See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/234896580

Permeability tensor of magnetized ferrites at microwave frequencies: A

comparison between theory and experiment

Article  in  Journal of Applied Physics · May 2003

DOI: 10.1063/1.1555393

CITATIONS READS

7 759

3 authors, including:

Stéphane Mallégol P. Queffelec

Thales Group Université de Bretagne Occidentale
77 PUBLICATIONS   300 CITATIONS    174 PUBLICATIONS   1,872 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Electromagnetic Modeling and Characterization of Anisotropic Ferrite Materials for Microwave Isolators/Circulators View project

Metamaterials for antennas View project

All content following this page was uploaded by P. Queffelec on 27 May 2014.

The user has requested enhancement of the downloaded file.

JOURNAL OF APPLIED PHYSICS VOLUME 93, NUMBER 10 15 MAY 2003

Ferrites and Garnets Gerald Dionne, Chairman

Permeability tensor of magnetized ferrites at microwave frequencies:

A comparison between theory and experiment
Stéphane Mallégol,a) Patrick Quéffélec, and Marcel Le Floc’h
LEST, UMR-CNRS 6165, 6 Av. Le Gorgeu, BP 809, 29285 Brest Cedex, France
共Presented on 13 November 2002兲
The permeability tensor of partially magnetized ferrites or composite materials, given by a recent
mathematical model developed in our laboratory, is compared with measured one over the 50
MHz– 6 GHz frequency range. The characterization method used to measure the tensor components
␮, ␬ is based on the broadband determination of the S parameters of a nonreciprocal strip
transmission line. The quasistatic electromagnetic analysis of the measurement cell permits the
expression of ␮ and ␬ analytically from the S parameters. The theoretical approach used to calculate
␮ and ␬ is based on the extension of the effective medium approximation 共EMA兲 to the case of
anisotropic magnetic materials. The interactions between magnetic domains and the hysteresis
phenomenon are taken into account in our calculations. The good agreement observed between
simulated and measured ␮ and ␬ data for polycrystalline ferrites and ferrimagnetic or ferromagnetic
loaded composites provides an experimental validation of the model. © 2003 American Institute
of Physics. 关DOI: 10.1063/1.1555393兴
I. INTRODUCTION fore, highly dispersive permeability of ferrite materials can-
not be satisfactorily resolved with those methods. For broad-
Over the years, ferrites in a partially magnetized state band measurements, an X-band rectangular waveguide
have found important applications in the area of microwave technique was first developed in our laboratory.7 A numerical
circuits for electronic and communication systems. Their an- optimization procedure is needed to solve the inverse prob-
isotropic properties are used in the design of components lem and then to obtain the diagonal ␮ and off-diagonal ␬
where nonreciprocity or tunability is needed 共circulator, iso- permeability tensor components of the rectangular magnetic
lator, filter, etc.兲. To optimize performances of such a device, test sample. Their calculation times are then important. To
the exact knowledge of the permeability tensor components avoid this drawback, we have worked out a new character-
of the constituent specimen is required. ization method based on a nonreciprocal strip transmission
Many theoretical approaches have been described to line.8 It enables one to analytically determine the terms ␮
characterize the permeability tensor. Existing models for un- and ␬ of magnetized rectangular magnetic samples from the
saturated ferrites are emperical1 or do not simultaneously S parameters of the loaded measurement cell, i.e., with the
provide all the tensor terms.2,4 Moreover, several simplifica- magnetic material, over the 50 MHz– 6 GHz frequency band.
tions are necessary to perform calculations. Indeed, these This article presents an experimental validation of the
theories assume either a nonrealistic magnetic domains above permeability tensor model5 through the comparison of
configuration2,3 or neglect interactions between domains.4 In theoretical ␮ and ␬ results with measured ones by using this
other words, they are not predictive. A predictive model new experimental technique.8 The presence of porosity and
based on the extension of the effective medium approxima- nonmagnetic volume fractions will be taken into account in
tion 共EMA兲 to the case of anisotropic magnetic materials was the model in order to precisely characterize polycrystalline
recently proposed by our research team.5 It provides the ferrites and magnetic composites.
whole tensor terms of unsaturated polycrystalline ferrites or
magnetic composites in a single phase. The model considers
II. THEORETICAL AND EXPERIMENTAL
the shape of the magnetic domains or particles, their interac-
DETERMINATION OF ␮ AND ␬
tions, the dilution of the magnetic matter, and the hysteresis
phenomenon. A. Mathematical model
In addition, experimental techniques giving the perme- The model described in Ref. 5 gives the permeability
ability tensor of magnetized ferrites at microwave frequen- tensor of heterogeneous magnetic materials in the unsatur-
cies have been implemented. Resonant methods6 offer the ated state. Calculations start with choosing the material mag-
highest accuracy in the magnetic parameters-determination. netization state (M /M s , where M s is the saturation magne-
But, they only allow single frequency measurements. There- tization兲. This fixes the distribution of the magnetic domains
magnetization direction over space. The EMA is then used to
a兲
Electronic mail: [email protected] determine the effective tensor components when a limited

