Micromechanical Resonant Arrays and Methods of Making

The present invention relates to arrays of micromechanical resonant members, and to methods of making such arrays. The resonant arrays may for example be used as memory devices, and may for example be used in tagging devices and in sensors.

WO 2004/084131 and WO 2004/083798, both to the inventor of the present application, describe memory devices and temperature sensing devices that utilize arrays of resonant members to represent data. These members may take the form of cantilever and bridge structures, and may have different resonant frequencies from one another so that the presence or absence of a vibratable resonant member of a particular frequency may be equated to a logical "1 " or "0", and may represent binary code, a status flag or the like. A determination of the presence or absence of a member may be made by applying an excitation signal to the array and by analysing the response to determine if it is indicative of a particular member's frequency.

Data may be encoded in these arrays in a number of ways, e.g. by fabricating members only of particular frequencies, by making members with a full range of frequencies and by then destroying particular members, or by enabling or disabling the ability of particular members to vibrate, e.g. by using a removable tether.

The arrays may be fabricated using MEMS technology (microelectromechanical systems technology), which is also known as MST (Micro System Technology) and micromachining. MEMS technology includes fabrication technologies for integrated circuits, and technologies specifically developed for micromachining. It generally relates to the fabrication of components with dimensions in the range of micrometers to millimeters.

Resonant members are also disclosed in US 5481 102, US 5552778, US 5563583, US 5565847 and US 6819246, and the contents of these documents, and of WO 2004/084131 and WO 2004/083798, are incorporated herein in their entirety by reference.

An array of resonant members of different frequencies from one another can be fabricated by varying the dimensions of the members, e.g. to vary the

lengths and thicknesses of the members, which would also vary their overall masses. However, if these methods are used with standard masks and procedures, the tolerances in the resulting member dimensions and so in their frequencies can be low. This need not necessarily be a concern. However, the nominal frequencies of the members must be spaced apart to a greater extent than their tolerances to ensure that the actual frequencies of adjacent members do not overlap and provide false readings. Therefore, a large tolerance requires the total frequency range over which the array frequencies are spaced to be correspondingly large, and requires interrogators to sweep through a large frequency band in order to read the full set of members. This problem increases as the numbers of resonant members increases.

High resolution masks and laser or electron trimming of the members can improve tolerances, but can be expensive. It would be desirable to provide alternative resonant arrays and fabrication methods, which may provide resonant arrays of different frequencies to good tolerances.

Viewed from one aspect, the present invention provides a method of fabricating an array of micromechanical resonant members, including the steps of: using MEMS techniques to form a set of resonant members of a common first/basic resonant frequency; and using MEMS techniques to provide said resonant members with frequency trimming masses, said frequency trimming masses reducing said first resonant frequency of each resonant member by a different amount, such that said resonant members resonate at frequencies different to one another.

MEMS techniques may include for example masking, deposition and etching steps, amongst other well-known lithographic and micromachining processes, and the making of the arrays with a common first or basic resonant frequency may for example be achieved by making the resonant members of the same dimensions, geometries (shape) and materials. Actual lengths, thickness, geometries and materials may be varied to provide a desired basic resonance.

In the present array, rather than rely on changes in the lengths of the array members or the like, and their bulk properties, all of the resonant members are designed to resonate at the same basic frequency, and differences in resonant frequency may be obtained by associating discrete frequency trimming masses with the resonant members. The use of these frequency trimming masses, which can be small in size, to provide the frequency differences between the resonant members enables the frequency differences to be at higher tolerances than could be achieved by fabricating the members of different length. Therefore, the resonant members can be fabricated to have accurate relative resonances to one another, and so may be designed to have resonant frequencies that are close to one another, without their tolerances overlapping.

Generally, in MEMS fabrications, multiple arrays are formed at the same time in a batch on a silicon wafer, and are then diced into separate individual arrays. The same set of frequency trimming masses may be used for each array on a wafer, as the actual length and frequency of the members of the arrays generally can be made constant across the wafer at a particular length and frequency within the tolerance, the latter also being constant across the wafer. The present method may provide resonant arrays that have relatively narrow frequency bands, using cost-effective and practical techniques and equipment.

Preferably, the basic common frequency of the untuned resonant members is set to be equal to or higher than the highest array frequency that is required. This allows all of the required array frequencies to be set by applying a frequency trimming mass to each of the members to lower each member's basic frequency by an appropriate amount.

In one preferred embodiment, the method includes the steps of forming the resonant members at a common nominal resonant frequency and with a common frequency trimming mass, and then removing at least part of the frequency trimming mass so that the common resonant frequency of the resonant members is set to the basic/first resonant frequency which the masses are designed to alter. This will involve determining an actual resonant

frequency of the resonant members, and the fabrication process preferably involves the steps of forming the resonant members to have a common nominal resonant frequency equal to or higher that the first/basic resonant frequency; adding a frequency trimming mass to each resonant member to ensure that their actual common resonant frequency is equal to or below the first/basic resonant frequency; determining an actual resonant frequency of the resonant members; comparing the actual resonant frequency with the desired basic/first resonant frequency; and removing at least part of the frequency trimming mass so that the resonant members have an actual common resonant frequency equal to the basic/first resonant frequency.

