Optical Metrology Breaks Barriers

HANK HOGAN, CONTRIBUTING EDITOR, hank.hogan@photonics.com

At one time, the diffraction limit, half the wavelength of light, acted as a cutoff to optical measurement. That’s no longer the case. Superresolution techniques now allow optical metrology well below that point in the XY direction, while interferometry and other approaches produce precise optical measurement in the Z, or vertical, direction.

Other innovations involve laser speckle analysis and the combination of different technologies, enabling better metrology on ever smaller features. On the horizon are techniques that could enable wide-field superresolution.

Confocal-based optical metrology enables characterization of a thin aluminum nitride film on silicon. Courtesy of Leica Microsystems.

Optical measurement is typically employed for two reasons, according to Oscar Rodriguez, manager of application specialists for industry and material science in Europe for Leica Microsystems GmbH. The Wetzlar, Germany-based company makes subdiffraction limit resolution systems based upon selective deactivation of fluorophores in biological samples.

“You need to use optical technologies because of the [feature] size. We’re entering into the microscopic world. Or you’re using optical technologies because they’re noncontact and fast,” Rodriguez said.

With standard resolution techniques, optical measurements of a few hundreds of nanometers are achievable in X and Y. In Z, technologies based on interferometric, confocal and focus variation techniques enable much more precise metrology.

Confocal imaging (above) enables the creation of a three-dimensional view (at left) for optical metrology of bond bumps on semiconductor chips. Courtesy of Nikon Metrology.

Interferometry exploits the interference between light beams, which start as one, are split and then brought back together after traveling down different paths. Interferometric systems can detect pathway differences in nanometers.

Other three-dimensional metrology techniques are confocal or focus variation based. Both depend on the blurring of features outside a focal point, with this being particularly acute for a confocal system. Sweeping the focal point through space enables measurements to be made.

The different techniques offer distinct advantages and end users are interested in having flexibility. “The trend that we’re seeing is to combine different measuring technologies,” Rodriguez said.

Another trend is to combine a laser with a broadband source, like an LED. By doing so, systems can do coherent- and noncoherent-source metrology and merge that with high-speed imaging.

Interferometry forms the basis for many of the metrology products from Middlefield, Conn.-based Zygo Corp., according to Peter de Groot, executive director of research and development. He noted that building optics depends upon making measurements in the microns or nanometers.

High-volume examples of this can be found in today’s phones, which have cameras constructed with low-cost, high-precision optics. Those cameras, in turn, can be used to make optical measurements. In an SPIE 2014 paper, J.H. Burge and others from the University of Arizona showed that an app on a camera-equipped phone could produce results comparable to a standard interferometer. Another example of phone-based metrology is an affordable, internet-connected imaging system in portable microscopes for field work in the developing world, de Groot said.

Optical metrology helps make the optics that enable optical measurement possible, as shown here with a confocal system image of a lens surface. Courtesy of Nikon Metrology.

The need for measurements in the micron range is now showing up in nonoptics areas, such as car fuel injectors. Emission and fuel efficiency requirements have resulted in tolerances of a half micron or less, far smaller than the tolerances of thousandths of an inch, or 50-plus microns, of years past.

“That’s a trend which has caused that entire industry to switch over from mechanical tools to high-performance optical noncontact tools that give them aerial surface measurements on the scale of a few nanometers,” de Groot said.

Software compensates for optics’ shortcomings

For optical metrology systems, the increasing use of software means that lenses do not have to be as high quality as in the past because software can oversample the incoming signal, do noise filtering or do a better job weeding out weak signals. Software and optical techniques also minimize the coherent noise created by a laser beam scattering off dust and particles in the light path, thereby improving measurements.

According to Bhanu Singh, confocal product manager for Nikon Metrology of Brighton, Mich., the demand today is for automated multisensor systems. These might combine different optical techniques, like confocal and bright-field images, to create the capability for wide area X, Y and Z measurements. Such a combination can offer significant advantages in industrial or other high throughput settings.

Many industrial components, like fuel injectors, now require measurement tolerances well below a micron, making them candidates for optical metrology. courtesy of Zygo Corp.

“You have a single setup. It reduces your setup time. You don’t need multiple fixtures across different systems. You don’t need different tools,” Singh said.

