Superconductivity for Improved Microwave Ferrite Devices€¦ · Superconductivity for Improved Microwave Ferrite Devices VOLUME 9, NUMBER 1, 1996 THE LINCOLN LABORATORY JOURNAL 21

• DIONNE, OATES, TEMME, AND WEISSSuperconductivity for Improved Microwave Ferrite Devices

VOLUME 9, NUMBER 1, 1996 THE LINCOLN LABORATORY JOURNAL 19

Superconductivity for ImprovedMicrowave Ferrite DevicesGerald F. Dionne, Daniel E. Oates, Donald H. Temme, and Jerald A. Weiss

■ We have demonstrated a new class of radio frequency (RF) devices thatcombine low-insertion-loss superconductor circuits with variable magnetizationferrites. The devices are designed to prevent penetration of the superconductorby DC magnetic fields associated with the ferrite, thus preserving the low RFloss of the superconductor while allowing interaction of the RF magnetic fieldswith the magnetized ferrite. A prototype ferrite-superconductor phase shifterthat employs high-Tc YBa2Cu3O7 (YBCO) superconductor meanderlines hasproduced over 700° of differential phase shift at 9 GHz with a figure of meritexceeding 1000°/dB over a 2-GHz-wide frequency band. We have alsodemonstrated a three-port switching circulator by combining three low-losssuperconductor meanderline sections with alternating T junctions on a ferritesubstrate. This innovative adaptation of two important technologies promisesperformance advantages that include lower insertion loss and reduced size andweight, compared to conventional techniques. By optimizing ferrite propertiesfor use at cryogenic temperatures, we also expect to achieve lower switchingenergies and faster switching speeds.

S of high-Tc superconduc-tors in 1986, a major thrust in the develop-ment of these materials has been for microwave

applications. With increasingly refined thin-filmdeposition techniques, the radio frequency (RF) sur-face resistance of superconducting YBa2Cu3O7(YBCO) films at 77 K is now much lower than that ofconventional conductors, up to frequencies of at least100 GHz [1]. This achievement has been exploited todevelop a wide variety of miniature passive micro-wave components, as described in this journal in acompanion article by W.G. Lyons et al. In this articlewe present work on a new class of active planar RFdevices that exploit the low insertion loss of supercon-ductor circuits and the unique magnetic properties offerrite substrates. These devices, which include mi-crowave phase shifters and circulators, have applica-tions in a broad range of military and commercialmicrowave systems, and offer important size, weight,and performance advantages over conventional mi-crowave technology.

Ferrite components have been widely employed inmicrowave systems for many years. Adjustable phaseshifters, for example, are the principal beam-steeringcomponent of agile-beam phased-array radars such asthe ground-based Patriot and shipboard Aegis sys-tems. Ferrite circulators are used in radar systems toisolate transmitters from reflected energy and to di-rect return signals into receiver channels. Circulatorscan also be employed as switches in selective filterbanks. These devices are based on unique microwavephase-shift properties that are the result of gyromag-netic effects. The phase shift of an RF wave traversingthe ferrite can be controlled by applying a small mag-netic field to adjust the ferrite magnetization.

Superconductive RF circuits overcome the limita-tions of present-day planar circuit components by vir-tually eliminating conduction losses. To avoid exces-sive conduction losses, devices based on normal metalconductors such as copper or gold must be config-ured in waveguides rather than the more compactmicrostrip or planar stripline geometries. In our work


20 THE LINCOLN LABORATORY JOURNAL VOLUME 9, NUMBER 1, 1996

with superconductor microstrip circuitry, we havedemonstrated that miniature low-loss ferrite devicescan be realized, and these devices offer an attractivereplacement for bulky waveguide systems. As weshow, these devices provide dramatically improvedfigures of merit, or degrees of phase shift per dB ofloss, compared to conventional microwave devices.Furthermore, the reduced device size results in lowerswitching energies and shorter switching times.Devices based on high-Tc superconducting films op-erate at temperatures between 50 and 90 K, and thuscan utilize small, reliable, power-efficient closed-cyclecryocoolers that have been developed for operation inthis temperature range.

An advantage of operating ferrite devices at cryo-genic temperatures is the improvement in key proper-ties of the ferrite material. At low temperatures, forexample, ferrite saturation magnetization limits canbe increased substantially, extending the range of effi-cient phase-shift operation to RF frequencies higherthan those available at room temperature.

One of the technical challenges in realizing super-conductor ferrite devices is the development of com-ponent designs that avoid magnetic-field penetrationof the superconductor. Applied magnetic fields arecentral to device operation because these fields con-trol the ferrite RF properties. However, magnetic-field penetration of a superconductor can lead to de-terioration of the superconducting properties [2]. Fordevice designs in which significant magnetic fieldspenetrate the circuit, superconductivity would haveno advantage over the normal conductivity of copperat 77 K [3]. This article describes our development offerrite-superconductor devices that eliminate mag-netic-field penetration, thus preserving the favorablesuperconducting properties. The results of our experi-ments confirm that ferrite-superconductor devicescan be realized and they can operate with negligibleconduction losses in the microwave frequency bands.

Gyromagnetic Effects

Gyromagnetism is the basic phenomenon that en-ables magnetic control of the phase of RF waves ingyrotropic media such as ferrites. In applyingMaxwell’s equations to analyze the propagation of aplane wave in a magnetized ferrite medium, we must

first recognize that an RF component of magnetiza-tion is induced by the RF magnetic-field componentand that the associated RF permeability contains dis-persive and dissipative terms. These terms arise from atensor that results from gyromagnetic effects (ferri-magnetic resonance), but the permeability can be ap-proximated as a complex scalar quantity for the pur-poses of our device analyses [4]. Therefore, we definethe RF permeability as

µ µ µ= −′ ′′j .

