Performance of Zr and Ti adhesion layers for bonding of platinum metallization to sapphire substrates

https://doi.org/10.1016/S0925-4005(01)00756-0Get rights and content

Abstract

Single crystal sapphire wafers with <1 nm root mean square (RMS) roughness are ideal substrates for chemiresistive sensors that utilize ultra-thin (<50 nm thick) semiconducting metal oxide (SMO) films. Platinum metallization on a highly polished sapphire platform to form electrodes, heater, and a resistive temperature device (RTD) requires the use of a very thin (<20 nm) buffer layer, such as Ti or Zr, to achieve good adhesion at the Pt/sapphire interface. Using AES, secondary ion mass spectroscopy (SIMS), XRD, and wire bond tests before and after annealing treatments, we have found that Zr has superior performance as an adhesion layer compared to Ti. At temperatures of 200–700°C, required for RTD and SMO film stabilization as well as prolonged sensor operation, there is significant migration of Ti through the Pt film, whereas the Zr layer is less mobile. The Pt/Zr/sapphire architecture also minimizes delamination failure of wire bonds to the sensor device.

Introduction

The stability, chemical sensitivity, and selectivity of thin film SMO sensing materials is largely determined by film composition, structure, and purity. These factors are in turn strongly affected by the method of film growth, growth parameters, and substrate material. For chemiresistive sensors, the substrate must be electrically insulating, or there must be an insulating layer between the substrate and the sensing film. Substrate materials reported in the literature have included alumina, sapphire, and oxidized silicon [1], [2]. A potentially important advantage of sapphire is that highly polished surfaces with less than 1 nm root mean square (RMS) roughness may be obtained. These sapphire substrates can be well characterized prior to film growth, have extremely low levels of impurities, and are highly insulating. They also provide an excellent platform for the development of thin film sensors, since a wide range of SMO film structures can be obtained, ranging from amorphous to epitaxial [3].

Electrically interrogating SMO thin films requires an electrode pattern that is in contact with the SMO film. These electrodes must be oxidation resistant and typically include bond pads for connecting wires. The SMO chemiresistive sensors must also operate in the range 200–600°C to enhance gas sensitivity and selectivity. A convenient means of achieving this is to locate a serpentine heater element fabricated from the same oxidation resistant material as the electrode on either the sensor side or the backside of the substrate [1], [2], [4]. To measure temperature, a separate serpentine resistive thermal device (RTD) can also be integrated on the sensor platform. Fig. 1 shows representative examples of electrode, RTD, and heater patterns that we have used on our sapphire platform.

Platinum metal is an excellent material for these functions since it is oxidation resistant and can be patterned by microelectronic fabrication techniques. Alternatively, a physical shadow mask may be used to pattern a device, although linewidths are inferior to those obtained with microfabrication techniques. Achieving strong adhesion between patterned Pt films and highly polished sapphire substrates, and maintaining good adhesion following wire bonding and operation at elevated temperatures, has proven to be a challenge.

Platinum metallization has been used in a variety of microelectronic devices such as thin film RTDs, capacitors, ferroelectric devices, and sensors [5], [6], [7], [8], [9], [10], [11]. However, the substrate material has been primarily silicon or alumina. Pt adhesion to sapphire has been studied in some detail by Schierbaum et al. [2] in the context of a NO2 sensor using a SnO2/Pt/Ti/sapphire sensor structure with front side interdigitated Pt electrodes and backside Pt RTD traces. Following 800°C annealing in N2 to stabilize the Pt RTD, secondary ion mass spectroscopy (SIMS) analysis showed significant migration of Ti within the Pt and the presence of TiO2 on the top Pt surface. While the presence of small amounts of Ti or TiO2 on the Pt may not cause sensor failure, it does introduce the possibility of inadvertent doping or delamination failure due to reduction in the Ti adhesion layer thickness at the Pt/sapphire interface.

Titanium has the tendency to migrate and form oxides and other compounds at interfaces with other substrates (e.g. titanium silicides) following high temperature annealing [2], [5], [6], [7], [8]. Often, migration and compound formation are associated with decreased Pt adhesion to the substrate. A number of other procedures and adhesion promoting layers have been proposed for use with Si/SiO2 or alumina substrates to enhance Pt adhesion including variations in the gas mixture used during sputter deposition of the Pt layer [9], [10], using Ti, Ta, Hf, or Zr adhesion layers [8], [11], and use of various metal oxide adhesion layers including Ni, Ta, Ti, Cr, and Al oxides [12].

The quality and reliability of wire bond connections to Pt bond pads is also dependent on the metal adhesion layer. Bonding in conventional IC processing is concerned primarily with wire bonds whose operating temperatures do not exceed 125°C [13]. After processing temperatures in the range 200–700°C required for RTD and SMO film stabilization, delamination can occur as shown in Fig. 2. In this example, the Zr layer process was not yet optimized and SEM–EDS analysis indicated that wire bond delamination occurred both at the Zr/sapphire (dark areas) and Pt/Zr (light areas) interfaces. Similar results were obtained for delaminated Pt/Ti devices.

