Thin-Film Phase Shifters for Low-Cost Phased Arrays

Drastic cost reduction of phased-arrays will require a shift in design strategy. The emphasis must be on removing expensive active components, not developing more highly integrated components or exotic packaging techniques. The phase shifter circuit—an essential active component—is the primary obstacle in decreasing the cost and number of active components in a phased-array. The problem is multi-faceted: (1) current phase-shifter components are expensive MMICs (~40% of a typical receive-array cost); (2) MMICs require careful packaging, further increasing costs; and (3) current phase shifters have significant RF loss and therefore additional amplification must be provided to compensate for this loss, which in turn may require limiting circuits (in a receive application in situations with strong local interference). If a phase shifter could be designed with little or no loss, it is believed that up to 80% of the active devices in a typical phased-array could be deleted from the design. Two new technologies—thin-film nonlinear dielectrics and MEMS—have recently emerged and show significant promise for implementation of low-loss and low-cost phase shifters. Both technologies appear capable of ultimately providing <3dB insertion loss for a 360 degree phase shifter at considerably lower cost than existing MMIC designs. This paper will focus mostly on developments in thin-film phase shifter circuits using monolithic Barium Strontium Titanate (BST) varactors on inexpensive substrates. Enhanced MEMS circuits using such thin-film dielectrics will also be discussed and contrasted with the nonlinear dielectric approaches in terms of electrical performance and practical considerations for use in low-cost phased-arrays. The thin-film nonlinear dielectrics have been grown by both MOCVD-grown and RF magnetron-sputtering, and can now be obtained commercially. They are very inexpensive, can be deposited on inexpensive Silicon or ceramic substrates, and are compatible with most standard monolithic processes with considerably looser lithography requirements as compared with MMICs. Current progress in materials and projected performance limits will be discussed. Thin-Film BST Barium Strontium Titanate (Ba1-xSrxTiO3, or simply BST) thin films have an electric-fielddependent permittivity, making them suitable for several linear and nonlinear circuit applications at microwave frequencies. Until recently, the use of BST in microwave circuits was impractical due to the high microwave losses associated with these materials. However, there have been numerous advances in the deposition technology for BST in thin-film form, mostly for use in small-area storage capacitors in high-density silicon DRAM [1], which have subsequently lowered microwave losses in BST thin films to levels where they are now practical for circuits. In addition to tunable permittivities and low losses, these materials also have high breakdown field strengths leading to high power handling capability. Most of the research on microwave applications of BST has focused on wave propagation in bulk crystals or thick films [2]. Although bulk (ceramic) BST has been demonstrated with low RF loss tangents, the bulk approach can nevertheless result in high microwave losses as a consequence of the extremely high permittivity of the bulk material (a large substrate dielectric constant indirectly leads to large ohmic losses on transmission lines due to changes in the conductor geometry required to maintain a specified characteristic impedance). Bulk BST has a very strong temperature dependence, exhibiting a well-defined and compositiondependent Curie temperature near room temperature. In contrast, thin-film BST can have remarkably different electrical properties than bulk BST. Figure 1 compares the temperature-dependent zero-bias permittivity of bulk material and MOCVD-grown thin-film BST (on Pt-coated Silicon) of roughly the same composition. The low-field permittivity is dramatically reduced from >4000 at room temperature for bulk material to something in the range of 200-350 for the thin-film material. It is believed that mechanical stress in the film due to growth on thermally mismatched substrates is largely responsible for this behavior. Figure 1— Comparison of zero-bias dielectric constant for bulk BST and thin-film (MOCVD-grown) BST versus temperature. Attractive features of BST materials and devices are described in figure 2, along with promising circuit applications. Integrated parallel-plate or interdigital BST-based capacitors may be used as replacements for semiconductor varactor diodes in several linear and nonlinear microwave circuit applications. Thin-film varactors offer several advantages over semiconductor varactor diodes: They are easy to fabricate and can be made at lower cost than semiconductor varactor diodes; They have higher breakdown field strengths and thus have higher power handling capability than semiconductor diodes. Like semiconductor diodes, the nonlinear capacitance can also be exploited to make frequency conversion devices (multipliers, mixers, etc.). Large