When µ is introduced into the expression for thecomplex RF propagation constant Γ = +α βj in thefrequency regimes of interest, the absorption constantα depends primarily on ′′µ , and for ′µ ≈ 1 is given by

απ ελ

µ≈ ′′12

2,

where ε is the dielectric constant of the ferrite and λ isthe free-space RF wavelength. The phase constant βdepends primarily on ′µ , and for ′′µ ≈ 0 is given by

βπ ελ

µ≈ ′2

. (1)

The corresponding phase change φ of a wave travers-ing a distance L through the ferrite equals βL . Withthe approximation in Equation 1 we can express φ as

φ π ε µλ

≈ ′2L

. (2)

The RF absorption by the ferrite is usually character-ized by the magnetic loss tangent

tan δµµµ = ′′

′. (3)

The relations for the frequency dependence of ′µand ′′µ originate from the classical analysis of a mag-netization vector precessing at an angular frequencyω0 = 2πν0 about a DC magnetic-field vector of mag-nitude H. The sense of the precession is determinedby the direction of the magnetic-field vector, and thefrequency ν0 is proportional to H. When the fre-quency ν of the RF magnetic field matches the preces-sion frequency (i.e., when ν = ν0), ferrimagnetic reso-nance occurs. This effect can be visualized if a linearlypolarized wave with the RF magnetic-field compo-



nent normal to the magnetization vector is analyzedin terms of two counterrotating circularly polarizedcomponents, one rotating in synchronism with theprecession (labeled +) and the other in opposition toit (labeled –). The propagation of the two polariza-tions differs as a function of frequency, with the +component undergoing the resonance effect. Distinctpermeabilities µ+ and µ− can then be defined for usein Equations 1, 2, and 3. For devices that utilize cir-cularly polarized waves, the difference between thetwo permeabilities can be exploited to produce sig-nificant phase-shift effects, as we show later.

As derived from a mathematical analysis of the per-meability tensor, the real and imaginary permeabilitycomponents for the two circular polarizations are ex-pressed as

′± = ++

µγ π ν ν

ν ν ν1

4 0

02

02

( )( )

( ) ( )

M m

m ∆ (4)

and

′′± =+

µγ π ν

ν ν ν

( )

( ) ( )

4 0

02

02

M ∆∆m

, (5)

where ν γ0 = H is the ferrimagnetic resonance fre-quency, ∆ν0 is the intrinsic half-linewidth resultingfrom magnetic resonance damping, γ = 2.8 GHz/kGis the gyromagnetic constant, and M is the magnitudeof the DC magnetization vector [4]. Recalling thatthe + and − polarizations are determined by the senseof the precessing magnetization vector, we note that areversal in the direction of the vector will cause µ+and µ− to be interchanged. From Equations 4 and 5,at a given frequency there are two ways to alter thepermeability: (1) by varying ν0 through control of theexternal magnetic field H , or (2) by changingγ (4πM ) through adjustment of the magnetization M,which can be varied over a range from zero (completedemagnetization) to 4πMs (full magnetic saturation).

In our work on ferrite devices, we emphasize RFwaves having circular polarization, since this polariza-tion can provide much larger phase shifts than linearpolarization. Figure 1 shows an example of the varia-tion with respect to ν0 of the ferrite complex perme-ability for circularly polarized waves of fixed fre-quency ν equal to 10 GHz. To avoid high absorptionnear the resonance peak in ′′+µ , the resonance fre-

quency ν0 is set far from ν. To maximize the usefulfrequency band of operation, the ferrite is chosenwith ∆ν0 small enough to be ignored in the denomi-nator of Equations 4 and 5. When the resonance fre-quency ν0 is small with respect to ν , there is a signifi-cant difference between the permeabilities ′+µ and ′−µfor the two senses of circular polarization. Underthese conditions, Equation 4 can be reduced to

′± ≈µγ π

ν1

4m

( )M. (6)

Devices that operate in this regime with circularly po-larized waves are therefore described as nonreciprocal,because forward and reverse directions of propagationhave different permeabilities and propagation con-stants. The corresponding difference in phase,

∆φ φ φ π ε µ µλ

= − = −( )+ − + −′ ′2L

,

obtained by reversing the direction of the signalpropagation or, more conveniently, reversing the di-

FIGURE 1. Variation of the four complex permeabilitycomponents for circular polarization as a function of theferrimagnetic resonance frequency ν0 for a fully magne-tized ferrite with γ(4πMs) = 5 GHz and the half-linewidth∆ν0 = 0.5 GHz (broadened for illustration purposes). TheRF frequency ν for this example is 10 GHz. The magni-tude of the difference between ′+µ and ′−µ , which is re-lated to the magnitude of the differential phase shift, isgreater and less frequency sensitive in the regime whereν0 << ν , compared to the regime where ν0 >> ν. In this ex-ample, the shaded region indicates the range of reso-nance frequencies acceptable for device operation.

10

100 20–10

0 (GHz)

(4 Ms) = 5 GHz

0 = 0.5 GHz

0

−′

+

+′ −

πν

γ

ν

∆

′′′

′′

′′

= 10 GHzν

µµ

µ

µ

µan

d

µ



rection of the magnetization vector is called the differ-ential phase shift.

