Those actions respectively create the binary 1s and 0s of digital information. Transistor performance critically depends on how reliably a designated amount of current will flow.
These defects can manifest themselves immediately or over a period of time while the device is operating. But defects become harder to identify as transistor dimensions become almost unimaginably small and switching speeds very high. For some promising semiconductor materials in development — such as silicon carbide SiC instead of silicon Si alone for novel high-energy, high-temperature devices — there has been no simple and straightforward way to characterize defects in detail.
They published their results on October 6th in the Journal of Applied Physics. When a transistor is functioning correctly, a specific electron current flows along the desired path. Holes can also form a current. This research explored electron current, the most common arrangement. If the current encounters a defect, electrons are trapped or displaced, and can then combine with holes to form an electrically neutral area in a process known as recombination.
Each recombination removes an electron from the current. Multiple defects cause current losses that lead to malfunction. The goal is to determine where the defects are, their specific effects, and — ideally — the number of them. Manufacturers are able to produce products that perform to carefully worked out strategies. Some faults will occur due to the product exceeding its "designed life" whilst others will occur prematurely. Designing an electronic product for a particular life span, under conditions that will be very variable e.
However such faults as do occur usually follow a distinct pattern, and careful recording of previous faults can be a good indication of future ones. These failures can affect transistors just as easily as any other component.
When considering an item of faulty equipment, always remember that the reliability of any component is proportional to the power it dissipates. In other words, "If it normally gets hot it normally fails". Such a rule suggests that a failed transistor is more likely to be in the output stages of a circuit than the low voltage, low power stages that precede it. Any circuit which uses either high voltages, high current or both, puts much more stress on semiconductors than low voltage, low current circuits.
Although the devices used in these circuits are designed to withstand such use, they do so less well than those devices having a relatively easy life in low power situations. Main problem areas are power supplies and output stages. When faced with a faulty circuit and very little circuit information, a quick check on semiconductors in these stages can save much work. Of course this list could be extended to include that junctions may become leaky slightly low resistance , though this is rare.
Holes can also form a current. This research explored electron current, the most common arrangement. If the current encounters a defect, electrons are trapped or displaced, and can then combine with holes to form an electrically neutral area in a process known as recombination. Each recombination removes an electron from the current. Multiple defects cause current losses that lead to malfunction. The goal is to determine where the defects are, their specific effects, and—ideally—the number of them.
We then conducted proof-of-principle experiments that confirmed our model. In a classic metal oxide semiconductor design see figure , a metal electrode called the gate is placed atop a thin insulating silicon dioxide layer. Below that interface is the bulk body of the semiconductor.
On one side of the gate is an input terminal, called the source; on the other is an output drain. Scientists investigate the dynamics of current flow by changing the "bias" voltages applied to the gate, source and drain, all of which affect how current moves. In the new work, the NIST and Penn State researchers concentrated on one particular region that is typically only about 1 billionth of a meter thick and a millionth of a meter long: The boundary, or channel, between the thin oxide layer and the bulk semiconductor body.
The detection method we investigated was previously unable to determine how many defects were within this layer. One sensitive method to detect defects in the channel is called electrically detected magnetic resonance EDMR , which is similar in principle to medical MRI. Particles such as protons and electrons have a quantum property called spin, which makes them act like tiny bar magnets with two opposite magnetic poles.
In EDMR, the transistor is irradiated with microwaves at a frequency about four times higher than a microwave oven. Experimenters apply a magnetic field to the device and gradually vary its strength while measuring the output current.
At exactly the right combination of frequency and field strength, electrons at defects "flip"—reverse their poles. This causes some to lose enough energy that they recombine with holes at defects in the channel, reducing the current.
The channel activity can be hard to measure, however, because the high volume of "noise" from recombination in the bulk of the semiconductor. To focus exclusively on activity in the channel, researchers use a technique called bipolar amplification effect BAE , which is achieved by arranging the bias voltages applied to the source, gate and drain in a particular configuration see figure.
We can select just defects that we care about within the channel. The exact mechanism by which BAE operates was not known until the team developed its model. Before the model of BAE, the scheme was used strictly as a resource for applying voltages and controlling currents for EDMR measurements, which is useful for a more qualitative defect identification.
The new model enables BAE as a tool to quantitatively measure the number of defects and to do so with just currents and voltages. The parameter of importance is the interface defect density, which is a number that describes how many defects are within some area of the semiconductor-oxide interface. The model, which the researchers tested in a set of proof-of-concept experiments on metal oxide semiconductor transistors , makes quantitative measurements possible.
This would be an opportunity to further enhance design of the chip circuitry and device performance leading to better performing products. The others either vanished unlikely or were not produced in this wafer. What that looks like at high magnifications as seen by transmission electron microscopy can be seen in the link. The question is: Will the transistors work?
The answer is: It depends. The transistors without any stacking faults will work, but their leakage currents may still be considerably higher than the leakage currents in transistors without any stacking faults in the neighborhood.
0コメント