High-definition X-ray Inspection
Understanding the uses and limitations
BY MIKE COOKE
Non-destructive failure analysis of semiconductor devices is an important capability offered by X-ray inspection tools, which allow a look inside defective devices without reopening or dismantling components. High-resolution X-ray inspection can be used to examine wire bonds, wafer bumps, wire sweep, chip scale packaging (CSP), ball grid arrays (BGAs), multi-chip modules and micro leadframes (MLFs). Newer applications include searching for voids in solder balls, and whiskering and “champagne” voiding in lead-free, tin-based solders.
High-resolution X-ray inspection is achieved through a combination of capabilities provided by magnification, X-ray tube, and detection systems. Although magnification is a main process goal, a system’s effectiveness is conditioned by the source and detection subsystems.
X-rays are absorbed at different levels by different materials (Figure 1). The image at distance b from the source is constructed from the shadows cast by the object at a. The magnification is given by m=b/a. In X-ray inspection, this is often referred to as the “geometric magnification” to distinguish it from the “magnification” carried out by other parts of the system, such as in the detection system.
Figure 1. Geometric magnification.
In terms of resolution, X-ray wavelengths are negligible - a 50-keV photon has a wavelength of 0.025 nm; a subatomic distance. The main factor reducing clarity is the size of the source. Another calculation shows that the image of a point in the object at x is smeared by (m-1)f, where f is the diameter of the source. This smearing is referred to as the “geometric unsharpness” or “Ug”, and represents a distance of f(1-1/m) in the object. For large magnifications, smearing in the object is of the order of the source size.
A typical X-ray inspection system for electronics has the object placed as close as possible to the source, with a=0.25 mm being a practical minimum. The detector distance is of the order b=500 mm, giving a magnification m=2000. This results in a smearing of 1999f in the image, representing a distance of 0.9995f in the object, a difference from f of only 0.05%.
The lower X-ray intensity at the detector due to the inverse square law is one penalty for large magnifications. An image where the detector is moved to give a tilted view is another factor to be considered. These are useful in giving depth information about defects and failures. Since the effective distance from the source to the sample is increased, the magnification is reduced.
The radiation sources used in inspection accelerate electrons onto a target to produce the X-rays. A point-like source is achieved through focusing the electrons in one or two stages, with apertures, or “skimmers,” further reducing the spot size.
Electrons lose energy as they hit the target, mainly producing heat (Figure 2). However, some electrons produce X-rays; 90% of which result from “bremsstrahlung,” a German word meaning “braking radiation.” The resulting X-ray spectrum is broad and continuous with a couple of sharp peaks superimposed from “characteristic” X-rays of the target material’s spectrum.
Figure 2. Creation of X-rays from an electron beam hitting a target.
The maximum photon energy possible in bremsstrahlung is equal to the energy of the incident electron. The broad spectrum peaks at around two-thirds of this energy, enabling the X-ray energy content to be altered, and allowing the contrast of an image to be optimized. Various materials absorb X-rays differently, but as the energy increases, all materials absorb less radiation. Using X-rays for IC package inspection involves reducing the energy so the feature of interest (e.g., a wire bond) is visible without extraneous materials (e.g., the plastic packaging). An accelerating voltage of 40 kV is typical for die attach inspection.
“Feature recognition,” “minimum feature recognition,” “detail recognition,” or “detail detectability” are all terms used among industry experts to characterize the whole image chain. The measurement of the focal-spot size in the X-ray tube is the common connection of these terms.
A number of X-ray inspection tool companies have managed to reduce the focal spot to what they describe as the “nanofocus” level. In this context, “nano” is defined by experts as a focal-spot size of under 1 µm.
Focal-spot size is at the root of most manufacturers’ determination of system performance. Since X-rays are not visible, this usually involves the imaging of some test object.
One test sample and methodology from the Japan Inspection Instruments Manufacturers Association (JIMA RT RC-01) consists of an easy-to-handle, 2 × 2-cm frame containing a periodic structure of bars separated by 1 µ. The sample image is recorded on high-definition film for contrast resolution. The developed film is examined with a densitometer.
Another approach is the European EN 12543-5:1999 standard. This involves imaging a 1-mm wire on a high-resolution detector. Both methodologies have been found to give comparable results, although the JIMA technique yields the parameter faster.*
These techniques¹ yield results under the best possible conditions, using the best detection techniques and unlimited time. Paradoxically, when manufacturers come to quote values for feature recognition, these are generally smaller than the focal-spot size.
Experts in the industry estimate minimum feature-recognition values at roughly half the focal-spot size, or smaller. Detail recognition is expected to be better than theoretical “resolutions” around half the focal-spot size.
One company** uses the term “sub-micron” when describing a tube with a spot size of less than 1 µ and has adopted a new method of determining this, which can be used for both optical and X-ray inspection resolution tests. Since most open-tube systems employ roughly the same source to detector distances and the same types of imaging systems, the relationship between focal-spot and feature recognition should apply to all such systems equally. Thus, the focal-spot size can be estimated from the feature recognition in the same way that feature recognition can be estimated from the focal-spot size.
The test piece consists of a series of tapering lines in a starburst pattern formed from 1-µ-thick gold on a silicon nitride substrate. The test will qualify sources with focal-spot sizes from 5 down to 0.5 µ, giving a quick estimate of the system’s resolution and contrast sensitivity; visually for users and mathematically for manufacturers.
Pushing an X-ray tube to its limits does not usually provide the most useful images in a production environment. To get the smallest focal-spot, the electron current, and thus the X-ray dose must be reduced, increasing the amount of time needed to acquire decent images. A high current in the X-ray tube de-focuses the beam due to the mutual repulsion of the negative charges of the electrons - the “space-charge effect” - which becomes worse at lower-beam accelerating voltages, making it more difficult to reduce the focal spot for lower-energy X-rays.
Heat transfer from the electron beam increases as the spot size shrinks. Tungsten is the most common target because it has a high melting point. Finding alternative materials that could allow even smaller spots is strictly limited. A thin target is used for the smallest spots, but produces fewer X-rays and needs more frequent replacement because of electron-beam damage.
Figure 3. Electron microscope image of a test piece.
The ability to form X-ray images depends on the different radiation absorptions of materials. Elements of low atomic weight become invisible to X-rays because of the lower electron density. Therefore, by varying the X-ray energy, different materials can be distinguished. The detector casing is commonly aluminum, which filters out lower energy X-rays used to distinguish materials composed of lighter elements. Silicon is only slightly heavier than aluminum and is almost impossible to see in X-ray images using such detectors. Fortunately, it is the heavier materials - such as gold, lead, and tin - that are of interest in package inspection.
*According to Viscom.
- Further details can be found in European Semiconductor, July 2005, p.27.
MIKE COOKE, Ph.D., is a UK-based technology journalist; for more information, please contact Nick Hadland, president, X-Tek Group Inc., 100 Tyngsboro Business Park, Tyngsboro, MA 01879; 978/649-6333; E-mail: firstname.lastname@example.org.