Progress In Electromagnetics Research Symposium 2006, Cambridge, USA, March 26-29 143

Mathematical Modeling of Nonlinear Waves and

Oscillations in Gyromagnetic Structures by Bifurcation
Theory Methods

G. S. Makeeva1 , O. A. Golovanov2 , and M. Pardavi-Horvath3

1
Penza State University, Russia
2
Penza Military Institute of Artillery, Russia
3
The George Washington University, USA

Abstract—The vector field bifurcation approach and its numerical implementation for the rigorous mathemat-
ical simulation of nonlinear phenomena in microwave and mm-wave ferrite or composite semiconductor/ferrite
devices are developed. The bifurcation points of nonlinear Maxwell’s operator for the three-dimensional bound-
ary problems, stated and solved rigorously (i.e., considering the full Maxwell’s equations together with the
nonlinear equations of motion for magnetization in ferrites and transport carriers in semiconductors) are ana-
lyzed using numerical methods. The electromagnetic field is represented as decomposed into a series of weakly
nonlinear wave fields. The solutions of a linearized Maxwell’s operator matrix equation are determined. The
propagation constants of weakly nonlinear waves in waveguiding structures (WGS) or eigenfrequencies of weakly
nonlinear oscillations in resonator structures (RS) are found. Using the bifurcation dynamics of Maxwell’s equa-
tions the nonlinear wave interactions in the strongly nonlinear planar ferrite insert, loaded into strip-slot RS,
are analyzed (from the harmonic frequency terms at the ‘soft’ non-linear stage into the region of ‘hard’ non-
linearity). The nonlinear propagation of electromagnetic waves in the strip-slot ferrite RS are modeled. The
nonlinear wave phenomena, including the parametric excitation of oscillations and the wave instability process
are investigated taking into account constrained geometry WGS and RS.

1. Introduction
The research of bifurcations in nonlinear dynamical systems with distributed parameters, described by non-
linear differential equations in partial derivatives, involves serious mathematical difficulties. As for distributed
systems a characteristic determinant is an analogue of frequency characteristics, that’s why it is possible to
analyze distributed self-sustained oscillation systems using the linearization method combined with the charac-
teristic determinant analysis (at first it was shown in [1] for the one-dimensional case). Hitherto the bifurcation
analysis was used to investigate nonlinear dynamical systems with lumped parameters, described by nonlinear
ordinary differential equations (ODEs). When the ordinary differential equation is of second order a qualitative
analysis is possible on the two-dimensional phase surface [2]. The linearization method in combination with the
frequency-domain analysis is used for the analysis of self-sustained oscillating systems and automatic control
systems [2]. Determining the solutions of nonlinear differential equations in fixed points using numerical compu-
tation is a very complicated problem even for ODEs, because at the branching points qualitative modifications
of solutions can happen due to variation of parameters.
The behavior caused by the instability of waves and oscillations in nonlinear or parametric systems, con-
taining nonlinear magnetic or semiconductor media, is complex [3]. The physical theories of the instability of
magnetostatic or spin waves were developed using the approximate analysis of the equation of motion of the
magnetization vector in ferromagnet for one-dimensional structures only [4, 5]. The analysis of the transition re-
gion from the stable regime to the onset of labile oscillating mode caused the instability is the most complicated
problem. This analysis can only be based on the solutions of full nonlinear Maxwell’s equations, complemented
by the nonlinear equations of motion of the magnetization vector in a ferromagnet [3]. The goal of this pa-
per is to develop a new approach based on the bifurcation theory [6, 7] for accurate electromagnetic modeling
of nonlinear wave phenomena in gyromagnetic or semiconductor waves in waveguiding structures (WGS) or
resonator structures (RS) using a numerical approach for the analysis of the linearized matrix equation and
bifurcation points of the nonlinear Maxwell’s operator. It opens up new prospects of bifurcation analysis and
rigorous mathematical modeling of strongly nonlinear electrodynamical systems using the bifurcation dynamics
of Maxwell’s equations.
144 Progress In Electromagnetics Research Symposium 2006, Cambridge, USA, March 26-29

