Nuclear Instruments and Methods in Physics Research A 474 (2001) 197–208 Suppressing electron multipacting in ceramic windows by DC bias Pasi Yla. -Oijala*, Marko Ukkola Rolf Nevanlinna Institute, University of Helsinki, P.O. Box 4, (Yliopistonkatu 5), FIN-00014 Helsinki, Finland Received 3 January 2001; accepted 26 February 2001 Abstract Electron multipacting can cause breakdown in high RF power components like couplers and windows. The phenomenon starts if certain resonant conditions for electron trajectories are fulfilled and if the impacted surface has a secondary electron yield larger than one. A general cure against multipacting is to avoid the resonant conditions by a proper choice of the geometry. In many cases, however, it might not be possible to change the RF geometry sufficiently. Therefore, other methods are required to suppress multipacting, like grooving the surfaces or static electric and magnetic perturbations. In this paper, we investigate by numerical simulations the effect of a biasing DC voltage to multipacting in two special ceramic window geometries. Those are the DESY and FNAL cylindrical symmetric cold window designs of the TESLA input power coupler. We show that in both window designs multipacting can be avoided by choosing an appropriate biasing voltage. # 2001 Elsevier Science B.V. All rights reserved. PACS: 52.80; 07.05.T Keywords: Multipacting; Input coupler; Ceramic window; DC bias 1. Introduction Multipacting is a phenomenon of resonant electron multiplication, in which a large number of electrons build up an electron avalanche, leading to remarkable power losses and heating of the walls, so that it becomes impossible to increase the cavity fields by rising the incident power [1]. In superconducting structures a large rise of temperature can lead to a thermal breakdown. Furthermore, a heavy bombing of multi*Corresponding author. Fax: +358-9191-22779. E-mail address: pasi.yla-oijala@rni.helsinki.fi (P. Ylä-Oijala). pacting electrons may break the ceramic windows leading to disastrous consequences. A general cure against multipacting is to avoid the resonant conditions by either a proper choice of the geometry [2] or by coating the critical areas with a material with a lower secondary yield [3,4]. In many cases, however, it might not be possible to change the RF geometry sufficiently and the coating, which is typically used to reduce the secondary yield of a ceramic window, does not suppress completely multipacting and the success rate of the reduced secondary yield is often unsatisfactory. In those cases, other suppressing methods must be applied, like static electric [5] or magnetic perturbations, or grooving [6] the surfaces. 0168-9002/01/$ - see front matter # 2001 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 8 - 9 0 0 2 ( 0 1 ) 0 0 8 8 2 - 8 198 P. Ylä-Oijala, M. Ukkola / Nuclear Instruments and Methods in Physics Research A 474 (2001) 197–208 In Ref. [7], we analyzed multipacting (without the bias) in the cylindrically symmetric cold windows of the TESLA [8] input power coupler. Both the cylindrical DESY (Deutsches Elektronen-Synchrotron) and conical FNAL (Fermi National Accelerator Laboratory) window designs were considered. It was found that in the DESY design multipacting can be avoided by placing the window at the right distance from the coupler end. In the FNAL window design this was not possible because of a very wide existence of multipacting. Therefore, in the FNAL design other methods are required to avoid multipacting. A static electric perturbation, i.e., a biasing DC voltage, is an attractive choice in the coaxial structures [5]. Since both TESLA input power coupler designs include long straight coaxial sections and a DC voltage is typically applied to avoid multipacting in coaxial lines, one should know also in the DESY design whether a new type of multipacting is possible with the DC bias. Our earlier computations in straight coaxial lines with the DC bias [9] showed that the voltage has to be chosen properly and a wrong DC voltage may cause new low order multipacting with a high secondary yield. Furthermore, we were able to find simple rules for multipacting resonances with respect to the dimensions of the line, frequency of the field and the biasing voltage. In the ceramic windows the multipacting analysis with a DC bias becomes much more complicated, because of the varying field conditions during the operation of the system and of the complexity of the window designs. In this paper, we investigate the effect of a biasing DC voltage to multipacting in both TESLA cold window designs and, in particular, show that with the developed computational methods [9] we are able to find the optimal DC voltages in the DESY and FNAL window designs by which multipacting can be avoided. 