A novel way to detect latent flaws
Semiconductors are widely used in electronic devices, often along with liquid crystal display panels based on glass substrates. For both semiconductors and glass substrates, fine polishing techniques are a very important part of the manufacturing process.1, 2 However, the mechanical friction used for fine polishing may induce micro/nanoscale cracks (latent flaws) on the surface of the polished products.1–3 If latent flaws are not detected before the washing process, they may become large dimples or visible cracks because of the erosive effect of the cleaning liquid.
Glass substrates and the interlayer dielectric film of semiconductors are conventionally inspected by a light scattering method, which uses incident laser light and photodetectors, such as a photomultiplier tube, to detect light scattering from tiny particles (for instance, dust particles or pieces of foreign matter) and latent flaws with high sensitivity. However, this method detects both tiny particles and latent flaws on the surface as light scattering intensities, and so cannot distinguish between latent flaws and tiny particles. Additionally, ‘closed state’ latent flaws hide under the surface of products, and are thus difficult to detect by this method.
We have proposed and evaluated a novel technique to detect latent flaws, which we name SILSM (stress-induced light scattering method). SILSM enables us to classify and separately detect latent flaws and particles nondestructively for both glass substrates and the dielectric interlayer of semiconductor surfaces.4, 5
SILSM detects the change in refractive index at the crack tips of latent flaws as a change in light scattering intensity before and after stress is applied (the photoelastic effect). It detects latent flaws only, and not particles, on which stress does not act and so the particles' light scattering intensity does not change with stress and they are undetected.
To test SILSM, we indented a glass substrate at a single point using a Vickers hardness tester. We fine polished away several micrometers of the sample surface so that only one latent flaw remained. Figure 1 shows a latent flaw, a residual fracture layer, and particles in an optical microscopic image of the indenter's contact point on the sample surface. A light scattering image of a latent flaw shows that the flaw gives rise to the highest light scattering intensity: see Figure 2. Many particles are also observed in this image, and it is apparent that it is difficult to classify the latent flaw and particles by the light scattering method alone.
Figure 3 shows the SILSM experimental setup, which differs from conventional light scattering methods by including an actuator to bend the sample. The laser beam point was shaped to the beam plane and we irradiated the observation area uniformly with the beam plane. A cooled CCD camera captured light scattering intensities from the sample, which was deformed in 5μm increments along the z-axis by indenters. We captured light scattering images for each bending load.
Figure 4 shows the relationship between increased stress and changes in the light scattering intensities of a latent flaw and a particle. We integrated the intensities with 3×3 pixels around the maximum point of the light scattering intensities of scatterers. As can be seen, the latent flaw's light scattering intensity changes significantly with surface stress, but the particle's light scattering intensity hardly changes, and consequently SILSM detects only latent flaws.
In summary, we have shown that SILSM is an effective technique for nondestructively detecting latent flaws, and not particles, on a glass substrate surface5 and the interlayer dielectric film of semiconductors.4 We are now applying SILSM to other polished products, such as the surfaces of silicon bare wafers, the edge face of optical fibers, and silicon carbide, which are manufactured in a similar way.
