deflecting cavity developed at FNAL for use in an RF- separated kaon beamline is also useful as a longitudinal bunch profile diagnostic, as a crab cavity candidate for the ILC2, and in light sources3. Unlike the kaon separator application, these uses involve beams with substantial time structure, and the impact of RF energy left in the cavities by such a beam needs to be evaluated.
At frequencies above the cutoff frequency for the beampipe, energy may propagate out the beamline. It need not do so however; there can be energy stored in parts of the cavity that are relatively distant from the beampipe and this energy (called “trapped modes”) will stay in the cavity for long periods of time. The basic analysis strategy is to handle modes under the cutoff frequencies in a semi-analytic frequency domain method, and to confirm that these are indeed the dominant contributions by doing a numeric time domain calculation.
When used as a bunch profile diagnostic, the cavity will reside in a beamline that is used for a number of beam physics studies; in this case the real problem is to know what the cavity will do to the beam when unpowered and left at 4.15K. To this end, the calculation allows for the inclusion of an estimate of the surface resistance at various frequencies at 4.15K. The power loss is estimated from the geometry constant G = Q0 RSURF where RSURF has a frequency-squared dependence. The surface resistance at 4.15K and 3.9GHz is a free parameter; it is set to 6370n, (as determined by fits to DESY data from 1.3GHz TM010 cavities) by default.
The default beam parameters are taken as typical-to-aggressive values for the A0 photoinjector at FNAL: 1MHz bunch spacing, 12nC bunches of 3ps length.
1 FNAL Technical Memos 2060, 2144; N.Solyak, L.Bellantoni et.al. at LINAC 2004, Lubeck Germany, Aug 2004. 2 P.Goudket, G.Burt at Snowmass 2005; C. Adolphsen, private communication. 3 G.Waldschmidt, private communication; A.Zholents, Advanced Photon Source Strategic Planning Meeting, August 30, 2004. The upgraded beam energy of 40MV is taken as the default, and the calculation allows for up to 60 bunches in a train. This last number is larger than is actually used but it clarifies some of the resulting plots. There is a full second between bunch trains, and the transverse spot size is about a millimeter.
The first mathematical problem The problem is to find the interaction of a bunch train upon a following bunch as mediated through the fields inside the cavity that are below the cutoff frequency of the beampipe. The beam has velocity c zˆ , i.e. is exactly parallel the longitudinal (z) axis of the cavity. Space charge effects are negligible. The longitudinal extent of the cavity is L, in the sense that electric boundaries at z = L will not perturb the fields of the resonant modes. With this, integrals over z
from –L to +L can be replaced with integrals from – to + exciting bunch train and the trailing bunch are allowed to enter the cavity at the same (r,) value. All the exciting bunches are taken to enter at the same position with the same charge and with uniform spacing. The trailing bunch has the samer charge and timing. (For most of the calculation we will carry the positions r = (r,f ) and r 1 1 1 r2 = (r2,f 2) as independent variables, along with the charges q1 and q2, and call the interbunch distance s.) The charge of the bunch makes a negligible contribution to the electric field, i.e., ~ 0, and the electric field is generated r only by the time derivative of the magnetic potential A .
The electric boundary condition is CAVITY WALL = constant;r the constant is here set to zero. In MKSA units with the Coulomb gauge A = 0, the familiar Maxwell С Ч equations are 2 r r м 2 1 ¶ ь 1 ¶ 2 r н- С + 2 2 э A = m0J - 2 СF - С F = c ¶t c ¶t e0 о ю
where r ¶ r r r E = - A - СF B = С ҙ A . ¶t
The wake potential is defined as
+ r r r 1 Ҙ r r r r W (r1 ,r2 ,s) = dz[E (r2 ,z,t) + czˆ ҙ B (r2 ,z,t)] . q т 1 -Ҙ t=(s+z)/ c r r r = r It is the forcer uponr a trailing bunch entering at 2 created by a leading bunch entering at r = r1 and s meters ahead of the trailing bunch, normalized to the
charges q1 and q2.
The first appendix gives the derivation of the following results:
r r r r 1 Vm (r1 )Vm (r2 ) жw sц 1) W r ,r ,s = cos m // ( 1 2 ) е з ч 2 m Um и c ш
2 r r 2 r r 2) С ^ (1) W// (r1 ,r2 ,s) = С ^ (2) W// (r1 ,r2 ,s) = 0
r r 3) W r ,r ,s = W (n) s rm rm cos m f - f cos m f - f // ( 1 2 ) е // ( ) 1 2 [ ( 1 n )] [ ( 2 n )] n r r r s r r 4) W (r ,r ,s) = - ds W (r ,r ,s ) ^ 1 2 С ^ (2) т // 1 2
where W// and W are the longitudinal and transverse components of the wakefield, sums over m and n are sums over all modes (regardless of frequency degeneracies) below cutoff. The convention is that (as there are multiple modes with the same azimuthal order) if the sum over modes is indexed with n, the azimuthal order of mode n is given by m. A specific set of eigensolutions are
required, and Um and m are the cavity energies and angular frequencies for (n) these modes. Equation (3) is an expansion theorum; W// and n are the coefficients of the expansion. It would take a somewhat different form in the case where the cavity does have azimuthal symmetry. The two dimensional r gradient operator with respect to ri is С ^(i) . The longitudinal integrated + r Ҙ r r V r ә dx - iw zˆ · A r ,z eiw mx c accelerating voltage is m ( ) т [ m m ( ) ], where the origin of -Ҙ the axis is selected so as to make this quantity real; that phases the bunch to the cavity field.
