ims workshop wmfb - truebner · ims workshop "wmfb" electromagnetic wave interaction with...

IMS WORKSHOP "WMFB"

ELECTROMAGNETIC WAVE INTERACTION WITH WATER AND MOIST SUBSTANCES

Chairman: Andrezj W. Kraszewski USDA-ARS, Athens, GA, USA

Monday, June 17, 1996, 8:00 -17:00

Programme

08:00 - 08:10 A. Kraszewski 08: 10 - 08:40 Udo Kaatze

(Germany) 08:40 - 09: 10 Constantino Grosse

(Argentina)

09:10 - 09:50 Craig Dobson & Kyle C. McDonald

09:10 - 09:50

09:50 - 10:20

10:50 - 11 :20 Clarke Topp (Canada)

11 :20 - 12: 10

12:10 - 13:00

13:00 - 13:30 Ray King (USA)

13:30 - 15:00

15:00 - 15:40

15:40 - 16:10 Mike Kent (Scotland)

16:10 - 17:00

17:00

Welcome and Introduction The dielectric properties of water in its different states of interaction. Relaxation Mechanisms of cells suspended in aqueous electrolytes

Microwave Dielectric Properties of Soil and Vegetation and their Estimation from Spaceborne Radar Discussion

COFFE BREAK & Poster I

Time Domain Reflectometry techniques for soil water content measurements Discussion

LUNCH BREAK

On-line Industrial Applications of Microwave Moisture Sensors Discussion

COFFEE BREAK & Poster II

Simultaneous Determination of Composition and other Material Properties Using Dielectric Measurements Discussion

CLOSE

LIST OF CONTENT Page

number

Welcome to the Workshop A. Kraszewski, USDA, ARS, Russell Research Center, Athens, GA 1

1. "The dielectric proper tis of water in its different states of interactions" (Invited) Udo Kaatze, Georg-August Universitiit, Gottingen, Germany

2. "Relaxation mechanisms of cells suspended in aqueous electrolytes" (Invited) Constantino Grosse, Universidad National de Tucuman and C.N.I.C. T., Argentina

3. "Dielectric studies of hydration properties in hydrogels" - P. Pissis, A. Kyritsis, A.A. Konsta

3

9

and D. Daoukaki, Natl. Technical University of Athens, Greece . . . . . . . .. 11

4. "Hydration water in myoglobin solutions, studied by dielectric relaxations" - Lucia De Francesco, Franco Wanderlingh and Ulderico Wanderlingh, Universita di Messina, Italy . . . . . . . .. 15

5. "Dielectric relaxation versus water potential" - Max A. Hilhorst, IMAG-DLO, Wageningen and Jozua Laven, Technical University of Eindhoven, The Netherlands . . . . . . . . .. 19

6. "The effect of ionic conductivity and dipole orientation on the dielectric loss of the hevea rubber latex" - Kaida bin Khalid, Jumiah bt Hassan and Wan Daud bin W. Yusof, Universiti Pertanian Malaysia, Serdang, Selangor, Malaysia . . . . . . . . . .. 23

7. "Dielectric relaxation study of water in biological system - honey", Sanjay M. Puranik, T.S. Chanakya, Karave, Nerul, New Bombay, India

8. "Multifrequency microwave probing method. An ion and water concentration determination in salt solutions and biological tissues" - S.Y. Semenov and R.H. Svenson, Carolinas Medical

27

Center, Charlotte, NC . . . . . . . . .. 29

9. "A further study on material dependence of microwave attenuation for moisture determination over wide range of moisture content and density" - Zhi-Hong Ma and Seichi Okamura, Shizuoka University, Hamamatsu, Japan . . . . . . . . . . .. 33

10. "A study of the effects of sample temperature on microwave moisture content monitoring" Frank Thompson, Manchester Metropolitan University, Manchester, England . . . . . . .. 37

11. "Dielectric relaxation measurements of living organs by time domain reflectometry" T. Umehara, N. Miura, N. Shinyashiki, S. Yagihara and S. Mashimo, Tokai University, Kanagawa, Japan . . . . . . . . .. 38

12. "Measuring microwave permittivities of rice weevils" - Stuart O. Nelson, Philip G. Bartley, Jr. and Kurt C. Lawrence, USDA, ARS, Russell Research Center, Athens, GA . . . . . . .. 39

13. "On the localized microwave overheating in an organic water solution" - A.F. Korolev, A.1. Kostienko, A.P. Sukhorukov, Moscow State University, LV. Timoshkin, Russian Academy of Sciences, Moscow, Russia and A. Pulino, Gamma Tel., Cambridge, MA . . . . . . . . . 41

14. "Effects of microwave exposure on the extraction efficiency and temperature profiles of solvent/material mixtures" - M.M. Punt, G.S.V. Raghavan, M. Fakhouri, McGill University, Montreal, PQ, J.M.R. Belanger and J.R.J. Pare, Environment Canada, Ottawa, Ont., Canada .. 45

15. "Microwave absorption of wet snow" - Christian Matzler and Thomas Weise, University of Bern, Switzerland . . . . . . . . . . . . . . 46

16. "Microwave dielectric properties of soil and vegetation and their estimation from spaceborne radar" (Invited) - M. Craig Dobson, University of Michigan, Ann Arbor, MI, and Kyle C. McDonald, Jet Propulsion Laboratory, Pasadena, CA . . . . . . . . . . . . . . .. 47

- ;; -

17. "Time-domain reflectometry techniques for soil water content measurements" (Invited), G. Clarke Topp, Agriculture and Agri-Food Canada, Ottawa, Ont., Canada .............. 53

18. "Dielectric soil water content determination using time-domain reflectometry (TDR)" Rudolf Plagge, Christian H. Roth and Manfred Renger, Technical University Berlin, Germany . .. 59

19. "In-situ measurement of water content profiles in geological formations" - B. Oswald, H.R. Benedickter, Jeroen de Keijzer, W. Ba.chtold and H. Fliihler, Swiss Federal Institute of Technology, Zurich, and P. Marschall, NAGRA, Wettingen, Switzerland ............. 63

20. "New Time-Domain Reflectometry probes for water content determination in porous media" Markus Stacheder, Kurt Koehler and Robin Fundinger, IMKO GmbH, Ettlingen, Germany .... 67

21. "Distinguished problems in soil aquametry" - Alex Brandelik and Christof Hubner, Forschungszentrum Karlsruhe GmbH/ Universitiit Karlsruhe, Germany ............... 71

22. "On-line industrial applications of microwave moisture sensors" (Invited) - Ray J. King, KDC Technology Corp., Livermore, CA . . . . . . . . . . . . . . . .. 75

23. "Measurements of liquids, semisolids and powders: an overview of principles and techniques", James Baker-Jarvis, National Institute of Standards and Technology, Boulder, CO ......... 79

24. "Coplanar sensors for moisture measurements" - S.S. Stuchly and C.E. Bassey, University of Victoria, Victoria, B. C., Canada . . . . . . . . . . . . . .. 83

25. "Microwave sensors for measuring moisture content of solid and liquid materials" Quang Xiang Song, Huanzhong University od Science f3 Technology, Wuhan, P.R. of China. .. 86

26. "Portable moisture measurement system" - F. Menke and R, Knochel, Universitiit Kie1, Germany 87

27. "The dielectric permittivity meter for soil moisture determination", Wojciech Marczewski, Polish Academy of Sciences, Warszawa, Poland . . . . . . . . . . . .. 91

28. "Microwave industrial sensor for continuous measurement of moisture content in fat liver emulsions", G. Cottard and J. Guillon, Satimo, Les Ulis, B. Piriou and C. Sourioux, Bizac, Brive, France . . . . . . . . . . .. 95

29. "Accurate resonant frequency estimation of a dielectric ring resonator of a moisture measurement system", Seichi Okamura and Shigekazu Miyagaki, Shizuoka University, Hamamatsu, Japan ............ 97

Quebec, and 30. "Development of microwave methods to measure traces of water in gases"

Jacques Goyette and Tapan K. Bose, Universite du Quebec a Trois-Rivieres, Michel F. Frechette, Hydro-Quebec (IREQ), Varennes, Quebec, Canada ... 101

31. "Microwave moisture sensors and their application in civil engineering" - Klaus Kupfer, Hochschule fUr Architektur und Bauwesen, Weimar, Germany ........... .

32. "Microwave resonator moisture meter in the automated system of dumping grain before grinding" - P.D. Kuharchik, LA. Titovitsky, A.Ch. Belyachits and N.L Kourilo, Byelorussian State University, Minsk, Belarus

33. "A low cost microwave sensor for moisture measurement of textile webs" - T. Lasri, D. Glay, A. Mamouni and Y. Leroy, IEMN, Domaine Universitaire et Scientifique de Villeneuve d'Ascq, France .......... .

34. "Microwave instrument for high water content measurement in pulp and paper industry"

105

109

112

Pekka Jakkula and Esko Tahkola, VALMET A utomation Measurements, Oulu, Finland. . . .. 115

35. "A new application of WLAN-concept: complex permittivity monitoring of large-sizeed composite boards" - Ferenc Volgyi, Technical University of Budapest, and Balazs Zombori, University of Forestry and Wood Science, Sop ron, Hungary . . . . . . . . .. 119

- iii -

36. "Guided microwave spectrometry for on-line moisture measurement of flowable materials" Buford Randall Jean, Epsilon Industrial Inc., Austin, TX . . . . . . . . • .• 123

37. "Microwave linear sensors for on-line moisture detection and measurement" - G. Cottard, and P. Berthaud, Satimo, Les Ulis, and J.Ch. Bolomey, Supelec, Gil-sur- Yvette, France .. 127

38. "Use of measurements at two frequencies for phase-shift determination" - Samir Trabelsi, Andrzej W. Kraszewski and S.O. Nelson, USDA, ARS, Russell Research Center, Athens, GA .. 129

39. "Simultaneous determination of composition and other material properties ising dielectric measurements" (Invited) - M. Kent, Kf9S Associates, Edinburgh, Scotland . . . . . . . . . 131

40. "Multi-channel microwave resonator moisture-mass meter of paper web" - P.D. Kuharchik, I.A. Titovitsky, A.Ch. Belyachits and N.!. Kourilo, Byelarussian State University, Minsk, Belarus .. 135

41. "Density independence of microwave moisture determination using two-parameter measurements" R. Knochel and F. Menke, Universitiit Kiel, Germany . . . . . . 138

42. "Temperature-insensitive and density-independent grain moisture content determination from microwave measurements", A.W. Kraszewski, S. Trabelsi and S.O. Nelson, USDA, ARS, Russell Research Center, Athens, GA . . . . . . •. 143

Welcome to the Workshop !

This booklet contains summanes of the papers to be presented at the Second

Workshop on Electromagnetic Wave Interaction with Water and Moist Substances held

in conjunction with the 1996 IEEE Microwave Theory and Techniques Society

International Microwave Symposium in San Francisco, CA, on June 17. There are two

kinds of papers: first, papers invited to cover certain important areas of microwave

aquametry, deliberately chosen by the Organizer, and, second, those selected from an

avalanche of papers received in response to the invitation sent world-wide to people

involved in the subject.

The main topics for this Workshop are: physics of water in vanous stages of

binding, relaxation mechanisms of water in cells, in solutions and in soil, review of

microwave sensors for moisture determination in various materials and in different

industrial situations, and multi-parameter measurements of material that can be used

for simultaneous determination of various properties of the material, moisture content

being only one of them. Each topic will be discussed in an invited paper, tutorial in

character, and in a series of short contributions following it, which should be considered

as voices in the discussion rather than the usual paper presentation. The number of

excellent papers accepted for the discussion forced the Organizer to strictly limit the

time allocated for such presentations. Even introducing short poster sessions during the

breaks does not solve the time problem.

A special presentation is devoted to the fascinating subject of remote sensing of

water in soil and vegetation from airborne stations and satellites. Both the hardware

and the software and procedures developed for this purpose are highly specialized -

using radar and radiometry techniques in the broad frequency range from UHF through

microwaves to infrared radiation. Spectacular achievements of this branch of metrology

are extensively presented on sessions devoted to remote sensing at microwave

conferences, as well as at subject-oriented annual symposia and scientific conferences.

Such a presentation should be of value to individuals involved in research on physical

and technical aspects of microwave aquametry.

Essentially, the order in which papers are listed in this booklet will be the order

of oral presentations. Short poster sessions during breaks should allow participants to

- 2 -

get acquainted with interesting work which are not presented from the podium. They

are also important voices in the discussion. It is obvious that last minute changes in the

program may require reversions in the proposed order, but this is inevitable in holding

any large meeting. However, one of the principle aims in organizing this Workshop was

to provide a broad forum for exchange of information and ideas, no matter the

presented order. The presence of almost 40 speakers from 19 countries, and the serious

interest of many more who for various reasons were not able to attend, indicates that

there is an increasing interest in microwave aquametry in many countries of the world.

When planning the first Workshop in Atlanta, it seemed that five hours of

meeting time should be enough to accommodate the contributions devoted to various

aspects of microwave aquametry. We had 21 presentations, and 30 papers were

published in the resulting book. This year, nine hours were devoted for this purpose -

well over 40 summaries were submitted, and 36 of them were accepted for presentation.

It seems that the existing interest in the subject could justify the organization a two

day meeting in the future, possibly three years from now. Any suggestions on this

matter are welcome.

As you may already know, the book of the proceedings from the first Workshop,

held three years ago in Atlanta, entitled "Microwave Aquametry", has been recently

published by the IEEE Press. This is the first publication on microwave aquametry on

the market, and its success is important for the future of the papers presented at this

Workshop. It's my sincerest hope that publishing these manuscripts will be possible in a

much shorter time period than before, perhaps even during the next year.

I want to thank my supervisor and friend, Dr. Stuart Nelson, for his help and

encouragement during the whole 15-month preparation period, and Dr. Ronald Ham,

Chairman of the MTT-ll Microwave Measurement Technical Committee, for sponsor

ing the Workshop.

In spite of the shortage of time, I believe we will have an interesting Workshop. I

wish everyone of you a fruitful and enjoyable meeting and a pleasant stay in the

beautiful town of Saint Francis.

Andrzej Kraszewski, D.Sc.

WMFB Workshop Organizer.

- 3 -

The Dielectric Properties of Water in its Different States of Interaction

Uclo Kaatze Drittes Physikalisches Institut, Georg-August-Universitat

BiirgerstraBe 42-44, D-37073 Gottingen, Germany

Abstract

This paper summarizes results from broad-band spectrometric studies of the dielectric properties of aqueous solutions. The interest is focussed on indications for interactions of water with various solutes including low-weight inorganic and organic electrolytes, polyelectrolytes, small organic molecules and polymers, as well as amphiphiles that associate to form supramolecular structures like micelles and liposomes. Molecular mechanisms that may act an influence on the dielectric spectra are discussed, among them dielectric saturation, kinetic depolarization, positive, negative, and hydrophobic hydration, as well as the kinetics of the hydrogen bond networks in a variety of organic solute/water mixtures.

1 Complex Dielectric Spectra

Dielectric spectra that are characteristic of many aqueous systems are presented in Figure 1 where the complex (electric) permittivity

E( v) = E' (v) - i E" (v) (1)

is displayed as a function of frequency v for an aqueous solution of sodium chloride and also for water at the same temperature. Also shown for the solution is the dielectric loss defined by

(2)

where (j denotes the specific electric conductivity at low frequencies and EO

the electric field constant. Around a frequency Vw = (2rriw )-1 of about 20 GHz at 25°C the dielectric

spectrum of water exhibits a distinct dispersion (dE'(V)jdv < 0) and dielectric loss (E~(V) = E"(V) > 0) region. This behaviour is characteristic for a relaxation mechanism. It is due to the hindrance of the orientational motions of the dipolar water molecules by interactions with neighbors, particularly by

- 4 -

-

3' '", 40- -

20 I- -

60

3' :."

'" 40 ~

'" 20

2 5 10 20 GHz 50100

I'

Figure 1. Real part C.'(l/) and negative imaginary part tl/(ll) of tile complex permittivity semilogarithmicly plotted versus the frequency v for water (full points [1]) and for a 0.5 M aqueous NaCI solution (circles [21) at 25"C. Halfclosed symbols show the dielectric loss spectrum t~(v) for the salt solution (eq 2).

hydrogen bonding. The time Tw corresponding with the relaxation frequency Vw (dt"(v)/dvlllw = 0, d2t"(v)/dv21I1w < 0) is called dielectric relaxation time. For water at 25°C T10 = 8.27.10- 125 [IJ. The microwave dielectric relaxation properties of water are characterized by a discrete relaxation time. Hence the spectrum can be represented by a Debye-type relaxation spectral function given by

(3)

where tw( (0) denotes the permittivity extrapolated to high frequencies (Figure 1). Changes in the water properties on addition of solute are reflected by three changes in the characteristics of the spectrum: (i) The dispersion/ dielectric loss region of the solutions extends over a broader frequency range, indicating that there exists a distribution of relaxation times T. (ii)

- 5 -

-

-«.. 40 r

20 r

30

~ :", 20 «..

2 !Trw

10

o~~--~==~~--~~--~--~~

0.003 0.01 0.03 0.1 0.3 3 10 GHz 100 v

Figure 2. Semilogarithmic plot of the real part C:'(I/) and negative imaginary part (without conductivi ty contributions) c:~( 1/) of the complex dielectric spectrum for water (points [1 J) and for a 0.17 M aqueous solution of dipolar 1,2-dimyristoylglycero-L-3-phosphatidylcholine (circles [3]) at 30°C.

The principal relaxation time T s , as derived from the frequency I/s at which the dielectric loss of the solution adopts its relative maximum, differs from the pure water value at the same temperature, thus reflecting different molecular interactions. (iii) The extrapolated low-frequency (static) permittivity c(O) is smaller than the static permittivity cw(O) of water. In many liquid systems this finding is partly due to the fact that the water in the solutions changes its orientational polarizability with respect to the pure solvent.

Dipolar solutes may contribute their own relaxation characteristics. Hence, as illustrated by Figure 2, the dielectric spectra of such solutions normally look more complicated. It is then neccessary to discriminate between the static permittivity c(O) of the liquid on the one hand and the solvent contribution Csw to the static permittivity on the other hand (Figure 2).

- 6 -

2 Static Permittivity

The static dielectric decrement &(0) = €w(O) - €sw, where at vanishing solute contribution to the spectrum €sw = €(O), contains an effect of dilution of the dipolar water by the solute particles. The magnitude of the effect of dilution in €sw, however, is not just given by the volume fraction of solute but also by internal depolarizing electric fields within the dielectrically heterogeneous liquid. Unfortunately, even for perfectly spherically shaped solute particles the depolarizing electric fields cannot be calculated rigorously so that numerous empirical and approximate mixture relations exist [4]. Hence there is some ambiguity in selecting one. This is particularly true for solutions of nonspherical solutes. Nevertheless, experimental €sw data indicate that two different molecular mechanisms may contribute to the decrement O€(O) of aqueous systems. One mechanism, the so-called kinetic depolarization reflects an interesting coupling between the macroscopic dielectric properties and a microscopic hydrodynamic flow [5]. The other (positive or negative) contribution to oc(O) results from alterations in the orientation correlation of the solvent dipole moments [6]. Water around inert (hydrophobic) particles shows a tendency toward an enhanced dipole orientation correlation and, as a result, an enhanced static permittivity [7,8]. On the contrary, the direction of the electric dipole moment of water molecules around small cations appears to be largely fixed in the strong Coulombic field so that dielectric saturation [8] occurs.

3 Principal Relaxation Time

Depending on the solute, the dielectric relaxation time Ts of the water in the solutions may be smaller or larger than the relaxation time Tw of the pure solvent. Monovalent ions with radius between about 1.5 and 3 A induce relaxation times smaller than Tw [9]. This finding is called negative hydration in contrast to the positive hydration that is connected with the preferential radial orientation of water molecules around small cations [8,9]. Ions with radii larger than about 3 A and organic molecules show hydrophobic hydration characteristics giving rise to Ts > Tw [7,9]. Hydrophobic hydration is an effect of entropy rather than energy.

The hydration properties of aqueous solutions of organic molecules, including polymers, are largely governed by the interplay of hydrophobic hydration (around inert groups) and hydrogen bonding interactions at hydrophilic sites. The resulting water characteristics thus also depend on steric properties of the solute, its size and overall shape with respect to the water structure. A dominating parameter in determining the hydration water properties of organic particles, however, is the density of the hydrogen bonding sites offered by the solute. There do not seem to exist special hydration properties around polymers.

- 7 -

4 Relaxation Time Distribution

Interesting details of the structure and microdynamics of solutions are reflected by the type of the relaxation time distribution function that underlies the dielectric dispersion and absorption. For many organic molecule/water mixtures the assumption of an almost homogeneous hydrogen bond network with a continuous distribution of dielectric relaxation times turns out to be appropriate. Other solutions can be consistently represented by a hydration model in which the water that surrounds solutes particles is assumed to be affected while pure water characteristics are attributed to solvent apart from solute molecules or ions. There are also systems, like aqueous solutions of carboxylic acids [10], which appear to be composed of two subphases with different dielectric properties, one phase with low the other with high water content.

5 Summary

A more recent review on microwave dielectric properties of liquids, with special emphasis on aqueous systems, is given in ref. [11 J.

References

[lJ Kaatze, U. J. Chern. Eng. Data 34 (1989) 371.

[2J Kaatze, U. J. Phys. Chern. 91 (1987) 3111.

[3] Kaatze, U.; Gopel, IC-D.; Pottel, R. J. Phys. Chern. 89 (1985) 2565.

[4J van Beek, L.K.H. in Progress in Dielectrics, Vol. 7 (J.B. Birks, Ed.) Heywood, London, 1967.

[5] Hubbard, J.B.; Onsager, L.; van Beek, \tV.M.; Mandel, M. Proc. Natl. Acad. Sci. USA 74 (1977) 401.

[6J Kirkwood, J.G. J. Chern. Phys. 7 (1939) 911.

[7] Kaatze, U.; Pottel, R. J. lv/ol. Liquids 52 (1992) 181.

[8] Kaatze, U. Z. Phys. Chern. (Nliinchen) 135 51.

[9] Pottel, R.; Giese, K.; Kaatze, U. in St'ructure of Water and Aqueous Solutions (\tV.A.P. Luck, Ed.) Verlag Chemie-Physik Verlag, Weinheim, 1974.

[10] Kaatze, U.; Menzel, K.; Pottel, R. J. Phys. Chern. 95 (1991) 324.

[l1J Kaatze, U. Radiat. Phys. Chern. 45 (1995) 549.

- 9 -

RELAXATION MECHANISMS OF CELLS SUSPENDED

IN AQUEOUS ELECTROLYTES (Invited)

Constantino Grosse

Instituto de Fisica, Universidad Nacional de Tucuman, Argentina and

Consejo Nacional de Investigaciones Cientificas y Tecnicas.

Abstract.

The main relaxations observed in cell suspensions in the whole frequency rane;e accessible to dielectric measurements - from a few Hz up to the THz region, are examined. These encompass the alpha, beta, delta and gamma relaxations. The physical mechanisms leading to the counterion polarization, charging of the cell membranes, Maxwell-Wagner polarization and dipolar relaxations, are discussed. The formalisms which permit the description of these relaxations and the determination of various cell properties from dielectric measurements are presented.

Summary

The aim of this work is to present an unified treatment of the main dielectric

relaxations observed in biological materials, complementing, in the tutorial aspect,

existing Reviews on this subject [1-4].

The method used consists of choosing a specific system: a cell suspension with

specified values of all its characteristic parameters, and determining its dielectric

behavior. This is done in four stages:

1) A presentation of the main results and definitions necessary for a description of the

dielectric properties of suspensions.

2) The usual calculation based on the Maxwell mixture formula which leads to a

complicated expression in which the influence of system parameters on the different

relaxations cannot be analytically obtained.

3) A qualitative discussion of the different relaxation mechanisms in time domain,

showing how the system evolves when a DC field is applied. The successive changes of

the dipolar coefficients are illustrated with a series of figures which show the field

around the cell (these figures are precise representations of the field lines in the chosen

system drawn in such a way that the proportionality of the line density is everywhere

rigorously proportional to the field strength).

4) A calculation in the frequency domain of the characteristic parameters of the

different relaxations treated separately.

- 10 -

In all this presentation, the importance of the dipolar coefficient of a suspended

particle is stressed. This make it possible to show origin of the second component of the

gamma relaxation, and to calculate in a completely independent fashion the delta and

the beta relaxations.

The low frequency alpha relaxation, which is only mentioned in the Reviews, is

then discussed in length. The physical mechanisms leading to the counterion

polarization are first discussed in the time domain, and illustrated with figure showing

the onset of the formation of the neutral salt clouds and the final flows of counterions

and coions. The model of Shilov [5] is then presented, and the analytical results for the

dielectric properties of the suspension are deduced.

Finally, the permittivity and conductivity spectra valid in the whole frequency

range accessible to dielectric measurements are calculated, and the interpretation of

measured data is discussed.

[1] H.P. Schwan, "Electrical properties of tissue and cell suspensions", Advances in

Biological and Medical Physics, Academic Press, New York, vol. 5, 1957.

[2] R. Pethig, "Dielectric and electronic properties of biological materials", John Wiley

& Sons, Chichester, 1979.

[3] K.R. Foster and H.P. Schwan, "Dielectric properties of tissues and biologucal

materials: a critical review", CRC Critical Reviews in Medical Engineering, vol. 17,

CRC Press, 1989.

[4] S. Takashima, "Electrical properties of biopolymers and membranes", Adam Hilger,

Bristol, 1989.

[5] S.S. Dukhin and V.N. Shilov, "Dielectric phenomena and the double layer in disperse

systems and polyelectrolytes", Kerter Publishing House, Jerusalem, 1974,

- 11 -

DIELECTRIC STUDIES OF HYDRATION PROPERTIES IN HYDROGELS

P. Pissis, A. Kyritsis, A.A. Konsta and D. Daoukaki

National Technical University of Athens, Physics Department, Zografou Campus, 157 80 Athens, Greece

Hydrogels are cross-linked hydrophilic polymers which imbibe substantial amounts of water, but which swell rather than dissolve in water [1]. The investigation of their properties, which are strongly influenced by water, attracted much interest in recent years, tightly related to their numerous applications in pharmaceutics and in medicine [1]. From a more fundamental point of view, hydrogels are of special interest for hydration studies.

The physical structure and the properties of water in a hydrogel are primarily determined by the existence of binding sites in the polymer [2] and by geometrical confinement. It is now well established that, in the structural sense, polymer hydro gels are an analogue of a pore sponge with pore sizes corresponding to the distance between neighbouring cross links of the polymer network. , A variety of experimental techniques has been employed in the study of the structure and the properties of water in hydro gels. The experimental data are often interpreted in terms of classification of water into different fractions (free and bound, freezable and nonfreezable, molecularly distributed and clustered, depending on the specific technique employed) and, alternatively, in terms of homogeneous and heterogeneous mixtures [3].

We report here on the hydration properties of three different hydrogel systems in a wide range of water content, studied by means of dielectric techniques. The systems are (a) poly(hydroxyethyl acrylate) (PHEA), (b) interpenetrating networks of poly( ethyl acrylate) and poly(hydroxyethyl methacrylate) (PHEA/PHEMA IPNs), and (c) polyacrylamide (PAA). The dielectric techniques include broad band dielectric relaxation spectroscopy (DRS) in a wide range of frequency (10- 10 Hz - 20 GHz) and temperature (173 -363 K) by means of several setups, and thermally stimulated depolarization currents (TSDC) techniques in the temperature range 77 - 300 K. They allow to look at the mobility of water itself and at the influence of water on the relaxation and conductivity mechanism of the polymer matrix. Additional relevant information is obtained from equilibrium and dynamic water sorption isotherm measurements.

In Figs 1 and 2 we show, in log-log plots, the real s' and the imaginary s" part of the complex permittivity s* of PHEA hydro gels at different temperatures, T, and water contents, h, expressed in grams of water per gram of dry material, respectively. An increase of T at constant h in Fig. 1 has the same effect as the increase of h at constant T in Fig. 2, namely an overall increase of molecular mobility (T-h-fsuperposition principle [4]). At low frequencies and high Tlhigh h we observe the effects of space charge polarization and the so-called conductivity relaxation. Detailed analysis of the data, transformed to the formalisms of complex modulus, conductivity and impedance, allows to investigate the effects of water content on DC conductivity and on the glass transition dynamics. The glass transition temperature Ta is obtained from DSC and TSDC measurements. At high water contents, freezing and ~elting events obscure the Tg signals recorded with TSDC, but not the TSDC ones [5].

At high frequencies we observe in Figs 1 and 2 a broad loss peak due to the secondary y and ~sw mechanisms. A detailed study of the behaviour of this band has shown that peaks y and ~sw coexist for h < 0.10 gig, whereas for h> 0.10 the ~sw mechanism grows at the expense of y and is plasticized. Combined DRS and TSDC studies allow to determine the parameters describing the mechanisms and their hydration dependence. DRS measurements in the GHz frequency region and TSDC measurements at low temperatures allow to look at the relaxation of water itself in the liquid and the crystalline phases respectively.

- 12 -

10' a 10' 0.32 a 293K

105 105 0.17

273K

10' £' 0.16

£' 10'

103 243K 0.120 103

102 223K

213K 102

10' 193K 0.095 173K 10'

10'

10' 293K b

10' 0.32 b

273K 0.17 105

105 0.16 £" 10'

10' 243K £" 0.120

103

103

102 223K

213K 102

0.095 10'

1113K 10' 100

173K 100

10-'

10-2 10-'

100 10' 102 103 10' 105 10' 107 10' ,01 ,0'0 10-1 ,00 101 102 103 10' 105 10' 10' 10' 101

f (Hz) f (Hz)

Figure 1. Real part, e' (a) and imaginary part, Figure 2. Real part, e' (a) and imaginary part, e" (b) of the dielectric function, e* , plotted vs. e" (b) of the dielectric function, e* , plotted vs. frequency, f, for a PHEA sample at h = 0.30 gig frequency, f, for a PHEA sample at T = 297 K and at several temperatures between 173 and 293 and at several water contents given on the plots K. as g of water per g of dry material.

The following picture emerges from the detailed DRS, TSDC and water sorption measurements on PHEA. The hydrogel is a homogeneous mixture for h ~ 0.30, whereas a separate water phase exists for h ~ 0.30. Water in excess to h 1:::1 0.30 does not contribute to plasticization of the glass transition and of the secondary ~sw mechanism, whereas it gives rise to an additional relaxation mechanism, characteristic of loosely bound water. No dielectrically free (bulk) water was observed in the range of water contents studied (h < 0.46). Moreover, our dielectric results suggest strong interactions between water molecules and PHEA side groups at low water contents, in apparent disagreement with water sorption data, these latter indicating that clustering occurs at much lower water contents than phase separation as detected by DRS.

TSDC and AC spectra obtained with polyacrylamide (PAA) / water hydrogels as a function of water content and temperature are shown in Figs 3 and 4, respectively. Figure 3 displays TSDC spectra recorded with a PAA hydrogel with four different water contents. They reveal in· general two peaks, the first, at low temperatures (L T peak), being attributed to the reorientation of loosely bound water molecules, whereas the second one, at high temperatures (HT peak), is a manifestation of a glass transition. For very high water contents (not shown) a supplementary peak is observed at about 120 K, due to the relaxation of free (bulk) water molecules in the sample. It is observed that the LT peak disappears for h values lower than about 0.15 g/ g, suggesting that for lower hydration all water is tightly bound into the hydrogel network. The HT peak, attributed to the glass transition, shifts to lower temperatures with increasing b, due to the plasticizing effect of water.

- 13 -

0.35 3.5 a h = 0.05 gig

0.30 -,- b h = 0.13 g/.g 3.0 c h = 0.29 gig

d d h = 0.38 gig

____ 0.25 2.5 ____

~ c ~

~ 0.20 2.0 ~ ...... ......

0.15 1.5 b

0.10

d 1.0

0.05 0.5 a ".--....... ...

0.00 0.0 100 150 200 250

T (K)

Figure 3. TSDC spectra of a PAA hydrogel with four different water contents h, in g of water per g of dry material.

The evolution of AC conductivity spectra with temperature of a P AA hydrogel, recorded for decreasing T, is shown in Figure 4. It is observed that the low frequency conductivity value changes abruptly for -5 °c < T < -10 °c, evidencing crystallization of, at least, part of the bound water at this T-range, according to similar results obtained with samples of natural origin [6]. A second discontinuity region, corresponding to practically vanishing of DC conductivity, is detected between -20 and -30 °C. If the heating process is reversed (Fig. 4, dashed curve, increasing 1), the first discontinuity is shifted well above -5 °C, showing the irreversibility of the freezing-de freezing process. Finally, the drop of the apparent conductivity below its DC value for frequencies lower than 1 kHz is a clear indication of the electrode polarization effect [7].

10 -2

10 -3

10 -8

• ••.••.•.•.••.....••••••••• : ! : • • • • : : : : : + + + + + + + + + + + + + + + + + + + + + + + + + + + • • •

+ + + + •••••••••••••••••••••••••••••• • • • • • . -5 ·C, deer. T

. . . . . .' •••• . . . . i •

• •

... ...

,/

+++++T= O·C • •••• T = -5·C 00000 T = -10 ·c 00000 T = -15 ·C ••••• T = -20 ·C • •••• T = -30 ·C ••••• T = -40 ·C ••••• T = -60 ·C - - T = -5 ·C,lner. T . 10 -7 ••

+-~~~Tnrr--'~~~~-'~-rnTnr--~~··r·"·~·.-~-r~ 10 2 10 3 10 • 10 e

f (Hz) 10 8

Figure 4. Dependence of AC conductivity spectra on temperature for a PAA hydrogel, recorded with decreasing T, except for the dashed-line plot, which was recorded with increasing T. The crystallization of water for temperatures lower than _5°C, as well as the non-reversibility of the freezing-thawing process is clearly demonstrated.

- 14 -

References

[1] N.A. Peppas, Ed., Hydrogels in Medicine and Phannacy. New York: Wiley and Sons, 1987. [2] P.H. Corkhill, A.M. Jolly, C.O. Ng and B.J. Tighe, "Synthetic hydrogels: 1. Hydroxyalkyl acrylate

and methacrylate copolymers-water binding studies", Polymer, vol. 28, pp. 1758-1766, September 1987.

[3] G. Smyth, F.X. Quinn and V.J. Mc Brierty, "Water in hydrogels. 2. A study of water in poly (hydroxyethyl methacrylate)", Macromolecules, vol. 21, pp. 3198-3204,1988.

[4] A. Kyritsis, P. Piss is and J. Grammatikakis, "Dielectric relaxation spectroscopy in poly(hydroxy ethyl acrylates) / water hydrogels", J. Polym. Sci. Polym. Phys. Ed., vol. 33, pp. 1737-1750, 1995.

[5] A. Kyritsis, P. Pissis, J.L. Gomez Ribelles and M. Monleon Pradas, "Depolarization thermocurrent studies in poly(hydroxyethyl acrylate) / water hydrogels", J. Polym. Sci. Polym. Phys. Ed., pp. 1001-1008, 1994.

[6] A.A. Konsta, P. Pissis, A. Kanapitsas and S. Ratkovic, "Dielectric and Conductivity studies of the hydration mechanisms in plant seeds", Biophys. J., vol. 70 (in press) 1995.

[7] H.P. Schwan, "Linear and nonlinear electrode polarization and biological materials", Ann. Biomed. Eng., vol. 20, pp. 269-288, 1992.

- 15 -

HYDRATION WATER IN MYOGLOBIN SOLUTIONS, STUDIED BY DIELECTRIC RELAXATION

Lucia De Francesco, Franco Wanderlingh, Ulderico Wanderlingh

Dipt. di Fisica, Universita di Messina, C.P.55 S. Agata, 98166 Messina, Italy

Abstract In this note we present a measurement of the number of the water molecules in the hydration shell of small

a protein. We investigated myoglobin acqueous solutions at intermediate and high concentrations. We

have exploited the dielectric tecnique illustrated by Wei et al., already applied at low and intermediate

concentrations. The method is based on the measure of the dielectric excluded volume fraction and

accurate density and concentration measurements.

Introduction The term hydration water indicates the water molecules in a aqueous solution that are adjacent to

macromolecules and have different properties from the remaning water, called bulk water. In biological

solution these molecules gain a particular importance because they are a crucial factor for stabilizing

different conformational states of biomolecules, providing them with the flexibility required for the func

tionality [1,2]. It is now widely understood that hydration water show properties different from those of

bulk water. In general the translational and rotational motions are slowed down compared to those of

free water and also the spatial structure is modified. But an unique quantitative description of such a

modifications is not present in literature, probably due to the many different environments in which the

hydration water can be found.

A suitable technique for the investigation of this subject is Dielectric Spectroscopy. It is in fact possible to

distinguish between bulk and hydration water and, moreover, quantify their respective amount. Recently

Wei et al. [3] have developed a novel dielectric method for measuring the number of water molecules in

the hydration shell of macromolecules. In their paper they calculate the volume fraction of the dielectrical

voids present in the solutions by considering only the relaxation of the free water molecules. They have

applied the tecniques to several proteins at low concentration (few mM), obtaining consistent results.

In this note we intend to extend that method to higher concentrations for two main reasons. From a

practical point of view an higher population of solute molecules with their hydration shells make easier

to determine the properties of these latter. Furthermore the actual physiological condition in which bio

logical macromolecules operate are crowed environment [4]. In such a situation the characteristic figure

obtained in idealized low concentration case are likely to be not very correct.

In order to extend the method of Wei et al. at higher concentration we had expressly taken into ac

count a contribution that describe the relaxation of the hydration water. The results we obtained are

encouraging, but further investigation are necessary.

Experimental 'vVe studied solutions of myoglobin at four different concentrations, from 4 to 11 [mM], that correspond

to about 5-15% by volume. We used lyophilized sample of lIorse Skeletal Muscle Myoglobin (MW 17640

Da) purchased from Sigma Chemical Company. Known quantities of it were dissolved in bidistilled water

and then the solutions wert, centrifuged. The exact concentrations were found by spectrophotometric ab

sorption measurements, based on note extinct.ion coelIicients. The density of each sample was measured

by a pychnometer, using all high precision balance with an error of 1O-5g.

The dielectric spectra[5-6] were taken by all IlP8720C Automatic Network Analyzer, which measure the

reflection coellicicnt from a flanged flat ended probe (II P85070A) directly immersed in the solutions under

- 16 -

investigation. This procedure require a calibration by means of three known standards: a short block,

the air and, bidistilled water, as "matched load" . We choose water because its permettivity is near to

solution's one and also because its reflection coefficient is middle between the others two standards. A

thermal bath was used to stabilize the temperature within O.l°C. A devoted software routine is able to

extract the real and imaginary part of the dielectric function from the measured reflection coefficients in

the frequencies range from 100 MHz to 20 GHz.

80

70

60 ..........

8 '-'

50

w 40

30

20

10

. . .

0 2 1010 4 1010 61010 8 10lD 1 lOll 1.2 lOll

0) (Hz)

Figure 1. From top to bottom: f' versus frequency for pure water and protein solution at increasing concentration. The continuous line are the fitted two Debye model.

Data analysis and discussion The real parts of the obtained relaxation curves are shown in Figure 1, along with that of pure water.

As can be observed the contribution due to the bulk water is present even at the higher concentration, and

an other dispersion appears at smaller frequencies, due to the hydration water. Moreover a modulation

of the experimental points is present, mainly at the higher concentration, whose origin is not clear. These

data have been then tentatively described in terms of two Debye type relaxation terms:

"( ) Ah AJ f W = + + foo

1 + JWTh 1 + JWTJ (1)

The latter dispersion is related to the free water, and the parameter TJ and foo are clamped to the value

obtained in the bulk sample: TJ = 8psec.; foo = 6. The former dispersion is related to that fraction of

the hydration shell water, whose relaxation falls ill the time window of our experiments.

The results of the fitting procedure are shown in Table 1. It can be observed that the values of AJ and

- 17 -

D..h show a linear dependence from the concentration. From these data the value of the static dielectric

constant of bulk water Cw = D..j + coo, can be derived extrapolating the data at zero concentration. Then,

by using the same procedure described by Wey et a1.[3], we are able to obtain the dielectric excluded

volume fraction: R - (cw -Cj)*(2cw+ coo)

v - ( 2cw + C j ) * (cw - coo) (2)

Since in the above expression we use for C j the values of the dispersion coming only from the water

contribution, the second term in eq.(l), it follows that the value for Rv refers to the total volume of the

hydrated protein molecules per liter of the solution. This is related to the volume of the dielectric void

originated by the protein plus its hydration water, by the simple relation: Vv = Rv/cNo. The volume of

the hydration shell and the number of the water molecule, Nhyd, by which it is formed can be derived

via accurate measurement of the density, d, for the various solutions and for the solvent, togheter with

the molecular weight of the protein and its concentration, [c], in the solution. The calculated quantities

are summarized in Table 1.

Table 1

[c]mM D..j D..h 1'h psec d g/cm;! Nhyd

4.14 63.0 4.6 42 1.0130 279 6.65 57.8 6.3 58 1.0200 341 9.79 52.5 7.7 64 1.0380 370 10.97 50.2 7.9 57 1.0410 379 Table 1: results for the parameters in eq.1, the measured density and

calculated number of the molecules in the hydration shell at the various concentrations.

The values we have obtained for the number of molecules in the hydration shell are consistent with

literature values [7) thus indicating that the extention of the method of Wey et al. at higher concentration

can be successfully used. r ... loreover from our results the thickness of the hydration shell seems to sligtly

increase at increasing concentration suggesting a cooperative effect of the solute protein in modifying the

structure of the solvent.

As for the bound water contribution, from our analysis we can state that part of the water in the

hydration shell show an average relaxation time of 55· 1O- 12sec. If we consider the literature data for

the relaxation time of water molecules bound to biological macromolecules, we find values ranging from

10-11 - 1O-7sec. Our findings are just at the lower limit of that range, and agree with the results

found by Grant et a1.[8] in similar solutions. We can than suppose that this relaxation time is to be

attribuited to a less bound fraction of water molecules in the hydration shell. These considerations are

at the moment speculative, mainly because of the not yet explained deviation from the fitted equation

of the experimental data at the higher concentrations, and because of the possibility to fit a two Debye

dispersion by a single dispersion with a distribution of relaxation time. Nevertheless such a results are

encouraging and worth of further investigation.

- 18 -

REFERENCE [1] J.A.Rupley, "G.Careri Protein hydration and function" in Adv. Prot. Chem. Vol.41 pp37-171, 1991

[2] LD.Kuntz, W.Kauzmann "Hydration of proteins and polypeptides" in Adv. Prot. Chem. Vol.28 pp236-

345, 1974

[3] Y. Wei, A. C. Kumnharkhane, M. Sadeghi, J. T. Sage, W. D. Tian, P. M. Champion, S. Sridhar and M.

J. McDonald, "Protein Hydration Investigations with High-Frequency Dielectric Spectroscopy" J. Phys.

Chem. 1994, Vol. 98, pp. 6644-6651

[4] S.B.Zimmerman, A.P.Minton "Macromolecular Crowing" in Ann. Rev. Biophys. Biomol. Struct. vol.22,

pp.27-65, 1993

[5] Y. Wei, S. Sridhar Rev. Sci. Instrum., 1989, Vo1.60, pp.3041-3046

[6] Y. Wei, S. Sridhar IEEE Trans. Microwave Theory Tech., 1991, Vol.39, pp.526-531

[7] R. Pethig, "Protein-Water interaction determined by dielectric method" Annu. Rev. Phys. Chem., 1992,

Vol. 43, pp. 177-205

[8] E.H.Grant, V.E.R. McLean, N.R.V.Nightingale, R.J.Sheppard, M.J.Chapman Bioelectromagnetics vol.7

pp. 151-1621986

- 19 -

DIELECTRIC RELAXATION VERSUS WATER POTENTIAL Max A. Hi/horst' and Jozua Laven2

IIMAG-DLO, P.O. Box 43, NL-6700 AA Wageningen, The Netherlands 2Technical University of Eindhoven, P.O. Box 513, NL-5600 MB, Eindhoven The Netherlands

Abstract The ratio between a reference pressure and water potential is shown to be inversely proportional to

the ratio between the relaxation frequency of bulk water and that of bound water. This relationship is found using thennodynamic properties of water bound to the material matrix, and is illustrated using dielectric and water retention properties of materials found in the literature. The first mono-layer of water molecules adsorped on a particle surface of an soil dried in air with an average relative humidity

== 50 %, which is equivalent to a water potential == -1 kbar, has an activation enthalpy == 35 kllmol with a corresponding relaxation frequency == 50 MHz. The increasing number of water layers with activation enthalpies between 35 kllmol and 20 kllmol for bulk water, correspond with water potentials between -1 kbar and -1 bar and relaxation frequencies between 50 MHz and 17 GHz. Therefore, the dielectric spectrum of a wet porous material is a function of both the sum of the various stages of bound water or total water content and water potential. It is almost flat below 50 MHz and fonns a continuously declining curve between 50 MHz and 17 GHz depending on the water retention characteristic. As a consequence, the hysteresis observed for the water retention characteristic of a porous material applies to the dielectric spectrum too.

