Dynamic Dichroic Mirrors and Single-Shot Spectroscopy

Optical spectroscopy involves splitting light up into its component wavelengths, exactly like how a glass prism splits sunlight up into a rainbow (or 'spectrum') of colours. While this may, at first, just appear to be aesthetically pleasing, the rainbow of colours does include some very useful information about the light it came from. For example, if one were to look very carefully at the orange part of the spectrum of the sun (roughly around the colour corresponding to old orange street lights), one would find a pair of dark lines where the light appears to be missing. This corresponds to absorption due to sodium, which tells us that there is sodium in the upper atmosphere (or 'chromosphere') of the sun; we have, in effect, learned part of what the sun is made of, from 93 million miles away, using nothing more than a glass prism. This is the power of optical spectroscopy.

Optical spectroscopy has many other uses; it can investigate the chemical composition of forensic samples, help locate a tumour, identify chemical weapons from a distance, monitor deforestation from orbit, authenticate artwork, and many more besides. Nevertheless, if we try to take pictures like we would do with a camera, there's a problem; a camera can only capture 2D information, and if we have a spectrum at each pixel, we either need to illuminate one line at a time (and use the other axis of the camera to measure the spectrum) or use a sequence of filters to get the data one wavelength at a time. In fact, if we know what spectrum we're looking for, the filter approach is much faster, but that requires carrying a stack of filters for each thing you might want to measure. That might be OK for a few things, but for portable or space-based applications, or if there are a lot of potential analytes, that can become infeasible. The alternative is to make a new filter each time, using a photorefractive polymer.

This new system can create any filter that the user might want, writing it into a material similar to that used to make holograms. This is unlike normal tunable filters, which are typically only capable of tuning a single transmission band's width and center wavelength. This new approach can create any filter profile the user might want, including multiple independent bands and different band shapes. It works by recording an interference pattern in a holographic plate, using two laser beams. The angle between the laser beams is changed, and another interference pattern is recorded. After doing this many times, the filter profile is recorded in the hologram; this entire process takes less than a second. Once the filter isn't needed any more, it can be rewritten, and a new pattern created; any number of analytes can be searched for, including those which the system has never seen before; they need only be programmed in by the control computer.

The holographic material is a new type known as a photorefractive polymer. Unlike normal holograms, this material is rewritable; it can be erased, and a new pattern written into it. While this has been used to create rewritable holographic displays before, this is the first example where the material will be used to create a holographic filter. Nevertheless, synthesizing it is not difficult; it requires just two components to be made from scratch, and these are both easy synthetic procedures with high yields.

Overall, this project offers to create a kind of 'Instagram for spectroscopy'; rather than being limited to a small selection of physical camera filters, a user can digitally apply any one that might be needed, including programming a new one from scratch if necessary. This makes the spectroscopic imaging process faster, and more efficient, allowing the user to gather data over larger areas, and with more precision, than ever before.

Aside from the academic impact, this technology has the potential to benefit a large number of other stakeholders. As noted elsewhere, while this is intended purely as a prototype instrument, once it is constructed and characterized, altering it for other applications is trivial. The Rowlands Lab research program will be pursuing some of these applications, but the designs and characterization data will be disseminated freely to encourage users to adapt the design for other purposes. Some example stakeholders include:

1) Soldiers and homeland security (standoff threat detection). The ability to detect and monitor small molecules in the environment is important when these molecules may be hazardous. Chemical warfare agents such as chlorine and mustard gas can be identified by their vibrational spectra and hence avoided; some toxins even have a pronounced colour, hence can be detected by their UV/Vis spectra alone.

2) Environmental monitoring. By enabling earth scientists to study larger areas in more detail, the resulting information directly informs political decisionmaking, commercial development and pollution control. The population as a whole all benefit from these efforts, since it enables natural resources to be preserved, and the population's exposure to toxic chemicals to be minimized. Similarly, monitoring of emissions from power stations, factories and other potential polluters can be done over a larger area and at lower cost, further protecting the population from potentially harmful contaminants.

3) Forensic investigators. The ability to rapidly sweep a crime scene for signs of blood, semen or other forensic samples is clearly of interest in criminal investigations. Investigators could tune the filter to that of a vibrational spectrum corresponding to an analyte (such as blood) and allow a computer to analyse the image to determine when there are stains that might be worth further study. The ability to rapidly switch between different filters, and to process the entire scene as an image rather than having to take spectra one line at a time, minimizes the possibility that a sample is missed, or isn't classified correctly.

4) Doctors / surgeons. A great deal of ongoing research is focussed on the use of hyperspectral imaging in medicine. For example, Raman microscopy is under investigation as a means of locating tumour cells, and image-guided fluorescence surgery is being trialed in order to locate a tumour margin. Bacteria and other toxins can be identified by their optical spectra as well. By developing a high-speed hyperspectral imaging system, these techniques could be sped up in order to make them of practical use in the clinic, rather than just curiosities to be investigated in the laboratory.

5) Manufacturing. As noted elsewhere, one potential application of dynamic hyperspectral imaging would be quality control. Being able to monitor a reaction, assess the composition of a polymer, or determine the contents of a dye sample are of widespread interest throughout the manufacturing sector, and by increasing hyperspectral image throughput to match the rate of production of a modern industrial process, these applications can be pursued, rather than being avoided because of a lack of speed.

6) Robotics. Robots use many different types of hyperspectral imaging to sense the environment, from near-infrared LIDAR to multiple cameras operating in the visible, to long-wave infrared thermal imaging. Being able to rapidly filter the light to, for example, image at wavelengths which easily penetrate foliage, or reject interference from lasers and other bright sources would enable the robot to be more robust and to cope with a larger number of environmental conditions.

Along with the benefits highlighted above, the project has a number of opportunities for training researchers in precision optical engineering and high-tech manufacturing, and these will be a source of additional impact.