Microelectronics Reliability xxx (2015) xxx–xxx

MR-11657; No of Pages 6

Contents lists available at ScienceDirect

Microelectronics Reliability

j ourna l homepage: www.e lsev ie r .com/ locate /mr

Reliability of platinum electrodes and heating elements on SiO2 insulation layersand membranes

R. Rusanov a,⁎, H. Rank a, J. Graf a, T. Fuchs a, R. Mueller-Fiedler a, O. Kraft ba Applied Research 1, Microsystem Technology, Robert Bosch GmbH, Gerlingen, Germanyb Institute for Applied Materials, Karlsruhe Institute of Technology, Karlsruhe, Germany

⁎ Corresponding author.E-mail address: radoslav.rusanov@de.bosch.com (R. R

http://dx.doi.org/10.1016/j.microrel.2015.06.1060026-2714/© 2015 Elsevier Ltd. All rights reserved.

Please cite this article as: R. Rusanov, et al., ReMicroelectronics Reliability (2015), http://

a b s t r a c t

a r t i c l e i n f o

Article history:Received 25 May 2015Received in revised form 22 June 2015Accepted 24 June 2015Available online xxxx

Keywords:PlatinumElectromigrationReliabilityPECVDSiO2Thermal conductivity

In thiswork, failuremechanisms of Pt electrodes including adhesion problems,materialmigration due to thermallyinduced compressive stress and electromigration that could occur in the platinum electrodes and heater structuresat temperatures above 600 °C have been systematically studied, after the deposition. Lifetime determination,scanning electron microscopy and XRD analysis have been applied for samples which have experienced differentloading conditions in order to qualitatively and quantitatively understand the phenomena. Electromigration testingis performedwith the aim to enable time-to-failure prediction for sensor elements and compare different platinumlayers in terms of their stability. Dedicated, application-related test structures are used so that the results are appli-cable to sensor lifetime estimations. Furthermore, a method for the determination of thermal conductivity of thininsulating films has been adapted for the characterization of plasma-enhanced chemical vapor deposition(PECVD) silicon oxide and successfully applied on two materials with different deposition recipes. These two ma-terials are used for the fabrication of platinum-based heating elementswith PECVD SiO2 as insulation ormembranelayer. The results for the two recipes are similar but with a significant difference. A slight increase of the conductiv-ities has been observed due to a thermal anneal of the test structures at temperatures above 700 °C.

© 2015 Elsevier Ltd. All rights reserved.

1. Introduction

Human-caused environmental polluting is one of themain challengesof our modern society. One key aspect of environmental protectionconcerning personal mobility is fuel consumption reduction in combus-tion engine vehicles by sensor aided engine control [1]. Moreover, thesoot particle emission is suspected to cause many respiratory systemdiseases, so that its reduction by using a diesel particulate filter andmon-itoring is at least of great importance as well. Gradual improvement ofboth factors is embedded in the new EU/US emission standards [1],and, thus, the development of appropriate sensors is necessary to reachthe according goals.

A simple schematic of a soot particle sensor is shown in Fig. 1 togetherwith the sensor's position in the exhaust gas system. The functionalprinciple of the soot particle sensor is based on the deposition of sooton its surface, which leads to the formation of electrically conductivepaths between the interdigitated thin-film platinum electrodes. Whena voltage is applied between the electrodes, an electrical current canbe measured that is correlated to the total amount of deposited soot.The sensor can thus be used formonitoring the condition of the particlefilter of a combustion engine vehicle. The sensor needs to be regeneratedperiodically by heating up to a temperature of at least 700 °C to burn the

usanov).

liability of platinum electroddx.doi.org/10.1016/j.micror

soot deposited on its surface. Such sensors are typically fabricated on aceramic substrate with sintered platinum electrodes and backside heaterelements [2]. Using micromachining methods for the fabrication of suchsensors allows for the reduction of the electrode spacing and thus an in-crease of the sensitivity.

The materials of the soot particle sensor are exposed to the harshenvironment of the exhaust gas system of a combustion engine vehicleincluding temperatures of up to 500 °C and a corrosive atmosphere.During the regeneration phase the temperature can reach values evenas high as 800 °C. Therefore, robust materials are required for the sootparticle sensor and thin-film platinum has been chosen as electrodeand heater material, and silicon oxide as insulation and/or membranematerial. The silicon chip of the sensor is mechanically stable for temper-atures of at least 900 °C and thin silicon oxide film forms on its surface,which protects it from further degradation in the corrosive atmosphere.