2. The Numerical Method of Linearization of Nonlinear Maxwell’s Operator in Combi-

nation with the Analysis of the Characteristic Determinant
The numerical method to determine the propagation constants of weakly nonlinear waves in WGS (or eigen-
frequencies of weakly nonlinear oscillations in RS) loaded with strongly nonlinear gyromagnetic or semiconductor
boundary media consists in the following.
The three-dimensional boundary problems, stated rigorously (i.e., considering the full Maxwell’s equations
with the nonlinear equation of motion for magnetization in ferrites or the equation of transport carriers in
semiconductors, with boundary conditions following from conditions of non-asymptotic radiation) was reduced
to the boundary problem for a system of nonlinear DEs together with the system of the nonlinear algebraic
equations using the cross-sections method in [8, 9].
The system of nonlinear DEs together with the system of nonlinear algebraic equations [8, 9] is represented
in a symbolic form, as:

dyi
= Fi (y1 , y2 , . . . , yn ), Ψj (y1 , y2 , . . . , yn ) = 0, (1)
dz

where i = 1, 2, . . . , m; j = m + 1, m + 2, . . . , n; yi = yi (z) are unknown functions of the longitudinal coordinate

z compiled on the functions atn (ωm ), btn (ωm ), azn (ωm ), bzn (ωm ), given in references [8, 9].
Let yi = 0 (i = 1, 2, . . . , n) be the solution of the system (1), satisfying the boundary conditions as given
in reference [8, 9]. Then the functions Fi and Ψj (i = 1, 2, . . . , m; j = m + 1, m + 2, . . . , n) identically vanish,
consequently, the solution yi = 0 (i = 1, 2, . . . , n) of the system (1) is fixed (stationary) relative to the coordinate
variable z.
As the first approximation, reduce the system of nonlinear differential equations (1) to a system of linear
differential equations. For this purpose it is necessary to represent functions Fi and Ψj by their generalized
Taylor’s series in the neighborhood of fixed (stationary) points xi = 0, and to take into account the first order
partial derivatives. This procedure results a system of linear differential equations:

n
X n
X
dyi ∂Fi (0, 0, . . . , 0) ∂Ψj (0, 0, . . . , 0)
= · yK , · yK = 0, (2)
dz ∂yK ∂yK
K=1 K=1

where i = 1, 2, . . . , m; j = m + 1, m + 2, . . . , n.
Let us represent the system of differential equations (2) in expanded form:



 a11 (z) · y1 + a12 (z) · y2 + . . . + a1n (z) · yn = y10 ,



 a21 (z) · y1 + a22 (z) · y2 + . . . + a2n (z) · yn = y20 ,



 ··· ··· ··· ··· ··· ··· ··· ··· ··· ···
 0
am1 (z) · y1 + am2 (z) · y2 + . . . + amn (z) · yn = ym ,
(3)

 a m+1,1 (z) · y 1 + am+1,2 (z) · y 2 + . . . + a m+1,n (z) · yn = 0,



 a m+2,1 (z) · y 1 + am+2,2 (z) · y 2 + . . . + a m+2,n (z) · y n = 0,



 · · · · · · · · · · · · · · · · · · · · · · · · · · · · · ·

an1 (z) · y1 + an2 (z) · y2 + . . . + ann (z) · yn = 0,

where the coefficients aij (z) (i, j = 1, 2, . . . , n) compiled on the partial derivatives from (2). The system of
equations (3) can be represented in matrix form as:

dy
A·y = (4)
dz

dy
where y is the vector with components y1 , y2 , . . . , ym ; 0
is the vector with components y10 , y20 , . . . , ym ;
dz
Progress In Electromagnetics Research Symposium 2006, Cambridge, USA, March 26-29 145

A = A11 − A12 · A−1

22 · A21 ,
   
a11 a12 . . . a1m a1m+1 a1m+2 . . . a1n
 a21 a22 . . . a2m   a2m+1 a . . . a2n 
A11 = 
 ·
,
 A12 = 

2m+2 ,

· · · · · · ·
am1 am2 . . . amm am,m+1 am,m+2 . . . amn
   
am+1,1 am+1,2 . . . am+1,m am+1,m+1 am+1,m+2 . . . am+1,n
 am+2,1 am+2,2 . . . am+2,m   am+2,m+1 am+2,m+2 . . . am+2,n 