Introduction For wet porous materials, such as soil or ceramics, a relationship can be found between its dielectric

properties and water content, S. This relationship is also a function of frequency, f. At frequencies, f> 50 MHz, this frequency dependence is dominated by the various stages of water binding.

There are two main types of dielectric measuring techniques: TDR (Time Domain Reflectometry) and discrete Frequency Domain measurements. In practice, the measuring frequencies for both types can range from 1 MHz to over 15 GHz, depending on the application or the technique used. To be able to interpret the spectra, or calibration curves for discrete frequencies it is important to understand the impact of the binding of water on the dielectric spectrum.

In this paper, low frequency if < 50 MHz) effects on the dielectric properties, resulting from ionic conduction, will be ignored.

Dielectric Polarization and Relaxation The dielectric properties of a material can be described by a complex representation of the dielectric

constant, the complex electrical permittivity, E*,

E*=e' - jeff (1)

where the real part of the pennittivity, E', is a measure of the total polarizability of the material constituents. The imaginary part of the permittivity, E", represents the dielectric energy absorption.

The reorientation of water molecules in an alternating electromagnetic field, although rapid, takes some time. At very low frequencies E' = Es , where Es is the static dielectric constant representing the permittivity for a static electromagnetic field. With increasing frequency, J, the molecules become too slow to follow the fast alternating field. For f ~ 00, the real part of the permittivity will decrease to £.,., ,

a value for the polarization at atom and electron level. The frequency for which the permittivity is decreased to (Es - £"")/2 is called the relaxation frequency J,.. According to experimental results of Kaatze and UhlendOIf [1] this frequency dependence is best described by the Debye [2] function for a single relaxation process:

* e, -e~ e = +e~ l+jflfr

(2)

- 20 -

Dielectric Relaxation and Water Potential Two of the reasons for the frequency dependence of E' of wet porous materials, are the

intermolecular bonding between water molecules and the bonding of water molecules to the pore surfaces of the solid material. The degree of binding to a particle or pore surface varies from unbound for bulk water at great distance from the surface to heavily bound for adsorbed water. The binding forces acting on a water molecule are also affected by other causes such as the presence of ions.

A water molecule is connected to its neighboring water molecules or to the particle surfaces, by one or more hydrogen bonds. This prevents the molecule from reorienting in a fast changing electromagnetic field. The relaxation frequency can be related to the probability of making or breaking such bonds during one period, t = 1I21t fr , of the relaxation frequency. Using the theory of rate processes [3] the relaxation frequency can be related to the Gibbs free energy:

kT _ flG·

Ir = --e NT 2rch

(3)

where tl.G* is the molar Gibbs free energy for breaking a bond, h is Planck's constant, k is Boltzmann's constant, T is the absolute temperature and R the molar gas constant. When a molecular bond is broken, this is almost immediately followed by the creation of a new bond. The Gibbs function is dermed as G = H - TS where H is the enthalpy and S the entropy. Thus t!.G* = tl.H* - TM*, where tl.H* is the molar activation enthalpy and M* the molar activation entropy. According to Kaatze [1] and Grant [4], the weak temperature dependence of tl.H* and M* is negligible between -5°C and 60 °C. For liquid water M* is approximatly constant. Thus tl.H*, corresponds to the energy to make or break a hydrogen bond.

In the following, the suffix 0 in fro, t!.G*o etc. denotes values for bulk water at a reference atmospheric pressure Po = 1 bar and 20°C, while values without this suffix refer to molecules experiencing a certain level of binding force at 20 0c. With formula (3), the ratio between fro and J, becomes

Iro -=e Ir

flG·-dGO·

RT =e /lH*-/lHo*

NT (4)

It is shown in textbooks on thermodynamics that the chemical potential (also termed partial molal free enthalpy), J..L, of a pure substance is equivalent to t!.G* [6]. The thermodynamic equilibrium condition between liquid and vapour phases at given temperature and total pressure is J..Lvap = J..Lliq' Thus the difference between the chemical potentials in the vapour phase and in the liquid phase are equal; i.e., J..Lvap - J..Lvap 0 = J..Lliq - J..Lliq 0 =::: tl.H*- tl.Ho*. According to Raoult's law [6] for the present situation, the partial pressure of a component is directly proportional to its mole fraction in the liquid, and the proportionality constant is the vapour pressure of the pure substance. Therefore the ratio between the partial vapour pressure, P, and the reference pressure, Po, is related to tl.H*- tl.Ho* of the liquid by

P /lH*-/lHo *

-=e RT

Po (5)

Under equilibrium conditions for the water vapor and liquid water interface, P is identical to the inverse of water potential \f' ; the amount of energy needed to transport water to or from the material matrix. With P = -\f' and equations (4) and (5) the relationship between relaxation frequency and matrix potential becomes

(6)

where the minus sign takes account of the inverse nature of the processes of \f' and P. This implies a direct relationship between the real part of the permittivity as function of the relaxation frequency, and as function of water content and water potential: E'(fr) = £'(8, \f').

With (6) the permittivity of a wet porous material as a function of frequency can be calculated from the contributions of the volume fractions of all stages of water binding using (2) for each component

- 21 -

and mixing fonnulas. However, up to now many mixing formulas have been published [8] and none of them is universally applicable, in particular not at lower water contents. In addition the permittivity of a porous material depends also on parameters such as pore size and shape distributions [5,9].

Kaatze and Uhlendorf [1] found for bulk water at 20°C and atmospheric pressure MIo* = 20.5 kllmol and irlJ = 17 GHz. The activation entropy calculated from this data is Mo* = 36.8 J/mol K. They also found Es = 80.2 and E.. = 5.6. Hasted [5] found for ordinary ice MI* = 55 kllmol with ir = 7 kHz, Es = 92 and E.. = 3.1. For ice all three bonds are active, while for bulk water statistically one bond is active. The static permittivity for liquid bound water is nearly independent of the binding forces at constant temperature, Es is even higher for ice than for bulk water. The high-frequency permittivity of ice E.., is a little lower than that of bulk water. It is reasonable to consider both Es and E.. constant as a function of binding state for analyzing bound water effects at constant temperature. Thus E'(f)e.T=constant is a function of the water potential only.

Conclusions and Discussion The ratio between a reference pressure and water potential is inversely proportional to the ratio

between the relaxation frequency of bulk water and that of bound water. Therefore, the dielectric spectrum of a wet porous material is a function of both the sum of the various stages of bound water or total water content and water potential.

This conclusion for porous materials in general, can be illustrated using the dielectric and water retention properties of materials as found in the literature. For bulk water MIo = 20.5 kllmol with corresponding f,. = 17 GHz. The fraction of water in the first layer bound to the surface of the particles for coarse sand is low because of the small effective surface area. In this case the effect of '¥ on the ir is small and the permittivity as a function of frequency is dominated by the volume fraction of (bulk) water. The mean relaxation frequency of wet sand is only a little lower than that of bulk water. Wet coarse sand will show an almost flat dielectric spectrum for frequencies of less than a few GHz. For increased silt or clay content the specific surface area is increased and consequently the bound water fraction as well. Dirksen and Dasberg [10] showed that, for a broad range of soil types, there is little difference between the water content calculated for the first mono-layer of water molecules and the measured hygroscopic water content. This was done under average ambient conditions with a relative humidity of = 50 % corresponding to a water potential '¥ = -1 kbar and an relaxation frequency calculated from (6) J,. = 50 MHz. This water content varied between a few percents for sandy soils up to 11 % for Vertisol and Bentonite. Hoekstra and Delaney [11] showed, that for a low conductive Goodrich clay there was almost no increase in permittivity for frequencies below 100 MHz and a continuously declining permittivity for 100 MHz and up. Hasted [5] found for ordinary ice MI = 55 kllmol with ir = 7 kHz, and for water bound to protein MI = 29 kllmol withJ,. = 500 MHz. Note that the latter is an average value over 2 or 3 layers water. Rolland and Bernard [13] found MI = 52.5 kllmol for water adsorbed to silica gel (= one layer). Brown et al. [14] found for water adsorbed to settled concrete MIbetween 54 kllmol and 155 kllmol. However, most of the literature [15] shows values around 50 kl/mol. According to Kent and Meyer [12] the mean activation enthalpy of bound water in micro-crystalline cellulose (= one layer of water molecules) can be estimated between 31 kllmol and 42 kl/mol. Calculated from (6), the corresponding relaxation frequencies are between 200 MHz and 2 MHz. Nearly all water molecules in the first mono-layer are removed from most soils using the oven-dry method at 105°C. The oven-dry method can remove water to ,¥:::: _104 bar, corresponding to a calculated MI = 40 kllmol and ir:::: 6 MHz at 20°C. For some clays and concrete, however, a temperature of more than 400 °C is needed to remove all bound water. For practical use this water fraction is small and it is often regarded as part of the soil matrix.

Summarizing the hydrogen binding forces. expressed by Ml. for the first mono-layer of bound water in a porous material. has a Ml between 35 kJ/mol and 55 kJ/mo!. the corresponding ir between 50 MHz and 7 kHz, and '¥ between -1 kbar and -1 O~ bar. In extreme cases, such as for concrete, this value can be as high as 155 kllmol with correspondingJ,.« 1 Hz. Apart from polarization phenomena arising from ionic conduction. the frequency response for EO is almost nat for ir < 50 MHz (the first adsorbed mono-molecular water layer. thus for constant water content) and forms a continuously declining curve (increasing number of water layers. thus increasing water content and decreasing water

- 22 -

potential) between 50 MHz and 17 GHz. Depending on the total surface area of the particles, the slope of the curve is most pronounced for colloidal particles « 2 J.I.In) such as clay and negligible for coarse sand (> 0.1 mm).

A consequence of the foregoing is that the hysteresis, the difference between adsorption and desorption, observed for the water retention characteristic of a porous material, applies to the dielectric spectrum as well.

References [1] U. Kaatze and V. Uhlendorf, "The Dielectric Properties of water at Microwave Frequencies",

Zeitschrift fUr Phys. Chern. Neue Folge, Bd. 126, pp 151-165, 1981. [2] P. Debye, Polar Molecules. Reinhold, New York, 1929. [3] S. Glasstone, K.1. Laidler and H. Eyring, Theory of Rate Processes, McGraw-Hill, New York,

1941. [4] E.H. Grand, R.J. Sheppard and G.P. South, Dielectric Behaviour of Biological Molecules in

Solution. Oxford University Press, Oxford, 1978. [5] J.B. Hasted, Aqueous Dielectrics, Chapman and Hall, London, 1973. [6] L.P. Harrison, "Fundamental Concepts and Definitions Relating to Humidity", in Humidity and

Moisture, editor A. Wexler, vol. 3, Chapman & Hall, Ltd., London, 1963. [6] G.N. Lewis and M. Randall, Thenno-dynamics, 2nd ed.,revised by K.S. Pitzer and L. Brewer,

McGraw-Hili, New York, 1961. [7] L.P. Harrison, "Fundamental Concepts and Definitions Relating to Humidity", in Humidity and

Moisture, editor A. Wexler, vol. 3, Chapman & Hall, Ltd., London, 1963. [8] M.C. Dobson, F.T. Ulaby, M.T. Hallikainen and M.A. EI-Rayes, "Microwave Dielectric Behavior

of Wet Soil-Part IT: Dielectric Mixing Models", IEEE Transaction on Geoscience and Remote Sensing, GE-23, No.1, pp 35-46, 1985.

[9] A. Sivhola and L.V. Lindell, "Polarizability and Effective Permittivity of Layered and Continuously Inhomogeneous Dielectric Spheres" 1. Electromagnetic Waves and Applcations, vol. 2, No.8, pp 741-756, 1988.

[10] C. Dirksen and S. Dasberg, "Improved Calibration of Time Domain Reflectometry Soil Water Content Measurements", Soil Sci. Soc. Am. J., vol. 57, pp 660-667, 1993.

[11] P. Hoekstra and A. Delany, "Dielectric Properties of Soils at UHF and Microwave Frequencies", J. of Goephysical Research, vol. 79, No. 11, pp 1699-1708, 1974.

[12] M. Kent and W. Meyer, "Dielectric Relaxation of Adsorbed Water in Microcrystalline Cellulose", J. Phys. D: Appl. Phys., vol. 16, pp 915-925, 1983.

[13] M.T. Rolland and R. Bernard, C.R. Acad. Sci., Paris, 232, 1098, 1951. [14] P.W Brown, "SMDDHC", MRS, Proc. vol. 85, pp. 83-90,1986. [15] K. van Breugel, Simulation of hydration abdfonnation of structure in hardening cement-based

materials, Thesis, Delft University of Technology, Civil Eng., 1991.

Abstract

- 23 -

The Effect of Ionic Conductivity and Dipole Orientation on the Dielectric Loss of the Hevea Rubber Latex

Kaida bin Khalid, Jumiah bt Hassan and Wan Daud bin W. Yusof Physics Department, Universiti Pertanian Malaysia

43400, Serdang, Selangor Malaysia.

The experimental results of the effect of ionic conductivity and dipole orientation on the dielectric loss of hevea rubber latex are presented. The results show that the conductive loss is dominated in the region less than 2 GHz and for the frequencies greater than 2 GHz the loss mechanism is dominated by the dipole orientation of water molecules. These effects are clearly observed as we plot the dielectric loss versus moisture content (40% to 100% wet basis) and temperature (-30°C to 50°C) at 0.2 GHz, 2 GHz and 10 GHz.

Background and Experimental Set up

Hevea rubber latex is a biological product of a complex composition. The basis components of freshly tapped natural rubber latex consist of 50-80% water, 18-45% rubber hydrocarbon an approximately 2-5% non-rubber constituents l

. The typical compositions of non-rubber (excluding water) are proteins, lipids, quebrachitol and inorganic salts2 . The total concentration of inorganic salts is approximately 0.5% of which consist of potassium (0.12-0.25%) and phosphate ions(-0.25%). Small amounts of copper, iron, sodium, calcium and magnesium are also present. Field latex can be concentrated to a higher rubber content(-60%) to make it more uniform in its quality and economically more attractive. Normally 0.3% to 0.6% of ammonia are added to latex concentrate as a preservative.

This paper will review some of previous work3•4 and current fmdings5 on the effect of ionic conductivity

and dipole orientation on the dielectric loss ofhevea latex. This aspect is studied by looking at the variation of dielectric loss with frequency, moisture content and temperature.

A series of latex solutions with moisture content ranging from 40% to 98% is prepared from latex concentrate, diluted latex concentrate, field latex, diluted field latex. Measurements of dielectric properties in frequencies ranging from 200 MHz to 20 GHz are done by using open-ended coaxial-line probe (HP85070B).

Experimental Results and Discussions

(I) Variation with Frequency

The spectrum of dielectric loss, g" ofhevea latex concentrate, field latex, diluted solutions, deionized water and solid rubber is shown in Fig.!. In the frequency range of 0.2 GHz to 20 GHz, the g" can be divided into two regions. The first region is for frequencies less than 2 GHz where g " is dominated by conductive loss while the second region with frequencies greater than 2 GHz, the loss mechanism is dominated by dipole orientation. It is clearly shown that at low frequency the value of g" depend on the strength of conductive phases in the solution while at higher frequency (-10 GHz) the g" increasing as moisture content in the solution increases.

- 24 -

50,-----------------------------------,

40

30

1·!Mac......tnU()9%) 2· DiI~ !Maea-..n..(",,) """"'!Ma(''''') 4.~FrahLo!=(lI") '~DIkm.l.zedW&Ur(Ioo%)

20\\ .~.~ _,

10V' ~~: ;:-

10 15 20 FREQUENCY (GHz.)

25

Fig (1) Dielectric properties of natural rubber latex at 26°C as a function frequency.

(ii) Variation with Moisture Content

Fig.s 2(a-c) show the variation of E" with moisture content at 0.2 GHz, 2 GHz and 10 GHz. At 0.2 GHz, E"

shows a spreading in its value which depends very much on the type of solution used. For the same moisture content, the magnitude of E" is found to be higher for the solution originated from latex concentrate compared to the solution from fresh latex or diluted fresh latex. This is due to the ammonium ions, associated with ammonia added to the latex concentrate acting as preservative.

~T,----.--------------------------~ ~ P~D.2GH&. =~ T~:t(C

40-1 ~o 44 i I

0 .... I .. a :J

30., .... C I .. .ur ++ + + 'fa n ! .. + 't-+++ lj:

20 1 .. + + ++ *+++ 'Il

I

¥ I +

IO~ Jilo+

I ljl

ffi+ o I 30 40 ~ 60 70 80 90 100

MOISTURE CONTENT (WET) <"'J

(a)

" fnqIamcy 10.9 G&. _lOt ..

'" g u ~

'" ~

g '" '" 20 f-U t.:; w is

10

10 .. " 60 70

~ I_~OOJU. ~T""""'ut

40 I ,

30 ~

:J I 01 .~~~~~~~~~ljlJ

m 40 ~ 60 m 80 90 ~

14r_~ -f-o.i.,o...tr.-t...a

I DOooIt.udI.&a.I-~'~

I·~c_ l'~w_

90

MOISTURE CONTENT (WET BASIS) (%)

(b)

100

MOISTURE CONTENT lWET BASIS) I")

(c)

Fig (2) Dielectric loss ofhevea rubber latex as a function of moisture content at 26°C and specific frequency of (a) 0.2 GHz. (b) 2.0 GHz. (c) 10 GHz ..

At transition frequency (- 2 GHz) there is no spreading of E" and E" lmost lies on a single line and almost independent of moisture content or type of solution used.

At higher frequency ( -10 GHz) E" is increasing as moisture content increases and this loss is due to the dipole orientation of water molecules. This relationship is almost unaffected by non-rubber constituents and

- 25 -

preservative. This fme relationship has been previously demonstrated by the accuracy and reproducibility of the measurement of moisture content or rubber content by transmission method6 and reflection method? . At

. the operating frequency of 10 GHz, these techniques of measurement are able to achieve an accuracy of about 1 % (as compared with standard laboratory method) and reproducibility of about 0.5%.

(iii) Variation with Temperature

The temperature dependence of e" of hevea latex at 0.2 GHz, 2 GHz and 10 GHz are summarized in Figs. 3(a-c). Over the temperature range -30oe to _3°e (solid state region), at 0.2 GHz and 2 GHz, e" shows a spreading in its value which depends very much on the type of solution used and lines coincide at about -200e with e"-O.4. This spreading is due to the different degree of concentration of dissolved ions in the samples. Since at high frequency (-1OGHz) the e" is due to the dipole orientation of water molecules, therefore as moisture content in the latex become solid at around _2°e the e" of all samples become constant at this temperature.

50~----------~----------------~

40 a::

~ 30 til g V

~ 20 v W ....J W is

10

f'REO.-O.2CHz

50

40 a:: 0 l-V .:: til 30 til 0 ....J

U

~ 20 u W ....J W is

10

0 -30

FREQ.2.2CHz 'LU.1..I'DEIONIZED ~OIL FRES w.t.t FRESH LA ....... LAT[X C

,

I J

~J -20 -10 0 10 20 0_ 30 -20 -10 0 10 20 30 40 50

TEMPERATURE (e) TEMPERATURE (e)

(a)

50,------------------------------.

Ul..t..!' DEIONIZED WATER (100.) &.&.&.&.I. OIL. FRESH LATEX t8g.} UW FRESH LATEx 56:11 u.u.JI LATEX CONCEN~AAT (3&1:)

FREO._1OCHz:

40 a:: 8 ~ til 30 til g V

~ 20 v W ....J W

is

10 ., J

./j :

o~~~~~~~~~~~~~~ -30 -20 -10 0 10 20 30 40 50

TEMPERATURE (e)

(c)

(b)

