2014 PROCEEDINGS - International Microwave Power …impi.org/wp-content/uploads/2014/01/IMPI-2014-Proceedings_Final... · 2014 PROCEEDINGS. International ... Don’t forget to take

June 18-20, 2014

Doubletree Hotel, Canal StreetNew Orleans, Louisiana, USA

www.IMPI.org

Presented by theInternational Microwave Power Institute

PO Box 1140, Mechanicsville, VA 23111Phone: +1 (804) 559 6667 Email: [email protected]

www.impi.org

IMPIs48th Annual Microwave Power Symposium (IMPI 48)

ISBN: 978-0-956274748

International Microwave Power Institute, 2014

2014 PROCEEDINGS

International Microwave Power Institute, 2014 3

ChairRaymond Boxman, Tel Aviv University, Israel

MembersJon Binner, University of Birmingham, U.K.Georgios Dimitrakis, University of Nottingham, U.K.Ulrich Erle, Nestle R & D, USAJohn F. Gerling, GAE, Inc., USAMarilena Radoiu, SAIREM, FranceBob Schiffmann, R.F. Schiffmann Associates, Inc., USABecky Schulz, Corning Incorporated, USAPaolo Veronesi, University of Modena & Reggio Emilia, ItalyVadim V. Yakovlev, Worcester Polytechnic Institute, USA

ChairBob Schiffmann, R.F. Schiffmann Associates, Inc., NY, USA

MembersJustin Balousek, H.J. Heinz, PA, USASohan Birla, ConAgra Foods, NE, USAUlrich Erle, Nestle, OH, USARic Gonzalez, ConAgra Foods, NE, USAMarie Jirsa, Hillshire Brands, IL, USAJuming Tang, Washington State University, WA, USA

TECHNICAL PROGRAM COMMITTEE FOOD SCIENCE AND TECHNOLOGY PROGRAM COMMITTEE

Each year, IMPI brings together researchers from across the globe to share the latest findings in microwave and RF heating theories and applications, and this year we have an outstanding array of researchers in attendance. If you are not yet a member of IMPI, we strongly encourage you to con-sider joining onsite. IMPI membership connects you to microwave and RF academia, researchers, developers and enthusiasts across the globe. Talk to an IMPI member today to learn more about the value of joining our out-standing organization!

Thank you for joining us. We hope you learn, interact, and enjoy the compa-ny during the meeting. Dont forget to take the opportunity to participate in some of the many leisure activities and visit the unique venues New Orleans has to offer.

Special thanks to the following partners for printing these Proceedings:

IMPI wishes to express its gratitude to the following individuals:

Welcome to New Orleans for the 48th IMPI Symposium

Purchasing Information: Copies of the 48th Annual Microwave Power Symposium are available to individuals and organizations for purchase. Back issues of prior year symposia are also available for purchase. Contact Molly Poisant, Execu-tive Director of IMPI, at +1 804 559 6667 or [email protected] for more details.


Table of Contents

Systems, Components, Modeling and Safety

Automatic Impedance Matching in High-Power Microwave ApplicationsVladimir Bilik

2.45 GHz Solid State Microwave Generators: From Laboratory to Industrial Applications David Guillet, Marilena Radoiu and Louis Latrasse

Improving Microwave Oven Heating Uniformity Using Grid WallsRobert L. Eisenhart

Application of Slow-Wave Structures for RF and Microwave HeatingYuriy N. Pchelnikov and A. V. Mamontov

Controlling Microwave-Induced Temperature Distribution in a Wooden Load Through a Two-Source ExcitationAlexander V. Mamontov,Vladimir N. Nefedov and Yuriy N. Pchelnikov

Improving Microwave Cooking Performance by Source Phase ShiftingF. Gambato, F. Moro and M. Guarnieri

Aura-Wave Microwave-Assisted Plasma Source Using 2.45 GHz Solid State GeneratorLouis Latrasse, Marilena Radoiu and Bertrand Depagneux

Electrical Characteristics of a Plasma Excited by an Azimuthal Microwave Electric FieldD. Tsimanis and R.L. Boxman

Microwave Oven Injuries: Review of Actual IncidentsRobert F. Schiffmann

A Preliminary Investigation on a Solid-State Multi-Source Microwave OvenF. Bressan, M. Bullo, F. Dughiero

Food, Agriculture and Biology

Low Power Microwave Heating to Control Insect Pests on Tomato PlantsSandeep V. Gaikwad, Rajesh Harsh, A. N. Gaikwad and Anurag Gupta

A Crop-Ecology Based Assessment of Microwave-based Weed Management when Herbicide Resistance is PresentGraham Brodie

Effect of Microwave-Assisted Hot Water Treatment on the Quality of GrapefruitsNohemi Soto-Reyes, Aurelio Lpez-Malo and Mara E. Sosa-Morales

6

10

14

18

22

26

30

34

38

102

42

46

50


Table of ContentsComputer Simulation Analyses to Improve Radio Frequency (RF) Heating Uniformity in Dried Fruits for Insect ControlBandar Alfaifi, Juming Tang, Shyam Salbani, Shaojin Wang and Barbara Rasco

Microwave Synthesis of Sulfonated Cyanine Dyes for Use as BiosensorsMargaret E. Grow, Angela J. Winstead and Rachael Matthews

Optimized Microwave-Aassisted Extraction of Oils from Mango (Mangifera indica) Kernel Using Response Surface MethodologyDivine B. Nde, Dorin Boldor and Pranjali Muley

Microwave-Assisted Extraction of Black Pepper Essential Oil and its Performance as Antioxidant during FryingYoselin A. Snchez-Prez, Aurelio Lpez-Malo, Nohem Soto-Reyes and Mara E. Sosa-Morales

Microwave Processing of Food Products: Three Case StudiesRobert F. Schiffmann

Material Processing and Properties

New Tools in Biomedicine: iCrystal System and New Crystallization Platforms for Rapid Drug DevelopmentMuzaffer Mohammed, Anginelle Alabanza, Adeolu Mojibola, Kevin Mauge-Lewis, Gilles Dongmo-Momo, Taiwo Ogundolie, Mohammad Giwa, Tabassum Kabir and Kadir Aslan

An Artificial Neural Network Technique for Determining the Volume Fraction of Solids in Particulate MaterialsAlexander V. Brovko, Ethan K. Murphy and Vadim V. Yakovlev

Dielectric Spectroscopy Studies in Granular FlowSamah Yousif, Juliano Katrib, Olaosebican Folorunso, Paul A. Langston and Georgios A. Dimitrakis

Dielectric Properties of Beans at Ultra-Wide Band FrequenciesRichard Torrealba-Melndez, Jos Luis Olvera-Cervantes, Alonso Corona-Chvez, Mara E. Sosa-Morales and Nohem Soto-Reyes

Dielectric Properties and Thermal Conductivity of Peanut ButterSoon K. Lau and Jeyamkondan Subbiah

Microwave Douglas Fir Log Modification for Preservative TreatmentG. Torgovnikov and P. Vinden

Effectiveness of Microwave Lumber Modification Technology Application in IndustryA. Leshchinskaya

54

58

62

66

70

74

78

82

86

90

94

98


IMPI 48 MICROWAVE POWER SYMPOSIUM New Orleans, Louisiana, USA, June 2014

6

IMPIS 48 MICROWAVE POWER SYMPOSIUM New Orleans, Louisiana, USA, June 2014

(c) INTERNATIONAL MICROWAVE POWER INSTITUTE, 2014

Automatic Impedance Matching in High-Power Microwave Applications

Vladimir Bilik

Slovak University of Technology in Bratislava, Faculty of Electrical Engineering and

Information Technology

This paper is an overview of automatic impedance matching (tuning) in high-power industrial microwave applications. We first review the general theory and summarize the needs for and specifics of tuning in such applications. We compare the most commonly used types of matching devices (tuners) and explain the principle of operation of the automatic tuner (autotuner) based on the three-stub waveguide impedance transformer. We discuss basic characteristics of such autotuners. We analyze questions of optimal matching in the challenging signal conditions in which autotuners operate. We introduce the problem of tuning stub swapping and discuss possible methods of its suppressing. Keywords: Autotuners, impedance matching, industrial electronics, waveguide tuner. INTRODUCTION BASIC THEORY One of the means to improve productivity of industrial processes utilizing microwaves is employing a system of automatic impedance matching. The term impedance matching refers to a state in which the power Pa transmitted from a microwave generator to a load (applicator) is maximized [1], [2]. A basic matching arrangement and its equivalent circuit are shown in Figure 1. The load is defined by reflection coefficient L at plane T2. The generator is characterized by power PG deliverable to a load L = 0 and its internal reflection coefficient G at plane T1. To achieve the match, a lossless reciprocal two-port network with adjustable parameters, known as impedance transformer or tuner, is connected between generator and load. The tuner, defined by its scattering parameters Sij, transforms L to input reflection coefficient I, presented to the generator. Power

( ) 22 11 IGIGa PP = (1) absorbed in load is maximum when

= GI or, equivalently, = L2 (2)

i.e. when the reflection coefficients looking towards the generator and load are mutual complex conjugates (for this, impedance matching is also called conjugate matching). High-power magnetron generators contain circulators, making them well-matched sources with G 0. The input matching condition (2) then reduces to the known rule I = 0. Since by definition 2 = S22 for a matched source, the output condition becomes

= LS22 (3)



7



This is a useful tuning criterion: for a given L, adjust the tuner until its S22 satisfies (3). Because S22 is complex, a tuner must contain at least two adjustable elements.