In this embodiment, as well as being able to provide high tolerances in the relative frequencies of the array members, the method is also able to provide high tolerances in the absolute frequencies of the array members. The relatively low tolerance involved in manufacturing the resonant members at the common nominal resonant frequency is compensated for by measuring an actual array resonant frequency, and by altering the frequency trimming mass across all of the members accordingly, e.g. through an etching process, so that the measured actual frequency is modified to the first/basic resonant frequency that the frequency trimming masses are designed to modify. The measurement of an actual resonant frequency and the subsequent trimming compensate for the initial poor tolerances associated with the initial fabrication of the members at the common nominal resonant frequency, and allows the members to be fabricated with absolute resonant frequencies of high tolerance. Thus, the relative frequencies of the members can be defined accurately using frequency trimming masses.

In one embodiment, the first/basic resonant frequency (that the frequency trimming masses are designed to modify) is set to the highest required array frequency, the common nominal frequency to which the resonant members are initially formed is set to the sum of the highest required array frequency and a tolerance associated with the nominal frequency, and the initial amount of frequency trimming mass is set to twice the tolerance of the common nominal frequency.

The amount of frequency trimming mass to be removed may be determined based on a comparison of the measured array frequency and the basic/first resonant frequency, and the determined amount of frequency trimming mass may be removed in one step so as to provide the first/basic resonant frequency. Alternatively, frequency trimming mass may be removed in increments. In this case, an actual frequency measurement may be compared with the desired first/basic resonant frequency after each incremental reduction in frequency trimming mass, so that the desired first/basic resonant frequency can be approached gradually in a stepwise manner. The frequency trimming of each resonant member to a different frequency may be achieved by adding different amounts of frequency trimming mass to each resonant member, with higher masses providing lower frequencies and lower masses providing higher frequencies.

Different amounts of frequency trimming mass may be provided by utilizing frequency trimming masses of different dimensions (e.g. by varying length along a member, width across a member or thickness) or by using different numbers of frequency trimming masses. Preferably, the width of the frequency trimming masses is altered.

Different frequencies may also be achieved by fabricating the frequency trimming masses in an array of different materials, e.g. of different density.

In another embodiment, frequency trimming masses may be added to the members at different locations. Thus, a tuning mass provided further from an anchor point of a resonant member (e.g. towards the centre of a bridge member or towards the free end of a cantilever member) will provide a greater frequency reduction than the same mass located towards an anchor point of the member (e.g. at either end of a bridge structure or at the fixed end of a cantilever structure). It has been found that masses may have a near linear relationship between location along a member and resulting member frequency, and so may be positioned accordingly. A combination of these methods may be used to tune the resonant frequencies of the members, and, preferably, both frequency trimming mass dimension, e.g. width, and mass position are varied across the members of an array.

Frequency trimming masses that provide a relatively coarse frequency adjustment are preferably offset from their most effective positions on the resonant members, and are preferably offset from locations that are most distal from the anchor points of the members (i.e. from the centre of a bridge structure or from the free end of a cantilever structure). Thus, it is preferred to provide a set frequency change by using larger "coarse trim" masses at positions that have less effect than to use smaller "fine trim" masses at positions that have more effect. This is because the expense of a fabrication process typically depends on a "critical dimension" of the required masks (the smallest feature dimension or highest positional accuracy required), and the "coarse trim" masses can set this critical dimension. By using larger "coarse trim" masses at less effective positions, lower critical dimensions can be used. Thus, frequency trimming mass size may be chosen to maintain the critical dimension at values that are attainable by standard MEMS technology and standard masking techniques. For example, masses may be chosen such that the critical dimension of the fabrication process is greater than 0.5 microns, or greater than 1 micron.

Frequency trimming masses may be provided symmetrically on the resonant member, along the member and/or across the member, although this has been found not to be necessary. Frequency trimming masses may be provided on both sides of the centre point of a bridge structure, between the centre point and the two ends.

The appropriate initial frequency trimming mass may also be achieved by a particular mass amount, e.g. defined through dimensions (length, width and/or thickness), and/or through materials, and/or by appropriate positioning on the member. It is preferred that the "fine trim" mass is provided at a portion of the member where it has most effect, e.g. at a position most distal from an anchor point (e.g. at the centre of a bridge member or at the free end of a cantilever member). This allows the amount of mass to be kept to a minimum. As examples, the "coarse trim" masses may be made from high density materials such as gold, tantalum or tungsten. The "fine trim" masses may be made from lower density materials, such as titanium, copper or nickel.

Frequency trimming masses may be removed using an etching process.

They may be removed by a wet or dry etch process, and may be removed by ion beam milling or reactive ion etching.

In one embodiment, the "fine trim" masses are masked when the "coarse trim" masses are removed, so that the tuning masses are protected from the removal process, e.g. etchant. For example, metal or polymer masks could be used to protect the tuning masses. For example, a gold mask may protect the tuning masses from etchants used to remove tungsten trim.

In another embodiment, the trim and tuning masses are made of different materials, such that the tuning masses are resistant to an etchant of the trim mass. Again, therefore, the tuning masses will be unchanged as the trim mass is etched. The tuning masses could for example be made from copper, whilst the trim masses could be made from tungsten.