Fig (3) Dielectric loss ofhevea rubber latex versus temperature at specific frequency (a) 0.2 GHz. (b) 2.2 GHz. (c) 10 GHz ..

~~~) (Jao)

30 40 50

In the transition region (_3°e to 3°C) the figures show a steep increase in e" as a phase of latex is changing from solid to liquid. This is due to the changing in the physical state of water in latex changing from bound water to free water.

- 26 -

In the liquid state region (> 3°C) at 0.2 GHz elevations in temperature raise the mobility of ions in the solution resulting in e" increases. The gradient of the curve is dependent on the degree of concentration of dissolved ions in the sample. At 2.2 GHz , e" is decreasing as temperature increases which is similar to the trend of the deionized water. The shapes of the latex curves show a depression as compare with the deionized water which might be due to the water binding by dissolved ions. This depression increases as the conducting phases in the solution increases. The above phenomenon is almost similar to the temperature dependence for a few food products at 2.8 GHz and temperature from -20°C to60°C as reported earliers. The variation of £ " at 10 GHz is almost similar to the trend at 2.2 GHz, however with less depression as the effect of ionic conductivity is becoming ineffective.

Conclusions

In this study the effect of conductive mechanism and dipole orientation on the dielectric loss of hevea rubber latex can be clearly observed as we plot e" versus frequency, moisture content and temperature. In the range of frequency from 0.2 GHz to 20 GHz conductive loss and dipole orientation contribute additively to the total dielectric loss. Conductive loss is dominated in the region less than 2 GHz and the loss due to dipole orientation of water molecules dominated at frequencies greater than 2 GHz. This means that the frequency above 2 GHz is a suitable operating frequency for microwave moisture meter for hevea latex as dielectric loss is mainly dominated by dipole orientation of water molecules.

Acknowledgment

The work described was supported by IRPA research grant (4-07-05-021), Ministry of Science, Technology and Environment of Malaysia. The authors wish to thank research assistance, Miss Mariani Ngah and technical staff, Roslim Mohd of Applied Electromagnetic Lab. Physics Dept. UPM for their invaluable help. The authors also would like to thank Rubber Research Inst. Malaysia and Farm Dept. UPM for supplying hevea latex.

References

1. Training Manual on Analytical Chemistry, Rubber Research Inst. Malaysia (1979),63 2. Ibid. ,44 3. Khalid K.B. and Daud W.M.(1992), Dielectric Properties of Natural Rubber Latex at Frequencies from

200MHz to 2500MHz, J. Nat. Rubb. Res., 7(4), 281-289 4. Khalid KB. , Hassan J. and Daud W.M. (1994), Dielectric Properties of Hevea Latex at various

Moisture Contents J. Nat. Rubb. 9(3), 172-189 5. Hassan 1., Khalid K.B. and Daud W.M. , Microwave Dielectric Properties of Hevea Rubber Latex at

Temperatures from -30°C to 50°C (will be published in J. Nat. Rubb.) 6. Khalid KB. (1982), Determination of Dry Rubber Content of Hevea Rubber Latex by Microwave

Technique, Pertanika, 5(2) 192-195. 7. Khalid KB. (1994) Portable Microwave Moisture Meter for Lossy Liquids, Asia Pacific Microwave

Conference Proc. Dec. 6-9, Tokyo, Japan, 477-481 8. Bengtsson, N.E. and Risman, P.O. (1971), Dielectric Properties of Food at 3 Ghz as Determined by

Cavity Perturbation Technique. II. Measurements of Food Materials, J. Microwave Power, 6, 107-123.

- 29 -

MULTIFREQUENCY MICROWAVE PROBING METHOD. An ION and WATER CONCENTRATION DETERMINATION

in SALT SOLUTIONS and BIOLOGICAL TISSUES.

Serguei Y.Semenov Ph.D., Robert H.Svenson M.D.

Laser and Applied Technologies Laboratory, Carolinas Medical Center 1000 Blythe Boulevard, Charlotte, NC 28203.

Correspondence should be addressed to: Serguei Y.Semenov Ph.D., Laser and Applied Technologies Laboratory, Carolinas Medical Center, 1000 Blythe Boulevard, Charlotte, NC 28203, tel (704) 355-3215, fax (704) 355-7217.

SUMMARY. A method of multifrequency microwave probing of the dielectric properties of

solutions and tissues has been studied. The possibility of reconstruction of the ion and water concentration by this method had been demonstrated.

Theoretically the method is based on the relaxation theory of the dielectrics. Several groups have experimentally investigated the dielectric properties of biological tissues [Johnson,Guy(1); Stuchly, et al.(2); Foster, Schwan(3)]. Various models of tissue dielectric properties have been proposed [Foster, Schwan(3); Grant, et al.(4)].

Analysis was performed at a frequency range from 0.15 to 6.0GHz. Dielectric properities of water were described on the basis of the Debay theory of relaxation [Debay (5)] as a one component of the relaxation process:

where: E", - infrared permittivity; Eo - static permittivity; fo - relaxation frequency.

An ion conduction component was added to the imaginary part of the Debay equation for the modelling of the dielectric properties of the salt solutions:

(2) where:

O'ion - ion conduction component; Ey - dielectric constant of vacuum.

In addition to this, free water, bound water and protein components with correspondent volume fractions were added into the dielectric relaxation model of the myocardium:

E* m = E", + (Eo + E", ){Kj(1 + j(f/fo» + Kbj(1+ j(f/fbw» + 1/(1 + j(f/fpr»}+ j(O'io/roEy) (3)

where: Kw - water volume fraction; Kbw - bound water volume fraction; fbw - bound water relaxation frequency;fpr - protein major relaxation frequency.

More detail model description had been reported [Semenov (6)]. The experimental setup and procedures are reported. The results of reconstruction of the ion conductivity and DC resistance of the salt solution

are presented on figure 1. The sodium concentration and DC resistance of the solution have been

100.0

[%]

99.5

99.0

98.5

98.0

o 20 40

- 30 -

300 1.6 a

p

[ohm'em] 1.4

250 1.2

1.0

200 0.8

150 0.6

0.4

100 0.2

0.0

[1/ohm/mJ 0 20 40 60 80 10Q i 20 140

Na+ [mmol/LJ

Figure 1. Reconstructed ion conductivity and resistance of the salt solution.

I J[ea.sured I

I Reconstructed I ' .. ' ..

,

•

60 80 100 120 i 40

Na+ [mmol/L]

Figure 2. Reconstructed and measured water fraction in the salt solution.

- 31 -

measured independantly. The reconstructed and measured water fraction is presented on figure 2. As can be seen from the figures a good agreement was observed between the reconstructed and experimentally measured DC solution resistance. The water fraction is qualitatively reconstructed. The reasons for the observed deviation between measured and reconstructed water fraction are discussed.

The results of the reconstruction of the ion conductivity, free water and total water fractions of the myocardium in vitro under three different conditions are presented in Table 1. The correlations between the reconstructed myocardium properties and those reported by Scholtz and colleguages [7] using NMR have been observed and discussed.

The accuracy of the reconstructed properties depend on a number of factors, such as frequency range, accuracy of measurement, number of tested frequencies etc. The influence of those factors on the accuracy of the reconstructed properties have been investigated and discussed.

The proposed method of determining of ion and water concentration have been succesfully varified.

Table 1. Ion conductivity and water content of the myocardium in vitro under three different conditions. The samples temperature were 21°C.

Myocardium crion[/ohm/m] I Free Water I Total Water I Condition. Ll[%] from 1. Ll[%] from 1. Ll[%] from 1.

1 0.604 0.897 0.975

2 0.546 I -9.6 0.868 I -3.2 0.949 I -2.7

3 0.802 1+32.8 0.941 1+4.9 1.000 1+2.6

Authors gratefully acknowledge the technical assistance of Michael E.Quinn, Kathy R.Dezem and Michelle Thompson.

REFERENCES. 1. Johnson C.C., Guy A.W., "Non-ionizing EM wave effects in biological materials and systems," in Biological effects of EM radiation, Osepchuk J.M. Ed., New York, IEEE Press, 1984,47-73. 2. Stuchly M.A., Kraszewski A., Stuchly S.S., Smith A.M., "Dielectric properties of animal tissues in vivo at radio and microwave frequencies: comparison between species," Phys. Med. BioI., vol. 27(7), pp. 927-936, 1982. 3. Foster K.R., Schwan H.P. "Dielectric properties of tissue and biological materials: a critical review" in "Critical Reviews in Biomedical Engineering", 1989,17,1,25-104. 4. Grant E.H., Sheppard RJ., South G.P. "Dielectric behavior of biological molecules in solution", Clarendon Press, Oxford, 1978. 5. Debay P. "Polar molecules", The Chemical Catalog, Dover publications, Inc., New York,1929. 6. Semenov S.Y. "Comparative modelling of water, saline and myocardium dielectric properties in the microwave spectrum", in Proceeding of the 25 European Microwave Conference, 4-8 September 1995, Bologna, Italy. 7. Scholtz T.D. et all "NMR relaxation times in acute myocardium infarction: relative changes in tissue water and fat", in "Magnetic Resonance in Medicine", 1992, 23, 89-95.

- 32 -

- 33 -

A Further Study on Material Dependence of Microwave Attenuation for Moisture Determination over Wide Range of Moisture Content and Density

Zhi-Hong Ma and Seichi Okamura

Shizuoka University, Dept. of Electrical and Electronic Engineering 3-5-1 Johoku, Hamamatsu, Shizuoka, 432, Japan

Abstract. The material dependence of microwave attenuation over wide range of moisture content and density has been studied with sawdust. When the moisture content is kept unchanged, the microwave attenuation is linearly related to the density of material. The slope and cross value of this linear relationship change with the moisture content and they can be expressed as functions of moisture content. Both of the functions are nonlinear in the lower range below about 30% of moisture content and linear in the higher range above this moisture content.

I. Introduction. Many studies have been performed to obtain the relationship between the microwave parameter and the

properties of material for its moisture determination. These studies can be roughly classified as 1. The microwave attenuation is a function of moisture content of material [1]-[3]. 2. The microwave attenuation is a function of both of the water concentration in material and density of dry

material [4]-[7]. 3. The ratio of microwave attenuation versus phase shift or real part versus imaginary part of permittivity is a

density independent function of moisture content [4 H 11 ]. All of these studies reveals that the microwave attenuation measurement as a function of the properties of

material is the most important thing. However, in these studies, the microwave attenuation has not been measured over wide range of moisture content and density of material.

In fact, if the material is soft and water-absorptive, it is possible that the moisture content and density of the material take wide dynamic variation range during measurement. For this reason, it is unreliable to apply the results of these studies to such cases of moisture determination. Thus, there is a need to study the relationship between the microwave attenuation and the properties of material over wide range of moisture content and density.

II . General consideration The properties of material have the relationship expressed as eq. (1), where they are defined as follows.

M - q:>xp (1)

water concentration M: weight of the water in material per volume of the material moisture content q:>: weight of the water in material per dry weight of the material

density p : weight of the dry material per volume of the material

For obtaining the relationship between the microwave attenuation and the properties of material, it would be noticed that the microwave attenuation characteristic of the moist material is different from that of the pure water and dry material because the binding of molecules presents in the moist material [12]. However, for given material and its moisture content, even how complicated the binding of molecules is, the state of the biding of molecules is unchanged when the density of material changes. Thus, it is reasonable to think that the relationship between microwave attenuation and density of material is linear when moisture content is kept unchanged.

According to this consideration, by changing the density of material to some different values but keeping the moisture content unchanged and measuring the microwave attenuation corresponding to each of the different densities, the relationship between the microwave attenuation and the density of material can be obtained for this constant moisture content from the measurement data.

Changing the moisture content of material to some other values and repeating the same measurements, the relationship between the microwave attenuation and both of the moisture content and density of material can be obtained as eq. (2) from the measurement data, where Am is the microwave attenuation as a linear function of the density of material, s( Ip) and c( Ip) are the slope and cross value of this linear function and they are related to the

moisture content. Am - S( Ip)p + C( Ip) (2)

- 34 -

ill. Experiments 1. Samples: Sawdust (dry weight is 127.5g) was selected to make the samples. The samples of different moisture content were obtained by controlling the quantity of water to be put into the material. The samples of different density were made by pressing the piston shown in Fig. 1. 2. Measurement system: The measurement system is shown in Fig. 1. The signal of 9.4 GHz made by the oscillator is divided into input and reference signals (Pi and Pc) by the dual directional coupler. The input signal is attenuated by the sample and becomes the output signal (Pt). The output and reference signals are fed into the amplitude analyzer for measuring the microwave attenuation (dB).

horn a.n:tenn.a 9.4GHz sarnp1e

a:;cillatnr I---'--+--'---~+---I--+EH tuner e-_--{

Ff t Pr L ______________ --------------------J

Fig. 1 Experiment system The reflection occurring at the surface of the material varies over wide range because the moisture content and

density of the material are changed over wide range during the measurement. For reducing the influence of the reflection to the microwave attenuation measurement, the EH-tuner is used. The reflection signal is put into the amplitude analyzer for observation. In whole experiment, it can be reduced to about 35 dB by adjusting the HEtuner. 3. Experiments: The moisture content of the sample was made at 13 levels in the range from 0 to 153.3%. The data of these moisture content of the sample were obtained from eq. (3) by measuring the weight of wet sample mw and the weight of dry sample ffid, respectively. With each moisture content, for obtaining the relationship between the microwave attenuation and the density of material, the density of the sample was changed to different values of number of 5 or 6 in the range from 0.095 to 0.148 glcm3

, and the microwave attenuation corresponding to each of the different density of the sample was measured. The data of the microwave attenuation (Am) caused by the material was obtained from eq. (4), where Ac and Acm are the microwave attenuation caused by the empty holder (a plastic case with a cross section of 120 x 40 mm2

) and the holder filled with the sample, respectively. The data of the density of the sample was obtained from eq. (5), where V. is the volume of the sample measured at the same time of microwave attenuation measurement. In all the experiments, 13 data of moisture content, 6 data of density and 72 data of microwave attenuation were obtained as shown in Fig. 2.

m -m IP - w d X 100 (3) (4) (5)

md

IV. Experiment results and mathematical approximations The experiment results are shown in figure 2 and the followings can be known from it:

1: When the moisture content of material is kept constant, the relationship between the microwave attenuation and the density of material can be approximated by linear function expressed as eq. (2). 2: The slope S(IP) and cross value C(IP) are shown in Fig. 3 and 4, respectively. The characteristics of them

change at about 30% of moisture content. 3: In lower range of moisture content below 30%, both of the functions s( IP) and C( IP) are shown in Fig. 5 (a) and

(b), respectively. They are nonlinear and can be approximated as eq. (6) and (7). In higher range of moisture content above 30%, they are shown in Fig. 6 (a) and (b) , respectively. Both of them are linear and can be approximated as eq. (8) and (9), where SLj, CLj, SHj, and CHj are constants determined by the material.

SL(IP)-SL21P2

+SUIP+SLO (6) CL(IP)-CLlIP2

+CUIp+CLO (7)

SH(IP)-SHIIP+SHO (8) CH(IP)-CHlIP+CHO (9)

4: The experiment equation of the microwave attenuation, as the function of both of the moisture content and density of the material, can be derived as eq. (10) and (11) by substituting SL( IP), CL( IP) and SHe IP), CH( q» into

- 35 -

eq. (2) for lower and higher moisture content range, respectively.

35

30

25 ....... !8 20 '"" 8 15 .... § 10 is 5 .... .... -<

0

-5

-10

AmI. = (SL2QJL 2

+ SLlQJL + SLO)P +(CL2 QJL 2

+ CLlQJL + CLO )

AmH - (SHlQJH +SHO)P+(CH1QJH +C HO )

(10)

(11)

lIoisture content

t:. 0.0 x 4.5 x 9.0 o 14.5 + 17. 4

20.9 31. 4

<> 37. 5 o 50.2 A 71. 8 • 97.6 .. 124. 9

0.00 0.02 0.04 0.06 0.08 O. 10 o. 12 O. 14 0.16 • 153.3 Densi ty (g/cms)

Fig. 2 Relationships between microwave attenuation and density of material at 13 moisture content levels

300

250

200 § 150 -V)

100

50

o

• •

• • • •

~

o 30 60 90 120 150 180 lIoisture content (~d.b.)

Fig. 3 Relationship between the slope and the moisture content

2

o .3 -2 ~ -4

] -6 -8

-10

Fig. 4

. ; ~ . "

o

• I.

~

30 60 90 120 150 180 lIoisture content (% d. b. )

Relationship between the cross value and the moisture content

60 • 2.0 50

./ 40 ji

§ 30 20 ~

1.0 ".. g «l J : o. 0 f-~rt--+--+--i -~ 10

0 0 10 20 30 40

Moisture content (~d.b.)

~ -1.0 o 10 20 30 40

Ioisture content (~d.b.)

(a) (b) Fig. 5 Relationships between the slope and

cross value and the moisture content in lower moisture content range

300 250 200 "

~ 6 3 g ~

§150 'I'

" «l 0 > -3 (/J

!, -~100 41

~ ~ -6 50 0

0 50 100 150 200 Moisture content

(~d.b.)

u -9 -12

"\

o 50 100 150 200 Ioisture content

(~d.b.)

(a) (b) Fig. 6 Relationships between the slope and

cross value and the moisture content in higher moisture content range

- 36 -

VI. Discussion about the application of the experiment results A calibration measurement method for moisture determination of material has been given in this paper. The

calibration equations of water concentration M and moisture content Ip can be obtained as eq. (12) and (13) for

the lower moisture content range, and as eq. (14) and (15) for the higher moisture content range from the eq. (10) and (11), respectively, where F and G are functions determined by the eq. (10) and (11).

ML -FL(AmL,p) (12) IpL -GL(AmL,p) (13)

MH -FH{AmII,p) (14) IpH -GH{AmII,p) (15)

These calibration equations can be applied to the water concentration or moisture content determination over wide dynamic variation range of the moisture content and density, because the data used to derive these calibration equations have been obtained over wide range of moisture content and density of material.

However, for this moisture determination from the measurement value of microwave attenuation caused by the material, the density of material must be measured by other method because the microwave attenuation depends on both of the moisture content and density of material. This will be studied in next time.

References [1] T.Okabe, M.T.Huang and S.Okamura, A New Methodfor the Measurement of Grain Moisture Content by

the Use of Microwaves, J. agric. Engng Res. (1973) 18, 59-66 [2] Kraszewski, Microwave Aquametry - A Review, Journal of Microwave Power, 15(4), 1980 [3] Seichi Okamura, High- Moisture Content Measurement of Grain by microwaves, Journal of Microwave

Power, 16 (3&4), 1981 [4] Kraszewski, Microwave Monitoring of Moisture 'Content in Grain-Further Considerations, Journal of

Microwave Power and Electromagnetic Energy, vol. 23 No., 1988 [5] A. Kraszewski, Microwave Aquametry-Needs and Perspectives, IEEE Transactions on Microwave Theory

and techniques, Vo1.39, No.5, May 1991 [6] A. W. Kraszewski and S. O. Nelson, Wheat moisture content and bulk density determination by microwave

parameters measurement, Canadian Agricultural Engineering, Vol. 34, No.4, October / November / December 1992

[7] Meyer and W Schilz, A Microwave Methodfor Density Independent Determination of the Moisture content of Solids, 1. Phys. D: Appl. Phys., 13 (1980) 1823-30

[8] Jacobsen, R. W., W. Meyer and B. Schrage. Density Independent Moisture Meter at X-band. Proceedings 10th European Microwave Conference, 216-220. Warsaw, Poland. 1980.

[9] Kress-Rogers, E. and M. Kent. Microwave Measurement of Powder Moisture and Density. 1987. Journal of Food Engineering 6:345-356.

[10] Wolfgang Meyer and Wolfram M. Schilz, Feasibility Study of Density-Independent Moisture Measurement With Microwaves, IEEE Transactions on Microwave Theory and Techniques. VOL. MTT-29. NO.7, July 1981

[11] Wolfgang Hoppe, Wolfgang Meyer, and Wolfram M Schilz, Density-Independent Moisture Metering in Fibrous Materials Using a Double-Cutoff Gunn Oscillator, IEEE Transactions on Microwave Theory and Techniques. VOL. MTT-28. NO.12. December 1980

[12] Ebbe Nyfors & Pertti Vainikainen, Industrial Microwave Sensors, Artech House 1989 (pp.66-67)

- 37 -

A STUDY OF THE EFFECTS OF SAMPLE TEMPERATURE

ON MICROWAVE MOISTURE CONTENT MONITORING

Frank Thompson

Dept. of Mathematics & Physics, Manchester Metropolitan University Chester Street, Manchester, England.

The interaction of mIcrowave radiation with moist substances, where the

temperature of that substance changes, is of general practical interest in applications

involving drying, heating or moisture content measurements. For example, if a microwave

moisture measurement system is used at a cement producing plant for measuring the

water content of sand, it is likely that sand temperature in winter months will be close to

zero degrees Celsius whereas temperature readings in the summer may be above 30° C.

The moisture measurement system must be able to provide a temperature-independent

result so that concrete with optimum mechanical properties is always ensured. In addition

to this practical interest, there is also interest in being able to gain a deeper understanding

of the dielectric behaviour of moist substances. Various theoretical models can be establi

shed to describe the complex permittivity of these substances in which parameters such as

density, temperature and anisotropy are varied.

The presentation will contain results of permittivity measurements for a number of

materials at different temperatures. Wide band measurements using coaxial line and fixed

frequency measurements using multimode cavity will be reported. Attempts will be made

to explain the results using dielectric modelling techniques.

- 38 -

Dielectric relaxation measurements of living organs

by a time domain reflectometty

T. Umehara, N. Miura, N. Shinyashiki, S. Yagihara and S. Mashimo

Department of Physics, Faculty of Science, Tokai University,

Hiratsuka-shi, Kanagawa, Japan, 259-12

Recent development of time domain reflectometry (TDR) is remarkable. It is

shown that the TDR can be used for the permittivity measurement ofbiopolymers,

living organs, food etc. over a frequency region from 100kHz to 40GHz. In general

biopolymer such as DNA shows three obvious absorption peaks. The first observed

around 10GHz is due to reorientation of free water. The second around 100MHz is

due to a bound water to biopolymer through hydrogen-bonding. The third comes from

surface polarization. The second peak was fIrst observed by the TDR. It is shown

that the bound water plays an important role in keeping structure of biopolymers. If

biopolymer such as globular protein is frozen, another kind of water can be observed,

which is called as an unfreezable and intermediate water between free and bound

water. A molecular motion was also fIrst observed for trypsin, which is related

directly to enzyme activity. If trypsin inhibitor is added to trypsin, the motion

disappears. The TDR technique can be applied to measure the water content in living

organs such as lunge, brain, skin and also that in foods.

- 39 -

MEASURING MICROWAVE PERMITTIVITIES OF RICE WEEVILS

Stuart O. Nelson, Philip G. Bartley, Jr., and Kurt C. Lawrence Russell Research Center, Agricultural Research Service

U. S. Department of Agriculture Athens, GA 30604 USA

The permittivities or dielectric properties of materials are of major significance in the consideration of any new high-frequency or microwave energy applications involving energy absorption in biological materials and agricultural products. These properties are highly dependent on the amount of water in the materials. In considering the potential use of selective radio-frequency (RF) or microwave dielectric heating to control insects in cereal grains, the need has been evident for data on the permittivities of the insects as well as that of the grain [1], [2]. Although assessments of potential use of RF and microwave energy have not looked promising from an economic viewpoint [3], continual concern about health hazards of chemical pesticides necessary to control stored-grain insects suggests that exploration of alternative physical insect control methods may be advisable. Thus, obtaining information on the permittivities of insects over wider frequency ranges may be justified to examine opportunities for more efficient control of insects by these means.

The permittivities, E' - jE", where E' is the dielectric constant and E" is the loss factor, of bulk samples of adult rice weevils were measured over the frequency range from 0.2 to 20 GHz at temperatures from 10 to 65 ·C with an open-ended coaxial-line probe, network analyzer, and a sample temperature control assembly designed for the measurements (Fig. 1). The adult insects were collected from laboratory cultures of rice weevils, Sitophilus oryzae (L.) , reared in hard red winter wheat, Triticum aestivum L., and, after anaesthitization with carbon dioxide, were placed in the stainless steel sample cup as shown in Fig. 1 and compressed by insertion of the coaxial-line probe. The insects do not consitute a very homogeneous sample, because their dimensions are of the same order as the 3-mm diameter of the coaxial line. The mean sample density was calculated from the sample weight and the volume occupied in the sample cup with the coaxial probe in place.

Repeated measurements were highly variable because of the nonhomogeneity of the insect sample, even when compressed, and because the mean sample bulk densities did not accurately reflect effective densities of the bulk rice weevil samples in the small volume of sample sensed by the coaxial-line probe. Density corrections based on earlier permittivity measurements on bulk rice weevil samples at 9.4 GHz in waveguide sample holders, at known sample densities (Fig. 1) [4], removed much of the variability. The corrections utilized the linear relationship between the cube root of the dielectric constant and bulk density, which also permitted estimates of the weevil body permittivities to be obtained with the Landau & Lifshitz, Looyenga equation for dielectric mixtures [2]. The necessary value for the adult rice weevil density was also available from the earlier studies [4]. The moisture contents of the rice weevils used in the new measurements, as determined by oven drying insect samples for 16 h at 105 ·C, averaged 49%, which agreed closely with that of insects used for the earlier measurements. Estimated dielectric constants and loss factors of the insects from averages of eight different measurement sequences carried out over the 0.2- to 20-GHz frequency range are presented graphically for temperatures from 15 to 65 ·C in Fig. 3.

Estimated permittivity values for the rice weevil bodies from this work are somewhat lower than those estimated from the previous measurements [4], where extrapolation of a quadratic fit of permittivity-vs.-density data gave an estimate of 31.5 - j12.7 at 9.4 GHz and 24 ·C. The mean estimated value at 9.4 GHz and 25 ·C, from the eight measurement sequences described here, is 25.3 - jI2.0.

More accurate determinations of the permittivities of these insects and other species of interest will require measurements by more appropriate techniques. However, the measurements reported here established that the insect loss factors at frequencies up to at least 20 GHz do not provide opportunities for better selective dielectric heating of the insects with respect to a host medium of grain than the frequency range from 10 to 100 MHz identified earlier as the best frequencies for differential heating of the insects for lethal purposes [1], [5].

- 40 -

Fig. 1. Sectional view of stainless steel sample cup, Delrin water jacket, and supporting platform for measurements on bulk rice weevil samples with coaxial-line probe, showing portions of supporting clamps.

rz 2.3

;5 '" a 2.2 u u DC g 2.1

w ~ 2.0 a .... a a 1.9 a: UJ CD a 1.8

0.50 0.55 0.60 0.65 0.70 0.75 o.eo 0.85 0.90

BULK DENSITY, GlCM'

Fig. 2. Linear relationship between ((/)1/3 and bulk density, p, for adult rice weevil samples at 9.4 GHz at 24 "C. ((/)1/3 = 0.71125 + 1.93049 p, r2 = 0.9996.

eo s: ~ 70 z a ~ 60

~ ~ 50

1 o 40

~ ~ 30

'" w

20

a: 50 g u 0( u. 50 ., ~ .... u 40

~ frl m 30 i5 o i 20

~ w 10

10'

10'

10'

FREQUENCY, HZ

10'

FREQUENCY, HZ

o 65·C

o 55

" 45 o 35 .. 25 o 15

Fig. 3. Mean values of estimated permittivities of adult rice weevils at indicated temperatures.

REFERENCES

[1] S. O. Nelson and 1. F. Charity, "Frequency dependence of energy absorption by insects and grain in electric fields," Trans. ASAE, vol. 15, no. 6, pp. 1099-1102, 1972.

[2] S. O. Nelson, "Dielectric properties of agricultural products -Measurements and applications," CEIDP Digest of Literature on Dielectrics, A. deReggie, Ed., IEEE Trans. Electr. Insul., vol. 26, no. 5, pp. 845-869, 1991.

[3] S. O. Nelson and L. E. Stetson, "Possibilities for controlling insects with microwaves and lower frequency RF energy," IEEE Trans. Microwave Theory Techn., vol. MTT-22, no. 2, pp. 1303-1305, 1974.

[4] S. O. Nelson, "Microwave dielectric properties of insects and grain kernels", J. Microwave Power, vol. 11, no. 4, pp. 299-303, 1976.

[5] S. O. Nelson and 1. E. Stetson, "Comparative effectiveness of 39- and 2,450-MHz electric fields for control of rice weevils in wheat," J. Econ. Entomol., vol. 67, no. 5, pp. 592-595, 1974.

- 41 -

ON THE LOCALIZED MICROWAVE OVERHEATING IN AN ORGANIC WATER SOLUTION

A.F. Korolev*, Associated Professor A.I. Kostienko*, Associated Professor A.P. Sukhorukov*, Professor I. V. Timoshkin * *, Researcher A. Pulino***, Doctor

*Physical Department, Moscow State University, Moscow, 119899, Russia, fax: 7 (095) 932-8820, e-mail: [email protected] **Theoretical Department, Institute for High Temperature of Russian Academy of Sciences, Izhorskaya st. 13/19, Moscow, 127412, Russia, e-mail: [email protected] ***Gamma Tel., Cambridge, MA, 02139, USA

Understanding of the process of microwave action on the biological tissue and materials is important for prevent of hazards at the microwave processing of a products and materials and other practical situations. Knowing of the features of this processes is necessary for exception of the microwave hazards for a humans dealing with the electromagnetic radiation. In this connection the influence of the microwaves on the biological materials, water solutions and pure water is the subject of many experimental and theoretical investigations. Some aspects of this problem including the so-called "non-thermal" effects and arising of metastable states in liquid water, under the influence of microwaves are discussed by the authors previously [5].

The aim of the present paper is the investigation of the localized overheating of the protein macromolecules. Usually this effects are considering at the electron transfer and .temperatures of the overheating can reach 20 K [6]. We will consider the problem of the localized overheating in water protein solution under the action of microwaves. The theoretical investigations of this effect have been carried out and first experimental confirmation has been obtained.

At the considering of the process of the interaction of electromagnetic radiation with the dilute protein water solutions the macroscopic shell model suggested on the base of the results of dielectric studies of organic solutions [3,4] is used. In this model the protein macromolecule is represented as a spherical core with the radius r 1 and it is surrounded by a concentric shell of a hydration (bound) water molecules with the strong hydrogen bonds. The hydration shell radius is r2' The macromolecule with the hydration shell is immersed in a continuum of free water with dissolved ions. The physical properties of the hydration water are different from the properties of free water. The relaxation time of pure water "Cj is smaller than the relaxation time of ~ound water "Cb , and the relation {"Cbl"Cj} does not depend on the concentration of organic molecules for dilute solutions [3] and its value is closed to 2.5. This fact leads to the difference in the dielectric properties between hydration and free water. The imaginary parts of the relative permittivity of hydration and free ",:,~ter are Eb" and E/ correspondingly. The imaginary part of permittivity of the protein is E] and in the frequency range of microwaves from 100 MHz up to least 1 GHz the average energy deposition per unit volume in hydrogen water is greater than that in free water [2]. The average energy deposited in one cycle at any point of the solution is given by the following expression

- 42 -

(1)

where CJn is the ionic conductivity, En is the electric field strength in the n-th region,

The average energy deposited per unit volume

- I r -Po =-Jl PdV Vo 0

where V n is the volume of the region considered.

(2)

For the calculation of the temperature of the heating of the albumin molecule in water under the influence of the microwave with the frequency of 900 MHz we use the system of the equations of the nonstationary heat conduction

(3)

where the indexes 1, 2 and 3 correspond to the temperatures of protein molecule, hydration water and free water; Tj is the temperature; a{I'J is the temperature conductivity; Cj , Cb are the heat capacity and Pj . Pb are the density of the free and hydration water, f and I are the heat sources in the regions of the pure and hydration water; functionf(t) defines the character of the microwave radiation.

The initial and boundary conditions in this model are given by the equations

Ii (r,O) = 12 (r,O) = 13(r,O) = To

Ii (rb t ) = 12 (rl ,t)

T2(r2,t) = 13(0.,t)

T3(r3,t) = To

11 (O,t) < 00

-xlCli) 811 I = -X2(T2)812

1

8r r. t 8r r. t ., I'

-X2(12)8T2

1 = -X3(13)8T3

1 8r r. t 8r r. t

l' l'

(4)

where Xn is the heat conductivity.

At the calculation we suggest that

Xl ~ X2 = X

- 43 -

PfI:::Pb=P

cf I::: cb = C

The sources in the right hands of Eq.3 characterize the energy deposition and can be write by using the following expressions

(5) 2

2 1 ( ")-Iq = -!t (j +E<fOE3 E av 2V3 J

For determining of the dielectric properties of the free and hydration water one can used the Debay equation

(6)

The volume V3 is defined by the overall volume of the irradiated sample ~ and protein concentration n

V3 = VS(1- V2n) where V 2 is the volume of the hydration shell.

(7)

The computer calculations have been carried out for the egg albumin water solution with the concentration n=0.2 (2%) . In the calculations the microwave source has the following parameters: the pulse duration 't =10 ms and power P= 1350 kWat the frequency 900 MHz. The frequency of the pulse period-to-pulse duration ratio was 1 rr. The numerical magnitUdes of the using values are listed in Table 1 [1]. It has been obtained, that for this parameters of the microwave radiation the mean temperature of the overheating of the protein macromolecule is 5-1 OK.

Table 1

a [m"ls] P [kglm"'] X [JIm s K] C [J/kg K] rob [GHz] rof [GHz] (j [ohm/m] 7 10-6 1000 O.l 1.2 10"' 1.5 20 0.68

The experimental investigation of the localized overheating has been carried. The parameters used in computer calculations correspond to the original experimental microwave source. The protein solution has been treated by two method: microwave irradiation and heating in the thermostat unit. The temperature of the heating was 47C. The

- 44 -

changing of the denaturated part of the protein M in the treated solution relative to the control untreated sample (J 00%) has been detennined. The value M is measure of the influence of the external factors (microwave energy, temperature) on the protein.

We have established that the heating by the microwaves at the frequency 900 MHz leads to appearance of the excess part of the denaturated protein (M= 15%) in compare with the conventional heating in the thennostat unit (M=10% ) up to the same temperature (47C). This effect probably results from the excess microwave energy deposition in the hydration shell of the protein molecule, it is possible due to the difference in the dielectric properties between pure and hydration water.

References [1] H. Heurath, K. Bailey, The protein chemistry. Biological activity and method, vol. 1, p. A,B, Academic Press. [2] V.E.R. McClean, R.G. Sheppard and E.H. Grant, "A Generalized Model for the Interaction of Microwave Radiation with Bound Water in Biological Material", JMicrowave Power, vol. 16, no. 1, pp. 1-7, 1981. [3] U. Kaatze "On the existence of bound water in biological system as probed by dielectric spectroscopy", Phys. Med. Bioi., vol. 35, no. 12, pp. 1663-1681, 1663-1681. [4] S. Mashimo, N. Miura and T. Umehara "The structure of water detennined by microwave dielectric study on water mixtures with glucose, polysaccharides, and L-ascorbic acid", J.Chem.Phys. vo1.97, no. 9, pp. 6759-6765, 1992. [5] M.G. Gapochka, L.D. Gapochka, A.F. Korolev, A.I. Kostienko, A.P. Sukhorukov, LV. Timoshkin and A. Pulino, "The effect of microwave radiation on liquid water", in Proc. 25th European Microwave Conf, vol. 2, pp. 849-851, 1995. [6] V.D. Lakhno, L.A. Uvarova, "Effects of localized heating of the proteins at the electron transfer", J.Chem.Phys., vol. 69, no. 8, pp. 1387-1390, 1995 (in Russian).

- 45 -

EFFECTS OF MICROWAVE EXPOSURE ON THE EXTRACTION EFFICIENCY AND TEMPERATURE PROFILES OF SOLVENTIMATERIAL MIXTURES

M. M. Punt1&2,G. S. V. Raghavant, M. Fakhourit, V. Yalayant, J. M. R. Belange~ and J. R. J. Pare2

Environment Canada's patented Microwave-Assisted Process (MAPTM) is a solvent extraction technique whereby target compounds are quickly extracted from solid materials by applying microwave energy to a mixture of the solid material and a low dielectric constant solvent. The technology has been proven effective for analytical sample preparation and industrial extraction purposes with applications varying from the extraction of natural products (essential oils, pharmaceuticals, pesticides) from plant material to the extraction of contaminants from soil.

In conventional solvent extraction processes, the extraction of compounds is enhanced by the heating and mixing of the solid/solvent slurry. Preliminary research has shown that the rapid extraction rates obtained using the MAP technique is not merely a result of the microwave energy increasing the overall temperature of the slurry but, because of the difference in the dielectric properties of the components of the mixture, it is rather a result of the selective and localized heating within the slurry. It has been demonstrated that this phenomenon causes physical changes to the structure of the solid thus facilitating the transfer of the target compounds from the solid to the solvent. It is uncertain, however, to what extent the localized heating contributes to the enhanced diffusion rate of the target analyte into the solvent. This parameter will be investigated further indirectly through measurements of the effect of microwave energy on extraction efficiencies and the temperature profiles ofvarlous solvent/material matrices.

1 Department of Agricultural & Biosystems Engineering, McGill University, MacDonald Campus, 21111 Lakeshore Road, Ste-Anne-de-Bellevue, PQ, CANADA H9X 3V9

2 MAP Division, Environmental Technology Centre, Environment Canada, 3439 River Road Ottawa, ON, CANADA KIA OH3

Abstract:

- 46 -

Microwave Absorption of Wet Snow

Christian Matzler and Thomas Weise

Institute of Applied Physics

University of Bern, CH-3012 Bern

e-mail: [email protected]

Fax: +41 31 6313765

When liquid water is formed in snow, microwave absorption is drastically increased. This effect has been confirmed many times by both active and passive remote sensing studies. The transition occurs very frequently in nature, e.g. during the diurnal cycles of warming and cooling. However, it is very difficult to actually measure the opacity of a slab of wet snow because the attenuation is so large, especially above 10 GHz. For the study of the dielectric properties of wet snow such measurements would be valuable. They can be used for testing dielectric mixing models which then allow to determine the amount and shape of the liquid particles.

A new and simple technique to be applied in the field was recently developed!, and first results were obtained from melt-freeze cycles observed in the surface layer. The technique consists of measuring the emitted microwave radiation of the layer with portable radiometers at one ore more frequencies. In order to enhance the effect of snow wetness, a flat metal reflector was inserted into the snow at the lower boundary of the slab to be investigated.

First measurements were made at 21 and 35 GHz in a fresh snow layer in the Austrian Alps in April 1994. During dry-snow condition the brightness temperatures Tb were low (=60K). With the occurrence of tiny amounts of liquid water Tb rapidly increased by tens of degrees, sometimes reaching near-blackbody values. The increase was larger at the higher frequency.

Based on physical mixing theory, the measured increase of Tb and thus of the opacity is interpreted in terms of the depolarization factors, of the volumetric liquid-water content Wand of the liquid water column height dw=d· W where d is the thickness of the wet snow layer. The observed spectral response requires extremely eccentric water particles, and the retrieved dw

values were on the order of 0.1 mm. The liquid water was present at a depth between 2 and 10 cm below the snow surface. The minimum detectable dw was less than 0.001 mm. An interesting phenomenon during the warming phase was the observation that the hand test (by forming a snow-ball) sometimes indicated the presence of liquid water in the fresh snow before liquid water was detected by the microwave method.

1 T. Weise, "Radiometric and structural measurements of snow", Doctoral Thesis, Institute of Applied Physics, University of Bern (1996).

! .

- 47 -

Microwave Dielectric Properties of Soil and Vegetation and Their Estimation from Spaceborne Radar

M. Craig Dobson and Kyle C. McDonald

Radiation Lab, EECS Dept., The Univ. of Michigan, Ann Arbor, MI 48109-2122 Tel.: 1+313-747-1799, Fax.: 1+313-747-2106, email: [email protected]

*Jet Propulsion Lab, MS 300-233,4800 Oak Grove Dr., Pasadena, CA 91109-8099 Tel.: 1+818-354-3263, Fax.: 1+818-354-9476, email: [email protected]

Abstract - This paper is largely tutorial in nature and provides an overview of the microwave dielectric properties of certain natural terrestrial media (soils and vegetation) and recent results in estimating these properties remotely from airborne and orbital synthetic aperture radar (SAR). Sections present (1) instrumentation for laboratory and in situ measurement of the relative dielectric constant, (2) a synopsis of laboratory measurements, (3) examples of in situ measurements, (4) the relationship between dielectric properties and radar backscatter, and (5) a summary of recent progress in estimation of surface dielectric properties from SAR observations.

1.0 Introduction

The dielectric resonance of both pure and saline water lies in the microwave portion of the spectrum, and the relative dielectric constant of liquid water can be 1 to 2 orders of magnitude greater than anhydrous terrestrial media. Hence, the microwave dielectric constant is very sensitive to water content and can be highly diagnostic of other system attributes. There are secondary dependencies of the dielectric constant on dry density, chemical composition and the temperature of the medium.

Because of the near-transparancy of the atmosphere, the potentials of active radars (scatterometers and imagers) and passive radiometers have been (and continue to be) explored for earth observation. The emission and scattering behaviors of a medium are controlled by two sets of factors: (1) the geometrical properties describing boundary conditions and (2) dielectric properties. Most theoretical developments have concerned description of the forward problem, that is solving for scattering and emission as functions of media properties, whereas the inverse problem is of greatest interest to the application of remote sensing techniques and has received less attention to date. The key in the inverse problem is often to decouple the geometric (structural) effects of the medium from the dielectric effects on the measured signal.

2.0 Instrumentation for Dielectric Measurements

Several methods may be used to measure dielectric properties of natural media and include transmission, reflection, and resonance techniques. The development of automatic network analyzers and sweep frequency measurement techniques has lead to the development of faster measurement techniques. Emphasis is placed on techniques specifically suitable for vegetation and soil media. Most transmission techniques utilize either a waveguide or a free-space system. The amplitude and phase of the transmission coefficient through a substance of unknown dielectric constant is measured and an iterative procedure is used to express the unknown dielectric constant in terms of the propagation constant of the dielectric filled sample holder. Full details of the procedure are given in [1].

Reflection techniques utilize slotted line or coaxial probe systems. These methods measure the magnitude and phase of the reflection coefficient at the end of a transmission line and relate it to the

- 48 -

dielectric constant of the medium placed at the end of the transmission line. Broadhurst [2] utilized a slotted line system in which he measured dielectric constant by relating it to the admittance of a coaxial transmission line with the material specimen occupying some of the space between the coaxial conductors. The measurement of the admittance of a coaxial line is equivalent to the measurement of reflection coefficient at the same plane of reference. Open-ended coaxial lines have been used successfully by many researchers [3-5]. This approach applies a standard reflection coefficient measurement system with the probe tip in contact with the dielectric sample acting as the termination load.

Resonant cavities may also be applied to measure dielectric constant. In the filled-cavity approach, the dielectric constant of a material is determined by the shift in the resonant frequency and the quality factor of a resonant cavity [6]. Although theoretically this technique is straightforward to apply, from a practical standpoint it may be very difficult to completely fill the resonant cavity with the solid dielectric material (e.g. vegetation) without some air pockets remaining. This complicates the inference of the dielectric constant of the material.

The partially-filled cavity technique, also known as the perturbation technique, makes use of small changes in the cavity's resonant frequency attainable by proper selection of the sample size. The derivation is based on the assumption that either the sample size or its dielectric constant is small enough that the field structure inside the cavity is not substantially changed by insertion of the sample. The shape of the sample is also an important factor in determining the appropriate formula to be used. Spheres, discs, and needles are the most commonly used shapes.

Various techniques have been applied for in situ characterization of the dielectric properties of vegetation canopies. For measurement of vegetation dielectric constant, the waveguide and filled cavity techniques are not suitable because it is not possible to avoid air gaps between the vegetation sample and the measurement assembly. Also, it is impossible to achieve the smooth surface required of the sample for free-space system measurements. Slotted line and partially filled cavity measurements always suffer from inaccuracies in thickness measurements of the sample. Soil and vegetation dielectric properties are typically monitored using time domain reflectometry or coaxial probe systems.

A field portable dielectric probe (PDP) has been developed and marketed by Applied Microwave Corporation [7] and has proven very useful in many remote sensing studies in which microwave dielectric properties of vegetation and soil must be characterized. The PDP has proven useful in characterizing dielectric properties in a number of studies [8-10]. However it is an inefficient instrument for purposes of studies involving more intense analyses of the temporal response of physiologic and dielectric properties of vegetation. Consequently, a design effort was undertaken in order to facilitate the measurement scheme necessary to carry out experiments that involve continuous long-term monitoring of canopy dielectric properties [11]. The dielectric monitoring system developed for this application incorporates the PDP unit with a switching network and data logger assembly that permits autonomous monitoring of dielectric constant in a near-continuous fashion.

3.0 Laboratory Measurements of Dielectric Constant

The average dielectric properties of the soil medium depend upon bulk soil characteristics, such as moisture, density, particle-size distribution, mineralogy and fluid chemistry. Soil is commonly considered to be a four component system consisting of soil solids, air and water in turn consisting of 'bulk' and 'bound' phases.

A number of studies [12,13] have reported on the dielectric properties of rocks mostly as functions of composition (mineralogy) and bulk density (i.e., the mass to volume ratio of powdered specimens). A reflection technique using an open-end coaxial probe was used to measure the relative permittivity of 80 terrestrial rock samples over the range from 0.5 - 18 Ghz [13]. The dielectric loss factor of these same rocks was measured using resonant cavity techniques at five frequencies from 1.6 - 16 GHz. The real part of the permittivity was found to be independent of frequency. The effects of the specific density of rock were found to account for approximately 50% of the variance between rock samples using a geometric mean fromulation. An additional 28% of the variance was found to be related to differences in bulk chemical composition of the samples that are related to mineralogy. The loss tangent was always small and found to decrease with frequency and vary by mineral class: carbonate or silicate (subdivided into volcanic, plutonic and sedimentary) and also the bulk chemical composition. Dielectric loss was found to be poorly correlated with rock density.

The microwave dielectric properties of bulk soils were examined in the laboratory by a number of investigators in the 1970s [14, 15]. Theoretical dielectric mixing formulations treating a two-component

- 49 -

system of soil and water could not account for all of the observed effects. As a consequence, relatively simple empirical models were developed relating the dielectric behavior to readily measured bulk soil physical properties such as density and particle size distribution [1, 17-19]. In addition, more complex semi-empirical and theoretical mixing models incorporating up to four-components (soil, air, bulk water and bound water) were developed [19]. These provide better agreement with observations but are more complicated and can require knowledge of additional soil properties such as cation exchange capacity, salinity and specific surface. These studies show that the soil dielectric properties are primarily controlled by moisture content, bulk density, texture and clay mineralogy. No studies have reported on the effects of the organic content of soils.

Similar laboratory studies have performed on the dielectric properties of vegetation as functions of density, moisture content and temperature and used to generate fairly simple expressions to estimate vegetation dielectric properties from 1 - 18 Ghz via a dual-dispersion model that uses a sucrose solution to model the behavior of 'bound' plant fluids [20]. The model has been found to work well for many plant materials and the results compare favorably to in situ observations.

4.0 In situ Measurement of Dielectric Constant

In situ measurements of dielectric porperties have been made as a component of a number of recent remote sensing investigations. The EOS Synergism Study examined the temporal variability of the optical reflectance and microwave backscatter caused by diurnal change in canopy properties of interest to ecosystem modelers [21]. The experiment was designed specifically to address diurnal changes in canopy water status, including water potential and vegetation water content, that relate to canopy transpiration. As part of the synergism study, an L-band (1.2 GHz) field portable dielectric probe was used to obtain measurements of the dielectric constant of the vegetation and soil. Diurnal observations were made of the soil surface, the main tree stems, the bark and the higher order green stems. The most dramatic diurnal changes in dielectric constant were found in the main stems and the soil surface. The diurnal behavior observed in the dielectric of the soil surface is explained by irrigation and subsequent drying processes. To quantify behavior of the vegetation dielectric constant, a single tree was instrumented to monitor the time dependence of e in the main stem (trunk) at several selected heights. Probe tip depth increments were selected to obtain a depth profile at depths of 1 cm, 2 cm, 4 cm, and 7.5 cm for a 15 cm diameter trunk.

The time dependence of er was found to be greatest at the 2 cm depth (nominally in the hydroactive xylem tissue) and to vary by an order of magnitude over a 24 hour period. There was little diurnal variation observed in the bark and the diurnal variation decreased with increasing depth into the main stem. The drop in the trunk dielectric constant in the afternoon occurs coincidentally with the drop (to greater negative values) in xylem water potential. Following that, the rise in dielectric constant in the late afternoon and evening nearly coincides with the rise in water potential.

In March 1988, a campaign was carried out at the Bonanza Creek Experimental Forest (BCEF) near Fairbanks, Alaska, in order to examine the seasonal transitions in boreal forest with synthetic aperture radar (SAR) [10]. As part of that campaign, a PDP was used to measure L-band dielectric properties in white spruce, black spruce and balsam poplar trees. Data were collected as a function of depth into the tree

trunks on two dates with air temperatures of 20C and -140C. The freezing of the liquid water within the trunks caused the dielectric constant of the trunks to drop from 35 to below 5 for the species measured.

To develop a better understanding of the relationship between vegetation dielectric constant and vegetation physiologic activity, a number of field exercises have been performed using the single and multi-channel dielectric monitoring systems in concert with xylem sap flux and microclimate sensors that allow for characterization of the hydraulic response of the vegetation. During 1994, several stands at the Boreal Ecosystem-Atmosphere Study (BOREAS) in Canada were instrumented with this equipment throughout much of the growing season [22, 23]. Mean xylem dielectric constant showed a decrease with higher vapor pressure deficit and sap flux density but the trend was not significant. Individual trees varied widely in trend and diurnal amplitude of xylem dielectric constant changes. Similar results were observed in previous studies in Alaska [24].

5.0 Radar Backscatter Response to Dielectric Constant

Results of field measurements carried out with dielectric and hydrologic monitoring equipment have shown that a link exists between canopy water status and dielectric constant. Thus, since dielectric

- 50 -

constant is sensitive to changes in the canopy water status, and since radar is sensitive to dielectric constant, then it should be possible to couple radar backscatter to canopy water status via the dielectric constant. This would greatly improve our ability to estimate canopy carbon, water and energy budgets using remote sensing techniques.

A three day series of scatterometer observations of a walnut orchard was obtained during the August 1987 EOS Synergism Study [9]. As part of the analysis of those measurements, the vegetation dielectric constant measurements and corresponding soil and branch dielectric measurements were used as input to the MIMICS radiative transfer model [25]. The model successfully predicts the level of the measured backscatter along with the decreasing trend in backscatter observed over the three day period.

Furthermore, MIMICS predicts the 1-2 dB dip seen in aOvv and aOhv in the early afternoon each day.

5.0 Retrieval of Dielectric Constant from Backscatter Measurements

Recently, a constellation of earth-orbiting synthetic aperture radars has been put into place (i.e., ERS-l/2, JERS-l and Radarsat), and research interest has begun to focus on the inverse problem and the development and demonstration of applications. One such area of application is the estimation of soil moisture via its dependence on the dielectric properties of soil. The critical part is to deconvolve the effects of geometric attributes (the soil surface roughness) and the effects of any overlying vegetation from the dielectric effects. All of these SAR systems are single polarization and frequency systems, so multiple channels of information are not available to estimate these three sets of parameters. However, on the basis of single frequency polarimetric data, an empirical approach was developed to estimate both moisture (via the dielectric) and roughness by using polarization ratios [26] for bare soil (non-vegtetated conditions. This approach was modified to use only hh and vv polarizations and applied to airborne polarimetric SAR data and SIR-C data obtained at L-band over a watershed in Oklahoma [27]. In applying the inversion algorithm in the image domain for a time series of data, the results show the spatial pattern of the dry-down sequence over a large area after saturating rains. The technique did not work very well for vegetated conditions. A different inversion algorithm that also uses polarization ratios at L-band (hh and vv) was developed using empirical coefficients applied to a simplified first order solution to the integral equation method [28]; this algorithm solves for dielectric constant and two roughness parameters. This approach has also been tested using the same data from Oklahoma. The strengths and weaknesses of the two approaches are compared by Wang [29]. All approaches solve for dielectric constant and then infer soil moisture using an empirical model along with some assumptions about properties. At present, there are no algorithms that opperate well for soils covered with vegetation (other than short grass).

References:

[1] M.T. Hallikainen, F. T. Ulaby, M. C. Dobson, M. A. EI-Rayes, and L. Wu, "Microwave dielectric behavior of wet soil - Part I: Empirical models and experimental observations from 1.4 to 18 GHz," IEEE Trans. Geosci. Rem. Sens., vol. GE-23, no. 1, pp. 25-34, 1985.

[2] M.G. Broadhurst, 1970: "Complex dielectric constant and dissipation factor of foliage," NBS Report 9592, October 1970.

[3] E.C. Burdette, F. L. Cain, and J. Seals, "In vivo probe measurement technique for determining dielectric properties at VHF through microwave frequencies," IEEE Trans. Micro. Theory and Tech." vol. MIT-28, no. 4, pp. 414-427, 1980.

[4] T.W. Athey, M. A. Stuchly and S. S. Stuchly, "Measurement of radio frequency permittivity of biological tissue with an open-ended coaxial line: Part I," IEEE Trans. Micro. Theory and Tech., vol. MIT-30, No.1, pp 82-86, 1982.

[5] M.A. Stuchly, T. W. Athey, G. M. Samaras, and Ge. E. Taylor, "Measurement of radio frequency permittivity of biological tissues with an open-ended coaxial line: Part II Experimental results," IEEE Trans. Micro. Theory and Tech., vol. MIT-30, no. 1, pp. 87-92, 1982.

[6] F.T. Ulaby and R. Jedlicka, "Microwave dielectric properties of plant materials," IEEE Trans. Geosci. Rem. Sens., vol. GE-22, pp. 406-414,1984.

[7] D.R. Brunfeldt, "Theory and Design of a Field-Portable Dielectric Measurement System," Proc. ofth e 1987 Int. Geosci. Rem. Sens. Symp., Ann Arbor, Michigan, May 18-21, pp. 559--563,1987.

[8] M.C. Dobson, R. de la Sierra and N. Christensen, "Spatial and temporal variation of the microwave

- 51 -

dielectric properties of loblolly pine trunks," Proc. of the 1991 Int. Geosci. Rem. Sens. Symp., 3-6 June, Espoo, Finland, 1991.

[9] KC. McDonald, M. C. Dobson and F. T. Ulaby, "Modeling multifrequency diurnal backscatter from a walnut orchard," IEEE Trans. Geosci. Rem. Sens., vol. 29 no. 6, pp. 852--863,1991.

[10] J.B. Way, E. Rignot, K. McDonald, R. Oren, R. Kwok, G. Bonan, M. e. Dobson, L. Viereck and J. E. Roth, "Evaluating the type and state of Alaska taiga forests with imaging radar for use in ecosystem flux models," IEEE Trans. Geosci. Rem. Sens., vol. 32 no. 2,1994.

[11] KC. McDonald, W. Chun, R. Zimmermann, 1. B. Way, and A. Guerra, "Automated instrumentation for continuous monitoring of the dielectric properties of woody vegetation," In prep., 1996.

[12] D.W. Strangway, G. R. Olhoefi, W. B. Chapman and J. Cames, "Electrical properties oflunar soils: Dependence uopn frequency, temperature and moisture," Earth Planet Sci. Lett., vol. 16, pp. 275-281,1972.

[13] F.T. Ulaby, T. H. Bengal, M. C. Dobson, 1. R. East, 1. B. Garvin and D. L. Evans, "Microwave dielectric properties of rocks," IEEE Trans. Geosci. Rem. Sen., Vol. 28, No.3, pp. 325-336, 1990.

[14] F.E. Geiger and D. Williams, "Dielectric constants of soils at microwave frequencies," Nat. Aeronautics and Space Admin., Washington, D. C., Rep. NASA TMS-65987, 1972.

[15] P. Hoekstra and A. Delaney, 1974: "Dielectric properties of soils at UHF and microwave frequencies," J. Geophys. Res., vol. 79, pp. 1699-1708, 1974.

[16] D. Wobschall, "A theory of the complex dielectric permittivity of soil containing water, the semidisperse model," IEEE Trans. Geosci. Electron., vol. 15, no. 1, pp. 29-58, 1977.

[17] J.R. Wang,"The dielectric properties of soil-water mixtures at microwave frequencies," Radio Science, vol. 15, pp. 977-985, 1980.

[18] J.R. Wang and T. J. Schmugge, "An empirical model for the complex dielectric permittivity of soils as a function of water content," IEEE Trans. Geosci. Rem. Sen., vol. GE-18, pp. 288-295,1980.

[19] M.C. Dobson, F. T. Ulaby, M. EI-Rayes and M. Hallikainen, "Microwave dielectric behavior of wet soil, Part II: four-component dielectric mixing model," IEEE Trans. Geosci. Rem. Sens., vol. GE-23,no. l,pp.35-46, 1985.

[20] F.T. Ulaby and M. A. El-Rayes, "Microwave dielectric spectrum of vegetation - Part II: Dualdispersion model," IEEE Trans. Geosci. Rem. Sen., vol.GE-25, no. 5, pp. 550-557,1987.

[21] J.B. Way, J. Paris, M. C. Dobson, K C. McDonald, F. T. Ulaby, J. A. Weber, S. L. Ustin, V. C. Vanderbilt and E. S. Kasischke, "Diurnal change in trees as observed by optical and microwave sensors: The EOS Synergism Study," IEEE Trans. Geosci. Rem. Sens., vol. 29,pp. 807-821, 1991.

[22] KC. McDonald, R. Zimmermann, R. Oren and J. B. Way, "Dielectric and hydraulic response of selected forest canopies at the BOREAS test sites in Canada," Proc. of the 1995 Int. Geosci. Rem. Sens. Symp., Fierenze, Italy, 1995.

[23] R. Zimmermann, K. e. McDonald, R. Oren and 1. B. Way, "Xylem dielectric constant, water status, and transpiration of young Jack Pine (Pinus banksiana Lamb.) in the southern boreal zone of Canada," Proc. of the 1995 Int. Geosci. Rem. Sens. Symp., Fierenze, Italy, 1995.

[24] R. Zimmermann, KC. McDonald, J. B. Way and R. Oren, "Microclimate, water potential, transpiration, and bole dielectric constant of conifers and deciduous tree species in the continental boreal ecotone in central Alaska," Proc. of the 19941nt. Geosci. Rem. Sens. Symp., August 8-12, Pasadena, California, 1994.

[25] F.T. Ulaby, K Sarabandi, K McDonald, M. Whitt and M. C. Dobson, "Michigan Microwave Canopy Scattering Model," Int. J. Rem. Sens., vol. 11, no. 7, 1223-1253, 1990.

[26] Y. Oh, K Sarabandi and F.T. Ulaby, "An empirical model and an inversion technique for radar scattering from bare soil surfaces," IEEE Trans. Geosci. Rem. Sens., vol. 30, pp. 370-381, 1992.

[27] P.e. Dubois, J. van Zyl and T. Engman, "Measuring soil moisture with imaging radars," IEEE Trans. Geosci. Rem. Sens., vol. 33, no. 4, pp. 915-926, 1995.

[28] J. Shi, J.R. Wang, A. Hsu, P.E. O'Neill and E.T. Engman, "Estimation of bare soil moisture and surface roughness parameters using L-band SAR image data," IEEE Trans. Geosci. Rem. Sen., in review, 1996.

[29] J.R. Wang, A. Hsu, J.C. Shi, P.E. O'Neill and E.T. Engman, "Estimating surface soil moisture from SIR-C measurements over the Little Washita watershed," Rem. Sens. Env., in review, 1996.

This work was carried out at The University of Michigan and at the Jet Propulsion Laboratory, California Institute of Technology, under contract to the National Aeronautics and Space Administration. BOREAS is co-sponsored by the National Aeronautics and Space Administration and Energy, Mines and Resources, Canada.

- 52 -

- 53 -

TIME-DOMAIN REFLECTOMETRY TECHNIQUES FOR SOIL WATER CONTENT MEASUREMENTS

G. Clarke Topp Soil Physicist, Eastern Cereal and Oilseed Research Centre

Agriculture and Agri-Food Canada 960 Carling Avenue

Ottawa, Ontario, Canada, K1A OC6

The electrical dipolar nature of the water molecules results in their mutual interaction in the liquid state to give a high polarisability to water, which is evidenced as a high static (Le. low frequency) relative permittivity, ep or dielectric constant. The relative permittivity of water at 25°C is 78.5. Comparable values are 1 for air and a range of 3 to 5 for most commonly-found soil. solids. This large disparity between the relative permittivity of water and the other soil components has long been recognized and suggested that the measurement of the relative permittivity of soil may be useful for monitoring soil water contents. The currently available TOR instruments, having both portability and ease of operation, have enhanced the TOR applications in soil for the measurement of both water content and electrical conductivity.

In TOR, a fast rise step voltage pulse is propagated along a transmission line in the soil. The voltage pulse propagates as an EM wave, traveling in the soil and guided by the conductors, which may have a variety of configurations. The properties of the soil which govern the propagation of EM waves are contained within the propagation constant of the soil. For the early applications of TOR the propagation constant was divided into velocity, v, and attenuation, 0, which can be written as (Topp et aI., 1980; Topp et aI., 1988),

_ c _2L ct2 v - - - - from which E = (_) (1)

r;::- t ra 2L VEra

(2)

where c is the velocity of an EM wave in free space, era is the apparent relative permittivity measured by TOR, L is the length of the transmission line in the soil, t is travel time, W is the frequency, e"r is the imaginary component of the relative permittivity, 00 is the zero frequency (dc) electrical conductivity of the sample, eo is the permittivity of free space. Although it is possible to apply equations (1) and (2) sequentially to determine both water content and electrical conductivity, this paper focuses only on water content.

- 54 -

The high relative permittivity of water imparts to wet soil a very strong dependence of the relative permittivity on water content. Using coaxial transmission lines, Topp et al. (1980) obtained from laboratory measurements an empirical relationship between relative permittivity and volumetric water content, 8. The equation now widely used for a calibration curve is

e = -5.3xlO -2 + 2.92xlO -2€ra - 5.5xlO -4€~a + 4.3xlO -6e;a . (3)

In a surprising number of instances, a variety of researchers have confirmed that eq. (3) is quite broadly applicable.

The widespread use of TOR has resulted, however, in a number of applications in which eq. (3) cannot be used. An alternative to the use of empirically derived calibrations is to make use of dielectric mixing formulae. In the mixing law approaches, the soil is considered a mixture of three phases: water, soil and air, where the volume fraction of each phase and its shape and orientation are the major factors affecting the electrical response of the whole soil. From these analyses and related experimental work, it has been shown that the square root of the TOR-measured relative permittivity is essentially a linear function of the water content. The empirical equation (3) is very closely approximated by the mixing formulae over the usual water content range. From the assumption that the mixing law is analogous to a refractive index model, it has been shown that the slope of the relationship between relative permittivity and water content can be used to infer the apparent relative permittivity of the water within the soil. These results have shown that the water in soil has an apparent relative permittivity that is above that

