Progress In Electromagnetics Research Symposium, Cambridge, USA, July 2–6, 2008 425

Particle-in-cell Simulation of a Novel High Power Terahertz Vacuum

Electron Device
Hai Zhang1 , Jianguo Wang1, 2 , and Changjiang Tong2
1
School of Electronics and Information Engineering, Xi’an Jiaotong University, Xi’an 710049, China
2
Northwest Institute of Nuclear Technology, Xi’an 710024, China

Abstract— New micromachining techniques now provide the technology to fabricate the vac-
uum electron devices with dimensions suitable for operation in the terahertz region of the elec-
tromagnetic spectrum. In this paper, results of theoretical and numerical simulation studies of a
MW-class, large diameter terahertz backward wave oscillator are presented. The device consists
of a large cross-section (overmoded), slow wave structure with a unique profile of wall radius
specifically designed to support surface wave and to provide a strong beam-wave coupling at
moderate voltage. A novel full electromagnetic Particle-in-cell simulation tool UNIPIC is also
developed. With this method, we investigated the characteristics of the device at the frequency
of 0.14 THz. It is found that the steady-state, single-frequency operation in this overmoded sit-
uation could be realized by the property of surface wave at “π” point. And the comparison of
computational results between UNIPIC and KARAT demonstrates that UNIPIC is a reliable and
promising software for the design of high power terahertz vacuum electron devices.

1. INTRODUCTION
“Terahertz (THz) fields” is a generic term for the electromagnetic waves within frequencies between
0.1 and 10 THz [1]. A wide variety of scientific and commercial applications, such as imaging of
biological tissue, spectroscopic identification of complex molecules, broad-band communications,
which rely on the physical properties of terahertz waves are currently under investigation [2].
These systems all require THz frequency power sources which, ideally, are compact and inexpensive.
Recent advances in micromachining technologies suggested that it was a “natural” route to fabricate
small structures, such as cavities, waveguides, and slow wave structures (SWS) that might be used
in a scaled-down vacuum electron tube operating at THz frequencies.
In order to achieve and optimize the performance of these THz vacuum electron devices (VEDs),
apart from a detailed understanding of its operating principle, it is also necessary to develop
accurate simulation tools where the main peculiarities derived from operating conditions and the
device dimensions are taken into account. Although there are many available codes devoted to the
simulation of VEDs, they are limited in their application to terahertz frequency devices. Therefore,
we have recently developed a Particle-in-Cell (PIC) simulation software UNIPIC, which is based
on the Monte-Carlo algorithm and is specifically designed for the analysis and optimization of
micromachined VEDs. In this paper, a 140 GHz MW-Class backward wave oscillator (BWO, one of
the most important and promising devices of THz VEDs) was designed and investigated using this
software, and the simulation results were compared with that of KARAT (a traditional, famous
PIC code in HPM field) under the same operating conditions.

2. PHYSICAL CONSIDERATIONS
2.1. Operation Principle of THz BWO
The BWO consists essentially of an electron beam confined radically by a strong longitudinal
magnetic field and propagating axially through a cylindrical resonant cavity containing SWS, which
provides a set of electromagnetic wave modes with phase velocities parallel to the beam less than the
speed of light. These slow waves interact resonantly with the slower space-charge wave supported by
the beam, which leads to an instability that transfers energy from the beam to the electromagnetic
field. It is characterized by the fact that the SWS mode involved in the resonance has a negative
group velocity; accordingly, the wave energy transfer is backward along the beam, with a Poynting
vector antiparallel to the beam velocity. The structural configuration of BWO system is shown in
Fig. 1, and a spatially periodic structure with the rectangular profile was employed as the SWS
in BWO, because it can supply both slow waves and strong coupling, and even can provide the
fabrication convenience in the millimeter range.
426 PIERS Proceedings, Cambridge, USA, July 2–6, 2008

(a) (b)

Figure 1: (a) Physical configuration of BWO, where parts 1 to 3 represent the outer-shell related components;
part 4 is the kernel of BWO, rectangular SWS; parts 5 is the drift tube; parts 6 and 7 denote the focusing
magnetic system; parts 8 to 12 constitute the foilless diode system of electron source; and part 13 is connected
to the S-5N electron accelerator. (b) Model of SWS with rectangular rippled wall.

2.2. Characteristics of the SWS

For a TM01 mode propagating in the SWS, an approximate relation [3] between the maximum
power Pmax and the maximum strength of electric field allowable at the wall, Emax , can be given
by
µ ¶ r ³υ ´
Emax · λ 2 π 2 0 4 0, 1 2 1
Pmax = 8.707 σ 1− 2 (1)
511 2 2πD0 υ0, 1

