NIST Study Aims to Improve Utility of the Scanning Electron Microscope

Spherical Retarding Field Analyzer

Using an electron beam to image the tiniest of defects and patterns on microchips, the scanning electron microscope (SEM) has long been a mainstay of the semiconductor industry. But as the industry continues to miniaturize chip components--essential for computers, implantable drug dispensers, cell phones, and other devices--the demand for ever-more detailed information from SEM images has grown.

Although the exquisite, atomic-scale resolution of an SEM can’t be improved, researchers at the National Institute of Standards and Technology (NIST) have begun a multi-year study to reduce uncertainties in the measurements inferred from SEM images. To do so, NIST physicist John Villarrubia and his colleagues are conducting a series of experiments in which electrons scatter off different materials, By comparing the results of the scattering experiments to theory, the team hopes to establish a more precise connection between SEM images and the features of the object under study.

An SEM produces atomic-scale images of a sample by scanning the surface with a focused beam of electrons. The interaction between the beam and the sample generates additional electrons that emerge from the sample with a wide range of energies. Those with the lowest energies, known as secondary electrons, are critical for creating SEM images because they originate at or just beneath the surface of the sample. The more energetic electrons (those with energies higher than 50 eV) are of less interest because most of them consist of electrons from the source beam that were backscattered by collisions with nuclei deep within the material.

Determining just how many secondary electrons are generated – and how many are actually recorded by detectors – is key to properly interpreting an SEM image. But computing those two numbers isn’t easy.

For instance, secondary electrons originating from a depression in the sample may be reabsorbed by surrounding material instead of reaching the detector. On the other hand, more secondary electrons emerge from a sloped region than a horizontal one. These effects must be taken into account in order to decipher the true size and shape of surface features in an SEM image. However, physicists have only an approximate understanding of electron scattering, especially at low energies, which leads to uncertainties in interpretation.

“Because our knowledge about electron scattering is incomplete and may have important errors, so will the mathematical models metrologists rely on to interpret the SEM images,” Villarrubia said.

To make sure they are properly accounting for all secondary electrons in SEM images, he and his NIST collaborators, Olga Ridzel and Glenn Holland, designed a simpler but novel scattering experiment. In their study, a beam of electrons strikes the surface of a sample, creating both secondary and backscattered electrons similar to the way an SEM operates.

However, the experiment differs in two important ways from an SEM study. First, the sample is fabricated to be perfectly flat, which makes it easier to analyze the intensity and energy of scattered electrons. Second, the sample is placed inside a device, known as a retarding field analyzer (RFA), that filters backscattered and secondary electrons according to their energy. By adjusting the filters so that only electrons above a certain threshold energy can reach detectors, the team can measure the total number of secondary electrons with high accuracy, as well as the number of secondaries within a particular range of energies.

The team plans to repeat these measurements using different beam energies that fall within the range at which SEMs operate. The researchers will also make the same measurements with the flat sample tilted at different angles in order to assess how changing the slope can alter the number of electrons that are collected.

The scientists will then compare their measurements to the predictions of various theoretical models of electron scattering. One possibility, Villarrubia said, is that one of the existing models may prove to be correct. But more likely, he noted, the comparison “will show us inaccuracies in our best physics models.” The new data then becomes a tool to refine new and existing models of electron scattering so that they match the results of the team’s electron scattering experiment.

Once the researchers identify the best model, they can apply it to the more complex scattering process that occurs when the beam from an SEM scans a transistor or other chip component that has an irregular surface.

Industrial users can then be assured that the SEM images they rely on can truly determine the size of a fracture, the width of a hole no bigger than ten hydrogen atoms, or the topography of a logic gate, the size of a fracture or the width of a hole no wider than a billionth of a meter, or the topography of a logic gate.