Gen Tuner LoadWaveguide

T1 T2

GI Tuner(Sij)

1 2G2 GLGG

GG

GLPG

T1 T2

(a) (b)

Pa

Figure 1. Basic impedance matching arrangement (a) and its equivalent circuit (b).

IMPEDANCE MATCHING IN INDUSTRIAL APPLICATIONS An important motivation for impedance matching in industrial applications is increasing energy efficiency. Practical applicators are for various reasons poorly matched, often reflecting as much as 50% of the incident power. The reflected energy is wasted to heat. Since the installed power in production plants may be on the order of megawatts, reducing this power loss is imperative. An equally important benefit of impedance matching is shorter processing time. Other reasons are quality of production (e.g. stable plasma) and the need for extremely rapid volumetric heating in some applications.

Matching in high-power microwave applications differs from that in common electronics, in which the networks typically employ lumped elements or TEM lines, signal power is low and its spectrum wide, source and load impedances are frequency-dependent but unchanging in time. The matching is often broadband, accomplished by complicated but permanently tuned filters. In contrast, industrial applications use waveguides, power is very high, spectrum is relatively narrow, and the center frequency is variable and unknown. Load impedance is frequency-dependent and time-variable. The matching is narrowband, implying electrically simple tuner circuits, but these circuits must be adaptively adjustable, requiring advanced and rapid tuning algorithms.

Traditionally, the magic tee and the three-stub waveguide tuner are used for matching. The magic tee [3] is a 4-port hybrid junction with two ports terminated in sliding shorts, serving as tuning elements. The circuit behaves as a simple ladder network with variable series reactance and shunt susceptance. Theoretically, such circuit can match arbitrary load impedance. The peak working power is limited by the shorts. Contacting shorts are unsuitable due to the unreliable contact; the gaps of noncontacting shorts are prone to leakage and arcing. A drawback common to all tuning elements based on transmission line sections is substantially stronger frequency dependence compared to tuning stubs inserted into waveguides.

THREE-STUB WAVEGUIDE AUTOTUNER The three-stub waveguide tuner has been the most popular matching circuit due to its good electrical performance and compactness. A basic diagram of a three-stub automatic impedance matching device (autotuner) is shown in Figure 2a. The system consists of a vector reflectometer, a tuner, and a computer. The tuner consists of three noncontacting



8



metallic cylinders (stubs) inserted by means of stepper motors into the waveguide interior from the center of its broader wall. Electrically, a stub represents a capacitive shunt susceptance. Mutual stub distances are equal to a quarter of guide wavelength g at the highest working frequency fd. The matching principle is described e.g. in [2], ch. 5.3. Matching is accomplished by only two of the three stubs at a time (1+2 or 2+3), delineating two subareas of matchable reflection coefficients in the complex L plane (Figure 2b). For autotuning, an accurate tuner electric model is necessary in order to compute tuner S-parameters as a function of stub insertions and frequency.

1 3

Tuner

2

GLGI

Reflectometer

PrPG Pal g/4l g/4

f = fd

Gen

erat

or

Load

Computer(a)

1Re

j.Im

|GL|max321

2+3

1+2321

Gen

Load

0321

321

(b)

GL

Figure 2. A block diagram of a three-stub waveguide autotuner (a) and its matchable area (b).

The basic tuning procedure is a sequence of the following steps: (a) Measure I

and frequency f; (b) from known f and stub extensions, compute tuner S-matrix Sij; (c) using Sij and I, determine L; (d) use L in (3) to obtain tuner S22 minimizing |I|; (e) compute new stub extensions to realize the S22; (f) move the stubs to the new extensions.

The most important autotuner parameters are: Matching range. Expressed as the matchable area or, more simply, as the

highest |L| that can be matched regardless of phase (also shown in Figure 2b). Matching accuracy. Expressed as the residual input reflection coefficient after

setting the stubs to the determined extensions. Depends primarily on the reflectometer measurement accuracy and generally deteriorates with increasing load mismatch.

Matching speed. Expressed as the time elapsed from the start of a tuning cycle to the instant the match has been reached. Depends on the measurement time, computation time, and stub travel time. Measurement time must often be long (0.1 1 s) in order to smooth out fluctuations. Full travel time is typically 0.5 5 s. If only small corrections are needed, typical tuning cadence is 5 cycles/s.

Maximal working power. Limited by three factors [4]: electric breakdown (arcing) due to high local E-field; stubs overheating due to high surface current density; radiation leakage. Arcing is usually the most critical factor; its danger can be eased by reducing stub extensions (the penalty is smaller matchable area). The leakage can be suppressed by properly designed stub chokes.

SPECIFIC ISSUES Two specific autotuner issues of importance are optimal matching and stub swapping.



9



Optimal matching. Input to tuning algorithms is a single complex number L, expected to vary slowly due to the continued processing. In many applications, however, both L and magnetron power PG are modulated with repetition period Tr in the range 0.1 1 s, caused by mode stirrers or nonuniform objects moving through the applicator. Such rapid fluctuations are too fast for autotuners to handle, and hence a question arises as to what L value should actually be used. This problem was addressed in [5]. It was shown that the mean absorbed power Pa is very nearly maximized when the input to the tuning algorithm is the weighted average of L(t) over period Tr with the weight being the incident power PG(t). To configure a tuner accordingly, signal waveform analysis for each particular case should precede its employment. Failing to do so can lead to poor tuning and fluctuating data.

Stub swapping. If L crosses the boundary between the two matchable subareas (Figure 2b), an abrupt transient effect called stub swapping occurs in that one of the lateral stubs (1 or 3) is being greatly extended and its counterpart retracted. If L fluctuates across the boundary, stubs 1 and 3 will be incessantly swapped. During the transient the input reflection coefficient I is uncontrolled, which may lead to peaks in the reflected power reducing the overall energy efficiency of an industrial process. Two methods have been proposed in [6] to suppress the effect. A recent study [7] has shown that excursions of I can be eliminated by an appropriate control of stubs motion.

CONCLUSION Autotuners are advanced and sensitive devices, increasingly employed in the harsh conditions of production factories. To apply them successfully, a user should understand the basics of their operation, parameters, limitations, and correct settings. This overview attempted to contribute to such understanding. Further development includes improving the electric model of the tuner, handling of short pulses with high peak power, intelligent stub control to suppress arcing and glitches of reflected power, including additional features (arc and flood detectors, humidity sensors) as well as improving reliability and robustness for the work in environmentally challenging conditions.

REFERENCES [1] G. F. Engen, Microwave Circuit Theory and Foundations of Microwave Metrology, Peter

Peregrinus, London, 1992. [2] D. M. Pozar, Microwave Engineering 4th Ed., Wiley & Sons, New York, 2012. [3] R. E. Collin, Foundations for Microwave Engineering, McGraw Hill, New York, 1966. [4] V. Bilik and J. Bezek, High power limits of waveguide stub tuners, Journal of Microwave

Power and Electromagnetic Energy, vol. 44, no. 4, pp. 178-186, 2010. [5] V. Bilik, Optimum impedance matching of periodically time-varying loads, Proc. 2nd Global

Congress on Microwave Energy Applications, Long Beach, CA July 2012, pp. 405-419. [6] V. Bilik, Stub swapping in automatic three-stub impedance matching systems, Proc. 13th

COMITE Conf. on Microwave Techniques, Pardubice, CZE, April 2013, pp. 200-204. [7] V. Bilik, Optimizing tuning stubs motion in automatic three-stub impedance matching

systems, 2013, Proc. 14th AMPERE Int. Conf. on Microwave and HF Heating, Nottingham, Sept. 2013, pp. 222-225.



