Dit artikel is enkel beschikbaar in het Engels
Pol Van Dorpe, principal member of technical staff at the life science group of imec and Harrie Tilmans, program manager at imec, explain how the new concept works and how it can transform many industries.
From medieval paintings to Otzi
Raman spectroscopy was discovered by C. V. Raman in 1928. A discovery that earned him the Nobel Prize for Physics in 1930. Raman Spectroscopy is used to identify materials – fluids, powders, solids – and to learn more about the materials composition. By shining laser light on the material and analyzing a specific part of the scattered light, a spectral chart can be made. The scattered light that is analyzed is the so-called ‘Raman scatter’, which is characterized by a different wavelength than the laser wavelength. It originates from the molecular vibrations in the material. In the spectral chart, one can recognize the spectral ‘fingerprint’ of each material and compound.
Pol Van Dorpe: “Raman spectroscopy is used in many different fields – in fact, it can be used in any application where non-destructive material analysis is needed.
For example: in pharmaceutics, Raman discloses the distribution of active compounds in tablets; Raman can also classify meteorites in chondrites and achondrites based on its mineral composition; it can even dive into the kind of hybridization in a (carbon) molecule and tell whether it’s an sp2 or sp3 hybridization; in semiconductor R&D, it can be used to determine the electrical properties and the number of layers of graphene, or characterize the stress in specific layers; in life sciences, Raman is able to tell something about the interaction between cells and specific drugs; in art work, the technique is used to identify which pigments were used which in turn can give an indication about the artist’s identity, his working method, and the age of the work; even the mummified skin of Otzi the Alpine Iceman was examined with Raman spectroscopy.”
Raman spectroscopy can be used for a wide range of applications: to study old paintings, meteorites, mummies or – more industry relevant – to determine the quality of milk, food, medicines, electronic structures etc. Each compound has its own Raman ‘fingerprint’ and can be identified in this way.
Existing devices: desktop and handheld
Numerous Raman spectroscopes are available today, specifically tuned to the varied applications. Raman systems are often implemented as a microscope where the sample is illuminated with a diffraction limited spot and the Raman photons are collected with a high-NA objective (NA: numerical aperture, a measure of the objective’s ability to gather light and resolve fine specimen detail at a fixed object distance). Next to these desktop devices (which make up the largest part of the market), also handheld counterparts have emerged. These are really convenient for research ‘in the field’ like in the case of art work studies or archeology.
As is often the case with miniaturized versions of measuring equipment, existing handheld Raman spectroscopes are not as good as the desktops.
For the study of the pigments in paintings, they work fine because the pigments generate strong Raman signals, but for more complex samples such as opaque fluids (such as milk), skin or powders, the current systems are not adequate or need a long measuring time. Also, the high pricing of handheld Raman spectroscopes inhibits its widespread use.
What if a cheap but high-performant handheld Raman spectroscope was available? Which new applications would be made possible?
Harrie Tilmans: “Think of food screening ‘on the go’, skin screening for melanoma at the general practitioner’s office, screening for drug authenticity throughout the supply chain, by authorities, Within the food and beverage industry, it can be used for characterization purposes and for identifying and evaluating the authenticity, safety and quality attributes for a broad range of food and agricultural products. For instance, Raman is an ideal technique to identify and indicate the adulteration of edible oils or to determine the level of alcohol in beverages. Multiple component analysis using Raman spectroscopy is key in for example understanding the quality of milk, e.g., by determining the fat, protein and water content of the milk a cow or a goat is producing (in real time). For cosmetics, Raman spectroscopy can be used for verification of the purity of ingredients for use in personalized moisturizing creams and essential oils (eucalyptus, ginger, lavender, basil, vanilla). Furthermore, it can be used to determine the remaining concentration of solvents in pharmaceutical preparations, to decide if the level of nutrition is sufficient for a certain cell culture to maintain their growth.”
What if a cheap but high-performant handheld Raman spectroscope was available? Then complex samples such as milk could be analyzed 'in the field'.
Why is it difficult to further improve current handheld devices?
“There are two main challenges in developing Raman systems,” explains Pol Van Dorpe. “First of all, spontaneous Raman scattering is typically very weak, and as a result the main difficulty of Raman spectroscopy is separating the weak inelastically scattered light from the intense Rayleigh scattered laser light.
Secondly, in strongly scattering media, such as food or human tissue the incident photons are not confined in a small spot, but instead generate a blurred spot, with sizes up to several mm. This increases the “optical throughput” or “etendue”, which is a measure for the spread of the light in space and angle. The spectrometer usually limits the etendue, which drops for compact devices.
In commonly used dispersive spectrometers, the light is focused on a slit, and its spectral components are separated using a diffraction grating. Miniaturization of a high spectral resolution device (<1 nm) requires a reduction in the slit width, thus reducing the etendue.”
Existing dispersive handheld spectrometers make use of entrance slits, mirrors, gratings, ... like is indicated in this scheme. In this way, the etendue and spectral resolution are coupled, through the size of the entrance slit and downscaling of the spectrometer leads to lower spectral resolution or acceptance etendue.
“Ultimate miniaturization of optical devices can be realized using integrated photonics and waveguides.