To write the specific expansion coefficients for equation (3), define
2 r m (n ) 1 2 {Vn (r ) r cos[m(f - f m )]} R ә . Q 2m r w 4Un r From equations (1) and (3) directly, we have V (r ) ө rm cos[m(f - f )], and so r m m R(n)/Q is independent of r . The value of this definition is that these numbers can
be computed with a finite-element analysis.
Introducing a term to allow for exponential damping through e.g., external
couplers,
r r м (n) ь R жw n sц ж s ц m m W// (r1 ,r2 ,s) = н w n э cosз чexp з- ч r1 r2 cos[m(f 1 - f n )] cos[m(f 2 - f n )] е Q c ct n n о ю и ш и ш
and then the multipole expansion of equation (3) shows
R(n) w s W (n) s = w cosж n цexpж- s ц // ( ) n з ч з ct ч. Q и c ш и n ш
For the transverse wake potential, apply equation (4), obtaining r r r W ^ (r1 ,r2 ,s) = s мR (n) ь w s ж n ц ж s ц m m - ds н w n э cosз чexp з- ч С ^ (2){r1 r2 cos[m(f 1 - f n )] cos[m(f 2 - f n )]}. т е Q c ct n n о ю и ш и ш The integral over d is simplified by n>> 1/n, leaving
r r r мR (n) ь w s ж n ц ж s ц m W ^ (r1 ,r2 ,s) = - н cэ sinз чexp з- ч r1 cos[m(f 1 - f n )] ҙ е ct n n о Q ю и c ш и ш m- 1 ˆ m r2 {rˆ cos[m(f 2 - f n )] - f sin[m(f 2 - f n )]}.
The multipole expansion is then r r r r W r ,r ,s = W (n) s rm m rm- 1 cos m f - f ^ ( 1 2 ) е ^ ( ) 1 ( 2 ) [ ( 1 n )], n
(the factor of m means there is no transverse monopole term) with the expansion coefficient defined as
r R(n) w s W (n) s = c sinж n цexpж- s ц fˆ sin m f - f - rˆ cos m f - f . ^ ( ) з ч з ct ч{ [ ( 2 n )] [ ( 2 n )]} Q и c ш и n ш
This definition of the expansion coefficient permits also the relation
r s W (n) s = - ds W (n) s ^ ( ) т // ( ) .
With these expansions of the wakefield potential in terms of numbers than can be computed with a finite-element analysis, it is possible to compute the impact of the wakefields on a trailing bunch. The energy injected into the trailing bunch is the integral as the bunch passes through the cavity of the longitudinal force of the +Ҙ r r r r field: DE = q1 dzq2(E + czˆ ҙ B ) = q1q2W// (r1 ,r2 ,s). For the problem at hand, q = q1 = т z r r -Ҙr q2 and r = r1 = r2 , but we must sum over all the previous bunches in the train, spaced at distances si from the trailing bunch:
DE = ж ц q2 W (n )(s ) + W (n )(s )r2 cos2 (f - f ) + W (n )(s )r4 cos2 [2(f - f )] + ... . е з е // i е // i n е // i n ч bunchesиn monopoles n dipoles n quadrupoles ш r r r r q q W (r ,r ,s) Similarly, angular deflection is given by DQ = 1 2 ^ 1 2 , summed over the E 0 previous exciting bunches:
r 2 ж r r ц q (n ) (n ) 3 DQ = з W ^ (si)r cos(f - f n ) + W ^ (si ) 2r cos[2(f - f n )] + ...ч. E е з е е ч 0 bunchesиn dipoles n quadrupoles ш
Mode, R(n)/Q spectra
The values of the R(n)/Q quantities were computed with MAFIA, as follows.
First I found the spectrum of modes with periodic boundary conditions for a single cell with phase advances of 0, /4, /2, 3/4 and , up to 16GHz. The beampipe cutoffs are
4.880 GHz for TE11 evanescent wave in beampipe
6.375 GHz for TM01 "
8.096 GHz for TE21 "
10.157 GHz for TM11 "
11.137 GHz for TE31 "
14.132GHz for TE12 "
so 16GHz is 2 to 3 times the cutoffs for m = 0,1,2. Figures 1 through 3 show the dispersion curves for these three azimuthal numbers; for m = 0, only TM modes are shown, as TE modes all have zero R(n)/Q anyway. 16000.00
14000.00
12000.00
10000.00
8000.00
6000.00
4000.00
2000.00
0.00 0 13.8 27.7 41.5 55.4 69.2 83.1 96.9 111 125 138 152 166 180
Figure 1. MAFIA periodic-boundary simulation for m = 0 modes.
16000.00
14000.00
12000.00
10000.00
8000.00
6000.00
4000.00
2000.00
0.00 0 13.8 27.7 41.5 55.4 69.2 83.1 96.9 111 125 138 152 166 180 Figure 2. MAFIA periodic-boundary simulation for m = 1 modes.