2. DC bias in coaxial lines We begin by shortly summarizing our previous results on multipacting simulations in straight coaxial lines with a biasing DC voltage. In a straight coaxial line, we define the DC field corresponding to a voltage U as follows [9] EDC ðr; zÞ ¼ U 1 er : ln b=a r Here a and b stand for the inner and outer radii of the line and ðr; zÞ is a field point in a cylindrically symmetric structure. In other words, a positive voltage, U > 0, corresponds to an electric field pointing from the inner conductor to the outer conductor and a negative voltage corresponds to an opposite field. In the ceramic windows, the DC field is defined accordingly. Obviously, in the ceramic windows the DC field does no longer have an analytical solution and numerical methods must be applied for the field calculations. In Ref. [10], we studied DC voltage in coaxial lines by varying the dimensions, frequencies and wave forms. The results were summarized in Ref. [9]. Two main points were recognized. Firstly, the DC voltage which is required to suppress multipacting in the standing wave (SW) operation suppresses multipacting also for all other wave configurations (traveling waves (TW) and mixed waves). Secondly, the suppressing DC voltage satisfies the following scaling law U dZf ð1Þ where d is the diameter of the line, Z is the impedance of the line and f is the frequency of the field. In other words, the DC voltage which is required to suppress multipacting depends linearly on the dimension, impedance and frequency. Table 1 lists the DC voltages, both negative and positive, by which multipacting can be suppressed up to 1 MW in the straight coaxial sections of the DESY and FNAL input coupler designs [10]. Note that the coaxial lines have different dimensions on the cold (the section from the cold window to the cavity) and warm (the section from the cold window towards the generator) sides of the couplers. We may conclude that in the DESY coupler either 3:5 or þ2:7 kV is needed to suppress multipacting. In the FNAL coupler, the corresponding numbers are 3:5 and þ2:5 kV. Naturally, these results do not hold in the ceramic windows and the suppressing voltages have to be redefined. However, we may conclude that the analysis in the ceramic windows can be restricted P. Ylä-Oijala, M. Ukkola / Nuclear Instruments and Methods in Physics Research A 474 (2001) 197–208 Table 1 The suppressing DC voltages and dimensions of the coaxial lines of the TESLA input coupler designs. Here Z is the line impedance, d is the outer diameter and U is the voltage DESY Z ðOÞ d (mm) U (kV) U (kV) Cold Warm 70 50 40 62 2:9 3:5 þ2:7 þ2:5 FNAL Cold Warm 50 50 40 62 2:1 3:5 þ2:3 þ2:5 to the voltages equal or higher than 3:5 kV and þ2:7 or þ2:5 kV. 3. DC bias in ceramic windows Before we can start the multipacting analysis the distribution of the electromagnetic fields must be computed. The DC field in the ceramic windows is computed by a finite element method with a third order approximation [11]. For the trajectory calculations, both the DC and RF fields are given in a grid and interpolation is used elsewhere. During the operation of the RF pulse, the reflection conditions on the coupler vary. Therefore, we have to consider a large number of simulations with various RF field conditions. In a similar fashion as in Ref. [7] (without the bias), we define the field conditions by a complex reflection coefficient R with jRj41. The magnitude of R describes the amplitude ratio of the reflected and forward waves, and the phase of R gives the phase difference of the reflected and forward waves. In reality only part of the reflection conditions occur and the analysis may be restricted to these areas. Here, however, the analysis is carried out for all possible reflection conditions and the real operation conditions, given in Ref. [7], and indicated in the figures (in the DESY case). The coordinate systems are fixed similarly as in Ref. [7]. That is, R ¼ 1 gives a SW with a voltage maximum at the one quarter wave length from the center of the window, R ¼ 0 is a TW and R ¼ 1 is a SW with a voltage minimum at the distance of one quarter wave length from the center of the window. 