For the case in which both ν0 and γ (4πM ) aremuch less than ν , we can use Equations 3 and 6 toexpress the differential phase shift and the magneticloss tangent as

∆φ π εγ π

ν λ≈ 2

4( )M L

and

tan( )

δγ π

νννµ ≈ 4 0M ∆

.

From these relations, we see that the design objectiveis to employ the ferrite with the largest 4πMs , formaximum phase shift, and the narrowest ∆ν0, forminimum loss tangent. There is a limitation on thevalues of 4πMs , however, arising from the magneticfields that occur among the domains in partially mag-netized ferrite. These internal fields range in magni-tude from zero to 4πMs and create absorption lossesover the frequency band from zero to γ(4πMs). Sincethis low-field-loss regime must be avoided for devices,

the ferrite selected for a particular application requiresa value of 4πMs low enough to allow efficient opera-tion in the desired frequency band.

For most applications, the relation

γ πν

( ).

40 5

Ms ≤ (7)

defines the practical limit to avoid low-field-loss ef-fects, and is generally used as a device design goal formaximum efficiency. Ferrites with saturation magne-tization 4πMs ranging up to 5 kG at room tempera-ture can be obtained commercially for the lower mi-crowave frequency bands. Since this limit on 4πMscorresponds to a gyromagnetic frequency of 14 GHz,the maximum operating frequency for which theabove ratio can be achieved is 28 GHz. For frequencyincreases beyond 28 GHz, concerns about low-fieldloss disappear because the relation in Equation 7 is al-ways satisfied. However, the frequency limit for opti-mum phase-shift efficiency can be raised substantiallyhigher than 28 GHz by cooling the ferrite to cryo-genic temperatures where values of 4πMs exceeding7 kG are attainable.

Magnetization Control

For typical phase shifters and junction circulators thatoperate in the partially magnetized state, the value of∆φ is directly related to 4πM, which in turn is estab-lished by the device operating point on the ferritehysteresis loop. Figure 2 illustrates the key features ofthe ferrite hysteresis loop, which shows the depen-dence of magnetization on the magnetic field. The re-manent magnetization 4πMr , which has a value inthe magnitude range 0 < 4πMr < 4πMs , is defined asthe level of magnetization that remains after a magne-tizing field is removed. The remanence ratio, which isthe ratio of 4πMr to the saturation magnetization4πMs , is typically 0.7 for ceramics with randomly ori-ented crystal grains. The coercive field Hc should beas small as possible to ensure low switching energiesand switching times. In systems applications that em-ploy remanent-state variable phase shifters, such asphased-array radars, the parameters of the hysteresisloop are critically important.

The desired magnetization level can be establishedthrough the use of fixed-voltage impulses of con-

FIGURE 2. Hysteresis-loop principle for magnetic-fluxswitching (flux-drive method) of microwave ferrite phaseshifters. The important features of this loop are the re-manence ratio 4πMr /4πMs, which should be as large aspossible, and the coercive field Hc , which should be assmall as possible.

–Hs

HcH

Mr

Ms

M

1

2

34

34

34

Reference

Reset point

π4

π4

π4

Msπ−4



trolled duration applied through a magnetizing coilto generate variable amounts of magnetic-flux reversal(the flux-drive method) [5]. The ferrite is set to rema-nence at –4πMr on its major hysteresis loop by firstapplying a saturating impulse of field –Hs (reset point1 in Figure 2) and then allowing the magnetization torelax to its maximum remanence value (referencepoint 2, the latching condition). Selected voltage im-pulses of shorter length bring the magnetization topoints 3 on a minor loop by reversing predeterminedamounts of flux. The magnetization then attains thecorresponding new remanent states (points 4) aftereach pulse is completed. We can follow this procedureto obtain any value of 4πM within the range availableto the device. By resetting to point 1 and reestablish-ing point 2 before each new pulse is applied, we canexploit the full range of magnetization (–4πMs to+4πMs) without the use of holding currents.

Ferrite-Superconductor Devices

To join a ferrite to a superconductor without allowingmagnetic fields associated with the magnetization ofthe ferrite to penetrate the superconductor and de-grade its properties, we used a design based on mag-netic-flux confinement within a closed magnetic cir-cuit, e.g., ferrite in a toroidal geometry, as illustratedin Figure 3. In effect there are no magnetic poles onthe ferrite surface to produce an external magnetic

field, and the superconductor can be in contact withany surface of the magnetized toroid. The only otherbasic requirement for gyromagnetic interaction is forthe RF magnetic field to be normal to the magnetiza-tion vector of the ferrite. This relationship betweenmagnetization and field is usually achieved inmicrostrip circuits by maintaining the magnetizationalong the direction of signal propagation. The mag-netization level, and hence the amount of phase shift,is controlled by applying electrical impulses to themagnetizing coil, as described in the previous section.

Phase Shifters

We investigated several different configurations offerrite-superconductor phase shifters to demonstratetheir performance advantages. These configurationsall comprised compact low-loss superconductingtransmission lines in contact with a ferrite magne-tized in a closed flux loop, and included a means toadjust the magnetization level for controlling theamount of phase shift. In this section we describethree of these device structures.

Figure 3 illustrates the simplest device design. Aone-inch-long niobium superconducting transmis-sion line, deposited on an Al2O3 dielectric substrate,is pressed in direct contact with a ferrite (yttrium irongarnet, Y3Fe5O12, or YIG) rectangular-shaped toroid.A ground plane is deposited on the top surface of the

FIGURE 3. Structure of an experimental reciprocal phase shifter showing the coupling relationship between magnetizedferrite (yttrium iron garnet, Y3Fe5O12, or YIG) in the shape of a planar rectangular-shaped toroid and the niobium super-conductor transmission line on an Al2O3 dielectric substrate. Magnetic flux is confined within the YIG toroid.