2. Thermal properties of PECVD SiO2

In the micromachined version of the soot particle sensor, siliconoxide is used as an electrical insulation layer between theplatinumelec-trodes and the silicon-chip substrate. Additionally, a membrane can beformed under the interdigitated electrodes of the sensor by etchingthe silicon of the chip from the back by KOH or dry reactive etching(DRIE). In this case, a slightly tensile stress condition is required forthe stability of the oxide membrane, which should have a thickness of

es and heating elements on SiO2 insulation layers and membranes,el.2015.06.106

http://dx.doi.org/10.1016/j.microrel.2015.06.106mailto:radoslav.rusanov@de.bosch.comJournal logohttp://dx.doi.org/10.1016/j.microrel.2015.06.106http://www.sciencedirect.com/science/journal/www.elsevier.com/locate/mrhttp://dx.doi.org/10.1016/j.microrel.2015.06.106

Fig. 1. Schematic view of the soot particle sensor.

2 R. Rusanov et al. / Microelectronics Reliability xxx (2015) xxx–xxx

at least 5 μm. These two requirements can bemet by applying a plasma-enhanced chemical vapor deposition (PECVD) process, which allows forhigh deposition rates and the adjustment of many material parametersby recipe variation. An Oxford Instruments Plasmalab System 100(PL100) with 13.56 MHz driven parallel plate reactor has been usedfor the deposition of the PECVD SiO2 layers from nitrous oxide N2Oand silane SiH4 at a temperature of 400 °C. Two recipes are studiedhere: a stoichiometric one and a Si-rich SiO2 layer with increased silaneflow. The platinum thin-film electrodes with titanium adhesion layerfor the test structures are deposited, like in the case of the soot particlesensor itself, by magnetron sputtering and are patterned by the lift-offtechnique.

The design of the heater element of the sensor has the aim to ensurea homogeneous temperature distribution. This is required, since theminimal soot burning temperature needs to be reached over thewhole area of the interdigitated electrodes, but no parts of the heatershould exhibit a significantly higher temperature in order to reducethe load and to ensure its stability. Therefore, the thermal conductivityof the SiO2 deposited by PECVD is an important parameter for the sensordesign. The understanding of its anneal and aging behavior is necessaryin order to ensure the sensor reliability and its stability over lifetime.

The method for the determination of thermal conductivity, adoptedand successfully applied in this work, is similar to the one presented in[3]. It is based on a micro-machined test structure consisting of a 5 mmlong, narrow stripe with a PECVD SiO2 layer packed between two plati-num layers. A schematic cross-section of the test structure is shown inFig. 2. The upper platinum stripe is operatedwith high electrical currentin order to generate a considerable amount of Joule heat, which flows tothe substrate through the structure leading to a significant rise of the

Fig. 2. Schematic cross-section of themulti-layer stripewith a PECVD SiO2 layer sandwichedbetween two platinum layers used for heating and temperature measurement.

Please cite this article as: R. Rusanov, et al., Reliability of platinum electroMicroelectronics Reliability (2015), http://dx.doi.org/10.1016/j.micror

temperature difference between the two platinum stripes. The meantemperatures of each of the stripes can be determined from their elec-trical resistance. The amount of heat flowing through the structurecorresponds to the electrical power dissipated in the upper stripe. Thelower platinum stripe is operated with very low current to avoid anyinternal heat-up. The temperature difference ΔT and the heat flowdQ / dt are connected by the equation ΔT= dQ / dt · Rth where the ther-mal resistivity is defined as Rth=1 /κ ·Gwith a thermal conductivity κ ofthe PECVD SiO2. G is a geometry factor. Fig. 3 shows the simulatedtemperature distribution over the cross-section of the test structurewith a temperature gradient from theupper to the lowerplatinumstripesleading to the heat flow. The terrace like geometry, necessary to facemask misalignment generated by the exposure tool, leads to colderedges of the lower platinum stripe, which do not contribute to the heatflow. These colder edges reduce the mean temperature of the stripeand this way increase the measured temperature difference to theupper stripe. This effect needs to the corrected for in order to obtain thetemperature difference relevant for the heat flow. The correction factorsdepend on the geometrical parameters of the structure and the actualthermal conductivity of the PECVDoxide layer. Theyhave been calculatedto be quite significant between 0.45 and 0.90 using ANSYS FEM simula-tions and are applied recursively on the measurement results. The FEMsimulation shows that heat dissipation to the environment is negligiblefor temperatures up to 500 °C.