10


INTERNATIONAL MICROWAVE POWER INSTITUTE, 2014

2.45 GHz Solid State Microwave Generators: From Laboratory to

Industrial Applications

David Guillet, Marilena Radoiu and Louis Latrasse

SAIREM SAS, Neyron, France Present industrial microwave applications require more microwave sources, and those sources must meet stiffer demand in terms of the quality of the emitted power and ability to be integrated in complex systems. Most industrial microwave sources are magnetron based and thus have inherent inconveniences such as frequency shifting, spectral purity, and reproducibility from one magnetron to another, and large footprint. To meet industrial requirements, we developed the GMS - a new microwave generator based on solid state technology. Unlike magnetron-based generators which are using electronic tubes, the new solid state microwave generators use semiconductor devices, e.g. transistors. The advantages of the new technology particularly in terms of frequency control, power stability and spectrum quality will be reported and compared to the magnetron technology, Furthermore, GMS has a built-in autotuning feature based on the controlled frequency adjustment from 2425 MHz to 2475 MHz that allows manually or automatically changing of the frequency to obtain the maximum power transfer to the load by keeping the GMS and the load matched even if the load changes during processing. Keywords: Solid state technology, GMS, autotuning, transistor, semiconductor INTRODUCTION

During the past decade the high frequency (HF) power market was driven by the

telecom requirements. HF power transistors were to improve telecom equipment and to suit these specific applications; consequently, transistors were far from optimal for other applications like ISM. Unlike telecom applications which transmit information, ISM applications generally require continuous supply of microwave energy. Since the interest for ISM power sources is growing the situation of HF transistors designed especially for ISM has been changing slowly. Medical and plasma applications for instance can use the advantage of the solid state generator technology which offers many advantages over magnetrons. At present a few HF transistors have been developed for the ISM market and their reported performances are very encouraging.

The use of HF transistors to replace the magnetron eliminates the need for high voltage and filament power supplies inherent to magnetron generators thus, significantly



11



reducing size. This allows the development of much lighter and more compact products, both useful in the development of microwave-assisted mobile equipment. As it does not use a filament/cathode (known to be frail) like the magnetron, this technology is also much more robust in terms of mechanical stress and lifetime.

At present the power level of existing HF power transistors is limited to a few hundred watts but this level is constantly increasing due to the technology and research done by the major developers of the market. ISM high power HF transistors differ from the telecom transistors: HF transistors are used to transfer power to well-defined loads (like antennas) but also to energize changing loads such as plasma. The changing load cannot be handled by telecom transistors which are designed to drive an optimized antenna a very stable load compared to plasma, for example. From a thermal point of view, ISM applications like microwave plasmas often require continuous wave (CW) mode, and hence the equipment has to be designed to run smoothly at full power in CW. If needed, the pulsed operation can be also achieved by using a proper power control loop.

This paper presents the design of the GMS solid state microwave generator, and its performance in comparison with the magnetron tube technology. Advantages of the GMS generator with applications to microwave-sustained plasma sources and microwave-assisted chemical reactors for fundamental research will be presented. RESULTS AND DISCUSSION

Generally speaking, the spectrum quality of solid state generators is considerably

superior to that of a magnetron generator as shown in Figure 1. The magnetron is an electro-mechanical device whose frequency is mainly dependent on its geometry. While in operation, the frequency of the magnetron shifts due to metal dilatation (usually, cathode = thoriated tungsten, anode = copper) and therefore the frequency value depends highly on the temperature, and equally on the output microwave power. These variations may create problems especially when the magnetron is used to deliver power to high quality factor cavities/applicators, for example when the frequency shifts beyond the resonant frequency of the cavity in which the treatment takes places.

The change of the output power of a magnetron also impacts the frequency spectrum quality. Magnetrons are known to produce an acceptable spectrum between 10 % of their nominal power (maximum power) up to full power. Due to the fact that solid state generators act like amplifiers, they are able to produce an excellent frequency spectrum over the full range of power from the very first watts.

In addition, GMS takes advantage of the extremely precise control of the frequency to have a built-in autotuning system that keeps reflected power as low as possible by adjusting the frequency to the cavity [1]. When powering on the GMS the optimum match can be found in a very little time (less than 10 ms) and the frequency will be constantly adjusted to the cavitys response to maximize the energy transfer from the microwave generator to the load in the cavity. This way, even when the load changes over time the good matching can be achieved and the power transfer maximized.



12



2.45 GHz Magnetron spectrum*

600 W 1200 W

2.45 GHz solid state generator 200 W

Figure 1. Microwave spectra obtained with magnetron tube and solid state generator

* using a switch mode very low ripple HV power supply

This property was successfully applied and tested in two applications: a

microwave-assisted chemical synthesis reactor and a microwave-assisted plasma ECR source. The microwave reactor was successfully used in fundamental research in experimental batch or continuous flow chemical applications as well as in validating modeling results for graphite powder [2, 3].

In connection with plasma applications Figure 2 argon microwave plasma is created by the Aura-Wave, an electron cyclotron resonant coaxial plasma source powered by the GMS generator. Similarly to the chemical reactor, the concept of Aura-Wave was designed around the GMS to allow a wide range of operation conditions (pressure, gas type etc.) without any matching device. The autotuning feature of the GMS allows for significant extended operating conditions - unachievable otherwise, without any impact on the physical setup of the plasma reactor/source.

Figure 2. Aura-Wave plasma source



13



CONCLUSIONS

The GMS microwave generator using solid state (semiconductor) technology proved to be very promising not only because of the reported performances that makes it the almost perfect ISM microwave source, but also by the ease of integration in laboratory or industrial design due to:

- Compact size & light weight, microwave energy transmitted via coaxial cable; - Stable operation from microwave power levels as low as 0.5 W & power adjustable

in 1 W step; - Semiconductor technology, no magnetron and therefore longer lifetime & no high

voltage; - Very good frequency spectrum even at low power; - Built-in internal protection against mismatching and reflected power interlock; - Built-in isolator with automatic power reduction or switch off; - True RMS detector with linear measurement of reflected and forward power; - Very low ripple (0.2 % RMS); - Possibility to adjust manually or automatically the microwave frequency 25 MHz

from the central frequency 2450 MHz. Furthermore, for high microwave power requirements the use of multiple GMS

generators is envisaged; should this be the case, as each generator can be precisely controlled in terms of both microwave power and frequency, it creates many new opportunities in a wide range of applications requiring precise process control. Equally it allows simplifying some existing designs by removing many of the adjustments and increasing reproducibility which consequently reduces production costs. REFERENCES [1] Grandemenge, A., Jacomino, J.-M., Latrasse, L., Radoiu, M., Facility for microwave treatment of a load, WO 2012/146870 [2] Moon, E.M., Yang, C., Patel, M., He, H., Yakovlev, V.V., Microwave-induced temperature fields in graphite powder heated in a waveguide reactor, IEEE MTT-S Intern. Microwave Symp. Dig. (Tampa, FL, June 2014) (to be published). [3] Holmes, A.O., Murphy, E.K., Yakovlev, V.V., Scaling up reactors for microwave-assisted chemistry via ANN optimization, Proc. 14th AMPERE Conf. Microwave and High Frequency Heating (Nottingham, U.K., September 2013), pp. 316-319.



14

IMPI 48 - MICROWAVE POWER SYMPOSIUM New Orleans, Louisiana, USA, June 2014


Improving Microwave Oven Heating Uniformity Using Grid Walls

Dr. Robert L. Eisenhart

Eisenhart & Associates

This paper demonstrates a way of changing the reflection characteristics of the walls of a microwave oven to smooth out the 3-Dimentional standing waves, i.e. the nulls and peaks of the E-field. Simulated results using an Electromagnetic Field Simulator computer program (HFSS) are displayed, showing that considerable improvement in the uniformity is possible and could be particularly useful in industrial applications. Keywords: Electromagnetic fields, Electromagnetic modeling, Electromagnetic Heating, Microwave ovens, HFSS Modeling.

INTRODUCTION

Typical microwave ovens are designed with flat metal walls, the result of which are 3-dimensional modal patterns in the electric field, contributing to uneven heating (cooking) of food. To smooth out the heating characteristics a rotating turntable is utilized. While this does provide better average heat distribution there still is significant variation in the cooking.