These curves are useful indicators of where one might expect to find problem modes; when the dispersion curve corresponds to modes with a phase velocity 16000.00
14000.00
12000.00
10000.00
8000.00
6000.00
4000.00
2000.00 0 13.8 27.7 41.5 55.4 69.2 83.1 96.9 111 125 138 152 166 180 TM010 TM020 TM011 TMwierd TM021 TMwierd TM031 TMwierd TM023 TM041 TMwierd TMwierd Light Cone . . . .
Figure 3. MAFIA periodic-boundary simulation for m = 2 modes. close to the speed of the bunches, the beam can strongly excite the cavity and vice versa.
To find R(n)/Q values the cavity with 0.5m beampipes was simulated using both EE and BB boundary conditions. To include the intrinsic damping from surface resistance, G values were also tabulated. For m=0, modes up to 12GHz were found; beyond that the band structure of the periodic single simulation was gone. For m=1, modes up to 5GHz were found; beyond that the energy went out the beampipe very quickly. For m=2, modes up to 10GHz were found; beyond that the mode structure was murky and there was a lot of beampipe energy in the solutions. For m=2 over 9GHz, there were some cases where repeat calculations with different frequency spans and number of requested solutions did not converge to the exact same values.
To identify and discard modes where there is a lot of energy in the beampipe, a plot of a simulated beadpull of a round metal bead, 5mm off axis at 22.5 degrees was made for each mode. This is used to determine how much energy was in the beampipe and to select the numbers from the EE or BB boundary condition which has less energy in the pipe. The two solutions were also compared vis-à- vis the computed frequencies. The requirement that EE and BB boundary conditions give the same frequencies to less than 0.05% was met for modes below 6183.97MHz for m=0, 4829.32MHz for m=1, 8220.46MHz and for or m=2. As a rule, some subjective judgement was involved in interpreting MAFIA solutions over these frequencies. The software contains a flag to determine if these modes are to be included. Figure 4 shows the resulting values of R(n)/Q, in units of for m = 0, /m2 for m = 1, and /m4 for m = 2. Disturbingly, R/Q values for the dipole bands show an upward tendency with frequency. Comparison with an HFSS calculation not reported on here4 revealed that with such closely spaced modes in so many dispersion curves, having the correct field flatness (as was ensured for these MAFIA runs and meshings) will have a large effect upon R(n)/Q values. Modes with low R(n)/Q do not have low R(n)/Q because there is little beam-cavity coupling in each cell. Rather, the beam-cavity coupling is the sum of beam-cell couplings that cancel. The R(n)/Q numbers are sensitive to field flatness because if the cells are slightly different, the fine-tuned cancellation falls apart. But in the worst case, R(n)/Q in any such mode cannot be as bad as the worst R(n)/Q for that band.
R (m)/Q in MAFIA for CKM 13 cell
1.0E+12
) 1.0E+08 (2m) /m W1.0E+04 ( Q / (m) R1.0E+00
1.0E-04 0 2000 4000 6000 8000 10000 12000 14000 m=0 modes m=1 modes m=2 modes f (MHz)
(m) Figure 4. R /Q vs. f in MAFIA for 13 cell TM110 mode cavity.
The finite element calculation was done in 2.5 dimensions, i.e., the mesh extended over the cylindrical coordinates r,z and the dependence was assumed to be sinusoidal, as it in fact is for an azimuthally symmetric cavity. A free parameter for the mode splitting, with a default value of 10MHz was used for the dipole band. An ad-hoc formulation was made for the R(n)/Q values in this case; the additional frequency change was treated as being the same as using a beam with velocity slightly less than one. So e.g., for the TM110 mode, a 10MHz polarization shifts the frequency to 3910MHz, and a bunch that would have traversed the entire 13 cell cavity in (13/2) times the base period of 256.4ps will instead have traversed (3900/3910) = 99.74% of the cavity in (13/2) times 256.4ps. In effect, the beam has = 0.9974.
4 I. Gonin, private comunication. For the quadrupole bands, no assesment of mode splitting was made, but the mades were taken to be pinned to the cavities’ symmetry-breaking mechanical features.
No attempt to model manufacturing defects has been made.
Results of the first calculation:
A spreadsheet (Wakes.xls) which uses the formulas above, and the outputs of the MAFIA runs described above to determine the impact of the wakefields on the beam is available. Beam parameters are adjustable inputs, including the position of the beam relative to the cavity center. Also adjustable are the values of QEXT from damping couplers, and the surface resistance at 4.15K for 3.9GHz. The latter is used to determine a contribution to the damping from surface resistance, should a flag to do so be set. As discussed above, there is a flag to determine whether or not one should include modes that are above cutoff, but not so far above cutoff that with some subjective element entering into the matter, reasonable guesses as to the correct R(n)/Q values might be made. r The spreadsheet calculates E and D Q as a function of bunch number in the train. Beam breakup condition, where the magnitude increases without limit with bunch number, is readily identified.