0021-8979/2003/93(10)/7471/3/$20.00 7471 © 2003 American Institute of Physics

Downloaded 12 May 2003 to 193.54.246.42. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp
7472 J. Appl. Phys., Vol. 93, No. 10, Parts 2 & 3, 15 May 2003 Mallégol, Quéffélec, and Le Floc’h

number of parameters are known. These are M s , effective

anisotropy field in domain (H eff), as well as the damping
factor 共␣兲, and the domain shape 共sphere or cylinder兲. In this
study, H eff is deduced from the applied dc magnetic field
(H 0 ), and the anisotropy field (H a ) via the internal energy
minimization principle in each domain.9 The demagnetizing
effects are considered in our calculations by introducing in
the homogenization law a demagnetizing tensor linked to the
domain shape tensor.5
The EMA enables one to advantageously take account of
magnetic particles concentration 共q兲 in the material. If so, the
self-consistency of the model is reinforced. It, indeed, allows
FIG. 1. Cross section of the measurement cell containing the magnetic
to characterize polycrystalline ferrite and ferrimagnetic or
sample under test and porous titania (TiO2 ) sample.
ferromagnetic loaded composite through the consideration of
porosity and nonmagnetic volume fractions, respectively. In
this case, the theory accounts for physical phenomenon in
共⫽5 mm兲, and ␧ its relative permittivity. ␥ ⫹ and ␥ ⫺ denote
heterogeneous media, e.g., a shift of the gyromagnetic reso-
the propagation coefficients for the forward and backward
nance frequency toward higher values further to the reduc-
senses of propagation. They come from S parameters of the
tion in magnetic particles concentration 共q兲.
loaded cell by using analytical relations.8