The materials of the tuning (and trim) masses may be chosen so that sufficiently large amounts of mass are required that the dimensions can be kept above a desired critical dimension associated with the fabrication process.

The measured array frequency may be determined for a particular calibration member. This calibration member may be a member of the array, or may be provided separate from the array. The calibration member may for example be provided in a separate portion of a fabrication wafer from the arrays, e.g. in a wafer indexing area.

The frequency of more than one resonant member could be measured, and, for example, an average result could then be taken. A number of calibration members could be provided at locations across a fabrication wafer or other manufacturing substrate, so that they provide information on the actual frequency across the arrays. Again, an average result could be taken, and also frequencies could be compared to ensure that the fabrication process is working correctly and that beams are being produced consistently across the arrays.

Preferably, calibration members are provided on the outer periphery of an array fabrication wafer, e.g. four at ninety degrees to one another. A calibration member may also be provided at the centre of the wafer.

The calibration members may be of the same configuration (e.g. dimensions and masses) as the basic untuned array members. They however

need not be, so long as the nominal dimensions and masses of the calibration members are known and the resultant measured frequency can be compensated and extrapolated to provide information on the actual frequencies of the array members. Where frequency trimming mass is to be removed in increments, the calibration member preferably includes a frequency trimming mass also, which can then be removed with the removal of mass from the array members.

Calibration members could be provided that include one or more frequency trimming masses on them, and this could provide information on the mass deposition process.

If calibration members are provided at a number of frequencies, then they should be sufficiently spaced that they do not overlap in their tolerances.

In order to measure the frequency of a resonant member, it should be able to vibrate under an excitation signal. Therefore, the frequency measurement is preferably taken after the body of a calibration member (e.g. the basic bridge or cantilever structure) is released from the underlying fabrication substrate, e.g. silicon wafer. This preferably occurs at the same time as the array resonant members are released, so that only a single processing step is required. It would also be possible however to release the calibration members firstly, and release the array members later, e.g. after a trimming and tuning process.

In one embodiment, e.g. as used in WO 2004/084131 and WO 2004/083798, the resonant members of an array are initially tethered and then selectively released during an encoding or enabling process. In this case, the calibration members used in the frequency measurement are preferably fabricated without tethers, or have their tethers removed before measurement.

In order to resonate, the resonant members need to be sufficiently fabricated that they may react to an excitation signal. For a Lorentz force-type array that is mounted in e.g. an RFID tag, an electrical conductor is provided across the array members and attached to an antenna, and a magnetic field is applied normal to the conductor. When an excitation signal is applied to the antenna, an a.c. signal is induced in the conductor, and a Lorentz force acts on the current-carrying conductor due to their presence in a magnetic field. This

force then vibrates the members, and the members will resonate if the a.c. current corresponds to the resonant frequency of the members.

In order to measure the frequency of a Lorentz force member in the present process, an external magnetic field is applied to the fabrication wafer, at least in the region of the calibration members, and the calibration members are fabricated with a conductor across them, the conductor including connection points for connecting with an external excitation device that applies an a.c. current to the conductor and measures the response. Thus, in the fabrication process, the antenna of a Lorentz-type array device may be represented by an a.c. supply circuit. The external excitation device will include suitable circuitry for monitoring the response of the calibration member to the applied a.c. current, and may apply a range of a.c. frequencies to the member, about the member's nominal fabrication frequency.

Other types of resonant member may also be fabricated using the present processes, including magnetic, piezoelectric, electrostatic, magnetostrictive and electrostrictive force members. For each type, suitable external excitation devices, e.g. current applicators and magnetic field applicators, may be provided in order to resonate the members, and to substitute, where necessary, for antennae or magnetic elements of the completed array devices.

An external excitation device may include a pair of electrical probes for applying an a.c. signal to the contacts of a member's conductor. An external excitation device may include a magnetic probe for placement of a magnetic element adjacent a calibration member, or may include a larger magnetic element to apply a magnetic field over the whole of a test station associated with the calibration element or across the whole of a silicon wafer or the like.

As well as providing external excitation devices, it would also be possible to fabricate a calibration member, e.g. of a Lorentz type array, with an antenna on the wafer. This would then remove the need for an external a.c. current supply, and, instead, the calibration member could be interrogated with an excitation signal, e.g. a swept rf signal, at the same time as a magnetic field is applied to it.

The resonant frequency of the calibration members may also be determined in other ways. For example, the resonant members could be excited acoustically or electrostatically. Also, optical vibrometry could be used to detect the resonant frequencies, and may do so at high resolution, as it uses interferometry.

The frequency trimming masses may be added to the resonant members at any time during the fabrication process, and may be added before or after the trim frequency measurement. Preferably, the masses are provided before the trim frequency measurement. There are then fewer fabrication steps after the frequency measurement. Also, the body of the calibration member needs to be released from the surrounding substrate before it can vibrate, and this is preferably done at the same time as releasing the bodies of the array members. It is preferred to add the masses before the member bodies are freed. The masses may also be added before the frequency measurement. If the "fine trim" masses and "coarse trim" masses are made of the same material, then they may be deposited at the same time, using the same masking process.

