The last quarter-century has seen extraordinary growth in commercial wireless technology. In the late 20th century, advances were confined primarily to wireless telephony. In the last 15 years, however, there has been an explosion of applications, including GPS, wireless data transfer and currently, the Internet of Things.

Cellular handsets have become more sophisticated, embracing more functionality in very small spaces. Enabling this wireless communication revolution are parallel developments in wireless infrastructure. Ceramic materials have long been used for cellular base stations and other wireless infrastructure applications. Microwave dielectric ceramics were essential components in auto-tuned combiners for 2G (and earlier) systems. They are still used as filters for modern 4G- and LTE-based systems, where the trend is to carrier aggregation, requiring more filters for multiplexing.

Technological Requirements

Maintaining signal directionality in these cellular base stations is vital and requires the use of isolators and circulators. Isolators and circulators consist of a non-conducting ceramic ferrite that is biased by a permanent magnet. These devices rely on gyromagnetic polarization splitting (Lenz’s law) to direct the radio frequency (RF) signal and prevent RF power from leaking into unwanted areas. Travel in the opposite direction is vigorously attenuated.

Isolators and circulators can be operated either above or below the frequency of the gyromagnetic resonance of the ferrite. The frequency of operations is determined by the saturation magnetization and size of the ferrite, as well as the strength of the biasing field; above-resonance operation is limited by available magnet strength to below 3 GHz. It is necessary to completely saturate the ferrite to prevent unwanted intermodulation products. It is desirable that the Curie temperature be maximized to limit temperature drift. A narrow resonance linewidth is desirable in that it maximizes the frequency range of operations where there are minimal magnetic losses. Therefore, magnetic oxide materials are desired with very sharp gyromagnetic resonance linewidths.

The requirements for the ferrite-based materials for isolator and circulator applications are stringent. First, the material has to be an insulating material with low dielectric losses (loss tangent < 0.002). This requirement limits metallic magnetic materials and brings oxides to the forefront of materials for this application. The mechanism of magnetism is different between metals and oxides as well. Metals show spin interaction between neighboring metal atoms, while oxides show a magnetic interaction called superexchange in which the non-magnetic oxygen ion couples the magnetic interaction between second-nearest neighbor magnetic ions. In addition to being an excellent electrical insulator, the gyromagnetic resonance linewidth has to be very narrow. This second requirement limits the choice of materials to cubic ferrite-based structures, particularly those of the garnet structure due to their extremely low magnetic anisotropy.

The Role of Ceramics

Yttrium iron garnet (Y₃Fe₅O₁₂, or YIG)-based materials have been the industry standard for isolator and circulator ferrites for several decades. Prior to the commercial wireless communication boom, they have had numerous military applications for the same function. YIG was first synthesized by Geller et. al.¹ in 1957, and soon his group and other researchers found that they could substitute a large number of ions into the garnet lattice to create a suite of materials with useful magnetic properties.^2-3 This structure has three distinct cation crystallographic sites (see Figure 14): a dodecahedral (or eight-coordinate) site, where the larger yttrium ions reside (three of these sites per formula unit of garnet); octahedral (six-coordinate) sites, where two out of five of the iron atoms in the basic formula unit reside; and the tetrahedral (four-coordinate) site, where the remaining three iron atoms in the formula unit reside.

The magnetic response in garnets is due to the ferromagnetic interaction between the tetrahedral iron sites and octahedral iron sites of opposite spins (3 Fe ↑ - 2 Fe ↓), which leaves a net magnetization of 1 Fe ↑. Since each Fe³⁺ ion has five unpaired electrons, there is still a significant amount of net magnetization in YIG. That saturation magnetization can be adjusted down by substituting non-magnetic ions in the tetrahedral site (such as V⁵⁺) or up by substituting non-magnetic ions in the octahedral site (such as Zr⁴⁺), for example.^5-7

The advantage of garnet-phase ceramics lies in the cubic crystal structure. There is very little difference in the magnetic response of single crystal Y₃Fe₅O₁₂ based on the direction of an externally applied field relative to the crystallographic planes. Because of this, polycrystalline YIG ceramics (single crystals are extremely expensive to fabricate) with random crystallite orientation may still show a sharp gyromagnetic resonance. Since all of the resonances overlap at the same frequency regardless of the orientation of the grain, polycrystalline garnet ceramics tend to show sharp gyromagnetic resonances and are thus able to be used over wide bandwidths, including at frequencies close to the resonant frequency.

Although polycrystalline YIG is frequently used for isolator and circulator applications (it has a 3 dB linewidth < 20 Oersted and the highest Curie temperature of all garnet materials at 280°C), substitutions are often made for the following reasons:

• To adjust the saturation magnetization of the garnet material (YIG has a saturation magnetization close to 1,780 gauss). Adjusting the saturation magnetization of the material allows for the material to be used below resonance over a range of frequencies. Typically, the saturation magnetization may be reduced relative to 1,780 gauss for pure YIG by substituting a non-magnetic ion for iron on the tetrahedral site. Examples include aluminum (Al³⁺), gallium (Ga³⁺), silicon (Si⁴⁺) and vanadium (V⁵⁺). Note that for silicon and vanadium with formal oxidation states greater than 3+, there has to be a second substituent ion to compensate for the excess positive charge. In these cases, Ca²⁺ is frequently substituted for Y³⁺ to maintain charge balance. Vanadium is a desirable substituent because it is the only non-magnetic iron substitute that does not dramatically reduce the Curie temperature.
   