of free water. This is interpreted to indicate that "bound" water does not have a major effect on the TOR-measured relative permittivity of most soils. The linearity of the square root of the relative permittivity with water content allows the use of two point calibration of TOR for water content.

TOR IN THE LANDSCAPE The portability of the cable testers and the flexible and varied possibilities for transmission lines or wave guides as soil probes permitted very rapid development and testing of in-field applications. The initial laboratory research made use of coaxial transmission lines varying in length from 0.1 to 1 m, with diameters in the range of 0.03 to 0.08 m (Topp et aI., 1980). Although the coaxial configuration was electrically well-defined, it was, however, unsatisfactory for field use. The balanced pair transmission line, consisting of two parallel rods, soon became the probe of choice for use in field measurements. Oalton

- 55 -

(1992)and Zegelin et al. (1989) discuss various criteria and related probe configurations.

Two additional probe configurations have appeared recently and both are being evaluated and used. Zegelin et al. (1989) introduced the multi-wire probe which emulates a coaxial transmission line. In simplest wire arrangement, the three parallel conductors form a plain along their length axis. The centre wire acts as the "centre" conductor and the outer two as the "shield" of the simulated coaxial line. These probes offer the advantage of improved impedance matching between the usual coaxial output of the TOR instruments and the line in the soil being measured. Hook et al. (1992) showed that much better signal-to-noise ratio was obtained by taking differences between open and shorted parallel pair transmission lines. They exploited this advantage in a soil probes which use PIN diodes for selectively shorting positions along parallel pair lines. This allows depth profiling of soil water content using a single probe, which is a desirable option.

Accepting that the soil immediately surrounding the probe can be disrupted by installation of the probe, one recognizes that the distance to which the electric field measures around the probe becomes a very important consideration for adequate interpretation of the TOR measurement with all types of probes. Experimental and theoretical analyses have

demonstrated that the distribution along the length of probes has an effect which is represented by a linear-weighted average. For the lateral distribution the situation is more complex as the weighting functions are generally not linear. Theoretically based weighting functions have been proposed for parallel pair and multi-wire "coaxial" probes inserted into a medium of nearly uniform permittivity. These expressions can be used to optimize the design of soil probes.

The capability now exists to use TOR effectively for monitoring hydrological water balance, measuring agricultural or forest water use efficiency, or monitoring changes in water content for irrigation scheduling. The effective use of TOR water content values for water balance monitoring depends on rapid, reproducible recovery of data from a number of representative locations. These requirements led to the development of automated analyses of the TOR trace. Associated with this automation was the development of multiplexing, or switching, capabilities which allow interrogation of as many locations as required using a single TOR instrument. Although a number of custom systems have been developed and used in recent years, most commercial instruments now offer automated analysis as a part of the basic instrument with the multiplexing capability as an option.

A number of field and landscape scale water balance studies using TOR have been carried out in recent years and demonstrate the efficacy of the current state of the art. Growing

- 56 -

season monitoring under corn was able to account for the amount of incoming rainfall to within 5%, even though the within row variation was much greater. In a forested plot monitored throughout the year, the changes in water content recorded by the TOR tended to overestimate the precipitation events by an average of 13%. In a wheat field, comparisons were made between TOR-measured changes in soil-water and those found from a weighing Iysimeter and the Bowen ratio technique. The trends in soil water changes were similar for these techniques giving an average deviation of less than 10%.

Another monitoring example from our laboratory reported the water content changes of soil immediately adjacent to growing corn roots. The associated calculations from the TOR response gave a reasonable measure of the diurnal root-to-soil exchange of water. Oiurnal fluctuations in water content were detected around the roots (Fig. 1) where the maxima occurred at night and were about 0.015 m3m-3 higher than minima for the days.

0.08 . ..-----..----........----....------.--.......----.----.-........--. PI~nt~

•• * •••••• ~........ .--- •• --+---.. -.--.~-.---.-.-.~--.. ---.-. .; ..... -.... -:- .. -.. --.--~-.. -...... ~---.... ---~----.. ----.}.---------.;..-----.-.. -~.-.. ------~--..... ---.;---.------

0.04·. . .

12 24 36 48

Date 31 Feb. 1

60 2

72 84 3

96 108 4

Figure 1: Recordings of TOR-measured soil water contents as average from four soil cylinders with growing corn plants and the average of the three control cylinders (without plants). The record is from day 3 to day 7 of the period the seedlings were growing in the cylinders. Variations in initial values reflect probable differences in soil packing. Upward pointing arrows indicate sunrise (07:00) and the downward pointing arrows indicate sun set (17:00).

There was an overall uptake of water from this dry soil layer, resulting in a decreasing water content even though water at a higher, more available, potential was present in the upper and lower portions of the cylinders. The quantity of water calculated as exuded from

- 57 -

the roots to give the observed TDR response was in the range of 4 to 6 III per cm of root per night. This amount of water compares favorably with 8 III per em estimated to stabilize the soil observed to be adhering to the roots.

CONCLUSIONS Over the past 20 years, considerable progress has been achieved toward successful application of TDR to the measurement and monitoring of soil water content. The demonstrated validity and efficacy of TDR in this application to soil has stimulated instrument developments aimed directly at soil measurements. TDR is becoming a standard procedure which will supplant gravimetric sampling as the primary reference method for water content measurement in soil. As water content is so important to our understanding of the physical and chemical behavior of soil, TDR will become an integral part of other measurement procedures. The future promises many exciting developments resulting from the experiences with TDR.

REFERENCES CITED F.N. Dalton, "Development of time-domain reflectometry for measuring soil water content and bulk soil electrical conductivity", Advances in Measurement of Soil Physical Properties: Bringing Theory into Practice, ed. by G.C. Topp, et al. Soil Sci. Soc. Am., Madison, WI, SSSA Spec. Publ. 30, pp. 143-167, 1992.

W.R. Hook, N.J. Livingston, Z.J. Sun, and P.B. Hook, "Remote Diode shorting improves measurement of soil water by time domain reflectometry" Soil Sci. Soc. Am. J., v. 56, pp. 1384-1391, 1992.

G.C. Topp, J.L. Davis, and A.P. Annan, "Electromagnetic determination of soil water content: measurements in coaxial transmission lines", Water Resour. Res., v. 16, pp. 574-582, 1980.

G.C. Topp, M. Yanuka, W.D. Zebchuk, and S.J. Zegelin, "The determination of electrical conductivity using TDR: Soil and water experiments in coaxial lines", Water Resour. Res., v.24, pp. 945-952,1988.

S.J. Zegelin, I. White, and D.R. Jenkins, "Improved field probes for soil water content and electrical conductivity measurement using time domain reflectometry", Water Resour. Res., v. 25, pp. 2367-2376,1989.

- 58 -

- 59 -

DIELECTRIC SOIL WATER CONTENT DETERMINATION USING

TIME DOMAIN REFLECTOMETRY (TDR)

Rudolf Plagge Christian H. Roth and Manfred Renger

Institute of Ecology, Dept. Soil Science, Technical University Berlin, Salzufer 11-12,

0-10587, Berlin, Gennany

SUMMARY

The applicability of different conversion functions for the dielectric detennination of soil water content using Time Domain Reflectometry (TOR) was evaluated. A database comprising 63 soils and soil-like materials of mineral and organic origin was established to derive conversion functions using stepwise multiple regression analysis. Parameters included easily to determine properties such as texture, organic matter content, bulk density and porosity. These conversion functions were then compared with other existing conversion functions, being models based on the theory of mixed dielectrics [10, 3], the model of [12] and the approach of [6].

Time Domain Reflectometry (TOR) is a non destructive measurement technique for dielectric water content detennination of soils. Because water (e "'" 80) has a higher dielectric number (e) as for instance mineral soil material (e <= 5), dolomite (e <= 6.8), gypsum (e <= 6) and air (e <= 1), it is possible to derive the volumetric water content (0) of a soil by applying a dielectric constant-moisture content conversion function (6(KaJ), where Ka is the apparent dielectric constant. It is known that the 6( KaJ relation can differ remarkably between different soils and a unique relationship between dielectric constant and volumetric moisture content of soils does not exist [2, 6, 7, 8, 9]. Particularly, when using empirical third degree polynomial regression calibration curves [7, 8,9, 12], the error can be considerable. To avoid this problem, other empirical approaches use easy measurable soil data, like texture, porosity or bulk density to correct the 6(KaJ relation [4, 6, 7, 8]. As an alternative to empirical attempts, it is possible to derive the moisture content of soils by applying semi-empirical mixing models [3, 10] or phenomenological mixed dielectric approaches [1, 10). However, these models require measurements of material parameters that are difficult to obtain. Thus, the needed parameters often become fitting parameters and the results then depend on the chosen data set used, a problem which is encountered for all empirically derived conversion functions. To quantify and to improve the 6(KaJ calibration, the conversion functions proposed by [3, 6, 10, 12], were tested on a wide range of soils and soil-like materials. This base data set was used to perform stepwise multiple regression analyses, taking into account easily measurable soil data.

MATERIALS TESTED

A total of 63 soils and artificial materials of mineral and organic origin were used as the base data set (n=81O). Table 1 shows the range of some typical characteristic parameters of the base data set. To derive the conversion functions the base data set was splitted into groups of mineral and organic origin (organic carbon content> 30%). Samples were dried and those with mineral components were sieved through a 2 mm sieve. The samples were moistened with demineralized water in 8 to 15 steps, to cover the range of moisture content from air dry up to saturation. A moistened sample was then repacked into a beaker as unifonn as possible, in small increments up to full volume. The relative dielectric constant was measured using a computer-aided TOR-device (CAM!, EASY TEST) with needles of 53mm length, O.5mm diameter and separated by 5mm. For coarse materials rods with lOOmm length, 2mm in diameter and separated by 16mm were used. More details about the setup are given in [5,7]. After taking readings of e, the sample volumetric moisture (8) and the bulk density (p) were detennined gravimetrically.

- 60 -

Table 1: Description and range of relevant characteristics of the studied mineral and organic soils and soil-like materials of the base data set.

Range of characteristic material properties

Base data set

Material Corg Sand Silt Clay Bulk density

0.063-2mm 0.063-0.(l02mm <0,002mm p

[%] [%] [%] [%] [g cm-3]

Mineral soil horizons 0-20.0 2-98 2-78 0-81 1-1.8

Organic soil horizons 32.7-54.8 0.1-0.7

Mineral materials 0-24.6 1-100 0-87 0-63 0.6-1.7

Organic materials 30.2-50.0 0.2-0.6

DERIVATION OF CONVERSION FUNCfIONS

Three levels of increasing information on soil predictor variables were assessed for developing the relation between 8(KaJ and the other soil properties for soils with mineral origin. For organic materials two levels of increasing information were used. Scatter plots and correlation analyses were performed to examine the functional relationship between regressors and possible predictors. Stepwise multiple regression analyses were run on the different information levels with an entry probability level of p = 0.15 and a probability level of p = 0.05 for retention of variables as predictors. For the mineral data set only the bulk density was considered in the first level, while in the second level the texture and the organic carbon content was taken into account. The third level considered p, texture and Corg as additional predictor variables. For the organic data set C org was only used in the first level and in the second level the organic carbon content and the porosity were considered as additional predictors. The multiple regression equations are given in table 2.

EV ALUA TION OF THE CONVERSION FUNCfIONS

Standard error of estimate (SE) was chosen as a parameter to evaluate the quality of the respective models. The results of the statistical analyses are given in table 3. Furthermore, a comparison with existing mixed dielectric models and other empirical conversion function was carried out. The semiempirical a-3-phase- and a-4-phase dielectric mixing models described in [10] and [3] were tested separately for the mineral and organic materials. The dielectric number of the solid phase was set to es=3.9 for mineral soils and to es=5 for organic materials. In the 3-phase model the empirical geometry factor a was fitted, while in the 4-phase model the fraction of the bound water (Obw) and the empirical geometry factor a were fitted simultaneously. The dielectric constant of the bound water was set to elJw=3.2 [2]. The 8(KaJ relationship determined by [12] was tested only on the mineral data set. The matrix-sensitive conversion function of [6] was used for materials of mineral as well as of organic origin.

For the mineral materials the quality of prediction of the volumetric water content was best with the multiple regression equations, presenting a SE of about 0.02. Precision increases with increasing number of predictor variables [8]. The matrix-sensitive approach of [6] yielded a standard error of estimate of 0.027. The semi-empirical a-3-phase and a-4-phase models delivered comparable results with a SE of about 0.033. For the 3-phase model the best fit was found for a mean et;::0.505, confirming the results of [10]. For the 4-phase model the best fit was found for et;::0.605, which is lower than found by [3] and can be explained by the range of the porosity of the used data set. The empirical polynomial conversion function of [12] leads to the biggest error. but on the other hand, no additional data are needed. For material of organic origin, all conversion functions lead to comparable results. with the SE rangeing between 0.032 and 0.040. Organic materials are characterised by a great degree of shrinking and swelling. so that porosity and bulk density can not be considered as constant. As prediction depends on the quality of the input parameters. a safe evaluation in this case is not possible.

- 61 -

Table 2: Multiple regression equations with the predictor variables bulk density (p, g/cm3), porosity (tP, cm3/cm3), total organic carbon content (Corg, %), clay (C, <2mm, %), silt (U, 2 - 63mm, %) and sand (S, 63 - 2000mm, %).

Conversion functions for mineral soils and mineral soil like materials

Level 1 Regression model 1, predictor variable p 8= -0.0914 - 0.067p + 0.148ve - 0.0354pve + 0.0214* p2ve- 0.00244p 2vil

Level2 Regression model 2, predictor variable S, U, C and Corg 8= -0.131 - OA08ve- 0.00124T - 0.000393S - 0.000235Corg + 0.00571 veT +

0.00437veU + 0.00619veS + 0.00268veCorg - 0.00011 vil T - 0.000054Vils + 0.0000043T2 ve + 0.OOOOO9U2 "Ie - 0.0000078S2 "Ie - 0.0000604Corg2 "Ie

Leve13 Regression model 3, predictor variable p, S, U, C and Corg 8= -0.00763 - 0.124p - OA34ve - 0.000937T + 0.0102Corg + 0.00507 veT +

0.00419veU + 0.00571 veS + 0.00222veCorg - 0.OOOO772ve2T - 0.0000305vils + 0.OOOOO52T2ve+ 0.OOOOO77U2ve - 0.OOOOO81S2ve+ 0.0619pve+ 0.00405ve2p - 0.00487p2Ve2 - 0.0169pCorg + 0.00841p2Corg-0.000392Corg2p

Conversion functions for organic soils and materials (>30% Corg)

Levell Regression model 4, predictor variable Corg 8= -0.0878 + 0.116ve- O.OOlCorg + 0.000092ve2Corg + 0.0000155Corg2ve-

0.OOOOO32Corg2ve2 Level2 Regression model 5, predictor variable tP and Corg

8= -0.191 + 0.398tP ve- 0.188ip2ve-0.0131tPve2 + 0.0139ip2Corg-0.000263ip2Corg2 - 0.00337veCorg + 0.000815ve2Corg + 0.000107veCorg2 -0.OOOOI9ve2Corg2 - 0.00321tP Corgve+ 0.OOOOO71ip2Corg2ve2

Table 3: Derived and used conversion functions, their required data, found functional parameters for the dielectric mixing models and standard error of estimate using the base data set for mineral and organic materials. Porosity (tP) was calculated from bulk density (p) and particle density (ps).

Data Parameter Standard error requirements fittet of estimate

Regression model 1 p 0.0238 ...... Regression model 2 T, U,S, Corg 0.0198 ~ ·c Regression model 3 p, T, U, S, Corg 0.0186 e~

(1)~ a-3-phase model tP(p,ps) a=0.505 0.0344 .S S

::;E:::: a-4-phase model tP (p, ps) a=0.605,8bw=0.03 0.0328 0 CI.l Malicki et al. (1994) p 0.0269

Topp et al. (1980) - 0.0468 Regression model 4 Corg 0.0396

CI.l (,) .- Regression model 5 Corg, tP(p, ps) 0.0317 .- ~

!a'5 a-3-phase model tP (p, ps) a=0.327 0.0378 e!>~ OS a-4-phase model tP (p, ps) a=0.445, 6bw=O.047 0.0354

Malicki et al. (1994) p 0.0362

CONCLUSIONS

The statistical analyses lead to the following conclusions: - The empirical polynomial conversion function of [12] shows the biggest error in predicting the

volumetric water content.

- 62 -

- The developed multiple regression equations deliver the most confident predictions of volumetric moisture. With increasing number of predictor variables the standard error of estimate is reduced. For materials of mineral origin the absolute standard error is about 0.02 cm3 cm-3, for organic materials approximately 0.035 cm3 cm-3•

- The matrix-sensitive conversion function of [6] can be used for both mineral and organic materials, showing an absolute standard error of estimate of about 0.027 cm3 cm-3 and 0.036 cm3 cm-3.

- The mixed dielectric models deliver an absolute standard error of estimate between 0.033 cm3 cm-3

and 0.038 cm3 cm-3 and yield comparable results for mineral and organic materials. The a-4-phase model performs slightly better.

- Precision of prediction also depends upon the accuracy of the additional input variables. Soils with swelling and shrinking properties do not allow for accurate predictions because of uncertainities in determination of bulk density and porosity. Furthermore, spatial variability may also affect the correct assignment of input data. In such cases, conversion functions employing input variables with the smallest error of determination should be chosen [8].

REFERENCES

[1] G.P. de Loor, "Dielectric properties of heterogeneous mixtures with polar constituent", Appl. Sci. Res., vol. 11, pp. 310-320, 1964.

[2] C. Dirksen and S. Dasberg, "Improved calibration of Time Domain Reflectometry soil water content measurements", Soil Sci. Soc Am. J., 57, pp. 660-667, 1993.

[3] M.C. Dobson, F.T. Ulaby, M.T. Hallikainen and M.A. EI-Rayes, "Microwave dielectric behaviour of wet soil. Part II: Dielectric mixing models", IEEE Trans. Geosci. Remote sensing, GE-23 (1), pp.35-46, 1985.

[4] O.H. Jacobson and P. Schjonning, "A laboratory calibration of time domain reflectometry for soil water measurement including effects of bulk density and texture", J. Hydrol. 151, pp. 147-157.

[5] M.A. Malicki, R Plagge, M. Renger and RT. Walczak, "Application of Time Domain Reflectometry (TDR) soil moisture microprobe for the determination of unsaturated soil water characteristics from undisturbed soil cores", Irrigation Sci., 13, pp. 65-72, 1992.

[6] M.A. Malicki, R Plagge and C.H. Roth, "Influence of matrix on TDR soil moisture readings and its elimination", Symposium on Time Domain Reflectometry in Environmental, Infrastructure and Mining Applications, Evanston (Chicago) USA, in Special Publication SP19-94,United States Department of Interior Bureau of Mines, pp. 105-114, 1994.

[7] R Plagge, C.H. Roth and M. Renger, "Kontinuierliche Messungen des Bodenwassergehaltes im Feld und Labor mittels der Time Domain Reflectometry (TDR)", DFG-Abschluj3bericht. Re 1433, pp. 1-133, 1994.

[8] R Plagge, C.H. Roth and M. Renger, "Bestimmung des Wassergehaltes von BOden mit Hilfe der Time Domain Reflectometry (TDR) " , Zeitschrift far Kulturtechnik und Landentwicklung, submitted,1995.

[9] C.H. Roth, M.A. Malicki and R Plagge, "Empirical evaluation of the relationship between soil dielectric constant-water content as the basis for calibration of soil moisture measurements by TDR", Journal of Soil Science, 43, pp. 1-13, 1992.

[10] K. Roth, H. Schulin, H. FIuhler and W. Attinger, "Calibration of Time Domain Reflectometry for water content measurement using a composite dielectric approach", Water Resour. Res., 26, pp. 2267-2273, 1990.

[U]W.R Tinga, "Mixture Laws and Microwave Materials Interactions". In PIER 6: Progress in Electromagnetics Research, A. Priou (ed.), Elsevier, New York, pp. 1-40, 1992.

[12]G.C. Topp. J.J. Davis and A.P. Annan. "Electromagnetic determination of soil water content: Measurements in coaxial transmission lines" Water Resour. Res .• 16, pp. 579-582, 1980.

- 63 -

IN SITU MEASUREMENT OF WATER CONTENT PROFILES IN GEOLOGICAL FORMATIONS / SUMMARY

B. Oswald! , H. R. Benedickter, Jeroen de Keijzer and W. Bachtold Laboratory for Electromagnetic Fields and Microwave Electronics,

Swiss Federal Institute of Technology, CH-8092 Zurich

H. Fliihler Soil Physics, Institute for Terrestrial Ecology,

Swiss Federal Institute of Technology, CH-8952 Schlieren

P. Marschall National Cooperation for the Disposal of Radioactive Waste, NAGRA,

CH-5430 Wettingen, Switzerland

Abstract A novel approach for the measurement of volumetric water content profiles B( T) in geological

formations is presented under special consideration of the encountered installation problems. We applied the TDR, Time Qomain .R.eflectometry, method, which is well known in transmission line theory, on a laboratory scale. The resulting dielectric constant <res of any mixture consisting of a solid phase, an aqueous phase and a gaseous phase strongly depends, amongst others, on the volume fraction of the aqueous phase. Due to the dependence of the characteristic impedance of a transmission line section on the resulting dielectric constant, <res, a layered geological structure can be modeled as cascaded sections of transmission lines. In our application, additionally severe field installation problems have been attacked by the construction of an inflatable depth TDR probe. As an example, in our investigation a shear zone with a width Az = 50 mm in the longitudinal direction of the TDR probe and having a volumetric water content of B = 0.15 has reliably been detected.

KEYWORDS: locally variable volumetric water content, geological formation, transmission line theory, time domain reflectometry, resulting dielectric constant, dielectric mixing law;

1 INTRODUCTION

1.1 MOTIVATION

In recent year there has been considerable interest in methods for the determination of the volumetric water content of rocks and soils as defined in the following equation 1

B(r) d~ VH2 0 (1) VREV

where VREF means a Representative .Elementary 'y'olume, REV, which is a characteristic quantity of the investigated geological medium, and VH20 the water volume contained in VREF , [1], [3], [8]. However, these methods do not consider spatial variations of the volumetric water content B(z) but assume a uniform distribution, i.e. B(z) = Bo. Few attempts have been made, in order to adapt them to spatially varying water content profiles B(z), e.g. [9]. Especially in geological formations the volumetric water content profile B(r), is the state variable which controls virtually all transport processes: Water, gas, heat and soluble matter transport depend sensitively on the spatial distribution of the liquid phase. For example in geological sites, that might be potential candidates for the deposition of nuclear waste, transport phenomena related to the specific geologic structure must be investigated carefully before any deposition be made in order to prevent potential leakage of radionuclides to the outside world inhabited by man. These radionuclides are mainly transported by the liquid phase, therefore the spatial distribution of the volumetric water content has to be monitored under different constellations, i.e. tunnel ventilation on or off etc. This is the point for depth time domain reflectometry. Due to severe installation problems (for example it is difficult to install conventional TDR probes in geological strata in a distance exceeding 60 cm below the surface) another approach must be chosen to avoid these installation problems.

I Corresponding author, can be contacted by email at:[email protected]

- 64 -

1.2 FORMULATION OF THE PROBLEM

The general problem of the one-dimensional spatial determination of the volumetric water content is formulated as follows: in a direction perpendicular to a layered geological structure there is a spatial variation of the volumetric water content B(z) as shown in figure 1.

Water Content: arbitrary profile

6 (distance)

Distance from Rock Surface

General Volumetric Water Content Profile

Granodiorite Fault Gauge Granodiorite

Matrix Matrix

6 = variable 6=0.0 6=0.0

Shear Zone Specific Three Section Model

Figure 1: GENERAL AND SPECIFIC MEASUREMENT SITUATIONS

However, in this paper the situation examined is a simple three section model, i.e. there is, in that order, a layer of dry granodiorite rock, a shear zone consisting of fault gauge material, [2], partially saturated with water and another layer of dry granodiorite rock. The quantity to be measured is the volumetric water content O(r) in the middle section which may vary considerably from site to site and at different times of measurements, according to figure 1

2 METHODS

2.1 WORKING PRINCIPLE

The method of choice for the in-situ measurement of the volumetric water content in rocks 0 and soils is the TDR methodology, see also [5], as already mentioned. If there is additionally a spatial variation of the volumetric water content in the longitudinal direction of a transmission line, then this causes a corresponding change in the resulting dielectric constant f re • of the medium surrounding the TDR probe:

0= O(z) --..... f re • = fre.(Z) (2)

By the usage of traditional TDR analysis methods, see [1], [4] [3] and [8], such variations are only considered in an integral manner and no local differences in water content can be recognized. However, if the TDR analysis methods used in cable testing applications, see [5], are remembered, local inhomogeneities in terms of the volumetric water content 0 can be localized by corresponding variations in f re • in the longitudinal direction of a TDR probe due to the fact that the characteristic impedance Zw of any type of transmission line strongly depends on the resulting dielectric constant of the medium surrounding the line, according to:

Z ( ) Zw(fre.(z) = 1.0)

w fre.(z) = Jfre.(z)

(3)

and hence can be visualized by the signal on the TDR display which accounts for local changes in the characteristic impedance of a transmission line system, i.e. Zw(z) is actually observed on the TDR screen.

In transmission line theory the coefficient of reflection pnm at the interface of two transmission lines with characteristic impedance ZW n and ZW m is defined as the ratio of the amplitude V; of the backward wave traveling wave, that is generated at each discontinuity of the characteristic impedance in the longitudinal direction of the transmission line, and the amplitude Vr of the forward traveling wave:

(4)

Therefore if the functional dependence of the characteristic impedance on the geometrical configuration of a transmission line system is known the distribution of the resulting dielectric constant f( z) in the

- 65 -

longitudinal direction of the TDR probe can be calculated. By the usage of a dielectric mixing law describing the relationship between the (volumetric) fractions of a system consisting of several phases, in our case a solid phase (= rock), a gaseous phase (= air) and a liquid phase (= water), the volumetric amount of water contained in a geological structure at a specific position on the TDR probe can be computed, see [6] for a detailed discussion of dielectric mixing laws, and [7] for an empirical justification. A typical approach for the determination of the resulting dielectric constant of a mixture can be found in [7]:

(5)

where Q' is a fit parameter ~ 0.5, Cre> is the resulting permittivity of the mixture, Cw the permittivity of the aqueous phase, C$ the permittivity of the solid (granodiorite) phase, Ca the permittivity of the gaseous (air) phase, 8 is the volumetric water content and 1J is the porosity, i.e. the empty space contained within the solid phase that can be filled with water and air respectively. If equation 5 is solved for the volumetric water content 0 then:

8 = c~e$ - (1 -1J)c~ - 1Jc~ c~ - cg

(6)

Therefore it is possible to compute the spatial distribution of the volumetric water content, 8(z). However, in this paper the situation being examined is restricted to three layers, {£1' £2, £3} which differ in terms of 8(z). The first and the third layer are assumed to be completely dry, i.e.

01 = 83 = 0.0 (7)

whereas the middle layer, £2 has a variable water content level 82 as shown in figure 1 which is to be determined.

2.2 FIELD INSTALLATION PROBLEMS

In general any type of transmission line can be used as probe for depth TDR applications if its electrical properties are sufficiently determined by the surrounding medium (which is treated as a dielectric in terms of transmission line theory), i.e. there must be enough contact area between the conductors and the medium.

In this application however there is one severe constraint: the only means of access to the shear zone is a single borehole perpendicular to the middle layer. Although it would be very advantageous to drill two small holes for the insertion of a two-wire transmission line, as e.g. the electric field vector must penetrate the rock under investigation, this type of installation is impractical and economically not feasible. It is, however, no problem to drill boreholes with diameters starting from 20 mm. Consequently, the whole TDR probe must be housed in a cylinder that does not exceed the diameter of the borehole and which has enough clearance to allow for smooth insertion due to the fact that the inside of such boreholes is likely to be very coarse. If therefore the probe has a considerably smaller diameter it would lie in the borehole as shown in figure 2.2 which introduces a considerable error into the determination of O(z) due to the air gap that is neither known nor equally distributed around the probe and into its longitudinal direction: In order to avoid that problem a new type of inflatable probe has been developed which

Granodiorite Matrix. .

PROBE ECCENTRIC BEARING IN BOREHOLE INSURING ELECTRICAL CONTACT IN BOREHOLE

Figure 2: DEPTH TDR PROBE INSTALLATION PROBLEM AND SOLUTION

consists of a body onto which the transmission line conductors have been mounted. After being inserted into the borehole, the probe body is inflated and consequently the flat band conductors are tightly pressed against the cylindric rock face of the borehole and therefore ensure good electrical contact, as shown in figure 2.2b. A second advantage of this approach is the possibility to attach an appropriate termination at the end of the depth TDR probe.

- 66 -

3 EXPERIMENTAL RESULTS

TDR traces that have been obtained from the experimental setup are shown in figure 3. When analyzed in terms of two connected sections of transmission line with different Zw's, then the values from table 3 result. In this context POI and Pl2 are the reflection coefficients between the 50[2 coaxial cable from the TDR and the dry rock section and the shear zone respectively.

p Shear Zone p Shear Zone

80 8p 90

Time in Nanoseconds Time in Nanoseconds

TDR TRACE WITH WATER CONTENT B = 0.00 TDR TRACE WITH WATER CONTENT () = 0.15

Figure 3: EXPERIMENTAL RESULTS

POI Zw of dry section Pl2 Zw of shear zone 0.42625 124.29 0.041 134.99 0.42625 124.29 -0.063 109.64

Table 1: CHARACTERISTIC IMPEDANCES AND REFLECTION COEFFICIENTS

4 CONCLUSIONS

A new approach for the determination spatially varying volumetric water content profiles, B(z), by the application of the TDR methodology has been presented. Experimental results on a laboratory scale show that it is possible to discriminate locally different values of () in a layered geological structure due to local variations in the characteristic impedance of the TDR probe. It will be the subject of further work to extend the three-section model used in this paper to arbitrary water content profiles B.

References [1] A. Alharthi and J. Lange. Soil water determination: Dielectric determination. Water Resources Research,

23(4):591-595, April 1987.

[2] P. Bossart and M. Mazurek. Structural Geology and Water Flow-Paths in the Migration Shear Zone. nagra / N ationale Genossenschaft fur die Lagerung Radioaktiver A bfiille / National Cooperation for the Disposal of Radioactive Wastes, Hardstrasse 73, CH-5430 Wettingen, Switzerland, December 1991.

[3] S. Dasberg and F. N. Dalton. Time domain reflectometry field measurements of soil water content and electrical conductivity. Soil Sci. Am. J., 49:293-297, 1985.

[4] S. Dasberg and J. W. Hopmans. Time domain reflectometry calibration for uniformly and nonuniformly wetted sand and clayed loam soils. Soil Sci. Am. J., 56(5):1341-1345, September-October 1992.

[5] Hewlett Packard. HP Application Note 62, TDR Fundamentals: For Use With HP 54120T Digitizing Oscilloscope and TDR, April 1988.

[6] Ebbe Nyfors and Pertti Vainikainen. Industrial Microwave Sensors. ARTECH HOUSE, INC., 685 Canton Street, Norwood, MA 02062, USA, 1989.

[7] Kurt Roth, Werner Schulin, Hannes Fluhler, and Werner Attinger. Calibration of time domain reflectometry for water content measurement using a composite dielectric approach. Water Resources Research, 26(10):2267-2273, October 1990.

[8] G. C. Topp, J. L. Davis, and A. P. Annan. Electromagnetic determination of soil water content: Measurement in coaxial transmission lines. Water Resources Research, 16(3):574-582, March 1980.

[9] M. Yanuka, G. C. Topp, Zegelin S., and W. D. Zebchuk. Multiple reflection and attenuation of time domain reflectometry pulses: Theoretical considerations for applications to soil and water. Water Resom'ces Research, 24(7):939-944, July 1988.

- 67 -

NEW TIME DOMAIN REFLECTOMETRY PROBES FOR WATER CONTENT DETERMINATION IN POROUS MEDIA

Markus Stacheder, Kurt Koehler, and Robin Fundinger IMKO Micromodultechnik GmbH, 1m Stoeck 2, 76275 Ettlingen, Germany

1. Introduction In the last decade the TDR method has gained wide acceptance in water content determinations for a variety

of porous materials. But the bigger the number of applications, the more obvious become the limitations of conventional TDR-systems and especially TDR-rod-probes. The disadvantages are for example the complex and heavy hardware of the cabletesters, the relatively small measurement volume of the probes or the influencing of the porous medium by the penetrating rods. Also the procedure when water content profiling is desired is rather fussy and the HF-cable length of the probes is limited. Nonetheless TDR measures water content in % by volume and not % by weight.

To avoid these disadvantages and to enhance the utility ofTDR, we have developed a new TDR-system and a variety of different TDR-sensors to provide adequate probes for different problems.

2. TDR-Systcm The TRIME-system (TDR with Intelligent Micromodul Elements) builds the basis for the development of

new sensor designs. It is a specially designed TDR-technique for the determination of water content in porous materials. The difference from conventional cabletesters is the way the reflected voltage pattern is sampled [1]. A special algorithm measures time of arrival of specific predefined voltage levels, the converse of conventional TDR cabletesters. These direct time measurements are easier to obtain and no complex and expensive electronic devices are necessary. With the help of ASICs (Application Specific Integrated Circuits) and microprocessors, it is possible to build small devices with low power consumption, so that the system is well suited for mobile and field use.

Fig. la shows the read out box and its dimensions. The display not only shows the measured water content, but also yields information on the attenuation of the voltage amplitude (from which bulk electrical conductivity can be deduced) as well as on battery capacity and status messages. The measured data can be sent via an RS 232 link to a data acquisition system or a (portable) PC.

T o ..... ..... 1

15 r- 115 ---I, I

35 8

c.)

a (0 .-

1 d.)

1---- 160 ---i-- 190 ----i

Fig. /; nVME read Ollt box and the dijJel'en/ rod-probes (dill/ensions inll1ll/).

Beside these advantages the new system also allows considerable flexibility in TDR-probe design.

3. TDR-probes 3.1 Rod-p robes

- 68 -

To date the TDR-probes are in the majority parallel sets of wires (two, three or four) that are inserted into the moist medium. We have also constructed parallel 2-rod and 3-rod-probes with standard lengths of 100 mm and 150 mm (Fig. Ib,c). The probe rods need a coating of PVC (0.5 mm thick) in order to enable the above mentioned trace evaluation algorithm. This coating is not only elementary for the functioning of the TRlME system (it acts as a high pass filter and raises the reflected trace), it also extends the application range for materials with higher bulk soil electrical conductivities. There is furthermore no need for impedance transformers (baluns) often causing unwanted reflections and signal losses.

The new algorithm also offers the ability to miniaturize the probe, a problem with conventional TOR because of a superimposition of the reflections on the displayed trace. With the direct time measurement this problem docs not occur and rod lengths down to 50 mm are possible (Fig. ld). Due to the small dimensions of the high frequency electronics and built-in microprocessor it is possible to integrate the electronics behind the probe rods (Fig. Ie). Therefore the probes are independent of HF cable length, a limiting factor of common probes and systems. They can be interlinked and located along a serial data network up to 3 km in length. These probes are well suited for in situ long-term monitoring of water content. In case of rod damage, the rods can be changed with screws. A multiplexing device connecting up to six 2-rod-probes has also been developed.

3.2 Tube-I)robe A TDR-tube probe has been developed to allow easy vertical water content profiling and to eliminate errors

potentially introduced to the water content profile by the method of differencing average water contents of adjacent continuous TOR rod pairs of different length. The new probe also overcomes the delicate handling problems associated with the use of radioactive measurement sources such as neutron and gamma ray probes. The sensor (Fig. 2) consists of a cylindrical PVC-body (0 44 mm) and two metall shells (J 75 mm x 17 mm) on opposite sides, which act as electrical transmission lines. The measurements are carried out in glassfibre access tubes, installed in the soil before the measurement (maximum length is 3 m). To ensure a close contact between

plates and inner tube wall, spring bearings are used to press them on the tube.

glassfibl'e access tube I "~-3 __

PVC body ------f

1

Fig.]: lRfME lube probe (dimensions il1l11l11).

An effective penetration depth of the measurement field of about 60 mm was determined, with the highest sensitivity in the immediate vicinity around the access tube and decreasing exponentially with distance from the tube surface. Therefore the crux of the whole measurement is the contact between soil and access tube. To guarantee a close contact and a minimum of soil compression, a drilling set and procedure described previously by Bell et a!. [2] has been changed and adapted. The access tube has a steel cutting shoe and is driven directly into the soil without preboring. Simultaneously the soil material is drilled out with a screw auger, while an inner steel tube protects the access tube. After installation the protection tube is removed and the bottom of the tube is scaled with a rubber bang.

Fig. 3 shows a comparison of the tube probe measurements with gravimetrically determined results and neutron probe measurements in a loessy field plot. A measurement accuracy of ± 2 % by vol. is possible.

0

0

·10

·20

·30

·40

E ·50 ..!2.. ·60 .r= i5.. ·70 w "0

·80

·90

-100

·110

·120

- 69 -

water content (% by voLI 10 20 30 40 50

., A··.·O

A.·O·

«J

to

00

.I!l .... ; ...

· .. ·OM .. ;

.~ .... ; ....... ·· .. ··ca· .........

···0 Ao·

O' ~ ....

II. neutron probe • TRIME probe o gravimetric

Fig 3: Comparison o/tube probe measurements to neutron probe and gravimetrically determined water contents in a loessy field site.

However problems may arise in inhomogeneous soils or when drilling under very dry conditions. Also measurements in swelling and shrinking soils can be problematic, since cracks develop especially along the access tube. A tube probe with 0.5 m long transmission lines was successfully used to measure the saturation of a granite core with a maximum porosity of 1 % by volume.

3.3 Sample ring probe In civil engineering water content is an important measure, though - unfortunately for TDR - the water

content must be detennined in % by weight according to most instructions. Therefore we ha\'e developed a sample ring probe (Fig. 4), consisting of two aluminum shells, which are mounted with spring bearings and are used for signal transmission.

sample ring probe

PVC· body glasstibre sample ring

auger set

,,/

steel cylinder for sample ring

Fig. -I: TIUlvfE sample ring probe to determine the water content in % by weight (dimensions in !!llll).

The sample ring (same material as the access tubes of the tube probe) with a known volume of soil is placed between the shells and the volumetric water content can be detennined. In combination \\ith a pair of scales, the gravimetric water content 8w (and the density) of the material can be detennined according to cq. (1)

(I)

with Ov as volumetric water content, v as volume of the sample ring, mw as mass of thc wet sample and Pw as density of water. The sample ring can either be filled by hand or an undisturbed sample can be cut out of the soil with an auger set. Fig. 5 shows a comparison of the sample ring probe with gravimctrically determined values.

- 70 -

~ 40 01 'iii

35 ~ >-.Q 30 ~ 25 C ~ 20 0 15 u ID 10 ~ UJ 5 ~ 0 a: I-

0 5 10 15 20 25 30 35 40

grav. water content [% by weight)

Fig. 5: Comparison of gravimetrically determined water contents and sample ring probe measurements for a clayey field soil.

3.4 Surface probe Some materials that do not allow a penetrating by probe-rods or tubes, as for example ancient brick walls,

require non-invasive TDR-sensors. Therefore we are also developing a TDR-surface probe (Fig. 6).

I:::U-::=---~.;;f- handle bar

PVC-body

>--- 75 ----, copper stripes Fig. 6: TRiME surface probe (dimensions in mm).

It consists of a simple PVC-body with handlebar and two coated copper stripes at the bottom of the body. The first measurements were overall promising, yielding a penetration depth of about 20 mm-30 mm, but requiring an absolute planar surface of the porous material.

4. Conclusion Though the presented system offers a lot of possibilities in designing new probes, the crux of all TDR

sensors is the contact between sensor and material. Especially the development of non-invasive TDR-probes demands an increase in penetration depth of the electromagnetic field. Therefore future work should be focussed on the determination and description of the measured volume and the interaction of the porous media and the TDR wave guides.

5. References III M. Stacheder, R. Fundinger, and K. Kohler, "A new TDR-system (TRlME) to measure soil moisture and

electrical conductivity", in Symposium and workshop on TDR in environmental. infrastructure. and mining applications. u.s. Bureau of./vlines Spec. Publ., SP 19-94, pp. 56-65, Sep. 1994.

[21 J.P. Bell; T.J. Dean; & M.G. Hodnett, " Soil moisture measurement by an improved capacitance technique, part II. Field techniques, evaluation and calibration.- J. Hydrol., vol. 93, pp. 79-90, 1987.

- 71 -

DISTINGUISHED PROBLEMS IN SOIL AQUAMETRY Alex Brandelik and ChristofHiibner

Institut fUr Meteorologie und Klimaforschung, F orschungszentrum Karlsruhe GmbH / Universitat Karlsruhe

Postfach 3640, D-76021 Karlsruhe, Germany

ABSTRACT

We suggest the dialog about the following distinguished problems.

a.) We need the concord on volume related moisture measurements. b.) The optimum measuring frequency for soil aquametry should lie higher than approx.

100 MHz and lower than approx. 1 GHz. c.) The often used mixing rule by Birchak [1] shows a discrepancy to electric field

calculations with finite elements. d.) Our measuring method allows the partition in free and bound water. Unmovable water

bulks can be detected as free water. e.) Our bound water mobility measurement can provide numeric indicators for soil erosion

and for contamination catchment. f.) We extended the time domain reflectometry for some ten meters in length and for

detecting moisture change on large areas.

INTRODUCTION

Despite of the great industrial, economic and scientific interest on water content in different solid and liquid mixtures the accuracy of their in-situ measurements is rather poor compared with other physical quantities like voltage, mass, current, force a.s.o. For instance, the determination of water in jet-kerosene tolerates an uncertainty below the range of 10-3. The same class of accuracy is needed in water content measurements of the fresh cement if we want to predict the mechanical strength of concrete constructions. The most important part of a meteorological energy balance calculation is the condensed or evaporated soil water. A realistic demand on this field tolerates an uncertainty of 10-2. We do not have the ideal aquametric solution; a general accepted measuring method with sufficient accuracy, in-situ and on line. It is not stand of today technology. So, we need to split the tasks in specific problems. Our specific field is the soil moisture determination, the dielectric soil aquametry. We pick up and prefer the name aquametry as it was recommended by Kraszewski [2] We modified it with the attribute dielectric instead of microwave because the measuring frequency depends on the specific task itself. Experiences in soil allow us to compress the above distinguished problems in one paper even if they were separately investigated in our laboratory and are appreciated for individual report.

RELATION to VOLUME

The relation base in aquametry depends on profession, as well. For instance, in the food industry, soil mechanics, civil engineering the mass (dry or wet) related data is used. In hydrology (due to the use of the continuity equations) the volume related water content is in handling. In the building material industry the level of the air pore saturation is the measure. The only common accepted direct measuring method is the gravimetric one. But, its volume related accuracy is not better than 3 % [3]. The dielectric aquametry (including capacity measurement, high frequency, radio frequency, microwave and TDR technique) has better accuracy despite of its indirect character. In these cases the measuring interaction is always a

- 72 -

volume phenomenon and delivers the volumetric water content. So, it is consequent to talk always in volume related sense, which we strongly recommend.

OPTIMUM MEASURING FREQUENCY

Every mixing rule uses the dielectric coefficient of water as parameter. The diel. coeff. of water is a well known function of frequency. It is nearly constant up to approx. 1 GHz. Its degradation over this frequency is known for pure water. If the water includes soluted ions, the degradation starts at lower frequencies and its change is unknown and unpredictable. This uncertainty burdens the calculated water content accuracy. Consequence, the measuring frequency has to be lower than the awaited degradation (below 1 GHz).In the aquametry, especially in the soil aquametry, the TDR method and devices are very frequent. Several manufacturers try to build devices with very short rise time to get an accurate time measurement. The drawback of these efforts is that the high frequency limit is neglected, consequently the result is not accurate enough, or a very sophisticated calibration has to be made for the special soil. On the other hand, the most mixing rules use the real part of the diel.coeff. Its measurement is accurate if the imaginary part of it, the loss factor, is small. A practical limit is an electric quality factor of 1. Due to several field measurements we find that for frequencies lower than 100 MHz the quality factor is less than 1. So, the frequency has to be higher than 100 MHz for natural soils. Airborne or space borne remote sensing systems mostly use frequencies higher than 1 GHz. (L, C and X bands). Only recently is the P-band introduced. Sophisticated frequency conversion of the ground truth data has to be carried out to compare in-situ and remotely sensed soil moisture data. This is one of our urgent tasks today.

MIXING RULE

Several empiric, semi-empiric and simplified theoretical mixing rules are recommended in different problem fields. One of the best reviews are collected by Pirou [4]. In soil physics is often recommended the mixing rule by Birchak. We also find it as the best one for soil moisture sensing. We already reported [5], that a modified model by Birchak delivers very good correlations between measurements and model. However, the physical explanation by Birchak is not sufficient good. He closed out the real inhomogenity in the direction parallel to the electric field. Now, we were challenged to compare the Birchak's model with numeric electric field calculations with finite elements. A 3-dimensional rectangular mesh was laid over a cubic plate capacitor and a random generator assigned diel. coeff. for each elementary cell. Then the Maxwell's equations were solved with the program by Weiland [6] and lead to effective diel. coefficients of the mixtures between water content 0 % and 40 % at 60 % solid material. Convergence of results was tested with different mesh sizes. The comparison to Birchak's model shows differences up to 5 %! Further investigations are required! Our next steps are careful laboratory measurements of soil probes and the refinement of the electrical simulation with clustering algorithms.

BOUND WATER AND IT'S MOBILITY

Water exists not only as free water in mixtures. A part of the whole water content is more or less bound on the surface of other materials. What is bound water? The present situation is rather chaotic in this nomenclature. Several forces can bind water e.g. Coulomb-, Van der Waals-, capillary-force a.s.o. It is complicated to describe the transition and decay of these

- 73 -

forces. We need a more simplified model for it. We prefer the dielectric aquametry, consequently, the only reasonable new definition for bound water must be related to the possible measurement, namely to the change in its dielectric coefficient. We call water "bound water" if the real part of its complex dielectric coeff. is less than that of free water at the same frequency and temperature [5]. Binding of a water molecule is strongest in an ice crystal. The diel. coeff. of ice Ei=3.2 compared to Ewf=80 in case of free water The first monolayer on a solid surface is as strongly bound as in ice. The binding force diminishes successively, in our model exponentially, with growing distance from the solid surface or growing water content W. At a sufficiently far distance from the solid the water is free. The continuous function of the dielectric coefficient of water Ew, which replaces Ewf in mixing rules, will be

w

Ew = Ewf -(Ewf -Ei)·e-h (1)

The slope h of the function is characteristic of the surface of the soil particle and of the sum of its binding forces. Our measurement method enables the determination of the slope hand so the partition in bound and free water. Some closed water bulks can be measured as free even if they are unmovable at 105 oe. If after a drying and rewetting procedure the characteristic slope h of the bound water has changed, the bound water was mobile. The bound water mobility in soils can provide a numeric indicator for wind erosion threatening due to the philosophy: The internal soil structure is destroyed if the bound water is attacked. If the lost bound water is not replenished (only free water has risen as a consequence of a rewetting), the soil deterioration is not reversible, the soil has lost its former matrix structure, it has lost its quality. Similarly, remedy selection of contaminated soil needs detailed information about water mobility onsite. The bound water mobility can provide the numeric indicator for the catchment capability. If the bound water content can change, the kept poison or nutrient is mobile. The degree of bound water change gives measures for mobility classification [7].

EXTENDED TDR TECHNIQUE

A new sensor was introduced in our institute for the moisture measurement of soil or snow. It consists of a simple flat band cable. The diameter of the electromagnetic field extension is about 10 cm around the cable. Impulse propagation time and characteristic impedance of the cable are affected by the dielectric properties of the surrounding medium (soil or snow). Furthermore, if a line is pulsed, it can induce voltages in an other line, which is crossing in a distance of less than 20 cm. This way we can lay a mesh of wires in or on the soil surface respectively the snow cover, where a cable is crossed by several others. Then, the distribution of the dielectric coeff. in the field can be measured with a reasonable resolution. The calibration from diel. coeff. into volumetric water content can be made e.g. with our former cryo-soil moisture sensor which measures absolutely. Fig.l shows the measurement with two flat band cable sensors on the same field. The first one with a length of about 18 m is buried in depth of 30 cm. The second cable is 20 cm deep and 30 m long. The relative changes of the pulse traveling time are well correlated with rainfall events.

- 74 -

600 45

40 500

.. CIl C ... loa rI' - 35

.5 - -- loa - I dl 400

E E 30 E - .. Cable 1 '':: .5 c 25 Cable 2 .2 c

300 0 «l .~ 01) 20 0 Net radiation <':I ..... c.. '0..

Precipitation 0 200 15 '0 ....

c.. dl ... dl Q.,

.!!l 10 ::J Q., 100

5

0 0 V') V') V') V') V') V') V') V') V') V') V') V') V') V') V') V') et- et- et- et- et- et- et- et- et- et- et- et- et- et- et- et-et- et- et- et- et- et- et- et- et- et- et- et- et- et- et- et-

c-..: c-..: c-..: 00 00 00 00 00 00 0\ 0\ 0\ 0\ 0\ 0\ 0 c: c: c: c: c: c: c: c: 0 c: 0 0 0 0 0 \0 V') 0 V') 0 V') 0 '<t 0\ ..,f 0\ ..,f 0\ ..,f

N N M 0 N N M 0 0 N N 0

Fig.l Pulse propagation time versus rainfall events

REFERENCES

[1] Birchak, J. R., Gardner, C.G., Hipp, J.E. and Victor, J.M. (1974). High Dielectric Constant Microwave Probes for Sensing Soil Moisture. Proc. IEEE, Vol. 62, No.1, pp. 93-98.

[2] Kraszewski, A.W. (1991). Microwave Aquametry- Needs and Perspectives. IEEE Trans. on Microwave Theory and Techniques. Vol. 39, No.5, pp. 828-835.

[3] Halbertsma,1. and van den Elsen (1991). Calibration and Accuracy Analysis of the Water Content Measurement of the CAMI TDR Unit. Int. Rep. 146, The Winand Staring Centre, Wageningen, The Netherlands.

[4] Priou, A (1992). PIER 6 Progress in Electromagnetics Research, Elsevier, New York, Amsterdam, London.

[5] Brandelik, A and Krafft, G. (1994). Ground Truth for Soil Moisture Sensing. 7th Australasian Remote Sensing Conference. 1.-4. March, 1994, Melbourne, Australia.

[6] Weiland, T. (1985). On the Unique Numerical Solution of Maxwellian Eigenvalue Problems in Three Dimensions. Particle Accelerators, Vol. 17, pp.227-242.

[7] Brandelik, A, HUbner, C., Angler, O. and Ruppert, P. (1995). On-site Free and Bound Water Determination of Contaminated Soil. Proc. of the Fifth Int. FZKlTNO Conference on Contaminated Soil. 30. October - 3. November, 1995, Maastricht, The Netherlands.

- 75 -ON-LINE INDUSTRIAL APPLICATIONS OF

MICROWA VB MOISTURE SENSORS

Ray J. King KDC Technology Crop.

Introduction: Economic benefits from real-time, on-line determination of bulk moisture content (mc) in agricultural commodities, processed foods, manufactured wood products, minerals, polymers and many other products are exceptional. Such information is useful for determining the value of raw materials, for front-end processing, for in-process control and for output quality control. Microwave methods have found applications for monitoring the mc of solids, powders and particulates, and liquids, and can do so with accuracies and resolution that are only limited by the accuracies of the standards used to calibrate the measurement system. Besides moisture, microwave sensors are also useful for simultaneously determining material density or basis weight from which mass flow can be monitored, for identifying contaminants, for monitoring chemical reactions and ionic content, and for nondestructive testing of manufactured wood and nonconducting polymeric products.

In keeping within the scope of this workshop, this paper exclusively addresses the continuous determination of moisture content and density or basis weight of materials in an on-line industrial setting. Moreover, it deals exclusively with two-parameter methods, meaning that two independent microwave parameters are measured simultaneously, e.g., attenuation (M) and phase (~<p) changes in the case of transmission sensors, or resonant frequency (fr) and normalized input resistance (ro) in the case of resonant reflection sensors. From these two measured microwave parameters, two independent physical properties of the test material can be simultaneously determined. For example, in the case of through transmission sensors, the partial water and dry material basis weights (mw and md; mass per unit area) are determined from which mCa = mw/ffid (absolute or dry basis) or mCr = mw/mtot (relative or wet basis) where mtot (= mw + ffid) is the total bulk basis weight. In the case of contacting resonant sensors, the partial water and dry material densities (pw and Pd) are determined and mCa = Pw/Pd (absolute or dry basis) or mCr = Pw/Ptot (relative or wet basis) where Ptot (= pw + Pd) is the total bulk density.

Determination of the partial and total densities or basis weights from which the mc is computed is termed the density or basis weight compensated method. It is essential that density or basis weight of the dry material be accounted for if the mc is to be determined with good accuracy, especially at low moisture content. This is because the microwave fields depend on the dielectric (E') and loss factors (EIt) of the combined water and the dry material, even though the dielectric and loss factors of water are generally an order of magnitude larger than those of the dry material. Consequently, both of measured microwave parameters (M, ~<p) or (fn ro ) depend on the partial m's or p's of the water and the dry material. These partial p's or m's depend on several factors; settling during production and handling, swelling/shrinkage due to moisture and temperature, variations in the amount and size of particulates, e.g., crumbling, and cultivars in the case of grain commodities, etc. For these reasons, any electromagnetic method should be sufficiently robust to properly account for such vagaries in density.

Motivations for On-Line Moisture Monitoring: Producers and handlers of agricultural and processed food products are aware of the immense economic impact that moisture has on the production, transportation, storage, processing, texture/taste and distribution of their products. Motivations for online continuous moisture/density monitoring also include:

-determining the value of raw materials -real-time information for feedback/feedforward control of drying and manufacturing -energy savings resulting from tightened tolerance ranges and avoiding overdrying. Overdrying can

reduce the nutritive value of foods, and underdrying affects storage and shelf life.

- 76 -

-elimination of expensive and time-consuming off-line (manual) sample testing; reallocation of assigned personnel

·improved, consistent product quality -elimination of out-of-specification waste and/or the need for product recycling ·sorting and grading of a material according to its moisture ·establishing and maintaining a competitive advantage -maintaining a log of product properties, particularly moisture contact

Advantages and Benefits of Microwave Methods:

-Bulk moisture/density (or basis weight) are determined simultaneously. In contrast, IR methods only sense surface moisture.

-Micro-gradients in moisture have little effect on moisture determination, whereas low frequency capacitance and resistance moisture meters are sensitive to such gradients.

-Logistical flexibility--microwave sensors can be installed on conveyors, or in hoppers, shakers, pipes, chutes, etc.

-Rapid response--generally milliseconds to one second provides ample data to perform short-term averaging.

-Insensitive to dissolved salts, due to low mobility of ions at microwave frequencies. -Environmentally safe and free of government regulation. -Immunity to environmental conditions such as dust, humidity, vibrations (with time averaging),

and vapors.

Noncontact Transmission Sensors: The principles of transmission sensors are based on simultaneous measurement of changes in the attenuation (M in dB) and phase delay C~<p in deg.) experienced by a microwave beam having propagated through a layer of test material. Both of these measurable microwave parameters are linearly related to the partial basis weights. An emperical Basis Weight Compensated Model expressing these relationships can be constructed [1, 2].

M - Ao = alIIld + a2mw

= IIld [al + a2 mcJ

CIa)

(lb)

and ~ = a3IIld + 84mw

= IIld [a3 + 84mcJ

(2a)

(2b)

where al-4 are calibration constants, and Ao is chosen to minimize the variance of mc. when correlated with mCa by weighing. In (l) and (2), note that the material thickness is implicit, i.e., IIld, mw, M and ~

a2a3 - at a4 and

a4(~A - Ao) - a2~

a l a4 - a2a3 (3)

The dielectric properties (E', E") of most materials are temperature dependent, primarily due to the temperature dependence of the contained water. Fortunately, this dependence is linear and easily accounted for by letting at become al + b1 ~T, etc., where ~T = T - To is the difference between the test and the calibration (To) temperatures.

Bistatic (two horn) noncontact through transmission systems are well known. A recent innovative variation is the monostatic noncontact through transmission system [2] shown in Fig. 1. Here, a single transmit/receive horn antenna is used to receive the amplitude modulated signal that is backscattered from a modulated reflector (MR). The modulated backscattered signal is coherently detected after having passed through the material twice. Consequently, for the same frequency, the measured M and ~<p are twice as large as for the bistatic method. Thus, assuming that M and ~<p are approximately proportional to frequency, the operating frequency of the monostatic system can be about half of that for the bistatic system for the same M and ~<p.

The m's as detennined using (3), and hence the mc correspond to the integrated effects through the entire thickness of the test material. Then it is straightforward to detennine the corresponding rate of mass transport (dMJdt = mVII\i) passing by the sensor, where v is the velocity of the moving material and w is the effective width of the transporter.

- 77 -

TransmiV Receive

" Transmitted 'ff Reflected Signal '" ,Signal

Modulated Reflector

KDC Technology

Corp.

MMA-2000

Fig. 1. Monostatic two-way transmission system.

Remote PC

Fig. 2(a) illustrates the linear dependence of M and ~cp with mCa for soybeans, according to (1 b) and (2b). Note that such linear behavior only obtains for mc. (dry basis); if plotted vs. mCr (wet basis), the curves would show a characteristic parabolic shape. Fig. 2(b) correlates the corresponding mc. as detennined by the basis weight compensated model with that measured by weighing as the test sample was incrementally dried. The standard deviation from the ideal 45" line is 0' = 0.67%, which includes errors in microwave determination of mc., as well as weighing measurement errors. When testing moving materials, the random microwave errors would be essentially eliminated by computation of running averages over time windows that are short compared to variations in material properties.

30 800

........ -Ideal 25 ~50 ......

600

-.20 I>- ::::-40 to (1)

>9 " " 0 -15 4000: E30 <t (I) >- cr = 0.67% <1

10 cc - e20

200 (l' 5 (a) E 10 (b)

0 0 0 0 1 0 20 30 40 50 60 0 10 20 30 40 50 60

mCa (by weighing) [%] mCa (by weighing) [%]

Fig. 2 (a) Linear dependence of M and ~<p vs. mc. for soybeans. (b) Correlation of mCa as determined by the basis weight compensated model with that measured by weighing.

f = 4.9 GHz, To = 1S'C.

Contacting Open Reflection Resonator Sensors: Open reflection resonator sensors [3, 4] are the newest addition to the family of microwave sensors for industrial applications, particularly for on-line, continuous monitoring of mc. Open coaxial resonators are the most obvious physical realization of this type of sensor, although microstrip and strip line realizations offer the additional advantage of low profile flush mounting, simplicity and low cost. The two measured microwave parameters are the resonant frequency (fr) which decreases linearly with e' and the normalized input resistance (ro, or SWR) which decreases inversely with e" of the test material.

- 78 -

Development of the density compensated model parallels that for the basis weight compensated model, under the hypothesis that e' and e" are linearly related to the partial densities Pd and Pw, i.e., .

(4a) (4b)

where al-4 are calibration constants that are determined from a least squares fit to the ensembles of (independent) {fr-Z,(fx!o)-l} data and (dependent) {Pd, Pw} data. The constant A is chosen to minimize the variance of the least squares fit. Solving (4a) and (4b) simultaneously,

1015

1010

..... 1005 N

(5a)

1.8 (') 0 c:::

1.6 "C

::::J 1.4 cc

Pw

..... ;:;.g 0 '-'

-(I)

>

a3(fr-Z - A) - al (frfo)-l a2a3 - ala4

25

20

(5b)

,...., J: ::1000 '-'

1.2 ~

co ~ .=: 15

0.65«e, () ...

- 995 (")

1.0 0' cJ" 990 .:r E

10 985

........ I....I.-......... ...L-........ ..I-L ...... ~ ............... ~O. 55 10 15 20 10 15 20 25

m c;. [%] m cr (by weighing) [%]

Fig. 3 Testing of wheat using a micros trip open reflection resonator sensor. (a) fr and ro decrease linearly and inversely, respectively, with mc •.

3", .......

(b) Comparisons of Pd and mCr as determined by the density compensated model and by weight and volume measurements.

Note that while {Pd, Pw} are assumed linear with {e', e"} in this model, neither of these sets are necessarily linear with mCa or mc;.. In particular, Pd may vary with mc due to material shrinkage (or swelling).

To illustrate the performance of a typical 1 GHz microstrip resonator sensor Fig. 3(a) plots fr and ro for wheat, and Fig. 3(b) compares Pd and mCa as determined using (5) with corresponding values as determined by weight and volume measurements.

[1] A. Kraszewski, "Microwave Aquametry--Needs and Perspectives," IEEE Trans. on Microwave Theory and Techniques, Vol. 39, No.5, pp. 828-835, 1991. [2] R. J. King, "Microwave Sensors for Process Control. Part I; Transmission Sensors," SENSORS, Vol. 9, No.9, pp. 68-74, 1992. [3] R. J. King, "Microwave Sensors for Process Control. Part II; Open Resonator Sensors," SENSORS, Vol. 9, No. 10, pp. 25-30, 1992. . [4] R. J. King, J. C. Basuel, M. J. Werner and K. V. King, "Material Characterization Using Microwave Open Reflection Resonator Sensors," Electromagnetic Wave Interaction with Water and Moist Substances, A. Kraszewski, Ed., Proc. 1993 IEEE MTT-S International Microwave Symposium Workshop, 1996. .