The desired frequencies to which the members are set by the masses may not always be exactly those required of the final device within which a resonant member array is incorporated, as account may need to be taken of any processing after the trimming that may affect the resonant frequencies. Any such processing should only affect the resonant frequencies in a predictable manner that does not increase tolerance levels past those that are required in the final device.

The required array frequencies may relate to fundamental frequencies of the members, or may relate to higher order modes.

An array of resonant members may be formed from one set of resonant members and one set of frequency trimming masses. Alternatively, an array may be formed of two or more sets of resonant members, each set having a different basic/first resonant frequency. Each set may use the same set of masses, or may use different sets of masses, and each set may be of the same number of resonant members or may have different numbers of resonant member. This may for example be useful when a large number of resonant members are required.

Once the arrays have been trimmed, they may be processed in a standard manner. For example, a wafer of trimmed arrays may be diced to provide individual arrays, and these separate arrays may then be mounted with appropriate excitation circuitry and the like, e.g. antennas, magnets and the like. The array fabrication process may utilise any suitable MEMS technology, and may include for example steps of lithography, deposition and etching. It may include for example photolithography and thin film deposition or growth. Typically, the process results in a laminate structure. A number of structural layers can be formed on a substrate, and required components can be formed by selective etching of the substrate and/or sacrificial materials and component materials deposited thereon. The resulting micromachined components may be combined with electronics that are fabricated using standard integrated circuit processes.

Viewed from another aspect, the present invention provides a method of fabricating an array of micromechanical resonant members, including the steps of: using MEMS techniques to form a set of resonant members of a common resonant frequency equal to or above a desired resonant frequency; using MEMS techniques to provide a common frequency trimming mass on each of the resonant members to ensure that the resonant members have a common resonant frequency below the desired resonant frequency; determining an actual common resonant frequency of the resonant members; and removing at least part of the frequency trimming mass from each of the resonant members in accordance with the determined resonant frequency, so that the resonant members resonant at the desired frequency.

Viewed from a further aspect, the present invention provides a method of fabricating an array of micromechanical resonant members at a set of desired resonant frequencies, including the steps of: using MEMS techniques to form a set of resonant members of a common resonant frequency within a first tolerance; measuring an actual resonant frequency associated with the resonant members; and

processing the set of resonant members in accordance with the determined resonant frequency, using processing steps having a combined second tolerance lower than the first tolerance, such that the resonant members resonate at said set of desired resonant frequencies. The higher tolerance steps could also include the etching of mass from a removable mass layer of a resonant member. The etch could be fully across the mass layer or only in relation to specific portions of the layer.

The higher tolerance steps also may not include etching steps, and could include only the addition of masses. For example, after a measurement of the actual resonant frequencies of the members, a set of appropriate frequency trimming masses could be calculated and placed on the resonant members to take each to a target frequency. This would require a different set of masses for each batch of arrays, e.g. for each fabrication wafer, as the actual basic frequency of the members will be different for each wafer within the production tolerance of the nominal fabrication frequency. Alternatively, after the measurement of an actual resonant frequency, a specific trim amount could be calculated and applied to each resonant member. Again, the trim amount would need to be different for each array batch, but in this case the masses could stay the same between batches. An advantage of the etching of trim is that the same trim masses can be added to each batch of arrays, e.g. wafer, with compensation for tolerance differences being provided by the subsequent etching of the trim masses. Also, the deposition processes can all occur before the main bodies of the resonant members are released from the underlying fabrication structure, e.g. wafer, which is advantageous from a fabrication point of view.

In a further embodiment, after the frequency trimming masses are applied, each array could be provided with a calibration member for use in the interrogation process. The frequency of the calibration member could be set apart from the frequency band of the other array members by an amount greater than the tolerance with which the nominal frequency of the members is fabricated. The specific frequency of the calibration member could then be determined by an interrogator through a sweep of the tolerance band around

the nominal frequency, and then the interrogator could calibrate its sweep of the other array members based on the measured frequency.

Viewed from a further aspect, the present invention provides a method of fabricating an array of micromechanical resonant members the resonant members having resonant frequencies within a desired frequency band, including the steps of: fabricating a plurality of resonant members, the resonant members being designed to have a common resonant frequency to within a set tolerance; depositing a set of fine trim masses on the resonant members, the fine trim masses being designed such that a set of resonant members having the common resonant frequency would form an array of resonant members having a set of desired resonant frequencies within the desired frequency band; depositing a coarse trim mass on each of the resonant members to ensure that an actual common resonant frequency of the resonant members is below the designed common resonant frequency; determining an actual common resonant frequency of the resonant members; and etching at least part of the coarse trim mass from the resonant members based on the determined resonant frequency such that the resonant members have the set of desired resonant frequencies.

The present invention also extends to micromechanical resonant arrays made in accordance with any of the above methods.

Viewed from another aspect, the present invention provides a micromechanical array of resonant members, the resonant members having discrete fine trim masses provided thereon.

Viewed from a further aspect, the present invention provides a micromechanical array of resonant members, the resonant members having discrete coarse trim mass provided thereon.

It should be noted that any one of the aspects mentioned above may include any of the features of any of the other aspects mentioned above and may include any of the features of any of the embodiments described below, as appropriate.