• To further reduce the gyromagnetic resonance linewidth (at 3 dB below the peak). Van Hook⁶ and later Winkler⁷ showed that replacing octahedral site iron with a non-magnetic ion with a larger ionic radius reduces the magnetic anisotropy to values lower than in pure YIG. In³⁺ (0.94 Angstroms) and Zr⁴⁺ (0.86 Angstroms) substituted for Fe³⁺ (0.69 Angstroms) in magnetic garnets, may show 3 dB linewidths below 10 Oersted. Due to the high expense of indium, zirconium is more frequently used, although it does require charge compensation with non-magnetic calcium.
   
• Both of the previous reasons. Industry-standard narrow linewidth garnets for circulators tend to be vanadium and zirconium-substituted YIG. Since both vanadium and zirconium have formal charges greater than 3+, calcium is substituted for yttrium to balance the charge. The general formula for these garnet materials is Y_3-x-2yCa_x+2yFe_5-x-yZr_xV_yO₁₂.

Garnet materials with the general formula shown here tend to show dielectric constants in the range of 13-16. To further reduce the size of the circulators or isolators, higher dielectric constant materials would be required. In fact, this provides advantages in that a reduced ferrite disk size leads to reduced isolator sizes and smaller footprints for these components in the base station. The diameter of the magnetic disk in the circulator or isolator is proportional to the inverse square root of the dielectric constant. Therefore, an effort was made to increase the dielectric constant of the material to allow for smaller devices working at the same frequency.

To increase the dielectric constant of garnet materials, atoms with a very high ionic polarizability need to be substituted for the non-magnetic ions in the structure. The obvious site for the substitution is the dodecahedral site occupied by yttrium, since it is the only site that does not contain the magnetic iron atoms. Bismuth is a highly polarizable ion with a non-bonding lone pair. Although bismuth-based magnetic garnets have been known for some time,^4,5 they have not been used for narrow linewidth garnets for circulators or isolators. Since bismuth is a much larger ion than yttrium and asymmetrical due to the lone pair, there is a limit of the amount of bismuth that may be substituted for yttrium in the structure.

For a material containing zirconia on the octahedral site, up to 1.4 formula units of Bi³⁺ can be substituted for Y³⁺. Depending on the amounts of zirconium or vanadium, the saturation magnet, the saturation magnetization, the 3 dB linewidth and the Curie temperature may be adjusted. These substituted materials may have dielectric constants ranging from 25 to 31. To illustrate the effect of the higher dielectric constant on the size of the circulator, we can look at the reduction of the diameter of the ferrite disk on going from a Bi-free garnet (dielectric constant = 15) to a Bi containing garnet (dielectric constant = 27). The diameter of the ferrite disk would be reduced by a factor of (15/27)^1/2 or 0.72. As the area is proportional to the square of the diameter, the area is reduced by 48%, which is a significant number. A linewidth penalty is paid along with the reduced size; this is an acceptable trade-off for many modern applications. However, with improvements in ceramic processing, the penalty in linewidth is expected to decrease in the future.

Composite disks may be made as well, consisting of an inner core of a magnetic oxide material along with an outer ring of a non-magnetic insulating dielectric material. The two materials may be glued or co-fired together into an assembly, depending on the chemical compatibility of the two materials.⁴ The composite material ensures complete saturation of the inner core magnetic oxide by the applied field from the permanent magnet. If saturation does not occur, problems from intermodulation are likely to result. As with the garnet materials, the use of a higher dielectric constant material in the assembly will result in a composite material with a smaller diameter.

Ongoing Development

An active area of research continues to be designing materials with even higher dielectric constants for further miniaturization. In addition, ongoing research aims to fill out the suite of different magnetizations with high dielectric constant counterparts. The technology may also lead to other applications, including tunable filters. This is an area of ceramics where innovations in ceramic materials design enable enhanced device functionality in a burgeoning commercial market.  

For more information, contact the author at (301) 874-6422 or mike.hill@skyworksinc.com, or visit www.trans-techinc.com.

Author’s acknowledgement: Thanks to David Cruickshank for his insight and helpful discussions.

References

1.         S. Geller, M.S., Gilleo, “The Crystal Structure and Ferrimagnetism of Yttrium-Iron Garnet, Y3Fe2(FeO4)3,” J. Phys. Chem Solids, 3, 30 (1957).

2.         S. Geller, “Magnetic Interactions and Distribution of Ions in Garnet,” J. Appl. Phys., 31, 30S (1960).

3.         G. Winkler, P. Hansen and P. Holst, “Variation of the Magnetic Field Parameters and Lattice Constants of Polycrystalline Yttrium-Iron Garnet by Incorporation of Non-Magnetic Ions,” Philips Res. Repts., 27, 151 (1972).

4.         D. Cruickshank, Microwave Materials for Wireless Applications, Artech House (2011).

5.         G.R. Harrison, L.R. Hodges, “Microwave Properties of Polycrystalline Hybrid Garnets,” J. Amer. Ceram. Soc., 44, 214 (1961).

6.         H.J. Van Hook, J.J. Green, F. Euler, F.R. Czerlinsky, “Linewidth Reduction through Indium Substitution in Calcium-Vanadium Garnets,” J. Appl. Phys., 39, 730, (1968).

7.         G. Winkler, “Substituted Polycrystalline YIG with Very Low Ferromagnetic Resonance Linewidth and Optical Transparency,” IEEE Trans. Mag., MAG-7, 773, (1971).