The first appendix The following derivation follows that of Wanzenberg and Weiland5. The basic procedure is to: r a. Expand the magnetic potential A into a set of orthogonal functions that are the eigenfunctions for the infinite-conductivity limit of the empty cavity (i.e., sans beam).
5 T.Weiland, R.Wanzenberg, in “Frontiers of Particle Beams: Intensity Limitations”, the joint US-CERN Particle Accelerator School, Hilton Head S.C. U.S.A., 1990; and R.Wanzenberg, TESLA note 2001-33. See also appendix A of FNAL TM 2144. b. Write the magnetic potential, the beam current, and ultimately the electric field, in the frequency domain. c. Write the wake potential in terms of the frequency domain electric field and simplify it. This will give equation (1). Equations (2) through (4) will be shown subsequently.
To find the set of orthogonal functions that are the eigenfunctionsr for the infinite- conductivity limit of the empty cavity, note that when J = 0, the vector potential 2 r м 2 1 ¶ ь satisfies the wave equation - 2 A = 0, which responds well to the нС c 2 э о ¶t ю method of separation of variables. So each of the m solutions are r r r r iw m t A m (r ,z,t) = a m (r ,z)e , with the convention that m may take negative values, making it unnecessary to write e- iw m t solutions explicitly. In the frequency domain
2 r r w 2 r r the wave equation is just С a (r ,z) = - ( 2)a (r ,z). c
In any expansion, there must be some orthonormality condition. In this case the energy stored in the cavity for a given mode provides the normalization constant. For any specific mode m with a corresponding given field level, the energy in the cavity is r r e0 2r * r r Um = d r dz{E m (r ,z,t)· E m (r ,z,t)} 2 ттт t= t0 cavity volume
where t = t0 means that the electric field for that mode is to be evaluated at the time when the magnetic field (which is 90 out of phase) is zero. Each eigensolution may also be scaled by some arbitrary constant, thereby changing Um, but here we just assert that some particular scaling has been chosen and do not have to belabor the issue further. Then the normalization is r r e0 2r * r r Um = d r dz{E m (r ,z,t)· E m (r ,z,t)} ттт t= t0 2 cavity r r r * r r e0 2 м ¶ iw m t ь м ¶ iw m t ь = d r dzн- [a m (r ,z)e ]э · н- [am (r ,z)e ]э 2 ттт ¶t ¶t cavity о ю о юt= t0 e r r r r r = 0 w 2 d 2r dz a * (r ,z)· a (r ,z). 2 m ттт m m cavity
Next, expand the vector potential in these modes. If we allow the coefficients of
the expansion (t) to depend on time, it will be possible to allow for phenomena such as ring-down caused by external couplers: r r r r A r ,z,t = a t a r ,z eiw m t ( ) е m ( ) m ( ) . m
This being the first time we expand in modes, it is worth belaboring what the sum over m includes. It includes in principle a full set of eigenfunctions up to infinite frequency; that would be both a discrete spectrum at lower frequencies (f < a few times c/, where is the size of the largest aperture of the structure) and a continuum at higher frequencies. There is also the possibility of a nearly discrete spectrum at higher frequencies that have a relatively narrow width because the spatial characteristics of the mode prevent its rapid dissipation out through the cavity apertures. These “trapped modes” and high frequency continuum are ignored in this treatment. Their effects can be understood with time domain modeling. The sum over m includes cases of frequency degeneracy. In particular, for a pillbox-like cavity, there will be two dipole modes of each type. In a real cavity, manufacturing asymmetries will give them slightly different frequencies; in the cavity studied here, an intentional azimuthal asymmetry separates the modes by some 10 – 20MHz. Both modes are included in the sum. Similarly, quadrupole modes have a two-fold degeneracy. Unlike the situation when we take the Fourier transform of quantities such as the beam’s time structure, we do not have to include modes of negativer frequencies. Here we are generalizing rthe real field quantity denoted by A into a complex quantity cleverly denoted by A , with the intent of later taking the real part. For real numbers a and b, the real part of aeit + be-it is just (a+b)eit, and so we may get by with only positive frequencies.