Embodiments of the present invention will now be described, by way of example only, with reference to the accompanying drawings. It is to be understood that the particularity of the drawings does not supersede the generality of the preceding description of the invention. In the drawings:

Figure 1 is a schematic plan view of an RFID tag including an array of resonant members and an antenna;

Figure 2 is a cross-sectional view of a micromechanical resonant member in the form of a bridge member, showing tuning masses provided on the beam;

Figure 3 is a cross-sectional views of a micromechanical resonant member in the form of a bridge member, showing tuning and trimming masses provided on the beam;

Figure 4 is a cross-sectional view of a micromechanical resonant member in the form of a cantilever member, showing a tuning mass provided on the beam;

Figure 5 is a cross-sectional view of a micromechanical resonant member in the form of a cantilever member, showing tuning and trimming masses provided on the beam; Figure 6 is a plan view of a wafer on which multiple arrays are formed simultaneously in a batch process, showing also the use of calibration members on the wafer;

Figure 7 is a flowchart of main processing steps in a fabrication method for a batch of resonant arrays; and Figures 8a and 8b to 16a and 16b are schematic cross-sectional and perspective views of a Lorentz-type resonant member of an array at various stages in its fabrication.

Referring to Fig. 1 , a memory device 100, which may be an RFID tag, includes an array 1 10 of micromechanical resonant members 1 12. A common electrical conductor 1 14 extends in a serpentine manner across the resonant members 112, and connects to an antenna 1 16 at connection points 1 18.

The resonant members may be caused to vibrate by a signal on the conductor 1 14, e.g. an ac current induced by an excitation signal applied to the antenna 1 16, e.g. as discussed in WO 2004/084131.

In one preferred form, the resonant members are configured to be acted on by a Lorentz force, and a magnetic field is applied to the resonant members

112 in a direction perpendicular to the current direction, e.g. through the use of a magnet element (not shown) mounted opposite to the array 1 10 in the RFID device 100.

The resonant members 1 12 are designed to have different frequencies from one another, and, in use, an interrogator 120 applies an excitation signal 122, e.g. a swept RF signal, to the antenna 116 to induce an a.c. current to flow in the conductor 1 14. This causes the conductor 1 14 to experience a Lorentz force perpendicular to the direction of the induced current and the magnetic field and to vibrate the resonant members 112 across which it extends. When the frequency of the a.c. current in the conductor 1 14 corresponds to a resonant frequency of one of the members 1 12, the member 1 12 will resonate and will alter the impedance of the antenna circuit. This change will be reflected back to the interrogator 120, and so the interrogator can determine which of the resonant members exist and are able to vibrate. The presence or absence of a resonant member and its ability to vibrate can represent a logical "1 " or a logical "0", and can be used to represent binary code, a flag status or the like. The memory device can therefore be encoded by fabricating only specific ones of the resonant members, by destroying specific members or by enabling or disabling the members, e.g. by providing a restraining element, such as a tether, and releasing the restraining element, e.g. ablating the tether.

Other forms of resonant member can also be used in a similar arrangement to the above, e.g. capacitive or piezoelectric, and the memory device could also be made using magnetic, magnetostrictive or electrostrictive resonant members.

The resonant members 1 12 may for example take the form of bridge or cantilever structures, and are formed by MEMS technology, e.g. using masking, deposition and etching techniques. They will generally have dimensions in the

order of micrometers to millimeters. For example, for resonant member frequencies in the range of about 10 MHz to about 20 MHz, the resonant member lengths may be of the order of 20 microns.

An important feature of the device of Fig. 1 is that the resonant members 112 should have different resonant frequencies, and that these different frequencies should be resolvable by the interrogator 120.

One aspect of the present invention is to provide micromechanical arrays of resonant members of different resonant frequencies. These arrays may be used in RFID tags, such as shown in Fig.1 , as well as in any other appropriate applications, including in sensors, e.g. as flags, and in memory devices in general, e.g. as discussed in WO 2004/084131 and WO 2004/083798.

In one embodiment, as shown in Fig. 2, the resonant members of an array are formed as a micromechanical bridge-type resonant member 200 that include a bridge structure 210 on which are mounted discrete frequency trimming masses 220.

The resonant members 200 of the array are all designed to have a common basic nominal resonant frequency, e.g. by fabricating the bridge structures 210 so that they all have the same overall dimensions, e.g. are fabricated to have the same length and thickness. One or more frequency trimming masses 220 are then added to each of the resonant members 200, e.g. through a deposition process, so that the resonant frequency of each member 200 is modified (reduced) and so that the resonant members 200 have different resonant frequencies from one another.

The difference in resonant frequency may be achieved by placing a different amount of frequency trimming mass 220 on each resonant member 200. This may be achieved by varying the dimensions, area, or material of a single frequency trimming mass 220. The amount of frequency trimming mass may be varied for example by varying the length of the tuning mass (how far they extend in the longitudinal direction of the member), and/or by varying the width of the frequency trimming masses (how far they extend in the width direction of the member), and/or by varying the number of frequency trimming masses 220 applied to the resonant member, and/or by varying the material of the tuning masses 220, e.g. the use of materials of different density.