Niobium superconductor transmission linedeposited on the substrate

Magnetizing coil

Al2O3 dielectric substrate

Input Output

YIG toroid with silver groundplane on top surface

M



toroid. Since a portion of the RF signal in the super-conducting transmission line traverses the ferrite, thegyromagnetic interaction between the magnetic-fieldcomponent of the RF wave and the magnetized ferriteresults in a phase shift that varies with the ferrite mag-netization level. In this case, because the RF wave islinearly polarized, the phase shift is reciprocal. Ex-periments to validate the superconducting RF prop-erties of this initial design of the phase shifter werecarried out at 4 K [6]. Measurements indicated thatno increase in transmission-line RF loss occurred be-cause of the presence of the magnetized toroid.

As we discussed earlier, much larger phase shiftscan be achieved in ferrites by exploiting the non-reciprocal propagation of circular polarization. In pla-nar configurations, we can produce regions of circu-larly polarized RF field by using a meanderlinestructure of the type illustrated in Figure 4. When the

FIGURE 4. Meanderline structure showing the internal generation of circular polarization. (a) Plan view of a meanderlinecircuit on a ferrite substrate, with magnetization vectors of magnitude M parallel to the meanders. (b) Cross-sectionview through the meanderline circuit in part a, showing the generation of circular polarization at a point P centered be-tween adjacent meanders. The quantities h1 and h2 are adjacent-leg RF components of the magnetic field 90° apart inphase along the centerline of the meanderline circuit.

meander length is near a quarter wavelength, as indi-cated in the figure, there is approximately a 90° phasedifference between signals on adjacent legs of the me-ander near the centerline of the structure. Overlap ofthe fields results in circular polarization over a por-tion of the ferrite cross section between them. Themagnetization vector of magnitude M is directed par-allel to the meanders to provide a gyromagnetic ef-fect. While meanderline structures have been previ-ously demonstrated for ferrite phase shifters [7–9],the high transmission losses that occur with normalmetal conductors have prevented their widespreadapplication.

To demonstrate further the enabling capabilities ofsuperconductivity, we designed a ferrite phase shifterthat incorporates a superconducting meanderline cir-cuit to produce circular polarization effects withminimal conduction losses [10–12]. The design was

M M M M

/4

(a)

(b)

Cross-section line/4

h1 h2

λ

λ

Cross-section viewP

M M



FIGURE 5. Experimental layered phase shifter with meanderline superconductor (niobium or YBCO) deposited directlyon a LaAlO3 substrate rather than directly on the ferrite substrate. The ferrite is held in close proximity to the supercon-ductor meanderline. This configuration is necessary in the case of the YBCO circuit.

implemented, as shown in Figure 5, with a high-TcYBCO superconductor transmission line depositedon a lanthanum aluminate (LaAlO3) substrate, foroperation at 77 K. The magnetic circuit, consisting oftwo parts joined together, forms a toroid of rectangu-lar cross section that establishes the necessary magne-tization direction parallel to the meanders. Magneti-zation control is provided by applying voltageimpulses to the magnetizing coil. The meanderline isheld in contact with the ferrite by “spring fingers,” asshown in Figure 6. Ultimately, as materials technol-ogy advances, YBCO deposited directly on ferrite willsimplify the device design by eliminating the LaAlO3substrate and therefore the need to hold the YBCOcircuit in close contact with the ferrite by mechanicalmeans [13]. At present, YBCO films deposited onLaAlO3 substrates provide superior performance.

We measured both the differential phase shift andinsertion loss of the YBCO meanderline phase shifter,confirming the superior figure of merit of ferrite-superconductor devices. The differential phase shift∆φ , shown in Figure 7, was determined by measuring

the difference in phase between forward-travelingwaves and backward-traveling waves for a fixed mag-netization direction. The phase shift ranged from600° to 700° over the frequency band of 8 to 10GHz. The insertion loss, shown in Figure 8, was lessthan 1 dB over the 8-to-10-GHz band. After sub-tracting reflection losses from the data, the absorptionlosses were less than 0.5 dB over this frequency band.The high absorption region at lower frequencies iscaused by ferrimagnetic resonance that occurs whenthe ferrite is partially magnetized [14, 15]. As dis-cussed earlier, the upper frequency limit of this low-field-loss regime is determined by γ(4πMs ), which inthis case is approximately 5 GHz for the ferrite with4πMs = 1.8 kG at 77 K. The combination of mea-sured phase shift and insertion loss yielded a figure ofmerit exceeding 1000°/dB over the 2-GHz-wide fre-quency band.

The phase-shifter results shown in Figure 7 weremeasured with the ferrite in the remanent state. Mea-surements to compare the differential phase shift ver-sus frequency between the remanent state and the

LaAIO3 substrate

Niobium or YBCOmeanderline

Ferrite flux return path

Input

Magnetizing coil

Magnetized ferrite

Metal ground plane

2.5 cm

Output

M

M



saturated state were performed on a phase shiftercomprising a niobium meanderline deposited onLaAlO3. The results in Figure 9 show that the phaseshift versus frequency in the saturated state is about afactor of two greater than the phase shift versus fre-quency in the remanent state. We can infer from thesedata that the ferrite remanence ratio is reduced to lessthan 0.5, which is probably the result of the demag-netizing effects of air gaps in the toroid. Given thatthe remanence ratio of ferrite is typically 0.7, im-proved design of the toroid structure should lead tolarger phase shifts at the remanent point.

