Sharp Corners and Kaleidoscopes Lead to Novel Microscopy Techniques

Hank Hogan

Researchers at Harvard Medical School in Boston have developed two microscopy techniques: a near-field method that offers potentially better performance than other superresolution techniques and a “mirror tunnel” approach that promises high-resolution, wide-field imaging.

The first, dubbed differential near-field scanning optical microscopy, overcomes one of the problems with conventional aperture-based subwavelength imaging. The resolution of such methods is set by the aperture size, with smaller openings capable of imaging correspondingly smaller features. However, the signal and contrast fall off sharply as the aperture diameter shrinks.

A numerical simulation shows how sharp corners can make for sharp images using differential near-field scanning optical microscopy. At upper left (a) is a microscope image of a 2-μm² tissue sample with an overlaid V. The upper-right image (b) is the simulated output intensity of a 1-μm² aperture scanned over the surface. From the two-dimensional derivative (c), it is possible to recover the image (d). Even if noise is added to each pixel (e), it is still possible to extract the image (f). Images courtesy of Aydogan Ozcan, Harvard Medical School.

The differential technique — developed in conjunction with a Harvard University group led by professor Federico Capasso — differs in that it uses a relatively large (0.3 to 2.0 μm wide) rectangular opening, which provides superb resolution because of its sharp corners. When located within 20 nm of a sample, the corners interact with evanescent waves from the object. A superhigh-resolution image can be extracted by recording the resulting power of the collected light.

This approach increases overall light throughput because the aperture can be large and, thus, the signal-to-noise ratio should be better. Moreover, there is a manufacturing benefit. “It is easier to fabricate sharp corners in a relatively large aperture, or detector, than to fabricate small area apertures or detectors,” said team leader Guillermo J. Tearney, associate professor of pathology at the medical school.

A limitation of the differential near-field approach is that the sample can be only twice as large as the aperture. That constraint is not a problem for some applications, and in other cases, it could be overcome by screening off flat, large samples.

The group used the technique to image a 100-nm hole, using the 532-nm light of a frequency-doubled Nd:YAG laser from Witec Instruments Corp. of Savoy, Ill., for a source and a photomultiplier tube for a detector. They demonstrated a transverse resolution of 50 nm with a square aperture 1 μm wide, which they fabricated by cutting a hole in a gold film on a glass substrate.

Since their proof-of-principle demonstration, the scientists have fabricated square apertures on the ends of cantilever-mounted tips, making it possible to use the technique to scan across a sample. They also are planning to improve performance by changing the detection scheme. “We hope to demonstrate even better resolution with more complex objects in the very near future,” Tearney said.

Kaleidoscope inspired

The second imaging system had its genesis in a toy, according to researcher Aydogan Ozcan. In a kaleidoscope, multiple mirrors scramble images and transmit different k-vectors, part of the electromagnetic wave. “These k-vectors could potentially enable wide-field imaging that would have some advantages for full-slide scanning digital microscopy,” he noted.

The group used this idea to build a mirror tunnel microscope, which is composed of two nearly — or completely — parallel mirrors and a low numerical aperture lens. With a test setup, the group achieved an effective fivefold increase in a lens’s NA, enabling the blurless imaging of a pinhole.

The mirror tunnel microscope shown here in a side view (a) and in a conceptual schematic (b) improves the numerical aperture of a lens without diminishing the field of view. Light bouncing between the mirrors and passing through the lens emerges as a series of images that can be summed, potentially providing wide-field capabilities.

A higher-NA lens, however, typically improves resolution at the expense of narrowing the field of view. In the mirror tunnel microscope, light bounces between the facing mirrors and passes through the lens, forming a series of low-resolution images. These can be summed to create a higher-resolution image, increasing the NA without sacrificing the field of view. In theory, the mirror tunnel design could enable relatively simple, digital wide-field microscopy.

In a demonstration of the technique, the researchers imaged a 20-μm-diameter pinhole using two 35-mm parallel planar mirrors spaced 1.2 mm apart. For a light source, they used a helium-neon laser from Melles Griot Inc. of Carlsbad, Calif., operating at 633 nm and a CCD sensor from QImaging of Burnaby, British Columbia, Canada, as a detector. They used a 0.02-NA lens and improved that figure by a factor of five along one direction by summing up the zeroth-, first- and second-order images.

With a four-mirror tunnel, the researchers expect that the device will enable imaging with better than 1-μm resolution over an area of 2.0 × 4.0 cm. Work is under way to implement such a setup. “Our next step is to extend this system to generate wide-field images of microscope slides, which will form the basis of a prototype instrument,” Ozcan said.

Because each image has low resolution, each could be captured with a sensor that has a low pixel count, which would lower the cost of the system while making high frame rates possible.

As for the future, the researchers are considering the strengths and weaknesses of the two techniques: The mirror tunnel microscope cannot resolve objects below the classical diffraction limit because it detects propagating waves. The differential near-field scanning optical microscopy technique, in contrast, resolves objects below the diffraction limit, converting nonpropagating evanescent waves into propagating ones.

They are considering ways by which the two separate techniques could be joined into one. According to Tearney, linking the methods might provide some unique capabilities. “We have some ideas of combining these two inventions of ours in order to achieve superresolution over wide fields of view without requiring a scan.”

NanoLetters, November 2006, pp. 2609-2616; Applied Physics Letters, Sept. 25, 2006, 131124.