WO2018081092A1

WO2018081092A1 - Saw device with improved thermal conductivity

Info

Publication number: WO2018081092A1
Application number: PCT/US2017/058050
Authority: WO
Inventors: Tomasz Jewula; Maria Claudia CUSTODIO KAJIYAMA; Yoshikazu Kihara
Original assignee: Snaptrack, Inc.
Priority date: 2016-10-27
Filing date: 2017-10-24
Publication date: 2018-05-03
Also published as: DE102016120566A1

Abstract

It is proposed to use a doped lithium niobate as a substrate material for SAW devices having an improved thermal conductivity. MgO is a given as a preferred example for manufacturing SAW devices having a substantially reduced self-heating.

Description

SAW DEVICE WI TH IMPROVED THERMAL CONDUCTIVI TY

CROSS-REFERENCE TO RELATED APPLICATION

This application claims priority to German Patent Application No. 102016120566.4, filed October 27, 2016, which is expressly incorporated herein by reference in its entirety.

Description

The self-heating of micro-electronic devices, which is caused by dissipative losses generated in the device under operation conditions, needs to be reduced. When reducing the maximum operation temperature of a device by avoiding dissipative losses, the performance of the component or device can be improved .

In case of a surface acoustic wave (SAW) RF filters, the frequency stability and the efficiency of the RF signal transmission improves. Furthermore, reduced device

temperature has a positive impact on its reliability. Since most of the ageing mechanisms are temperature dependent, low self-heating leads to significantly improved operation lifetime of the devices.

There are already known different methods for the reduction of self-heating in micro-electronic devices that are formed on substrates with low thermal conductivity. In such devices, the heat cannot be dissipated or guided away because of the low conductivity of the substrate. One of the common ideas is to increase the area of the active parts of the device. This approach leads to reduced energy density and hence to lower losses. Furthermore, the increased component area helps to improve the heat dissipation because of a larger heat flux to the ambience. However, this method has a great disadvantage, which is its large device size, which is contrary to the trend of miniaturization.

Another approach is to introduce heat dissipation layers made by a material of high thermal conductivity and located close to the active area of the device. Preferred are metallic layers but also dielectric layers made of silicon nitride, aluminum nitride or alumina are examples for such materials having higher thermal conductivity.

Another approach to reduce self-heating is using multi-layer substrates. A thin functional layer that is usually the topmost layer of the substrate is deposited on or in a carrier. The supporting material below the functional layer can be a material of a thermal conductivity that is higher than the conductivity of the functional layer. Due to the low thickness of the functional layer, the thermal resistance between the device and the ambience can be reduced compared to a device on a substrate with lower thermal conductivity acting as functional and supporting material. However, a disadvantage of this approach are the high costs of the multi-layer substrates. In the case of surface acoustic wave devices, interaction between the supporting wafer and the acoustic wave are possible.

An object of the present invention is to provide a SAW device having an improved thermal conductivity that avoids the disadvantages of the above-mentioned known solutions.

This object is met by a SAW device according to claim 1.

Further improvements and embodiments are subject of sub^¬ claims . The principle idea of the invention is to reduce the self- heating in micro-electronic devices by improving the heat dissipation ability of the device by an increased heat flux. In currently used SAW devices, the thermal energy, which is generated in the acoustic tracks by transducers, is

transferred via a piezoelectric substrate, bump and package to the ambience that is normally a printed circuit board or another circuit environment. The heat flux is mainly defined by the thermal conductivity of the substrate and by the distance between the heat source (that is the acoustic track) and the heat sink (which, in current technology, is the bump providing the path having the lowest thermal resistance) . Since the distance between heat source and heat sink is design-dependent and cannot be shorted without changing the design, the only possible approach is to increase the heat flux by improving the thermal conductivity of the substrate itself. The invention provides a substrate with improved thermal conductivity, which can be yielded by using a doped substrate material.

In the case of a SAW device, the invention comprises a substrate and metallic transducers arranged on the substrate. The substrate comprises a piezoelectric crystalline material having a doping such that the thermal conductivity of the doped material is higher than the thermal conductivity of the respective piezoelectric crystalline material without doping. With such a material, the heat flux can be improved to about 20%.

Lithium niobate is a piezoelectric material that is commonly used for SAW devices but has only a poor thermal

conductivity. It has been found that lithium niobate, which can be short referred to as LN, that has been grown from a melt by congruent crystal drawing, for example by the

Czochralski method, has crystal defects. As thermal

conduction is done by phonon transport, the presence of defects gives a rise to phonons scattering. As a result, thermal conductivity is reduced.

Hence, for improving the thermal conductivity of such a crystalline material having crystal defects modifying the density of the non-stoichiometric defects is required by any method that is useful for doing that. The crystal defects are mostly due to the composition of the congruently growing LN crystal. This material has a lithium/niobate ratio of about 0.94 resulting in a lithium-deficient material comprising intrinsic defects such as cation vacancies and niob-anti- sites .