199 The multipacting analysis is carried out as follows. First, we fix the DC voltage. Then we scan through a representative sample of reflection conditions with jRj41 and power levels up to 1 MW as described in Ref. [7]. For each fixed reflection coefficient and RF power combination, we send a large number of electrons from different points on the surface of the structure, compute the electron trajectories and count the total number of secondary electrons after 20 impacts. This count is called an enhanced electron counter function. The results of the simulations are illustrated by plotting relative enhanced counter function, which is the number of secondary electrons after 20 impacts divided by the number of initial electrons, as a function of the reflection coefficient. For each complex R, we plot the maximum of the relative enhanced counter with respect to the RF power, when the RF power is scanned from 0 to 1 MW. By this analysis, we are able to locate those operation conditions at which multipacting may occur. In order to find the multipacting power levels we repeat the analysis by fixing the reflection coefficient and by varying the DC voltage and RF power. The latter analysis is, however, carried out only with a couple of reflection coefficients. In the figures below (Figs. 2–7 and 9–12) the shade indicates the change of the total number of electrons after 20 impacts. In bright areas the Fig. 1. The secondary electron yield at different impact energies (in eV) for a niobium surface baked at 3008C. 200 P. Ylä-Oijala, M. Ukkola / Nuclear Instruments and Methods in Physics Research A 474 (2001) 197–208 number of electrons is increased and multipacting occurs, whereas in dark areas the number of electrons is decreased. (See the colorbar on the right hand side of the figures.) The multipacting level, i.e., when the relative enhanced counter is Fig. 2. The relative enhanced electron counter for the cold side of the DESY coupler in 10 base logarithmic scale as a function of the complex reflection coefficient after 20 impacts when the DC voltage is 3:5 kV and for the RF power range 0–1 MW. The real operation conditions are denoted on the plot by dashed lines (see Fig. 9 in Ref. ½7). Fig. 3. The relative enhanced electron counter for the warm side of the DESY coupler for the RF power range 0–1 MW when the DC voltage is 3:5 kV. one (zero in the logarithmic scale) is indicated in the figures by a solid line. Here, we assume that the secondary yield of the ceramic window is the same as in the metallic surfaces. This corresponds to a coated ceramic, coated e.g. by TiN. The secondary yield curve applied in the calculations is displayed in Fig. 1 [12]. This curve corresponds to a niobium surface. However, the couplers are made of copper. The Fig. 4. The relative enhanced electron counter for the warm side of the DESY coupler for the RF power range 0–1 MW when the DC voltage is þ3:0 kV. Fig. 5. The relative enhanced electron counter for the warm side of the DESY coupler for the RF power range 0–1 MW when the DC voltage is þ4:0 kV. P. Ylä-Oijala, M. Ukkola / Nuclear Instruments and Methods in Physics Research A 474 (2001) 197–208 201 Fig. 6. The relative enhanced electron counter for the warm side (top) and cold side (bottom) of the DESY window in 10 base logarithmic scale as a function of the RF power and DC voltage, when R ¼ 1, i.e., a SW with a voltage maximum at the window region. secondary yield curve for copper is rather similar, the maximum yield 1.3 is reached at Ekin ¼ 600 eV, and the secondary yield curve exceeds unity between 200 and 1500 eV [1,13]. Thus, the major difference between the secondary yield curve for the copper and niobium is that the copper has a lower maximum yield. This effects only to the strength of multipacting, not to the existence. More details about the multipacting analysis in ceramic window are presented in Ref. [7]. 