AFMs pinpoint surface defects

To map the topography of surface features, an atomic-force microscope (AFM) uses physical contact between a fine mechanical probe tip and a sample. This microscopy technique can reveal small defects and failures, and when used in an ultra-high vacuum (UHV), it can locate lattices of atoms, and in some cases, individual atoms.

Every microscope interacts with a sample undergoing observation. A scanning electron microscope (SEM), for example, relies on the interactions between a beam of electrons and atoms in a sample. Light microscopes produce useful information only when a sample alters light passed through it or reflected from it. Likewise, an AFM requires an interaction between atomic forces on a small mechanical tip and those on a sample's surface. By scanning such a tip across a surface, an AFM produces a "map" of heights.

Figure 1. This commercial SPM also performed AFM measurements. The microscope system comes with the scanning and probing stage, control electronics, a data-acquisition system, and a computer that produces an image. Courtesy of JEOL USA.

AFMs form a subset within a larger group of instruments, called scanning-probe microscopes, or SPMs (Figure 1). An SPM can accommodate many types of tips that can measure electrical, mechanical, or magnetic properties. Some tips can perform simultaneous dimensional and electrical measurements. When an SPM uses a tip to sense properties at an atomic level, researchers refer to the instrument as an atomic-force microscope.

Chuck Mooney, the AFM product manager at JEOL USA (Peabody, MA; www.jeol.com), notes, "Collecting information about different forces depends on how you set up the tip. Basically, you have a cantilever—a spring—with a force sensor—the tip—on one end. Usually, the difficult part comes when users must figure out how to set up the sensor so it detects the force they want to look for."

F. Michael Serry, senior applications scientist at Veeco Instruments (Santa Barbara, CA; www.veeco.com) says one of the key features of AFMs is their ability to nondestructively measure 3-D features with resolutions of a few Angstroms. "Nondestructive testing increases in importance as semiconductor device sizes get smaller," adds Serry. "Manufacturers need to perform nondestructive measurements in all three dimensions to ensure their device geometries fall within ever-smaller tolerances."

Jim Jacobs, senior process engineer at Phonon (Simsbury, CT; www.phonon.com), uses an AFM to monitor and detect flaws in surface-acoustic-wave (SAW) devices the company manufactures. "To make a SAW device, we 'print' many fine parallel metal lines, ranging anywhere from 0.5 micron to 6 microns in width," says Jacobs. "And we use an AFM to find where a break occurs in a line or to find a short circuit from one line to another. Often, we can't see those defects with an optical microscope."

Figure 2 . AFM scans from SAW devices show a) a good device with a 4-µm lines, b) a device with contamination between the 4-µm lines, c) 1.2-µm lines with ragged ledges caused by poor "lift off," and d) 4-µm lines with scratches caused during cleaning. Courtesy of Phonon.

Jacobs also uses an AFM to check the results of photolithography process steps. The company uses a lift-off method in which holes in a photoresist mask allow a metal to bond to an underlying substrate to form the lines for a SAW device. At the end of processing, the mask—which now has metal on it, too—lifts off the substrate to leave the deposited lines. "When we remove the photoresist, if we've over- or under-exposed it, instead of leaving sharp edges, we get ragged edges. Edge conditions affect performance, so if we observe signal loss in a device, we use the AFM to take a few scans. The scans show whether or not the SAW devices have clean edges," Jacobs explains. Figure 2 shows scans from portions of several devices.

Teraxion (Sainte-Foy, QC, Canada; www.teraxion.com) also uses an AFM to measure the quality of the photomasks used to control the etching of quartz substrates (Figure 3). After measuring the depths and widths of openings in the mask, the company can decide whether to proceed with etching. In this application, the AFM provides information about the quality of masks, and it lets engineers determine the causes of any mask failures.

Learning to correctly interpret an AFM's data presents users with a challenge. "It took time to learn how to recognize when we had a good scan and to know the difference between good and bad data," says Phonon's Jacobs. "If you try to measure a depth of a channel, and the tip doesn't go all the way to the bottom, you don't get an accurate measurement." Typical depths for the Phonon SAW devices range from 1200 to 3000 Å. AFM tips have a height range of about 10 µm.