This paper presents some fundamental ideas that can be applied to creating a more uniform electric field distribution within a microwave oven. The contributions are: 1) Modify the wall(s) to be polarization selective so that vertically and horizontally

polarized E-fields will be reflected differently, so that when integrated, the waves will produce a more uniform 3-dimensional electric field profile.

2) Excite the oven with dual polarization with respect to the rectangular cavity to take advantage of the reflective differences.

3) Offset the input position of the excitation aperture to optimize the E-field uniformity. FUNDAMENTAL CONCEPT FOR DESIGN

The basic electric field distribution within a microwave oven is a complex set of standing waves, primarily determined by the positioning of the metal walls and resulting in a large variation in field values between the nulls and peaks. This causes uneven cooking in the food. Many patents have proposed multiple sources, movable elements etc. to try and create more even field distribution without much success because the position of the static walls are the dominant factor. Therefore, the main method in use today to smooth out the cooking effect is to have a rotating table so the cooking averages within the field due to moving the food over time.



15



The proposed concept here is to have grid walls replacing two of the walls, these grids providing different reflection depending upon the polarization of the incident wave, essentially creating standing waves in two different positions depending upon the wave polarization. This allows taking advantage of the fact that when standing waves of separate horizontal and vertical waves are properly positioned with respect to each other (offset by wavelength/4) and on the same axis, the energy sum is a constant, i.e. uniform.

It should be understood that this only works with respect to one reflecting surface (wall), and that the energy distribution within a microwave oven is also highly dependent upon the size, type and position of the food placed inside. Therefore, it is not suggested that there will ever be a perfectly uniform field including the food. However, it makes sense that if the field distribution prior to introducing food is much more uniform than for an uneven distribution, then it is likely that cooking with the uniform distribution will result in a more uniform result than for the course distribution. It would still be worthwhile to include the rotating table to additionally average the heating.

To make this concept work, the microwave source must provide both polarizations to the oven cavity. This is done by tilting the input waveguide 45 degrees.

Secondly, knowing there will be waves scattered all throughout the cavity, particularly when food is inside, it makes sense also to use the grid on two walls, contributing to the smoothing of the energy distribution.

A third aspect of this approach is to consider the variations in distribution as a function of positioning the source at various locations inside of the oven. Only horizontal offset was considered here although vertical movement could also be reviewed.

DETAILED DISCUSSION

Consider the E-field of a typical microwave oven. Fig. 1 shows the field distribution through the horizontal mid-cut plane in perspective and top down views. This oven is 15.5w x 15.5d x 8.25h inches with flat metal walls on all sides, and the input source waveguide is centered in the sidewall and is 1.7 inch x 3.4 inch (WR 340). Frequency is standard at 2.45 GHz. The blue represents near zero field and the red is the peak value, with the green as a midrange value.

Fig. 1. This typical type and size microwave oven makes no effort to even out the fields. It does, however, contain a rotating table.

Fig. 2. Using a few modifications to the oven design will result in a much more evenly distributed E-field.



16



Fig. 2 shows the E-field distribution of the proposed design. Not only are the

values of the nulls and peaks less extreme than for Fig. 1, they are much closer together spatially providing better uniformity (green) of heating through thermal conduction.

There is a lot more to see in this figure 2. In the picture on the left side, the perspective view shows the two back walls have vertical grids which reflect vertically polarized waves. The horizontally polarized waves pass by these grids and reflect off the outside walls. Both horizontal and vertically polarized waves are excited by the source waveguide which is tilted at 45 degrees. The right hand picture shows clearly the offset of the wave source (2.5 inches), the position chosen empirically through simulation to create the most even distribution. Also you can see the lines of the vertical grids on the top and right hand walls in this picture. Next let's address the issue of combining the standing waves of horizontal and vertical waves. Fig. 3 shows the two standing waves with a quarter wave offset, determined by the positioning of the grid in front of the metal outer wall. Using the grid position as a reference, the two waves appear as rectified Sin and Cos functions. Remembering that the waves are orthogonal to each other, the total field is given by: = ( ) + ( ) (1) Which leads to: () + = 1 (2) The result is a constant with respect to time and position.

This all sounds fine but let's see if it will actually work. Fig. 4 shows the HFSS simulations of two plane waves, vertically and horizontally polarized, incident upon a vertical grid, resulting in two standing waves offset by a quarter wavelength. This is just a top view of the waves of Fig. 3.

Fig. 3. The grid reflects the vertically polarized wave while letting the horizontally polarized wave pass through.

Fig. 4. This shows the same effect as in Fig. 3. with the simulated standing wave E-fields for two incident plane waves

The left picture in Fig. 5. shows two cut planes through the combination of

vertically and horizontally polarized waves when both waves are reflected from the same surface. The result is a strong standing wave pattern on axis. However, when the



17



reflection of the vertically polarized wave is made with a polarized grid at a quarter wavelength in

Fig. 5. Cut planes showing simulations of the total E-field from the combination of the vertically and horizontally polarized waves in an oversized waveguide for different end-wall conditions.

front of the wall, the resulting combination is different, as shown in the right picture. Here we see very uniform field strength region along the axis due to the complementary nature of the two waves. This design can be used in specialized cases to create a very large uniform heating zone. Interference nulls due to side wall reflections are only near the side walls.

CAN IT BE BUILT?

Incorporating the suggested design is actually pretty simple. The main change is the addition of the vertical grid in front of two walls. Multiple mechanical design approaches have been run in electromagnetic simulations with the same success. The source waveguide is tilted 45 degrees to create dual polarization.

CITING PREVIOUS WORK

Although multiple patents exist with techniques to improve the field distribution in a microwave oven, none were found that related to changing the wall reflections.

CONCLUSION

This paper has proposed and demonstrated through the use of HFSS, a method and steps to make the E-field within a microwave oven more evenly distributed. It is reasonably speculated that starting with a more uniform field prior to introduction of food in the oven will result in a more uniform heating of the food, which is the desired goal.



The technique discussed focuses on redesign of two of the walls of the oven to alter the reflection characteristics. The discussion deals with only one small size oven but could easily be applied to any other using microwaves for heating, including industrial ovens.



18



Application of Slow-Wave Structures for RF and Microwave Heating

Yu. N. Pchelnikov1 and A. V. Mamontov2

1 Slowaves, Inc., Cary, NC, USA

2National Research University Higher School of Economics, Moscow, Russia Wide application of radiofrequency (RF) and microwave (MW) heating of dielectric materials is restrained by the relatively small specific RF losses and inhomogeneity of the MW energy penetration in the treated objects. These disadvantages may be overcome by using applicators based on slow-wave structures which support modes in which the phase velocity is less than the free-space velocity of light. These structures concentrate the electromagnetic field relatively homogeneously along the system axis. Diverse applications have been found including food heating, disinfecting agricultural products, and electro-coagulation.

Keywords: Applicator, hybrid wave, microwave heating, radiofrequency heating, slowing factor, slow-wave structure.

INTRODUCTION Heating technology based on the application of slow-wave structures (SWSs) is described in this paper. SWSs combine properties and advantages of transmission lines and lumped elements, concentrate electromagnetic energy in a given volume, and distribute it homogeneously in the propagation direction, enabling their use with RF and MW sources.

Previously, SWSs were used mostly as delay lines [1] and as interaction circuits in MW vacuum devices, and their properties were studied for these specific applications. Spreading industrial, medical, and military MW applications have encouraged study of SWS properties [2]. Analysis, experiments and practical realization have shown that the SWSs have many previously unknown properties, which can be used for creating novel technologies for measurements, domestic and industrial heating, plasma generation, diagnostics, and physiotherapy [2, 3]. This paper briefly describes the advantages and potential of SWS application for heating.

DEFINITIONS AND PROPERTIES

Transmission lines in which electromagnetic waves propagate with a phase velocity

vp less than the light velocity c in free space are called SWSs. The ratio N=c/vp is called the slowing factor. In general, SWSs are formed by two electrodes, one of which, the impedance electrode, is formed by a periodic row of transverse conductors connected in series in the direction of the wave propagation [2]. Although slow waves can propagate



19



without radiation along one impedance electrode, a second screen electrode is used for simplifying the wave excitation and screening its field, if needed.

The main property of slow waves is the proportionality of the energy concentration in the longitudinal direction to the slowing factor N and its concentration near the impedance conductor surfaces, i.e., the energy is concentrated in a small volume with a characteristic length much less than the free space wavelength. The same energy distribution in the transverse direction can be achieved at different values of frequency by changing N, or vice versa, the energy concentration at constant N can be changed by changing the frequency.