r Insert the expansion into the inhomogeneous Maxwell equation for A (neglecting the space charge termr r ) and then apply the frequency domain wave equation for each particular a m (r ,z):
r r 2 r r 1 й 2 ¶ щ iw m t J (r ,z,t) = w m + ( a m (t)e am (r ,z)). e е к ¶t 2 ъ 0 m л ы
Apply the orthonormality condition 1 r r r r r r r r й ¶2 щ r r d 2r dz a * (r ,z)· J (r ,z,t) = d 2r dz a * (r ,z)· w 2 + a (t)eiw m t a (r ,z) ттт n ттт n е к m 2 ъ( m m ) e0 cavity cavity m л ¶t ы й ¶2 щ r r r r r = w 2 + a (t)eiw m t d 2r dz a * (r ,z)· a (r ,z) е к m 2 ъ m ттт n m m л ¶t ы cavity й ¶2 щ 2U = w 2 + a (t)eiw m t d m е к m 2 ъ m mn 2 m л ¶t ы e0wm 2 2Un й 2 ¶ щ iw n t = w n + ( a n (t)e ) e w 2 е к ¶t 2 ъ 0 n m л ы
+Ҙ - iwt ˜ and use the Fourier transform F [ f (t)] = т f (t)e dt = f (w). Since -Ҙ
¶ r r iw n t ˜ F [ f (t)] = (i) F [f (t)] and F [e f (t)] = f (w - w n ), the Fourier transform of ¶t r iw n t a n (t) e is given by 2 r r r r 1 ж w n ц 2 * ˜ a˜ n (w - w n ) = d r dz an (r ,z)· J (r ,z,w) 2U зw 2 - w 2 чттт n и n шcavity
˜ r where J (r ,z,w) is a vector quantity, the current density after a Fourier transform in the time but not the spatial coordinates. The similarly transformed vector r r r A˜ r ,z,w = a˜ w - w a r ,z potential is ( ) е m ( m ) m ( ). The beam current produced by the r m 2 r r exciting beam is J = c q d(z - ct)d (r - r )zˆ , which has Fourier transform r r 1 r 1 ˜ - iw z / c 2 ˆ J (r ,z,w) = q1 e d (r - r1 )z . Now we can write the electric field in the excited cavity in the frequency domain: r r E˜ (r ,z,w) = - (iw)A˜ (r ,z,w) r r = - iw a˜ w - w a r ,z е m ( m ) m ( ) m 2 й r r r r щr r ж- iw цж w m ц 2 * ˜ = 2 2 к d r dx am (r ,x )· J (r ,x,w)ъa m (r ,z) е з2U чзw - w ч ттт m и m ши m шлкcavity ыъ 2 й щ ж- iq1 ц жw цж w m ц 2 r r * r - iwx / c 2 r r r r = к d r dx am (r ,x )· zˆ e d (r - r1 )ъa m (r ,z) з 2 че зU чзw 2 - w 2 ч ттт и шm и m ши m шлкcavity ыъ r r 2 + Ҙ * жq 1 ц жam (r ,z)цж w m ц r r iwx / c = dx [- iw zˆ · am (r1 ,x ) e ] з 2 че з U чзw 2 - w 2 ч т и шm и m ши m ш -Ҙ
A similar expression for the magnetic field is not needed, because the Panofsky- Wenzel theorum, equation (4), lets us express all the desired quantities for the stiff beam case with just the z component of the electric field. r In the stiff beam approximation, zˆ · zˆ B = 0, so ( ҙ )
r r 1 +Ҙ r W (r ,r ,s) = dz E (r ,z,t = (s + z) c) // 1 2 q т z 2 1 -Ҙ r r +Ҙ +Ҙ 2 1 dw iwt жq ц жzˆ · am (r2 ,z)цж w ц = dz e 1 m q т т2p з2 че з U чзw 2 - w 2 ч 1 -Ҙ -Ҙ и шm и m ши m ш + Ҙ r r * dx - iw zˆ · a r ,x eiwx / c т [ m ( 1 ) ] -Ҙ +Ҙ 2 1 1 iw m iw s c = dw 2 2 e 4p е Um т m -Ҙ w(w - w m ) + + Ҙ r r Ҙ r r * dz - iw zˆ · a r ,z eiw z c dx - iw zˆ · a r ,x eiwx / c т [ m ( 2 ) ] т [ m ( 1 ) ] -Ҙ -Ҙ
Now let us do the integral over , the frequency spectrum of the beam. There
are two poles, at m. We’ll do a loop in the complex plane, as shown in
Figure A-1.
Although we write the integrals over z and as over , the contributions occur within the finite range L. Therefore, for s >> 0, the return path that contributes negligibly to the integral is in the upper half-plane; in this case, by causality the integrand must be non-zero. For s << 0, the return path is in the lower half-plane; in this case, by causality the integrand must be zero. For s within L, the upper/lower assignment is indefinite as it depends on the sign of s+z+. But we may insist that the contributions to the integrals over z and be made so as to keep (s+) = 0, so the causality may be enforced for all s by use of the contour shown.
Figure A-1.
