3380 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 69, NO.

6, JUNE 2022

Thermal Design of Megawatt Power-Level

X-Band Coaxial Magnetron
Sandeep Kumar Vyas

Abstract — Thermal analysis of X-band high-power RF source for the development of X-band LINAC [3]. There
(∼2 MW) coaxial magnetron is presented. The X-band are some recent works that have been reported on the elec-
magnetron is used for a medical linear accelerator tromagnetic and particle-in-cell (PIC) simulations of X-band
(LINAC). The performance of LINAC depends on the RF
performance of the magnetron. This magnetron consists magnetron [4]–[6]. The oscillation efficiency of high-power
of an oxygen-free copper anode with 40 vanes coupled magnetron generally lies between 50% and 60%, thereby
with the coaxial cavity, cooling channels, and an indirectly dissipating the remaining kilowatts of power in the form
heated oxide-coated cathode. The operating temperature of heat inside the magnetron. The pulse power magnetron
of the cathode is about 850 ◦ C, while the tungsten heater generally has an oxide-coated indirect heated cathode, which
temperature is greater than 1800 ◦ C. Therefore, the thermal
design of X-band magnetron is a challenging task and it operates at a temperature above 850 ◦ C [7], [8]. The anode
is essential for reliable operation. This article deals with also gets the heat radiation from the cathode surface. Thus,
the step-by-step thermal design procedure for each part of overall, a very large amount of power is dissipated inside the
this magnetron. The analytical (MATLAB code) calculation magnetron. Therefore, efficient thermal design is required for
of heater temperature for various heater voltages is also the safe operation of magnetron.
carried out for verification of simulation results. By analyz-
ing the temperature uniformity on the emitting surface of In the previous work for industrial 60-kW L-band mag-
the cathode and the effect of the heat film coefficient, it was netron [9], they estimated the heat convection coefficient of
found that the maximum temperatures on the magnetron forced-air cooling to be 500 W/(m2 ·K). The L-band magnetron
anode block, cathode surface, and heater are 220 ◦ C, had the directly heated tungsten cathode and the thermal analy-
1045 ◦ C, and 1842 ◦ C, respectively, for an applied 5-L/min sis was done with the simplified model: nonhelical cathode
flow rate of water cooling.
shape and power were delivered by heat flux. Their aim was to
Index Terms — Cathode support, coaxial magnetron, confirm the experimental results and find the heat convection
helical heater, oxide-coated cathode, thermal analysis. coefficient in order to apply the efficient cooling rate. The good
agreement between simulation and experiments was observed.
I. I NTRODUCTION
However, it is well known that the dimension of all most

T HE magnetron oscillator is a high-efficiency cross-field

vacuum electron device. Here, the beam–wave interaction
takes place in the azimuthal direction with an applied axial
microwave devices decreases with frequency increase causing
more difficulty for thermal heat management. For example,
in 3.1-MW S-band magnetron, the cathode length is around
continuous magnetic field that makes the device very compact 40 mm, but here, in X-band, it is around 10 mm [10].
and energy-efficient. Since its invention during World War II, This article deals with the thermal study of ∼2-MW pulsed
magnetrons had been used and are still being used in varieties power, frequency-tunable X-band coaxial magnetron using the
of industrial, scientific, and medical applications [1], [2]. 3-D code ANSYS [11]. The complete geometry and the 3-D
A linear accelerator (LINAC) system requires high-power sectional view of the cathode–anode assembly of simulated
(megawatts) microwave source for the RF field to beam energy coaxial magnetron with the cooling channels are shown in
transfer. The magnetron is the most economical and suitable Fig. 1. The different parts (and corresponding materials) of
for this. the magnetron are the anode [oxygen-free high conductiv-
The nowadays trend is to develop a lightweight, compact, ity copper (OFHC Cu)], coaxial cavity (OFHC Cu), heater
and easily transportable X-band LINAC system [3]. The (97% tungsten and 3% rhenium alloy) with the radius of rh ,
megawatt power-level X-band magnetron is required as an cathode (nickel) with the radius of rc , ceramics (Al2 O3 ), the
Manuscript received January 8, 2022; revised March 20, 2022; pole piece (iron), and cathode supports (Kovar). An iron pole
accepted March 30, 2022. Date of publication April 13, 2022; date of piece and yoke are used to confine the magnetic field in the
current version May 24, 2022. The review of this article was arranged by interaction region between the anode and the cathode. There
Editor L. Kumar.
The author is with the Department of Physics, Shree Chavo Veero Girls are three main heat sources that exist for heat generation in
PG College, Bagar, Rajasthan 333023, India (e-mail: sandeepvyas19@ the magnetron namely cathode temperature, electron beam
gmail.com). collision with the anode, and the RF loss due to the finite con-
Color versions of one or more figures in this article are available at
https://doi.org/10.1109/TED.2022.3164643. ductivity of cavity material. There is some other heat source
Digital Object Identifier 10.1109/TED.2022.3164643 that also exists such as slot mode and TE121 cavity mode.