In the most cases, the boundary conditions on the impedance electrodes are satisfied only by a sum of TM- and TE-waves [4], which exist in the slow-wave structures only together and have the same phase and group velocities. At the same time, these waves behave quite differently at boundaries of dielectric and magnetic materials as well as at anisotropic surfaces. For example, the distribution of the TE-wave in the transverse direction insignificantly depends on dielectric objects placed near the impedance electrode, while the TM-wave distribution changes significantly. The depth of the TE-wave penetration in the absorbing or lossy materials significantly exceeds that of the TM-wave. This effect relates to the transverse distribution in the SWSs and has no relation to the incident waves [5].

Unlike the TM- and TE-waves in waveguides, the TM- and TE-waves in SWSs differ by the electric and magnetic energy stored in each of them - the difference exceeds N2 times [2]. This property, together with the anisotropic properties of slow waves and the high concentration of energy, makes it possible to significantly increase the heating effectiveness for materials with various electromagnetic properties including high temperature plasma.

Volumetric objects can be heated relatively homogeneously by distributed "pumping" of energy from the slow wave into the treated object. This occurs when the phase velocity vp in a SWS exceeds the light velocity c in the surrounding material or object [6]. In the absence of an object, the intensity of the electric and magnetic fields near the open surface of the applicator impedance electrode decreases exponentially with the distance to this surface. In the presence of an object placed parallel to the impedance electrode and separaged from it with a gap, and implementation of the inequality

cvp > , a part of the wave energy radiates into the object as a plane wave propagates at an angle to the impedence electrode, depending on the difference in the velocities [6].

APPLICATORS BASED ON COUPLED SLOW-WAVE STRUCTURES

Coupled SWSs are formed by two parallel impedance electrodes with mirror configurations (Figure 1). Unlike SWSs with one impedance electrode, the coupled SWSs can be excited in different modes: in-phase and opposite-phase. These modes are characterized by different distributions of TM- and TE-waves in the transverse direction. In the opposite-phase mode, the TM-wave, having most of the electric energy, is concentrated between the impedance electrodes with a relatively homogeneous distribution, and the object to be treated may be placed there.



20



(a) (b) (c)

Figure 1. Coupled SWSs: (a) meanders shifted at half-period, (b) coupled helices, (c) coupled radial spirals.

The use of coupled impedance electrodes approximately increases the electric field intensity fourfold. Both relative homogeneity and increased intensity are achieved by superimposing the exponential distributions of the fields excited by the impedance electrodes.

PRACTICAL IMPLEMENTATION

SWS-based applicators were industrially applied at 27, 40, 915, and 2,450 MHz. A diaphragm waveguide was used for ground meat heating [7], a radial comb was used in a portable microwave oven [8], coupled meanders were used for heating food, disinfecting seeds and tubers in agriculture, fabricating wood boards, drying wood stokes, and defrosting food fat [9]. SWS-based applicators were also used for thermal processing of flat dielectric materials, for drying textiles, transfer printing, hardening concrete slabs, and surface treating semiconductors. Some proposed and realized SWS applicators are shown in Figures 2 and 3.

(a) (b) (c) Figure 2. SWS-based units for heating (and their corresponding patent numbers): (a) RF chamber

with three parallel impedance electrodes with a meander configuration, (b) transfer printing apparatus, (c) chamber with two coupled meanders.



21



(a) (b) (c)

Figure 3. SWS-based heating devices with patent numbers: (a) chamber evaporating chemicals, (b) applicator for melting road coating , (c) applicator for treating flat dielectric material.

In a significant break-through, RF and MW heating was applied to medical

applications using SWSs. Applicators, radiators, and coagulating scalpels were proposed and fabricated. MW radiators for internal and external therapy were approved and manufactured for clinical application.

CONCLUSION

SWS technology is characterized by high efficiency, simplicity and flexibility. SWS-based applicators and radiators can operate in very wide frequency bands with dimensions significantly less that a wavelength in free space. They do not radiate in free space and can be used in the open volumes. This technology has been successfully used in the chemical, semiconductor, agricultural, medical and food industries. REFERENCES [1] D.A. Watkins, Topics in Electromagnetic Theory, Willy & Sons, Inc., N.Y., 1958. [2] Yu.N. Pchelnikov, Features of slow waves and potentials for their nontraditional application,

J. of Communications Technology and Electronics, vol. 48, no 4, pp. 450-462, 2003. [3] Yu.N. Pchelnikov and V.A. Kholodnyi, Medical application of surface electromagnetic

waves, Bioelectrochemistry and Bioenergetics, no 47, pp. 283-290, 1998. [4] L.N. Loshakov and Yu.N. Pchelnikov, TWT Theory and Gain Calculation, Soviet Radio,

Moscow, 1964 [in Russian]. [5] Yu.N. Pchelnikov, Anisotropy of a semiconductor film in the field of slow wave, J. of

Communications Technology and Electronics, vol. 39, no 10, pp. 66-69, 1994. [6] Yu.N. Pchelnikov, Radiation of slow electromagnetic waves in a magnetic insulator, J. of

Communications Technology and Electronics, vol. 40, no 6, pp. 25-30, 1995. [7] Yu.S. Archangelskiy and I.I. Devjatkin, Microwave Heating Devices for Raising Rates of

Technological Processes, Saratov State University, Saratov, 1983 [in Russian]. [8] I.I. Devjatkin, G. V. Lysov, V. N. Makarov, Microwave oven on a traveling wave, Electron

Industry, no 7, pp. 27-28, 1983 [in Russian]. [9] A.Yu. Mitskis, Engineering Repots, Vilnius's Teacher Training Institute, 1981-1989 [in

Lithuanian].



22



Controlling Microwave-Induced Temperature Distribution in a

Wooden Load Through a Two-Source Excitation

Alexander V. Mamontov1, Vladimir N. Nefedov1,

Yuriy N. Pchelnikov2

1National Research University, Higher School of Economics, Moscow, Russia 2Slowaves, Inc., Cary, NC, USA

Obtaining even temperature distribution in the volumetric materials with dielectric losses exposed to the microwave radiation was addressed in this work. Two types of radiators, (1) with a rectangular aperture and (2) with a teardrop aperture, were used for heating a multi-layer wooden bar. The temperature distribution in the bar was measured in the direction of radiation. With a rectangular aperture, the maximum temperature reached inside the bar was at some distance from the surface of the bar facing the radiator. With a teardrop aperture, the maximum temperature was at the surface of the bar facing the radiator. It's expected that an even temperature distribution in a treated object can be obtained by simultaneous using both types of radiators

Keywords: electric field intensity, microwave heating, microwave power, volumetric

material. INTRODUCTION Obtaining a uniform temperature distribution in volumetrically heated dielectric material is one of the most important tasks in creating industrially applicable microwave processes [1]. Often, microwave systems used for volumetric heating of dielectric materials, are fabricated in the form of a rectangular chamber with radiation sources connected to one or more generators. An open end of a rectangular waveguide working at the dominant, TE10, mode is the most simple microwave radiator. Positioned at different places of the walls, the radiators create a complex filed distribution, which significantly changes after placing an object to be treated in the chamber. Despite the success in computer modeling of the heating processes, measurement of the temperature distribution in a real object, can be very useful. This was the objective of preliminary experiments, described in this paper.



23



(a) (b) Figure 1. Microwave unit for heating a volumetric dielectric material. 1 - microwave chamber; 2 - absorbing material; 3 - support from a radio transparent material; 4 - wooden multi-layer sample; 5

- microwave generator; 6 - rectangular radiator; 7 - teardrop radiator. TEMPERATURE DISTRIBUTION MEASUREMENT SCHEME Experimental apparatus for volumetric heating of a homogeneous dielectric material by a single microwave power source with a frequency 2450 MHz, output power 600W, and two different radiators located on the upper wall of a 60x60x60 cm chamber is presented in Figure 1. The heated sample was a wood bar formed by 30 identical square boards of dry (9% humidity) pine wood with thickness 1 cm and length 20 cm. The temperature distribution was measured along the bar thickness. The bar was placed on a microwave-transparent stand at 24 cm below the radiator with its square side facing the upper wall of the chamber. The power reflected from the sample was not measured. Two types of radiators, with different apertures, were used for these measurements: (1) an open end of a rectangular waveguide (Figure 1a) and (2) a slot in the wide wall of a rectangular waveguide (Figure 1b).