Hence, these defects could be reduced by growing an LN material that is stoichiometric. This is possible with a complicated process that requires good and continuous control of the melt composition that has to be continuously adapted.

A better and simpler way to modify the thermal conductivity can be done by doping LN with suitable elements that modify the density of non-stoichiometric defects (anti-sites and cation vacancies) .

Congruently growing crystalline LN is lithium-deficient.

Hence, doping with an ion that can substitute lithium or niob is preferred. Further preferred are dopants such as divalent or trivalent metal ions.

A crystalline LN material having an improved thermal

conductivity comprises lithium niobate doped with magnesium or iron. Such a doping is possible and can be done in the melt phase during crystal growth in a controlled way.

Crystalline LN material having a homogenous doping and a reduced number of crystalline defects can be yielded just as desired. Hence, reduced phonon scattering and improved thermal conductivity of the crystalline LN results.

It can be shown that a SAW device can be manufactured from such a doped LN material by using the crystalline material as a substrate material. The properties of the doped LN material are very similar to those of undoped LN. Moreover, SAW devices can be produced on substrates of doped LN having nearly the same properties as devices have which are made on substrates of undoped LN. Hence, by providing doped LN material, the production of improved SAW devices can be done without any problems.

If applied on an LN substrate a layer that compensates the relative large TCF (Temperature Coefficient of Frequency) of LN improves properties of SAW devices. Such a layer can also be applied onto a SAW device according to the invention. Such a compensation layer can comprise S1O₂ or Ge02 which layer can be applied over substrate and transducer structures that are directly arranged on the LN material.

Manufacture of doped LN can be done by appropriate doping of the material in the molten phase the crystal ingots are drawn from. Starting from a congruent composition of molten

material, doping material comprising the desired ions are added in a desired concentration that may be between 3 and 5 mol%. The doped LN material too grows in a congruent

composition. Therefore, no control of the melting composition during the growing process is necessary. For comparing the effect of doping of an LN material on the thermal conductivity and the stability of SAW devices manufactured therefrom, different wafers are fabricated. A first wafer comprises lithium niobate doped with magnesia. A second wafer comprises lithium niobate doped with iron. A third wafer comprises a "normal" LN material called black LN as used for current SAW devices. On these wafers, SAW test structures are fabricated that behave similar like respective filters and are fabricated with the same parameters but do not provide full filter functionality. All test structures are fabricated with the same manufacturing parameters on the different wafers.

On the surface of the respective wafer, the test structure comprises a multi-layer metallization for a resonator

comprising a copper-based electrode system with a total height of 154 nm. The pitch is scaled to 0.85 μπι and the metallization ratio is set to 0.5. The test structure is similar to a series resonator of a ladder-type structure as used for an RX filter for mobile communication band 2. Over the substrate and the applied metallization, a dielectric material is deposited comprising 655 nm S1O₂ and 90 nm of silicon nitride. The test structure will result in a SAW resonator having a series resonance of about 2,054.1 MHz on the reference wafer. Due to slightly variating optical properties that may have an impact on the lithographic behavior during manufacture, the resonance frequencies of the different wafers differ slightly.

With these test structures, different measurements of electric properties are made to analyze the behavior of the doped material as well as that of the SAW device made

therefrom.

In the following, the invention is explained in more detail by referring to embodiments and the accompanied figures. The figures show the results of the measurements on the test structures and a schematic depicture of a SAW device as well.

FIG. 1 shows the maximum power on device measured for

various LN materials,

FIG. 2 shows in a diagram the time to failure dependent on the incident power,

FIG. 3 shows the self-heating of a SAW device realized on different substrates dependent on the applied RF frequency,

FIG. 4 shows dissipative losses over the frequency and

dependent on the power applied to the device,

FIG. 5 shows a SAW device and the preferred paths of heat dissipation from the heat producing device structures to a heat sink.

FIG. 1 shows in a bare chart the maximum power on device that is the power the device can stand without failure. The power is applied to the test structure as an RF signal of a

frequency within the right flank of the passband. Two

different frequencies are used for measurement that are localized on a point of 3 dB and 6 dB below zero. It is clear that a higher level of power can be applied at the 3 dB point than at the 6 dB point as the bare chart shows. As to the differently doped material, an LN material doped with

magnesia shows the highest power durability. Iron-doped LN is about 0.5 dB below. The power durability of the reference test structure made on undoped LN is about further 0.5 dB smaller .

In FIG. 2, the same power durability is analyzed with a slightly different measurement. Dependent on the applied power in dBm, the time to failure of the analyzed test device is measured. The vertical axis depicts time to failure in a logarithmic scale. It can be shown that on each power level, the test structures made on LN material doped with magnesia shows the highest time to failure better than iron-doped and better than the reference test structure. This is true for all applied power levels.