Four different layer thicknesses of the PECVD silicon oxide between400 nm and 1700 nm have been studied for each of the two PECVDrecipes, so that a total of eight wafers have been processed. Eachwafer contained twelve test structures separated in two groups: widetest structures with a width of 52 μm for the PECVD SiO2 stripe andnarrow test structures with a width of 22 μm. The measurements areperformed at wafer-level using the PA200 Suss probe station with aheat chuck. The temperature of the chuck can be controlled between10 °C and 300 °C with a precision of 2 K allowing for the calibration ofthe resistance-temperature characteristic of the platinum stripes. Thischaracteristic is used for temperature determination of the stripesduring the actual experiment when high electrical power is applied tothe upper stripe and Joule heating occurs. The Joule heating increasesthe temperature of the PECVD oxide reaching values almost 150 Khigher than those for the chuck temperature. The electrical measure-ments are performed by a Keithley 2602 source-meter featuring twofour-wire channels allowing the simultaneous measurement of the re-sistance of the upper and lower platinum stripes with variable testingcurrents. The test structures have been annealed at 500 °C before themeasurement of the thermal conductivity in order to stabilize thethermo-electrical properties of the platinum stripes.

des and heating elements on SiO2 insulation layers and membranes,el.2015.06.106

http://dx.doi.org/10.1016/j.microrel.2015.06.106

Fig. 3. ANSYS FEM simulation of the temperature distribution over the test structure's cross-section when electrical heat is dissipated in the upper platinum stripe.

3R. Rusanov et al. / Microelectronics Reliability xxx (2015) xxx–xxx

The thermal conductivities of the four different layer thicknesses andtwo structurewidths are presented in Fig. 4 as a function of the temper-ature for up to 450 °C for the stoichiometric PECVD SiO2 recipe. The re-sults for the Si-rich oxide layers show a similar behavior with a slightlylower thermal conductivity. All eight measurements agree well witheach other within their uncertainties. This shows that the FEM basedmodel of the test structures describes their behavior well since it isrobust against geometry variations. Additionally, it can be concludedthat the thermal conductivity of the PECVD silicon oxide layers showsno or only very little thickness dependence. The mean values of thespecific thermal conductivity for the two recipes are presented inFig. 5 together with the values for fused bulk quartz [4] and other mea-surement for PECVD SiO2 thin films [3]. They are lower than the valuesfor bulk quartz and exhibit a significant difference between each other,which shows the influence of recipe variation on thematerial properties.

In order to investigate the effect of a thermal anneal at high temper-atures on the thermal conductivity of the PECVD SiO2 layers, the teststructures have been tempered at 750 °C for 20h and themeasurementsrepeated. Such an anneal is expected to cause a reduction of the hydrogencontent of the layer through diffusion [5] and possibly additional oxida-tion, especially for the recipe with increased silicon content, whichwould change the thin-film material properties.

Fig. 6 shows the results of the thermal conductivity measurementsfor the stoichiometric PECVD SiO2 after the thermal anneal. While thevalues for the narrow test structures agree with each other and areslightly higher than the results for the non-annealed samples, the valuesfor the wide test structures are significantly lower and deviate stronglyfrom each other. The results for the test samples with Si-rich PECVDoxide show a similar behavior. The only difference to the results forthe stoichiometric PECVD SiO2 is that for the Si-rich layers also thenarrow structures with the thinnest oxide layer are significantly lowerthan the results for other narrow structures.

Fig. 4. Thermal conductivities of the stoichiometric PECVD SiO2 recipe vs. temperature forall geometry variations (layer thicknesses and stripe widths).