The temperatures was measured without radiation by a thermo-couple moved in the direction of the wave radiation through a small hole previously drilled in the center of the bar. The estimated measurement accuracy (including errors caused by positioning of the thermo-couple) did not exceed 1C. To exclude the effect of power reflected from metallic walls of the chamber, they were covered with the standard absorbing plates with thickness 2 cm. In the same time, results of measurements were practically the same in the presence and in the absence of the absorbing plates.

The first open waveguide radiator (1) operated with the TE10 mode and created radiation with electric field approximately parallel to the irradiated surface of the bar (Figure 1a). The second radiator was fabricated as an end section of a standard rectangular waveguide, shorted at one end, and with a teardrop slot through the middle of the wide side of the waveguide, operating with the dominant, TE10 mode [2].The slot has length 15 cm and the width 3.6 cm in its widest part.



24



(a) (b)

Figure 2. Relative temperature distribution inside the wooden sample at different values of moisture; (a) irradiation through the rectangular aperture, (b) irradiation through the teardrop

shaped aperture; T0 = 800C. TEMPERATURE MEASUREMENTS The measured temperature distributions in the direction perpendicular to the surfarce of the wooden bar facing the radiator is presented in Figure 2a for irradiation by the rectangular aperture and different values of moisture. It is seen that in this case the highest temperature occured in a layer inside the bar. This effect is known and was discribed previously [3]. Curves in Figure 2b present the temperature distribution obtained with teardrop aperture. It is seen that independently on the moisture content, the maximum temperature was at the surface of the bar.

It follows from comparing curves in figures 2a and 2b that the simultaneous use of both rectangular and teardrop aperture radiators can flatten the temperature distribution along a treated sampl. In practice, the effect of simultaneous heating by both radiators can be controled by the choice of a power or operating time for the MW source feeding one of the radiators. RESULTS The curves in Figure 3a were obtained for the same wooden bar subsequently exposed to the rectangular aperture radiator (curve 1) and teardrop aperture radiator (curve 2). The operating time for each exposure was chosen to achieve the maximum temperature of approximately 800C. Taking into account the temperature distribution superposition, one can expect that simultaneous heating by two microwave sources with the rectangular and teardrop apertures gives a uniform temperature distribution in the material



25



(a) (b)

Figure 3. (a)Temperature distribution inside the wooden sample, (b) possible realization of the irradiation by two radiators with the teardrop and rectangular apertures.

with dielectric losses. It is demonstrated by curve 3 in Figure 2, calculated by superposition of curves 1 and 2. It is seen from results of measurements obtained for each radiator, that at the simultaneous use of both radiators, the temperature deviation in the material heated to 80-900 may be limited by 3-50C. A possible positioning of the radiators is shown in Figure 3b. CONCLUSION A method for obtaining an even temperature distribution in the volume of dielectric objects was proposed and investigated. This method is based on the superposition of complementary temperature distributions in the material under the condition that the object is heated by two sources with different polarizations of the excited field. It is expected that in some cases, a practical realization of this method will improve the quality of MW heating. ACKNOLEGEMENT This study had been carried out under the Scientific Fund of HSE in 2013-2014, project 12-01-0136. REFERENCES [1] A. C. Metaxas and R. J. Meredith, Industrial Microwave Heating, (Power and Energy Series

4), 1988. [2] Morgan M.A. Finite element calculation of microwave absorption by cranial structure.

Electromagnetics, 1981, vol.1, 3, pp.309-327. [3] Y. V. Karpenko and V. N. Nefedov. Machines for microwave heating of asphalt-concrete

coatings. Motor roads, M: Issue 1, 1997.



26



Improving Microwave Cooking Performance by Source Phase Shifting

F. Gambato, F. Moro and M. Guarnieri

Dipartimento di Ingegneria Industriale, Universit di Padova, Padova, Italy

Microwave cooking is rapid, but non-uniform. The effects of thermal non-uniformity are hot spots, cold spots and microbiological safety issues. An approach, based on phase shifting of two microwave generators, is proposed in order to attain a more uniform temperature distribution without the need of moving parts. In this paper 3D FEM electromagnetic and thermal analyses are combined with a metamodel into an optimization method. The effectiveness of the proposed method is verified on a test case model of practical interest.

Keywords: Optimization, response surface, uniform heating, food safety, microwave oven.

INTRODUCTION

The challenge for the fast food market is to provide quality food in a hurry and to this end high speed cooking equipment with microwave (MW) heating is essential. MW heating is non-uniform mainly because of the inherently uneven distribution of the electromagnetic field inside the oven cavity [1]. Moreover, the energy absorption process is strongly affected by shape, size, dielectric properties of materials, temperature distribution, position of the workload, as well as by the cavity geometry and dimensions.

Several methods of making the temperature more uniform, such as hardware-based methods (e.g. mode stirrer) and modeling-based optimization ones [2] have been developed. In this paper a model-based method, which does not involve moving parts, is presented. It applies when two MW sources are provided, which can be fed by phase-coherent currents in order to produce MW destructive and constructive interference zones, thus reducing the temperature of the hot spots and increasing the temperature of cold spots. The effectiveness of the phase shift method is studied with a 3D coupled FEM analysis on a realistic test case model. The optimal phase shift for uniform cooking is estimated by a metamodel based optimization (MBO) approach in order to limit computing costs.

MODELING AND SYSTEM CONFIGURATION

The MW oven geometry, presented in [3] and shown in Fig. 1, corresponds to a test case model of practical interest. The MW heating system is fed by two MW generators and the workload consists of a cylindrical vessel made of Pyrex containing 250 cc of distilled water, which is represented by a solid volume neglecting fluid dynamics. The dielectric properties of water and Pyrex considered in the model are



27



Figure 1. Configuration of the MW heating system.

reported in [3] and [4]. The model also includes both excitation currents and waveguides. MW generators are modeled as antennas fed by phase shifted sinusoidal currents at 2.45 GHz, which is a typical frequency used in domestic appliances. In the numerical model two physical phenomena, i.e. electromagnetic wave propagation and heat transport, are coupled together by the thermal effects of MW energy deposition and the temperature-dependent material parameters. The coupled problem is solved by means of a FEA commercial software (COMSOL).

OPTIMIZATION

The optimal phase shift for uniform cooking is obtained by MBO in order to limit the number of FEM analyses which are computationally demanding. Generally speaking, MBOs consist in running the simulations at a set of points (experimental design) and fitting the response surface (RS) [5] to the resulting input-output data, so that a metamodel (MM) is obtained [6]. The MM provides an approximation of an objective function (OF) and can be used for optimization. Time is saved by running in parallel the computation to fit the surface. These runs can be started even before stating the optimization problem and can be used again if this problem is modified (e.g. setting different objective function). Finally, the RS provides a way to compute a map between input and output, which can be used for robust design.

In our model, the key steps of the MBO algorithm are as follows: step 1. Design of experiments (DOE). The selected DOE scheme is a uniform sampling in the phase shift . Each simulation, corresponds to a different phase shift (selected location). The initial set of the design variable is [ 0, 3,2 3,, 4 3,5 3,2 ] rad. step 2. Numerical simulations at selected locations. 3D coupled FEM simulations are executed in parallel for all phase shift values as designed in the DOE. step 3. Evaluation of the objective function at selected locations. The OF is defined as the time average over 60s of temperature variances, i.e.

mean T2 ( )( ) (1)

where T2 ( ) is evaluated every second by sampling the temperature at 995 points of

a Cartesian 3D mesh uniformly distributed within the water load.



28



step 4. Construction of metamodel. From the set of data points (step 3) a MM is generated that provides a good approximation of the OF between data points [7]. The MM is an interpolation function, that is a linear combination of a non-parametric Gaussian radial basis functions (GRBF):

MM ,( ) = i e2 i

2

i=1

NDOE

, (2)

where coefficients i are determined by enforcing interpolation conditions on (2). DOEN is the number of selected locations, and is the shape parameter, whose initial value is evaluated on a test function. step 5. Metamodel optimization. The MM function (2) is minimized with respect to providing

best = argmin 0,2[ ]

MM ,( )( ) . (3) step 6. Convergence condition. 3D FEM simulation is executed at best . Equation (1) is computed and the discrepancy with the MM is estimated. The MBO algorithm is said to converge when the discrepancy is within a given tolerance, e.g., 10 3. step 7. Metamodel-error optimization. The set of selected locations is updated with best (step 5). The MM is updated with the new data point (step 6) and a more accurate is obtained by minimizing the root-mean-square error (RMSerror ):

RMSerror =1K

MM j,*( )MM + j,( )"# $%

2

j=1

K

(4)

where j are the evaluation points, K is the number of j , * is the last computed

shape parameter, and the superscript + indicates the MM update. The best value of (4) is:

( )( )

errorbest RMS%10*

minarg

= . (5)

The procedure is repeated from step 4 until convergence. The MBO algorithm finds the optimal phase shift opt and shape parameter opt ,

which produce the most uniform heating. The optimization algorithm has been implemented in the MATLAB environment. In particular, the optimization problem (steps 5 and 7) has been solved by the golden section search with successive parabolic interpolation, implemented in the fminbnd MATLAB function [8]. In order to overcome the local minima problem, the search interval has been divided into smaller search intervals.

