Strain Visualisation Workstation Embodies Advances In Image Correlation
Metrology systems for out-of-plane profiling and deformation measurement abound, finding wide application in all branches of engineering and science. The measurement of in-plane deformation, however, is more challenging, especially in the regime in which the deformation itself is substantially less than the wavelength of visible light. This is the regime of the microsystem, of which the most ubiquitous example is the IC package. Moir interferometry is often used for out-of-plane warpage measurements, and is currently the technology most commonly used by major IC manufacturers for measuring in-plane deformation and strain. However, as an in-plane-strain measurement technology, Moir suffers from a number of limitations, not least of which is the requirement to attach a grating to the sample by means of an adhesive. The grating acts as an intermediary between the metrology head and the sample, introducing an uncertainty as to the extent to which all strain within the sample asserts itself. It is the strain of the grating that is actually measured. The grating/adhesive system, itself a perturbation of the sample material structure, also introduces a temperature ceiling of typically 120¡C on the measurement range. For certain types of studies - especially in the detection of hairline cracks after solder shock stressing - the grating can be considered as an interference with the measurement. The ability to directly measure the deformation of the package, rather than that of an applied phase grating, is a major advantage of optical digital image correlation technology. Using three-dimensional (3D) digital image correlation technology, embodied in the new system developed by Optical Metrology Innovations, in-plane deformation and strain characteristics can be measured directly on the bare cross-section or exterior surface of an IC package, without the need for the adhesion of a grating or other auxiliary device.
Principle of operation
The basic principle of the technology is shown in Figure 1. It works by recording two or more sets of optical images of a sample under different conditions of an applied thermal stress to the sample, and correlating the images in focus rendered from each set of images.
In its simplest form, an image of the specimen under test is taken in two different states of stress, in this example cold and hot. A small region of the image (called the sub-image) is observed in the cold state and its match is then found in the hot image also. Using numerically intensive algorithms the size of the displacement between the two is then calculated to accuracies equal to small fractions of the size of each pixel of the CCD camera used to capture the images. One motion vector per sub-image is calculated. This surface displacement is found by calculating the position of maximum cross-correlation between the cold and hot sub-images. Thus an array of sub-image deformation vectors is calculated, showing the local relative deformation of different parts of the workpiece within the field of view of the optical system, under the thermal stress.
The system has two scales, and two associated resolutions. The image scale is depicted on the captured video microscope images. Its resolution is limited by the optical system and the camera characteristics, with diffraction as an ultimate limit. The deformation vector scale is that of the deformation vectors superimposed on the reference image, and has sub-pixel resolution. To display the measured sub-micron deformation, overlay vectors are plotted on a scale of typically 5 to 20 times that of the image.
The scale and resolution depends on a number of factors, but primarily on the magnification of the optics used. Table 1 illustrates the achievable resolution obtainable with the different options on the standard system. A system with a large camera field of view is also available. The X5 objective is most commonly used with the system, providing the optimum balance of field of view and resolution.
The technology is however more complex than the simple correlation of a pair of images and uses a full 3D image stack to achieve the best image in focus at each measurement point in the thermal stress cycle. If required, the image in focus is composed of sub-images from more than one layer of the 3D stack. This technology also allows the extraction of 3D information using a single camera, and obviates the need for stereoscopic or multiple camera/off-axis imaging systems.
Figure 2 shows the manner in which a stack of images is made through the region of focus of the optical system, and the way in which the image in focus is selected or composited. The focus function is the key to the compositing process and can be evaluated for every part of an image.
Investigation of a thin ball grid array
The system's deformation and strain analyses can be used to build up a picture of the thermo-mechanical behaviour about a die attach joint, near the die edge and within a ball grid array package.