0018-9383 © 2022 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.

Authorized licensed use limited to: POLO BIBLIOTECARIO DI INGEGNERIA. Downloaded on June 10,2024 at 10:51:25 UTC from IEEE Xplore. Restrictions apply.
VYAS: THERMAL DESIGN OF MEGAWATT POWER-LEVEL X-BAND COAXIAL MAGNETRON 3381

Fig. 2. 3-D simulation model of heater.

from the cathode is dependent on the heater temperature

remotely. Generally, the heater temperature is calibrated in
terms of applied heater voltage. Fig. 2 shows the 3-D ANSYS
workbench simulation model of meshed heater only. The ther-
mal analysis of the heater is important because it decides the
cathode temperature magnitude and variation of temperature
on the emitting surface of the cathode. The temperature-
dependent material properties of heater wire are taken in the
simulation to account for the nonlinear effects. The heater wire
and heater diameters are denoted as “d” and “dh ,” respectively.
For all simulations here and in Section II-B, the number of
turns is fixed as 9.
The electric voltage has been applied between the end
terminals of the heater to create an electric current, which
is responsible for resistive heating. The radiation boundary
condition has been applied on the heater surface. A current
reaction probe is used to find out the heater current value in the
postprocessing of ANSYS workbench. The simulations were
done for the different applied voltage and wire parameters:
wire diameter “d” swept from 0.5 to 0.8 mm (d = 0.6 mm
Fig. 1. Schematic of the complete (top) and 3-D sectional (bottom) view is typical for the S-band magnetron heaters [10]); the heater
of the cathode–anode assembly of coaxial magnetron.
diameter “dh ” swept from 8 to 12 mm, and the later one is
limited due cathode sleeve inner diameter, which is discussed
If magnetron is not designed properly, then the slot and TE121 in Section II-B; the heater voltage varied from 1 to 10 V
mode create a lot of thermal problems. An indirectly heated (10 V is the typical heater voltage for the 1.7-MW X-band
oxide-coated cathode is used to obtain the desired current, magnetrons [12]).
which in the present design is ≈90 A at 37 kV. The cathode The analytical calculation of heater temperature (1) is also
is heated at the temperature above 850 ◦ C using a heater carried out to validate the simulation results [13]
 
(filament). The heater is made using a tungsten–rhenium (3%) dT I 2 R0 (1 + αT ) − πdεlσ T 4 − Ta4
wire. =   (1)
dt ρd cl π d
2
This article is organized into four sections, including the 4
introduction. Section II presents a coupled electric-thermal
where R0 , α, T , ρ, l, ε, σ , d, T , Ta , ρd , and c are cold resis-
nonlinear simulation of heater and cathode. The analytical tance, temperature coefficient of the wire, temperature gradi-
calculation of heater temperature has been done by solving the
ent, resistivity, length of wire, emissivity, Stephen–Boltzmann
heat equation using MATLAB too. The MATLAB code for the
coefficient, wire diameter, final temperature, ambient tempera-
solution of the heat equation has been given in the appendix for ture, material density, and specific heat capacity, respectively.
the reader’s help. Section III contains the thermal simulation
The left-hand side of (1) should be zero for steady-state
of complete magnetron, the calculation of power loss, and heat
condition, and after some mathematical manipulations, the
transfer coefficient. final time-independent equation is represented as