- 79 -

MEASUREMENTS OF LIQUIDS, SEMISOLIDS, AND POWDERS: AN OVERVIEW OF

PRINCIPLES AND TECHNIQUES *

James Baker-Jarvis t

ABSTRACT

Introduction

There is a real need for accurate dielectric measurements of liquids, powders, and semisolids from d.c. to microwave frequencies. Applications include the biotechnology, food, chemical, and petroleum industries [1]. In this lecture we overview liquid measurement techniques with an emphasis on lossy liquids. Liquid measurements are complicated by the high permittivities and losses encountered, particularly at low frequencies. The relative permittivity can range from around 2-3 for oils to well over 1000 for colloidal solutions. Loss can be low, for example, for heptane or very high, for example, in saline solutions. Temperature changes as small as 10 C can strongly influence measurements.

In the next sections I will briefly overview the different measurement fixtures used for lossy liquid measurements. Dielectric measurement procedures will be discussed in the frequency range of 100 Hz to 20 GHz. Results will be presented on liquids and powders of varying permittivity and loss characteristics. We will highlight three measurement methods. These are the shieldedopen circuit holder, the coaxial probe, and the capacitor technique. Shielded open-circuits are useful from 1 MHz to 10 GHz, coaxial probes can be used for measurements from 50 MHz to 20 GHz, capacitor devices are useful for accurate liquid measurement for frequencies up to approximately 1 MHz. The problem of low-frequency electrode polarization will be discussed and correction algorithms will be outlined. Uncertainties of the various measurement techniques will be summarized and a list of candidate reference liquids will be presented.

Open-circuit Holders

The shielded open-circuited coaxial line sample holder (OCL) has been used for years for dielectric measurements of liquids and powders in the microwave band [2, 3, 4]. The sample

·Contribution of the National Institute of Standards and Technology and not subject to copyright in the United States.

tElectromagnetic Fields Division, National Institute of Standards and Technology, MS 813.08, Boulder, CO 80303-3328

- 80 -

0 Outer Conductor 0 I II

Port 1 I z=o

0 Outer Conductor 0

Figure 1: A shielded open circuit termination.

8

4

2~~~1--~~1--~~~--~~~ o 240 480 720 960 1200

Frequency (MHz) Figure 2: The real part of the permittivity of carbonaceous sand mixture as a function of frequency and water volume fraction (maximum rms uncertainty ~f~(3) ± 0.35).

holder operation is based on an accurate model of a coaxial line terminated in a shielded open circuit. The advantage of an open- circuited holder is the ease of sample installation, the broad frequency capability, and the strong electric field in the sample region. The OCL is composed of an outer conductor that extends beyond the end of the inner conductor as shown in figure l. The model we use in our laboratory is based on a full-wave solution. The full-mode model is much more general than the more common capacitance expansion [2].

Von Hippel [5] used an open-circuited sample holder for liquid measurements. Bussey extended the OCL technique [2]. Scott [6] studied the instabilities encountered in solving the relevant nonlinear open-circuit equations. Hill [7] studied in situ measurements of soils using open-circuited transmission lines. Jesch [8] used the shi~lded open-circuited holder for measurements on shale oil. Biological tissues have been measured using the shielded open-circuited line, for example by Stuchly and Stuchly [9]. The sample holder has been found useful in hightemperature measurements. We will overview the theory behind the shielded-open circuit holder, present measurement results, and uncertainties.

- 81 -

p

zr

(1)

tRe ground plane

z=O .. ________________ ~~ilR~g~tR~g~ru __ r~g~a~p __ ~(~2~) ________________ z=d

~RstRs sample (3) __________ -...,-_.,.-----;---,,---,.._---,--_____ z=L + d termination (4) ti'ui, ~i'ui

Figure 3: The open-ended coaxial probe over a sample with an air gap between sample and probe.

Open-ended Coaxial Probe

Open-ended coaxial probes are commonly used as nondestructive testing tools and for liquid measurements over frequencies of 100 MHz to 20 GHz. In most applications the coaxial probe is pressed against a sample, and the reflection coefficient is measured and used to determine the permittivity of the sample. Over the years, the open-ended coaxial probe has been studied extensively both theoretically and experimentally [10, 11, 12]. The method, although nondestructive, does have limitations. For example, the fields at the probe end contain both Ez and Ep components. If there is an air gap between sample and probe, the discontinuity in the normal electric field causes a large error in the predicted permittivity. For this reason the probe has been used primarily for liquid and semisolid measurements, where good contact can be obtained. In process control, such as monitoring water content for rolling stock on assembly lines, a non contacting probe may be required. For this reason it is important to have a coaxial probe model which includes lift-off [11].

Capacitive Techniques

Capacitance techniques are useful at frequencies extending from 1 Hz to 1 MHz [13]. The air gap problems encountered in solid transmission-line measurements is not encountered in liquid measurements. In these techniques the electric fields are nearly normal to the sample plane. The difficulty with capacitor measurements resides in minimizing fringing field effects. The fringe field is usually partially eliminated by measuring the capacitance with and without sample and using guards to protect from fringing fields. Measurements on conducting liquids are complicated by electrode polarization effects [14]. Electrode polarization is caused by the build up of conducting ions on the capacitor plates producing a double-layer electric field. Electrode polarization influences primarily the real part of the permittivity. Since this capacitance is not a property of the material under test is must be removed ~rom the measurement to obtain the correct permittivity.

References

[1] S. O. Nelson and A. W. Kraszewski, "Grain moisture determination by microwave measurements,"

- 82 -

ASAE, vol. 33, p. 1303, 1990.

[2] H. E. Bussey, "Dielectric measurements in a shielded open circuit coaxial line," IEEE Trans. Instrum. Meas., vol. IM-29, pp. 120-124, June 1980.

[3] J. Baker-Jarvis, M. D. Janezic, and R. Stafford, "Analysis of an open-circuited sample holder for dielectric and magnetic measurements of liquids and powders," Nat!. Inst. Stands. Tech. Technical Note 5001, National Institute of Standards and Technology, 1992.

[4] S. Jenkins, T. E. Hodgetts, R. N. Clarke, and A. W. Preece, "Dielectric measurements on reference liquids using automatic network analyzers and calculable geometries," Meas. Sci. Technol., vol. 1, pp. 691-702, 1990.

[5] A. R. V. Hippel, Dielectric Materials and Applications. Cambridge, MA: M.LT. Press, 1954.

[6] W. R. Scott and G. S. Smith, "Error analysis for dielectric spectroscopy using shielded open-circuited coaxial lines of general length," IEEE Trans. Instrum. Meas., vol. IM-35, pp. 130-137, June 1986.

[7] P. N. Hill and H. E. Green, "In situ measurement of soil permittivity and permeability," Journal of Electrical Electronics Engineering, Australia, vol. 2, pp. 205-209, December 1982.

[8] R. 1. Jesch, "Dielectric measurements of oil shale as functions of temperature and frequency," IEEE Trans. Geosci. Remote Sensing, vol. GE-22, pp. 99-105, March 1984.

[9] M. A. Stuchly and S. S. Stuchly, "Coaxial line reflection methods for measuring dielectric properties of biological substances at radio and microwave frequencies-A review," IEEE Trans. Instrum. Meas., vol. IM-29, pp. 176-183, 1980.

[10] 1. 1. Li, N. H. Ismail, L. S. Taylor, and C. C. Davis, "Flanged coaxial microwave probes for measuring thin moisture layers," IEEE Trans. Biomedical Eng., vol. 39, pp. 49-57, January 1992.

[11] J. Baker-Jarvis, M. D. Janezic, P. D. Domich, and R. G. Geyer, "Analysis of an open-ended coaxial probe with lift-off for nondestructive testing," IEEE Trans. Instrum. Meas., pp. 711-718, October 1994.

[12] S. Jenkins, T. E. Hodgetts, G. T. Symm, A. G. P. Warhamm, and R. N. Clarke, "Comparison of three numerical treatments for the open-ended coaxial line sensor," Elect. Lett., vol. 24, pp. 234-235, 1992.

[13] M. G. Broadhurst and A. J. Bur, "Two- terminal dielectric measurements up to 6 X 108 Hz," J. Res. Bur. Stands. Tech., vol. 69C, no. 3, pp. 165-172, 1965.

[14] H. P. Schwan, "Linear and nonlinear electrode polarization and biological materials," Annuals of Biomedical Engineering, vol. 20, pp. 269-288, 1992.

[15] J. A. Schellman and D. Stigter, "The double layer, zeta potential, and electrophoretic charge of double-stranded DNA," Biopolymers, vol. 16, pp. 1415-1434, 1977.

- 83 -

COPLANAR SENSORS FOR MOISTURE MEASUREMENTS

S.S. Stuchly and C.E. Bassey

Dept. of Electrical and Computer Engineering, University of Victoria P.O. Box 3055, Victoria, B.C. V8W 3P6, Canada

Abstract. A concept of using coplanar sensors for dielectric measurements at microwaves is presented. The basic theory, advantages and limitations of planar sensors are outlined. The use of this technique for measuring moisture content is considered.

1. Introduction

The propagation of electromagnetic waves in dielectrics is accompanied by absorption of energy and variation in the phase of the waves which depend on the permittivity of the medium. The permittivity may in turn depend on other physical properties of the medium such as moisture content, density and composition. It also varies with frequency and temperature. Thus, if the relationship between the permittivity and a physical property of a material is known, such property can be determined by studying the propagation of waves in the medium using an appropriate technique.

An appropriate sensor is an integral part of instrumentation required for such measurements. The use of coplanar sensors is proposed in view of their advantages over other sensors. While discrete sensors provide measurements in one location only, distributed sensors respond to the average values along the sensor. This is important in practical applications such as moisture measurements. Coplanar sensors can be placed conveniently in contact with the test material and therefore provide nondestructive measurements. Although coplanar sensors produce fringing field, that is nonuniform field distribution, they also guide the waves travelling along the sensor.

The objective of this research is to develop a theory that may facilitate the development of techniques based on coplanar sensors for material measurements. The propagation velocity of the wave is approximately inversely dependent on square root of the dielectric constant of the test material.

2. Basic Principles

The propagation constant 'Y of an TEM electromagnetic wave in a dielectric is given by

'Y = 0: + jf3 = j 2; ({/- j{II)1/2 (1)

where 0: is the attenuation constant, f3 is the phase constant, A is the free space wavelength, {I is the dielectric constant and (" is the loss factor of the material. From (1) the components of the propagation constant are given as

{I ( ({1I)2) 2" 1+ {' -1 f3 - 211" - A

For low-loss materials ({II/{I < 0.1) parameters 0: and f3 are approximately equal to

Thus the propagation velocity {) is {) = ~ where c is the speed of light in vacuum. y{1

t

~ T h

1 /

/ /

I

I z

- 84 -

y

t

Er

(0)

ELECTRIC FIELD LINES

MAGNETIC FIELD LINES

(b)

Figure 1: Three Coplanar Strips Sensor

Er

- 85 -

50r------r------.------.------.------.------r------,

45

40

35

c C1l 30 1ii c o o .g 25 "0 Q)

]? 20 o

15

10

5

£ =?

£0= I

O~--~~----~----~--__ --L-----~--__ -L----~ 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

Normalized phase velocity

Fig. 2. Dielectric constant of the test material vs normalized phase velocity

The propagation velocity in the TEM transmission systems is the same as that in free space provided that the wave propagates in the same medium. Thus the propagation velocity in low loss dielectrics depends mainly on the dielectric constant and little on the loss factor.

3. Coplanar Sensors

Coplanar guides are transmission lines for which all the conductors are in the same plane. These guides include three types of transmission lines (coplanar waveguides, two and three strip lines). The most advantageous configuration consists of three parallel strips of equal width (6.25 cm) in contact with the test dielectric (h -+ (0) as shown in Figure 1.

Figure 2 presents the dielectric constant of the test material, placed in contact with the sensor, as a function of the normalized -preparation velocity. The latter can be measured using suitable frequency and time domain methods. Since the propagation velocity is related to the density, composition and moisture content of the test material it may be possible to determine these properties by measuring the normalized propagation velocity of the wave passing through the sensor.

Conclusions

In this paper a concept of using coplanar sensors for material moisture measurements is outlined. A major advantage of using coplanar sensors in that -the averag~ (along the sensor) value of the dielectric constant (and moisture content) is obtained. The propagation of lowloss materials is inversely preportional to the square root of the dielectric constant.

- 86 -

MICROWAVE SENSORS FOR MEASURING MOISTURE CONTENT

OF SOLID AND LIQUID MATERIALS

Quang Xiang Song

Hangkou Branch of Huangzhong University of Science and Technology

\\Juhan, 430012 P.R. of China

ABSTRACT

Two types of microwave sensors designed for determination of moisture content are discussed in the paper. One sensor contains a section of rectangular waveguide with a slot cut along the middle of the broader wall of the waveguide. Solid materials in form of sheet or plate are inserted through the slot into the waveguide. It was found out that the attenuation introduced by the sensor into microwave power transmission from a source to the detector is linearly related to the moisture content of the sheet materia!. To increase the sensitivity of the system, the slotted section of the rectangular waveguide was filled with a lossless material - polyethylene of dielectric constant ( = 2.54. Wavelength in the waveguide section is reduced by 76% [1], thus its electric length is increased by the same amount. It was observed that the slope of the relationship between output power and moisture content of a paper board increased from - 0.60 to - 1.06, as shown in Fig. 1. The slope increased by 76%, as expected.

10

.. ~ QJ

~ 5 0..

o

I

5 10 15 Moisture content, %

rectangular waveguide

/ -<:"

r---.. t::A plastic pipe

Fig. 1. CharacterIstics of slotted sensor Fig. 2. Rectangular waveguide sensor

The second sensor consists of a lossless dielectric pipe located in the middle of a section of rectangular waveguide operating at the wavelength A = 3.2 em, as it is shown in Fig. 2. The amount of microwave power absorbed in the liquid flowing through the pipe is proportional to the amount of water in the fluid [2]. Experiments carried out with a liquid glue indicated good sensitivity and feasibility of the sensor. At frequency f = 9370 MHz, the permittivity of water is ( = 61.5 - j3.14, while for rubber ( = 2.4 - j 0.012. Thus even variation by a 'small amount of water content in a glue can be effectively detected by microwave sensor.

[1] Zhi Yushen, Microwave Techniques, Nat!. Defense Industry Press in China, 1984. [2] Bian Yinuo, Testing emulsions by microwave and its sensor, 6th Annual Meeting on Non-electric

Quantity Electronic Testing, China, Oct. 1991.

- 87 -

Portable Moisture Measurement System

Abstract

F. Menke, R. Knochel

Lehrstuhl fUr Hochfrequenztechnik, Universitiit Kiel, Technische Fakultiit Kaiser-Str. 2, 24143 Kiel, Germany

This paper presents a portable moisture measurement system that makes use of phase-shift and attenuation of the electro-magnetic field in a material under test where a plane wave is introduced. Only single sided access is needed making the system useful for many measurement situations. Examples are presented showing the feasibility such as measurements on granular materials in bulk like oats, on a wall of a building and on soil water content. For homogeneous samples the average moisture value is found, but for moisture distributions a depth profile can be obtained, too.

1 Introd uction

The importance of moisture measurement and control in nearly all industrial environments is well known. Commonly used indirect moisture measurement systems are based on various physical interactions with the water content of the material under test. Use of microwaves has a lot of advantages in this field and therefore microwave moisture measurement procedures are frequently applied. A survey is given by Kraszewski in [1]. The method benefits from the great difference in the dielectric constant Er = E' - j/ of dry material with E' ~ 2 - 5 and water with E' ~ 80 in microwave frequency region. The moisture measurement methods can be divided into those using resonant set-ups, where a dielectric material detunes a cavity resonance, into transmission or reflection set-ups, where the transmission through or the reflected signal of a material is used, and into TDR-set-ups, where the travel-time of a short impulse is measured. Usually such measurement systems are installed directly at the production process equipment and can hardly be removed for a fast measurement at an other place. Just this feature is important in many measurement situations like test production in industry, in medium-sized firms with fast changing production processes, in agriculture for animal food storage control, in meteorology, in civil engineering and so on. Important features for such a moisture meter are:

• portability, that means easy transportability and installability,

• requirement of only single sided access,

• delivery of representative measurement result for a significant volume of the sample,

• possibility of long time observation, e.g. in civil engineering.

Regarding the methods discussed above an open resonator set-up could fulfill some of the requirementsl.

However, the measurement results using an open resonator sensor are extremely dependent on air gaps between probe and material. Additionally the set-up delivers only moisture values for a small material volume near the surface of the material.

2 Method and Examples

A measurement system fitting the above discussed demands is shown in Figure 1. A network-analyser (NWA) is connected to a planar antenna, irradiating a free space plane TEM wave which then travels into the material. In a central hole a field probe is moved, protected by a thin PE pipe with a diameter of 15mm. As the field probe a small electric or magnetic dipole connected to the receiving port of the NWA via a balun and a thin semi-rigid cable is used. Applying this set-up, attenuation and phase-shift

Ie.g. the HUMY 100 from Miitech, 21218 Seevetal, Germany

r-----------, I I I I I I I

Hand I Held I Unit I

a)

I I I I I I Planar I Antenna I L __________ ..J

- 88 -

Phase Planes

b)

Figure 1: a) schematic measurement set-up and b) schematic field propagation with a dielectric disturbance

are monitored while the field probe moves into the material, driven by a stepping motor. NWA and stepping motor are controlled by a PC. The insertion-hole in the material and hence the traveling path of field probe cable are arranged orthogonal to the plane wave electric field and do not disturb the field distribution.

The measurement frequency is selected at 2.45GHz. This frequency is a compromise in several ways:

• at 2.45GHz a planar antenna has proper dimensions of less than 30cm in diameter, because size is important for a transportable device,

• with a dynamic range for the NWA of about 70dB, reachable with reasonable effort, the propagation depth for an average moisture content is ::::::lm. Higher frequencies only achieve less penetration depth,

• at 2.45GHz the influence of ionic conductivity, important at low frequencies, has already decreased signifi can tly,

• due to the high dielectric constant inside the material the wavelength reduces to ::::::5cm. The influence of inhomogeneities in the material is therefore small as long as the dimensions are less than approximately lcm.

The planar antenna is realised as 2x2 patch antenna on Teflon substrate. The power distribution network for the patches is arranged on the back of the antenna. Measurements show, that far-field conditions in air are reached at a distance of about 10cm in front of the antenna. This distance even decreases during measurements according to the higher permittivity of the material. The stepping motor enables a depth resolution of at least O.35mm/step and a maximum measurement range of 70cm. A PEpipe protects the field probe against mechanical damage and prevents drying or moisture transport in the insertion-hole. Using locally fixed pipes long time observation at specific locations becomes possible.

If the material layer is thick enough or the attenuation in the material is high, the reflected wave on the back of the layer can be neglected. In this case a one dimensional model is valid where the electric field component is radiated into the wall with .

E(z) :::::: Eo . e-'z (1) z

Eo is the electric field strength at the surface, z the depth coordinate and 'Y the propagation constant with 'Y = j.':!!... /f:;: = a + j (3. Here w means the angular frequency, Co the velocity of light in vacuum, Co V'-T

a the attenuation and (3 the phase constant. With the measured phase shift = (36.z and attenuation A = a6.z between two positions (z) and (z + 6.z) the dielectric constant can be calculated from

, (~)2_(~? f. = lL LL (2)

.':!!... Co

and

- 89 -

II 2(~)(..LL) € = L\.z Az

~ Co

(3)

As already mentioned the amount of water affects attenuation and phase shift and therefore the dielectric

-1

-2 ............ , .... "["."" .... "." ... '1''' ·""·,,:=:T.-·~:~:~:·::·7:.

~-3 ""'"''''''''''''f''''''''''''''''''-]--''' """"'"''\'''''''''''''''''''

'-' -4 <

.. "."."" ... ""j"'" ...... "." ... j" ... " ... ""j ... " ..... " .. ,. -5

a)

-6 ."" .. "" ...... y"''''' .. '''''' .. j-''''''''''.''''.''T'''."." ..... ". -7~--~~--~~--~~--~

0.1 0.2 0.3 0.4 0.5 z [m]

,.......,'-'~ -20 .. "".""".".+'." .. " .. " ... "+ ........ ::.~.'-'~<::~:~:~" .. - ''''~ ; : ;

~-30 ·"'"'''''''''·''[''''·, .. ,,·''''·'''1 .. '' .... ·''''''·;'''''' ........ ''' ; : ! . . .

-40 ·" .... ·, .... " .. [ .. ""·,· .. "· .... j·,,·,,··,,,,, .. l .. ,,· .. ·· , , .

-5Ql~~~--~--~~~ -0.1 0.2 0.3 0.4 0.5 b) z [m]

Figure 2: Attenuation and phase shift in soil

constant as shown in eq. (2) and (3). Usually the influence of attenuation on the dielectric constant is much smaller than that of phase shift. Instead of using the permittivity of the material other quantities could be used such as attenuation- and phase constant or the relation between attenuation and phase ~, which itself is independent of layer thickness and mass or density of the measured volume.

Next, the applicability of the measurement set-up will be demonstrated. Figure 2 show an example, where soil is analysed. Soil water content is an important information used in meteorological models. The ground surface is located approximately 32cm in front of the antenna, Closed lines show the measurements, dashed lines represent calculations with air as the dielectric. Attenuation in air is derived from AdB = K - 20 log· 4;Z [2) where the correction factor K considers the gain of patch antenna and dipole, cable and balun losses. Phase shift in air can easily be determined from cI> = ~. z. The differences co between the two curves are caused by the dielectric material. Looking at the absolute values of A and cI> it is obvious that cI> has a dominant influence on the real part in eq. (2). Inhomogeneities such as small stones can be the reason for the deviation from the expected continuous shape of the curves, especially with respect to the attenuation curve. The average permittivity value for the moist soil material is €' =9.03 and 8.97 when neglecting the attenuation and deriving / only from phase.

A second measurement example is visualised in Figure 3 a) which is taken from a layer of oats. The dielectric loss of oats with a moisture content of'ljJ = 10% is small. The measured layer thickness is about 80cm leading to a considerable backward travelling wave in the material. In this case wave propagation can be described by the one dimensional wave equation where the common solution is given by forward ( +) and reverse (-) traveling waves

(4)

The propagation constant and therefore the dielectric constant can be derived from the second derivative of E [3)

€r = _ (CO)2 2. 82E

w E 8z2 (5)

The second derivative, when realised with a difference equation, makes this calculation sensitive for noise influences. Using smoothing or estimation algorithms [4], this influence can be reduced.

Figure 3 a) shows €' of oats. Due to the assumption of constant 1 in eq. 4 the sharp step in the dielectric constant at the surface causes a peak in the calculated €'. With only slowly varying conditions in the material the measurement is capable of delivering the depth profile of the permittivity and hence moisture within the material.

A moisture profile inside a body can be determined this way, which can be useful for searching moisture sources or hidden water collections, which is of importance especially in civil engineering. Figure 3 b) show a measurement example for a dry wall.

The scanning of the electro-magnetic field with the field probe results in a vectorial superposition of field components reaching the probe location. When a significant dielectric disturbance is placed inside the dielectric body as sketched in Figure 1 b), superimposed waves can result in destructive interference. Then measurement quality deteriorates. To reduce this effect, multiple measurements at various positions

6 .---~--~----~--~--~

5 ····i················!···············{···············

4 ···· .. ·····t····· ...... \""" ........... ""\" ............ "]"" ........... . "'vJ3

2

1

o ~--~~~~~~~~~~ 0.1

a)

- gO -

10.---~----~--~--~----~

8 .............. ; ................ ; ............... > ............... ; ............. ..

6

VJ 4

2

O~---L~~~~~~~~--~

-0.05 0.2

b)

Figure 3: a) dielectric constant of oats with 'Ij; = 10% b) dielectric constant profile of a wall

in the vicinity of the first hole could be carried out and evaluated. This would lead to an averaging of the interference pattern in the lateral plane. Measurements at various places, however, are tedious and time consuming. Alternatively a planar antenna with multiple separately fed patch configurations could be used as indicated in Figure 1 b). Switching between the sub-arrays with respective amplitude and phase measurement is then equivalent to moving the field probe within the material. The thus derived mean value is supposed to lead to an improved measurement result. Using this method it would be possible to decide whether the virtual gap in the wall in Figure 3 b) at z =0.15m is really an air inclusion or caused by interference.

A further improvement can be obtained by eliminating the mechanically moving field probe replacing it by a probe array longitudinally extended across the measurement path. Each small probe across the array is loaded by a switching diode which modulates its reflection coefficient, thus producing an array of modulated scatterers for measuring the field distribution. Measurement speed can be significantly increased this way.

3 Conclusion

In this paper a portable moisture measurement system is presented evaluating the local distribution of electro-magnetic field profile in a material under test. A theoretical description and measurement results are demonstrating the practical functionality. The system needs only single sided access to the specimen and is capable of delivering depth moisture profiles with proper accuracy if only slight changes in propagation constant are present. An improved set-up using a planar antenna divided in sub-arrays is proposed to reduce the influence of interference due to inhomogeneities using the mean value of measured attenuation an phase-shift determined with sub-antenna.

4 Acknowledgement

This work is part of the project Kontinuierliche und zerstorungsJreie FeuchtemejJverJahren mit Mikrowellen. Financial support by the Volkswagen-Stiftung under contract I/67 193 is greatfully acknowledged.

References

[1] A.W. Kraszwski: Microwave Aquametry - Needs and Perspectives, IEEE Trans. MTT, vol. 39, no. 5, pp. 828-835

[2] H.G. Unger: HochJrequenztechnik in Funk und Radar, Teubner Verlag Stuttgart, p. 37, 1988

[3] F. Menke, R. Knochel, T. Boltze, C. Hauenschild, W. Leschnik: Moisture Measurement in Walls using Microwaves, IEEE MTT-S Digest, pp. 1147-1150,1995

[4] T. Boltze, C. Hauenschild, KD. Kammeyer, W. Leschnik: Robust Estimation Algorithms Jor the Determination oj Moisture Profiles with Microwaves, Proc. Feuchtetag 95 Berlin, pp. 47-56

- 91 -

THE DIELECTRIC PERMITTIVITY METER

FOR SOIL MOISTURE DETERMINATION

Wojciech Marczewski

Space Research Center, Polish Academy of Sciences, Bartycka 18A, 00-716 Warszawa, Poland.

Abstract. The system performs a measurement of input impedance of a short monopole antenna inserted in a medium under test. Input impedance depends on propagation properties of the medium and enables determination of the dielectric constant. The paper describes the system containing a 10 mW source operating at L-band, six port measuring arrangement and simple 8-bit digital circuitry, designed for this purpose. Data taken with the instrument presents ground-truth measurements to be used for soil moisture assessment from ERS-1 images.

INTRODUCTION

Investigations on soil moisture content in grass covered and cultivable fields have been carried out in Poland by the Remote Sensing Center OPOLIS and Space Research Center of the Polish Academy of Sciences in the frame of ESA ERS-1 Pilot Project PP PL4 [1]. The methods used for the field measurements are capable to determine soil moisture content up to limited depth. Results from ERS-1 (5.5 GHz) and from AVHRR radiometer (far infrared spectrum) installed on the board of NOAA satellite need to be compared with ground-truth measurements obtained by verified method. The method of short monopole antenna [2, 3] operating in L-band was developed for in-situ measurement of soil permittivity.

SHORT MONOPOLE ANTENNA

The permittivity measurements are carried out under the assumption that the medium under test is illuminated with a monochromatic continuous wave by a short monopole antenna. The antenna is not matched and it reflects and receives waves traveling backward. The ratio of the total reflected wave (a) to a radiated wave (b) has its contribution into a ratio of the incident wave (A) to the returning wave (B) measured in a feeder line. Thus, an input impedance of the antenna (or the related reflection coefficient) bears an information about propagation properties of the medium. The matter is to separate this contribution from the measured impedance and the antenna characteristics. The method proposed in [2] seems to be atractive because of its simplicity. It does not require taking samples of soil and loading them into transmission line, and even more important, it does not require a precise knowledge of the antenna characteristics. The method requires impedance measurements to be performed twice and at least at eight frequencies, once with the radiating element directed into free space and the second time with the antenna immersed into the medium under test.

Disadvantage of the method relates to the fact that the reflected wave contains also components reflected from near and far objects, located in all directions inside the antenna bandwidth. As the result, the assessment of an effective thickness of the layer under test can be hardly made. The method is based in fact on an integration of returning waves over certain volume of the illuminated medium. Short monopole antenna provides wide angle of illumination and works well for material located in its close vicinity. Short monopole antennas of the length of few millimeters backed by a grounded metal plane sheets being 60 by 60 cm wide, were used in this study. A portable instrument containing the antenna, six-port measuring system and controlled by a laptop PC-XT was developed for field operation. Schematic diagram of the measuring system explaining the principle of operation is shown in Fig. 1.

RF source freq=f(t)

incident

p\;---n dou~I~. 2 coupler

UT135 feeder line

short monopole ground plane

~ backward

'-'--~--p'" ,

;';' " ,,, " ;';' // ;';' '/ ;';' /' ;'" '/ ,,;' ,,, ;'" '/ ,,;' ,,/ " " ,,, " ;';' /' ;';' ;'"

Fig. 1. Diagram of the measuring system showing the principle of its operation: an L-band oscillator, directional coupler, two waves traveling in opposite directions, monopole antenna radiating into a medium under test and fields in the medium.

- 92 -

The antenna input impedance contains an information on the medium permittivity, but it has to be measured without phase ambiguity. Thus, the measurements should be performed for the medium of the permittivity in the range from 1 to 80 at different frequencies to avoid the ambiguity in phase measurement. Design of the instrument includes measuring of the complex values of the input impedance, performed in a sufficiently wideband six-port arrangement.

INSTRUMENT DESIGN

The instrument contains two channels: one channel dedicated to signal processing of an incident wave feeding the antenna and the other to process the wave reflected from the antenna. These two signals correspond to two components of a complex number of the antenna input impedance. They are extracted from the feeder line and separated by a double directional coupler. Each of the signals feeds the set of Schottky detectors terminating four outputs of the pentagonal ring in the respective channel. The pair of measured signals feed each set of detectors in an orthogonal mode. The sets of detectors are simultaneously fed with a reference signal supplied in a balanced mode. In effect, each diode in a set is fed with a different

pair of two components - measured and reference. In the result, one pair of detectors in the set supplies a sum of signal proportional to the real part and the second pair provides a difference that is proportional to the imaginary part of a product of complex conversion. The real and imaginary parts are further processed, respectively, in sine- and cosinechannels. Output products of the channels constitute a measured complex number proportional to the incident and reflected signals needed for impedance determination.

The instrument contains the VCO providing 10 to 20 m W output power in the frequency range from 0.9 to 1.7 GHz. The oscillator signal is split on two reference paths, as shown in Fig. 2. The third path is that feeding the antenna. This three way power distribution is provided by a ring of four directional couplers [4]. It was realized that broadband signal conversion is an important factor in the system design and can limit performance of the instrument, if done incorrectly.

To obtain a homo dyne conversion, a bi-static phase switch was introduced by providing a phase modulation in the path feeding antenna, as shown in Fig. 2. Thus, the measured components are phase modulated, while the reference components are monochromatic continuous wave. As a result, products of conversion are rectangular waves of frequency 5 kHz which are next filtered, amplified, sampled and sorted. All digital processing is performed at a low 10 kHz rate. Sampling is performed every half period of 5 kHz. Instrument operates with a portable PC performing data storage on a disk. Then the stored data is post processed to impedance, and finally to the dielectric constant of the tested medium. The instrument was calibrated in terms of the antenna input impedance. Permittivity tests with several types of soils, preliminarily, the uncertainty in permittivity is estimated to be less than 10%. Improved, temperature stable version of the instrument with a phase-locked loop is under development.

- 93 -

incident

incident ,-----, wove

RS232

Zo

* mod 5kHz

reflected

Fig. 2. Diagram of the six-port refiectometer measuring input impedance of a short monopole antenna radiating into the ground (soil permittivity measurement).

CONCLUSIONS

The soil permittivity meter using the short monopole antenna was discussed to presenta hardware aspects of the design. Further research on determining moisture content in various soils are in progress. Ground based field measurements are to be compared with conventional standard method. The program will be carried out on the agricultural site in Wolsztyn region of Obra river valley in central Poland.

References.

[1] K. Dabrowska-Zielinska, M. Gruszczynska, M. Janowska, K. Stankiewicz and Z. Bochenek, "Use of ERS-1 SAR data for soil moisture assessment", First Workshop on ERS-1 Pilot Project, pp. 79-84, Toledo, 22-24 June 1994.

[2] Y.D. He and L.C. Shen, "Measurement of complex permittivity materials using a monopole antenna", IEEE Trans. on Geophysics and Remote Sensing, vol. 30, no 3, pp. 624-627, 1992.

[3] G.S. Smith and J.D. Nordgard, "Measurement of the electrical constitutive parameters of materials using antennas", IEEE Trans. on Antenna and Propagation, vol. AP-33, no 7, pp. 783-792, 1985.

[4] W. Marczewski, "Ring of overlapped microstrip hybrid couplers", Proc. of PIERS-93, p. 309, JLP, Pasadena, CA, July 1993.

- 94 -

-95 -

MICROWAVE INDUSTRIAL SENSOR FOR CONTINUOUS MEASUREMENT

OF MOISTURE CONTENT IN FAT LIVER EMULSIONS

G. Cottard, J. Guillon (Satimo, ZA Courtaboeuf, 91152 Les Ulis, France)

B. Piriou, C. Sourioux (Bizac, Le Teinchurier, 91902 Brive Cedex, France)

1. Introduction

Fat liver (foie gras in French) emulsions are basically prepared from a mixture of duck or goose livers and water. The knowledge of the moisture content is of prime importance for satisfying the standards enforced by food regulations and for maintaining reliable product quality. Moisture content is usually deduced from weighing an emulsion sample, as it is at a given stage of production process and after drying it. Such a measurement is only conducted on mixtures sample during the fabrication process and requires about one hour to complete and to obtain the moisture content value. Such a duration time is evidently incompatible with a permanent control of the whole production. For this reason, a 3 GHz microwave sensor coupled to a temperature probe has been developed for providing continuous determination of the emulsion moisture content during its fabrication. By comparison to two series of reference measurements, the accuracy of the microwave measurements is shown to be about 1 % moisture in the specified moisture content range from 30 to 50%.

2. Sensor principle

The basic idea is that the complex permittivity of the emulsion is, in a given production configuration, mainly dependent on its temperature and water content [1, 2]. A mixed reflection/transmission arrangement has been selected for obtaining the reflection coefficient Sl1 of an antenna immersed in the emulsion, and the transmission coefficient S12 between two closely spaced such antennas. An empirical model has been developed for correlating the moisture content to the temperature of the emulsion and to the Sl1 and S12 coefficients. The use of both reflection and transmission coefficients provides an efficient way for compensating for possible fluctuations resulting from the emulsion flow in front of the sensor.

r ~~PENSATION ~ UNIT r-----~1IIoo-i

AMPLITUDE DETECfOR

EMULSION: MODEL

( D

(

j Fig. 1. Principle of the microwave sensor for the measurement of the moisture content

in fat liver emulsions

I:

- 96 -

The simple quadratic model has been used for calibration based on intensive preliminary measurements performed on the production line. The operating frequency is equal to 3 GHz. Such a frequency provides a good compromise with respect to penetration depth in the emulsion, size of the antennas, sensitivity to the moisture content and relative insensitivity to emulsion inhomogenities. For simplifying the electronic circuit assembled to the sensor, only amplitude measurements of the reflection and transmission coefficients are performed. For compensating possible drifts of the output signals, resulting from fluctuations of the power supplied to the microwave generator, a compensation unit has been added in parallel to the measuring channel. This reference channel is periodically used between two series of measurements to check stability of the generator.

3. Sensor integration in the production line

The sensor is introduced into a large toroidal bowl where fat livers and water are mixed. The section of this bowl is roughly semi-circular. The measurements are performed continuously while the bowl is rotating. Oppositely to the sensor, a rotating mixer is used for crushing the livers. In such a facility, a batch of 20 to 100 liters of emulsion is prepared within about 15 minutes. The antennas consists of two rectangular (3 cm by 2 cm) apertures covered with a teflon (PTFE) randoms. The distance between the two antennas is 3 cm. The antennas are matched with a VSWR less than 1.1 when radiating into emulsion of 44% moisture content. They are contained in a box of a shape specially designed for allowing a regular emulsion flow around the sensor. For providing contacts with food material, the box is cover with stainless steel envelop. At 3 GHz, the final product has the complex permittivity f = 20 - j5.0. The sensor contains also a temperature probe.

4. Results

The accuracy of the measurements performed with the microwave sensor has been assessed by comparison with reference data, obtained independently by Bizac and the French Technical Center for Food Industry (CTCPA) by standard weighing and drying techniques. The comparisons were conducted for the samples of fat liver mixtures of different moisture contents, before and after addition of water. The moisture content given for reference data is in the 35 - 38% moisture range, or in the 43 - 45% range, respectively before and after addition of water. The agreement between both series of reference data is always better than 0.2% moisture. With respect to these reference results, the error of the microwave measurements was less than 1.1% moisture.

5. Conclusion

The microwave sensor described in this paper has met all required specifications. By means of conveniently calibrated empirical model, the sensor provides the moisture content of fat liver emulsions with required accuracy. At the cost of a specific calibration procedure, this simple and low-cost microwave sensor could be used for other food mixtures involving water addition and of comparable viscosity.

6. References

[1] Y. Dufour, "Mise au point d'un humidimetre microonde applique au foie gras" , Memoire de DES, Universite de Bordeaux I, 8 Juillet 1992.

[2] R. Besnard, A. Le Blanc, G. Cottard, J. Guillou, B. Piriou and C. Sourioux, "Development of a microwave sensor for the control of an eml,llsion humidity", 6th European Conference on NonDestructive Testing, Nice, 24-28 October 1994.

- 97 -

ACCURATE RESONANCE FREQUENCY ESTIMATION ON A DIELECTRIC RING RESONATOR OF A MOISTURE MEASUREMENT SYSTEM

Seichi Okamura and Shigekazu Miyagaki

Shizuoka University, Dept. of Electrical and Electronic Engineering 3-5-1 Johoku, Hamamatsu, Shizuoka, 432, Japan

Abstract This paper describes the method how to detect accurate resonance frequencies in a moisture measurement system using a dielectric ring resonator operating at about 2.1 GHz. A curve fitting midpoint method is superior in examined three methods, especially for high moisture samples because it has the ability to prevent the noise superimposed on the resonance system. The cardboard samples having about 5 to 115 % moisture in dry basis are measured and the standard deviation of moisture content is 1.14 %.

I. INTRODUCTION Moisture measurement methods using microwaves have been studied for industrial or agricultural

production process such as veneer, grain, green tea, etc., and they become important instead of conventional measurement methods [1]-[5]. The principle of the measurement is based on the change of the attenuation or phase shift of microwaves caused by the moisture contained in the materials under test. These methods are very useful for a large sample because the whole beam of the microwaves transmitted from a antenna can propagate through the sample. However, they can not be used to measure a small size sample as one of grain seeds or wood chips. The microwave resonant method is very useful to measurement of such case. Kraszewski et al. measured single soybean seeds from the change of Q-factor and the resonance frequency shift using a resonant cavity made by rectangular waveguide[6]. Okamura et aI., using a dielectric ring resonator, measured small sample of paper from the change of Q-factor and the weight [7].

The dielectric ring resonator method has a merit that it can measure over wide range from low to high moisture content, but the Q-factor is affected by room temperature by the change of the gap between the resonator and the couplers. It is difficult to avoid the affect. However, the resonance frequency of the resonator is not so affected by the room temperature as the Q-factor. Then, the resonance frequency is used for moisture measurements[8].

In this paper we show the comparisons among the three detecting methods to get the accurate resonance frequency of the moisture measurement system using a dielectric ring resonator.

II. MOISTURE MEASUREMENT SYSTEM The system is made up of a dielectric ring resonator, a microwave sweep oscillator, a scalar network

analyzer, an electric balance and a personal computer as shown in Figure 1. A sample under test is set in the center of the hollow core of the ring resonator. Microwaves from a sweep oscillator, whose center frequency is about 2120 MHz, are fed to the resonator through the coupler-1 and lead to a scalar network analyzer through the coupler-2. The resonance curve is displayed on the screen of the analyzer and the resonance frequency is detected by using the methods mentioned bellow.

The resonator is made of a Teflon ring, whose relative permittivity is 2.08, and two conducting plates as shown in Figure 2. The resonant mode is TEolh the size of the dielectric ring is decided referring to the resonant mode chart of dielectric rod resonators [9]. The resonance frequency changes with the permittivity of the sample dependent on the moisture content. The resonance frequency is mainly decided by the real part of the permittivity. The influence of the imaginary part to the resonance frequency can be neglected when the quality factor of a resonator is higher than 353 [8,10]. As the Q-factor of the resonator used in this study has about 1000 or more, the imaginary part of permittivity can be neglected.

- 98 -

Sweep Oscillator

Dielectric Ring Resonator

REF. iGPIB

GPIB

Scalar Network Analyzer

Fig. 1 A block diagram of a microwave moisture measurement system using a dielectric ring resonator.

Dielectric Ring

D =400. Omm 2L= 56. 3mm

Conducting Plate

2a=199.9mm 2b= 40. Omm

Fig. 2 A dielectric ring resonator.

To reduce the fluctuation of the frequency caused by the change of room temperature, the moisture is measured from the difference of the resonance frequencies before and after inserting a sample into the resonator. It is shown in absolute value as the resonance frequency shift .6F.

III. ESTIMATION OF RESONANCE FREQUENCY It is important how to detect the exact resonance frequency from a resonance curve. In the

measurement of a sample with low moisture content or thin thickness, the Q-factor of the resonator is high enough. The resonance curve is so sharp that the exact resonance frequency can be decided easily from the peak value of the curve even if the noise in the detecting circuit is superimposed on the curve. In the measurement of high moisture or thick sample, however, the resonance curve is not so sharp that the position of the peak is influenced by the noise. We discuss the following three methods about the detecting of the exact resonance frequency. The first is the peak method which detects the resonance frequency from the position of the peak of a resonance curve. The second is the straight line fitting midpoint method. It decides a resonance frequency from the intersection of the resonance curve and the straight line which is connecting some midpoints as shown in Figure 3. The each midpoint, which is shown by a dark triangle in Figure 3, is the center frequency of the bandwidth corresponding to a level on the resonance curve. The third is the curve fitting midpoint method. It is obtained by fitting a second degree curve to the points.

The three methods have been examined using 108 cardboard samples of circle whose diameter and thickness are 35 mm and 0.67 mm, respectively. They are evaluated by the error distributions of the resonance frequencies obtained by each method and by eye method. The results are shown by histograms in Figure 4. The horizontal axis shows the difference from the true value which is divided by the 10 kHz interval. The vertical axis shows the number of samples. Figure 4(a) is for the samples below 60 % moisture in dry basis and (b) is for the samples beyond the moisture. The upper shows the peak method, the middle and the lower are the straight line fitting point method and the curve fitting midpoint method, respectively.

In the range below 60 %, the peak method is superior to the other methods for the number of the

Resonance Frequency

by Straight line fitting by Curve fitting midpoint method

1 _____ by Peak Method

+ midpoint method

Fig. 3

7-

Frequency (Hz)

Estimation of a resonance frequency from a resonance curve

40

'" 30

t 'i5 20

t .r> e :f 10

o

30

30

Peak method

Straight line fitting midpoint method

Curve fitting midpoint method

~ ~ ~ ~ ~ 0 10 W ~ ~ ~ @

Difference from true value (kHz)

(a)

- 99 -

40

'" 30 ., is. e ill 'i5 20

!:l .r> e " z 10

30

o

30

o

Peak method

Straight line fitting midpoint method

Curve fitting midpoint method

~ ~ ~ ~ ~ 0 10 W ~ ~ ~ w Difference from true value (kHz)

(b)

Fig. 4 Histograms of resonance frequencies measured by three methods. The upper is the peak method, middle is the straight line fitting midpoint method and the lower is the curve fitting midpoint method. (a) is for the samples below 60 % d.b. and (b) is for the samples over 60 % d.b ..

samples in the center range of the deviate frequency. As for the concentration to the true values, the straight line fitting midpoint method and the curve fitting midpoint method are superior to the peak method. Therefore, we can see little difference in the three methods. In the range over 60 %, there are remarkable difference among three methods. The peak method is good as for the number of samples in the center range of the deviate frequency, but the concentration is inferior to the other methods. The straight line fitting method is good in both characteristics of the number in the center range and the concentration except for the center of the distribution shifting into the lower frequency. The curve fitting midpoint method is as good as the straight line fitting method in the both characteristics. Furthermore, the error distribution coincides with the region of zero very much. It is obvious that the resonance frequency detected by the curve fitting midpoint method is closer to the true value. The method is useful for detecting the resonance frequencies automatically especially for high moisture samples.

IV. CALIBRATE EQUATION & MEASUREMENTS The resonance frequency shift is proportional to the thickness or the mass of a sample[lO]. It may be

- 100 -

expected that the resonance frequency shift per unit mass is independent of the sample quantity. We denote the normalized resonance frequency shift per unit mass as 11 F/M. The relation of the moisture content versus the 11 F/M at true resonant values is expressed by a approximate equation in a sixth order polynomial as follows,

q; = -6.9 + 121x -140x2 + 118x3 - 49x4 + 9.2xs - 056x6 ( 1 )

where q; is moisture content in percent and dry basis and x is 11 F/M in MHz/g. The correlation

coefficient of the equation is 0.998. A cardboard samples of various moisture contents are measured by the three methods. The predicted

moistures of cardboard samples were calculated by the eq. (1) and the true moisture contents were measured by oven method using a balance. The standard deviation of moisture content in dry basis are 1.14 % by the curve fitting midpoint method, 1.24% by the straight line fitting midpoint method and 1.68 % by the peak method at the samples having the oven moisture of about 5 to 115 %.

The curve fitting midpoint method can be measure within the least error in the three methods. The error of the curve fitting midpoint method is within ± 2 % for the samples under 60 % moisture and ± 5 % for the samples with the moisture from 60 % to 115 % in dry basis (d.b.). The whole characteristic of the system by the curve fitting midpoint method is as shown in Figure 5 (a) and the error is shown in (b).

_!--_-L __ '---_-,

o 20 40 60 80 100 120

True moisture (% d.b.)

i -4.0 e-.I --+-

!

-8.0 i

o 20 40 60 80 100

True moisture (% d.b.) 120

Fig. 5 The moisture content by the curve fitting midpoint method for cardboard samples with different thickness. (a) is the whole characteristic of the system by the method and (b) shows the difference from the true moisture value.

References [1 ]Nyfors and P. Vainikainen, Industrial Microwave Sensors. Artech House, 1989, chapter 4, pp. 201-230. [2]P.Vainikainen, E.G.Nyfors, and M.T.Fischer, "Radiowave sensor for measuring the properties of dielectric

sheets: Application to veneer moisture content and mass per unit area measurement", IEEE Trans. Instrum. Meas., vol. IM-36, pp.1036-1039, Dec. 1987.

[3]T.Okabe, M.Huang and S.Okamura, "A new method for the measurement of grain moisture content by use of microwaves", 1. agric. Engng. Res., VoLl8, pp. 56-66, March 1973.

[4]Y.Miyai, "A new microwave moisture meter for grain", 1. Microwave Power, Vo1.13, PP.163-166, June 1978. [5]S.okamura and F.Tomita, "Microwave moisture sensing system in drying process for green tea production",

in Proc. IEEE Instrum. and Meas. Technology Conference IMTC/94, Vo1.3, pp.1253-1256, 1994. [6]AW.Kraszewski, T.You, and S.O.Nelson, "Microwave Resonator Technique for Moisture Content

Determination in Single Soybean Seeds",IEEE Trans. Instrum. and Meas, Vol.38, pp.79-84, February 1989. [7]S.okamura and KOhishi, "Dielectric rod resonator for a moisture content measurements by microwaves",

IEICE Trans. (c), vol. J70-C, pp.1523-11528, November 1987. [8]S.okamura and T.Masuda, "A new moisture content measurement method by a dielectric ring resonator",

IEEE MTJ'-S International Microwave Symposium, WSMJ, pp.59-62, 1994. [9]Y.Kobayashi and S.Tanaka, "Resonant modes of a dielectric rod resonator short-circuited at both ends by

parallel conducting plates", IEEE Trans. Microwave Theory Tech, vol. MTT-28, pp.1077-1085, October 1980. [lO]M.Nakajima, Microwave Engineering. Morikita Publ. Japan, 1984, chapter 5, pp.152-153.

- 101 -

DEVELOPMENT OF MICROWAVE METHODS TO MEASURE TRACES OF WATER IN GASES

Jacques Goyette and Tapan K. Bose, Institut de recherche sur l'hydrogene, Universite du Quebec a Trois-llivieres

Trois-llivieres, Quebec, Canada, G9A 5H7

Michel F. Frechette, Hydro-Quebec (lREQ), 1800 Montee Ste-Julie, Varennes, Quebec, Canada, J3X lSI

INTRODUCTION

Detection of small quantities of water in gases like SF 6 and CH4 using a physical method has always posed a great challenge. Although there exist chemical methods to measure 1 or 2 ppm of water in gases, to the best of our knowledge, there is yet no in situ physical method that can reach such precision. In an effort to develop a physical method to measure traces of water in gaseous samples, we employed a microwave technique where the presence of contaminants in gases was measured by the shift in the resonance frequency of a microwave cavity.

DIRECT MICROWAVE METHOD

The resonant frequency of a cavity depends on its geometry and on the dielectric constant E of the material filling

it [1]. A rectangular cavity having dimensions a, b, d, filled with gas has a resonant frequency fr defmed by:

c m2 n2 p2 J, = -- -+-+- (1)

r 2~ a2 b2 d2

where c is the vacuum speed of light, J.!, the relative permeability of the gas and m, nand p are integers defming the resonance mode. Since f-l = I for most gases, it follows from equation (I) that

(2)

where fo is the resonant frequency of the cavity for mode (m, n, p) when it is under vacuum ( i. e. when

& = &0 = 1). The presence of water in the sample will change slightly the dielectric constant; this will induce a

small shift in the resonant frequency given by

/),& = _211fr . (3) & fr

One can see that the increase in our ability to measure a small shift in the resonant frequency of the cavity wiII enhance the precision for detecting small changes in the dielectric constant of the gas and hence to better determine traces of water.

The microwave generation and detection were made with a network analyzer (HP85 lOB); it was one-port connected to the cavity under study using a 2.5 mm coaxial line. Our cavity had its resonant frequency around 2.4 GHz. Taking into account the resolution of the synthesizer driving the network analyzer, the electric noise and the

repeatability of our measurements, we could measure fr with a precision of 12 kHz. We therefore had a resolution

of 10-5 for the dielectric constant.

Figure 1 gives a diagram of the experimental set-up. In order to preve.nt mechanical deformations of the resonant cavity, eight 2 mm holes were drilled in its faces for gas admission and exhaust. The cavity was put in a pressurized stainless steel cell filled with the gas under study. The cell was surrounded by a temperature controlled fluid which achieved a thermal stability of 0.04 K.

- 102 -

I Network Analyzer I Vent

I Thermometer Vs Display I

VI V2

Vent

?V, at T M -to G Cell

Cavity Vacuum Pump Compressor

Thermal I Insulator

'--- Oas

Temperature Controlled Bath

Figure 1. Schematic diagram of the experimental set-up. G: gas circuit, M: microwave circuit, P: pressure gauge, T: thermometer, VI to Vs: valves.

We have made measurements on samples of water contaminated methane. Pure methane and 3 mixtures of different water contents, 600 ppm, 300 ppm, 150 ppm were studied. In order to avoid any water condensation, we measured the dielectric constant at rather low pressure, up to 2.05 MPa. We were able to detect water vapor in methane down to a concentration of 75 ppm at 2 MPa. However, this detection threshold did not meet our objective of being able to detect less than 10 ppm of water; a more sensitive method needed to be developed.

DIFFERENTIAL MICROWAVE METHOD

In order to improve the water detection threshold, we have developed a microwave differential method. Our set-up is built around two microwave cavities connected to a 180° hybrid junction. A schematic diagram of our method is shown in figure 2. A carrier signal coming from a microwave synthesizer is divided in two parts. The first one is mixed to a low-frequency signal coming from a lock-in amplifier. At the output of the mixer, we have an amplitude modulated signal which is then sent at the input of the hybrid junction. The first half of that signal is directed towards point B where part of it is reflected by the measuring cavity. This reflected signal will also be divided in two parts; the one going back to A being absorbed there while the one going towards point D is phase-shifted by 180°. On the other hand, part of the signal reaching point C will also be reflected towards point D but it will not be phase-shifted. If the two cavities are identical, i. e. they are both filled with the same gas at the same temperature and pressure, two signal of equal amplitude but of opposite phase will reach point D; the signal at the output of the hybrid junction will be zero. But, if the gas in the measuring cavity is contaminated, this will change the impedance of that cavity, the reflected signal at B will also change and the two signals reaching point D will not balance out. There will be at the output of the hybrid junction a signal that can be related to the quantity of water present in the contaminated gas. This signal is then amplified to a level sufficiently high to permit it to mix with the carrier which has been phase shifted by an angle~. At the output of the last mixer, we a have a low-frequency signal that is filtered and then received by a lock-in amplifier.

Power

Divider

- 103 -

Figure 2. Block-diagram of the microwave differential method.

At the input of the lock-in amplifier, the signal V will be given by

V = C(coswmt)(LU"cos¢ + ilr"sinip) (4)

where C is a constant depending on the initial amplitude of the microwave signal, the gain of the microwave amplifier and the efficiency of the mixer, wm ' the modulation frequency, ilr' and ilr" are respectively the

difference in the real and imaginary parts of the reflection coefficients of the two cavities. If we adjust the phase shifter to get ¢ = 90° and if we sweep the carrier frequency over a range comparable to the width of the resonance,

the amplitude of the signal (in volts) is

11& Vsignal ~ 10- (5)

&

On the other hand, if il v is the width of the band pass filter of the lock-in amplifier, the noise signal will be given by

~ Vnoise = AVa l1v

where k is the Boltzmann constant, and A the gain of the microwave amplifier. Typically, we will have

Vnoise = 5.3 x 10-10 ~ . This will give us a signal to noise ratio of

Vsignal 2 x 1010 11&

Vnoise = ~ &

(6)

(7)

(8)

We can estimate the concentration of water that can be detected with our method. The virial form of the ClausiusMossoti equation is [2]

(&-1) 2 3 -- = Asp+Bsp +CsP + ... &+2

(9)

where p is the gas density and As, Bs et Cs are respectively the first, second and third dielectric virial coefficients.

As a first approximation, we can keep only the first term in the virial expansion; this permits us to compute the dielectric constant as

(10)

The first dielectric virial coefficient of a water contaminated gas (As) c is a function of the molar fraction of water x

and the first dielectric virial coefficients of the pure gas (As) p and of the water vapor (As)w [2]

- 104 -

(11)

The presence of water in a gas will therefore bring a relative change in the measured dielectric constant given by

Ae ec-ep 3xt{(Ae)w -(Ae)p) -;-= e = 1+3P(Ae)

p

(12)

The fIrst dielectric virial coeffIcient of gases can be derived from molecular properties [2]

N A [ Pb) Ae=- ao+-3eo 3kT

(13)

where ao is the polarizability of the molecule, Po, its permanent dipole moment and eo the permittivity of

vacuum. The polarizability of the water molecule is ao = 1.636 x 10-40 Fm2 while its permanent dipole moment

is Po = 6.188 x 10-30 Cm [3]. These values permit us to fmd

(Ae) = 73.6 x 10-6 m3 / mole water (14)

Water is a polar molecule, therefore its fIrst dielectric virial coeffIcient is a lot larger than the one of non-polar gases like SF6 or CH4• Consequently, a good approximation of equation (12) will be given by

Ae - = 0.089xP (15) e

where P, the pressure of the gaseous sample, is given in MPa and where we have also used the perfect gas equation of state.

From equation (8), we will get

Vsignal 1.8 x 109 xP

Vnoise = & If we want a signal to noise ratio of 10 and if we have a lock-in amplifier bandwidth of 1 Hz, the minimum concentration of water we can detect in a contaminated gaseous sample kept at a pressure of 0.1 MPa is

x = 5.5 x 10-8 = 0.055 ppm.

This is a very good sensitivity.

CONCLUSION

Although our fIrst microwave detection method did not show enough sensitivity to meet our objectives, we see that the differential method offers a lot of promise. With the sensitivity we can expect from that method, it could be used to detect water in CH4 and help prevent formation of hydrates in natural gas pipelines. It could also be useful to measure concentration of water in SF 6, a gaseous dielectric which is adversely affected by the presence of contaminants. Further work is in progress.

ACKNOWLEDGEMENT

We gratefully acknowledge the fmancial support of Hydro-Quebec for this project. We thank Olivier Le Noe for his help.

REFERENCES

[1] R. FHarrington,. Time-Harmonic Electromagnetic Fields, New-York, McGraw-Hill, 1961, p. 156. [2] N. E. Hill, W. E. Vaughan, A. H. Price, and M. Davies, Dielectric Properties and Molecular Behavior, Van

Nostrand, London, 1969, pp. 191-208. [3] C. G. Gray and K. E. Gubbins, Theory of Molecular Fluids, Vol. 1, Clarendon Press, Oxford, 1984, p. 582.

- 105 -

MICROWAVE MOISTURE SENSORS AND THEIR APPLICA nON IN CIVIL ENGINEERING

ABSTRACT

Klaus Kupfer Materialforschungs- und Priifanstalt an der

Hochschule fur Architektur und Bauwesen Weimar Amalienstr. 13 - 99423 Weimar - Germany Fax: +49-3643-564-201

Moisture measurement using microwaves has the advantage of being a non destructive measurement with high accuracy under severe environmental industrial conditions. Disturbances, such as bulk density variations, material grain size or salt content, can be reduced by both suitable data processing of measured values and choice of measuring frequency. In contrast to various applications in the building industry, there are only a few measurement units commercially available. The development of new sensor types, the application of new methods with high accuracy are a necessity for the future.

1 INTRODUCTION

Moisture is a major problem in many branches of the industry now and in the future. The building industry has to solve an immense variety of moisture problems because of different waterbindings in heterogeneous materials. Moisture measurement systems have to determine the exact quantity of the moisture content (MC) within a certain material, in order to control the dosage of water, the quality of products and the reduction of energy within the process. Microwave moisture measuring methods are based on the high value of permittivity of water but also on the utilisation of frequency characteristic material properties such as the relaxation process in polar liquids and lack of dependence on ionic conductivity. They enable simple automatic control of production processes - with high accuracy and sensitivity of measuring devices in severe environmental industrial conditions. Unlike nuclear measurement units, microwave devices are not dangerous and the measuring time is shorter. Dust and water vapour affecting the infra-red measurements adversely, have no influence on the microwave measurement [1]. The development of non-destructive microwave measurements are necessary to solve moisture problems in civil engineering.

2 DEMANDS FOR THE DEVELOPMENT OF MOISTURE MEASUREMENT DEVICES

A density independent moisture measurement is the basis for the quality control of many production processes for building materials such as the moisture measurement of aggregates and fresh concrete, the hydration process of concrete and the quality control of lime brick production. To analyse building defects in brickwork, historical buildings and bridges, moisture measurements at the surface and in the depth of brickwork are absolutely necessary to be connected with the analysis of salt content distributions in the brickwork and on its surface. The calculation of the moisture content, its distribution at present and in the future, the inspection of calculated values by measurement and drying out are the follow up steps. The measurement of moisture on the surface of walls can be done on wide areas with infra-red cameras. Microwave open resonator systems can help to complete these qualitative measurements. They have higher measuring accuracy and can be applied for detecting moisture in special areas of the wall. Probe techniques and radar are used to carry out moisture measurements within walls [2]. In civil engineering there are a lot of moisture measuring problems and applications, but as yet, there are only a few measurement devices available to solve these problems. In the future for the effective application of moisture measurement units the following developments are necessary:

• Sensor types which ensure an effective coupling of the electromagnetic field into the material from one side and which operate with high accuracy (open resonator systems)

• Measurement and evaluation of complex parameters of permittivity • Density, salt content and temperature independent measurements • Application of circular polarized waves to reduce influences of scattering • Development of methods for moisture measurement during drying of ceramic materials • Reduction of electromagnetic energy, radiated in the environment during the measurement • Standardization of measuring methods and measuring errors of moisture measurement

- 106 -

• Investigations of process technology and adaptation of the microwave measurement methods within the technological process.