A high degree of automation also provides data that can be tied to a specific part at a particular point in manufacturing. Measurement traceability is a requirement in medical device and aerospace manufacturing.

In an example of combining techniques, Nikon Metrology presented at a recent semiconductor industry event a solution to the problem of inspecting bond bumps on chips and probe tips on test boards. The confocal system sends the output of a high-intensity mercury xenon lamp through a pinhole in a spinning Nipkow disk. That is complemented by two additional sensors: one for bright-field optics and another using a laser. This yields two-dimensional bright-field and three-dimensional confocal images over a 650- × 550-mm area with a 10-nm resolution in the Z.

The system’s measurement capability is good enough to handle the 5- to 10-micron demands of today’s circuit board manufacturers, according to Singh. He added that measurement needs should move into the 1- to 5-micron range in three years or so.

More data = more challenges

A consequence of greater resolution, the use of multiple wavelengths, and increasing automation is that more data is generated. That presents challenges in moving, storing and analyzing it.

Singh hasn’t seen a demand for wide-field superresolution. He attributed this to the lack of a compelling reason to implement such a capability. There’s also the cost of such solutions, which currently have only been demonstrated in laboratory settings.

As for the future, even with the increasing use of software to make up for component deficiencies, metrology benefits from hardware advances and innovations. Thus, laser beam quality improvements help. So, too, can steps taken to reduce noise arising from the interaction of the laser and the object being measured. Such interaction produces speckle, created by diffuse scattering from the reflecting surface.

The resulting phase noise is a fundamental precision limitation in coherent laser ranging and so presents a problem to some optical metrology techniques, according to research from the National Institute of Standards and Technology (NIST) in Boulder, Colo. Esther Baumann, a NIST scientist, was lead author of a 2014 Optics Letters paper about this work.

She said that there are approaches to mitigate speckle noise, such as bright speckle tracking that is then used in laser feedback or to adapt the scanning pattern. Another solution is to filter out dark speckle in software. A third is a laser center frequency change when a dark speckle occurs.

In their paper, the NIST team demonstrated yet another approach: a laser with a high optical sweep bandwidth. Scanning the laser center frequency across a terahertz range, the researchers achieved range precisions below 10 µm, opening up the possibility of precise optical metrology at a distance.

A cactus imaged at almost 11 meters using a laser ranging system capable of 10-µm resolution shows the possibility of precise optical metrology at a distance. Courtesy of NIST.

However, the proof of principle system wasn’t in a configuration suitable for a commercial application. “Our setup was rather big and not very rugged,” Baumann said.

She added that shrinking the size of the demonstration system is a job best left to industry. So, too, is determining the best approach to reducing speckle noise.

A final likely future optical metrology development involves wide-field superresolution. There’s significant research going on aimed at demonstrating and then deploying systems that image and perform measurements in X and Y far below the diffraction limit of light. The present lack of commercial products that do this is not surprising, said Nicholas X. Fang, a professor of mechanical engineering at the Massachusetts Institute of Technology. Superresolution is one of Fang’s research interests.

From theory to lab

Fang noted that it may take 20 years or more for a technique to move from theory to lab demonstration to commercial practice. Based upon the theoretical discovery of superresolution in the mid-1990s, it may be five or 10 more years before a commercial product appears.

Some superresolution techniques improve measurement at the expense of scan time. For example, those that selectively turn on and off fluorophores require multiple passes. Other techniques, such as interferometry, only provide better-than-diffraction-limit metrology in one direction.

Possible techniques that avoid such issues use metamaterials. These are made of regular repeating structures that are much smaller than the wavelength of interest.

According to Fang, the semiconductor industry may both need superresolution and have the means to achieve it. The need is that the critical dimensions of advanced chip structures are closing in on 10 nm. Currently, imaging and measurement of these features is nonoptical. However, an ideal solution would be one that covers large areas economically, much like optical microscopes did for earlier chip generations.

As for the solution, the same structures — metal lines and connecting holes or vias — that need inspection and metrology can form the basis for a metamaterial. Semiconductor manufacturers routinely create regular repeating patterns in chips out of elements that are much smaller than the wavelength of light.

Discussing possible components for a superresolution metamaterial, Fang said, “Those interconnects and vias become the natural candidates.”