II. T HERMAL S IMULATION OF H EATER AND C ATHODE V2

αT 5 + (1 − αTa )T4 + αTa T − + αTa4 + Ta4 = 0.
A. Thermal Simulation of Heater R0 πdεσ l
(2)
A helical-type heater made of 97% tungsten and 3% rhe-
nium alloy is used in this magnetron to warm up the cathode MATLAB code is written for the solution of (2) (see
surface at a temperature above 850 ◦ C. The electron emission Appendix). There are some input parameters required to run

Authorized licensed use limited to: POLO BIBLIOTECARIO DI INGEGNERIA. Downloaded on June 10,2024 at 10:51:25 UTC from IEEE Xplore. Restrictions apply.
3382 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 69, NO. 6, JUNE 2022

Fig. 5. Heater temperature versus applied voltage for different heater

diameters (dh = 8, 9, 10, 11, and 12 mm) at a fixed wire diameter of
d = 0.6 mm.

Fig. 3. Analytical (dotted lines) and ANSYS (solid lines) simulated

results of heater temperature for different applied heater voltage and
heater wire diameters. (a) d = 0.5 mm. (b) d = 0.6 mm. (c) d = 0.7 mm.
(d) d = 0.8 mm.

Fig. 4. Temperature distribution on the heater surface.

Fig. 6. Cut view of simulated cathode model.

this code: wire diameter, heater diameter, pitch, number of

turns, resistivity, and temperature coefficient of resistivity. B. Thermal Simulation of Cathode
The output of this code comes as a graph for the voltage This section presents the thermal simulation of the cathode
dependency of the heater temperature. A good agreement with the heater. Fig. 6 shows the cut view of the simulation
between simulation and analytical calculation was observed model of the cathode. It consists of five parts made from
[Fig. 3(a)–(d)] for the different heater wire diameter and nickel and the W–Re heater, as described above. The electron
applied voltages. As expected, these analyses show that the emission surface has denoted by the red line. It is required that
heater temperature increases with the heater wire diameter the temperature of this surface should be more than 850 ◦ C due
increase because the heater current increases and resistance to the work function (1.2–1.5 eV) of triple carbonates of Ba,
decreases (P = I 2 R) for the same applied voltage. Sr, Ca, and coating material, for proper electron emission [13].
Fig. 4 shows the temperature distribution on the heater The cathode will be heated by the thermal radiation of the
surface for the applied voltage of 8 V, as an example, for heater. A surface-to-surface radiation boundary condition has
the case of d = 0.6 mm and dh = 8 mm. The maximum been defined in the ANSYS simulation. This boundary con-
temperature on the heater surface is found to be 1584 ◦ C. It is dition allows transferring heat from heater surface to cathode
quite below the melting point of tungsten and 734 ◦ C above surface and vice versa.
the required temperature of cathode. The main aim is to obtain more than 850 ◦ C on the
The temperature variation dependence on applied voltage electron emission surface for the minimum possible value
for the different heater diameters (dh ) for a fixed wire diameter of heater temperature, which depends on heater diameter,
of d = 0.6 mm is shown in Fig. 5. The heater temperature as mentioned above. Therefore, the effect of heater diameter
decreases with an increase of heater diameter due to an has been examined in terms of the temperature difference
increase of heater resistance. The cathode temperature distribu- between the cathode emission surface and the heater (Fig. 7).
tion uniformity is depending on the distance between the inner Here, all the parameters were fixed except heater diameter.
surface of cathode sleeve (dsleeve) and the heater diameter (dh ) The efficient cathode heating can be achieved by minimizing
(Fig. 6), as discussed in Section II-B. the difference in temperature between the heater and cathode

Authorized licensed use limited to: POLO BIBLIOTECARIO DI INGEGNERIA. Downloaded on June 10,2024 at 10:51:25 UTC from IEEE Xplore. Restrictions apply.
VYAS: THERMAL DESIGN OF MEGAWATT POWER-LEVEL X-BAND COAXIAL MAGNETRON 3383

Fig. 7. Temperature difference between the heater and the cathode for
different heater diameters at different applied voltages.