For this contour, enclosing two first-order poles, + 2 + + Ҙ - iww Ҙ r r Ҙ r r * dw m eiw s c dz zˆ · a (r ,z) eiw z c dx zˆ · a (r ,x ) eiwx / c т 2 2 т [ m 2 ] т [ m 1 ] -Ҙ (w - w m ) -Ҙ -Ҙ = 2pi Rs w = +w + Rs w = - w [ ( m ) ( m )]
and the residuals are
2 + + й Ҙ Ҙ *щ - iww m iw s c r r iw z c r r iwx / c (w -{ұwm }) к e dz [zˆ · am (r2 ,z) e ] dx [zˆ · am (r1 ,x ) e ] ъ w 2 - w 2 т т к( m ) -Ҙ -Ҙ ъ л ыw= ұw m
and so the integral is
+ + й Ҙ r r Ҙ r r * щ eiw m s c dz zˆ · a (r ,z) eiw m z c dx zˆ · a (r ,x ) eiw m x / c к т [ m 2 ] т [ m 1 ] ъ pw 2 к -Ҙ -Ҙ ъ m + + . к Ҙ r r Ҙ r r *ъ +e- iw m s c dz zˆ · a r ,z e- iw m z c dx zˆ · a r ,x e- iw m x / c к т [ m ( 2 ) ] т [ m ( 1 ) ] ъ л -Ҙ -Ҙ ы
Shunt impedances, which conveniently describe the interaction of the cavity with the beam, are defined in terms of longitudinal integrals of the voltage:
r +L r V r = dx E r ,x,t = x c . ( ) т z ( ) - L
This expression is phased, i.e. the origin of the axis is implicitly selected to maximize the integral. r r r r r r From E = - ¶A , E (r ,z,t) = - (iw )A (r ,z,t) in the time domain for any purely ¶t m harmonic field of frequency m, and so the last two integrals in the expression for r r +Ҙ r r iw mx c W// resemble to V(r ). DefineVm (r ) ә dx [- iw m zˆ · am (r ,z) e ] as the pure т -Ҙ r harmonic, single-mode complex generalization of V (r ). As with all cases where complex-valuedr quantities are used to represent oscillating quantities, only the real part of V (r ) is physical. Phasing so as to maximize the physical part m r requires that V (r ) is in fact real. With this, the integral over becomes m
r r * r * r r r p eiw m s c V (r ) V (r ) + e- iw m s c V (r ) V (r ) = 2p cos(w s c) V (r ) V (r ) [ m 2 m 1 m 2 m 1 ] m m 2 m 1
and the longitudinal wake field is just r r r r 1 V (r )V (r ) W r ,r ,s = cos w s/c m 2 m 1 // ( 1 2 ) е ( m ) . 2 m Um
Now for equationr (2), whichr states that W// is a two-dimensionally harmonic function of both r1 and r2 . Directly from the Maxwell equations,
r 2 r r 2 1 ¶ 1 ж 1 ¶ ц С E - E = С r + J . c 2 ¶t 2 e з c 2 ¶t ч 0 и ш Quite apart from the no-space charge approximation, which is about the electric field produced by the bunch charge, the z component of the term on the right disappears for any beam traveling at velocity c: r r ¶ ¶r ¶z 2 ¶r 2 ¶r J = c r (r ,z = z - ct) zˆ Ю Jz = c = - c = - c ¶t ¶z ¶t ¶z ¶z
leaving the wave equation
2 2 2 ¶ 1 ¶ 1 ж¶ r 1 ¶ ц С ^ Ez + 2 Ez - 2 2 Ez = з + 2 Jz ч = 0. ¶z c ¶t e0 и¶ z c ¶t ш
This equation is valid at point 2, which is at some distance from the exciting charge. Introducing then the subscript 2,
+Ҙ 2 r r 1 2 r С ^(2) W// (r1 ,r2 ,s) = dz С ^ (2) Ez (r2 ,z,t = (s + z) c) q т 1 -Ҙ +Ҙ +Ҙ +Ҙ 1 dw iw t dk - ikz 2 ˜ r = dz e e С ^ (2) E z (r2 ,k,w) q т т2p т2p z+s 1 -Ҙ -Ҙ -Ҙ t= c
˜ where E z is the z component of the electric field, Fourier transformed from both t to and from z to k. The advantage of using this quantity is that the wave 2 ˜ 2 ˜ 1 2 ˜ equation in the doubly-transformed space is С ^(2) E z + (- ik) E z - 2 (iw) E z = 0, so c +Ҙ +Ҙ +Ҙ 2 r r 1 dw iw t dk - ik z 2 w 2 ˜ r С ^(2) W// (r1 ,r2 ,s) = dz e e (k - 2)E z (r2 ,k,w) q т т2p т2p c z+s 1 -Ҙ -Ҙ -Ҙ t= c + + + Ҙ z Ҙ s Ҙ 2 1 i(w c- kz) 1 iw dk 2 ˜ r = dze dw e c k - w E˜ (r ,k,w) 2p т q т т2p ( c 2 ) z 2 -Ҙ 1 -Ҙ -Ҙ +Ҙ +Ҙ 1 iw s dk 2 r w c 2 w ˜ = d( - k) dw e k - 2 E z (r2 ,k,w) c q т т2p ( c ) 1 -Ҙ -Ҙ = 0.
r r r r 2 W (r ,r ,s) = 0 2 W (r ,r ,s) = 0 A symmetry argument that С ^(2) // 1 2 implies С ^(1) // 1 2 , because of the explicit form of equationr r 1. In general,r rwe should not expect a differential operator, D, to obey D 1 W// (r1 ,r2 ,s) №D 2 W// (r1 ,r2 ,s); the former has terms like r r r r ( ) ( ) r r r r DV r = r V r = r DV r = r V r = r [ ( 1)]Ч ( 2 ) and the latter has terms like [ ( 2 )]Ч ( 1) and both the derivativesr r and the value of V will be different at the two different evaluation points r and r . However, if the differential operator yields a constant, in this 1 2 r r r r r r r r case zero, then [DV (r = r2 )]ЧV (r = r1 ) = [DV (r = r1 )]ЧV (r = r1 ) is the same constant, r r just buy considering the case where r happens to equal r . By the same r 2 r r r 1 reasoning, the constant is also [DV (r = r1 )]ЧV (r = r2 ), and so r r r r r r r r D W (r ,r ,s) = D W (r ,r ,s) 2 W (r ,r ,s) = 2 W (r ,r ,s) = 0 (1) // 1 2 (2) // 1 2 , implying С ^(1) // 1 2 С ^ (2) // 1 2 .
