For semiconductors, the mantra is “smaller, faster, cheaper.” Continuing shrinkage of transistors has allowed industry to follow the axiom known as Moore’s law for decades. Eventually, transistors will be so small that the number of atoms present will make a difference. That day is looming, and so are some associated problems, according to the results of researchers from Imago Scientific Instruments Corp. of Madison, Wis., and from IBM Corp. in Hopewell Junction, N.Y.Researchers examined defects caused by doping semiconductor materials using laser-assisted atom probe tomography. They cooled a needle-shaped specimen with an apex radius of 50 nm in a vacuum and subjected it to high voltage. They hit the tip, which is under a high electric field, with a laser pulse, causing it to free a single ion. Released ions arrived at a detector, yielding location and depth information as well as mass resolution via time-of-flight data. Images courtesy of Imago Scientific Instruments Corp.Using laser-assisted atom probe tomography, the team looked at the controlled placement of dopants within semiconductors. They found defects that affect dopant distribution, which, according to Imago CEO Thomas F. Kelly, could be a problem someday. “The implications of nonuniform dopant distribution are underperformance or failure of devices.”Dopants are key to making semiconductors work. Replacing silicon atoms with arsenic, for example, makes what once was a resistor become an n-type conductor. Manufacturers often use ion implantation to introduce dopants because the technique allows them to place the impurities precisely.The laser-assisted atom probe tomography technique shows that arsenic atoms implanted into silicon become trapped in dislocation loops, even after a high-temperature annealing step. This situation could affect the function and uniformity of next-generation semiconductor devices. Silicon atoms are depicted as gray dots, with only 0.5 percent shown for clarity. Oxygen atoms are shown as light blue dots and arsenic dopant atoms as purple dots.However, implantation is not a gentle process because the ions knock around the silicon atoms. The resulting damage is healed by annealing the silicon at an elevated temperature. Defects, however, can remain.The researchers followed how such defects evolve using laser-assisted atom probe tomography. In this technique, a needle-shaped specimen is cryogenically cooled in a vacuum and subjected to high voltage. The result is a high electric field at the tip. When that tip is hit with a laser pulse, it temporarily heats up and frees a single ion. Under the influence of electric fields, the ion arrives at a detector, yielding X and Y location. The sequence of ions provides depth information. Because the laser is pulsed, the time of flight enables mass resolution. The material can be mapped with a spatial resolution of 0.4 nm.Using an instrument made by Imago, the investigators tracked defect evolution following arsenic implantation into silicon. They did this for samples that had been annealed for 30 min at 600 °C and for other samples that had undergone the same process plus a further 30 s at 1000 °C.They found that the interstitial arsenic atoms could interact with crystal dislocations. Such interstitials distorted the lattice slightly, causing a strain field. Near a dislocation, that field relaxed, pinning the dislocation and trapping the interstitials. Because they are energetically favorable, such defects, known as Cottrell atmospheres, did not disappear despite the annealing step.Although not a problem for today’s 65-nm generation of semiconductors, this could be an issue for the 22-nm technology that should be starting mass production within 10 years. For such small transistors, a Cottrell atmosphere could alter the electrical performance of a device or at least make it unlike a device without such a defect.The semiconductor industry has confronted problems before, however, and has always managed to find a solution. “It is to be determined how this can be overcome, but being able to see it makes fixing it easier,” Kelly said.Continuing research involves atom probing of an actual device. That will require sample processing so that a transistor is in the specimen’s tip.Science, Sept. 7, 2007, pp. 1370-1374.