The introduction ofISO DIN 9002 [3] and the quality certification of production processes are the basis for the necessity to determine quality properties such as moisture content with high accuracy. The moisture measurement devices of the future have to be more user-friendly, more intelligent, reliable and robust under severe environmental conditions but also adaptable to the traditional technological process. It is necessary to develop new custom-designed sensors for different applications. Microwave devices with high accuracy and several specific sensors have to be applied for sequential data processing of measurement. Due to the current state of research and technology in both computer techniques and integrated microwave circuits, such demands are becoming more realistic. Some examples of applications for moisture measurement in civil engineering are presented in the paper.

3 MOISTURE MEASUREMENT IN WALLS

The most present moisture measurement units for walls operate with the method of conductivity. Further developed systems use the real part of permittivity at low and medium frequencies. Disturbances of salt content and couplings of the field into the material influence the accuracy of measurement. Watson was one of the first who introduced microwave instruments for the moisture measurement in walls and bricks. He developed a transmission measurement unit with horn antennas between which the probe was placed and used the attenuation as the value of moisture at frequencies of 3 and 10 GHz [4;5]. Reasons of inaccuracy in a brick wall (> ± 30 %) were disturbances such as the conductivity of salt content and the influence of different structures in the material. A dielectric measurement unit was developed recently for moisture measurements at the surface of walls for different materials. The device is an open resonator system and operates at a frequency of 150 MHz. It has the advantage of the two parameter measurement. The frequency of 150 MHz is decreased proportionaly dependent on moisture and is related to the real part of permittivity. Disturbances like bulk density variations and ionic conductivity influence the quality of resonance line and consequently the bandwidth of the resonator system. The change of bandwidth is proportional to the value of the imaginary part of the permittivity. The equation for the two-parameter measurement method in a resonator is given as follows [6]:

MI_Mo

Q(\V) = 2 fl fo fo -fl

fl

E " ~_r_

E '-I r (1)

The resonator system is developed as a capacitivly loaded quarter wave resonator based on a coax line (Fig. 1). The shortening capacity is performed as an open stray field capacitance. The influence on the stray field by a moist material results in the change of frequency (fo - f l ) but also in the change of bandwidth from Mo to MI (Eq.l). The system shows a high repeatability and was used for recording calibration curves for different building materials i.e. bricks, natural stones, gypsum plates, wood, lime brick stones, mortar layers and concrete. Disturbances of salt content, temperature and density variations on the measurement have been investigated for different materials. The calibration curves are stored in the memory of the microcomputer. The system can be used for the moisture measurement in bulk-materials, paper etc. Problems of investigations are the coupling of the electromagnetic field into the building material and the repeatability of measurements for materials of different producers i.e. different structures and compositions. (Fig.l). A new type which has to be developed will operate at a higher frequency.

4 MOISTURE MEASUREMENT DURING THE CONCRETE PRODUCTION

Changes of moisture in aggregates in the region of 2 to 6% MC reduce the compression strength of concrete up to 25 %. In order to compensate for the losses extra charges of cement up to 120 kg per m3

of concrete have to be taken into account. For the concrete production in Germany of 25 millions m3

- 107 -

per year extra costs up to 150 mio DM would be necessary which can be reduced by application of an accurate moisture measurement. Different moisture sensors were examined for suitability of moisture measurement in the aggregate bin and in the mixer. They were based on open resonator systems with vertical polarized ring antennas and operating frequencies from 0.3 ... 1.2 GHz. Disturbances of the measurement arise from density differences in the aggregate silo or spikes from the blades of the mixer. It is possible to reduce these influences by arrangement of sensors in the flowing material stream and by filtration of output signal in a computer. Accuracies up to ± 0.3% MC could be reached. During the moisture measurement in the mixer different phases of moisture binding and moisture content occur in aggregates, in cement bounded aggregates and fresh concrete. In these phases the possibility exists for moisture control and shortening of mixing time. The different consistencies have to be taken into account. Due to the beginning of the hydration process the time for the reference-method of drying of fresh concrete has to be very short. The methods of microwave drying with simultaneous weighing have been applied as reference methods. The drying is finished when the value of weight remains relatively constant. The device was applied by the coal industry and first used for the reference moisture measurement of fresh concrete.

5 MICROWAVE MOISTURE MEASUREMENTS IN LIME-SAND BRICK PRODUCTION

The accurate determination of the moisture and its correct dosage of water are essential requirements for a high quality of lime-sand bricks. The results of conventional moisture measuring methods are falsified by interferences encountered during the process of carbonation. Measurement with microwave frequencies is absolutely necessary because of the influence of conductivity. Density independence is the basis for these moisture measurements in order to reduce the multitude of disturbances. During the carbonation process of a mixture for lime-sand brick production, the water contained in the sand will be in part chemically bound or evaporated due to the high temperatures created during the lime slaking. A change in the addition of calcium oxide causes changes in the polar properties of calcium hydroxide, but most of all in the lime sand mixture. They change the imaginary part of permittivity and thus falsify attenuation measurements at 3 GHz. The ionic conductivity in the material creates Maxwell-Wagner effects and binding of water molecules to the ions. These interference factors occur the other way round and, depending on the frequency, they may also falsify the real part of complex permittivity and thus the phase measurement too. In the range at 3 GHz, both effects compensate each other; at 9.3 GHz the binding of water molecules dominates. Despite the large number of interference factors encountered in the hydration process, microwave moisture measuring systems are able to achieve high accuracies during the phase measurement at 3 GHz, but also for the measuring units phase and attenuation at 9.3 GHz. A detailed analysis of the material to be examined as well as the process technology are necessary to ensure the optimum application of a microwave moisture measuring system in industrial process [7].

CONCLUSION

The building industry has to solve an immense variety of moisture problems because of different water bindings in heterogeneous materials. The introduction of ISO 9002 and the quality certification of production processes are the basis for the necessity to determine moisture content with high accuracy. Microwave moisture measurements exhibit the advantages of a non-destructive measurement with high accuracy, small influence of ionic conductivity and automatic process control within severe environmental industrial conditions. Disturbances, such as bulk density, material grain size and ionic conductivity, can be reduced by suitable data processing and choice of measuring frequency. Measurement of moisture in the surface of walls was performed with an open resonator system operating at a frequency of 150 MHz and providing of two-parameter measurement. An accurate moisture measurement during concrete production helps to reduce costs for extra charges of cement. Open resonator systems with measuring frequencies from 0.3 to 1.2 GHz have been tested for suitability of moisture measurement in aggregates and in fresh concrete. Disturbances of the measurement are caused by density variations in the aggregate silo and spikes from the blades of the mixer. By specific arrangement of sensors and application of intelligent data processOing accuracies up to ± 0.3 % MC could be reached. The microwave drying with simultaneous weighing was first used for the determination of reference moisture content in fresh concrete. Microwave moisture measuring methods for lime-sand brick production have been also examined. High accuracies in phase measurements at 3- and 9.3 GHz could be achieved, despite the multitude of disturbances, such as temperature, density, ionic conductivity, etc.

- 108 -

A detailed analysis of the moist material as well as the lime-sand brick technology are necessary to ensure the high accuracy of microwave moisture measurement but also the quality of the produced limesand bricks.

REFERENCES

[1] Kraszewski, AW.:Microwave Aquametrie - Needs and Perspectives.IDEE Trans. on MIT Vol. 39(1991)5, pp. 828-835

[2] Rudolph, M.; Schaurich, D.; Wiggenhauser, H.:Feuchteprofilmessungen mit Mikrowellen im Mauerwerk. Feuchtetag 1993 BAM Berlin DGZfP-Berichtsband 40 pp . 44-56

[3] ISO 9002 Quality systems - Model for quality assurance in production, installation and servicing. Quality assurance. Quality management 07.94

[4] Watson, A: The non-destructive measurement of water content by the microwave absorption method. Consil International de Beton Symposium No.3 (1960), pp. 15-16

[5] Watson, A: Measurement and control of moisture content by the microwave absorption method. Build International No.3 (1970), pp. 47-50.

[6] Meyer, W.; Schilz, W.: Microwave measurement of moisture content in process materials.Philips technical Review Vol. 40 (1982) No.4, pp. 112-119

[7] Kupfer, K.; Kupfer, H.: Mikrowellenfeuchtemessungen beim AufbereitungsprozeB der Kalksandsteinherstellung.- Wiss. Zeitschrift der Hochschule fur Architektur und Bauwesen Weimar Vol. B 38(1992) 112, pp. 13-25

Fig. 1 Moisture measurement unit with an open resonator system

- 109 -

MICROWAVE RESONATOR MOISTURE METER

IN THE AUTOMATED SYSTEM OF DUMPING GRAIN BEFORE GRINDING

P.D. Kuharchik, I.A. Titovitsky, A.Ch. Belyachits and N.!. Kourilo

Byelarussian State University, 200050 Minsk, Byelarussia

In the process of making flour from grain, the technological method of additional moistening is used to increase the output of finished products and to improve the flour quality. The major factor of accident-free operation of such enterprises is the accurate control of moisture content and its distribution in grain. In this respect, it is advisable to measure the moisture content of grain on-line continuously, and on the basis of these measurement, precisely moisten the grain according to the specifications. In this paper, a microwave resonator moisture meter for grain is presented, that is designed for operation in the automatic system of dumping grain before milling.

Moisture content of different materials and substances is related to a number of parameters, such as structure and composition, size of particles, etc. Microwave methods of moisture content measurements received the greatest development and practical application due to the high sensitivity of microwave radiation to water, to the possibility of achieving high accuracy in non-contact measurements, and to the integral moisture evaluation according to the volume of the substance under test.

When designing a microwave moisture meter intended for on-line operation in the industrial conditions, it is necessary to ensure the invariability of moisture indications with the changes of grain density, as well as with changes of the external conditions (temperature, humidity, power and frequency fluctuations of the microwave generator, changes of the parameters of electronic components, etc.

The presence of moisture in a substance affects the values of both real and imaginary parts of its permittivity ( = (' - j(". In majority of cases it is impossible to express ( as a function of the moisture content Wand the density of dry substance, p. The analysis of numerous moisture content measurements of not very dense and compressed organic materials, e.g., wool, tobacco, grain, cotton, etc., leads to the conclusion that at microwave frequencies and for constant moisture content, the value of A = «(' -1)/(" is constant as the density changes p = 0.1 - 0.5 gJcm3 [1]. Thus, to ensure the independence of the moisture indications upon the material density, it is necessary to measure two parameters of a material, one of which depends upon (' and the second - upon (". The ratio of these parameters is a function of moisture content only. Such an approach was realized when simultaneously measuring the wave amplitude and phase after it passed through the layer of material [2]. However, it should be noted that the practical implementation of such a device into industrial environment is rather difficult because it requires strict frequency and power stabilization of the microwave generator versus temperature. Introduction of an additional reference channel for the registration of the amplitude and phase drift when external conditions ar~ changing, leads to a considerable complication of the device and to an increase in prIce.

- 110 -

A simpler and more sensitive solution is to apply a resonant moisture content measuring method. Like the amplitude-phase shift method, the resonant method allows the measurement of two parameters simultaneously, related to i' and i", of the substance placed in the resonant cavity. These two are: the change of the natural resonant frequency and the quality factor of the resonator. To carry out the measurements using this method, a small volume of the analysed substance, small when compared to the total resonator volume, is usually used to provide acceptable changes in the quality factor and the resonant frequency, provided the moisture content of the substance changes within required limits. In our case, it was necessary to obtain the integral value of moisture over a quite big volume of the substance. For this purpose a special resonator was developed, being a ring resonator, containing a coupling slit with a dielectric tube through which flows the substance under test. The sensitivity of the resonator is, to some extent, controlled by the width of slit along the wide side of the ring resonator. In this sense, the sensor may be easily adapted for moisture measurement of various materials of different moisture content.

When measuring the moisture content of grain on-line, the sensor is filled with the material at all times, and resonant frequency of the ring resonator depends upon the ambient temperature (thermal expansion of the resonator) which causes moisture content error. The amplitude and frequency instability of the microwave generator and all other elements also causes additional errors. To ensure a stability of the readings, the double-mode regime of the resonator excitation was introduced to the meter. The design is such that, when the generator frequency changes from 4.0 to 4.5 GHz, there are two resonances in the ring. One mode of oscillation does not interact with the material under test, and it is called the fundamental mode. The other mode, interacts with the material and it is called the measuring mode. Since each of the oscillation modes are excited inside the same resonant ring, then all changes of the ambient conditions equally influence all their parameters. The periodic checking parameters of the fundamental mode makes it possible to correct the result of moisture measurement.

Let's designate F 0 and Ao as the resonant frequency and the amplitude of the fundamental mode, respectively, and FI and Al as the resonant frequency and amplitude of the measuring mode. Then the values F = F 1 - F 0 and A = (A~/ Ao)*A I ,

where A~ is the amplitude of the fundamental mode when calibrating the device, do not depend upon the external factors. The parameters F and R = In(l/ A) depend on the moisture content and density of the substance, and their ratio depends only on the moisture content. Then, assuming that the moisture content is a linear function, with a high degree of accuracy we can express it as

F/R=P.W+Q, where W is the moisture content of the substance under test.

To determine the coefficients P and Q, two samples of the grain at different moisture contents WI and W 2' determined by the standard method of drying, were used. When the sensor is filled with grain of moisture content Wll the values FI and RI are measured. Then the procedure is repeated with the sample of moisture content W 2 and the values of F 2 and R2 are measured. The coefficients P and Q are then evaluated by solving two equations with two unknowns:

(1_ F2) P _ .""R,.".I"'--..... Rri2'

- W I -W2 and FI

Q=Il- P .WI · 1

With known coefficients, the arbitrary moisture content calculated from the calibration equation (F _ Q)

W=....:R~pc--

of gram m the line, W, IS

- 111 -

It is well known that the dielectric properties of grain greatly depend on its temperature, because of water present in it. Therefore, in the process of an on-line moisture content measurement, it is necessary to monitor the grain temperature and carry out the corresponding corrections of the measured values of F and R.

The algorithm developed above for on-line moisture content determination in grain requires microprocessor techniques to be used. In the developed instrument, control of all modes is carried out by Siemens microprocessor SAB-C503-LN. To obtain the amplitude-frequency characteristics of each mode of operation, the microprocessor generates the sawtooth voltage to change the frequency of a microwave oscillator. The signal, proportional to the microwave power at the output of the resonator, is measured together with the grain temperature. The values of the resonant frequency and amplitude of the resonant curve are determined for each mode of oscillation, and the calibration algorithm is used for calculation of the grain moisture content and its temperature correction is performed. The output information is displayed on the LED panel and transmitted in the form of a serial numerical code into the computer that controls the sprinklers providing additional moistening of grain.

After the moisture meter calibration, a series of experiments for grain moisture content determination by standard method of drying was carried out to determine the distribution of moisture content in a grain batch. The moisture content of wheat, determined as an average of six dried samples, was 11.5 ± 0.2% moisture. The same batch measured in the calibrated meter gave the moisture content of 11.5 ± 0.3% moisture, as a result of 25 measurements. Thus, the real instrumental error in this series of experiments has not exceeded ± 0.1 % moisture.

The dependence of the moisture content readings on the glassiness of wheat grown in different regions was tested experimentally. For wheat samples of glassiness from 30 to 60%, coming from Belorus, Hungary, Russia and Kazakhstan, the instrument error was less than ± 0.3% moisture, in the range of moisture content between 10 and 17%. Taking into account that the result of drying method was taken only for one sample, and that the practical variation of the moisture content in one sample may reach ± 0.2% moisture, the perfect agreement between results obtained by the standard method and those obtained from the microwave resonator moisture meter has been achieved.

References.

[1] W. Hoppe, W. Meyer and W. Schilz, "Density-independent moisture metering in fibrous materials using a double-cutoff Gunn oscillator", IEEE Trans. on Microwave

Theory and Techniques, vol. MTT-28, no 12, pp. 1449-1452, 1980. [2] A. Kraszewski, "Microwave aquametry - a bibliography", Journal Microwave Power,

vol. 15, no 4, pp. 298-310, 1980.

- 112 -

A LOW COST MICROWAVE SENSOR FOR MOISTURE MEASUREMENT OF TEXTILE WEBS

T. LASRI, D. GLAY, A. MAMOUNI, Y.LEROY Institut d'Eiectronique et de Microeiectronique du Nord IEMN, UlvfR CNRS 9929

Departement Hyperjrequences et Semiconducteurs Domaine Universitaire et Scientifique de Villeneuve d'Ascq Avenue Poincare - B.P. 69 59652 Villeneuve d'Ascq Cedex FRANCE

Tel: (33) 20197938, Fax: (33) 20 19 7896, E-mail Tuami.Lasri@}EMN.Univ-Lillel.jr

Abstract The progress in Microwave technology allows us to think that original and low cost

sensors are now feasible. In this paper we present a sensor devoted to the measurement of the moisture content of materials webs. This sensor is based on the use of a complex correlator associated to a measurement cell. The first attempt is made in hybrid technology at a frequency of 10 GHz.

Introduction The study described here concerns a sensor for non destructive control (NDe) of non

metallic material sheets. The material is inserted in a longitudinal slot in the centre of the broad wall of a rectangular waveguide working in X band. Such a cell has already been used for Microwave heating [1], material control [2], and more particularly for non contact thermometry by Microwave radiometry [3].

In the present work, we are aiming to control the moisture content of a textile web, parameter which has a great influence on the scattering parameters. Obviously, the idea of moisture measurement by Microwave techniques is not recent. However, unlike the previous methods which use, for the most part, an Automatic Network Analyzer we have chosen to develop our own measurement systems. These devices must be transportable, low cost and able to work in a industrial environment with an acceptable precision.

The measurement system The sensor is dedicated to the moisture measurement of material webs and more

particularly for textile webs. So, considering that on one hand the textile webs thickness are, in general, lower than one millimetre, and on the other hand that the dielectric losses are low, a working frequency of ten GigaHertz looks suitable. Note that the position of the sample in the cell is favourable to a electromagnetic test because the material is placed where the electric field is maximum (Fig. 1).

Length

F=lOGHz

Slotted waveguide ;f

Material web ;f

Figure 1 : Measurement cell

- 113 -

The principle we have chosen for the system measurement, close to the six port method [4], is based on the use of a complex correlator (Fig.2).

e2 e1

V1 V2

V3 V4 Figure 2 : Basic scheme of the complex correlator

In this system, particularly simple, two signals el and e2 such as el (t) = El cos 27tft and e2(t) = E2 cos(21tft +<I» are applied to two couplers which are followed by two systems made of a coupler and two quadratic diode detectors. The fIrst system gives two values VIand V 2 which lead to a d.c.signal which can be written as SR = E1E2cos while the second one, where a phase shift of 90° has been added, gives two values V3 and V4 which lead to a d.c. signal SI = E1E2sin. So, for a constant value E1, these signals SR and SI, permit easily to detect the variation of E2 and which lead to the measurement of the scattering parameters.

The layout of this correlator, made on a dielectric substrate (Rogers RT Durotd 6010.8,

Er = 10.8, thickness = 0.635 mm), was created by means of the CAD software Microwave Design System (MDS).

Experimental results The fIrst step of this experimental study was dedicated to well known two-port in order

to valid our measurement system. The results of these experiments indicate that we are able to measure a transmission coeffIcient with a good precision.

The second step of this work concerns the characterisation of material webs. In fact, we can also take advantage of this system to determine the moisture content of textile webs. For this purpose, we have saturated with water a cotton sample in order to follow it's drying in time (at room temperature). The results of the experimentation made at once with the sensor and with an Automatic Network Analyzer (ANA) are presented in figure 3.

IS2l1(indB) 0

-10

. --20 • . • •

-30

0 30

-90

• : Sensor -180 - : ANA

-270

-360 60 90 120 150

<jI" (in degrees)

.-

0 30

-•

60 90

Tim. (minutes) Tim. (minutes)

• : Sensor -:ANA

120 150

Figure 3 : Transmission coefficient (S21) of a sample of cotton initially saturated with water versus time

- 114 -

Conclusion We have shown in this paper the feasibility of building a low cost microwave sensor

devoted to the moisture measurement of material webs. This investigation has been carried out for thin material sheets (about 1 rom) at a frequency of ten GigaHertz. Nevertheless, thicker materials could be studied if we choose a lower working frequency. In fact, if we adopt a frequency of 2.45 GigaHertz, the corresponding standard waveguide size is 86 mm x 43 mm, we can use a slot of about 1 cm without disturb the fundamental mode. Note that we have already developed a low cost system, made in hybrid technology, working at 2.45 GigaHertz able to acquire the S-parameters of a device under test [5].

References [1] Metaxas et aI., Industrial Microwave Heating, Peter Peregrinus Ltd, London, 1983. [2] : R.A. York and R.C. Compton, "An Automated Method for Dielectric Constant

Measurements of Microwave Substrates", Microwave Journal, March 1990. [3] : Y. Leroy, IC. Van de Velde, A Mamouni, B. Meyer, IF. Rochas, IIContactless

Thermometry of a Textile Web by microwave Radiometry", XVI European Microwave Conference - Dublin, 1986.

[4] : G.F. Engen, liThe Six-Port Reflectometer : An Alternative Network Analyser", IEEE Transaction MTT-25, pp 1075-1080, December 1977.

[5] : T.Lasri, B. Dujardin, Y. Leroy, IIA microwave sensor for moisture measurements in solid materialsll,I.E.E. Proceedings-H Vo1.138, N° 5, pp481-483, October 1991.

- 115 -

MICROWAVE INSTRUMENT FOR HIGH WATER CONTENT MEASUREMENT IN PULP AND PAPER INDUSTRY

Pekka Jakkula and Esko Tahkola

Valmet Automation Measurements, Teknologiantie lOB, SF - 90570 Oulu Finland

ABSTRACT

Two microwave methods have been developed for consistency measurement in pulp and paper industry. The first method is based on dielectric wavequide and the second one on the measurement of time delay of microwave propagating through the process pipe. The time delay was determined by using FMCW - techniques. Both methods were found to measure pulp consistency but the first one was sensitive to disturbing factors like flow velocity, pH and wood type. On the other hand the second method was found to be insensitive to most of the disturbing factors and superior compared to shear force and optical methods which are used today for consistency measurement in pulp and paper industry.

INTRODUCTION

Water is the dominating material in paper making process. Water is used to carry wood fibers from pulp mills digesters through bleaching stages to paper mill stock preparation and finally to paper machine wet end. Typical water content in paper making process is over 96 % (consistency less than 4 %). At this level the conventional moisture sensors do not work and the market is still dominated by mechanical shear force sensors. For years suppliers have tried to develope electrical consistency sensor mainly based on optics without wide succes. Optical sensors are sensitive to fiber length, wood species and colour and they have found applications only when consistency is below 1 %. The most important consistency measurement point in pulp and paper process is after the machine chest of paper machine where the consistency is around 3.5 %. After that point consistency cannot anymore be controlled and consistency variations will have direct effect on basis weight and quality of paper. The needed measurement accuracy is 0.01 %. In addition to the tightening quality demand there are some trends in paper making process which have made the need for better consistency sensor even more important. One is the increasing use of recycled fibers. The fiber characteristics of recycled pulp are variating and unknown causing problems both to the mechanical and optical sensors. The other is the increasing use of minerals in stock. Minerals are not seen by the shear force sensors and they disturb optical ones. Microwaves have been used for years to measure water content of different materials. Most common methods are based on attenuation, phase or resonance measurement [1] [2]. This paper discribes two different microwave methods developed for consistency measurement.

MEASUREMENT METHOD 1: DIELECTRIC W A VEQUIDE

The dielectric wavequide has been used to measure the moisture content of dry paper [3]. In that application microwave propagates on the dielectric and also in the paper web as a surface wave and so a high measurement sensitivity is achieved. When this kind of sensor structure is used to measure high water content materials the surface wave will attenuate away and the microwave attenuation is not anymore affected by material's dielectric losses but by its permittivity [4]. This can be clearly seen In Fig 1. where the attenuation of the developed sensor is presented as a function of moisture content. . The sensor structure is shown in Fig. 2 .. The dielectric is alumina and metal is titanium. The diameter of the sensor head is 44 mm. The frequency of the measurement system was choosen to be 10.525 GHz. This frequency was selected to keep the sensor size small and to avoid conductivity effects which will dissappear when frequency is over 2 GHz [5]. The method was first tested with good results in laboratory by using water and materials which are dissoluble into water ego sugar. The second step was to test the method with pulp in mill scale pilot plant where it was possible to control consistency from 0 to 15 %. The results are shown in Fig. 3. The results in Fig. 3. clearly show that the measurement is strongly dependent on wood species. This was not expected because the dielectric constants of different woods are very close to each other [6]. The problem was

- 116 -

studied by using a similar sensor head made of transparent glass and a video camera. It was found out that the reason to variations was the different behaviour of fibers on the sensor surface and so the consistency in the vicinity of the sensor was affected by pulp type, flow velocity, pH and many more variables. The conclusion was that in measurement of pulp consistency the measurement volume must be so big that the surface phenomenas don't have effect on the measurement result. The developed sensor can be used in applications where the components of the mixtures are dissolved and do not separate from each other. One example is fertilizer slurry.

Dielectric Sensor Response

I Low water content. 1\ Losses in m.edium dominate i \ the attenuatIOn.

!\~ ; \ High water content. I

Mismatch between I

0%

medium and wavequide I dominate.

Water Content 100%

Fig. 1. Response of the dielectric wavequide as a function of water content.

14

12

10 -:::!! ~ 8 .. 0 I/) 6 c Q) t/) 4

2

0

0 2 4

I i

L

/" Metal body

Dielectric wavequide

Fig. 2. Structure of dielectric wavequide sensor.

I t I x • CTMP I

I • Ix • Unbleached GW I I • .+

I .~I x Eucalyptus I X I

I x Bleached Pine :I.: I

I ~ Bleached Birch ~x

• Bleached GW

I + SAP

6 8 10 12 14

Laboratory (%)

Fig.3. Test results of dielectric wavequide sensor with different pulp types.

MEASUREMENT METHOD 2: FMCW - TECHNIQUES

FMCW - techniques have been used in short range radars where the distance is measured and the medium where microwave propagates is assumed to have stable moisture [7]. When using FMCW techniques in moisture measurement the distance of antennas is constant and the medium where microwave propagates has an unknown moisture content [8]. The velocity of microwave in lossy medium can be calculated if mediums permittivity is known [I]. The velocity is expressed in equation (I).

- 117 -

If material losses are low the equation (1) can be simply expressed as

v=c/...J--;:

where c is the speed of light.

Oscillator

IF

Transm itter Amplifier Switches 0 nReference

~Channel o

(1)

(2)

Transm itting Antenna

Process Pipe

Receiving ~ Antenna

Receiver Amplifier

Fig. 4. FMCW measurement arrangement.

~I-----'

The oscillator sweeps linearly a frequency band B in time T. If the length of the cable from oscillator to the LO channel of the mixer is adjusted so that the delay of microwave signal is the same as the delay in measurement channel without delay between antennas the IF frequency of the mixer will be ~f caused by microwave delay between antennas. ~f can be calculated like in radar application [9].

(3)

~'t is the time delay in measured material between antennas and it can be calculated from equation ~'t = d / v, where d is the distance between antennas. Using equation (2) we get

M = d * B * ...J7/ c * T r (4)

This equation gives the relationship between the output frequency of mixer and relative dielectric constant and thus the moisture content of the medium between measurement antennas. Based on this equation the consistency measurement instrument was developed. The frequency band B was selected to be 800 MHz, time T 10 ms and distance of antennas from 70 mm to 179 mm depending on the process pipe size. By using these values the typical IF frequency can be calculated to be about 300 Hz. To make the measurement easier the frequency can be made higher by increasing the length of the measurement cable. To compensate the drift of the components a reference channel which can be switched to the measurement instead of the measurement channel is used. Reference channel is realized by using a coaxial cable where the delay is stable and independent of temperature. The sensitivity of the measurement system can be evaluated by changing the length of the measurement channel by a known length. This is done in the production of the instrument and the deviation of the measured values gives indication how identical the products can be. The deviation of the sensitivity of the produced 250 measurement systems is 1.9 % which means an absolute error of 0.19 % in consistency if consistency change is 10 %. To make the measurement sensitivity more accurate each instrument has a special calibration number which is given to the instrument during start-up. Due to the linearity and accuracy of the instrument the calibration can be made by single point. This is extremely important when the process is stable and measured and laboratory values will form a cluster of points with no correlation. For example after the machine chest it is not possible to change the consistency for sensor calibration purposes.

- 118 -

The measurement system was evaluated first by electrical means. This was done by replacing antennas by an adjustable delay line. The delay of the line was changed 250 ps by lOps steps. The measured deviation was less than I ps which is also the accuracy of the adjustable delay line. The system was tested in the same pilot plant as method I. The results of these tests can be seen in Fig.5.

14

12

10 -~ ~ 8 ... 0 If) 6 c Q) If) 4

2

0

l

I i I I I

I !

I I

i ! i i .a< !

" , 1 Xl I i I

I : x"f i i 1 •• - ! ! I ~ ..

I I . • i ~' r----.~I i I

I i

o 2 4 6 8

Laboratory (%)

Fig. 5. Test results of FMCW sensor.

i I

I )(. ! i I

i I ,

i I

I 10

I

I

I : I

I I I

i I

i 12 14

I. Unbleached Pine (%) i I .. Bleached Pine (%) ! i Bleached Birch (%) I

Ix GW

1)( CTMP

I_ Eucalyptus (%)

It can be seen that the method is independent of wood type (pine, birch, eucalyptus), colour (bleached / unbleached) and the pulping process (mechanical pulps (GW and CTMP) and chemical pulps). Later it was found that the method was also independent of fiber lenght, freeness, colour, pH (when pH = 2.5 ... 11.5) and flow. The air had effect to the measurement but these difficulties were solved by using higher than 1.5 bar pressure in the process. The pressure will make air to dissolve into the water and dissolved air has only a neglible effect on the measurement. The stability of the system is proven to be very good. There are installations which have not been recalibrated during two years after the start-up.

CONCLUSION

It's been demonstrated that microwaves can be used to solve measurement problems in pulp and paper industry. The developed microwave consistency instrument based on FMCW techniques has remarkably better measurement capabilities than the conventional shear force sensors and it has already been used succesfully worldwide in pulp and paper industry.

REFERENCES

(1] A. Kraszewski, "Microwave Instrumentation for Moisture Content Measurement". Journal of Microwave Power, vol. 8, no. 3/4, pp. 323 - 335,1973.

[2] E. Nyfors, P. Vainikainen, Industrial Microwave Sensors. ARTECH HOUSE, INC.,1989. [3] S.T.Wiles, "Microwave Moisture Content Measuring Apparatus".British Patent Nr. 1447896, 1976. [4] P. Jakkula, "Method and Apparatus for Measuring the Moisture Content or Dry - Matter Content of

Materials using a Microwave Dielectric Wavequide". US Patent 4,755,743. July 5, 1988. [5] A. Klein, "Untersuchung der dielectrischen Eigenschaften feuchter Steinkohle im Hinblick auf die

Anwendbarkeit des Mikrowellenverfahrens zur Wassergehaltsbestimmung". Doctor Thesis 18th Juli 1978. Technische Hochschule Aachen.

[6] S. Boutros, A. Hanna, "Some Electrical Properties of Wood Pulp". Journal of Polymer Science: Polymer Chemistry Edition, vol.16, 1143 ... 1448. 1978

[7] P. Jakkula, P. Ylinen, M. Tiuri, " Measurement of Ice and Frost Thicknesses with a FMCW Radar". 10th European Microwave Conference, Warsaw 1980.

[8] P. Jakkula, E. Tahkola, "Method and Apparatus for Determining the Moisture Content ofa Material". US Patent Nr. 5,315,258, May 24, 1994

[9] M. I. Skolnik, "Introduction to Radar Systems", 2nd Edition, New York, McGraw - Hill, 1980.

- 119 -

A NEW APPLICATION OF WLAN-CONCEPT:

COMPLEX PERMITTIVITY MONITORING OF LARGE-SIZED COMPOSITE BOARDS

Ferenc VOLGYI

Technical University of Budapest, Dept. of Microwave Te1ecomm.,Goldmann Gy.-ter 3

H-lll1 Budapest,

and

HUNGARY

ABSTRACT

Balazs ZOMBORI

University of Forestry and Wood Science, Dept. of Panel Products, Bajcsy-Zs. u. 4

H-9400 Sopron,

Monitoring of the complex permittivity of large-sized composite boards has economic benefits. This paper describes a brand-new application of wireless local area networks using modulated backscatter technology for these measurements. The suggested free-space, reflection / double-transmission type, two parameter (attenuation and phase) basic measurement setup, which is reconfigurable, has several advantages over other earlier microwave methods. Recalling some basic equations, the calculation of dry wood basic weight and absolute moisture content, also are shown. Low cost MMICs and self-designed microstrip antennas and passive detector / backscatters are used in the instrumentation for the realization of the concept, mentioned above. Laboratory measurements and the results of on-line real-time industrial measurements will be presented.

QUALITY PROBLEMS IN THE PARTICLEBOARD PRODUCTION

The mechanical (MOE, MOR) and physical (MC, swelling) properties of particleboards are assessed by destructive standardised experiments on specimens taken from the boards in every hour. There is no information about the quality of the boards between two successive measurements. The continuous production line, in spite of the severe controll of the constituents and the process, might produce substantial amount lower grade particleboards within that hour. These boards are downgraded or rechipped again, causing a significant deficit to the company annually.

The solution is the continuous quality monitoring of the boards by a nondestructive method. The uneven density and moisture content distribution, voids and fissures, the uneven spreading of the glue on the particles and other less important factors have great effect on the quality of the boards. The majority of these factors can be related to the change of the properties of microwaves as it penetrates through the board.

INTRODUCTION

The complex permittivity measurements of dielectric substrates (composite boards, building materials, grains, etc)at microwave frequencies are suitable for the determination of moisture content, density and quality of these materials. Several measurement techniques [1] are used in practice, but the free-space techniques have many attractive features, viz. nondestructive and contactless and do not require special sample preparation [2]. During manufacture, or drying of wet substances, automatization is possible, by applying integrated microwave moisture sensors [3].

There are well known methods for the moisture content measurement and in-process control of composite boards. A set-up for measuring changes in the average volume of wooden chips for particle board manufacturing was described in [4]. The measurement at 8.3 GHz is based on a dual - polarization forward scattering measurement combined with the free-space measurement of phase and attenuation. A sensor array at the frequency of215 MHz was used in [5], for measurement of the thickness of 1.8 m broad particle mat. A stripline resonator array at 380 MHz was used for sorting the veneer sheets according

CABLE.

,-----'"'----..... Cl-1 MICROWAVE T/R UioJlT C 1'I.: ... s. FI<~D STATION: FXS)

SHAPE.~ BEAM RAJ). c.H .. Il.Ac.r. OF TIR ANTI!.N"AS -_..,.-

~ c.oMHOIJ IRI<AI>IATU 7? Iill882>j CJI.OSS $':OTIO>l (IT) iI SPEED oF BOJO.2l)

I

... /_---~ ...

--W-----i

Me"!. MOISTUR.E!. SUBSTANCe: (C.OHPOSITI!. BO""~1

- 120 -

to moisture and density. The use of a two-parameter (attenuation and phase) single frequency (10 GHz) freespace measurement system with homodyne microwave receiver was described in the paper [7]. We must mention the patent [8], which is related to a special measuring device containing two antennas and a diode array.

Microstrip antennas sometimes are used at the free-space measure-ments, for low power levels [9] and for drying [10], too. Good results are available with circular polarization [3]. A multichannel system which operates with six antennas (Moisture Analyzer of Solids) is mentioned in [11].

Some instruments have built-in microprocessors for control, but the sensors have special forms and they are not reconfigurable. This would'be highly important at the monitoring of large-sized wet substances, where a new application of wireless local area network (WLAN) concept [12] gives unlimited possibilities. The modulated backscatter technology offers inexpensive solutions for building low data rate wireless links, viz. automatic vechile identifier (A VI), road transport information (RTI) system, RF Tag, Smart Card, radio frequency identification (RF/ID), electronic shelf label (ESL) as written in [13], electronic retail systems. But we have no information about the application of this idea for measurement, especially the relation of the electromagnetic wave interaction with water and moist substances. Now we would like to call your attention to this. The realization of our suggested instrumentation is supported by the arsenal of low cost monolithic microwave integrated circuits [14], [15].

The objectives of this paper are: 11 to introduce a brand-new basic measurement set-up for measurement of attenuation and phase of the electromagnetic wave reflected / double-transmitted through moist substance, 2/ to show a microwave circuitry realization of this idea, and 3/ to give some measurement results.

BASIC MEASUREMENT SETUP

The basic measurement set-up using a new application of WLAN-concept is shown in Figure 1. This is a free-space, reflection / double transmission system, in which the attenuation (M) and phase (LicI» is measured by the receiver.

The microwave transmitter (Tx), which is working in one of the industrial-scientific-medical (ISM) frequency bands, radiates continuous wave (CW) during the measurement. After passing through the measured moist substance, this wave is reflected by one of the passive detector / backscatters (PDBs). The reflected wave reaches the microwave receiver (Rx), and from the measured data (aA,LicI» the relative complex permittivity (e* = e'-je"), moisture content (MC), wet and dry densities, etc. can be calculated.

The PDBs, having individual codes, are interrogated (Downlink) successively (in time share mode) by the transmitter using on-off keying (OOK) modulation. The matched detector of PDB demodulates the signal and transfers the data the to the digital circuits of PDB. The radar cross section (RCS or cr) of the PDB is changed by a frequency shift keying (FSK) encoder and switch driver (10.7 MHz) so that the backscattered (Uplink) signal from the PDB is binary-phase-shift-keying (BPSK) modulated and ultimately detected by the receiver antenna of the fixed station (FXS). The single-diode microwave circuit of the PDB offer simplicity, low cost and small dimensions. The distance (d) between the matchbox-like PDBs may be small, without perturbing each other, because only one is activated in a given time-slot, the others are passive and matched. The measure of the irradiated common volume (V), from which the information is gained, depends on the distance (D) and depends on the radiation pattern of PDB.

Circulary polarized microstrip antennas are used in the system, eliminating the disturbing effects of the reflected waves from the air-slab interface. The circular polarization of a wave reflected from an odd-bounce reflector surface is opposite from that of the incident wave (for example, RHCP incident wave yields LHCP reflected wave and vice versa), thereby the odd-bounce reflected waves will not appear at the receiver in Figure 1, and the reflected wave from the bottom side of the slab will not appear at the PDB. The TxIRx antennas have shaped-beam radiation patterns. At the measurement of travelling board, sufficient to irradiate only the array of PDBs, using a fan-beam. Broad-beams are used in the case of stopping slabs, when many PDBs are taken on the area under slabs.

The personal computer (PC) of the measurement system is in a common network (Ethernet) with other computers, and the data communication between fix stations and PC is transferred by the measurement control (MeT) unit in Figure 1.

The calibration of the computer controlled complex permittivity monitoring system is executed without moist substance, or using standard slabs with different moisture content and quality.

BASIC EQUATIONS

1/ Electrot.TIagnetic model and BPSK modulation Set off the radar equation, using the designations of Figure 1, the free-space attenuation is:

A.f=rt/Pr .. =(9:rc)lR4/(h~6'G2.rLM) (1)) R=(H+T+l»/sin.'" (2.)

where Pt is the transmitted power, Pr is the received power, Ao is the free-space wavelenght, cr is the radar cross section of the PDB, G is the gain of Tx and Rx antennas, r is the reflection coefficient of RF-diode circuit in PDB, M is the BPSK modulation rate. Using eq.(l), the calculated received power is Pr = -62.2 dBm,

- 121

supposing: H= 2 m, T= 2 cm, D= 10 cm, S= 60°, Pt= 100 mW,"A.o= 5.17 cm, 0= -20 dBm2, G= 12 dB,

M= -10 dB. The received power is suitably high for good-quality signal processing. Our measurement system measures the additional attenuation (M) and phase delay (.1<1» of the

electromagnetic wave after it passed throught twice the wet substance. In the case of not detailed conditions, by employing the results from [2] the relative dielectric constant and loss factor of the medium are as follows:

£.1 =' (~+ 2.e MY· (3)- ~I ~ re ~ t:.A (4) . 2.T 3Go) 2.'1.3 2.T

The PDB is a non-ideal BPSK modulator which has amplitude and phase error. In its ON and OFF states the reflection coefficients are: . ~ :: II; I e1f.. j G. = (G. I ejf2. (5)

Introducing the amplitude and phase errors: Dr"'I';l/lr2..li 1f11>lr'Z.l Llf=lf''\-'f2.+igO· (b)

The bit error rate for Uplink communication is: 13ER.::>~[ e\-rc(y%.) 1- u{cN o/tJ.,. ~r cos,c,'{')J ('f)

It was shown that in order to maintain the signal-to-noise ratio (ElNo) of 10.5 dB to 11 dB (then: BER-I0~) the amplitude error that can be tolerated is approximately 0.5 dB and the phase error that can be tolerated is 100

21 Linear intetpretive model for composite boards As it was mentioned in [7], both the insertion loss (M) and phase (.1<1» are highly linear with the mass of

the wood and of the moisture through which the electromagnetic waves must propagate: . D.A'" D..t ma -t- "2 tnw j L:><p= Cil Mel -t a4 tr'Iw (IS)

where a1-4 are calibration constants, md is the dry wood basis weight [g/cm2] and mw is the basis weigh of the

contained water. From measured values, the absolute (dry basis) fractional moisture content MC, and md, mw

are: Me.'" ~ = o.~4A- C\1 D ¢. tn ... Q4 AA - o.%.A¢. In: a3 DA- Q~Ll4> (9) Me:! Q2.C.rp-~L\.A J G{ Q,Q4- Ciaa~) . \At ~a3- Q,1l9

Because of the noticeable breakpoint in M at about MC= 6%, two piecewise linear models are given in [7]. From the measured temperature (t) dependencies of M and .1<1>, the calibration constants a l and a3 are linear with t, while 3z and a4 have definite quadratic behaviours, in the temperature range of 25 < t < 100°C.

MICROWAVE CIRCUITRY

The microwave elements of the basic measurement arrangement, shown in Figure 1 can be set-up in different ways. It seems feasible to use low cost monolithic microwave integrated circuits (MMIC) in the critical places [14]. The total area of the $hown 5.8 GHz transmitter and receiver chips is 7.2 mm2

, the total supply current is 210 rnA from a single 4V power supply, when the output power is +14 dBm. From these, completed with other elements (microwave antennas, hybrid low noise amplifier, power amplifier, dielectric resonator etc.) the fix station can be assembled.

For the performance of the PCB functions suitable MMIC also can be found [14] (Reflective PSK Modulator, OOK Demodulator). The size of the chip is 1 mm2

, the current consumption in transit mode is 140/200 J,JA. When there is no communication, the circuit is in a stand-by mode to reduce the power consumption. The typical stand-by current consumption is 0.5 J,JA.

The Figure 2. presents the block-diagram of a fix station (FXS) constituted from hybrid microwave circuits. It needs no more detailed explanation, because of the description given in the p.revious chapter.

Our microstrip antenna array, designed for the FXS, can be seen in Figure 3. It has sharper radiation pattern . towards the direction of the left-hand circular pohmzed (LHCP) circular shaped radiating elements, according to the requirements of the basic measurement set-up shown in Figure 1. The size of the antenna is: 105x 45x 2.4 mml. The self-designed PDB is presented in Figure 4. Its antenna is identical with one element of the previously mentioned array. The middle of the patch is grounded (RF and DC short circuit), the feedings of the patch are in the impedance of 100 Ohm points, and the 900 phase difference is produced by the microstrip

F{g.2.

ring-like branch-line hybrid, which is matched using ~~U~~6r 109 Ohm chip resistor at the radial sector. ~~ Isolator

designated matching circuit is used for· optimal (optional)

matching of the impedance of RF-diode in backscatter OOK

mode. A low-pass filter is connected to the diode and modulator

the in/out point is used as the 10.7 MHz (BPSK) input Directional Coupler

and DC-bias, and as the OOK detected signal output. Power

The substrate material is Duroid (D-5880), with the Ampl1fier

thicknesses of 1.6 mm (antenna) and 0.8 mm Trans.

(microstrip circuit), respectively. The size of PDB is: ~~~~~~SA) 40x 32x 2.4 mm3

• The complete PDB has been designed with the aid of a microwave CAD (MMICAD, Optotek Ltd).

+2OdBm (100 mW)

OOK or CW

"\.

Cable

,I npSl(

IF-in

IF-ampl.

~andpa9S Filter

Antenna (LHCP-liSA)

EXPERIMENTAL RESULTS

11 Measurement of microwave circuits. The measurement results of the microstrip antenna amiy (shown in Fig. 3) are: the resonant frequency is 5.8 GHz, bandwidth is 130 MHz, the 3 dB beamwidth in array direction is 22°, petpendiculary to the array the beamwidth is 82°, the. gain of the array is 12.5°.

The measurement results for the PDB are: gain of the patch is 6.8 dB, 3 dB beamwith is 82°,

- 122 -

. Fi9·"3 Fi9A

~::;;:U::::I""" ~:;::I: I· q,i

. the return loss of OOK detector is 14.5 dB, tangential signal sensitivity is -54 dBm, the reflection loss of the BPSK modulator is 1.8 dB, amplitude error is 0.3 dB, phase error is 4.3°, monostatic radar cross section (RCS) is -23.1 dBm2

, the common irradiated cross section is 2.5 dm2 (supossing: D= 9 em, -3 dB points). 2/ Measurement of particle board-. The development of the monitoring instrument is in progress.

Measurements are designed according to the next program: 11 single transmission measurements of different composite boards using a vector network analyzer, 2/ modulated backscatter (MBS) measurements in the laboratory, 3/ MBS system measurements in a particleboard company. These measurement results will be given in the final version of the manuscript and/or on the Workshop.

CONCLUSIONS

The free-space, reflection / double-transmission microwave measurement system using the WLAN-concept and modulated backscatter technology, shows considerable promise for on-line measurement of wood products. The monitored and measured quantities are: complex permittivity, moisture content, wet and dry densities, etc. The measurements are supported by an effective software. Mass produced monolithic microwave integrated circuits (MMICs) are used in the instrumentation, reducing cost and size and increasing reliability.

REFERENCES

[1] A.W. Kraszewski, "Microwave aquametry-needs and perspectives", IEEE Trans. on MTT, Vo1.39. No.5, pp.828-835, May 1991.

[2] A.W. Kraszewski, S. Trabelsi and S.O. Nelson, "Grain permittivity measurements in free space", 25th European Microwave Conference (EuMC), Bologna, Italy, 4-7 Sept.1995, pp.840-844

[3] F. VOlgyi, "Integrated microwave moisture sensors for automatic process control", Workshop on Electromagnetic Wave Interaction with Water and Moist Substances, in conjunction with the 1993 IEEE MTT-S International Microwave Symposium, Atlanta, Georgia, Jun 14, 1993, pp39-44

[4] A.P. Toropainen, "Set-up for measuring changes in the average size of wooden chips by microwave scattering", 24th EuMC, Cannes, Sept. 1994, pp1203-1208

[5] P. Vainikainen and E. Nyfors, "Sensor for measuring the mass per unit area of a dielectric layer: results of using an array of sensors in a particle board factory", 15th EuMC, Paris, Sept. 1985, pp901-905

[6] E. Nyfors, M. Fischer and P. Vainikainen, "On the permittivity of wood and the measurement of moisture and mass per area in veneer sheets", 1993 IEEE MTT-S Workshop WSMJ, pp79-80

[7] RJ. King and J.C. B.asuel, "Measurement of basis weight and moisture content of composite boards using microwaves", 8th Int.Symp. on Nondestructive Testing of Wood, Sept.l991, Vancouver, WA, pp21-31

[8] P. Berthaud and J.C. Bolomey, "Device for measuring, at a plurality of points, the microwave field diffracted by an object", U.S. Patent No. 5.128.621, Filed: Apr. 20, 1988

[9] F. VOlgyi, "Versatile microwave moisture sensor", SBMO'89 Int Microwave Symp., Brazil, Sao Paulo, 24-27 July 1989, Proc. Vol-II, pp456-462

[10] F. VOlgyi, "Microstrip antenna array application for microwave heating", 23th EuMC, Madrid, Spain, 6-9 Sept. 1993, pp412-415

[11] K. Kupfer, "Moisture measurement and its application in the building industry", MJ'94 Microwaves, London 1994, pp88-93

[12] S. Meyer, J. Guena, J.C. Leost, E. Penard and M. Goloubkoff, "A new concept of LANs: passive microwave links hooked onto a fiber optic backbone", IEEE MTT-S Digest, ppI549-1552, 1993

[13] M.V. Schneider, C. Tran and R Trambarulo, "Computer-aided design of modulated backscatt~r microwave modules", 24th EuMC, Cannes, 5-8 Sept. 1994, pp1745-1749 '

[14] M. Camiade, V. Serru, J.P. Brandeau and M. Parisot, "Low cost GaAs MMICs for 5.8GHz short range communications", MJ'94 Microwaves, Conf.Proc., ppI47-150, London 1994

[15] RR Buted, "Zero bias detector diodes for the RFIID market", HP-Journal, pp94-98, December 1995

- 123 -

Guided Microwave Spectrometry for On-line Moisture Measurement of Flowable Materials

Dr. Buford Randall Jean, Epsilon Industrial Inc.

Guided Microwave Spectrometry (GMS) is a new approach for measuring the moisture content offlowable materials, or more generally, for measuring any process parameter that can be inferred from a knowledge of a material's complex permittivity [1]. The material under test is placed in or caused to flow through and completely fill a measurement chamber that acts as a section of waveguide. A scalar measurement of the transmission response of the waveguide is made over a broad band of frequencies that includes frequencies below cutoff. Since the cutoff characteristic is governed by the dimensions of the waveguide and by the complex permittivity of the material filling the guide, the spectral response of the filled guide allows the dielectric properties of the material to be determined. By including frequencies below cutoff, it is possible to separate the real and imaginary parts of the permittivity without having to measure phase.

In typical practice, the permittivity is estimated by assuming a frequency response model for the material's dielectric behavior. Good results have been obtained by assuming that the material can be modeled by a Debye response over the frequency range of the measurement system. The signal processing technique seeks to fit the theoretical attenuation equation to the data by varying the model parameters. In the case of a Debye model, the parameters are the dc value of the dielectric constant E:., the dc value of the conductivity cr, and the molecular relaxation time 'to The high frequency limiting value of the dielectric constant is approximated as a function OfEs to simplify the curve fitting process. With the GMS technique, multiple component analysis becomes possible. For example, it is possible to make independent measurements ofthe fat, protein and moisture in processed meat or to measure the efficiency of a process that converts phenol into diphenyl oxide and water by separately measuring all three constituents.

The cutoff response ofthe waveguide dominates its transmission characteristic. This fact greatly simplifies the measurement task. Because the response is measured over a wide bandwidth, the curve fitting technique mitigates the errors that can be introduced by reflections in the various system components and by inhomogeneities that may exist in the process material. When installed in process piping, the GMS measurement system is also largely unaffected by upstream and downstream disturbances such as rotating augers, opening or closing valves, operating pumps, etc.

The transmission response of a typical GMS measurement cell can be analyzed as a parallel plate waveguide. Figure 1 illustrates a measurement cell that has a cross sectional area equal to that of a three inch diameter pipe. The overall length of the cell is approximately 18 inches and can be supplied with a variety of end connections. The cross section has transitions from circular to rectangular at each end, while the area is held constant. Coupling loops in the narrow walls of the rectangular section launch and detect a TEIO wave whose electric field is parallel to the long axis of the cell. Each loop is encased in a dielectric window whose inner surface is flush with the inside wall of the cell. Several options are available for sealing the windows. A clean-in-place gasket configuration is available for sanitary applications, and various window material and o-ring choices are available for meeting a wide range of temperature, pressure, and chemical compatibility requirements.

For perfectly conducting parallel plates filled with a loss less dielectric, the cutoff frequency is defmed as the frequency for which the propagation constant becomes zero. For frequencies below this value, the propagation constant is imaginary and no propagation can occur. For waveguides filled with real dielectrics, true cutoff does not occur because of the losses in the dielectric [2]. However, there is an abrupt transition in the propagation constant near the frequency for which the lossless cutoff condition would occur. This cutoff transition region yields information about both the real and imaginary parts of the complex permittivity. The location in frequency of the transition region is governed primarily by the real part of the permittivity, while the derivative or slope of the transition depends upon the loss factor. By examining frequencies farther out into the passband region of the

- 124 -

waveguide response, it is possible to distinguish between losses caused by the material's conductivity and losses caused by molecular relaxation effects.

The propagation constant y for a TE wave in a parallel plate waveguide filled with a lossy dielectric is

I

2 ,mn . 2 "

[ ()

2 ]2 Y( ffi) = ffi JloEoE - -;- - Jffi JloEoE (1)

where co is the radian frequency, !Io is the penneability of free space, Eo is the pennittivity of free space, E' is the real part of the relative pennittivity, E" is the imaginary part of the relative pennittivity, a is the spacing of the waveguide plates, and m is the order of the propagation mode.

If the transmission response is displayed in log magnitude fonn (dB), then the shape of the spectrum is that of the imaginary part ofy. Figure 2 shows the change in a typical spectral response for the conditions in which the values of cr, E., and't are increased one at a time from the initial values that produced the unlabeled curve. The labels on the remaining curves show which variable has been changed to produce that curve. The curves shown assume a cutoff dimension of I.S75 inches and a transmission path length of 3.945 inches. The dielectric constant is varied between 60 and SO. Conductivity is varied from 2000 to 3000 !IS/cm, and the molecular relaxation time, initially set at 6 ps, is changed to 14. It should be noted that these are very large parameter changes and were selected to make the plots easily distinguishable. Much smaller changes can be resolved by the GMS technique. The cutoff effect introduces a degree of orthogonality into the measurement of these variables that allows the real and imaginary parts of the pennittivity to be separated with only scalar measurements of waveguide attenuation.

Figure 3 shows real data from a measurement cell filled with tap water near room temperature taken over the frequency range from 200 to 1600 MHz. The vertical axis is in a log scale convenient for the microcontroller software used to collect and process the data. The dashed smooth curve in Figure 3 is the computed curve fit result. Parameter values for the curve are approximately 79.5 for Es" 3600 !IS/cm for cr, and 14.S ps for 'to The 't result differs significantly from the expected value, indicating a higher dielectric loss than should occur at these frequencies. It is believed that the errors are primarily due to the fact that the model does not include losses in the stainless steel walls of the guide. Tradeoffs have been made in the processing algorithm to disregard such factors that do not impact the operation of the sensor in order to achieve benefits in speed. Repeatability and sensitivity of the reading to small changes in parameter values of the process material are the attributes that make the instrument valuable for on-line monitoring and control. Typically, the update time between measurements that can be achieved by the current processor ranges from about 5 to 15 seconds. The update time depends to some degree upon the expected variability in the process stream and on choices made in spectral and output averaging. Work is in progress to further streamline the signal processing algorithm and to identify cost-effective processors that can do the math more quickly.

The model includes the frequency response of the coupling loops, and the signal strength measurements are corrected for cable losses using an extemal calibration procedure that substitutes a fixed attenuator in place of the measurement cell. An internal calibration is perfonned on every measurement at every frequency to compensate for power and gain variations inside the transceiver module. The measurement signal is generated by a synthesizer to assure frequency accuracy. Figure 4 is a plot of typical results for the measurement of moisture variation in hand lotion. The GMS approach can measure from 0 to 100 percent moisture in most flowable materials. The measurement accuracy depends upon the electrical properties of the material, but is typically better than 0.1 %

Operating near the cutoff frequency of the measurement cell offers many advantages that should be noted. First of all, the wavelength inside the cell for the fundamental mode is always larger than the cutoff dimension of the cell. Thus particle scattering effects that are encountered in quasi-free-space measurements are virtually eliminated. In most situations, the cutoff frequency of the measurement cell is lower than that of the process piping connected to the cell. Since the signal energy cannot leave the cell over the critical measurement frequency range,

- 125 -

upstream and downstream disturbances introduced by other process equipment cannot interfere with the microwave spectrum measurement. In addition, there are no moving parts and nothing protrudes into the process stream.

The frequency range of the microwave source and the pennittivity of the process material constrain the dimensions of the measurement cell. The three inch cross section cell shown in Figure 1 will accommodate materials falling within the pennittivity range from that of air to pure water when coupled to a synthesizer system that spans from 200 MHz to 3.2 GHz. Iflarger cross sections are required and the dielectric constant of the process is high, then higher order modes can be used to raise the measurement cutoff frequency into the range of the synthesizer. Dual loops that are driven out of phase easily produce the TE20 mode and suppress the fundamental. For some high loss materials, a much smaller cell is needed to allow sufficient energy to pass through the measurement zone. Fortunately, most high loss materials also have a high dielectric constant (e.g. salt water) and the smaller cross section of the cell does not shift the cutoff frequency out of the operating range of the measurement system.

Calibration of the system for an industrial process is best perfonned on-line. Most often, a simple multiple linear regression against the three electrical variables plus temperature yields excellent perfonnance over the expected range of process conditions. If the process cannot be easily manipulated over the full range of conditions for calibration purposes, then raw spectral data can be recorded over a period of time and processed off-line when enough calibration points have been collected. Feasibility evaluation can be perfonned in the laboratory if samples of the process material can be obtained.

For applications that deal only with binary mixtures, the signal processing can be simplified and the speed of response improved. A simple linear regression on the rise region of the log magnitude spectrum can be used to estimate a cutoff frequency that will correlate to moisture content. Measurement update time of less than a second can thus be provided with the existing processor design.

The measurement of moisture in granular materials is an important application domain for GMS technology. A granular material measurement typicallY involves at least three components: the solid material, water, and air. Since the GMS approach can separate the effects of both the real and imaginary parts of the mixture's pennittivity, such a three component measurement is often possible. Excellent results have been obtained in measuring moisture in harvested grains, ground com, chopped onions, processed cereals, phannaceutical powders, plastic pellets, coal, and minerals. It has also been demonstrated that the amount of moisture taken into a grain in a cooking operation can be measured with a GMS sensor without having to remove the grain from the cook water.

Limitations in the application of the GMS sensor typically involve trace sensitivity requirements at the part per million level, which is beyond the instrument's ability to provide at this time, or materials that are too conductive. The upper limit on conductivity is about 50,000 J-lS/cm. Applying more transmitter power and receiver amplification can not be used to mitigate the signal loss effects due to high conductivity. As the conductivity increases, the cutoff phenomenon is less distinct and the orthogonality between the real and imaginary parts of the pennittivity is lost. The current instrument supplies approximately one milliwatt of microwave power and has a signal strength measurement dynamic range of about 70 dB.

Guided Microwave Spectrometry has proven to be a versatile and robust technology for the online measurement offlowable process materials. Over one hundred systems are in service, with many of them operating in closed-loop automatic control for several years. The method has successfully addressed dozens of different applications with one basic sensor configuration. Work continues in evaluating new applications, in improving the signal processing operation, and in investigating ways to extract more infonnation from the spectral response of the measurement cell.

