In current microwave communication technology dielectric resonators (DRs) are key elements for filters, low phase noise oscillators and frequency standards. DRs possess resonator quality factors (Q) comparable to cavity resonators, strong linearity at high power levels, can possess low temperature coefficients, high mechanical stability and small size.

Ceramic dielectric materials are used to form DRs as key components in a number of microwave subsystems which are used in a range of consumer and commercial market products. These products range from Satellite TV receiver modules (frequency converter for Low Noise Broadcast (LNB)), Cellular Telephones, PCN'S.

(Personal Communication Networks Systems) and VSAT (Very Small Aperture Satellite) systems for commercial application to emerging uses in transportation and automobile projects, such as sensors in traffic management schemes and vehicle anti- collision devices. Dielectric Resonators may be used to determine and stabilise the frequency of a microwave oscillator or as a resonant element in a microwave filter.

New systems of satellite TV transmission based on digital encoding and compression of the video signals determine the need for improved DR components. The availability of advanced materials will also enable necessary advances in the performance of DRs used for other purposes as referred to above.

A review on dielectric resonators and materials is given by Wersing in"High Frequency Ceramic Dielectrics and their Application For Microwave Components"in "Electronic Ceramics"Ed BCH Steele Elsevier 67-119.

There are three key properties for a dielectric resonator. The first is the dielectric loss, or tan8. The Q of the resonator (which in the absence of other losses may be approximated to 1/tan8) determines the steepness of the filter skirts, the power requirements and the selectivity. The second is the dielectric constant (su). This determines the size of the resonator. Finally, the TCf, the temperature coefficient of the resonant frequency must be as near zero as possible. For high power filters it is very desirable for the dielectric to possess a high thermal conductivity, preferably greater than 20W/m. K.

A ceramic material for use in dielectric resonators should have a low dielectric loss (high Q), a high dielectric constant, a TCf near zero and a high thermal conductivity.

However it is difficult to obtain a material with all these properties and in Table 1 below the properties of polycrystalline dielectric materials are given Table 1 Material Er TCf Q f(GHz) Thermal ppm/K Conductivity W/mK Al203 9-60 50, 000 9. 2 >20 Ba (Mg1/3Ta2/3)O3 24 0 26,000 10 <4 Ba-Zn-Ta-O 30 -3....3 12,000 6 <4 Zr-Sn-Ti-O 38-3... +3 8000 7 <4 Ca-Nd-Ti-AI-O 43 0 4, 700 10 <4 Ti02 100 450 17, 000 3 <4 SrTiO3 270 1200 400 <4 The Ca-Nd-Ti-O material is disclosed in US Patent 5, 356, 844.

The materials which are currently used as DR materials are those with TCf close to zero. In order to achieve low TCf in dielectric resonator materials the chemical composition of the material is altered. In most temperature stabilised ceramic compositions, the ceramic composition is substantially single phase e. g. barium magnesium tantalate, barium zinc tantalate or zirconium tin titanate which are listed in Table 1. Whilst this is a very desirable approach, it is often the case that altering the chemical composition in order to achieve temperature stability causes a degradation in either the Q or the dielectric constant or both. Additionally, altering the chemical composition usually means that a second phase may be formed with undesirable TCf. For example, Ti02 in combination with Ba forms Ba2Ti09 which has an acceptable TCf of only 2ppm/K'but an inferior dielectric constant of 40 and a Q of 15, 000 at 2GHz. There are certain ceramic composites in which there exist two separate phases of opposite TCf. For example Bi203-Ti02 composites in which a Bi201 phase is formed which possesses a TCf of-533ppm/K, in opposition to the Ti02 phase which possesses a TCf of +450ppm/K. However, the Q factor is rather poor at Q = 1800 at 5GHz and with a TCf of 21 ppm/K.

In general it is well known that ceramic materials possess very poor thermal conductivity. Notable exceptions are beryllium oxide 270W/mK, silicon carbide 67W/mK, aluminium nitride 15W/mK. However, none of these materials possesses particularly low dielectric loss.