Fig. 9. Thermal analysis of the complete X-band magnetron without

beam collision. Temperature distribution on (a) heater, (b) electron
emission surface of the cathode, (c) cathode support, and (d) complete
magnetron for heater voltage of 8 V.

III. T HERMAL S IMULATION OF C OMPLETE M AGNETRON

There are two heating sources for the magnetron anode,
which was used for the thermal design of the complete
magnetron: one is the temperature of the cathode and the
other is electron beam collision with the anode. The thermal
analysis is carried out in two steps: without and with electron
Fig. 8. Temperature contour plot for (a) cathode sleeve, (b) electron beam collision. The estimation of maximum temperature in
emitting surface of the cathode sleeve (Fig. 6), and (c) heater.
each condition is required for the successful development of
magnetron. At the time of cathode activation, the electron
emission surfaces. This can be achieved by increasing the beam does not present, but during this time, the temperature
diameter of the heater and minimizing the gap between the of the cathode surface is more than 1000 ◦ C. This high
cathode sleeve radius and the heater diameter. The minimum temperature is required for conversion of the carbonates to
temperature requirement is elongating the heater life. oxide, e.g.,
It is found (Fig. 7) that T decreases by an increase
of heater diameter and more effective on the higher values 
BaCO3 −→ BaO + CO2
of heater voltage. Fig. 8 shows the temperature distribution 
for each part of the cathode for the case of dh = 8 mm CaCO3 −→ CaO + CO2

and d = 0.6 mm. The maximum temperature of 978 ◦ C is SrCO3 −→ SrO + CO2 .
found on the right-hand side end-hat of the cathode sleeve
[Fig. 8(a)]. This will produce 0.6% nonuniformity of the The ANSYS simulation model of the complete magnetron
emitting surface of the cathode [Fig. 8(b)]: temperature is geometry is shown in Fig. 9. The cathode is situated in the
uniform about 99.4%. This cathode operates in the space- center of the magnetron anode by the cathode support. The
charge limiting region and this temperature variation does not role of cathode support is to keep the cathode at the center
affect the performance of the cathode. It is also found that the with coaxes of the anode and provide the terminal to apply
heater temperature is increased by (8.3%) from 1584 ◦ C to the high voltage on it. One major aspect of cathode support in
1727 ◦ C [Fig. 8(c)] due to the encloser effect of the cathode the thermal design is that it should not drain the temperature
sleeve, compared to Fig. 4 where the only heater was used. from the cathode. The low thermal conductive [17 W/(m·K)]

Authorized licensed use limited to: POLO BIBLIOTECARIO DI INGEGNERIA. Downloaded on June 10,2024 at 10:51:25 UTC from IEEE Xplore. Restrictions apply.
3384 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 69, NO. 6, JUNE 2022