3.1. DESY cylindrical window design First we consider the DESY window design. The design is displayed in Ref. [7] and more details about the geometry are given, e.g., in Ref. [14]. As the computations in the straight coaxial lines suggest, first we consider the biasing DC voltages of 3:5 and þ3:0 kV. Figs. 2–4 display the results of the analysis, separately for both cold and warm 202 P. Ylä-Oijala, M. Ukkola / Nuclear Instruments and Methods in Physics Research A 474 (2001) 197–208 Fig. 7. The relative enhanced electron counter for the warm side (top) and cold side (bottom) of the DESY window in 10 base logarithmic scale as a function of the RF power and DC voltage, when R ¼ 0, i.e., a TW. sides. Figs. 2 and 3 show that when the DC voltage is 3:5 kV, multipacting takes place on a very wide area of the reflection chart (bright area). In fact, DC voltage has created a lot of new multipacting, especially on the warm side, which does not occur without the bias. Fig. 4 shows the relative counter function on the warm side of the DESY window when the DC voltage is þ3:0 kV. We can see that also in this case multipacting occurs on a rather wide area of the reflection chart. On the cold side, on the other hand, multipacting can be avoided with þ3:0 kV (the results are not presented here). These results suggest that a positive DC voltage higher than þ3:0 kV is required to suppress multipacting. Therefore, we repeated the analysis with þ3:5 and þ4:0 kV. When the bias is þ3:5 kV, the cold side is still free of multipacting and on the warm side the multipacting region (bright area in Fig. 4) becomes smaller, but the area of the real operation P. Ylä-Oijala, M. Ukkola / Nuclear Instruments and Methods in Physics Research A 474 (2001) 197–208 203 Fig. 8. Electron trajectories in the DESY window design when R ¼ 0: (a) U ¼ 3:5 kV and P ¼ 690 kW, (b) U ¼ þ3:0 kV and P ¼ 800 kW, (c) U ¼ 2:5 kV and P ¼ 606 kW, (d) U ¼ þ1:5 kV and P ¼ 138 kW. The dimensions are in meters and the trajectories are not to scale. All trajectories are one-point processes of order one and the corresponding impact energies are 270, 580, 785 and 460 eV. The window is denoted by shading. For the geometry, see Fig. 1 in Ref. [7]. conditions still hits the dangerous multipacting region. This takes place near the TW operation and therefore, it is not possible to avoid multipacting by a proper placing of the window with þ3:5 kV and a higher voltage is required. Thereafter, Fig. 5 displays the results of the analysis on the warm side when the voltage is þ4:0 kV. Once again the real operation conditions hit the multipacting region. Now the multipacting region is smaller than in the case of þ3:5 kV. A closer analysis of the counter functions shows that for the real operation conditions multipacting can be avoided up to 900 kW. A slightly higher voltage may suppress multipacting up to 1 MW. More- 204 P. Ylä-Oijala, M. Ukkola / Nuclear Instruments and Methods in Physics Research A 474 (2001) 197–208 Fig. 9. The relative enhanced electron counter for the warm side of the FNAL coupler in 10 base logarithmic scale as a function of the complex reflection coefficient after 20 impacts for the RF power range 0–1 MW when the DC voltage is þ2:5 kV. over, the analysis on the cold side with þ4:0 kV shows no multipacting. We repeated the analysis with higher negative voltages, too, and found that when U ¼ 4:0 kV the multipacting region becomes smaller, but the region is still very large (the results are not presented here). Thus, a positive voltage is clearly a better choice for the DESY window design. Next, we change the analysis so that we fix the reflection coefficient and let the biasing voltage vary from 4:0 to þ4:0 kV. Figs. 6 and 7 show the enhanced counter function as a function of the RF power and DC voltage when R ¼ 1 and R ¼ 0. In the figures all positive values correspond to multipacting (bright areas), while the dark areas show the multipacting free regions. By these figures one can determine the DC voltage so that the RF power can be raised from zero to the desired value without crossing any of the multipacting bands. For example, Fig. 6 suggests that on the warm side þ1:5 kV is required to suppress multipacting up to 1 MW, whereas on the cold side the needed voltage is twice as much, i.e., þ3:0 kV. Furthermore, Fig. 7 shows that in the TW case the corresponding voltage is þ4:0 kV. A closer analysis of Figs. 6 and 7 shows, analogously to the coaxial case, that when the RF power is Fig. 10. The relative enhanced electron counter for the warm side of the FNAL coupler for the RF power range 0–1 MW when the DC voltage is þ3:0 kV. increased from zero with any fixed DC voltage we always first ‘‘hit’’ the first order process and thereafter the second order process follows, etc. In order to get a complete picture of the phenomenon we would have to repeat the analysis for all reflection coefficients. Fig. 8 shows typical multipacting electron trajectories for four DC voltage and RF power combinations when R ¼ 0, i.e., for a TW. Figs. 8(a) and (b) shows the trajectories on the warm side and Figs. 8(c) and (d) shows the trajectories on the cold side. In the straight coaxial lines, we found that negative voltages generate multipacting to the outer conductor of the line and positive voltages generate multipacting to the inner conductor of the line [9]. The trajectories in Fig. 8 show that this is true in the DESY window design, too, in the TW case. For other wave forms, the situation is more complicated. 