Another approach is to use a composite dielectric resonator with improved frequency stability using two different zirconate materials with opposite TCf characteristics. A disadvantage of these materials is that the thermal conductivity of the zirconate materials is less than 4 W/m. K. Temperature compensated whispering gallery mode resonators have been examined for use at cryogenic temperatures. These use a single crystal sapphire disc and sandwiched the sapphire with thin rutile or strontium titanate plates. In the case of rutile plates, high Q and temperature stability was achieved at temperatures between 50-160K. The use of strontium titanate plates,

although achieving temperature compensation, considerably reduced the Q of the composite due to the high dielectric loss of the strontium titanate. The problems with this approach are the cost of the single crystals and the fact that attaching the single crystals effectively is not trivial.

Other attempts to obtain dielectric materials with improved properties are disclosed in US Patents 391672, 3798578, 4580116, 562556 and 5909160.

Aluminium oxide is a well known ceramic dielectric material and has a room temperature dielectric constant of approximately 10 and previous works has shown that extremely low tan8 can be achieved (tan8 = 2x10-5, 300K 10GHZ). The thermal conductivity of alumina is moderately high around 20-30W/m. K at room temperature and this is a great advantage in high power filters operating at around 100W.

Alumina has a TCf of-60ppm/K and it would be highly desirable to be able to tune the TCf while maintaining a low tan8.

We have now devised an alumina based ceramic material with good TCf and low tan8.

According to the invention there is provided a ceramic composition which comprises an alumina sintered body on which there is a layer of titanium dioxide.

In this specification ceramic means any solid inorganic particulate material, the particles of which can be caused to sinter together by the application of heat.

The alumina is preferably doped with from to 0. 05 to 0. 5wt. % of the titanium dioxide and more preferably with from 0. 01 to 1. 0 wt. % of the titanium dioxide. The compositions of the present invention can be made by mixing alumina powder with titanium dioxide powder and compressing and sintering the mixed powder using

titanium dioxide powder and compressing and sintering the mixed powder using conventional methods to obtain a doped alumina ceramic material. It is not a requirement for the present invention that the alumina ceramic be doped with Ti02.

The alumina sintered body or in the preferred form, the Ti doped alumina body will hereinafter be referred to as alumina in order to avoid confusion.

The alumina should contain the minimum of impurities as impurities can adversely affect the dielectric properties of the alumina.

The layer of the titanium dioxide preferably comprises a volume fraction between 0. 0001 to 0. 5 and more preferably between 0. 001 to 0. 05.

The alumina and titanium dioxide layer should be in intimate contact and preferably they are attached by solid state diffusion.

The titanium oxide layer can be formed on the alumina by forming a paste of the titanium dioxide and applying the paste to the alumina, drying and heating to form a dense layer of titanium dioxide on the doped alumina.

The paste of the titanium dioxide is preferably made with non-aqueous liquids such as polymers and non-aqueous solvents, the nature of the liquid is not critical.

The temperature of heating should be sufficient to densify the titanium dioxide and for example is preferably at least 1000°C.

The invention also provides a layered composite dielectric in which the two compositions, the alumina and the titanium dioxide layer, are attached by solid state diffusion and which has a very high Q a high thermal conductivity and a low temperature coefficient of the resonant frequency over a very wide temperature range.

The invention is described in the following Examples in which high purity alumina powder was doped with Ti02 at the 0. 2wt% level. (as described above, doping with Ti02 has been demonstrated to reduce the dielectric loss in sintered alumina considerably but it is not a necessity if a suitably high Q can be obtained in the alumina without doping with Ti02.) The powder was pressed at 100MPa in a 13mm stainless steel die set. The discs were sintered in air in a muffle furnace at a ramp rate of 5°Cmin~l to 1600°C, dwell for 1 hour and 10°Cmin~l to room temperature. The discs were weighed. The unloaded Q of 27 samples was determined at a frequency of approximately 10GHZ and found to be Q = 54, 772, 1 standard deviation = 2714, coefficient of variation = 4. 9%.

