BY DANIEL N. DONAHOE AND MICHAEL PECHT
As IC progress drives more functionality onto each IC die, fewer ICs on each PCB are required. At the same time, the trend toward lower IC voltage and to higher operating frequencies requires more passives to maintain signal integrity. For example, approximately 90% of electronics on a cell phone are passive. Of these, 60% are capacitors – most are multilayer ceramic capacitors (MLCC).
Capacitance has been improved by reducing the dielectric thickness between electrodes, and also increasing the number of electrodes. At the same time, smaller capacitors have been introduced. The smallest capacitor size in volume production has an area of only 0.6 mm × 0.4 mm × 0.2 mm, or approximately the diameter of a human hair.
 Figure 1. End termination. |
MLCC electrodes and terminations used to be composed of a palladium alloy and a lower-cost silver to make the melting temperature compatible with the sintering temperature of barium titanate (Figures 1 and 2). Beginning in the late 1990s, the price of palladium rocketed from $125 to almost $1,100 per troy oz. The capacitor industry responded by making electrodes with nickel and copper terminations, a design referred to as base metal electrode (BME). Today, most MLCCs are BME parts.
 Figure 2. BME MLCC cross section. |
Barium titanate (BaTiO3) is used in capacitors because it has a high dielectric constant, based on its atomic structure. Barium titanate, at room temperature, has a tetragonal (cuboid) shape, with one axis taking it approximately 1% from being cubic. The titanium atom in the center position of barium titanate cuboid is often described as a “rattling titanium” atom, because it can be in one of two positions along the unit cell’s longer direction. Within each crystalline grain of barium titanate there are domains that are separated by nanometer-scale transitions called “walls.” Within each domain, titanium atoms are positioned in one head-to-tail direction. In the adjacent domain, titanium atoms are positioned in the opposite direction. Therefore, each grain can be thought of as consisting of unit cells with their internal titanium atoms arranged in ordered up and down directions. The ordering can be envisioned as vehicles on strips of highways laid next to each other.
Application of a voltage to the MLCC generates an electric field between electrodes that forces individual titanium atoms to switch positions to line up with the field, creating a polarization. Many unit cell polarization vectors combine to generate what we measure as net capacitance.
It was already known that barium titanate capacitors lose capacitance over time, caused by the changes in mechanical stresses in barium titanate after firing. This “aging” effect involves atomic adjustment of stresses within the crystalline grains and tends to be gradual. Designers build in capacitance margin over product life to allow for this type of aging. When capacitors are exposed to moisture, however, they can exhibit another type of aging called “oxide vacancy.”
Oxide vacancy was first discovered when both precious metal and BME EIA 0805 capacitors were subjected to autoclave (120°C/100% RH) testing. It was determined that the precious metal capacitors aged according to the well-known aging mechanism (< 3% from their starting values), but the BME capacitors degraded to below the -30% criterion at 500 hrs. of exposure. Attempts to restore the capacitance after the autoclave exposure, using a standard industry method called “de-aging,” produced different results for the precious metal and the BME capacitors. The PME capacitors returned to their initial values, but the BME capacitors did not recover. This is because the humidity degradation mechanism is different than the mechanism for known aging.
The reasons for this new oxide vacancy failure mechanism are complex, and there two theories were hypothesized. The first hypothesis is that there could be oxidation or corrosion of the nickel plates. However, ion beam milling and electron microscopy of the electrode-to-dielectric-to-termination interface, electron backscatter diffraction (EBSD) of the polycrystalline grain structure of the capacitors, and dye penetrant found no possible interconnected path for moisture to flow into the capacitor body from the capacitor surface. Capacitors were also monitored for weight gain after various moisture exposures using balances and thermogravimetric analysis (TGA) with argon pure gas. No weight change was detected by either method, and it was concluded that moisture could not be entering the capacitor bodies. Finally, BME capacitors were subjected to long-term autoclave and then internally assessed using x-ray photoelectron spectroscopy. No nickel oxide was found in the body of the capacitors. In other words, the decrease in net capacitance was not caused by plate oxidation or increased plate spacing from an oxidation process.
The second hypothesis is that the loss of capacitance was caused by oxygen vacancies. When a barium titanate unit cell loses an oxygen atom, we call that an oxygen vacancy. If this oxygen vacancy is a random defect, there is no measurable change to the net capacitance. If many barium titanate unit cells lose an oxygen atom, however, there is a net reduction in capacitance. The capacitance loss can be significant if multiple cells affected by oxygen atom loss are lined up.
For oxygen ions to move within the capacitor, they must cross crystalline grain boundaries in the barium titanate. Since the size of the dielectric grains does not change as the industry shrinks capacitor size, and shrinking the size implies thinner barium layers between electrodes, there are fewer grains that ions must cross. With fewer grain boundaries to cross, there is less resistance to ion flow. In addition, as the MLCC size shrinks, the ratio of the surface area (where the barium atoms are removed from) to volume grows – bringing the surface charge and oxygen vacancies into more intimate interaction. In these two approaches, reducing the size makes the newer BME capacitors more vulnerable.
Experiments showed that oxide vacancy aging followed an exponential rule, just as would be expected in a diffusion process, and the movement of oxygen ions is a diffusion process. This process can occur in the autoclave testing and, perhaps, in the field because of long-term humidity exposure. The result is degradation of capacitance. Capacitor degradation caused by Donahoe aging is most problematic in high humidity environments, with high-value capacitors (thinner barium titanate layers). Unfortunately, standard humidity life testing, such as JESD-22, THB, HAST, or autoclave, likely will not uncover this problem. Poor reliability resulting from degradation of BME MLCC capacitance may catch manufacturers and consumers by surprise.
DANIEL N. DONAHOE, faculty research scientist, and MICHAEL PECHT, Ph.D., chair professor and director, may be contacted at CALCE Electronic Products & Systems Center, University of Maryland, College Park, MD 20742; (301) 405-5323; e-mail: [email protected], [email protected].