material—Kovar—was used as cathode support for this pur- TABLE I

pose. The surface area of cathode sleeve is 630 mm2 . H EAT T RANSFER C OEFFICIENT FOR D IFFERENT WATER F LOW R ATE
In the first step, 8 V (optimal emitting surface temperature)
of heater voltage and natural convection is applied. In this
case, the power loss due to electron beam collision is not
included. The maximum temperature on the heater and cath-
ode emission surface is found to be 1662 ◦ C and 930 ◦ C
[Fig. 9(a) and (b)]. The temperature decreases rapidly along
the length of cathode support [Fig. 9(c)] due to Kovar material
used as support, keeping the cathode temperature constant.
The emitting surface temperature of 930 ◦ C is 80 ◦ C above
the starting temperature of electron emission. This value of
temperature can be used during the magnetron operation.
However, it is 70 ◦ C below the required temperature of cathode
activation; it means that the cathode needs more power to
achieve the temperature of more than 1000 ◦ C. Therefore,
the heater voltage has been increased up to 10 V making the
cathode emitting surface temperature ∼1050 ◦ C. This ensures
that the heater voltage should be 10 V during the cathode
activation process.
Now, the power loss due to beam collision with the anode
block has been included in the simulation. The average power
loss due to beam collision for this magnetron is 1167 W
for 2 MW of output pulse power. The duty cycle of this
magnetron is 0.001. The value of power loss is obtained by the
PIC simulation. This power has been applied to the vane tip
(inner radius surface) of the magnetron in terms of heat flux
as 2.55 × 106 W/m2 . The heat flux is calculated as follows:
Area of single vane tip = 11.4 × 10−6 m 2 (3)
Total number of vanes = 40 (4)
1167W
Heat flux =
40 × 11.4 × 10−6 m 2
W
= 2.55 × 106 2 . (5)
m
The forced water cooling has been applied in terms of heat
transfer coefficient. A classic correlation for the convection
heat transfer for turbulent flow in a pipe is given by the
Dittus–Boelter equation
Nu = 0.0243 Re0.8 Pr0.6 (6)
where Nu, Re, and Pr are Nusselt, Reynolds, and Prandtl
numbers, respectively. The relation between Nusselt number
(Nu) and heat transfer coefficient (h) is given as follows:
hD Fig. 10. Thermal analysis of the complete X-band magnetron with
Nu = (7) beam collision. Temperature distribution on (a) complete magnetron and
k (b) anode block for heater voltage of 8 V, water-cooling flow of 5 L/min.
where D is the diameter of the cooling tube and k is the
thermal conductivity. Reynolds and Prandtl numbers could be
calculated by the following equations: summarized in Table I. The thermal simulation of complete
magnetron has been carried out by using the heat transfer
DV
Re = (8) coefficient values given in Table I. The remaining conditions
μ of heater voltage and heat flux are 10 V and 2.55 × 106 W/m2 ,
μC p respectively.
Pr = (9)
k Fig. 10(a) and (b) shows the thermal simulation results
where V , C p , and μ are the velocity, specific heat, and for the case of 5 L/min of water cooling. The flow rate
viscosity, respectively. The heat transfer coefficient (h) for the of 5 L/min is quite easily achievable for this type of magnetron
different flow rate of water cooling has been calculated and by a simple water pump. The maximum temperature on

Authorized licensed use limited to: POLO BIBLIOTECARIO DI INGEGNERIA. Downloaded on June 10,2024 at 10:51:25 UTC from IEEE Xplore. Restrictions apply.
VYAS: THERMAL DESIGN OF MEGAWATT POWER-LEVEL X-BAND COAXIAL MAGNETRON 3385

the outer surface of the magnetron is found at 66 ◦ C. The T = [a1 a2 0 0 a5 const];