A thick film paste of Ti02 was prepared with high purity Ti02 powder. The powder was thoroughly mixed with a vehicle comprising non-aqueous polymers and solvents on a Marchant 3 roll mill. A special jig was designed in which different thickness of shim ranging from 100, um to 300um could be placed on top of the sintered alumina disc. This enabled the manufacture of a range of depths for the Ti02 coating. The thick film paste was applied on to the surface of the alumina disc. The paste was dried at 80°C and then the composite was fired at 1400°C for 1 hour. This was sufficient to densify the Ti02. The result was a dense Ti02 layer on a dense alumina disc. The composite disc was then weighed and the weight of Ti02 was determined by difference. The volume fraction of the Ti02 and A1203 was determined by noting that the density for A1203 is 3. 97Mg m~3 and that for Ti02 is 4. 26Mg m~3.

The unloaded Q was measured at approximately 10 GHz by a resonant cavity method using the Tells mode. The sample was placed in an oxygen-free high conductivity copper cavity on a 4 mm high low loss quartz spacer. The cavity was 30mm in diameter, with adjustable height. The surface resistance of the copper was calculated from the Q of the TEol, resonance of the empty cavity to allow the results to be corrected for the loss due to the cavity walls. The TEols mode was examined using a

vector network analyser (Hewlett Packard HP8720C) with 1Hz resolution. Particular care was taken to ensure that the samples were dry. On removal from the furnace, samples were placed in a vacuum desiccator over silica gel, using plastic tweezers to minimise surface contamination ; when cool, samples were placed in the microwave cavity for measurement.

The TCf measurements were performed over a temperature range of 150K-320K.

The sample was placed on the floor of a copper cavity identical to that described above but without a quartz spacer. Cooling and heating the cavity was achieved by placing the cavity on the cold-head of a closed cycle Gifford McMahon two stage cryocooler (CTI) capable of a temperature range 320K-10K.

The results are shown below, The turning point temperature Tp is defined as that temperature where the TCf is zero Example 1 The volume fraction of Ti02 was determined to be 0. 0042 which resulted in a composite with a Tp of 134K and a Q of 4 1, 000 at room temperature.

Example 2 The procedure in example 1 was repeated except that the volume fraction of Ti02 was. 0086 which resulted in a composite with a Tp of 203Kand a Q of 34, 100 at 300K Example 3 The procedure in example 1 was repeated except that the volume fraction of Ti02 was. 0093 which resulted in a composite with a Tp of 200 and a Q of 34, 800 at 300K

Example 4 The procedure in example 1 was repeated except that the volume fraction of Ti02 was. 0115 which resulted in a composite with a Tp of 245K and a Q of 31, 400 at 300K Example 5 The procedure in example 1 was repeated except that the volume fraction of Ti02 was. 0116 which resulted in a composite with a Tp of 250K and a Q of 31, 800 at 300K Example 6 The procedure in example 1 was repeated except that the volume fraction of Ti02 was. 012 which resulted in a composite with a Tp of 272K and a Q of 30, 000 at 300K Example 7 The procedure in example 1 was repeated except that the volume fraction of Ti02 was. 012 which resulted in a composite with a Tp of 265K and a Q of 32, 200 at 300K Example 8 The procedure in example 1 was repeated except that the volume fraction of Ti02 was. 0128 which resulted in a composite with a Tp 298K of and a Q of 28, 200 at 300K Example 9 The procedure in example 1 was repeated except that the volume fraction of Ti02 was. 0135 which resulted in a composite with a Tp of 286K and a Q of 29, 100 at 300K

Example 10 The procedure in example 1 was repeated except that the volume fraction of Ti02 was. 017 which resulted in a composite with a Tp of 364 and a Q of 24, 100 at 300K Example 11 The procedure in example 1 was repeated except that the volume fraction of Ti02 was. 0225 which resulted in a composite with a Tp of 384K and a Q of 26, 600at 300K The results are summarised in Table 2 Table 2 EXAMPLE No. Sample no Volume Tp (K) Q at 300K TCf 300K fraction TiO2 I 1402D. 0042 134 41, 100-44. 4 2 1493D .0086 203 34, 100-23. 4 3 1401D. 0093 200 34, 800-20. 2 4 1401E. 0115 245 31, 400-10. 5 1401G. 0116 250 31, 800-10. 0 6 1403E. 012 272 30, 000-8. 2 7 1402F. 012 265 32, 200-8. 2 1401F. 0128 298 28, 200-4. 8 9 1402E. 0135 286 29, 100-1. 8 10 1403F. 017 364 24, 100 12. 5 11 1402C .0225 384 26,600 33. 7