~~~

'" .E i'i.

~ '£

<J

" '" til

Figure I. Sectional views of a typical OMS sensor.

5000

0

-5000 V -10000

-15000

/ /

-20000 ! -25000

200 400 600

--- -

~ r--

600 1000 1200 1400

Frequency (MHz)

Figure 3. Theoretical and actual cutoff spectral response characteristics for tap water at room temperature in a 1.875 by 3.945 inch measurement cell.

- 126 -

~ u

'" ~ ~ u

§

~ }; <J

" '" til

References

0 E.

-10 \ I' /::- -

.. -t--I / 'ff ~ ~(J ~ ~ r-~

-20 ; I £-: ~ /

.JO

-40

-50

I t

] :1

fl' .00

200 400 600 600 1000 1200 1400 1600

5

4

o

Frequency (MHz)

Figure 2. Spectral response for a typical lossy dielectric material as its parameters are changed one at a time.

I 1/ R'~.999

Std. F.rr. = .04%

/ ;7

/ /

o 2 3 4

Lab Analysis (% Moisture by Weight)

Figure 4. Predicted versus observed moisture deviation from control value for hand lotion.

5

[1] B. R. Jean, G. L. Warren, and F. L. Whitehead, "Meter and Method· for In Situ Measurement of the Electromagnetic Properties of Various Process Materials Using Cutoff Frequency Characterization and Analysis," U. S. Patent 5,455,516, Oct. 3, 1995.

[2] R. E. Collin, Field Theory of Guided Waves, Second Edition. Piscataway, NJ: IEEE Press, 1991, p. 341.

- 127 -

Microwave Linear Sensors for On-line Moisture Detection and Measurement

G. Cottard and P. Berthaud - Satimo, ZA Courtaboeuf, 91152 Les Ulis, France

J. Ch. Bolomey - Supelec, 91192 Gif-sur-Yvette Cedex, France.

1. Introduction

Microwaves are well known for their sensitivity to water content. For this reason, many industrial sensors have been developed either for detecting and localizing moisten areas or for measuring moisture content in various products [1, 2]. Most of theses sensors provide a local indication which is spatially integrated over the sensor area. More recently, arrays of sensors have allowed to obtain rapid and high spatial resolution inspection of industrial materials. Theses sensors are particularly convenient in the case of products conveyed at high speed. Real time measurement of transverse moisture profiles can be used for an effective control of the transformation/fabrication process or for reliably maintaining the quality of products. More generally, cost saving can be expected either from a reduction of material losses or from an optimization of the energy consumption. This paper reports the experience gained from the use of such sensors in different areas of applications: paper, wood, plaster composite, technical textiles, rock and glass wool, etc.

2. Principle of MTS linear sensors

The modulated scattering technique (MTS) has been shown to provide an interesting technical solution, as soon as significant speed and spatial resolution are required [3]. Typically, an MST sensor consists of two parts:

1) the retina which is an array of small dipoles loaded by PIN diodes, and 2) the collector which collects the modulated perturbation resulting from the modulation of a

given diode. Sequentially modulating the diodes with a LF signal allows to record the field over the retina surface, without any mechanical movement or microwave multiplexing unit. Measurement rates of the order of 1,000 to 100,000 points per second can be reached, according to the desired dynamic range. Such measurement rates allow for recording true transverse profiles, while mechanical scans of unique sensors only provide pseudo-transverse profiles. MTS linear sensors are then particularly convenient for highspeed conveyed products for which true transverse profiles are necessary to efficiently control the fabrication/transformation process. Spatial resolution can be adjusted from a few centimeters to a few millimeters. Such a flexibility can be used to meet specific requirements derived from the size and shape of the misted areas to be detected. An interesting feature of the MTS sensors is that they can be used in transmission, reflection and also double transmission operating modes. The MTS technology can be easily implemented for frequency ranges comprised between 100 MHz and 20 GHz. The selection of the operating frequency is dictated by the specifications in terms of spatial resolution and contrast, and hence, it is directly dependent on the product properties in the microwave frequency range. For many situations, the X-band (8.2 to 12.4 GHz) proved to be a convenient frequency range, providing an adequate sensitivity with respect to moisture, sufficient spatial resolution for most areas to be detected and localized, while still maintaining acceptable cost. Typically, an X-band linear sensor consists of an array of 128 probes, spaced by 8 mm, and extending over about 1 meter. Usually such sensors are operating at fixed frequency. However, broadband versions have also been designed and realized.

- 128 -

3. Applications

Roughly speaking, two kinds of microwave inspection have been identified for moist products, involving quite different requirements for the sensing equipment. Firstly, the objective may be to effectively measure the water content, i.e. the dry and the wet masses, of materials such as paper or wood, for instance, for obtaining real-time controlling capabilities on the fabrication/drying process. In these cases, requirements are: high accuracy (1%), moderate spatial resolution (1 to 10 cm) and speed. As in the case of single-point sensor, the derivation of the moisture content requires an adequate model incorporating the largest part of information available a priori on the material to be inspected [4, 5].

A second kind of inspection modality concerns the cases where some faults or defects to be detected are primarily related to the local water content. Examples are provided by rock- or glass-fiber materials used for thermal isolation, laminated composites, or more generally, improperly dried materials. This kind of inspection is necessary, for instance, for sorting purposes or for cutting optimization. In such cases quantitative measurements are usually not required, but high spatial resolution (in millimeter range) and contrast constitute stringent aspects. Consequently, the sensing problem has then to be considered as an imaging problem where the final performances are limited by strong diffraction mechanisms. The compensation of such diffraction effects may require specific wavefront transformations [6, 7]. This paper will show many practical examples of successful as well as unsuccessful developments, for illustrating the major capabilities and limitations of the MTS linear sensors.