The figure-of-merit values obtained in these ex-periments can be compared with the 130°/dB rema-nence value at X-band for a copper meanderline cir-cuit at room temperature [16]. This value is almost anorder of magnitude inferior to that of the supercon-ductor devices we have demonstrated. Even at cryo-genic temperatures, where the insertion loss from thecopper meanderline circuit would decrease by nomore than a factor of two, a phase shifter employingsuperconductors would have at least a factor-of-fiveadvantage in figure of merit. If the theoretical limitsof the ferrite properties can be realized in practice, anoptimized design with ideal impedance matching of

the superconductor circuit and full utilization of4πMs , a figure-of-merit value as high as 10,000°/dBis a realistic goal.

To miniaturize device dimensions and optimizethe integrity of the magnetic circuit, we have fabri-cated and tested a planar microstrip phase shifter witha YBCO meanderline circuit deposited on a LaAlO3substrate, based on the design illustrated in Figure 10.In contrast to a conventional 10-GHz waveguidephase shifter, which is typically about two inches inlength, this microstrip device is only a half-inch inlength, with a corresponding reduction in cross-sec-tional area and overall volume and weight. The planarstructure still confines the magnetic flux entirelywithin the ferrite but eliminates the added externalyoke. The return path of the flux is outside themeanderline so that the device functions magneticallyin a manner identical to the design shown in Figure 5.Because of the simplified structure, this planar deviceis readily compatible with compact structures inwhich the superconductor would be in contact withthe ferrite through multilayer deposition processes.Figure 11 shows the loss and differential phase shiftfor the in-plane magnetic circuit. These initial resultsindicate performance comparable to that achieved forthe two-part ferrite structures.

Circulators and Switches

The principles of magnetic and cryogenic design de-scribed for the ferrite-superconductor phase shifter

FIGURE 7. Differential phase shift versus frequency for aYBCO meanderline ferrite-superconductor phase shifteron a LaAlO3 substrate, as illustrated in Figure 5.

FIGURE 6. YBCO meanderline phase shifter on a LaAlO3

substrate. The ferrite yoke at the bottom has been re-moved to reveal the meanderline structure.

750

5

Dif

fere

nti

al p

has

e sh

ift (

deg

)

7 9 11 13

600

450

300

0

Frequency (GHz)

150

T = 77 K

1 in

Meanderline circuit



can also be applied to a three-port junction circulator.A style of circulator design that is well adapted to pla-nar circuits, and especially to ferrite-superconductortechnology, is the ring-network circulator. The con-cept was introduced in 1965 [17] and was the subjectof experimental investigation at that time [18]. Thespecific embodiment was a ring composed of threeidentical nonreciprocal phase shifters alternating withthree identical reciprocal T junctions, constituting athree-port junction circulator, as illustrated in Figure

12. This design principle is particularly applicable tominiature planar circuit design, including the ferrite-superconductor combination of superconducting cir-cuit path and planar magnetic-flux confinement, butthe formulation is general and valid for any type ofmicrowave or millimeter-wave transmission medium.

The analysis of the ring-network circulator [19,20] is based on a rigorous scattering formulation ofclockwise- and counterclockwise-propagating partialwaves. This analysis leads to characterization of the

FIGURE 10. A planar microstrip YBCO meanderline phase shifter designed to minimize size and weight and eliminate airgaps in the magnetic-flux path. Such a device would be only a half-inch in length. Tests on devices of this design showresults comparable to the larger two-part ferrite structures.

FIGURE 8. Insertion loss versus frequency for the YBCOmeanderline ferrite-superconductor phase shifter illus-trated in Figure 5. Loss due to reflection effects from im-pedance mismatches at higher frequencies has not beensubtracted.

2500

2000

1500

1000

500

0

5 7 9 11 13Frequency (GHz)

Dif

fere

nti

al p

has

e sh

ift (

deg

)

T = 4 K

SaturatedRemanent

0

–25

–50

–75

–1004 8 120

Inse

rtio

n lo

ss (

dB

)

Frequency (GHz)

T = 77 K

FIGURE 9. Comparison of differential phase shift versusfrequency for a niobium meanderline circuit on a LaAlO3

substrate for saturated and remanent states. The deviceconfiguration shown in Figure 5 was used to determinethese data.

Ferrite

YBCO superconductor circuiton LaAlO3 substrate

(inverted)

DC magnetic-flux loop

Magnetizing coil

Ground plane



nonreciprocal phase shift in relation to the T-junctionscattering parameters such that, when the phaseshifters and T junctions are assembled into a ring, thenetwork satisfies the imposed condition of perfect cir-culation. Circulator performance over the frequencyrange in that vicinity is governed by the dispersivecharacteristics of the components. Computational re-sults indicate that, by suitable design, the requiredmagnitude of differential phase shift can be madesmall, so as to minimize the amount of gyromagneticinteraction needed. Furthermore, with appropriatechoice of dispersive characteristics as specified by thetheory (for example, by incorporating reactive ele-

FIGURE 12. Diagram of a ring network. The labels T andPS denote the symmetrical T junctions and the non-reciprocal phase shifters.

ments at the T junctions), the circulator can achievebroadband performance.

Figure 13 shows photographs of an experimentalmeanderline ring-network circulator. A central holein the substrate creates the toroidal geometry for fluxconfinement and accommodates coil windings for es-tablishing and reversing the state of magnetization.The circulator design in Figure 13 is narrowband.Figure 14 shows 10-GHz-band performance data forthis device but with a niobium circuit cooled to 4 K.The insertion loss is below 1 dB and isolation isgreater than 15 dB over a 2% band. Because the con-duction loss due to the niobium superconductor isexpected to be less than 0.1 dB for this circuit, we be-lieve that the measured insertion loss is dominated byspurious impedance mismatch effects.