Please cite this article as: R. Rusanov, et al., Reliability of platinum electrodMicroelectronics Reliability (2015), http://dx.doi.org/10.1016/j.micror

For the investigation of this behavior of the annealed test structures,focused ion beam(FIB) and scanning electronmicroscopy (SEM) imaginghas been applied after the thermal conductivity experiments. An examplefor the images of thewide test structures is presented in Fig. 7, where thecross-section of the multi-layer stripe is visible. Next to the large, abnor-mally grown crystallites, which have been studied in detail in our earlierpublications [6,7], even larger, “bubble-like” structures are visible on thesurface of the upper platinum thin film. They are presumably due todelamination caused by the difference in the coefficients of thermalexpansion between the platinum and the silicon substrate leading tohigh compressive stress in the Pt layer [8]. These structures are visibleon all test structureswhich exhibit lowermeasured thermal conductivityvalues. The thermal flow through the interface of the two layers isreduced due to these defects, which artificially reduces the measuredthermal conductivity of the PECVD silicon oxide.

For the calculation of the proper values of the thermal conductivityof the SiO2 layers after the thermal anneal, only the results for thenarrow test structures have been used, excluding the test sampleswhere delamination has been observed. The resulting thermal conduc-tivities are plotted vs. temperature in Fig. 5. The values for the tworecipes have increased compared to the non-annealed samples andare closer to each other, probably due to the increased density of thePECVD silicon oxide [5].

3. Reliability of platinum thin films

As observed for the thermal conductivity test structures, extremeconditions similar to these in the exhaust gas system of a combustionengine vehicle can cause failure even to such a mechanically stableand chemically inert material as platinum. Temperatures above 700 °C

Fig. 5. Thermal conductivities of the two PECVD SiO2 recipes vs. temperature compared tobulk fused quartz and other measurements for PECVD silicon oxide [3,4].

es and heating elements on SiO2 insulation layers and membranes,el.2015.06.106

http://dx.doi.org/10.1016/j.microrel.2015.06.106

Fig. 6. Thermal conductivities of the stoichiometric PECVD SiO2 recipe vs. temperature forall geometry variations after the thermal annealing.

Fig. 8. XRD spectra around the platinum {111} peak measured for the pure-platinum testsamples on a silicon chip before and after anneal in air and in N2 atmospheres.

4 R. Rusanov et al. / Microelectronics Reliability xxx (2015) xxx–xxx

and high stress conditions due to the mismatch in the coefficients ofthermal expansion of the platinum layer and the substrate can lead notonly to adhesion issues but also to grain growth and material relocationdue to stress migration. Moreover, at these high temperatures the diffu-sivity of foreign atomswithin the platinum increases so that they can dif-fuse into the layer and alter its properties. In addition, electromigrationphenomena can occur in the platinum structures of the soot particlesensor due to the high current densities in the electrodes which are nec-essary to produce the high temperatures.

For the investigation of these phenomena and especially of theelectromigration-caused failure in the platinum structures, qualitativeand quantitative experiments have been designed and performed onvarious platinum thin films. These platinum films and their Ti or Taadhesion layers are deposited by DC magnetron sputtering to typicalthicknesses between 300 and 500 nm. Some results have been alreadypresented in [6,7]. Here, additional results will be presented and theelectromigration lifetime of the layers will be related to the materialevolution at high temperature. The parameters of Black's equation [9]will be shown for Ti and Ta adhesion layers, Ti alloying of the Pt layer,and for SiO2 passivation.

Fig. 8 presents the XRD spectra of platinum thin filmswithout an ad-hesion layer around the platinum {111} peak before and after thermalannealing at 730 °C for 20 h in air and N2 atmospheres. The shift ofthe main peak corresponds to a change of the distance of the crystallo-graphic planes parallel to the surface towards smaller values, probablydue to a more tensile stress condition of the layer. The narrowing andsplitting of the main peak into several peaks suggests the developmentof a multimodal grain size distribution, with different plane distances.

Fig. 7. SEM image of a FIB cut through a thermal conductivity test structure sho

Please cite this article as: R. Rusanov, et al., Reliability of platinum electroMicroelectronics Reliability (2015), http://dx.doi.org/10.1016/j.micror

The distribution may consist of platinum grains with higher mean grainsizes compared to the non-annealed samples. The grain growth has alsobeen observed in FIB/SEM images of “as deposited” and annealed plati-num layers presented in [7]. In addition to these results, a platinum{100} peak emerges after the thermal annealing as shown in Fig. 9. This{100} peak could correspond to the abnormally grown grains since thegrain orientation is expected to affect the growth rate [10]. The XRDresults of the platinum samples with Ti adhesion layer are very similarto the results for the pure-platinum samples shown here with the differ-ence that the {100} peaks are less pronounced.