Two thin ball grid array (TBGA) packages, one of which was subjected to a severe thermal/humidity solder shock, were cross sectioned in order to inspect in-plane deformation on a cross section through the package, and micro-textured using 1200 grade abrasive paper. The field of view was set over the die attach in the middle of the package using the software navigator. The field of view straddled the package top to bottom, excluding the solder balls. Figure 3 shows a schematic of the field of view region.
A single thermal stressing cycle, performed over an approximate 30 minute period, was used to reveal the deformation characteristic of the die attach joint. The sample was heated, in situ, by a Peltier (thermoelectric) heat pump stage from 25¡C to 95¡C with a 5 minute soak at each temperature before imaging.
Figures 4 shows an image of the microsection of a TBGA sample, at the location indicated. The in plane strain maps (Exx and shear) provide an easy method to visualise the occurrence of a delamination or cracks. Figures 5 and 6 show the strain behaviour of a virgin TBGA sample before solder shock stressing. At this stage, the system has only been exposed to the moderate thermal cycle of the system.
Figure 5 shows the Exx strain component derived from a deformation vector map of this region of the sample, after the removal of rigid body motion, and Figure 6 shows the shear strain component. While no significant local concentrations of strain are evident in the Exx map, the shear map gives cause for concern below the silicon die, where even this moderate thermal stress gives rise to shearing across the die attach adhesive and in the top of the printed circuit board.
The results obtained from an identical TBGA sample after solder shock stressing show why the moderate shear strain observed in the virgin sample is a cause for concern. Figure 7 shows the deformation vector map of this sample, with the reference point for zero motion located in the silicon die. The deformation vectors are plotted at 20 times the scale of the micrograph. A crack is evident running from the top of the image downwards (marked out by the red box), located near the interface between the die attach adhesive and the printed circuit board. The motion pattern shows a clear separation of the material on heating, and this "zip-fastener" separation pattern is a commonly observed characteristic of propagating cracks in this technology.
Figures 8 and 9 show the Exx and shear strain maps for the solder shocked TBGA sample, plotted on the same scales as Figures 5 and 6. The crack is clearly observed as a sharp deformation change across a short distance. Comparison of Figures 6 and 9 shows that the probability of failure occurring first at this region was predictable from the shear map of the virgin sample. The high shear strain location in the virgin package was revealed in the absence of an actual defect, and no indication of this was observable using acoustic microscopy.
Tool for reliable engineering
One of the values of this technology is to provide a metrology tool capable of measuring motion differences across thin layers of material such as die attach adhesive. This - coupled with its ability to reveal incipient defects invisible to other techniques - makes the optical digital image correlation system a powerful tool to assist in the reliable engineering of the next generation of IC packages. X-ray analysis, scanning acoustic microscopy and dye penetrant testing have relatively high resolution limits and are only useful in detecting defects which have already occurred, necessitating the performance of a costly thermal stressing cycle to induce defects in a fraction of a set of prototypes. Additionally, it is recognised by the packaging industry that defect detection by acoustic microscope requires a secondary independent inspection. This has been explicitly recognised in the recent JEDEC standard J-STD-020-B for substrate based IC packages, which prescribes a second cross-sectional analysis for packages shown to have failed acoustic microscopy inspection.
Over the last decade, the silicon industry has been shifting from post-mortem failure analysis towards process control. The use of technology which can detect incipient failures in virgin packages, without the use of stressing cycle to induce defects, opens the possibility of a similar movement for the IC packaging industry. A package engineering team can be in possession of such analyses within hours. The information gleaned can be used to review the design, and take decisions on whether engineering changes should be implemented before proceeding with a full reliability cycle. Improved engineering prior to the first reliability cycle has the potential to reduce the number of these extended stressing and post-stress analysis cycles needed to qualify a product. The value of pre-reliability metrology of the type described in this article is ultimately realised in a reduction of the time-to-market of a more reliable product.
Liam Kehoe, Vincent Gunebaut, Pat Lynch, Maura O'Sullivan, Pat Kelly, Optical Metrology Innovations.