r r Now for equation (3), which is an expansion theorum. First consider W // (r1 ,r2 ,s) 2 r r as a function that solves С(1) W// (r1 ,r2 ,s) = 0, where the Laplacian operator only r r applies to the two dimensional space 1 described with rpolar r coordinatesr r1 and W r ,r ,s = T(r ,s)y f G r . By the familiar separation of variables method, // ( 1 2 ) 2 ( 1) ( 1) and () = {A function of r1 only} cos[m(- . The integer m ensures that
(0) equals (2). The solutions to the separated differential equation for G(r1) +m -m are G0(r1) = A0 + B0 ln(r1) and Gm(r1) = A0 r1 + B0 r1 with Ai and Bi arbitrary
constants. The wakefield potential is finite at r1 = 0, so the Bi are zero for all i. r r Ҙ r W r ,r ,s = A r ,s rm cos m f - f (n) Subsume T into the Ai, leaving // ( 1 2 ) е i ( 2 ) 1 [ ( 1 1 )]. n= 0
It is at this point that we have to allow that multiple modes may have the same azimuthal order; the Ai are one-to-one mapped to solutions to the separated equation for G(r1), but there can be two cos[m(- s, with different values multiplied to each solution G(r1). Hence, n here indexes the modes, and m indexes the azimuthal order of the modes. r (n) Applying the same technique to r2 , and use W// (s) rather than Ai to denote the remaining functions of s,
r r Ҙ W r ,r ,s = W (n ) s rm cos m f - f (n ) rm cos m f - f (n) // ( 1 2 ) е // ( ) 1 [ ( 1 1 )] 2 [ ( 2 2 )]. n= 0 r r n n W (r ,r ,s) But equals . Consider // 1 2 rwhen rr1 = r2 but . From the explict form of equation (1) exchange of r1 with r2 , or equivalently exchange of
with , will not change the value of W//. From the expansion expression, this can n n only happen when = ; this is the quantity n. It will be determined by the
azimuthal geometry of the cavity. For the common case where the cavity is
azimuthally symmetric, the symmetry of the system is broken by the entering r r Ҙ W r ,r ,s = W (n ) s rm rm cos m f - f bunch at 2; then // ( 1 2 ) е // ( ) 1 2 [ ( 2 1)]. n= 0
Finally for equation (4), which is essentially the Panofsky-Wenzel theorum. From ¶ r v ¶ r v v a Maxwell equation, zˆ ҙ B (r ,z,t) = E ^ (r ,z,t) - С ^ E z (r ,z,t) so the transverse ¶t ¶z component of the wake potential is
+ r r r r 1 Ҙ й r м¶ r r ьщ W ^ (r1 , r = r2 ,s) = dz E ^ (r ,z,t) + c dtн E ^ (r ,z,t) - С ^ E z (r ,z,t)э q т к т ¶ z ъ z+s 1 -Ҙ л о юыt= c
where the subscript 2, indicating that the fields are evaluated at the trailing bunch, has been squelched. After finding the functional form of the indefinite integral, that form is to be evaluated at t = z + s / c, and the integration constant is to be discarded. So we can write an upper limit of z + s / c, and select a lower limit at some value where all the fields and their derivatives are zero. From d (z+s) c r (z+s) c ¶ r 1 r Lebnitz’ rule, dt E (r ,z,t) = dt E (r ,z,t) + E (r ,z,t) and so dz т ^ т ¶z ^ c ^
+Ҙ (z+s) c (z+s) c r r r r 1 й жr z + sц ¶ r r щ W ^ (r1 ,r = r2 ,s) = dz кE ^ зr ,z, ч + c dt E ^ (r ,z,t) - c dt С ^ E z (r ,z,t)ъ q т c т ¶z т 1 -Ҙ лк и ш ыъ (z+s) c (z+s) c c +L й d r щ c +Ҙ й r щ = dz к dtE^ (r ,z,t)ъ- dz к dt С ^ E z (r ,z,t)ъ. q т dz т q т т 1 - L лк ыъ 1 -Ҙ лк ыъ
The range of the first integral is constricted into the region of non-zero fields to show why the first integral is zero. At the electric boundaries that define the
integration range, E is zero, and consequently
z= +L +L й d (z+s) c r щ (z+s) c r (L +s) c r (- L +s) c r dz dtE r ,z,t = dtE r ,z,t = dtE r ,L,t - dtE r ,- L,t т к т ^ ( )ъ т ^ ( ) т ^ ( ) т ^ ( ) - L dz лк ыъ z=- L
vanishes. Again applying the Leibnitz rule,
d r r r r c d +Ҙ (z+s) c r W ^ (r1 ,r = r2 ,s) = - dz dt С ^ E z (r ,z,t) ds q ds т т 1 -Ҙ c +Ҙ d (z+s) c r = - С ^ dz dt E z (r ,z,t) q т ds т 1 -Ҙ c +Ҙ 1 r = - dz E r ,z,t = z + s c С ^ т z ( ( ) ) q1 - c Ҙ r r r = - W (r ,r = r ,s) С ^ (2) // 1 2
where the subscript 2 has been restored in the last expression. Re-integration r r r s r r W (r ,r ,s) = - ds W (r ,r ,s ) gives ^ 1 2 С ^ (2) т // 1 2 , where the lower bound on the integral is
omitted to indicate that we want the value of the antiderivative at a certain value of .
continues here: What are the requirements?