3.2. FNAL conical window design Next we consider the results of the multipacting simulations in the conical FNAL window. The design is shown in Ref. [7] and more details are presented, e.g., in Ref. [14]. The simulations in the straight coaxial lines suggest that in order to avoid multipacting the DC P. Ylä-Oijala, M. Ukkola / Nuclear Instruments and Methods in Physics Research A 474 (2001) 197–208 205 Fig. 11. The relative enhanced electron counter for the warm side (top) and cold side (bottom) of the FNAL coupler in 10 base logarithmic scale as a function of the RF power and DC voltage, when R ¼ 1, i.e., a SW with a voltage maximum at the window region. voltage has to be at least either 3:5 or þ2:5 kV. The analysis shows that multipacting does not occur in the window region when the voltage is 3:5 kV (the counter functions are not presented here). Thus, the same negative voltage which suppresses multipacting in the straight coaxial sections of the FNAL coupler design will also suppress multipacting in the ceramic window. Next, we study whether multipacting can be avoided with a lower positive voltage. Fig. 9 shows the enhanced counter function as a function of the reflection coefficient on the warm side of the FNAL design when the DC voltage is þ2:5 kV. The figure shows that multipacting occurs on a wide region on the reflection chart. Therefore, we repeated the analysis with þ3:0 and þ3:5 kV. Fig. 10 shows that the multipacting region of Fig. 9 becomes smaller and the intensity of multi- 206 P. Ylä-Oijala, M. Ukkola / Nuclear Instruments and Methods in Physics Research A 474 (2001) 197–208 Fig. 12. The relative enhanced electron counter for the warm side (top) and cold side (bottom) of the FNAL coupler in 10 base logarithmic scale as a function of the RF power and DC voltage, when R ¼ 0 ,i.e., a TW. pacting decreases when the voltage is increased from þ2:5 to þ3:0 kV. However, multipacting still takes place on a rather wide region on the reflection chart. This suggests that a little higher voltage is needed. An analysis with þ3:5 kV confirms this observation. Now the counter function stays clearly below the multipacting level for all analyzed reflection coefficients and power levels up to 1 MW and multipacting can be avoided. Furthermore, the simulations show that on the cold side all positive voltages equal or higher than þ2:5 kV suppress multipacting. Thus, we may conclude that either 3:5 or þ3:5 kV is required to suppress multipacting in the FNAL design for all reflection coefficients up to 1 MW. Next, we repeated the analysis in the FNAL window design by fixing the reflection coefficient and by letting the biasing voltage vary from 4:0 to þ4:0 kV. Figs. 11 and 12 illustrate the behavior of the multipacting power bands when R ¼ 1 P. Ylä-Oijala, M. Ukkola / Nuclear Instruments and Methods in Physics Research A 474 (2001) 197–208 207 Fig. 13. Electron trajectories in the FNAL window design when R ¼ 0: (a) U ¼ 2:0 kV and P ¼ 613 kW, (b) U ¼ þ2:0 kV and P ¼ 389 kW, (c) U ¼ 2:0 kV and P ¼ 536 kW, (d) U ¼ þ1:5 kV and P ¼ 219 kW. The dimensions are in meters and the trajectories are not into scale. All trajectories are one-point processes of order one and the corresponding impact energies are 460, 435, 745 and 750 eV. The window is denoted by shading. For the geometry, see Fig. 1 in Ref. ½7. and R ¼ 0 by giving the enhanced counter function as a function of the RF power and DC voltage. We may conclude that when R ¼ 1, i.e., the SW case, the voltage which is required to suppress multipacting is roughly of the same size as in the straight coaxial lines, i.e., 3:0 or þ3:0 kV on the warm side and 2:5 or 2:0 kV on the cold side (see Table 1 for the voltages in the coaxial case). Furthermore, Fig. 12 shows that when the wave is switched from a SW to a TW, the SW multipacting power levels scale upwards roughly by factor 4, similarly as in the coaxial case [10]. The trajectory calculations confirm