The difference in frequency may alternatively or also be achieved by varying the position of the frequency trimming masses 220 on the resonant members. Thus, a given frequency trimming mass 220 will reduce the frequency of a resonant member 200 by a greater extent the further it is from an anchor point 240 of the bridge structure 210, and will reduce the frequency by a lesser amount the closer it is to the anchor points 240. It has been found that there is a near linear relationship between the position of the frequency trimming masses on the resonant member 200 and the reduction in the frequency of the members, and the masses can be positioned accordingly. Preferably, both area of the frequency trimming masses, e.g. width of tuning mass across a member, and position of the tuning masses along a member are used to provide the required frequency shifts in the members of an array.

The use of frequency trimming masses 220 enables the relative frequencies between the resonant members 200 of an array to be set with high tolerance, so that the resonant frequencies of the members 200 can be set close together without overlap of the tolerance bands of adjacent frequencies. The frequency trimming masses 220 therefore allow for the resonant frequencies of an array to be closely packed into a narrow frequency band, so that the interrogator 120 for example need only scan a correspondingly narrow frequency band in order to read all of the possible resonant frequencies.

This contrasts to the manufacture of an array of resonant members of different frequency by making them of different bulk properties, e.g. lengths. With such arrays, it can be difficult to manufacture the lengths to high tolerances without using expensive masking equipment and the like. The use of small frequency trimming masses, however, allows for resonant members to be fabricated with high relative frequency tolerances using standard relatively inexpensive MEMS techniques.

Figure 3 depicts a variation to the embodiment shown in Figure 2, in which, as well as frequency trimming masses 220, resonant members 200 of an array may also include a frequency trimming mass 230. The frequency trimming masses 220 act to shift the resonant frequency of each resonant member 200 coarsecoarsely (for example by some hundreds of kHz) any may

therefore be referred to as "coarse trim" masses. By contrast, the frequency trimming mass 230 acts to shift the resonance frequency of each resonant members 200 finely (for example by some tens of kHz) and may therefore be referred to as "fine trim" masses. The "fine trim" mass 230 is used to provide the resonant members 200' with absolute resonant frequencies of high tolerance, and is a constant value across the resonant members.

During the fabrication process, an actual common resonant frequency of the array members may be determined, and a proportion of the coarse trim masses 220 and/or the fine trim mass 230 removed based on the difference in this actual frequency from a common frequency that the frequency trimming masses are designed to modify. The target frequency and frequency trimming masses will be set such that the combination produces a required set of array frequencies for the final device to which the array is to be applied.

By actually measuring a frequency of the array, and then modifying the frequency trimming masses accordingly, the tolerance associated with the fabrication of the lengths of the bridge structure 210' is removed, and replaced by the tolerance to which the frequency trimming masses can be put down and removed, and the tolerance to which the array frequency can be measured, and enable higher tolerance member frequencies to be achieved. The array frequency that is measured for the trimming process may be a basic resonant frequency of the bridge structure 210, absent the coarse trim masses 220' or fine trim mass 230. It may be measured by providing a separate calibration member that is not part of the array 110, but that is fabricated with the resonant members of the array. The calibration member may have the fine trim mass 230 provided on it, and this would allow for an iterative trim removal process, so that instead of removing the required amount of trim in one go, the fine trim mass and/or coarse trim masses are is removed in increments and the calibration member's resonant frequency is measured after each incremental removal of mass, so that a target frequency can be approached gradually in a stepwise manner.

Fig. 4 shows another micromechanical resonant member 300 that includes a cantilever structure 310 on which are also mounted a discrete coarse trim mass 320. Fig. 5 shows a variant to this embodiment in which a

micromechanical resonant member 300' that includes a cantilever structure 310' on which are mounted a discrete coarse trim mass 320 and in this case, a fine trim mass 330. The same considerations apply to this embodiment as to the bridge-type member 200 of Figs. 2 and 3. In Figures 2 to 5, elements referred to with a number followed by a prime (such as resonant member 200') are identical to elements having referred to by the same number only (such a resonant member 200).

The coarse trim masses and fine trim masses may be provided on the resonant members at any suitable locations, and could be combined together as combined trim/tuning masses. It is preferred however that they are separate. It is also preferred that, in the embodiments shown in Figs. 3 and 5, the fine trim mass is provided at a position of most effectiveness on the members, whilst the coarse trim masses are offset from such a position. Generally, the position of most effectiveness will be a position most distal from the anchor points 240, 240', 340 340' of the bridge or cantilever structures 210, 210', 310, 310' i.e. the centre 250, 250' of the bridge structure 210, 210' or the free end 350, 350' of the cantilever structure 310, 310'.

By having the coarse trim masses offset from the bridge centres and the cantilever ends, relatively more mass is required to effect a set change in resonant frequency. This can be advantageous, because the differential sizing and positioning of the masses over the resonant members may be a critical dimension of the fabrication process, and it is preferred to keep this above a threshold value, e.g. greater than about 0.5 or 1.0 microns, depending upon the optical resolution of the lithography system being used. A critical dimension is the smallest feature dimension or positional accuracy required in a fabrication process and dictates the tolerances of the masks and the like and therefore the costs of the processes. Accordingly, the larger the critical dimension generally the less expensive the processes are, and this may be controlled to some extent by varying the position of the tuning masses along the resonant members.