This class of circulator designs confers numerouspotential benefits. In contrast to the conventionalresonant-type junction circulator in which an exter-nal magnetic yoke must be used to magnetize the fer-rite element in the direction normal to the plane ofthe circuit—resulting in degraded performance of thesuperconductor [3]—the ring-network circulatorlends itself naturally to magnetization in the plane ofthe substrate. The advantages of such a configurationinclude reduced size, weight, complexity, and energyconsumption of the magnetizing structure. Further-more, the configuration allows for reduced coercivityrequirements in self-magnetized versions (i.e., havingno external magnet structure) as well as for high-speed energy-efficient switching capability in revers-ible versions of the circulator, which can function ef-fectively as switches. The minimization of circulatorinsertion loss through the use of a microstrip super-conducting circuit is a major advantage of this designconcept.

Materials

The choice of ferrite for a particular application atcryogenic temperatures is determined by the sameconsiderations that apply at room temperature. Thesaturation magnetization is selected so that γ(4πMs)is sufficiently lower than the operating frequency νsuch that low-field loss is avoided, as discussed in aprevious section. Because 4πMs values can increaseby more than a factor of two when the material is

FIGURE 11. Example of insertion loss and differentialphase shift versus frequency from a YBCO meanderlinecircuit on LaAlO3 substrate in a microstrip phase shiftersuch as the design shown in Figure 10. Excessive ripplesin the loss curve are the effects of impedance mis-matches in the test circuit.

T2 T3

2 3

PS12

PS23

PS31T1

1

0

–4

–2

–6

–8

–1.0

Inse

rtio

n lo

ss (

dB

)

Dif

fere

nti

al p

has

e sh

ift (

deg

)600

400

200

06 8 10 12

Frequency (GHz)

T = 77 K

Phase shift

Insertion loss



cooled from 300 K to 77 K, care must be taken thatlower room-temperature magnetization materials beused for the standard microwave frequencies (10GHz and below). This 4πMs enhancement, however,can be an immediate benefit for millimeter-wave de-signs in which the attainment of 4πMs values greaterthan 5 kG have been a longtime goal.

For cold environments, some commercially avail-able ferrimagnetic spinels and garnets can be used forapplications in which the magnetic state is fixed.Where rapid and frequent changes of the magneticstate are expected, however, available materials maynot be adequate for achieving high-speed, low switch-ing-energy performance. As shown by the hysteresisloops of yttrium-iron garnet (YIG) in Figure 15, thedecrease in temperature from 300 to 77 K causes al-most a tenfold increase in the coercive field, due to arise in the magnetocrystalline anisotropy. At low tem-peratures, increased stress effects in ferrites can also bea cause of deterioration of the hysteresis loops. If rare-earth impurities are present in any of the chemicalconstituents, the values of ∆ν0 can increase substan-tially at 77 K and raise the values of tan δ µ to unac-ceptable levels. For these reasons, alteration of thechemical compositions of standard room-tempera-ture microwave ferrites will be necessary to attain the

highest efficiency. Strategies for accomplishing thesematerials design changes have been previously estab-lished and are well within the capabilities of modernmaterials technology [21].

FIGURE 13. Photographs of a ring-network circulator: on the left is a laboratory prototype microstrip circulator with fer-rite substrate and meanderline nonreciprocal phase shifters; on the right is an enlarged view of the ring network. Thedevice shown here was fabricated with gold circuitry; the versions of the devices used in the experiments employed nio-bium superconductor circuits.

0

–10

–20

–3010.09.5

Mag

nit

ud

e (d

B)

Frequency (GHz)

Insertion loss

Isolation

FIGURE 14. Insertion loss and isolation versus frequencymeasured at the cryogenic operating temperature of 4 Kfor a ring-network circulator with a niobium supercon-ducting circuit on a ferrite substrate, magnetizedcircumferentially in its plane. This laboratory prototypeis narrowband (about 2% bandwidth at 15-dB isolation)but has favorably low insertion loss (less than 1 dB overthat band) and illustrates the feasibility of the ring net-work as a circulator design concept.



For latching operations in which square hysteresisloops are critically important, designs that employ asingle-piece ferrite toroid, such as that described pre-viously, might be required to avoid deterioration ofthe hysteresis-loop squareness caused by gaps that oc-

cur where the two ferrite pieces are mechanicallyjoined to form a closed path.

Another objective is to control the physical contactbetween ferrite and superconductor in order to opti-mize the interaction between the two and to establishthe highest possible reproducibility. Figure 16 illus-trates four proposed schemes for achieving the desiredconfigurations. With the use of suitable dielectricbuffer layers, the YBCO (or equivalent) superconduc-tor circuit can be established in a controlled proxim-ity to the surface of the ferrite, which can in turn bemagnetized with internal flux confinement.

Summary

We have successfully demonstrated the feasibility ofemploying superconductors in microwave ferrite de-vices without deterioration of the superconductorproperties. This new technology not only offers dra-matically improved efficiency, but also makes possiblethe use of compact lower-cost microstrip structures toreplace bulky ferrite waveguide phase shifters.

For systems such as phased-array radars, the order-of-magnitude advantages gained from the superiorperformance and reduced size and weight of the fer-rite-superconductor technology can make it the pre-

FIGURE 16. Four proposed arrangements for materials integration by layering ferrites and high-Tc superconductors. Thepurpose of these arrangements is to optimize the interaction between the ferrite and the superconductor.