where, D0 = πD/λ, D is the transverse diameter, λ is the free-space wavelength, and v0, 1 is the
first root of the zero-order Bessel function. From Eq. (1), we can find that the output power Pmax
is dramatically increased with the enlargement of D. If D/λ ≥ 1.76, the SWS is referred to as an
overmoded SWS.
In this paper, the transverse diameter of SWS is designed to be several times of the free-space
wavelength λ (D/λ ∼ = 3) thereby reducing the internal field stress for the same power flow, and
in other words, it can efficiently increase the power-handling capacities of the high power THz
devices. However, in this overmoded device there might simultaneously exist multiple eignmodes
for the same operating frequency, which could lead to the modes competition and the resulting
power decay.
One useful approach to solve this problem is to shift the operating point close to the upper edge
of pass band (π point) because of the following two reasons. First, in this region, the Q factor is
relatively large due to both large reflection and small group velocity, which can efficiently decrease
the starting current for oscillation. Secondly, operating in the near π mode, the device could behave
as a pure surface wave generator, where only the lowest mode (TM01 ) is excited on benefit of the
relatively large coupling impedance (or high efficient beam-wave interaction) compared with that
of other volumetric wave modes. Fig. 2 shows the dispersion curves of three main competitive
modes as well as the electron Doppler line corresponding to 550 keV beam energy near the π
point. The values of structural parameters of SWS in the calculation were set as follows. Inner
radius of cylindrical waveguide equals 3.0 mm; depth of each slot is equal to 0.2 mm; length of each
spatial period equals 0.9 mm. It can be seen that the operating frequency (point of intersection)
approximately equals 0.145 THz under this condition, and according to the further calculation from
the diagram, we found that the voltage range from 450 kV to 610 kV was necessary to excite the
oscillations with frequencies above 0.14 THz.
Progress In Electromagnetics Research Symposium, Cambridge, USA, July 2–6, 2008 427

3. SIMULATION TOOL
3.1. General Characteristics of UNIPIC
UNIPIC is a novel simulation programme that can be configured at run time by selecting appro-
priate geometry, material properties, boundary conditions, field algorithms, particle algorithms,
and output specifications. Default configurations provide good accuracy in reasonable time for the
novice who does not wish to or may not know how to specify some aspects of an electromagnetic
PIC simulation. The basic electromagnetic computational processing cycle is shown in Fig. 3.

240
220
550keV
200
180
160
Frequency(GHz)

TM03
140
120
100 TM02

80
60 TM01
40
Light Doppler line
20 Electron Doppler line
0
0.0 0.5 1.0 1.5 2.0

Normalized wave number , β L/ π

Figure 2: Dispersion diagrams of there competitive Figure 3: Computational flow chart of UNIPIC.
modes.

The fields in Maxwell’s equations are represented on a finite-difference grid. From an alternative,
engineering viewpoint, the grid creates a circuit element at each location in space, which is coupled
to elements at neighboring locations. UNIPIC can use a variety of field algorithms with a range
of speed, stability, and numerical smoothness; when advantageous, it employs the engineering
viewpoint to provide special models that treat material properties and complex, fine-structure
geometric details.
Particles are represented using the PIC approach, in which a computational particle (referred
to as a macro particle) represents a large number of physical particles of the same species (e.g.,
electrons). A variety of algorithms for emission processes, particle kinematics, and current density
allocation are available, all based upon the same mathematical foundation, but each individually
optimized for different density, velocity, and field strength regimes.
Material models provide accurate treatment of the complexities of real systems and can include
both field and particle effects. Some quantum effects are modeled using phenomenological models.
Material models are designed to be compatible with all field and particle algorithm combinations.
3.2. Volume-weighting Cloud-in-cell Technique of UNIPIC
Early PIC methods along with a zero-order weighting model usually produced high noises, when a
particle was passing through a cell boundary. Birdsall introduced an area-weighting cloud-in-cell
(AW-CIC) model, which gives a first order accuracy in a Cartesian coordinate system to reduce
the noise to an acceptable range. In 2.5D simulations of Cartesian coordinate system, the AW-CIC
model works very well. However, since most of the vacuum electron tubes are in axially symmetric
shapes, like the BWO we discussed in this paper, 2.5D simulations of these devices have to be done
in the z-r axially symmetric plane of cylindrical system, where the accuracy of AW-CIC decreases,
especially on the points near the axis.
Therefore, a volume-weighting cloud-in-cell (VW-CIC) model was developed to implement par-
ticle simulation in a cylindrical system. The main highlight of this algorithm is that both the
charge assignment on the four adjacent points and the current determination along cell edges can
be calculated by the volume-weighting method, which can give a second order accuracy superior to
that of area- weighting algorithm. The VW-CIC model is incorporated in the UNIPIC code and
the more details were discussed in another contribution [4].
428 PIERS Proceedings, Cambridge, USA, July 2–6, 2008

4. PARTICLE-IN-CELL SIMULATION AND RESULT ANALYSIS

The simulation is performed by using the full electromagnetic PIC software UNIPIC. The simulation
model is shown in Fig. 4, where proper boundary conditions were set as follows. The rectangular
rippled-wall waveguide is perfect conductor, and the z axis performs as the azimuthally symmetric
axis. The electromagnetic wave at the right end of the structure was truncated by using the
convolutional perfectly matched layer [5]. As the initial condition for the simulation, there were no
electromagnetic fields in the BWO, and the electron beam was just incident at the left-end side of
the structure. The values of the structural parameters were set as mentioned above, and 20 periods
were used in the simulation.

Figure 4: Simulation model of BWO.