We chose two routes to explore improved adhesion layers and Pt bondability: (1) the use of Zr metal as an adhesion promoter on sapphire, since studies using Zr to promote adhesion of Pt to SiO2 on Si [8], [7] have shown Zr to be far less mobile than Ti, resulting in improved Pt adhesion; (2) the use of low growth rate electron beam deposition to improve the texture and density of the Pt layer and reduce Ti or Zr migration. Two types of samples were investigated, unpatterned sapphire wafers prepared with blanket depositions of either 20 nm Ti or Zr, followed by 300 nm of Pt, and fully functional sensor devices prepared by metallization, photolithographic patterning, and dicing of an entire sapphire wafer. The unpatterned wafers were cut into 3 mm wide strips which were used for bonding tests and diffusion studies. The unpatterned samples were used to test the viability of Ti and Zr as adhesion layers on sapphire without the possibility of interference from the microlithographic patterning process. Patterned devices were used to evaluate the effect of the photolithography process on adhesion as well as to provide prototype samples for testing sensor lifetime during accelerated aging tests and actual gas testing.

Section snippets

Device fabrication

About 2 in. diameter, single-sided and double-sided polished, r-cut single crystal sapphire wafers were used for the blanket metallization tests. About 3 in. diameter, double-sided polished, r-cut single-crystal sapphire wafers were used for the patterned devices. The roughness of the polished wafer surfaces was <1 nm RMS over a 500 μm scan length as measured by a KLA/Tencor AlphaStep 500 profilometer with a stylus force of 54.8 mg and a stylus tip radius of 5 μm. As-received wafers were cleaned by

Bond strengths of Pt/Zr and Pt/Ti metallization on sapphire

The results of pull tests on 0.10 mm Pt wire loops bonded to Pt pads on blanket metallized samples as function of annealing treatment are presented in Fig. 3. The wires were attached prior to annealing and the samples were then subjected to air anneals at 500, 600, or 700°C. A failure load of 150 g indicates that the bond remained intact up to the 150 g limit of the load cell. For adhesion layers deposited at room temperature, the Ti samples demonstrated a significant decrease in bond strength

Conclusions

The elevated temperatures at which metal oxide chemiresistive gas sensors operate pose challenging materials compatibility issues to achieve integrity of metallization layers. Our investigations have shown that of the two metal adhesion layer materials investigated (Ti and Zr) at the Pt/sapphire interface, the Zr shows superior performance after annealing treatments. For unpatterned samples with wire loops attached prior to annealing, bond weakening begins sooner for samples with Ti adhesion

Acknowledgements

This work has been sponsored by the Department of the Navy, Naval Surface Warfare Center, Dahlgren Division, Grant No. N00178-99-1-9002.

G. Bernhardt received his BS from the US Air Force Academy in 1979, his MS from M.I.T. in 1983, and his PhD from the University of Maine in 1994. He currently is a research scientist at the Laboratory for Surface Science and Technology. His research interests include fabrication and characterization of thin film sensors, thin film synthesis, and materials characterization.

References (14)

There are more references available in the full text version of this article.

Cited by (0)

G. Bernhardt received his BS from the US Air Force Academy in 1979, his MS from M.I.T. in 1983, and his PhD from the University of Maine in 1994. He currently is a research scientist at the Laboratory for Surface Science and Technology. His research interests include fabrication and characterization of thin film sensors, thin film synthesis, and materials characterization.

C. Silvestre obtained a PhD in Electrical Engineering from Princeton University in 1991. He did postdoctoral work in the Center for Advanced Electronic Materials Processing, North Carolina State University, studying CVD deposition of CMOS gate structures, and postdoctoral work at the Naval Research Laboratory, Washington, DC, studying mbe deposition of silicon–germanium alloys. He is currently a research engineer in the Laboratory for Surface Science and Technology, University of Maine.

N. LeCursi received his BS from Kent State University in 1988. He has held several technical positions utilizing his skills in design, development, and construction of industrial and scientific instrumentation, most recently (1995–2000) as a research engineer in the Laboratory for Surface Science and Technology, University of Maine. He is currently pursuing a career in prosthetics/orthotics.

S.C. Moulzolf graduated with a BA in physics in 1993 from St. John’s University, MN. He received a PhD in Physics in 1999 with a specialization in surface physics from the University of Maine. He conducted his Thesis research at the Laboratory for Surface Science and technology at the University of Maine, where he is currently employed as a research scientist. His current research involves electronic and structural characterization of semiconducting metal oxides for chemiresitive sensor applications.

D.J. Frankel is a Senior Research Scientist at the Laboratory for Surface Science and Technology, University of Maine. He received his BS in Engineering Physics from Cornell University and his MS and PhD in Applied Physics from Stanford University. After receiving his PhD he carried out research in semiconductor surface physics at Montana State University. At the University of Maine he is involved in the development of sensor technology, thin films, and surface analysis instrumentation.

R.J. Lad received his PhD in materials science from Cornell University in 1986. He is currently a Professor of Physics and the Director of the Laboratory for Surface Science and Technology at the University of Maine. His research efforts have focused on thin film synthesis, materials characterization and sensor technology.

View full text