FIGURE 15. Enlargement of a YIG hysteresis loop when Tis decreased from 300 to 77 K. The coercive field in-creases by almost a factor of ten, and the magnetizationis enhanced 20% at the remanent point. The 77-K loop isnot fully saturated at H = 25 Oe.

0

B (

arb

itra

ry u

nit

s)

1

–1–25 0 25

H (Oe)

Dielectric

YBCO circuit

Ground plane

Ceramic ferrite film

Epitaxial ferrite film

Single-crystal ferrite

Bulk ceramic ferrite

Dielectric buffer

Dielectric

Ground plane

Ground planeGround plane

YBCO circuit

YBCO circuit

YBCO circuit



ferred technology, notwithstanding the added costand complications of a cryogenic enclosure. For filter-bank selector applications that would employ thering-network circulator device, these issues also apply,and can include even larger bandwidths and switch-ing speeds. For implementation of low-loss ferrite-su-perconductor phase shifters, circulator switches, andother devices that can utilize this low-loss principle,the adoption of cryogenically cooled array antennas isalso an important step and should be treated as partof the system design where appropriate.

Acknowledgments

The authors are grateful to the Lincoln LaboratoryInnovative Research Program Committee for its sup-port in this exploratory project. The authors are alsoindebted to R.P. Konieczka, D.J. Baker, L.S.DiPalma, J.F. Fitzgerald, C.D. Hoyt, D.L. Hovey,R.J. Magliocco, and C.M. Vanaria for technical assis-tance throughout the course of this work. Dr. A.C.Anderson provided the YBCO films. We give a spe-cial acknowledgment to Dr. L.M. Johnson for hishelpful assistance in the preparation of this article.This work was sponsored by the Department of theAir Force.

R E F E R E N C E S1. H. Piel, H. Chaloupka, and G. Müller, “High Temperature

Superconductors in High Frequency Fields—Fundamentalsand Applications,” Advances in Superconductivity: Proc. 4th Int.Symp. on Superconducting, Vol. 4, Tokyo, 14–17 Oct. 1991, pp.925–930.

2. M.S. Pambianchi, D.H. Wu, L. Ganapathu, and S.M. Anlage,“DC Magnetic Field Dependence of the Surface Impedancein Superconducting Parallel Plate Transmission Line Reso-nators,” IEEE Trans. Appl. Supercond., vol. 3, no. 1, pp.2774–2777 (1993).

3. E. Denlinger, R. Paglione, D. Kalokitis, E. Belohoubek,A. Piqué, X.D. Wu, T. Venkatesa, A. Fathy, V. Pendrick,S. Green, and S. Mathews, “Superconducting Nonreci-procal Devices for Microwave Systems,” IEEE Microw.Guid. Wave Lett., vol. 2, no. 2, pp. 449–451 (1992).

4. B. Lax and K.J. Button, Microwave Ferrites and Ferrimagnetics(McGraw-Hill, New York, 1962), pp. 150–157.

5. W.J. Ince and D.H. Temme, “Phasers and Time Delay Ele-ments,” Advances in Microwaves, Vol. 4, L. Young, ed. (Aca-demic Press, New York, 1969), pp. 1–189.

6. G.F. Dionne, D.E. Oates, and D.H. Temme, “Low-Loss Mi-crowave Ferrite Phase Shifters with Superconducting Cir-cuits,” 1994 IEEE MTT-S Int. Microw. Symp. Dig., Vol. 1, San

Diego, 23–27 May 1994, pp. 101–103.7. F.J. Rosenbaum, “Integrated Ferrimagnetic Devices,” Ad-

vances in Microwaves, Vol. 8, L. Young and H. Sobul, eds. (Aca-demic Press, New York, 1974), pp. 203–294.

8. S.K. Koul and B. Bhat, Microwave and Millimeter Wave PhaseShifters, Vol. 1 (Artech, Boston, 1991), pp. 338–347.

9. E.R.B. Hansson, S. Aditya, and M.A. Larsson, “PlanarMeanderline Ferrite-Dielectric Phase Shifter,” IEEE Trans.Microw. Theory Tech., vol. 29, no. 3, pp. 209–215 (1981).

10. G.F. Dionne, D.E. Oates, and D.H. Temme, “Ferrite-Super-conductor Microwave Phase Shifters,” IEEE Trans. Magn.,vol. 30, no. 6, pp. 4518–4524 (1994).

11. G.F. Dionne, D.E. Oates, and D.H. Temme, “YBCO/FerriteLow-Loss Microwave Phase Shifter,” IEEE Trans. Appl.Supercond., vol. 5, pp. 2083–2086 (1995).

12. G.F. Dionne, D.E. Oates, D.H. Temme, and J.A. Weiss, “Fer-rite Superconductor Devices for Advanced Microwave Appli-cations,” IEEE Trans. Microw. Theory Tech., vol. 44, no. 5, pp.1361–1368 (1996).

13. A. Piqué, K.S. Harshavardhan, J. Moses, M. Mathur,E. Belohoubek, T. Venkatesan, E.J. Denlinger, D. Kalokitis,A. Fathy, V. Pendrick, M. Rajeswari, and W. Jiang, “Micro-wave Compatible YBa2Cu3O7–x Films on Ferrimagnetic Gar-net Substrates,” Appl. Phys. Lett., vol. 67, no. 12, pp. 1778–1780 (1995).