B. Noniterative measurement method of ␮ and ␬

III. EXPERIMENTAL VALIDATION OF THE MODEL
We recently developed an experimental method permit-
A. Experimental setup
ting one to determine the tensor components ␮, ␬ of mag-
netic materials using no iterative technique.8 The measure- An HP 8720A network analyzer provides S parameters
ment cell is constituted by a rigid conducting strip and lower of the entire cell. The first step in the measurement process is
ground plane 共Fig. 1兲 with a thickness of 0.4 and 8.76 mm, to reduce systematic errors by using a SOLT calibration
respectively. To adapt the empty cell, i.e., with no sample to procedure.8 The S parameters of the region of the cell loaded
a characteristic impedance of 50 ⍀, the width 2b of the by the samples are then deduced from measured S param-
conducting strip is 9 mm and the distance between it and the eters of the entire cell after the appropriate phase corrections
ground plane is 1.8 mm. The rectangular magnetic sample to have been applied. From that time, Eqs. 共1兲 and 共2兲 of ␮ and
be tested is placed between the strip and the ground plane. ␬ are employed to determine the permeability tensor of the
This region of the cell is, indeed, the one where the micro- test sample.
wave energy is located. The nonreciprocal behavior of the The loaded cell is inserted in the air gap of an electro-
loaded cell is required in order to measure enough indepen- magnet to magnetize the sample under test. A Hall effect
dent S parameters to obtain ␮ and ␬. Two steps are needed to Gaussmeter obtains H 0 strength. The electromagnet we used
ensure nonreciprocity (S 12⫽S 21). First, the cell must be provides a homogeneous bias field in a wide region of its air
asymmetrically perturbed. To do so, a 5⫻5⫻1.8 mm3 rect- gap, especially in the one where the test sample is located.
angular piece of porous titania (TiO2 ) of relative permittivity For example, if the strength of H 0 is 500 Oe at the center of
(␧ d ) equal to 15.5 is set at one edge of the strip and the air is the electromagnet gap, it varies between 499.8 and 500.3 Oe
used at the other edge 共Fig. 1兲. Next, a dc magnetic field H 0 in a 20-mm-diameter circle centered at the center of the gap.
is applied along the y direction of a Cartesian coordinate Measured data are processed by a personal computer as-
system. Thus, a field displacement phenomenon10 comes out sociated with the network analyzer.
along the x axis.
By considering a new transmission-line parameter 共the
characteristic memductance兲11 and assuming magnetic walls IV. RESULTS AND DISCUSSIONS
at both edges of the strip, we analytically expressed the terms In this section, the diagonal ␮ ⫽ ␮ ⬘ – j ␮ ⬙ and off-
␮, ␬ of the magnetic sample to be tested by8 diagonal ␬ ⫽ ␬ ⬘ – j ␬ ⬙ effective permeability tensor terms
␮共 ␻ 兲 given by the above model are compared with experimental
ones. Various magnetic materials of well-known intrinsic
2a• ␥ ⫹ • ␥ ⫺
⫽ , properties were tested. For theoretical results, we assumed
␮ 0 ␧ 0 ␻ 2 关共 b⫺a 兲 • 共 1⫹␧ d 兲 ⫹2a•␧ 兴 ⫺2 共 b⫺a 兲 • ␥ ⫹ • ␥ ⫺ cylinder-shaped-magnetic domains.
共1兲
Figure 2 shows a comparison between theory and experi-
␬共 ␻ 兲 ment for a 5⫻5⫻1.8 mm3 polycrystalline ferrite sample
关共 b⫺a 兲 • ␮ 共 ␻ 兲 ⫹a 兴 • 共 ␥ ⫹ ⫺ ␥ ⫺ 兲 (4 ␲ M s ⫽1.5 kG, ␧⫽13.2) submitted to a dc magnetic field
⫽ , 共2兲 H 0 of 0.75 kOe. The input parameters of the calculation code
␮ 0 ␧ 0 ␻ 2 共 1⫺␧ d 兲 •a 共 b⫺a 兲
associated with the model are: f m ⫽4.2 GHz, f eff
where ␮ 0 and ␧ 0 are permeability and permittivity of free ⫽1.56 GHz, M /M s ⫽0.5, ␣ ⫽0.175, and q⫽0.9 with f m
space, ␻ is the radian frequency, 2a the test sample width ⫽ ␥ g .4␲ M s and f eff⫽␥g .Heff .␥g is the gyromagnetic ratio

Downloaded 12 May 2003 to 193.54.246.42. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp
J. Appl. Phys., Vol. 93, No. 10, Parts 2 & 3, 15 May 2003 Mallégol, Quéffélec, and Le Floc’h 7473

FIG. 3. Effective permeability tensor components ␮, ␬ vs frequency ( f m

FIG. 2. Effective permeability tensor components ␮, ␬ vs frequency ( f m
⫽2 GHz, f eff⫽0.95 GHz, M /M s ⫽0.475, ␣ ⫽0.22, q⫽0.7). Comparison
⫽4.2 GHz, f eff⫽1.56 GHz, M /M s ⫽0.5, ␣ ⫽0.175, q⫽0.9). Comparison
with measurements for a 5⫻5⫻1.8 mm3 ferrite loaded composite sample
with measurements for a 5⫻5⫻1.8 mm3 polycrystalline ferrite sample
(4 ␲ M s ⫽0.72 kG, q⫽0.7), partially magnetized (H 0 ⫽0.4 kOe): 共a兲 ␮;
(4 ␲ M s ⫽1.5 kG), partially magnetized (H 0 ⫽0.75 kOe): 共a兲 ␮; 共b兲 ␬.
共b兲 ␬.