that the corresponding SW and TW processes are very similar. Fig. 13 shows samples of multipacting electron trajectories in the TW case. Again the same phenomenon as in the straight coaxial lines and in the DESY design can be recognized, namely 208 P. Ylä-Oijala, M. Ukkola / Nuclear Instruments and Methods in Physics Research A 474 (2001) 197–208 a negative voltage generates multipacting to the outer surface and a positive voltage generates multipacting to the inner surface. As has been already mentioned, we have found that in the straight coaxial lines it suffices to consider the SW case only in order to find the optimal biasing voltage for all possible wave configurations. The present analysis shows that also in the FNAL window design the SW case gives a rather good estimate for the optimal DC voltage, when several SW cases with the voltage maximum at the window region are studied. Again, it has to be pointed out that in order to get a complete picture of the phenomenon we would have to repeat the analysis for all reflection coefficients. 4. Conclusions In this paper, we apply the numerical methods developed in Ref. [9] to study the effect of a DC bias to multipacting in two special ceramic window designs. We consider both the FNAL and DESY cylindrically symmetric cold window designs of the TESLA input power coupler. It is found that in the DESY design þ4:0 kV is required to suppress multipacting on the real operation conditions up to 900 kW. A little higher voltage will suppress multipacting up to 1 MW. In addition, in the DESY window design the sign of the DC field is important, since a negative voltage 4:0 kV creates multipacting. In the FNAL window design both 3:5 and þ3:5 kV suppress multipacting on the whole reflection chart. Thus, we may conclude that although our previous analysis without the DC bias showed that the FNAL design is clearly more sensitive to multipacting than the DESY design, in the case when a DC voltage is required, e.g., to suppress multipacting in the straight coaxial sections, the DC bias is more effective in the FNAL design and 500 V higher voltage is needed in the DESY design. In the FNAL design, multipacting phenomenon has some similarities with the straight coaxial case. For example, a rather good estimate for the optimal DC voltage for all reflection conditions can be found by studying the SW case with a voltage maximum at the window region. In the DESY window the phenomenon is clearly more complicated, because of the more complex design. Acknowledgements The authors wish to thank S. J.arvenp.aa. from Rolf Nevanlinna Institute for computing the DC fields. This work has been financially supported by the Helsinki Institute of Physics. References [1] H. Padamsee, J. Knobloch, T. Hays, RF Superconductivity for Accelerators, Wiley, inc., New York, 1998. [2] U. Klein, D. Proch, Multipacting in superconducting RF structures, Proceedings of the Conference of Future Possibilities for Electron Accelerators, Charlottesville, WU B 78-34, 1979. [3] A.R. Nayaiesh, E.L. Garwin, F.K. King, R.E. Kirby, Properties of thin anti-multipactor coatings for klystron windows, SLAC-PUB-3760, 1985. [4] D. Proch, D. Einfeld, R. Onken, N. Steinhauser, Measurement of multipacting currents of metal surfaces in RF fields, WPQ24, IEEE Proceedings, PAC 95, 1776, 1996. [5] J. Tu. ckmantel, Multipacting calculations for a power DCbiased 75 O coupler, Technical Report 94-26, CERN LEP2 Notes, Geneva, Switzerland, 1994. [6] H. Padamsee, D. Proch, P. Kneisel, J. Mioduszewski, IEEE Trans. Magn. 17 (1981) 947. [7] P. Yl.a-Oijala, Particle Accel. 63 (1999) 105. [8] R. Brinkmann, G. Materlik, J. Rossbach, A. Wagner, Conceptual design of a 500 GeV eþ e linear collider with integrated X-ray laser facility, Vol. I, DESY 1997-048, ECFA 1997-182, 1997. [9] E. Somersalo, P. Yl.a-Oijala, D. Proch, J. Sarvas, Computational methods for analyzing electron multipacting in RF structures, Particle Accel. 59 (1998) 107. [10] P. Yl.a-Oijala, Suppressing electron multipacting in coaxial lines by DC voltage, TESLA Reports 97-21, 1997. [11] S. J.arvenp.aa. , A 2D FEM code for computing DC fields in ceramic windows, Rolf Nevanlinna Institute, Helsinki, 1999. [12] R. Calder, et al., Nucl. Instr. and Meth. B, Beam Interact. Matter. At. 13 (1986) 631. [13] Handbook of Chemistry and Physics, 65th Edition, CRC Press, Boca Raton, FL. [14] D.A. Edwards (Ed.), TESLA test facility linac}design report, TESLA Reports 95-01, DESY Print 1995.