In FIG. 3, the self-heating of the test structures is

measured in order to prove the reasons for the different power durability. The graph shows three curves that depict the self-heating in Kelvin per Watt dependent on the

frequency of the applied signal. The three curves have different maxima according to the slightly different acoustic properties of the test structures made on different LN materials. Like the respective resonance frequency of the test structures, the maximum of self-heating appears at different frequencies that are rising from undoped LN to iron-doped LN to magnesia-doped LN. The greatest self-heating appears at the test structures made on undoped LN. Iron-doped LN shows a very similar behavior in this measurement.

However, test structures on magnesia-doped LN show a

particular reduced self-heating that is about 20% lower than the self-heating of the two other test structures. This result gives a proof that the different power durability of the test structures are due to the different self-heating.

FIG. 4 shows a measurement of dissipative losses dependent on the frequency of the applied signal. The measuring points are normalized to the applied power on device. Similar like in FIG. 3, the losses for each material show a maximum at different frequencies due to the different resonance

frequencies of the test structures. The figure shows that nearly the same losses result for the three analyzed

materials where only an iron-doped LN material, respectively the test structure manufactured thereon, show some scattering or dispersion, which do not influence the total result.

From the above measurements, it could be shown that a

magnesia-doped LN material is a preferred material for building up SAW devices having an improved power durability and thus an enhanced lifetime. The self-heating of devices made on magnesia-doped LN is substantially reduced.

FIG. 5 shows a cross-section through a SAW device mounted on a ceramic carrier and bonded to a printed circuit board. The substrate SUB is a chip cut from a piezoelectric LN wafer. A metallization forming device structures BES of the SAW device are applied on the surface of the substrate SUB. A dielectric layer comprising a TCF reducing material like S1O₂ may be applied onto the device structures BES. The chip with the device structure is then mounted onto a ceramic carrier in flip-chip arrangement by bonding via bumps. After connecting the substrate SUB to the ceramic carrier CER, this

arrangement may be sealed by a sealing cover that can be made by an epoxy resin, for example. The ceramic carrier with the mounted SAW device is bonded to a PCB that may be part of a mobile phone.

The device structures BES, when operated at RF frequencies produce dissipative losses that result in self-heating. As shown by the depicted arrows, the produced heat is

transferred via piezoelectric substrate SUB, bump BU and vias of the carrier CER to the printed circuit board PCB forming a heat sink. The heat flux is mainly defined by the thermal conductivity of substrate and distance between the heat source that are the operating device structures BES and the heat sink that are de facto the bumps because of their low thermal resistance. Since the distance is design-dependent, the only possible approach to increase the heat flux can be done by an improvement of the thermal conductivity of the substrate as the invention does.

The invention has been described by reference to one

embodiment only but is not restricted to the use of a

magnesia-doped LN material. It is not excluded that other dopants can be found that are improving the heat conductivity of lithium niobate. The given example of magnesia-doped lithium niobate is a preferred example as magnesia-doped lithium niobate has already found use in optical devices. A SAW device made on a substrate of magnesia-doped lithium niobate is not limited to the structure as shown in FIG. 5, for example. It is possible to manufacture and apply any SAW structure on a doped lithium niobate.

Claims

We claim

1. SAW device

- having a substrate (SUB) and

- a metallic transducer arranged on the substrate

wherein

- the substrate comprises a piezoelectric crystalline

material having a doping and a thermal conductivity that is higher than the thermal conductivity of a respective piezoelectric crystalline material without doping.

2. The SAW device of the foregoing claim,

wherein the substrate material comprises doped lithiumniobate LiNb0₃.

3. The SAW device of the foregoing claim,

wherein the substrate material comprises LiN doped with a divalent or three-valent metal ion.

4. The SAW device of one of the foregoing claim,

wherein the substrate material comprises lithium niobate LiNb0₃ doped with Mg ions.

5. The SAW device of one of the foregoing claim,

wherein a TCF compensation layer of S1O₂ or GeC>₂ is arranged over substrate and transducer.

6. Method of producing a SAW device having improved thermal dissipation, comprising the steps

- providing a substrate having improved thermal

conductivity

- applying metallic structures for forming transducers, conductor lines and connection pads - completing the SAW device comprising applying a

temperature compensation layer and making a package for the SAW device

wherein providing the substrate comprises

- providing a melt of lithium oxide and niobium oxide in a composition that allows growing a congruent crystal

- adding a doping material to the melt

- pulling a congruent crystal of doped lithium niobate

from the melt that has an improved thermal conductivity relative to a respective lithium niobate crystal without doping

- cutting wafers from the crystal.

7. The method of the foregoing claim,

wherein adding a doping material comprises adding solid MgO, melting the solid material and homogenizing the melt with the molten doping material.