A21 =  , 
A22 =  ,
· · · ·  · · · · 
an1 an2 ... anm an1 an2 ... ann

We find the partial solutions of the system of equation (4) in the form:

y = α · eλ·z , (5)

where α is the vector with components α1 , α2 , . . . , αm . Substituting (5) into (4), we obtain the following
eigenvalue matrix equation:
A ·α = λ ·α, (6)
where λ and α are correspondingly the eigenvalues and eigenvectors of matrix A. Using a numerical method
(for example, the QR-algorithm) to solve the matrix equation (6) the eigenvalues λm and eigenvectors α of A
can be determined.
The solutions (5) of the linearized Maxwell’s operator (6) are treated as weakly nonlinear waves. The
electromagnetic field in WGS is decomposed into a series of weakly nonlinear wave fields. The eigenvalues λm
of matrix A are the propagation constants of the weakly nonlinear waves in WGS (or the eigenfrequencies of
weakly nonlinear oscillations in RS). The components of the eigenvectors α of matrix A are the transverse and
longitudinal components of weakly nonlinear waves .
The computational algorithm, using the linearization of nonlinear Maxwell’s operator and the decomposition
into a series of weakly nonlinear wave fields, is more complex than those for the propagation constants and fields
of eigenwaves of WGS, filled with a linear medium. But the convergence of this algorithm and its stability for
rounding errors is better. It permits to solve the threedimensional diffraction boundary problems for WGS or
RS loaded with strongly nonlinear gyromagnetic or semiconductor insertions having sizes comparable to the
wavelength. This is important for CAD of prospective ferrite or composite semiconductor/ferrite devices at
microwave or mm-waves.

3. Numerical Simulation of the Parametric Excitation of Oscillations in Nonlinear Gy-

romagnetic Structure Using Bifurcation Points
The rigorous mathematical modeling of parametric oscillations in strip-slot RS loaded with a planar magne-
tized ferrite (Fig. 1) is based on solving the nonlinear diffraction boundary problem by the crosssections method
of [8], using the decomposition algorithm on nonlinear autonomous blocks [10].
For the computational algorithm the transverse and longitudinal components of weakly nonlinear waves
are used. It results a stable and computationally efficient algorithm for computing the instability of waves or
oscillations in WGS or RS containing strongly nonlinear gyromagnetic media.
There are two incident electromagnetic waves: the signal wave of frequency ω1 and the pumping wave of
frequency ω2 are incident on the input cross-sections S1 of RS (Fig. 1). The waves are the fundamental and
+ +
higher-order modes of strip-slot WGS, having magnitudes Cn(α) (ω1 ) and/or Cn(α) (ω2 ), where α is the index of
the cross-sections, n are the indices of eigenwaves of strip-slot WGS [8, 9].
The instability of parametric excitation process of oscillations in ferrite RS depending on the bifurcation
parameters is simulated using the numerical method of bifurcation points analysis, developed by us [11]. The
results of computing of the instability regions for parametric excitation of oscillations in ferrite RS by the inci-
+
dent pumping wave, depending on the magnitude C2(1) (ω2 ) and the normalized frequency (the signal frequency
ω1 with respect to the pumping frequency ω2 ) are shown in Fig. 2. The onset and the breakdown of parametric
oscillations caused the wave instability in nonlinear ferrite structure in the neighborhood of bifurcation param-
eters were simulated into the region of ‘hard’ nonlinearity taking into account constrained geometries RS, and
it is represented in Fig. 2.
146 Progress In Electromagnetics Research Symposium 2006, Cambridge, USA, March 26-29

It follows from the results of the mathematical modeling that the unstable regions for parametric excitation
of oscillations in ferrite RS are near the values of the eigenfrequencies of fundamental and higher-order modes
+
of oscillations of the strip-slot line resonator: ω1 = mω2 /2, m = 1, 2, 3, . . . The threshold magnitude C2(1) (ω2 )
+
is rising steeply as m increases. The minimum threshold of C2(1) (ω2 ) is given by ω1 = ω2 /2.