RESULTS

The effectiveness of the proposed method in finding the optimal phase shift has been verified by COMSOL FEM analysis. The OF value is 195.24 when 0= rad,



29



whereas it is 175.58 if 915.4=opt rad. Fig. 2 shows the temperature patterns, after 25 s,

on a vertical cross-section for 0= rad (Fig. 2a) and opt = (Fig. 2b). It can be noted

(a) (b)

Figure 2. Temperature patterns in a specific vertical plane of the workload after 25s: at 0= rad, ( )( ) 24.19502 =Tmean (a); 915.4=opt rad, ( )( ) 58.1752 =optTmean (b).

that the phase shift parameter variation significantly affects the temperature field. In fact the heating is more uniform and the hot spot temperature is lower in Fig. 2b.

CONCLUSION

A metamodel-based technique for solving the problem of MW heating uniformity has been proposed. Computer simulations show that the phase shift of MW power sources significantly affects heating patterns and hot spot temperatures. The particular choice of OF allows for properly assessing the heating rate and the spatial temperature distribution. The optimization algorithm uses the RS methodology with successive enhancement by selecting the next sampling point in the region of interest. The goal of this technique is the infill criterion, which is biased towards both local exploitation of promising basins of attraction and global exploration of the search space. Finally, the MBO method guarantees a good approximation of the FEM model and a limited computing cost since a small number of samples in the design domain is required. The next step will involve an experimental validation of the proposed method.

REFERENCES [1] R.J. Meredith, Engineers' Handbook of Industrial Microwave Heating, IET Power Series, vol.

25, U.K., 2007. [2] B.G. Cordes, E. Eves, and V. Yakovlev, Modeling-based minimization of time-to-uniformity

in microwave heating systems, Proc. 11th AMPERE Conf. on Microwave and High Frequency Heating, Oradea, Romania, 2007, pp. 305-308.

[3] F. Gambato and A. Morassut, Two Magnetrons - Microwave Ovens 3D Modelling and Simulations, Proc. Int. Conf. HES-13, Padova, Italy, 2013, pp. 219-225.

[4] COMSOL Multiphysics Material Library. [5] V.A. Mechenova and V.V. Yakovlev, Efficiency optimization for systems and components in

microwave power engineering, J. Microwave Power & Electromagnetic Energy, vol. 39, no. 1, pp. 15-29, 2004.



[6] N.V. Queipo, R.T. Haftka, W. Shyy, T. Goel, R. Vaidyanathan, and P.K. Tucker, Surrogate-based analysis and optimization, J. Progress in Aerospace Sciences, no 41, pp. 1-28, 2005.

[7] G.E. Fasshauer, Meshfree Approximation Methods with Matlab, World Scientific, Singapore, 2010.

[8] MATLAB users guide.



30



Aura-Wave Microwave-Assisted Plasma Source Using 2.45 GHz Solid

State Generator

Louis Latrasse, Marilena Radoiu, Bertrand Depagneux

SAIREM SAS, Neyron, France

Large-scale processing with high density uniform plasma is necessary for surface treatments that need highly uniform etching or deposition rates. To meet these industrial requirements, we designed a new ECR coaxial microwave plasma source called Aura-Wave with very high performances in terms of working pressure range, minimum plasma sustaining power and operating frequency range. With Aura-Wave plasma can be sustained from 10-4 mbar up to a few mbar, whatever the gas type, the minimum required sustaining power is only a few watts, and the plasma sources are matched over several decades of pressure. Furthermore, each plasma source is connected to its own microwave solid state generator. This transistor based microwave generator allows to produce a wave with variable frequency, thus the low mismatching created by a change in the operating conditions can be compensated automatically (the reflected power decreased to 0 W) by the variation of the frequency of the forward wave. The advantages of this new technology are reported in connection with the plasma scaling up requirements to distribute uniformly the electric field over large areas. The flexibility of plasma source distribution together with the accurate control of the microwave parameters of each individual plasma source allows to produce large, uniform, high uniform plasma without scale limitation. Keywords: Microwave plasma, Electron Cyclotron Resonance, Aura-Wave, Solid state microwave generator, auto-tuning frequency INTRODUCTION

Plasma scaling up requires distributing and applying a uniform electric field over large areas. In the low pressure range (103 102 mbar range), a recently proposed solution consists of producing large uniform plasmas from two or tri-dimensional networks of elementary sources sustained at electron cyclotron resonance (ECR). These multi-dipolar sources [1], where every elementary plasma source consists of a cylindrical magnet (magnetic dipole) fed with microwaves at the end of a coaxial line, can produce plasma densities between 1011 and 1012 electron/cm3. However, this kind of technology requires to have a good distribution of the microwave power between the magnetron tube



31



technology sources as well as the use of matching impedance devices that limits this technology to laboratory applications.

To meet industrial requirements, a new ECR coaxial microwave plasma source called Aura-Wave due to the shape of the created plasma, was designed This source avoided power-loss within the source and impedance matches over a wide range of operating conditions. Thus, no impedance matching system is required.

In parallel, a 200 W solid state generator was designed, which had very stable power and which allows to vary the frequency of the emitted wave function, for automatic impedance tuning [2]. The combination of the solid state generator and the Aura-Wave plasma source made it possible to control exactly the power transmitted to the plasma; the low mismatching created by changes in the limit operating conditions could be compensated automatically by the variation of the forward wave frequency, thus allowing to significantly extension of the operating condition range of the plasma source.

Aura-Wave was designed as an industrial plasma source and to be mounted on a KF DN40 flange Figure 1. An example of integration of Aura-Wave in a plasma reactor is presented below; the top of the reactor makes possible to integrate 16 Aura-Wave plasma sources in matrix configuration while a crown distribution is possible on the side.

Figure 1. Left: final design of the Aura-Wave ECR plasma source; Right: example of integration on the top of a plasma reactor distribution of up to 16 off x Aura-wave plasma sources in matrix

configuration and crown configuration RESULTS AND DISCUSSION

Operating conditions range of the Aura-Wave plasma source were investigated in several common gases like oxygen, nitrogen and argon. As shown in Figure 2, assimilated Paschen diagram, i.e. plasma breakdown requested power as a function of the pressure, and the minimum power requested to sustain the plasma are plotted in oxygen. Photos to the right show plasmas obtained at different operating conditions. O2 plasma can be sustained on a wide pressure range, from 104 mbar to a few mbar and with 2 W only - the third picture shows a 2 W microwave plasma. Moreover, the Aura-Wave plasma source is matched, i.e. no additional matching device is used and there is no

N-type connector

Water cooling

KF DN40 flange



32



reflected power at < 101 mbar. Generally 30 W is sufficient power to ignite the plasma at lower pressure. For pressure higher than 101 mbar, as the ECR heating influence decreases compared to the collisional heating, it becomes difficult to completely minimize the reflected power.

Figure 2. Minimum required power to breakdown (Paschen diagram) and sustain the plasma as a function of the oxygen pressure, pO2

Figure 3 shows the minimum power requested to breakdown at different

frequencies and to sustain the plasma as a function of the nitrogen pressure pN2. The plasma can be sustained from 104 mbar to 1 mbar at 5 W minimum microwave power. The lowest requested power to breakdown is obtained at 2.45 GHz on the entire pressure range. Aura-Wave was well matched between 103 mbar and 101 mbar and becomes difficult to tune at higher pressure. This was attributed to increased collisional heating outside the above nitrogen pressure range.

In argon, plasma could be sustained from 104 mbar to 10 mbar, at 1 W only over 3 decades of pressure. Aura-Wave was matched from 103 mbar to 1 mbar, 30 W microwave power was generally enough to ignite the plasma in this pressure range. At lower pressure mismatching could be easily compensate by varying the frequency.