To get time slices, the crabbed spot must thus be 1cm in vertical span. The screen will be at 17.2m in the reworked A0 beamline design; the cryovessel starts at 13.0m. Actually, Phillipe would be very happy if the device worked on a screen at 16.5m, but we decided to set the spec at 17.2m. If we use the existing cryovessel, we will need to find an additional 40cm or so in the beamline for it; maybe that is possible. The existing vessel is, according to the drawings, 1.823m long but we measured about 2m. The distance from its center to the downstream end of the upstream cavity is 3 inches, from the drawing. So 13.0 +1.823/2 –(3*0.0254 +0.100) = 13.7m, a mildly conservative estimate of the downstream end of the cavity. Throw arm is thus 17.2 – 13.7 = 3.5m. Required
V is thus, for a 40MeV beam, 40(0.5cm / 3.5m) = 57keV. For a bunch of 1ps length at zero crossing, the head of the bunch is at 2(0.5ps)(3.9 x109Hz) = 12.2 mradian, the sin of which is 12.3 x10-3. Then, sticking with 5MeV/m for safety’s sake, (5 MeV/m)(12.3 x10-3)ℓ = 57keV gives ℓ = 0.930m. This is conservative; 2ps or maybe even 3ps are more likely than 1ps, and we probably can get to 6MeV/m or maybe even 7MeV/m. With 2ps and 6MeV/m, the length requirement goes down to 0.388m 10 cells. For a transverse kick of 57keV, a bunch of 12pC will consume 684J of energy in the transverse kick. Were we to run a long train like this with 1sec spacing, that would be a lot of power, but such is not our intention.
What are the issues?
At these duty cycles, we do not have to be concerned with cryogenic loading.
Hassan’s Eqn. (15.27) gives a factor that quantifies resonant build-up of the cavity fields on the beam. The quantity Fr, given by T 1- expж- 2 b ц з T ч F = и d ш r T T й ж b ц ж b цщ 2 1 - 2expз- чcos(DwTb ) + expз- 2 ч лк и Td ш и Td шыъ where is 2 times the frequency shift to the nearest harmonic of Tb, the bunch spacing, and and Td is the decay time for the mode of concern, is the ratio of voltage seen by the bunch from all the proceeding bunches to the voltage induced by a single charge. For most modes, I have on the order of 21010 5 5 and Qext over 10 , leading to decay times on the order of 10 sec or longer. This -6 is much longer than Tb = 10 , so I expand in small powers of = Tb / Td. The function Fr is sharply peaked near zero, so expansion in small powers in Tb is also in order: the place where Fr = 1 is
exp(t ) + 3exp(- t ) cos(DwT ) = b 4 so the blowup condition is DwTb Ј t , or (for A0)
T f - 6 f Df Ј b n = 282 ҙ 10 n . 4p Q Q ext ext
In the case of the 10/13 TM010 mode, f = 2827.38 – (integer)*1MHz = 380kHz, -9 so the blowup condition is Qext < 1.6 x10 . That is, for any reasonable damping, 2827.38MHz is so far from the nearest exciting frequency of 2827MHz that resonant buildup cannot happen. Figuring a random HOM to have f on the order 10 6 of 10 , and Qext to be typically at or above order of 10 , f is on the order of a few mHz. I think that resonant buildup of this type is just not an issue. Previous MAFIA studies (14 August 2003) show that a 13 cell cavity introduces an energy shift proportional to the vertical distance of the beam from the axis:
+0.349m E r = z eiwz / cdz @(0.204MV /mm) y т z ( ) . - 0.349m
This is an overestimate in that the bunch will deposit some energy into the cavity corresponding to a voltage of 2kd(r=a) |q| /a (Padamsee 15.46) where a is a 2 specific radius, i.e., the iris radius, and the dipole loss factor is kd = Va / 4U, and Va is the accelerating voltage given by the above MAFIA study at the radius a. This loss parameter is thus 2 r 2 (0.204MV ) 2 (1) ( 1mm) - 3 2 k (r) = = 28.7 ҙ 10 (r ) MV /J 4(0.363J) 1mm where I am using Rainer Wanzenberg’s notation for the loss factors. The loss factor goes quadratically with the radius of the bunch passing through the cavity. The energy left in the cavity by a bunch is also quadratic in beam displacement and linear in the charge of a bunch:
жq ц 2 V = 57.4 10- 12 MV r = z q ҙ з 1nCч( 1mm) . и ш
The fundamental theorum of beam loading says that one half of this voltage acts to slow the charge as it passes through the cavity. However, even for our worst case 10nC bunch displaced out to 12mm – that is 3mm from the iris – this is 41.3 x10-9 MV, quite negligible.
Nonetheless, the energy dispersion introduced into the bunch, even due to it’s finite width, will have some issues later down the beamline where the bunch is focused.