Figure 1: Resonator structure with nonlinear ferrite insert: 1, 2, 3, 4 – coupled strips of strip-slot WGS; 5
– strip-slot resonator; 6 – planar magnetized ferrite insert (ε = 9, H0 = 278 A/mm; M0 /µ0 = 160 A/mm;
ωr = 3 ∗ 109 Hz; β = 45◦ ); 7 – dielectric substrate (ε = 9; µ = 1); 8 – point of field observation; f1 = 5 Hz,
f2 = 10 GHz; all sizes are in mm.

Figure 2: Instability regions for parametric excitation of oscillations in nonlinear ferrite RS, depending on
+
bifurcation parameters: C2(1) (ω2 ) – magnitude of incident pumping wave; ω1 – eigenfrequency of fundamental
modes of oscillations the strip-slot line resonator (length of the resonator = half-wave for signal wave at f1 =
5 GHz); ω2 – frequency of pumping wave.

4. Conclusion
Using the achievements of modern mathematics in the area of vector field bifurcation theory opens new
possibilities for computer analysis of the onset of nonlinear waves in WGS with bounded gyromagnetic media
having a strong nonlinearity. This approach has a high likelihood of success in investigating nonlinear phenom-
ena in new microwave/millimeter-wave ferrite devices [12] for frequency multiplexing/filtering, limiters, noise
rejectors, signal-noise ratio enhancers, and pulse compressing devices.
Progress In Electromagnetics Research Symposium 2006, Cambridge, USA, March 26-29 147

REFERENCES
1. Landa, P. S., Self-oscillations in Distributed Systems, Moscow, Nauka, 320, 1982.
2. Andronov, A. A., A. A. Vitt, and S. E. Haikin, The Theory of Oscillations, Moscow, Nauka, 568, 1981.
3. Gurevich, A. G. and G. A. Melkov, Magnetic Oscillations And Waves, Nauka, Moscow, 1994.
4. Su, J., S. A. Nikitov, R. Marcelli, and P. De Gasperis, “Parametric and modulation instabilities of magne-
tostatic surface spin waves in ferromagnetic films,” J. Appl. Phys., Vol. 81, 1341–1347, 1997.
5. Nazarov, A. V., C. E. Patton, R. G. Cox, L. Chen, and P. Kabos, “General spin wave instability theory for
anisotropic ferromagnetic insulators at high microwave power levels,” J. Magnetism and Magnetic Materials,
Vol. 248, 164–180, 2002.
6. Krasnoselskiy, M. A., Functional analysis, Nauka, Moscow, 1964.
7. Vainberg, M. N. and V. A. Trenoguin, Theory of Nonlinear Equations Solutions Branch Points, Nauka,
Moscow, 1969.
8. Makeeva, G. S. and O. A. Golovanov, “Accurate electromagnetic modeling of nonlinear wave interactions in
ferrite included in strip-slot waveguiding structures,” Proc. 9-th Int. Conf. Ferrites (ICF-9), San Francisco,
California, USA, 671–676, 22–27 Aug. 2004.
9. Makeeva, G. S., O. A. Golovanov, and M. Pardavi-Horvath, “Rigorous nonlinear analysis of propagating
waves in the strip-slot waveguiding structure with a strongly nonlinear semiconductor discontinuity progress
in electromagnetics research symposium,” PIERS 2005, Hangzhou, China, 22–26 Aug. 2005.
10. Golovanov, O. A., “Nonlinear autonomous blocks and their usage for studying irregular waveguides and
resonators with nonlinear media,” Izvestiya vysshikh uchebnykh zavedenii. Radiofizika, Vol. 33, No. 7, 793–
804, 1990.
11. Makeeva, G. S., O. A. Golovanov, and M. Pardavi-Horvath, “Numerical analysis of wave instability in
strongly nonlinear ferrite structures using bifurcation points of the nonlinear maxwell’s operator,” Progress
in Electromagnetics Research Symposium, PIERS 2005, Hangzhou, China, 22–26 Aug. 2005.
12. Marcelli, R. and S. A. Nikitov, Eds, “Nonlinear microwave signal processing: towards a new range of
devices,” NATO ASI Series, 3. High Technology, Kluwer Academic Publishers, Dordrecht, 20, 509, 1996.