Copper on Silicon Interface

Introduction

This page is concerned with the research that I (Eric Tavenner) conducted under Dr. Robert Mayanovic at the Southwest Missouri State University (SMSU) Material Science graduate program (with the help of two undergraduate students: Dan Busenbarrick and Clay Hainline). I worked on the instrumentation to deposit and characterize copper and tantalum films on silicon. The problem here is that the copper has a tendency to diffuse into the silicon. This diffusion creates a semiconductor integrated with the conductor that you are trying to make, thereby creating a new material with properties that you may not want. Tantalum is the proposed diffusion barrier between the silicon surface and the deposited copper film. This will keep your conductor (copper) a conductor and your semiconductor (silicon) a semiconductor.

Equipment

Spectrometer

Reflection Set-up

A system using Oriel's Merlin lock-in amplifier, MS257 monochromator, silicon detector (UV & Vis) and a cooled lead sulphide detector (IR) was developed to do UV-Vis-Near IR spectrographic characterization of various samples using transmission and reflection measurements. This system has scanned the visible portion of the mercury emission lines with reasonable accuracy, and it has characterized red cut-off filters with logical results. More importantly, it has also reproduced gold standard reflection data (more important than the red filter because it's based off of a reconized standard). It has been used for the characterization of various ion-implanted polymers and for the study of the deuterium line in the hydrogen emission spectra.

Transmission Set-up

Deposition System

The MBE Deposition Chamber
This system was developed to deposit various materials, most notably copper and tantalum, onto substrates. It is designed to accomplish this at ultra-high vacuum (typically < 10^-8 Torr) with an electron beam gun. The reason for using UHV and an e-beam gun is to keep the number of contaminants down to a minimum which should result in a better diffusion barrier (fewer impurities and defects → fewer paths for diffusion).

Pumping Components

The chamber is pumped down by three vacuum pumps: a sorption pump, an ion pump, and a sublimation pump.

Sorption Pump

The pumping action of a sorption pump is achieved by the cooling of a substrate material; this causes the air particles to stick to the substrate (this is similar to water vapour in the air condensing on a can of cold pop). The sorption pump is the roughing pump for the system, that is it takes the vacuum chamber from atmospheric pressures down to a low vacuum (in our lab we generally measure pressures in the units of Torr; atmospheric pressure is 760 Torr, and a low vacuum is around 10^-3 Torr).

Ion Pump

An ion pump works by ionising gas particles, and having those particles impinge into the walls of cathodes. An additional pumping process that occurs in an ion pump is that after the ionised particles hit the cathodes, some of the cathode atoms are discharged from the surface, and these deposit on all exposed surfaces of the pump. This is good because reactive gas particles will chemically attach themselves to the newly deposited cathode atoms. The ion pump is able to bring the chamber to a very high vacuum (10^-8 Torr).

Sublimation Pump

The sublimation pump assists the ion pump, and it works in much the same way as the additional pumping process of the ion pump. That is, the sublimation pump evaporates titanium from a filament that gets deposited on the inside surface of the pump; this titanium is then exposed to the reactive gases inside the chamber, and it pumps out the reactive gases by chemically combining with them. The sublimation pump should be able to bring the chamber to ultra high vacuum levels (10^-10 Torr). Even with these pumps, water vapour and some other gases are difficult to remove from the system. To remedy this problem we bake out the system (while the pumps are running, the chamber is brought to an elevated temperature causing the trapped gases to desorb from the walls), and a cold trap is used (the cold trap works by having liquid nitrogen run through tubes inside the chamber, which causes the gas particles to freeze to the coils).

Pressure Measuring Devices

The pressures inside the chamber are measured by two different instruments: an thermocouple gauge and an ionisation gauge.

Thermocouple Gauge

A thermocouple gauge works by the gauge heating itself up, and then the rate of the temperature change is measured. This relates to a pressure in that the higher the pressure, the more air particles that are in the chamber, the more thermally conductive is the atmosphere inside the chamber, and thus a quicker rate of change of the temperature of the gauge. A thermocouple gauge can measure pressures from atmospheric (760 Torr) to 10^-3 Torr.

Ionisation Gauge

An ionisation gauge works by a filament being heated (like in a light bulb) which then gives off electrons. These electrons are then collected, but if there are any air particles in the way of the electrons' path, the electron are either deflected away from the collector of the gauge or the particle is ionised by the electron. This causes the current produced on the gauge's collector to be different than if the gauge was in its reference pressure, and there is a direct relationship between the pressure and the current. This gauge can read from 10^-4 Torr down to 10^-10 Torr.

Evaporation Source

Tantalum and copper thin films are deposited onto silicon with an electron beam gun. An e-beam gun works by electrons being accelerated towards the material you want melted, and, by the electrons striking the surface (and thereby heating it), the material is evaporated away. This device is mounted on the afore mentioned vacuum chamber that is also equipped with a deposition monitor.

Electron Beam Gun

Results

This equipment proved reliable for the tasks it was developed for. The vacuum chamber was able to produce both copper and tantalum thin films at ultra-high vacuum. It proved difficult to e-beam evaporate copper rods in the horizontal position, so a technique was developed to do just that. The problem is that the copper would melt and then drip onto the chamber walls of the e-beam gun without evaporating. To remedy this, it was found that if thin tungsten wire was wrapped, in a shape similar to a spring, around the copper rod, then the copper was trapped by the tungsten wrappings and prevented from dripping (the tungsten, compared to copper, is such a high temperature melt metal that none of the tungsten was evaporated with the copper). Tantalum, however, did not suffer from the same problems as copper. It was able to evaporate in rod form (horizontally) without any support. Tantalum is such a high temperature melt metal that it did push the limits of the e-beam gun. It took 50 minutes to deposit ~50Å of the metal onto glass. Of course this would not be a problem for this study because only very thin films of tantalum would be used (~50-200Å), so not too much time would be used in producing a sample.

The spectrometer proved to be very reliable (after a great deal of time getting it working and calibrating it). Calibration of the spectrometer was carried out using a mercury lamp, and it proved to be highly accurate (within 5nm). Scans of other lamp sources (e.g. hydrogen, helium, nitrogen and sodium) gave similar results. Next, with the spectrometer in transmission configuration, a sample of red polymer film was measured, and reasonable results were obtained (without a good reference standard, this was about all we could do here). The real clincher was, with the spectrometer in reflection configuration, measuring a sample and comparing it to a known standard. What we did here was to make very thick gold films (~1500Å), measure their reflectance spectra, and compare them to published data for electropolished gold (from the CRC Handbook of Chemistry and Physics). The two graphs were very close to the same shape (within experimental error, and taking into account that our samples were produced differently than the standard). With this, and with the fact that a dark scan showed that our noise level was at least an order of magnitude lower than our signal, we felt that the system was extremely accurate. This was demonstrated when one of our gold reflectance samples showed signs of contamination. The spectrometer was able to tell a difference from the contaminated sample and a non-contaminated sample (in both transmission and reflection spectra).