The accord between simulated and measured ␮ and ␬,

( ␥ g ⫽2.8 MHz/Oe). As mentioned in Sec. II A, H eff has been showed in this study for partially magnetized polycrystalline
previously calculated by using the Stoner–Wohlfarth model.9 ferrites 共Fig. 2兲 and ferrite loaded composites 共Fig. 3兲, was
We notice that simulated and measured ␮ and ␬ points agree also observed for other samples. It experimentally highlights
with the whole exploited frequency range. For example, the the validity of the model we developed in the characteriza-
theoretical ␮ ⬙ value is 2.59 at gyromagnetic resonance fre- tion of such materials. However, the damping factor 共␣兲 en-
quency whereas the experimental one is 2.66 at the same tered in the model is stronger than the one usually deduced
frequency 关Fig. 2共a兲兴, with a relative uncertainty of 2.63%. from the measured resonance linewidth. Indeed, it should be
At this frequency, calculated and measured ␬ ⬙ values are of the order of 10⫺2 or lower.14 This difference likely results
equal to 1.84 and 1.79 关Fig. 2共b兲兴. Moreover, the extension of from the cylinder-magnetic-domains-shape assumption. In
magnetic losses up to f eff⫹f m experimentally observed on fact, domains have random shape. Therefore, a distribution
both ␮ ⬙ and ␬ ⬙ curves also appears in theory, as it can be over domains shape needs to be considered but, this will be
seen in Figs. 2共a兲 and 2共b兲, respectively. discussed in a future publication.
Currently, ferrimagnetic and ferromagnetic loaded com-
posite materials are widely used in planar microwave 1
J. J. Green and F. Sandy, IEEE Trans. Microwave Theory Tech. 22, 641
devices.12,13 They have numerous advantages compared with 共1974兲.
bulk ferrites: the nonrequirement of isostatic pressing, the
2
E. Schloemann, J. Appl. Phys. 41, 204 共1970兲.
3
M. Igarashi and Y. Naito, IEEE Trans. Magn. 13, 1664 共1977兲.
association of high saturation magnetization together with 4
G. T. Rado, Phys. Rev. 89, 529 共1953兲.
low dielectric loss when ferromagnetic powders are used, 5
D. Bariou, P. Quéffélec, Ph. Gelin, and M. Le Floc’h, IEEE Trans. Magn.
etc. To assist the device-design optimization, we applied the 37, 3885 共2001兲.
previous model and experimental technique to characterize
6
J. Krupka, IEEE Trans. Microwave Theory Tech. 39, 1148 共1991兲.
7
P. Quéffélec, M. Le Floc’h, and Ph. Gelin, IEEE Trans. Microwave
such a composite material. In Fig. 3, measured ␮ and ␬ data Theory Tech. 48, 1344 共2000兲.
for a 5⫻5⫻1.8 mm3 ferrite loaded composite sample 8
P. Quéffélec, S. Mallégol, and M. Le Floc’h, IEEE Trans. Microwave
(4 ␲ M s ⫽0.72 kG, q⫽0.7, ␧⫽13.8) are compared with the- Theory Tech. 50, 2128 共2002兲.
oretical ones. The applied dc magnetic field H 0 was 0.4 kOe.
9
E. C. Stoner and E. P. Wohlfarth, IEEE Trans. Magn. 27, 3475 共1948兲.
10
M. E. Hines, IEEE Trans. Microwave Theory Tech. 19, 442 共1971兲.
The following parameters were entered in the model: f m 11
R. Marqués, F. Mesa, and F. Medina, IEEE Microwave Wireless Compo-
⫽2 GHz, f eff⫽0.95 GHz, M /M s ⫽0.475, ␣ ⫽0.22, and q nents Lett. 11, 467 共2001兲.
⫽0.7. Simulations compare equally well with measurements.
12
J. B. Mills, C. J. Stevens, and D. J. Edwards, High Freq. Postgrad. Student
Colloq. 20, 59 共2001兲.
Indeed, as depicted in Fig. 3共b兲, the theoretical and experi- 13
E. Salahun, P. Quéffélec, G. Tanné, A. L. Adenot, and O. Acher, J. Appl.
mental ␬ ⬙ values, respectively, are 1.95 and 2.03 at gyro- Phys. 91, 5449 共2002兲.
magnetic resonance frequency. 14
N. Vukadinovic et al., Phys. Rev. B 62, 9021 共2000兲.

Downloaded 12 May 2003 to 193.54.246.42. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp
View publication stats