The coarse trim masses and fine trim masses need not be provided on the resonant members in a symmetrical manner, and in the embodiment of Fig.

3, a single coarse trim mass 220' could be provided on the bridge structure 210' to one side of the fine trim mass 230 only.

Typically, resonant member arrays 1 10 are formed in bulk on a silicon wafer 400, as shown in Fig. 6, where an array 1 10 will be formed at each die location 410 on the wafer. Once the fabrication process is complete, the silicon wafer 400 is diced to provide individual arrays 1 10, and these are then packaged with antennae, magnets and any other further components required in the memory device or the like that the array is to be part of, e.g. the RFID tag of Fig. 1. In accordance with one fabrication method, a set of test stations 420 are defined on the wafer 400. Each station 420 comprises a calibration member 422. The calibration member 422 is preferably identical to the array members 1 10, e.g. in length and the like, except without the coarse trim masses 220. Measurement of the resonant frequency of the calibration member 422 therefore provides the untuned basic resonant frequency of the array members 112, so that the required reduction in the masses of for example Figs. 2 and 3 can be determined.

The exact structure of the test stations 420 will depend on the types of resonant member being fabricated. In the case of a Lorentz force-type member, and also capacitive and piezoelectric members, the test stations 420 will include a conductor 424 and a pair of contacts 426. In order to measure the resonant frequency of the calibration member 422, an interrogation device will apply a.c. current of different frequencies to the contacts 426 of the test stations 420, e.g. through an appropriate electrical probe and test circuitry, and will monitor the response, e.g. for a change in impedance of the conductor 424 and test circuitry combination.

In the case of a Lorentz force resonant member, a magnetic field is also applied across the calibration member 422, and this may be done by an external magnetic field. The magnetic field could be provided as a large field over the whole test station or wafer, or could be more focussed and provided by magnetic elements positioned on a probe that places the magnets adjacent the member 422 itself.

With some resonant members, e.g. magnetic resonant members, a supply of current may not be required, and all that may be necessary may be an excitation signal as would be used to read the array 1 10 in the final device to which it is to be applied. The test stations 420 may be provided at a number of locations across the wafer 400, and a measurement may be taken for each. This may increase the accuracy of the measurement by allowing an average of the five measurements to be taken, and may also be used to confirm that the fabrication process, including for example the lengths of the members 422 are consistent across the wafer 400. Inconsistent results may alert to an unwanted trend across the wafer, which may then be quickly addressed.

More than one calibration member 422 could be provided at each test station 420, and it would be possible to use calibration members 422 of different design, e.g. length, to those of the array members 112, so long as the response of the calibration member 422 can be extrapolated to determine a frequency of the array members 1 12. The calibration member 422 could also for example include known frequency trimming masses, and this could provide a check that the deposition of mass has occurred correctly. The calibration member preferably includes the mass, although it need not do so if the mass is being removed in one go rather than incrementally.

One example of the main stages in a fabrication process for producing an array of resonant members is shown in the flowchart of Fig. 7. Firstly, the resonant members and associated componentry, including array members 1 12, calibration members 422, conductors 114, 424 and the like, are fabricated at step S510 such that the resonant members have a common resonant frequency within a set tolerance. The common frequency is chosen so that the actual basic resonant frequency of the members will always be higher than or equal to the highest required array frequency, so as to ensure that all required array frequencies can be provided by the addition of masses (which will lower the basic frequency of the members). Accordingly, the common resonant frequency at which the array members are fabricated is set equal to at least the sum of the highest required array frequency and the tolerance of the fabrication process at that common resonant frequency.

Next, at step S520, coarse trim masses and optically fine trim masses are deposited onto the resonant members. The coarse trim masses are designed to modify a target frequency (F _ta rget) to produce the required array frequencies, and optionally a sufficient common amount of fine trim mass may be applied to each member such that the resonant frequencies of the array members will be below the target resonant frequency no matter at which point in the tolerance band the common resonant frequency of the members is actually formed at. For example, the nominal frequency may be set to the sum of a highest required array frequency and a tolerance associated with that nominal frequency, the target frequency may be set to the highest required array frequency, and an initial amount of fine trim mass may be set to twice the nominal frequency tolerance.

At step S530, an actual resonant frequency (F _mθasurθd ) of the array 1 12 is determined. As discussed, this may be achieved by measuring the calibration member 422 of a test station 420, and the calibration member may correspond to an untuned basic resonant member of the array 1 10 having the frequency trimming mass.

Once the resonant frequency has been determined, it is compared to the target frequency (F _targθt ) at step S540. If the measured frequency is lower than the target frequency, a calculation is made as to how much frequency trimming mass must be removed in order to provide the target frequency. This amount could be immediately removed from each frequency trimming mass in one go, but in the present embodiment, at step S550 only a portion of the calculated mass is removed (e.g. 50%). This may occur by etching the mass, and this process will be applied over both the array and the calibration member, e.g. over the whole wafer.

After frequency trimming mass is removed, the resonant frequency of the calibration member is remeasured at step S530. This procedure is iterated until the measured frequency equals the target frequency (within acceptable tolerances), at which point the process proceeds to step S560, which the wafer is diced and the arrays 1 10 packaged with the various other componentry required, e.g. antennae and magnets.