Dave Brownell, a staff engineer in the Advanced Technology Group at NVE (Eden Prairie, MN; www.nve.com), agrees that it takes time to learn how to interpret AFM data. In an SEM, the image you see accurately represents the sample. But with an AFM, bad data can look good when displayed as a surface map. "I once had an AFM scan tell me a sample had a surface roughness of two picometers—that's two hundredths of an Angstrom!" says Brownell. "It turned out the last person to use the AFM had it set to keep the probe stationary. The surface map looked good, but the AFM just sampled the same point, over and over again, during the entire 'scan.'"

In addition to detecting defects after manufacturing, an AFM scan can help prevent defects during production. "An AFM helps us get good surfaces on which we deposit giant magneto-resistive—GMR—materials for the sensors we manufacture," says Brownell. "As we receive silicon wafers from suppliers, we measure their surface roughness. To produce a good tunnel junction, which has a barrier thickness of about ten Angstroms, we need a surface roughness of about five to seven Angstroms rms," adds Brownell. "Anything more and the junctions won't work—the barrier becomes discontinuous and shorts the device. But we really want to go down to a surface roughness of about two Angstroms rms. So we use the AFM to help us develop better surface-preparation processes."

JEOL USA's Mooney adds, "An AFM can measure height with great accuracy. If you have a one-nanometer sphere on a flat surface, you'll get a one-nanometer measurement. But the size of a tip limits the width resolution, so users typically reach a limit at about ten nanometers."

Users must remember a feature's measured size equals the dimension of the feature plus the dimension of the tip. Say a tip manufacturer quotes a tip's end radius at <10 nm. Scan a 1-nm feature with that probe, and the AFM will report something about 7–8 nm across.

Although most AFM scans take place under normal lab conditions, some users may need to vary operating temperatures to observe how surfaces change over time. Cooling below about 32°F (0°C) usually requires placing the sample in a vacuum to prevent moisture in the atmosphere from forming ice on it. Heating, however, can take place without a special environment.

Figure 3. The AFM image of an optical phase mask shows the depths of corrugations at the fused silica surface. Depths range from 0 at each end to 240 nm at center. The use of an apodization profile with a width of 40 µm lets researchers see the entire mask profile within the AFM's scanning range. Courtesy of Teraxion.

If AFMs have a weakness, it's their tips. If you view a sample in an optical microscope, you expect it to look the same from day to day. But AFM users must periodically change tips, due to normal wear from contact with a sample.

Unfortunately, every tip has its own characteristics, so scanning a sample with an old tip, replacing it with a new tip, and scanning again may produce a different image. Even tips within the same production lot may produce different results. And if you use coated tips for electrical tests, signal amplitudes can vary, too.

Often, users don't know what sort of a scan they'll get until they plug in a tip and try it. The lack of predictable or reproducible results may make it difficult to quantify data from a scan. Experience and training can overcome this problem, but it never goes away completely.

When an AFM produces an unusual or unexpected surface map, users should scan a sample with known dimensions and characteristics. If that sample also produces unexpected results, the tip may need replacement. Also, AFM users should replace tips when they run new samples. Material from a previous sample may adhere to a tip, thus contaminating the next sample with foreign materials.

Tips that contact surfaces wear at varying rates. A soft tip scanning a hard surface may last for only a few scans, but specially coated tips can last much longer. Some scanning methods sense proximity to a surface, rather than contact, so tips used in such scans do not wear out quickly.

Tip shape also can alter measurements. Imagine a sharp-pointed tip that scans across a rounded surface feature. The AFM produces an image of that rounded surface. Now turn your mental image upside down and imagine a rounded probe scanning a sharp-pointed object. In this case, the AFM also produces an image of a rounded surface—the tip. Users must choose tips geometries appropriate to the features they plan to scan. AFM manufacturers can suggest tips and measurement techniques for special tasks, such as measuring thin-film dielectric strength.

Like all high-quality measuring instruments, AFMs require periodic calibration. Standard grids and gratings available from several companies provide surface features of known depths and widths. The TGS01 set of gratings from MikroMasch (Portland, OR; www.spmtips.com), for example, includes structures with heights of 20, 50, or 100 nm. Calibration should take place with heights, or depths, close to those you plan to measure.