Under the condition of 500 kV beam voltage, 1 kA beam current and 4 T axial guiding magnetic
field, the oscillator began to work. The behavior of electron bunching in the device can also be
observed in Fig. 4. This phenomenon can be understood as follows. As the uniform electron beam
propagates through the tube, it reacts with the electromagnetic wave sequentially. Due to the
different phase effect, some of the electrons are accelerated by the field while other electrons are
decelerated till the bunching is formed. However, the average effect of this motion is that the beam
is totally decelerated and gives out its kinetic energy to the field. This is because the center of each
bunching is appropriately kept in the negative phase of wave all long time during the operation
although it slowly moves toward the positive phase. Therefore, the electromagnetic wave is excited
and the useful output power is generated.
Figure 5 shows the temporal behavior of the magnetic field component at a certain point in the
BWO as well as the corresponding Fourier transform of this signal. We can find that the signal
began to grow up at about 0.7 ns, and became saturated at 1.5 ns. After this time, the device
operated in a steady state at the frequency of about 0.143 THz.

Bp(100 ,20),Ave=-3.73 e- 005 FFT_Bp(100,20),Ave =-3.73e- 005

9.0
(E-2)
(E-1)

1.50
8.0
1.00 7.0
6.0
FFT_Bp/T

0.50
Bp / T

5.0
0.00 4.0

-0.5 3.0
2.0
-1.0
1.0
-1.5 0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0
t/s (E-9) f/Hz (E11)
(a) (b)

Figure 5: (a) Time plot of the generated THz signal, (b) spectrum of the generated signal.
Progress In Electromagnetics Research Symposium, Cambridge, USA, July 2–6, 2008 429

The 3D distribution pattern of the electric field as well as the magnetic field in the device at
3.0 ns are demonstrated in Fig. 6. It can be seen that both fields have its maximal value on the
surface of SWS (R = 3.0 mm), and gradually declines along the radial direction. This characteristic
proves that the device well operated in the TM01 mode and possessed the special property of surface
wave.

(a) (b)

Figure 6: (a) Electrical field distribution in the device, (b) Magnetic field distribution in the device.

The influence of period number (corresponding to the total length of SWS) on output power,
for two different voltages, 500 kV and 600 kV, is shown in Fig. 7. For each voltage, simulations were
respectively performed using two different programs, UNIPIC and KARAT, in order to compare
the computational results from each program. With the same method, the effect of external applied
magnetic field on the device behavior is illustrated in Fig. 8. It can be seen that the simulation
results of UNIPIC agreed very well with that of KARAT, which is a famous, reliable PIC simulation
software. Thus, we can conclude that the UNIPIC has the same computation precision as KARAT,
and it can be helpful and reliable in the design of high power terahertz electron devices.
55 55

50
50
45
45
40
40 35
Power(MW)

Power(MW)

35 30

25
30
20
25
15 UNIPIC:60 0kV
20 UNIPIC:600kV KARAT:600kV
10
KARAT:600 kV UNIPIC:50 0kV
15 UNIPIC:500kV 5 KARAT:500kV
KARAT:500kV
10 0
16 17 18 19 20 21 22 23 24 25 26 27 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5

Period Number Magnetic field(Tesla)

Figure 7: Dependence of output power on number Figure 8: Dependence of output power on guiding
of periods. magnetic field.

5. CONCLUSION
In this paper, we presented the theoretical analysis and numerical simulation results of an over-
moded, MW-class terahertz backward wave oscillator. It was found that an overmoded interaction
BWO must support surface waves that are synchronous with the electron beam, and simultane-
ously exhibit large values of the coupling impedance. The characteristics of surface wave and “π”
point operation are crucial to avoid mode competition. Furthermore, a novel full electromagnetic
430 PIERS Proceedings, Cambridge, USA, July 2–6, 2008

PIC software UNIPIC is developed, and the comparison of simulation results between UNIPIC and
KARAT proves that UNIPIC is a reliable, practical and promising software which must ease the
complicated design process of high power terahertz electron devices in future.
REFERENCES
1. Dragoman, D. and M. Dragoman, “Terahertz fields and applications,” Progress in Quantum
Electronics, Vol. 28, No. 1, 1–66, 2004.
2. IEEE Trans. Microwave Theory Tech., (Special Issue), Vol. 48, Apr. 2000.
3. Vlasov, A. N., A. G. Shkvaruntes, J. C. Rodgers, et al., “Overmoded GW-class surface-wave
microwave oscillator,” IEEE Trans. on Plasma. Sci., Vol. 28, No. 3, 550–560, 2000.
4. Li, Y. D., F. He, and C. L. Liu, “A volume-weighting cloud-in-cell model for particle simulation
of axial symmetric plasma,” Plasma Science & Technology, Vol. 7, No. 1, 2653–2655, 2005.
5. Wang, J. G., Y. Wang, and D. H. Zhang, “Truncation of open boundaries of cylindrical waveg-
uides in 2.5-Dimensional problems by using the convolutional perfectly matched layer,” IEEE
Trans. on Plasma. Sci., Vol. 34, No. 3, 681–690, 2006.