In order to investigate the effect of the phenomena taking place inthe platinum layer, the resistivities of five platinum samples havebeen measured before and after a thermal anneal at 730 °C for severalhours. Fig. 10 shows the resistivity values for the non-annealed samplesand Fig. 11 shows the results for the platinum samples annealed in air.The non-annealed pure-platinum sample shows a linear resistivity-temperature characteristic, with reduced resistivities after the thermalanneal due to the grain growth. The platinumwith a titanium adhesionlayer shows a resistivity that grows stronger than linear above 300 °C,probably due to diffusion of titanium atoms into the platinum andalloy formation, which reduces the electrical conductivity [11]. Afterthe thermal anneal the resistivity-temperature characteristic becomeslinear and the resistivity reduces. The titanium reacts with the oxygendiffusing into the platinum [7] and is, this way, removed from the solu-tion, and its effect on the resistivity is strongly reduced [12]. The platinumsamples with 10 vol.-% titanium addition have higher resistivity values,

wing large abnormally grown crystallites and delamination in the Pt layer.

des and heating elements on SiO2 insulation layers and membranes,el.2015.06.106

http://dx.doi.org/10.1016/j.microrel.2015.06.106

Fig. 9. XRD spectra around the platinum {100} peak measured for the pure-platinum testsamples on a silicon chip before and after anneal in air and in N2 atmospheres.

Fig. 11. Specific electrical resistivities vs. temperature of 730 °C annealed platinum thin-filmsamples with different adhesion or mixed layers or SiO2 passivation.

Fig. 12. SEM image in topographical mode of a platinum stripe failure, showing a narrow

5R. Rusanov et al. / Microelectronics Reliability xxx (2015) xxx–xxx

but show a similar behavior. For the non-annealed samples with tanta-lum adhesion a linear resistivity–temperature characteristic is observed,but the resistivity reduces after the thermal anneal to values close tothe ones of the Ti-adhesion samples, thus a similar diffusion-basedmate-rial evolution can be assumed.

The passivated platinum samples show a higher resistivity com-pared to all other platinum layers, since they have been heated up to400 °C for several minutes during the deposition of the passivationPECVD oxide and the titanium has already diffused into the platinum,reducing the electrical conductivity of the layer. During the heat-up ofthe non-annealed samples, the resistivity grows stronger than linear,probably due to further diffusion of titanium into the platinum. Theannealed samples show a higher resistivity compared to the non-annealed ones due to blocking of oxygen during the annealing by thepassivation, and the titanium remaining in solution in the platinum.

Electromigration [9] can cause failure in thin-film heater elementsdue to an effective depletion of material caused by an unbalanced flowof platinum atoms. This phenomenon can take place on a macroscopicscale due to temperature gradients or current density divergences or ona microscopic scale in the grain boundaries of the platinum. A detailedcomparison between the electromigration-driven failure on a macro-scopic and microscopic scale has been presented in [7]. Although it isalmost impossible to fully avoid temperature gradients and currentdensity divergences, the design of a heater element should aim to mini-mize them, since they are extremely critical for the device reliabilityand strongly reduce its lifetime.

Current density divergences and temperature gradients almost alwaysoccur in real heater elements. It is, nevertheless, necessary to study opti-mized test structures, where the damage forms on a microscopic scale

Fig. 10. Specific electrical resistivities vs. temperature of non-annealed platinum thin-filmsamples with different adhesion or mixed layers or SiO2 passivation.

Please cite this article as: R. Rusanov, et al., Reliability of platinum electrodMicroelectronics Reliability (2015), http://dx.doi.org/10.1016/j.micror

due to current density divergences in the grain boundaries. In this case,the parameters of Black's equation are specific for the studied material,and the results can be used for a comparison between layers with differ-ent fabrication processes.