Figure 3. Minimum microwave power needed to breakdown (Paschen curve) at several microwave frequencies and to sustain the plasma as a function of the nitrogen pressure pN2



33



CONCLUSIONS

The new Aura-Wave ECR coaxial microwave plasma source was designed to sustain microwave plasma over the range of several decades of pressure, 104 mbar to 1 mbar, from a few watt microwave power, whatever the gas. Equally, the coaxial plasma source was designed to avoid internal power-losses and has proved to be matched, i.e. no reflected power with no additional impedance matching system over 2 to 3 pressure decades, depending on the plasma gas.

The use of the solid state generator made it possible to control the power transmitted to the plasma with one watt increment and the variable frequency extended the operating conditions range.

Multi-sources can be used for scaling-up in crown distribution for volume plasma processing or in matrix distribution for planar plasma processing. The feasibility of creating a uniform plasma by using elementary plasma sources in matrix distribution was demonstrated a few years ago in the mbar pressure range [4], thus no difficulties are expected at 102 mbar because of diffusion. As each Aura-Wave has its own microwave generator, it is now possible to control exactly the transmitted power to each plasma source whatever the number. For example, in matrix distribution, the losses in uniformity due to edge effect of the peripheral plasma sources can be compensated by increasing their microwave power and thus increasing the surface of uniform treatment area.

Presently control software is developed for multiple Aura-Wave sources in a matrix configuration. In the first step, eight Aura-Wave sources will be set up on the plasma reactor and sixteen sources afterwards. Plasma parameters will be investigated using a generic deposition application on 300 mm diameter wafers. REFERENCES [1] Lacoste A, Lagarde T, Bchu S, Arnal Y and Pelletier J, Multi-dipolar plasmas for uniform

processing: physics, design and performance, Plasma Sources Sci. Technol. 11 pp.407412, 2002.

[2] Latrasse L, Radoiu M, Jacomino J-M, Grandemenge A, Facility for microwave treatment of a load, WO 2012146870.

[3] Bchu S, Bs A, Lacoste A, Pelletier J, Device and method for producing and/or confining a plasma, WO 2010049456.

[4] Latrasse L, Lacoste A, Sirou J and Pelletier J, High density distributed microwave plasma sources in matrix configuration: concept, design and performance, Plasma Sources Sci. Technol. 16 pp. 7-12, 2007.



34



Electrical Characteristics of a Plasma Excited by an Azimuthal Microwave

Electric Field

D. Tsimanis and R.L. Boxman

Tel Aviv University, Tel Aviv, Israel

Electrodeless discharge lamps are sought for improved lifetime and flexibility in formulating a lamp fill without mercury. In the present work, an Ar discharge contained in a cylindrical envelope mounted at the end of a cylindrical waveguide was excited by the circular TE01 mode produced by a small circular antenna driven by a domestic oven magnetron. The axial distribution of the azimuthal electrical field in the waveguide was sampled via a circular aperture in the waveguide wall and a short external monopole antenna and detector. Light emission from the discharge was measured with a spectrometer, and the temperature rise of the discharge envelope was measured at various locations. Heating of the envelope was azimuthally uniform, while the microwave electric field, the light emission, and the envelope heating decreased as a function of axial distance from the end of the envelope facing the source. With the Ar pressure maintained at 6.9 Torr by a mass flow controller and vacuum pump arrangement, it was found that 149 W of 2.45 GHz power was absorbed by the plasma, of which 12 W was radiated in the 200-800 nm range, and 105 W were dissipated in heating the envelope. The remaining 32 W are presumed to have been radiated outside of the spectral range of the spectrometer, or convected from the discharge via the vacuum pump. The plasma presented a reflection coefficient of = 0.84ej0.33 and a complex impedance of Z = ( j1113128+ ) towards the source. Keywords: microwave plasma excitation, electrodeless lamp, circular TE01 mode INTRODUCTION

The lifetime of high intensity discharge lamps is mostly limited by erosion effects and most of these lamps contain mercury. Several mercury-free non-LTE gas discharge lamps have been used or investigated. Dielectric barrier discharge (DBD) lamps were reported in which their UV and VUV radiation is converted by a fluorescent coating to visible light [1]. However, DBD lamps are not efficient. Simpson introduced an apparatus for exciting electrodeless mercury free lamps in which electromagnetic energy was generated by a magnetron and excited a plasma comprised of Ar gas and S vapor, the latter sublimated from solid S. The lamp was enclosed by a perforated metal screen and a circular flange. The system was impedance-matched to minimize frequency shift of the



35



magnetron. The metal screen around the lamp, connected to the common terminal of the magnetron, prevented microwave radiation leakage while permitting high intensity light output. The lamp was rotated about its axis to even out heating of the lamp envelope. [2]. Lutterbach presented a microwave lamp having a quartz envelope, whose inner surface was coated with metal halides. This lamp was also rotated inside a perforated screenError! Reference source not found.[3].

Boxman and Dikhtyar suggested using the circular TE01 mode to excitate plasma for lighting application in unpublished preliminary work. The plasma would be contained in a cylindrical envelope, whose lateral and distal surfaces are coated with a transparent conductive film. The electric field of the circular TE01 mode adjacent to a conducting wall is relatively low, and parallel to the surface. This arrangement should minimize envelope heating, and perhaps eliminate the need for the perforated screen and envelope rotation used in previous microwave lamps. Previously, measurements of the optical emission spectrum from an Ar discharge and simulations of the electrical field were reported [4]. However, the electric field was not measured quantitatively in this configuration, nor was the the Ar plasma electrically characterized.

The objective of the present work was to measure the circular TE01 electrical field and the excited argon plasma electrical properties. The axial distribution of the field, light emission, and power dissipated in heating the envelope will be presented and used to understand the power balance within the discharge. TECHNIQUES AND MATERIALS

A schematic diagram of the experimental setup is shown in Figure 1. A Samsung OM75P magnetron was fixed to the left end of a 160 mm i.d. Al waveguide. A 20 mm diam loop antenna was connected to the magnetron. The waveguide had a 10 mm wide axial slot on its lower end for diagnostic access, which was generally covered with Al tape. Plasma was excited in a Pyrex envelope, which was mounted on a shorting plate which was positioned near the right end of the waveguide. The envelope was filled with Ar gas, introduced into the envelope through a mass flow controller (MFC), and evacuated by a rotary vacuum pump. The pressure was monitored by a Thermovac TM101 gauge.

The circular TE01 mode inside the waveguide has only an azimuthal E-field. The field inside the waveguide was coupled by a 1 mm radius aperture to the outside, where a monopole antenna attached to a diode detector (S-Team DM213) was located. The L-shaped 1 mm diameter wire antenna (10 mm length, oriented parallel to the internal azimuthal E-field, and 35 mm length in the radial direction) was located 50 mm from the waveguide wall. The ratio between the maximum E-field at each axial position inside the waveguide and E-field at the detecting antenna input was found by CST simulation of a model which consisted of a waveguide excited by the loop antenna and a detecting antenna located in front of a 1 mm radius aperture, outside the waveguide. It was found that this ratio was 850. The E-field at the detecting antenna produced an input voltage at the diode detector which was determined by its interaction with the short antenna, and the voltage divider comprised of the antenna impedance and the detector input impedance.The diode detector output voltage was measured on an oscilloscope.



36



Figure 1. Schematic diagram of the experimental setup.

Light emitted by the plasma was measured via a Quartz window on the Pyrex chamber, a StellarNet CR2 cosine receptor placed 50 mm from the waveguide slot, and a StellarNet EPP-2000HR spectrometer, and was analyzed using StellarNet SpectraWiz v5.0 software. The axial distribution of the illuminance was measured by a MRD 3050 phototransistor, which could be positioned at various axial locations. RESULTS

Using the procedure described in previous section, the electric field inside the waveguide was determined as a function of axial position (defined in Figure 1). The electric field was sampled as a function of z at several points (discrete points in Figure 2). These measurements were inputted into the MATLAB curve fit tool cftool, which fitted them to a graph of maximum electric field inside the waveguide as a function of z asin(bz+c)+d. VSWR=17.7 was obtained as the ratio between maximum and minimum of the fitted curve. The reflection coefficient from the plasma was obtained using Equations 1 and 2 and was equal to 0.84ej0.33. The impedance of the Ar plasma load was found to be be Z = (128 + 1113j) fro

Documents

2014 PROCEEDINGS - International Microwave Power …impi.org/wp-content/uploads/2014/01/IMPI-2014-Proceedings_Final... · 2014 PROCEEDINGS. International ... Don’t forget to take