maximum temperature of 220 ◦ C is found on the tip of the sol = roots(T);
anode. This value of temperature is far below the melting for j=1:length(sol);
point of copper and baking (vacuum processing) temperature if imag(sol(j)) == 0;
(≈450 ◦ C) of the magnetron. This thermal analysis concludes temp(i) = sol(j)-273;
that this magnetron can be used safely without any thermal end
issues. end
hold on
IV. C ONCLUSION end
Results of the thermal design of a water-cooled 2-MW plot(v,temp);
X-band magnetron have been presented. The simulation and xlabel(‘Volatage(Volt)’);
analytical values of heater temperature are found in good ylabel(‘Temperature(C)’)
agreement. It is found that the selection of Kovar material
for the cathode support helps to stop the heat drainage from
the cathode emission surface. The heat transfer coefficient of R EFERENCES
different flow rates has been calculated and given in Table I. [1] R. M. Gilgenbach, Y. Lau, H. McDowell, K. L. Cartwright, and
T. A. Spencer, “Crossed-field devices,” in Modern Microwave and
The outer surface temperature is found at around 66 ◦ C. The Millimeter Wave Power Electronics, R. J. Barker, N. C. Luhmann,
maximum temperature on the anode block is found to be J. H. Booske, and G. S. Nusinovich, Eds. Piscataway, NJ, USA: Wiley,
220 ◦ C for the case of 5 L/min of water-cooling flow, which is 2005.
[2] S. K. Vyas, R. K. Verma, S. Maurya, and V. V. P. Singh, “Review of
in the safe operating range of magnetron. The overall step-by- magnetron developments,” Frequenz, vol. 70, nos. 9–10, pp. 455–462,
step procedure for the thermal design of magnetron has been Jan. 2016, doi: 10.1515/freq-2015-0196.
presented. [3] H. Obata, K. Furumoto, and H. Miyamoto, “State of the art advanced
magnetrons for accelerator RF power source,” in Proc. LINAC,
East Lansing, MI, USA, 2016, pp. 405–407, doi: 10.18429/JACoW-
A PPENDIX LINAC2016-TUOP09.
MATLAB CODE. The following code has been written for [4] Q. Jiaqi, C. Huaibi, and T. Chuanxiang, “PIC simulation of the coaxial
solution of heat equation (2) to obtain the heater temperature: magnetron for low energy X-band linear accelerators,” in Proc. 1st Int.
Part. Accel. Conf., Kyoto, Japan, 2010, pp. 232–234.
clc; [5] J.-H. Lee, G.-J. Kim, S. Kim, Y.-S. Lee, I. S. Kim, and J.-I. Kim,
close all; “Investigation of X-band coaxial magnetron using three-dimensional
clear all; particle-in-cell simulation,” in Proc. Int. Vac. Electron. Conf. (IVEC),
Paradise Hotel Busan, South Korea, Apr. 2019, pp. 1–2, doi:
Ts=300; 10.1109/IVEC.2019.8744923.
sigma=5.67e-8; [6] M. K. Joshi, S. K. Vyas, T. Tiwari, and R. Bhattacharjee, “Optimal
alpha=input (‘input the alpha_valu = ’); design of a coaxial cavity based on quality-factor maximization for high
power coaxial magnetron in X-band,” IEEE Trans. Plasma Sci., vol. 46,
emis =input (‘input the emissivity of wire = ’); no. 3, pp. 503–510, Mar. 2018, doi: 10.1109/TPS.2018.2800168.
d=input (‘input the wire diameter = ’); [7] I. G. Herrmann and P. S. Wagener, The Oxide-Coated Cathode, vol. 1.
N=input (‘input the number of turn = ’); London, U.K.: Chapman & Hall, 1951.
[8] E. Q. Zhang, “On the magnetron cathode,” IEEE Trans. Electron
D=input (‘input the diameter of heater = ’); Devices, vol. ED-33, no. 9, pp. 1383–1384, Sep. 1986, doi: 10.1109/T-
rho=input (‘input the resistivity of heater wire = ’); ED.1986.22674.
pich=input (‘input the pitch of heater = ’); [9] M. A. Sattorov et al., “Thermal analysis of a strapped magnetron,” IEEE
Trans. Electron Devices, vol. 58, no. 8, pp. 2784–2788, Aug. 2011, doi:
hleg=input (‘input the leg length of heater = ’); 10.1109/TED.2011.2150227.
L=(N∗ sqrt((pi∗D)^2+(pich)^2))+hleg; [10] S. K. Vyas and T. Tiwari, “Study of hole and slot type S-band
R0=(4∗ rho∗ L)/(pi∗ d∗ d); pulsed power magnetron,” J. Electromagn. Waves Appl., vol. 32, no. 8,
pp. 1008–1017, May 2018, doi: 10.1080/09205071.2017.1412358.
a1 = [alpha]; [11] ANSYS. Canosberg. [Online]. Available: http://www.ansys.com
a2 = [1-(alpha∗Ts)]; [12] Communications and Power Industry (CPI), USA. [Online]. Available:
a5 = [-alpha∗(Ts^4)]; https://www.cpii.com/docs/related/2/LINAC%20FAMILY%20BROCH%
208.5%20x%2011.pdf
v=[1:20] [13] T. L. Bergman, A. S. Lavine, F. P. Incropera, and D. P. Dewitt,
temp = []; Introduction to Heat Transfer, 6th ed. Hoboken, NJ, USA: Wiley,
for i=1:20 2011.
[14] A. S. Gilmour, Klystrons, Traveling Wave Tubes, Magnetrons, Crossed-
const = (-v(i)^2/(pi∗d∗ emis∗ sigma∗ R0∗ L)) + Field Amplifiers, and Gyrotrons. Norwood, MA, USA: Artech House,
alpha∗ (Ts^5)-Ts^4; 2011.

Authorized licensed use limited to: POLO BIBLIOTECARIO DI INGEGNERIA. Downloaded on June 10,2024 at 10:51:25 UTC from IEEE Xplore. Restrictions apply.