Feature Article

Technology that is widely used in semiconductor manufacturing has been successfully applied to the accurate, nondestructive, in-line measurement of coatings on medical devices.

Figure 1: Schematic of a white-light interferometer.

Firstly, techniques such as white-light interferometry and confocal microscopy independently image the top surface and buried interface of the coating, inferring thickness from the amount of z-axis translation needed to go from one to the other. Figure 1 illustrates a typical white-light interferometer: light that is reflected from either the surface or the buried interface is made to interfere with light from the same incident ray that has been reflected from a half-silvered beam splitter and a small mirror placed just beneath the main objective lens. The beam splitter just below this mirror forms a reference plane. As the distance between the sample surface and reference plane varies so the spectral composition of the light that returns to the detector changes because of the interference effects.

Figure 2: Schematic of a confocal microscope.

Figure 2 shows a confocal microscope arrangement. The principle here is that when light falling upon the sample surface or buried interface is in focus, then the reflected light is also brought to a focus at the pinhole beneath the detector and, thus, is able to pass through. When the light is out of focus at the surface, it is also out of focus at the pinhole and so is heavily attenuated. By scanning the focus across the surface and moving it down through the coating, a 3-D map of the coating surface and buried substrate can be built up.

Figure 3: General behaviour of light on encountering a coated surface.

An alternative approach is to intentionally combine light from the two interfaces and look for interference effects in the reflected light. As illustrated in Figure 3, overall surface reflectance is determined by the wavelength of incident light, the coating thickness, and the angle that the light travels relative to the surface. It is also affected by the refractive indices of both the coating and the substrate, and the polarisation of the light. Analysis of reflected light is usually carried out by keeping most of these factors constant and changing one or, at most, two of them in a controlled manner. In spectrophotometry and spectroscopic ellipsometry, the surface is illuminated by white light at a constant angle-of-incidence and the reflectance is measured as a function of wavelength. In the former case (Figure 4), normal-incidence light is used and the intensity of the reflected light is analysed. In the latter (Figure 5), a high angle-of-incidence is used and both intensity and phase are analysed. Even so, for medical device coatings these techniques face two significant problems. The first is that, as they depend on the assumption of a constant angle-of-incidence, it is difficult to align them on samples with complex shapes in such a way that the orientation and hence the angle-of-incidence is reproducible. This is a major source of error.

Figure 4: Spectrophotometry measures reflectance as a function of wavelength at normal incidence.

Secondly, although the techniques offer some capability of measuring the coating refractive index, this is limited by the phenomenon of optical dispersion, namely the variation of refractive index with wavelength. Hence, they must measure not just one value, but as many values as there are wavelengths. This renders it impossible to make a deterministic measurement, as the number of parameters to be measured always exceeds the number of independent data points that can be collected.

Figure 5: Spectroscopic ellipsometry measures reflectance as a function of wavelength and polarisation at a fixed angle of incidence.

As illustrated in Figure 6, BPR overcomes this limitation by using a high-magnification lens to bring a collimated laser beam to a sharp focus. At the focal point, which is typically less than 1 µm across, light falls on the sample with the whole range of different angles-of-incidence through which the lens bends the light to achieve focus. After reflection, the lens recollimates the reflected light and there is a one-to-one correspondence between the physical location of a ray of light within the recollimated beam and the angle at which that ray was reflected from the surface. It is therefore possible to measure reflectance as a function of angle of incidence simultaneously for a wide range of angles (typically, for a ~100× lens, a range of 0–60˚) with a very short data acquisition time using an apparatus with no moving parts.

Figure 6: Schematic representation of a beam profile reflectometry system.

Because all these measurements take place at a single wavelength, determined by the laser source employed, there is no need to take account of optical dispersion. This means that for each material in the film stack there is only one value of refractive index to find (or, at most, two in the case of a birefringent film), and yet there are hundreds of independent data points in the raw data. This data richness enables direct, deterministic measurements of refractive index to be made, in contrast with spectral techniques, which must rely on models and assumptions to account for the effects of dispersion.

Figure 7: Results obtained from an evaluation on curved samples performed at the UK National Physical Laboratory.

Where the surface is not only misaligned but also curved, there are further effects that need to be accounted for and to which, again, BPR is uniquely sensitive. Further details of the analysis involved are given elsewhere³, but Figure 7 gives an indication of the resulting capability of BPR by showing results obtained over a range of samples with different coating thicknesses and curvatures, up to an extreme case of 15-µm films deposited on 50-µm-diam wires. Extremely high (better than 99%) correlation was obtained relative to destructive measurements carried out by the UK’s National Physical Laboratory⁴.