4. Conclusion

The use of the MTS rapid linear sensors has been investigated for many applications during the last decade. As it is well known, the development of a dedicated sensor is a long process, from preliminary feasibility studies to installation in the production unit (in factory). Intermediate steps, involving systematic assessments and calibration, have to be performed on pilot units, when possible. As a matter of fact, a successful development requires not only an adequate microwave technology, but also a specific know-how concerning the material of interest. While the microwave performance of a sensor can be easily expressed in terms of amplitude and phase accuracy, the prediction of the final accuracy on a quantity of practical relevance, such as moisture content, drastically depends on the properties of the material to be inspected and its technological process. The result is that there is a significant work to be done before making a microwave sensor efficient for a given application.

5. References

[1] E. Nyfors and P. Vainikainen, Industrial Microwave Sensors, Artech House, Norwood, MA, 1989. [2] A.W. Kraszewski, "Microwave aquametry - needs and perspectives", IEEE Trans. on MTT, vol. 39,

(5), pp. 828-835, May 1991. [3] J.Ch. Bolomey, G. Cottard and B.J. Cown, "On-line transverse control of materials by means of

microwave imaging techniques", Material Research Society, vol. 189, pp. 49-53, 1991. [4] J.Ch. Bolomey, G. Cottard, P. Berthaud, A. Lemaitre and J.F. Portala, "Microwave on-line

measurements on conveyed products: application to paper and wood industries", Material Research Society, vol. 347, pp. 161-168, 1994.

[5] C. Lhiaubet, G. Cottard, J. Ciccotelli, J.F. Portala and J.Ch. Bolomey, "On-line control in wood and paper industries by means of rapid microwave linear sensors", 22nd European Microwave Conf., Espoo, Finland, vol. 2, pp. 1037-1040, August·1992.

[6] J.Ch. Bolomey, "New concept for microwave sensing", SPIE Proc. Series, vol. 2575, pp. 2-10, 1994. [7] J.Ch. Bolomeyand Ch. Pichot, "Microwave imaging techniques for non-destructive testing of

materials", Material Research Society, vol. 269, pp. 479-489, 1992.

- 129 -

USE OF MEASUREMENTS AT TWO FREQUENCIES FOR PHASE-SHIFT DETERMINA TION

Samir Trabelsi, Andrzej W. Kraszewski, and Stuart O. Nelson Richard B. Russell Agricultural Research Center, Agricultural Research Service,

U S. Department of Agriculture, Athens, GA 30604-5677, US.A

Phase measurements are important when used for the dielectric characterization of materials. They are required for industrial material monitoring applications where the phase is correlated with parameters such as moisture content and density which need to be determined in real time. However, when the thickness of the material under test is greater than the wavelength in the material, the phase ambiguity problem is encountered. Measurements at two selected frequencies can be used to solve this problem.

Because phase-angle measurements are only possible between -180 degrees and + 180 degrees, the actual phase shift is the reading, e.g., from the vector network analyzer, shifted by n times 360 degrees, where n is an integer to be determined. This can be achieved either by making measurements on samples of different thicknesses [1], or by using the delay-time if the wave is nondispersive in the observed frequency range [2]. Both methods can be applied successfully in laboratory investigations, but they are time-consuming and impractical for industrial implementation where the phase shift needs to be measured in real time.

Dielectric properties, usually expressed by the relative complex permittivity, E = E' - jE/I, where E' is the dielectric constant and E/I is the loss factor, are often used to study the structure and composition of materials. When the free-space transmission technique is used to characterize a nonmagnetic material (f..l = 1), the attenuation ~A and phase shift ~if> are obtained from measurements of the transmission coefficient [3]. The attenuation is the loss of energy that the incident wave undergoes as it passes through a layer of material of thickness d . The phase shift is the difference between the phase angle (if» with the layer between the two radiating elements and without the layer

(if>o), namely ~if> = - 360(#-)( 0' - 1) (1) o

On the other hand it can be expressed as ~if> = <P - 360 n (2) where <P is the reading of the instrument ( - 180° :S <p:S + 180°) and n is a positive integer to be determined.

Assuming a plane wave traveling through a layer of a low-loss material (E/I «E') with thickness d, the real and imaginary components of the relative complex permittivity are determined as follows:

, (1 t:.~)2 (3) d /I t:.AAo R E ~ + 360d an E ~ 8.68611" d

where I::!..if> is the phase shift in degrees, >'0 is the wavelength in free space, and ~A is the attenuation in decibels. I::!..if> and I::!..A are to be taken as positive.

(4)

Equations (3) and (4) show that phase information is necessary to determine the dielectric constant and loss factor. It is then obvious that if these properties are to be used in any industrial process, the phase ambiguity problem has to be solved in a way that can be easily automated.

Measurements at two frequencies can be used to solve the phase ambiguity problem. The frequencies are selected in a nondispersive region such that the difference between dielectric constants E\ at 11 and E'2 at 12 is small enough to permit the following assumption, from (2) and (3) :

>'01 (<PI - 360 nl) = >'02 ( '01 and >'02 are the wavelengths in free space at 11 and 12 , respectively with 12 > 11 ' and n 1 and n 2 are the integers to be determined. For this purpose, a second equation is needed. This equation can be

n2 - nl = k (6) where k is an integer. The integers nl and n2 can be either equal ( k = 0 ) or different ( k = 1, 2 ... ) depending on the frequency difference, and the dielectric properties and thickness of the material under test. Therefore, two cases can be distinguished :

l)k=O,

2) k '" 0,

(7)

(8)

(9)

- 130 -

1:1n = >'01+>'02 1:1 360(>'01->'02) cp

(10)

where 1:1cp is the error with which CP1 and CP2 are measured and is expected to be ± 3 degrees.

To illustrate this approach, some data obtained from measurements on hard red winter wheat [3] are used. The wheat kernels were poured into a container with a rectangular cross section so that they formed a homogenous layer of constant thickness d = 10.45 cm. Fig.l shows the absolute value of the difference between the actual (naed and calculated ( ncaz.) values of n, Inact. - neal. I as a function of the absolute value of the difference between the square roots of the dielectric constants, E'l at 11= 12.3 GHz and E'2 at 12 = l3.3 GHz, 1Ft - ~I. A

linear relation exists between the two parameters, with Inaet. - neal. I increasing as 1Ft - ~I increases. This confIrms the original assumption, the smaller 1Ft - ~I the more accurately n1 and n2 are determined. However, it is not the only requirement; an additional condition on the selection of the two frequencies can lead to better estimates. Equation (10) implies that the larger the difference A01 - A02 , the smaller will be the error in n. To verify this experimentally, some measurements were performed for various frequency differences 12 - 11 between

11= 12.3 GHz and 12 max = l3.9 GHz. Fig.2 shows that the ratio ~ -~ has an exponential-like decay with € 1- €2

12 - 11 which agrees with (10). Optimum choice of the two frequencies is possible provided that the assumption regarding the dielectric

constants remains valid. Although the use of measurements at two specifIc frequencies does not place any restriction on the sample thickness, the 10-dB attenuation criterion should be respected. This method can be easily programmed in an automated process where measurements of phase shift are needed in real time.

0.4 ,..-----.-----,-----,---r----,

0.3

...., tl

s;:

0.2 -.: <>

s;:<:3

0.1

O.O~----L---~-___ L-___ ~

0.000 0.002 0.004 0.006 0.008

I{;' -'W I 1 2

Fig. I. Absolute value of the difference between the actual and calculated values of n as function of the difference between the absolute value of the square root of the dielectric constants €'land 10'2 obtained for hard red winter wheat at 11 = 12.3 GHz and 12 = 13.3 GHz, and at T=24 CC.

120

100

u:l 80 Cl.. o u3 60

40

20

O~~_-L_L-~_-L_~_L-~ _ _L~ 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0

12 -1;. GHz

Fig.2. Slope nad. -neal. as function of the difference ,j7;-yR2

between two selected frequencies, 12 - 11' with 11 = 12.3 GHz and 12 max = 13.9 GHz.

REFERENCES

[1] M. Sucher and J. Fox, Handbook of Microwave Measurements. New York: Brooklyn Polytechnic Press, 1963. [2] A. Klein, "Microwave moisture determination of coal. A comparison of attenuation and phase measurement," in Proc. 10th Euro. Microwave Con!, voLl ,pp. 526-530, 1980. [3] A. W. Kraszewski, S. Trabelsi, and S.O. Nelson, "Grain permittivity measurements in free space," in Proc. 25th Euro. Microwave Con!, vol.2, pp. 840-844, 1995.

- 131 -

SIMULTANEOUS DETERMINATION OF COMPOSITION AND OTHER

MATERIAL PROPERTIES USING DIELECTRIC MEASUREMENTS

(INVITED)

M. Kent

K & S Associates, 28A Ann Street, Edinburgh EH4 1PJ, Scotland.

Abstract. The development of microwave methods for composition analysis from its early stages to the present state is covered by this paper. For the most part this has meant the determination of water content, but it is shown that later work has attempted to measure or eliminate other variables such as temperature, density, other compositional variables or even different kind of treatment. The possible direction of future work is suggested.

Summary

It can be shown that for particulate materials there is a strong dependence of the dielectric properties on bulk density. This is understood if one considers each particle as a non-interacting absorber, then it is obvious that increasing the number of these absorbers per unit volume will increase the power absorbed proportionally. In fact, it was noted [1] that within certain constraints, for a given mass of material filling a waveguide, the attenuation and the phase shift remained constant as the sample was compressed, i.e., its length was shortened. Such behaviour will obtain only while compression has no effect on the internal fields experienced by the individual water molecules. Because of the mutual relationships between real and imaginary part of the material permittivity encapsulated in the Kramers-Kronig relationship:

00 00

2f X("(w) 2 f(I(W)_(

(I(W) - (00 = 7f 2 2 dx and ("(w) = i( 2 200 dx,

x -w x -w (1)

o 0 where x = 27r f is the operating angular frequency, it would be expected that the phase shift through the material should also behave in the same way as the attenuation, i.e., remain constant. This of course applies only to dispersive loss and it is assumed that d.c. conductivity has a negligible effect in these samples at micro-wave frequencies. It is also assumed that the effect of increasing the density and thus increasing the d.c. conductivity through increased number of points of contact between particles has no significant effect at these frequencies. This observations of constancy in the attenuation and phase shift leads very directly to simple relationships between the real and imaginary components of the permittivity and the density, p. Under these circumstances they are seen to be simple quadratic functions:

(' = ap2 + bp + 1 and (" = cp2 + dp. (2)

It has been shown [2] that the dielectric material permittivity occupies a certain space in the complex plane with moisture content and density as parameters, and no crossing or folding of curves within each family. Figure 1 shows this for milk powder at 25°C measured at 10 GHz. It is obvious that any measurement of (' and (" represents an unique solution for a given pair of variables such as in this case, moisture content and density.

In practice, the complex permittivity is never measured directly, other variables which are dependent on this property being more amenable to direct observation. Such variables are also complex and a choice must be made as to which to use. We have already mentioned the phase shift and attenuation of a plane wave propagated through the material. The complex reflection coefficient or complex resonant frequency of the resonator, can be other choices.

0'20 - 132 -

0'810

0·10

%M.C

Density

OL-~~---~~~~~~---~~~---~~~---~~~

1'0 1'5 2'0 2·5 t'

Fig. 1. Complex plane plot of the dielectric properties of milk powder with varying density and moisture content

Although a great deal of attention has been paid to the problem of density of powders and its effect on moisture content measurement, this is not the only problem that can be solved by the measurement of more than one dielectric variable. In the food industry one that is often faced is the determination of the composition of food, say meat products, in terms of fat, water and other solids. Some data have been published [3] concerning the dielectric properties of ternary mixtures such as this and it has been pointed out that this proximate composition could be determined from such data. At this point it should be remembered that to measure a number of variables n, which have n degrees of freedom, requires at least n independent measurements. In case of ternary mixtures however, the number of degrees of freedom is n -1, which is 2, because determination of two components must also determine the third. It could be argued that even the problem of water content and density in particulate material is basically one of ternary mixtures, since in that case the components are water, dry material and air.

30 f :3'05GHz

25

Su,crose concentration (%)

20

Temperature (OC)

J5~--------~----------~----------~~--------~--------~ -2'5 -2'0 -J'5 -J·O -0'5 o

-5;'2 (radians)

Fig. 2. Sucrose concentration and temperature as parameters in the plane of the complex scattering coefficient S12

One problem encountered by the confectionery industry is in measuring the concentration of hot sugar solutions. Apparently anomalous results were obtained when the simple attenuation method

- 133 -

used. It seems no surprise since such liquids have dispersion because of their polar constituents and the relaxation frequency is very temperature dependent as is the viscosity, a highly correlated variable. Circumstances could dictate that loss increased and then decreased with increasing sugar content or with changing temperature. Using published data [4] the plot shown in Figure 2 was obtained with an assumption of a 10 mm pathlength between 20 mm PTFE windows. As can be seen, for this range of temperatures and concentrations, there is indeed ambiguity in the concentration for a given temperature using loss measurement alone, 1812 1. In fact, it would be perfectly feasible to measure phase, arg 8 12, and temperature alone to determine the sucrose concentration. It is known that the frequency and temperature are to a certain extent complementary in the absence of other changes in the system. The use of temperature as a parameter in these sucrose concentration measurements could readily have been replaced by frequency with similar results if multi-frequency measurements were made. Over a limited range the whole system could be calibrated in terms of the dimensionless product WT or f / f r' where the relaxation time T = 1 /27r f r'

It has been shown that dielectric properties vary with frequency and temperature and that the variations depend, among other things, on the relaxation frequencies of the polar molecules involved. The spectrum of these properties over a two to three decade range of frequencies can provide unique information about the material. In addition, when, for example, water is added to a foodstuff, the dielectric spectrum will change according to the way in which this water was added [5]. Water can be taken up in ordinary processes such as washing or can be added deliberately using polyphosphate solutions. The results are quite different as Figure 3 shows. It can be seen that water alone has the effect of diluting other solids, indicated by a sharp fall in conductivity at low frequencies. Polyphosphate on the other hand adds conducting ions as well as water and the resultant curve in the complex plane is quite different. It must be said that for the case of addition of water alone the reduction in conductivity is not entirely due to dilution, but also to loss of salt into the exterior water through diffusion.

The problem is how to automate such measurements when much of what is seen is qualitative. Measurements were made at 26 different frequencies and with both real and imaginary components and that represents a lot of information on one sample. Much of it seems redundant though because of the high degree of correlation between values at adjacent frequencies. A technique used in similar cases in the field of Near Infra-Red (NIR) spectroscopy is the multivariate approach of Principal Component Analysis (PCA) and Principal Component Regression (PCR) and these methods were applied to these data. PCA is a technique for applying a linear transformation to a set of correlated variables to produce a new set of uncorrelated and standardized variables called 'components'. These components are a linear combination of the original variables, for example in this case, the 26 complex spectral values. This is expressed as follows:

(3)

where Y j is the j-th principal component, X /s are the original variables (e.g., E', E"), and the a{.'s are constants. Various conditions are imposed such that the principal components are orthogona in p space and the constants are calculated in sequence by maximising the variance of each principal component. Often the first few principal components account for a large proportion of the total variance of the original set of variables. If this is the case, they may be used to summarise the original data and as an input to further analysis.

Principal Component Regression uses PCA to reduce the data to manageable proportions so that a regression can be performed. For this there have to be dependent variables and these can be, for example, the water content or any other compositional variable. The aim is to describe these in terms of a smaller number of principal components by regression methods. The orthogonality of the principal components makes this a relatively simple task since the addition of further components to the regression does not alter the coefficients of the equation already found. The correlation coefficient, however, is always increased and the process of adding components only stops when there is no more significant variable to add. After forming a regression equation, an attempt can be made to construct a calibration equation such that for any further spectrum obtained, the composition could be predicted after calculating the principal components from (3).

60.0

50.0

40.0

, w 30.0

20.0

10.0

- 134 -

• B.

• • •

•• • • • .' . . · .. ··.A

..... .....

• ~''''"''''''II. : .... II. JI .... '1- • ...,,- ....

........ .... ... ~ • Untreilted chicken breilst

.... Water soaked chicken breast • Polyphosphate soaked

chicken breast

0.0 ..... _....I. __ -'-__ L-_-'-__ ..L..._......I __ -I.._--J

0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0

E'

Fig. 3. Effects of adding water and polyphosphate solution on the dielectric spectra of chicken meat plotted in the complex plane

The principal components can be used in a variety of ways to discriminate the various treatments of the samples. When the values of the first two principal components are plotted against each other, there is clear distinction between samples which have had only water added, those which have had polyphosphate solution added and the controls. This may be the way to use the dielectric spectra of foods in the future, i.e. for the discrimination of processing effects or for the identification of abnormal properties of the food.

This field has been in existence for a long time and it cannot be doubted that microwave techniques offer a useful method for measuring the bulk properties of materials. The term bulk is stressed, because there are competing methods, such as NIR, which offer far more discrimination of components in a mixture and accuracy of measurement, but are nevertheless limited to either surface measurement from the reflected NIR spectrum or, because the IR energy is absorbed strongly in a material, can only measure thin samples in the transmission mode. The many other problems that assail NIR techniques such as a dependence on particle size, etc., are all resolved to certain extent by the application of numerical techniques and comp~ting power. It is possible that microwave methods could also achieve much more with the multivariate spectral approach. This kind of approach has begun, but there is a lot of work to be done for researchers in this field.

References

[1] M. Kent, "Complex permittivity of fish meal: a general discussion of temperature, density and moisture dependence", J. Microwave Power, vol. 12, pp. 341-345, 1977. [2] E. Kress-Rogers and M. Kent, "Microwave measurement of powder moisture and density", J. Food Engng, vol. 6, pp.345-376, 1987. [3] T. Ohlsson, M. Henriques and N.E. Bengtsson, "Dielectric properties of model meat emulsions at 900 and 2800 MHz in relation to their composition", J. Food Science, vol. 39, pp. 1153-1156, 1974. [4] M. Kent and E. Kress-Rogers, "The COST90 bis collaborative on the dielectric properties of foods", in Physical Properties of Foods, vol. 2, R. Jowitt, F. Escher, M. Kent, B. McKenna and M. Roques, Eds., London, Applied Science Publishers, Elsevier, pp. 171-197, 1987. [5] M. Kent and D. Anderson, "Dielectric studies of added water in poultry meat and scallops", J. Food Engng, vol. 15, in press, 1996.

- 135 -

MULTI-CHANNEL MICROWAVE RESONATOR

MOISTURE- MASS METER OF PAPER WEB

P.D. Kuharchik, I.A. Titovitsky, A.Ch. Belyachits and N.!. Kourilo

Byelarussian State University, 200050 Minsk, Byelarussia

Infrared and microwave sensors are most widely used for moisture control od paper web at the production stage, and radioisotope sensors are used for mass measurement. Each of the moisture content determining methods has its own advantages and disadvantages. Thus, when measuring the moisture content of a paper web of mass from 10 to 150 g/m2, infrared sensors are most widely used, but for thick papers and paper board, microwave moisture meters are superior. The distinctive feature of microwave moisture content meters is that their indications do not depend on paper color, its density, different technological additions, etc., while for infrared moisture sensors those factors are substantial and require additional instrument calibration, when switching from one paper to another. The main drawl>ack of radioisotope mass meter is the utilization and periodic burying of radioactive materials.

In this paper, a multichannel microwave resonator moisture-mass meter of paper web is described. The instrument can simultaneously measure the moisture profile in the range from 1 to 10% moisture, and mass profile from 40 to 300 g/m2, at the output of a paper-making machine, at six zones across the web width.

The main task of researchers was to develop a microwave sensor allowing measurement of low moisture content in materials of a little mass. For this purpose, a microwave resonator was developed which allows to intensify the interaction effect between the electromagnetic radiation and the material, because the microwave signal travel through the material under test several times. It is known that by introducing a dielectric material into a microwave resonator, the change of the resonant frequency and the amplitude of resonance, proportional to the value of the material permittivity and its mass, will take place.

To develop a calibration algorithm for percentage of moisture content and mass per unit area determination from the resonator parameters mentioned above, numerous experimental data for different types of paper were taken. It was find out that chan'ge of the resonant frequency, b.f, and the change of the attenuation of the resonance, b.r, are both linearly related to changes of the paper mass, while the moisture content is kept constant. However, when the paper moisture content is changed from 1 to 10% moisture, the character of changes in b.f and b.r is nonlinear. Numerous experiments showed that the dependence between these variables and moisture content can be approximated by the second degree polynomial, and it can be expressed by the following equations

b.f = (Af W2+ BfW +Cf)(Md+Mw )

b.r = (ArW2 + BrW + C r)(M d + M w)'

(1) (2)

where W is the moisture content of a sample, M d is the mass of dry substance in a sample, M w is the mass of water in the sample, and A,B, and C are the numerical coeffi ci en ts.

- 136 -

After inverting (1) and (2), the moisture content can be expressed from the second order equation

SW2+TW +V =0

where S = LlrAf-LlfAr ; T = LlrBf-LlfBr, and V = LlrCf-LlfCr., in the form

-T+ VT2-4SV W= 2S ' [%]

and the mass of paper from M=M +M = Llr

d W A W2+B W+C r r r

(3)

(4)

(5)

Thus, the calibration of moisture-mass meter turns to the measuring of the values of Llf and Llr for three moisture content values in the range from 1 to 10% moisture for one and the same sample, without regard to its mass changes due to varying water content. After the calibration, the instrument can be used for the moisture content determination in percent from (4) and mass per unit area from (5) within the indicated range of moisture content and the paper mass from 40 to 300 gJm2

•

The main objective in the design of an instrument to operate in a production mill, is to ensure the stability of readings when external factors are changing, e.g., ambient temperature and humidity. Since the natural resonant frequency of a resonator is determined by its volume, thermal expansion of the cavity and air humidity can affect the parameters of the system. To avoid this limitation, the double-mode regime of the resonator operation was applied. At one of the resonant frequencies, the antinode of the electromagnetic field in the gap is created, and the interaction with paper web is maximum. At the other resonant frequency, the node of the electromagnetic field exists at the gap location, and there is no interaction with the web. Since these two modes of oscillations are excited in the same resonator, then for changes in external temperature, the difference between the resonant frequencies of two modes remains constant. In the same way, the amplitude of the measuring mode is corrected with respect to the amplitude changes of the fundamental mode.

At the paper production mill, obtaining the profile of moisture content and mass distribution across the web is very important. At present, special expensive scanning systems are used to accomplish this task. The moisture-mass meter presented in the paper, allows obtaining moisture and mass profiles without scanning system, simple by using several sensors placed across the width of the paper web. In the instrument, the principle of the frequency channel separation is applied. Each resonator has its natural resonant frequency. When the frequency of the basic generator is changing in the range exceeding the natural frequencies of all resonator, the successive resonator switching in time takes place, that ensures a scanning mode without moving parts in the system.

The external view of the microwave moisture-mass meter is shown in Fig. 1. The instrument consists of the bearing construction made of two pipes (1) on which six resonators (2) are attached, providing a clearance of 8 mm for introducing a paper web into it. On one side of the bearing, the measuring unit (3) is located, and at the other side there is an electromechanic lock (4) which ensures an invariable gap clearance between the two parts of the resonators during the measurement. The bearing support is suspended at the mobile platform (not shown) which allows to move the whole system. In the initial position, the cramp is out of the paper web, and the el-mag lock is closed. In this position the meter is automatically calibrated before the start of the measurements. When rolling on the web, the lock opens and the paper web is introduced into the gap of the resonators. In the final position the lock closes and the moisture content and mass are measured in six zones of the web across its width.

- 137 -

3

~

O:~~6~~.~ ... Fig. 1. Diagram of the moisture-mass measuring system

The IBM compatible PC (5) is applied to control the measuring process and store the measured values of the moisture content and mass. It allows to display of the data in each of the six zones in a form convenient for an operator, to calculate of an average values for the web width, and to control the tendency of changes in the average moisture content and in mass in each zone versl,ls time, as well as to provide an alarm signal in cases when moisture content and/or mass are out of the given limits. The data transmission from the instrument to the computer is provided within the paper mill area over a communication line (6).

The instrument provides moisture content determination in the range from 1 to 10 % with an uncertaintl of ± 0.2% moisture. The weight per unit area of the paper web from 40 to 300 g/m is measured with an accuracy of ± 2.5%. The multi-channel microwave moisture and mass meter of paper web passed successfully industrial tests in a number of enterprises in Byelorus and Russia.

.,

Density independence of microwave moisture determination using two parameter measurements

R. Knochel, F. Menke

Lehrstuhl fUr Hochfrequenztechnik, Universitat Kiel, Technische Fakultat Kaiser-Str. 2, 24143 Kiel, Germany

1 Introduction

Many modern production processes, for instance processing of agricultural products or food-stuffs, are running at very high speed. Since products are marketed by weight, close control of the output in amount of mass per unit of time is extremely important for keeping the manufacturing costs at the desired level. Precise online gauges are required for monitoring the output in mass but also its relative or absolute moisture content, which is an important part of the overall mass balance.

Most widespread in use are ,-ray mass or density gauges. However, with respect to the mentioned food and agricultural market, application of such radiometric sensors, which radioactively irradiate the output mass-stream, becomes more and more undesired. This is due to objective and also psychological reasons of the customer, but also dictated by legal regulations, at least in some countries.

Microwaves are well known to offer an excellent possibility for monitoring the moisture content, because a representative measurement value can be delivered by the capability of penetrating a large volume and traveling throughout a layer of material. This is in contrast to other non-contacting measurement approaches such as infrared reflection at the surface of the specimen. The relative change of attenuation A or phase tl> caused by the moist material, compared to the empty path, can serve as the measurement signal. However, as moisture content determination using A or tl> is an indirect measurement, there is still a dependence on mass per area or density p. This density dependence of the one parameter microwave measurement is a significant disadvantage and again raises the undesired need for a separate density gauge.

Fortunately the observation can be made, that both attenuation and phase depend in a similar or virtually equal manner on density. The same seems to hold for the electric susceptibility X = f.' -1 and the loss factor / , which compose the complex permittivity f.r = f.' - j/. As a consequence, proper processing of a suitable two parameter measurement can lead to density independent moisture measurement, but also to moisture independent density measurement. Both can be obtained by exclusive application of microwaves, thus making radiometric gauges completely obsolete.

2 Two parameter microwave moisture measurement

Two parameter measurement for eliminating the density dependence in microwave moisture measurements has first been proposed in [1) and then been further developed [2), [3). The method applies to a microwave transmission setup, where attenuation and phase are described as

w A = - . 1 . K . p . F( 1/1)

c (1)

and w tl> = - . 1 . K . p . G( 1/1)

c (2)

In eq. (1) and (2), 7 is the propagation constant of free space, 1 the thickness (length) of the dielectric layer, J{ a constant and p the density. Eq. (1) and (2) are based on empirical knowledge about a great number of moist dielectrics, and tell us, that A and tl> depend linearly on density changes. F and G are functions of the dielectric mixture and reflect the kind of behaviour with changing moisture content (MC) 1/1 , where 1/1 may be expressed as the absolute MC

(3)

or relative M C .1. _ mw 'f/r -

mw+md (4)

Here mw and md mean the masses of dry material and water. The linear dependence of eq. (1) and (2) on density can be justified and shown by describing the moist dielectric mixture with the so called refractive index model [4).

It is evident from eq. (1) and (2) that the relation

A F(?jJ) q> = G(?jJ) (5)

is neither dependent on density p nor a function of layer thickness 1 Hoping that F and G differ significantly enough, the fraction (5) is still a function of moisture and can be used for density independent moisture measurements.

Another approach for achieving density independence using two parameter microwave measurements has been proposed in [5). It is claimed and empirically supported by measurements, that the quantity

II

M=~ X

(6)

is a density independent moisture function. This statement can be supported by the following consideration [6):

The permittivity of the moist material with respect to density changes can be modeled by a large number of thin parallel layers of moist material (thickness tl , permittivity fl) separated by air gaps (thickness to). Having the electric field vector polarized parallel to the layers, the effective permittivity derives as

to + tlcl f=---to + tl (7)

Separating f = C' - jf" into real- and imaginary parts leads to

, (f~ - l)h f - 1 = ":""'='---'--

to + tl (8)

Hence II

f f fl -=--=-,--X f' - 1 fl - 1

(9)

is independent of the filling factor or density, when varying to or t1. The same reasoning can be repeated for orthogonal polarisation of the electric field, revealing that e"

for that case is significantly smaller. Although a real material includes layers in all possible directions, the permittivity of the parallel polarisation dominates and eq. (9) is approximately valid.

The expression eq. (9) is very well suited for determining the moisture content by cavity perturbation methods, because the ratio of attenuation and detuning of a cavity after insertion of moist material is directly proportional to it. However, in an open transmission arrangement, where layer thickness 1 is seldom known, permittivity cannot be measured directly.

In that case, the relation

~c,;c".Vl+1 q> X 2N (10)

is used and claimed to be approximately density-independent, as long as e' :::::: 1 is valid. Therefore, it can be assumed, that within practical limits, eq. (5) and (10) lead to similiar results.

Returning back to eq. (5), assumptions are now made concerning the functional dependence of F and G on ?jJ . To this end, a polynomial representation is adequate, i.e.

F = io + 1I?jJ + h?jJ2 + ... (11)

and (12)

Often linear dependence of A and q> (F and G) on moisture is reported [3), [7). However, this seems to be only true for a limited range of MC. An example is given in Figure 1a and 1b for tbacco. Typically A and q> change only weakly at low MC (below ?jJr = 10% ) because of the strong binding of the first

20r-~~--~~--~~--~~--.

18 16

14

~ 12 to ~IO

~ 8

6· 4 ......... / ..

2 . ····A)~· .

.. , . ,

. _ .--0-

00 5 10 15 20 25 30 35 40 45

'V [%)

5r-~----~~--~~--~-T--'

4.5

4

~ 3.5 ~ .!:::. 3 e

2.5

2 ..

1.5 ... ~ _ .t:r /~< ; 10 5 10 15

'V [%)

./0··

35 40 45

Figure 1: a) attenuation and b) phase shift for tobacco versus moisture content (constant density p = 220~)

molecule layers of water. Increasing Me then produces steeper slope of the A- and q,-curves. Hence the functional dependence can be modeled better by using quadratic polynominals instead of straight lines. As a further characteristic feature, it can be seen, that the dry material is virtually lossless, i.e 10 = 0 , but produces a significant phase shift due to its nonzero X , i.e. go '" 0 . Now eq. (5) can be rewritten as

A _ h1/J + 121/J2 =-'---'--'--'--

q, go + gl1/J + g21/J2 (13)

From eq. (13), we can see, that the whole measurement approach is mainly based on the fact, that 10 ~ 0 and go '" O. If 1/J increases, q, increases also rather fast, due to the high permittivity of water, and relations establish in a way that

(14)

is valid. In that range eq. (13) changes only weakly with 1/J and becomes insensitive, i.e. accuracy deteriorates. These statements can be substantiated by the example of tobacco, where

(15)

with 12 = 0.01 , go = 0.6 and g2 = 0.002 can be used. The calculated functional dependence of eq. (15)

4r-~---~~--~~--~~--'

3.5 .

3 . :;:;' e 2.5 . ;:a ~ 2

~ 1.5

'V[%)

,. : 1 ... . /

0.5 .~.«~'

00 5 10 15 20 25 30 35 40 45

'V [%]

Figure 2: a) simulated and b) measured relation ~ for tobacco

is shown in Figure 2a, whereas Figure 2b shows a measurement. It is evident, that both curves behave similiar. The S-shaped curve is characteristic for most of the matl:)rials measured in our laboratory. In [7] is proposed to use 1/Ja instead of 1/Jr , in order to get more linear relations, because 1/Ja is directly proportional to mw . A graphical representation between 1/Ja and 1/Jr is shown in Figure 3.

When applied to the A-q,-ratio, a slightly more linear shape of the curves of Figure 2 can be achieved, but mainly at low Me. No essential changes are obtained, and a significant loss of accuracy above 1/Jr ~ 25% or 1/Ja ~ 33% remains. The solution of eq. (15) is shown in Figure. 4, where an error (noise) is added to the measured A-q,-ratio.

0.8 r---r----,~~__,.-_.___._-..,..___,___r>

0.7 .

0.6 .

0.5 .

;. 0.4 ..

0.3 .

0.2 .

0.1

%~~~~~~~~~ __ ~~--J 'V[%] 'V [%]

Figure 3: 'l/Ja versus 'l/Jr Figure 4: Error in MC calculation using eq. (15) with distorted~-values

As expected, significant uncertainty in the reading appears for MC below approximately 5% and above 25%, where the slope of the A-~-ratio is flat. From the given description, it can be deduced that twoparameter moisture measurement using the A-~-ratio suffers from a limitation in accuracy to a restricted range of MC, which is typically established between 'l/Jr = 5 ... 25% . The same reasoning applies to the utilization of / lx, because both / and X can be described by polynomials similar to eq. (11) and (12).

3 Frequency swept two parameter measurement

A solution is required which allows us to overcome the described limitation while maintainig the density independence. A possibility could be to modulate the water in the moist substance, for instance by using switched static electric fields of sufficient strength. This, however, is not successful in practice. Inspection of the dielectric behaviour shows, that all constituents of the dielectric mixture but water have a virtually frequency independent permittivity. This suggests to utilize broadband frequency swept measurements to increase sensitivity against MC. Such an approach also permits us to take advantage of differences in the frequency characteristics of / and X or A and ~ , respectively. Measurements show, that such a difference is not very pronounced in a practical material, but it is present and can be seen by sweeping over a broad frequency band. By proper choice of the frequency, a stronger relative increase of A as compared to ~ over the bandwidth can be attained. Further inspection of eq. (1) and (2) indicates, that frequency-differences of A and ~ have the same density dependence as the quantities themselves. Hence, as a new density independent two parameter measurement approach, we can use

LlA = (LlF) . (LlG)-l Ll~ Llw Llw (16)

A plot of this function, measured for tobacco as the example in the band from 8-12 GHz is shown in

Figure 5: Measured ~~ for various densities of tobacco

Figure 5. We can determine from this curve, that, as a favourable property, no saturation at high moisture

values takes place but a virtually linear increase with MC can be stated. The cited measurement scheme has further advantages: Frequency sweeping can be done at very high speed, i.e. it is possible to collect a great number of measurement data, even at a moving dielectric, as is present for example on a conveyor belt. These data can be used for averaging purposes and reducing measurement uncertainties, which may be caused by frequency ripple. Such a ripple is often present and due to standing waves (mismatches) in the measurement path. It remains unseen, if investigations are made only at a fixed frequency.

4 Conclusion

This paper discusses the well known two parameter moisture measurement methods and their usability versus possible range of MC. It is pointed out, that the principle only works, because the dry fraction of the moist material is virtually lossless but has a permittivity f.' > 1 , i.e. produces a phase shift against the empty measurement path. Sensitivity is weak at low MC because of strong binding of the water, but also at high MC, if the phase shift produced by the water content dominates that of the dry fraction. An analog discussion can be extended to f.' and / .

Finally, a new density independent approach is suggested, which uses data collected by frequency swept measurements and can be made usable even at very high MC.

5 Acknowledgement

This work is part of the project Kontinuierliche und zerstorungsJreie FeuchtemejJverJahren mit Mikrowellen. Financial support by the Volkswagen-Stijtung under contract 1/67 193 is greatfully acknowledged.

References [1) Kraszewski, A.; Kulinski, S.: An improved microwave method of moisture content measurement and control. IEEE

'Trans. on Industr. Electr. and Control Instrum. lECl-23, pp. 364-370, 1976.

[2) Kraszewski, A.; Kulinski, S.; Stosio, Z.: A preliminary study on microwave monitoring of moisture content in wheat. J. Microwave power 12 (3), pp. 241-251, 1977.

[3) Kraszewski, A.: Microwave monitoring of moisture content in grain-further considerations. J. Microwave power and electromagnetic energy, 23 (4), pp. 236-246, 1988.

[4) Birchak, J.R.: High dielectric constant microwave probes for sensing soil moisture. Proc. IEEE, vol. 62 (1), pp. 93-98, 1974.

[5] Meyer, W.; Schilz, W.: A microwave method for density independent determination of moisture content of solids. J. Phys. D: Appl. Phys. (13), pp. 1823-1830, 1980.

[6] Unger, H.G.: Bemessungsgrundlagen fur den Rundhohlleiter mit Wandbelag. AEU, Archiv der elektrischen Ubertragung, pp. 141-152, 1964.

[7] King, R.J.; King, K.V.; Woo, K.: Microwave moisture measurements of grain. IEEE Trans. Instr. Meas., vol. 41, pp. 111-115, 1992.

- 143 -

TEMPERATURE-INSENSITIVE AND DENSITY-INDEPENDENT

GRAIN MOISTURE CONTENT DETERMINATION FROM

MICROWAVE MEASUREMENTS

A.W. Kraszewski, S. Trabelsi and S.D. Nelson

u.s. Department of Agriculture, Agricultural Research Service, R.B. Russell Agricultural Research Service, Athens, GA 30604-5677, U.S.A.

ABSTRACT

Online moisture content monitoring is necessary for efficient control of many dynamic processes, and without such a possibility, the processes cannot truly be automated. Monitoring of material moisture content must be fast to allow automation in real time. It must also be continuous and nondestructive to overcome process disturbances, and it must be accurate to assure precise control of the processes. For some time, a new branch of metrology, microwave aquametry [1], has been developing; and its methods, based on microwave techniques, have proven to be successful in solving such problems. The purpose of the approach presented in this paper is to enhance the accuracy of continuous moisture content determination by limiting the disturbing effects existing in industrial situations.

Measurements of the attenuation and phase shift of a microwave beam passing through a layer of grain of given thickness were carried out for grain lots of different origins, at various levels of moisture content, and at different densities and temperatures. Typical results of these measurements for shelled field corn, Zea mays L., are shown in Fig. 1. At a given frequency both wave parameters are dependent upon moisture content, bulk density of the material and its temperature. For a given temperature, the attenuation and phase shift increase with moisture content as shown in Figs. 1a and 1b, respectively.

When the experiments were repeated with grain at different temperatures, both wave parameters increased with temperature, as shown in Figs. 2a and 2b. Data for selected moisture contents shown in Fig. 2 indicate that the temperature dependence of both parameters is highly linear. For a given moisture content and temperature, both wave parameters are dependent upon the grain bulk density, as shown in Fig. 3. For both parameters, the dependence is linear. Such results are typical for grain at microwave frequencies.

Closer analysis of the moisture content definition indicates that both wave parameters are dependent upon the water concentration in the material (mw/v, mass of water in the unit volume) and temperature. The mass of dry material in the unit volume (md/v) does not essentially affect the wave amplitude and is linearly related to the phase shift. It has been shown previously [2, 3], that two measurements are needed to determine md/v and mw/v independently and to permit calculation of the moisture content and the bulk density of the moist material. The attenuation A and phase shift ¢ of an electromagnetic plane wave traveling through a layer of material of thickness t, can be described by the following relationships

A = t (a1 + a2k + a3T + a4P) ¢ = t (b1 +b2k+b3T+b4P),

(1) (2)

where T is the material temperature in degrees Celsius and water concentration k and quantity pare defined by the following relationships in terms of bulk density p and percent wet-basis moisture content Me:

mw mw md mw Me d md (Me) ( ) Me = mw + md X 100, p = v- + tr' k = v- = P 100 an p = tr = P 1 - 100' 3

where k and p are in units of density, e.g., in kg· m - 3.

In many practical situations, it is not difficult to control the layer thickness of the material by simple mechanical means. Temperature of the material can be also easily measured by available methods in real time. Thus, both thickness t and temperature T can be considered as known variables in the above set of equations. Solving (1) and (2) for k and p provides the following

- 144 -

45

40

., 35

" Z 30 0 ;:: « ::> 25 z w 1= « 20

15

10 8 10 12 14 16 18 20

MOISTURE CONTENT, ~

800~--~----'-----.----'-----r----'

'" .g 700

t :;: V> W V>

~ 600 "-

CORN 24°C

15.2 GHz

ft804

00 o 0

~o.:Oo

o

o ~/. ;,.0 0 co

500~--~----~----~--~~--~--~

8 10 12 14 16 18 20

MOISTURE CONTENT, ~

Fig. 1. .Attenuation (a) and phase shift (b) in shelled field corn (t = 5.2 cm) vs moisture content

60 900

CORN 24°C CORN 24°C

50 800 ., '" MC= 17.5~ II

" " Z 40 t 0 ;:: :;: 700 « V> ::> w z 30 V> w « 1= :J: « "- 600

20

500 10 20 0 10 20 30 40 50

TEMPERATURE, dog C TEMPERATURE, dog C

Fig. 2. Attenuat~on (a) andp.h.~e shift (b) in shelled field corn as a function of temperature

40 800 CORN 24°C

MC = 17.5~ CORN MC = 17.5~

35 0 24°C

., 0 0

'"

~ " " 700 " Z 30 t 0

;:: 14.27- :;: « ::> V> z 25 0 0 0 w 11.87-~_ w 1= 9 8 00 0 0 0 V> _A-c:r-0 « 600 « :J: --11.87- "-

20 q.

0 u

15 500 700 720 740 760 780 800 700 720 740 760 780 800

DENSITY, kg· m -3 DENSITY, kg.m-·

Fig. 3. Attenuation (a) and phase shift (b) in shelled field corn as a function of its density

60

N = 259 e-

50 t-. 15.2 GHz -

V> I- 40 V> r-W I-

"-0 30

'" w m r-::Ii 20 ::> z

A rh 10

0 -2 -1 0 2

DIFFERENCE IN MOISTURE, %

Fig. 4. Distribution of differences in moisture content of shelled field corn. (Me = 9 to 19%)

100

N = 259 80

V> 15.2 GHz

l-V> w

60 I-

"-0

'" w 40 ., ::Ii ::> z

20

-40 -20 .0 20 DIF'FERENCE IN DENSITY, kg m-

3

Fig. 5. Distribution of differences in density of wet shelled field corn (p = 765 kg, m - 3).

- 145 -

and (4)

Next, the moisture content and bulk density can be determined from (3).

In the example used here for illustration of the procedure, attenuation and phase shift for three shelled corn hybrids were measured at 15.2 GHz for various moisture contents in the range from 9 to 19%, at temperatures from 4 to 45°C and three different densities for each sample. Layer thickness was kept constant at t = 5.2 cm. The set of 518 data points was divided into two equal subsets, a calibration set (even numbers) and validation set (odd numbers). The calibration set of 259 data points was fitted by the following equations:

A = 20.9605 + 0.2914k + 0.40975T - 0.05649p r = 0.9794 (5)

¢ = 39.707 + 2.8433k + 3.1853T + 0.37152p r = 0.9700

where r is the correlation coefficient. Following the procedure outlined above, the moisture content and bulk density can be expressed as follows:

Me _ 21.01¢ + 138.17 A -123.5T - 3730.4 [%] - 1.294¢ - 9.193A - 0.354T + 141.316

p = 1.294 (¢-7.104A-0.273T+109.21) [kg.m- 3].

(6)

The validation set of 259 data points, not used for deriving the calibration equation (5), was used for calculation of the moisture content and bulk density from (6) to evaluate the quality of an expected prediction. The results were compared with oven moisture tests and bulk density determinations from sample weights and sample holder volume. The histogram of Fig. 4 shows the distribution of differences between oven moisture content determinations and calculated moisture contents. The mean value of differences was 0.022% moisture, while the standard deviation of differences (standard error of performance, SEP) was 0.59% moisture. The SEP for the bulk density was 13.63 kg· m - 3 with bias of - 0.42 kg. m - 3. The respective numbers for water concentration and density of dry material are 4.67 kg· m - 3 and 13.07 kg· m - 3, with biases being 0.1 kg· m - 3 and - 0.53 kg· m - 3. The error distribution for the prediction of bulk density is shown by the histogram of Fig. 5. From the normal error distribution, it is evident that the sign of the error is not related to the range of measured values, and the errors have totally random character for both predicted variables.

It has been shown that compensation for the material temperature can be easily introduced into the calibration equations providing density-independent and temperature-insensitive determination of the material moisture content and its density. The obtained values of the SEP compare favorably with the SEP achieved for the same material of constant temperature at 9.4 GHz [3]. The uncertainty in static determinations for shelled corn is ± 1.15% moisture and ± 25 kg· m - 3, moist density, at the 95% confidence level, for sample temperatures changing from 4°C to 45°C.

References: [1] A.W. Kraszewski, "Microwave Aquametry - Needs and Perspectives", IEEE Trans. on Microwave

Theory.fj Techniques, vol. 39, no 5, pp. 828-835, May 1991. [2] A.W. Kraszewski and S.O. Nelson, "Wheat moisture content and bulk density determination by

microwave parameters measurements", Ganad. Agric. Eng., vol. 34, no 4, pp. 327-335, Nov. 1992. [3] A.W. Kraszewski and S.O. Nelson, "Determination of moisture content and bulk density of shelled

corn by measurement of microwave parameters", J. Agric. Eng. Res., vol. 58, no 1, pp. 37-46, 1994.

ims workshop wmfb - truebner · ims workshop "wmfb" electromagnetic wave interaction with...

Documents