A typical failure of a platinum stripe on oxidized silicon substrateafter loading with a current density above 20mA/μm2 at a temperatureabove 650 °C for several hours is shown in Fig. 12. The failure occurredat the middle of the stripe (hottest area) and additional damage isvisible on the stripe in its vicinity. This failure morphology indicates abehavior with void growth and not nucleation as the limiting

interception in the middle along the length of a massively damaged stripe.

Fig. 13. Cumulative distribution of the failure times of an electromigration experimentplotted together with the fitted cumulative and probability distribution functions.

es and heating elements on SiO2 insulation layers and membranes,el.2015.06.106

http://dx.doi.org/10.1016/j.microrel.2015.06.106

Table 1Parameters of Black's equation.

Material Act. energy Current exponent

Pt etched 1.22 eV 2.09Pt/Ta etched 1.76 eV 1.97Pt/Ti etched 1.67 eV 1.92Pt + Ti/Ti etched 2.02 eV 2.09Pt/Ti lift-off 1.29 eV 2.03Pt/Ti passivated 0.47 eV 2.43

6 R. Rusanov et al. / Microelectronics Reliability xxx (2015) xxx–xxx

mechanism. The cumulative distribution of the failure times of a set ofsamples under identical loading conditions follows a log-normal distri-bution (Fig. 13).

The parameters of Black's equation determined for differentplatinum layers are summarized in Table 1. A current exponent of ~2has been determined for most of the tested material systems similarto the classical Black experiments [9]. The activation energies dependstrongly on the fabrication process of the studied test samples. Thevalue for the pure-Pt samples is, with 1.22 eV, low compared to theself-diffusion activation energy in bulk platinum of 2.89 eV [13], sug-gesting that the electromigration-based damaging takes place in thegrain boundaries and at the surface of the layer. The titanium or tanta-lum adhesion layers, which lead to Ti or Ta oxide formation in the plat-inum grain boundaries, increase the activation energy of the samplesand improve their stability as already observed for aluminium additionto copper conductive lines [14]. An increase of the titanium content inthe layer increases further the activation energy and improves thelayer stability. The lift-off samples with Ti adhesion show a lower acti-vation energy compared to the corresponding etched samples, probablydue to theminor edge quality as an effect of the fabrication process. Thepassivation of lift-off samples increased their lifetime by an order ofmagnitude, but also altered the failure mode, so that the parametersof Black's equation cannot be compared to the other samples. Here,the failure occurred due to damage in the passivation by hillock forma-tion in the platinum stripe.

4. Conclusions

The test structures developed in this work are capable of measuringthe thermal conductivity of insulating layers at temperatures of up to400 °C with high precision. The measurements show that the different

Please cite this article as: R. Rusanov, et al., Reliability of platinum electroMicroelectronics Reliability (2015), http://dx.doi.org/10.1016/j.micror

recipes for the PECVD SiO2 and the aging of the layers can strongly influ-ence the thermal conductivity, which need to be accounted for in thedevice design. The method is also capable of detecting adhesion defectsat the platinum-oxide interface.

The diffusion of titanium and tantalum from the adhesion layerand of oxygen from the ambient atmosphere into the platinum donot only have an effect on the specific resistivity of the Pt layer but alsoimprove the stability of the heater elements in terms of electromigrationfailure.

Acknowledgments

We are thankful to the microTEC Suedwest Cluster and the GermanFederal Ministry of Education and Research for funding the joint projectbetween the CR/ARY Department of the Robert Bosch Company and theInstitute for Applied Materials from KIT, Germany under the projectnumber s16SV5126.

References

[1] T. Johnson, Review of diesel emissions and control, SAE Int. J. Fuels Lubr. 3 (1)(2010) 16–29.

[2] T. Ochs, et al., Particulate matter sensor for on board diagnostics (OBD) of dieselparticulate filters (DPF), SAE Int. J. Fuels Lubr. 3 (1) (2010) 61–69.

[3] M. Kleiner, et al., Thermal conductivity measurements of thin silicon dioxide films inintegrated circuits, IEEE Trans. Electron Devices 43 (9) (1996) 1602–1609.

[4] H. Schafft, et al., Thermal conductivity measurements of thin-film silicon dioxide,Proc. IEEE 1989 Int. Conf. on Microelectr. Test Str., 2(1) 1989, pp. 121–125.