In the procedure of Fig. 7, the basic untuned resonant frequency of the array members 112 approaches the target frequency in a gradual stepwise manner, so that the tuned frequencies also approach the required array frequencies in a gradual manner. It will be understood that the steps shown in Fig. 7 may be carried out using various MEMS techniques and that each step may include a number of substeps, including masking, deposition and etching techniques. Also, the steps need not necessarily be carried out in the order shown. For example, the frequency trimming masses may be added to the resonant members before the resonant members are fully formed or defined, e.g. before the bridge or cantilever structures are fully defined or released from the surrounding layers and underlying wafer substrate.

Figs. 8a and 8b to 16a and 16b show schematic cross-sectional and perspective views of a resonant member 1 12 at various stages of its fabrication. Firstly, a silicon wafer 600 is cleaned and prepared, and, as shown in Figs. 8a and 8b, a sacrificial layer 610 e.g. of polyimide is deposited (e.g. ProLIFT 100). Next, as shown in Figs. 9a and 9b, a photoresist layer 620 (e.g. AZ1514) is deposited to define anchor pad locations 630 for the bridge structure. In Figs. 10a and 10b, the sacrificial layer 610 is etched away at the anchor pad locations 630 and the photoresist 620 is removed. In Figs. 1 1 a and 1 1 b, a layer of aluminium 640 is sputter-deposited onto the sacrificial layer 610, and, in Figs. 12a and 12b, an image reversal photoresist 650 (e.g. AZ5214E) is used to define locations 660 for the coarse trim masses and a location 670 for the fine trim mass. In Figs. 13a and 13b, tungsten is sputtered onto the photoresist 650 to produce the coarse trim masses 680 and the fine trim mass 690, and lift-off of the photoresist 650 is performed (e.g. using AZ100).

In Figs. 14a and 14b, the aluminium bridge structures are defined from the surrounding aluminium layer 640 by providing a protective layer 700 (e.g. AZ1512) using photolithography, and in Figs. 15a and 15b, the unprotected aluminium layer 640 is removed by etching to define the aluminium bridge structures, and then in Figs. 16a and 16b, the sacrificial layer 610 is removed (e.g. using AZ726) and the bridge protective layer 700 is removed (e.g. using

AZ100). This releases the bridge members from the underlying silicon substrate 600. Plasma ashing follows to ensure that all residual remnants of photoresist and polyimide are removed.

Measurement of the calibration member can take place once the bridge members are released, and the structure of Figs. 16a and 16b can be etched to remove fine trim mass so as to provide the target frequency discussed above, e.g. using hydrogen peroxide. This may be done with the coarse trim masses masked, so as to protect them from mass removal, or the coarse trim masses may also be etched with the fine trim mass, and may include extra mass therein to compensate for this.

Various other steps are also possible in the above procedure. For example, as the resonant structures are fabricated on a wafer, a calibration member may be made at the same time. It may be made without the tuning masses. Each calibration member may be fabricated with a conductor thereacross and a set of contacts for connection of the conductor with a test probe. The test probe may supply an ac current of various frequencies across the conductor and monitor the response, e.g. through a change in impedance of the calibration member's circuit, and so may determine the resonant frequency of the calibration member. Other methods of detecting the resonant frequency would also be possible, e.g. the use of an interferometry method, such as an optical vibrometry device, as would electrostatic or acoustic excitation and detection.

The conductors 1 14, 424 may be formed at the same time as the resonant bridges, as they can both be formed of aluminium. A whole array of resonant members may be formed from a single set of resonant members and a single set of frequency trimming masses. It would also be possible to form an array from two or more sets of resonant members, each having a different basic frequency. Each set of resonant members could then be modified by a common set of frequency trimming masses, or each set could be modified by a different set of frequency trimming masses. Each set may have the same number of resonant members or different numbers. This may be useful when a large number of resonant members are to be formed.

The above embodiments all relate to the etching of frequency trimming mass after a measurement of an actual frequency of the array members, e.g. through a calibration member. As well as the above methods, the frequency trimming masses could be added after the frequency measurement is made. In this case, the precise amounts of material could be applied without requiring trim and etching, but this would require different sets of frequency trimming mass for each wafer.

In a further embodiment, without any fine trim mass, each array could include a calibration beam within it. This beam could have a frequency spaced from that of the other members by at least the lowest tolerance associated with the fabrication of the common untuned resonant frequency of the members. There would then be no need to measure the actual frequency of the beams or compensate for the lower tolerance during the fabrication process. Instead, an interrogator could first detect the actual frequency of a calibration beam by looking for the frequency in the tolerance range of the basic resonant frequency. The interrogator would then be able to determine the actual frequencies of the main array members by adding the known frequency changes of the tuned masses with the measured basic untuned common frequency of the calibration member. It is to be understood that various alterations, additions and/or modifications may be made to the parts previously described without departing from the ambit of the present invention, and that, in the light of the above teachings, the present invention may be implemented in a variety of manners as would be understood by the skilled person. The present application may be used as a basis for priority in respect of one or more future applications, and the claims of any such future application may be directed to any one feature or combination of features that are described in the present application. Any such future application may include one or more of the following claims, which are given by way of example and are non-limiting with regard to what may be claimed in any future application.