14. B. Lax and K.J. Button, Microwave Ferrites and Ferrimagnetics(McGraw-Hill, New York, 1962), pp. 540–544.

15. B. Lax, “Frequency and Loss Characteristics of MicrowaveFerrite Devices,” Proc. IRE, vol. 44, pp. 1368–1386 (Oct.1956).

16. G.T. Roome, “Thin Ferrites for Integrated Microwave De-vices,” Tech. Rept. AFAL-TR-69-149, Air Force SystemsCommand, Wright Patterson Air Force Base; also described inRef. 7.

17. J.A. Weiss, “Circulator Synthesis,” IEEE Trans. Microw. TheoryTech., vol. 13, no. 1, pp. 38–44 (1965).

18. S.D. Ewing and J.A. Weiss, “Ring Circulator Theory, Designand Performance,” IEEE Trans. Microw. Theory Tech., vol. 15,no. 11, pp. 623–628 (1967).

19. J.A. Weiss and G.F. Dionne, “Performance Capabilities ofthe Ring Network Circulator for Integrated Circuits,” 1995MTT-S Int. Microw. Symp. Dig., Vol. 2, Orlando, FL,15–19 May 1995, pp. 725–728.

20. J.A. Weiss, G.F. Dionne and D.H. Temme, “The Ring Net-work Circulator for Integrated Circuits: Theory and Experi-ments,” IEEE Trans. Microw. Theory Tech., vol. 43, no. 12,pp. 2743–2748 (1995).

21. G.F. Dionne, “Properties of Ferrites at Low Temperatures,” J.Appl. Physics, vol. 80, pp. 5064–5069 (1997).



. is a member of Expert ServicePersonnel, supporting theAnalog Device Technologygroup. He received a B.Sc.degree (summa cum laude) inphysics from the Université deMontréal (Loyola College), aB.Eng. degree in engineeringphysics from McGill Univer-sity, an M.S. degree in physicsfrom Carnegie-Mellon Univer-sity, and a Ph.D. degree inphysics from McGill Univer-sity. He served as a member ofthe physics faculty at McGillUniversity, worked on semi-conductor device developmentwith both IBM and theSylvania Division of GTE, andresearched electron emissionand cesium vapor ionizationfor thermionic energy conver-sion at Pratt & WhitneyAircraft. Since 1966 he hascarried out research at LincolnLaboratory, where he haspublished extensively in fieldsthat include magnetism theoryand properties of ferrimagneticmaterials for microwave,millimeter-wave, and mag-netooptical applications,secondary electron emission,and millimeter-wave andsubmillimeter-wave radiom-etry. From his studies of therole of large polarons in high-Tc superconductivity, herecently developed an effectivemodel for giant magnetoresis-tance effects. He has authoredpatents on microwave ferriteand superconductor devices.He is a Fellow of the IEEE anda member of the AmericanPhysical Society and the Soci-ety of Sigma Xi.

. is a staff member in the Ana-log Device Technology group.He received a B.A. degreefrom Yale University and aPh.D. degree from MIT, bothin physics. In 1971 he held anAlexander von Humboldtfellowship at the University ofBonn, Federal Republic ofGermany, and from 1973 to1978 he worked at Bell Tele-phone Laboratories. Since1978 he has been a staff mem-ber at Lincoln Laboratory,where his research has in-cluded investigations of sur-face-acoustic-wave and bulk-acoustic-wave devices. Since1988 his primary researchinterests have been in super-conducting electronics utiliz-ing the newly discovered high-transition-temperaturematerials, especially for appli-cations in frequency controland microwave-frequencyanalog signal processing.Motivated by the potentialapplications of superconduc-tors, he also has investigatedthe fundamental properties ofsuperconductors at microwavefrequencies and at high micro-wave power levels. Theseinvestigations have includedboth high-Tc and low-Tcmaterials. He is a member ofthe IEEE and the AmericanPhysical Society.

. is a former senior staff memberin the Communications divi-sion, and is now a consultantto the Analog Device Technol-ogy group. His research spe-cialty is in ferrite-supercon-ductor control devices. Hereceived a B.S. degree inelectrical engineering from theUniversity of Nebraska and anS.M. degree in electrical engi-neering from MIT. Beforejoining Lincoln Laboratory in1957, he was in the U.S. AirForce. He is a Life Member ofthe IEEE and a member ofSigma Xi.

. is a consultant in the AnalogDevice Technology group. Hereceived B.A., M.A., andPh.D. degrees, all in physics,from The Ohio State Univer-sity. In 1953 he joined the BellTelephone Laboratories (nowLucent Technologies, Inc.) as amember of the technical staff,and he worked in the develop-ment of microwave compo-nent principles and deviceswith applications of magneticmaterials. In 1962 he joinedthe physics faculty at Worces-ter Polytechnic Institute,where his work includedteaching and research in mi-crowave spectroscopy and themeasurement and applicationsof gyrotropic and dielectricmaterials. He was appointedProfessor Emeritus in 1988. In1962 he also was appointedconsultant to Lincoln Labora-tory for work on materials andcomponents relating to radardevelopment, and he contin-ues to serve in that role. He isthe author of numerous publi-cations and patents in areasincluding nonreciprocal mi-crowave and millimeter-wavedevices, transmission-linetheory, and magnetic measure-ments. He is a member of PhiBeta Kappa, Sigma Xi, and theAmerican Physical Society, andhe is a Life Fellow of the IEEE.

Superconductivity for Improved Microwave Ferrite Devices€¦ · Superconductivity for Improved Microwave Ferrite Devices VOLUME 9, NUMBER 1, 1996 THE LINCOLN LABORATORY JOURNAL 21

Documents