[5] B. Bhushan, et al., Stress in silicon dioxide films deposited using chemical vapordeposition techniques and the effect of annealing on these stresses, J. Vac. Sci.Technol. B 8 (5) (1990) 1068–1074.

[6] R. Rusanov, et al., Reliability characterization of a soot particle sensor, Design, Test,Integration and Packaging of MEMS/MOEMS (DTIP)2014. 1–6.

[7] R. Rusanov, Reliability Characterization of a Soot Particle Sensor in Terms of Stress-and Electromigration in Thin-film Platinum, Microsyst. Technol (2015) 1–13, http://dx.doi.org/10.1007/s00542-015-2576-6.

[8] A. Lee, et al., Stress induced delamination methods for the study of adhesion of Ptthin films to Si, Acta Mater. 52 (2004) (2004) 2081–2093.

[9] J.R. Black, Electromigration failure modes in aluminum metallization for semicon-ductor devices, Proc. IEEE 57 (9) (1969) 1587–1594.

[10] M. McLean, et al., Surface energy and the secondary recrystallisation of platinumsheet, Acta Metall. 13 (12) (1965) 1291–1297.

[11] P.J. Ding, et al., Effects of the addition of small amounts of Al to copper, J. Appl. Phys.75 (7) (1994) 3627–3631.

[12] S. Guruswamy, et al., Internal oxidation of Ag–In alloys: stress relief and the influ-ence of imposed strain, Oxid. Met. 26 (1–2) (1986) 77–100.

[13] F. Cattaneo, et al., Self-diffusion in Platinum, Philos. Mag. 7 (80) (1962) 1373–1383.[14] R. Spolenak, et al., Effects of alloying elements on electromigration, Microelectron.

Reliab. 38 (6–8) (1998) 1015–1020.

des and heating elements on SiO2 insulation layers and membranes,el.2015.06.106

http://refhub.elsevier.com/S0026-2714(15)30051-2/rf0005http://refhub.elsevier.com/S0026-2714(15)30051-2/rf0005http://refhub.elsevier.com/S0026-2714(15)30051-2/rf0010http://refhub.elsevier.com/S0026-2714(15)30051-2/rf0010http://refhub.elsevier.com/S0026-2714(15)30051-2/rf0015http://refhub.elsevier.com/S0026-2714(15)30051-2/rf0015http://refhub.elsevier.com/S0026-2714(15)30051-2/rf0065http://refhub.elsevier.com/S0026-2714(15)30051-2/rf0065http://refhub.elsevier.com/S0026-2714(15)30051-2/rf0020http://refhub.elsevier.com/S0026-2714(15)30051-2/rf0020http://refhub.elsevier.com/S0026-2714(15)30051-2/rf0020http://refhub.elsevier.com/S0026-2714(15)30051-2/rf0070http://refhub.elsevier.com/S0026-2714(15)30051-2/rf0070http://dx.doi.org/10.1007/s00542-015-2576-6http://refhub.elsevier.com/S0026-2714(15)30051-2/rf0030http://refhub.elsevier.com/S0026-2714(15)30051-2/rf0030http://refhub.elsevier.com/S0026-2714(15)30051-2/rf0035http://refhub.elsevier.com/S0026-2714(15)30051-2/rf0035http://refhub.elsevier.com/S0026-2714(15)30051-2/rf0040http://refhub.elsevier.com/S0026-2714(15)30051-2/rf0040http://refhub.elsevier.com/S0026-2714(15)30051-2/rf0045http://refhub.elsevier.com/S0026-2714(15)30051-2/rf0045http://refhub.elsevier.com/S0026-2714(15)30051-2/rf0050http://refhub.elsevier.com/S0026-2714(15)30051-2/rf0050http://refhub.elsevier.com/S0026-2714(15)30051-2/rf0055http://refhub.elsevier.com/S0026-2714(15)30051-2/rf0060http://refhub.elsevier.com/S0026-2714(15)30051-2/rf0060http://dx.doi.org/10.1016/j.microrel.2015.06.106

Reliability of platinum electrodes and heating elements on SiO2 insulation layers and membranes1. Introduction2. Thermal properties of PECVD SiO23. Reliability of platinum thin films4. ConclusionsAcknowledgmentsReferences