field dependent permittivity Compact tunable circuits Intrinsically fast field response Fast switching speeds High breakdown fields, > 3 x 106 V/cm High power handling capability Low drive currents (dielectric leakage) Low prime power requirements Symmetric nonlinearity Low cost high-power zero-bias multipliers Simple fabrication Low cost Circuit Implementation Voltage controlled capacitance phase shifters VCOs tunable filters Variable phase velocity transmission line phase shifters delay lines Nonlinear reactance frequency multipliers mixers What is required • Low loss tangents • Wide tunability • Low leakage, long lifetime • Reproducible growth Key BST Properties Figure 2 — Summary of BST merits and potential for use in circuits. Chemical Vapor Deposition (MOCVD) and RF Sputtering are the two most promising methods for deposition of high-quality BST thin-films on commercially viable substrates. At UCSB, RF sputtering is favored since it allows for growth at somewhat lower temperatures than MOCVD and hence affords greater flexibility in terms of substrate and electrode choices. The sputtered BST thin films can be deposited on a number of semiconductor or microwave substrates such as silicon, GaAs, MgO, alumina etc. which have reasonably low permittivity thus helping to reduce overall circuit loss. The ability to deposit sputtered BST thin films on standard microwave/semiconductor substrates also has benefits from a cost and compatibility standpoint since thin film processing techniques are widely used in the semiconductor and microwave industry. Microwave circuits also tend to be physically large, making them cost-prohibitive on precious semiconductors such as GaAs. Sputtered film deposition is also much cheaper than semiconductor epitaxy. Integrated Monolithic BST Varactors Currently, BST thin film based varactors have similar loss performance to GaAs Schottky varactor diodes, for a given set of lithographic design rules. But with further improvements in material quality it is likely that they will have superior loss performance. Recent studies [1] have demonstrated extremely low RF losses up to 20GHz, corresponding to loss tangents of less than 0.003, or device Q-factors of >300. UCSB has focused on the development and optimization of BST thin-films specifically for microwave integrated circuits, using both MOCVD-grown and RF magnetron-sputtered films. The material optimization efforts have concentrated on achieving high tunability and simultaneous low loss, and also developing suitable electrode systems for circuit fabrication on silicon substrates. Currently we have achieved 4:1 tunability in the dielectric constant with loss tangents of 0.003 (@ 1 MHz) using sputtered material. The materials have been incorporated as monolithic microwave integrated capacitors in thin-film phase-shifters circuits using Pt-coated silicon substrates, with direct growth of the BST on the Pt. Figure 3 illustrates a simple parallel-plate capacitor structure integrated in shunt along a coplanar transmission-line (CPW). For lowest possible loss, the silicon substrates can be micromachined in the CPW gap region. This also has the advantage of reducing the capacitance per unit length on the unloaded-line, allowing for larger capacitive loading in a circuit. Figure 3 also shows a typical capacitance versus voltage curve for the parallel plate devices fabricated at UCSB, on a 100nm film with a zero-bias dielectric constant of ~300. coplanar waveguide (CPW) etched troughs Si substrate BST capacitor ground ground signal 5 10 1 10 1.5 10 2 10 2.5 10 3 10 -15 -10 -5 0 5 10 15 Capacitance Density vs. Voltage C ap ac ita nc e/ A re a [p F /u m ^2 ] Voltage [V] Figure 3: a) Integrated thin-film capacitor structure on a coplanar waveguide. b) Low-frequency (1 MHz) C-V curve of a typical UCSB device showing 4:1 capacitance variation. Monolithic delay lines are then implemented using his technology. The delay lines are distributed circuits using a coplanar waveguide periodically loaded with thin-film BST varactors, as shown in figure 4. When designed correctly this structure is a synthetic transmission line with a phase velocity that can be controlled by changing the value of the external loading capacitors. The parallel plate capacitor topology utilizes the tunability of the BST film effectively and requires lower control voltages than interdigital designs. Conductor losses are low in this topology since the transmission lines are fabricated on low dielectric constant substrates such as HR-silicon (as opposed to bulk ferroelectric) substrates. Nonlinear applications of such varactor loaded lines have been previously demonstrated with considerable success [3,4]. The use of this topology for linear applications is now gaining popularity, and UCSB has been a leader in the development of low-loss phase distributed shifters [5-6]. UCSB has recently demonstrated analog phase shifter circuits using monolithic GaAs varactors loading a coplanar waveguide (CPW) transmission line in K-band, resulting in < 4 dB insertion loss at 20 GHz and a continuously programmable delay of 0-360 degrees at this frequency.