Adding Depth to Spectroscopy Imaging

Hank Hogan

Conventional spectrometers examine just one location that scientists hope is representative of a whole sample. With spectroscopic imaging, a single point of information can be replaced by measurements from many locations. Now researchers at the University of Tennessee in Knoxville have enhanced spectroscopic imaging by adding depth information. The technique could be useful, for example, in quality monitoring in the pharmaceutical or food processing industries, where samples come in complex shapes.

In a two-step process, researchers add depth information to infrared spectroscopic imaging. They capture a data cube of IR spectra of a sample at high spatial resolution, illuminate the sample with a visible light pattern generated by projecting a regular mesh onto 3-D samples and extract height information from the distortions. The inset on upper left shows a mesh projected onto a sample, while inset on right shows the added illumination unit. Images reprinted with permission of Analytical Chemistry.

In spectroscopic imaging, conventional two-dimensional pictures are supplemented with spectroscopy, with data acquired from a large number of neighboring sample locations. Each point is then a data cube, with an X and Y position and an associated spectrogram. Armed with this information, researchers can study the spatial distribution of chemicals and reactants.

However, a 2-D image of a 3-D object loses depth information. For some applications, that loss may not matter much. In other cases, however, the missing data is critical because topography affects important chemical or biological reactions. An example of this might be cancerous tissue samples subjected to drug treatments, where assessing growth in all directions is important.

Frank Vogt, assistant professor of chemistry at the university, said that the solution to acquiring 3-D data just popped into his head. He illustrated it to his students by placing a flashlight behind the meshwork of an old air filter. He projected this onto a racket ball, and the palm-size round object distorted the rectangular light pattern in predictable ways. From those distortions of a known mesh, it is possible to reconstruct the object’s topography.

A micromesh with holes measuring 100 μm² and a wire thickness of 30 μm produces a transmission ranging from 30 to 70 percent (left). The micromesh is used to project light on a (middle) sample, which produces distortions in the projected image. The white dotted line (A) traces a light pattern distortion and (B) a discontinuity in the pattern arising from height differences between sample and substrate.

The size and spacing of the mesh determine how small an object can be successfully handled by the technique, Vogt noted. “The mesh should cover the samples at a rather fine resolution, or the topography cannot be resolved at a sufficient level,” he said.

In their demonstration, Vogt and graduate student Michael Gilbert used a mesh with openings of tens of microns to examine samples that measured hundreds of microns. They used a Bruker Optics infrared spectroscopic imager and added a custom-made micromesh, a Sony digital video camera and an illumination unit, selecting a blue LED light source to avoid interference with the infrared measurements.

The micromesh had 100-μm holes and 30-μm-thick wires. Using this and objects of known height, the researchers established a translation from observed distortions into the sample’s height. They could have upped the resolution by stepping the mesh across the sample and recording multiple measurements.

In a first-analysis phase, the scientists used the imager in reflection to do Fourier transform infrared spectroscopy and then determined the chemical composition at various points of a pharmaceutical sample. In a second phase, they measured the sample’s surface topography using the distorted light pattern, determining that a ridge ran along one particular direction.

A drawback of the technique is that the mesh cuts down on transmitted light, and so care must be taken in balancing transmission against light pattern contrast. However, Vogt noted, the approach does not require special equipment.

“Any imaging spectrometer should work just fine. If spectroscopic information is of no importance and only surface structures or changes in them are of interest, any camera would do just fine.”

Analytical Chemistry, July 15, 2007, pp. 5424-5428.