Chip-Based Spectrometer to Reduce Chemical Sensor Cost

By eliminating the mechanical components of traditional spectrometers, the chip-based sensor is far more robust and less prone to damage. (Image courtesy of MIT.)

These days most types of sensors can be reimagined on a chip. Miniaturizing the spectrometer however has continued to elude engineering capabilities. Typical spectrometers are often expensive, and are usually only found in university labs, where they are used to determine the chemical composition of samples in the lab or objects in space. Constructing a spectrometer-on-a-chip would allow these chemical sensors to be produced as easily and inexpensively as any other silicon chip. Now, a team from MIT has published research showcasing successful chip-based spectroscopy. The research was published in Nature Communications.

Spectrometers are chemical sensors that have high potential for industries that work with organic materials, such as the oil and gas or the food and agriculture industries. Making spectrometers more affordable for industrial uses has been on the rise. MIT has previously made attempts to develop hand-held spectrometers. However, these models were still subject to the constraints of traditional spectrometers. The French research institution CEA Tech is also making an attempt at spectrometers-on-a-chip, but only for the infrared range, with the goal of developing a detector for gas leaks.

The new team of MIT researchers are going above and beyond both aforementioned designs, as their model is a spectrometer-on-a-chip that can measure a much wider chunk of the EM spectrum. They believe that their new approach to making spectrometers-on-a-chip could provide major advantages in performance, size, weight and power consumption, compared to the types of instruments currently on the market.

So how does the spectrometer-on-a-chip work? Spectrometers are essentially light sensors that can measure the amount of light that is being emitted or absorbed at each wavelength. The result is a type of graph known as a “spectrum,” where the y-axis is light intensity and the x-axis is the wavelength of the light. These spectra can be used to identify the chemical makeup of any object that is emitting or absorbing the light. As each atom and molecule in nature has its own unique fingerprint, reading a spectrum to identify a chemical component is similar to reading a barcode.

The difficulty with building a spectrometer-on-a-chip mainly has to do with breaking the light up so that its intensity can be measured by wavelength. Typically, this is done with mirrors, but adapting this traditional method to silicon is of course not directly possible.

Instead, the team took a whole different approach, instead using a technique that employs “optical switches.” These optical switches can divide the light according to wavelength by flipping a beam between a selection of different optical pathways, which can be of different lengths. According to doctoral student Derek Kita, who contributed to the research, these optical switches are essentially a silicon replacement for what mechanical mirrors do in traditional spectrometers.

“By eliminating the moving parts, there’s a huge benefit in terms of robustness,” explained Kita. “You could drop [the spectrometer] off the table without causing any damage.”

According to the researchers, reinventing the spectrometer-on-a-chip was aided by developments in artificial intelligence computing. The team used machine-learning techniques to convert the spectrometer’s output into detailed spectra. The method they developed can detect both broad and narrow spectral peaks, Kita said, opening up a wide range of potential development for various applications.

If the spectrometer-on-a-chip continues to perform well outside of the lab, its uses could go far beyond gas detection. The team says that sensing devices, materials analysis systems and optical coherence tomography in medical imaging are all on the list of potential applications. The spectrometer may also have a future in monitoring the performance of optical networks, upon which most of today’s digital networks rely.

Kita further stated that the team has already been contacted by some companies interested in possible uses for the microchip spectrometer. The research has also grabbed the attention of others in the field.

“[The work] is a very interesting approach, as it enables realizing a high-resolution spectrometer on a small footprint,” said Gunther Roelkens, a professor at Ghent University in Belgium, who is unaffiliated with the research. “This device enables applications such as on-chip spectroscopic sensors, which is a hot research topic.”

“The challenge for future research will be to extend the wavelength coverage while maintaining the same resolution,” Roelkens added. “Also, addressing different wavelength bands will enable many new applications.”