The etendue of a single-mode waveguide (which is the ultimate scaling of this approach) equals approximately 𝜆², with 𝜆the wavelength. A wavelength of for example 860nm, results in an etendue of 7.3e-7 mm²sr, which is 106-107 times lower than required for analyzing diffuse scattering samples,” concludes Pol Van Dorpe.
Existing photonic-based handheld spectrometers are limited in miniaturization by the etendue of single-mode waveguides.
The new concept: a spatially heterodyne spectrometer based on silicon photonics
Because of the limitations of current approaches in handheld Raman spectrometers, other concepts are investigated. Another class of spectroscopy is based on light interference and implemented for instance in Fourier transform spectroscopy or in spatially heterodyne spectroscopy. These concepts exhibit an intrinsically larger etendue and are therefore more ‘scaling proof’.
A well-known Fourier Transform spectrometer is based on the Michelson interferometer. A beam of light is divided into two beams that take different paths before coming together and interfering. This enables tiny differences in the wavelength to be measured. The disadvantage of this design – particularly if you want to miniaturize it – is that two mirrors are used, one of which moves.
Harrie Tilmans: “We developed an integrated photonics version of such a scheme, without moving parts: the integrated spatially heterodyne spectrometer. The etendue is less limited in this case and equals n x 𝜆², with n the number of interferometers. As mentioned above, if the used wavelength is for example 860nm, and the preferred etendue (for complex samples) is around 0,5 mm²sr, than about a million interferometers would be needed.
This massive parallelization is possible with integrated photonics. The patented solution monolithically integrates close to a million interferometers on top of CMOS image sensors and light is delivered using micromirrors.Each interferometer is a little smaller than the previous one so that tiny differences in wavelengths can also be measured, as is the case with the Michelson interferometer.”
Concept of the patented solution made by imec researchers for a handheld Raman spectrometer, using a million interferometers on top of a CMOS image sensor. This build-up allows for extreme miniaturization without compromising on the etendue of the sample. This way, also complex samples can be measured. Furthermore, by using chip technology, the price of the device can be much lower than current devices.
Pol Van Dorpe explains the working principle of the newly developed Raman spectrometer: “Laser light (785 nm wavelength) is focused on a sample and thescattered photons are collected and collimated by means of a compound parabolic concentrator (CPC). After filtering the Rayleigh photons (@785nm), Raman photons are directed to the on-chip waveguide (WG) accessports using a wedge-shaped light guide and chip-integrated micromirrors. By a proper choice of thewedge shape and the angle of the incident light, the micromirrors are able to redirect the light with alarge efficiency (>50%), where it is coupled into the WGs thru gratings linked to individual single modeWG interferometers. The interferometers exhibit a range of lengths, allowing for reconstruction of theoriginal spectrum. Their outputs are aligned with the pixel pitch of the integrated CMOS image sensorserving as a highly parallelized detector array. The chip is wire-bonded to a PCB and connected to acustom designed read-out board to capture the data and transfer it to a computing device, that reconstructs the spectrum and displays the required property.”
Silicon nitride is the material of choice for the waveguides
Harrie Tilmans: “Due to requirements of CMOS compatibility and visible transparency, silicon nitride was chosen as the WG material. The WG stack is built monolithically on top of the BEOL of 200mm front-side illuminated CMOS image sensor wafers. Post-processing is done in a 200mm CMOS pilot line, using 193 nm DUV lithography for patterning the WGs and grating couplers.
The spectrometer chip is based on SiN-based waveguide photonics, implemented on top of a CMOS image sensor (CIS) used for electrical readout. The spectrometer chip as used in the current design consists of an array of massively parallel evanescently-coupled Fabry-Perot interferometers, varying in length in the range 2.2-152.8µm (in linear steps of 0.2µm). Incident light is coupled into the waveguide structures using a grating based in-coupler (GC). Sloped metal output mirrors are used to couple the light from the waveguide to the readout pixels of the CIS (see also insert with a cross-section of the chip in figure above). An illustration of the layout showing a top view of the F-P resonators together with the grating incoupler and sloped metal output mirror is shown in the figure below.”
Part of the spectrometer chip layout, showing an evanescently coupled Fabry-Perot resonator (17.6µm long) together with the grating in-coupler(s) and the sloped output mirror for coupling the light to the pixels.
Raman spectroscopy is a powerful technique with numerous applications. Existing devices are rather bulky (desktop) and have a price range of a few hundred thousand dollars/euros. The handheld solutions that exist today fail to reach the desired performance for high-end applications. Thanks to imec’s new concept, it is now possible to overcome this performance barrier. By massive parallelization of waveguide interferometers integrated monolithically on top of a CMOS image sensor, both high optical throughput and high spectral resolution can be reached in a miniaturized device.This novel system is built in imec’s SiN biophotonics platform which guarantees robustness and compatibility with high-volume manufacturing.
Part of the work was performed under the EU-funded IoSense project: The authors acknowledge the funding received from the Electronic Component Systems for European Leadership Joint Undertaking (ECSEL-JU) under grant agreement No 692480 (Project Acronym: IoSense).This Joint Undertaking receives support from the European Union’s Horizon 2020 research and innovation programme and Germany, Netherlands, Spain, Austria, Belgium, Slovakia.