This article discusses the first demonstration of a new spectrometer, which can achieve < 4cm-1 resolution, when it is coupled to a 3 mm diameter optical fiber, without using a slit.
The spectrometer system is based on a static Fourier transform concept. The device has undergone tests by making transmission Raman observations, where the light was coupled to the spectrometer through fibers with different apertures.
Loss of resolution was not observed, and there was an increase in light corresponding to the fiber collecting area. Similar gains were noted when the system was tested in a stand-off Raman configuration. Compared to a conventional dispersive system, the total throughput gain is 500 times more.
Raman Observations
Over the past few years, Raman measurement has emerged as a major tool for many applications such as; quality assurance (QA) in the process industry, detecting counterfeit drugs and explosives by the security agencies, and enabling online measurements during the production of pharmaceuticals. Classic Raman observations are carried out in a backscatter configuration, by placing the target at a distance of 10 - 20 mm from the instrument collection optics. This technique is commonly utilized by hand-held devices.
However this technique is not suitable for several industrial applications, especially for a bulk measurement of the sample or when the sample is heterogeneous. Transmission Raman observations have once again become a suitable approach for several QA applications. In this approach, Raman photons are observed from the complete sample, instead of one localized surface point. However, this approach provides many challenges for the Raman spectrometer:
- Compared to a back scatter configuration, the signal strength is usually lower.
- Light emerging from the sample comes from a considerably increased Area (A) × Solid angle (Ω) product.
The second issue must be resolved by the design of the spectrometer. For a given resolution, a spectrometer can only accept light from the A Ω product or étendue. This can be controlled by the slit width in a Czerny-Turner arrangement. ISI’s HES range and other static Fourier transform spectrometers can deliver more than 100 times of the total throughput, compared to a dispersive system, with a potential increase in excess of 500 times.
The large étendue challenge also exists when conducting stand-off Raman observations at distances of more than 60 mm. This can restrict the ability of the Raman system to investigate processes in adverse conditions; including high temperature process vessels. This article also discusses the étendue advantage offered by the static Fourier transform spectrometers. Based on a spatial heterodyne spectrometer, a new device has been introduced that accepts fiber inputs from 50 µm core diameter up to 3 mm. This feature improves the system’s advantages, and can potentially offer a throughput gain of more than 1000 times, compared to an identical resolution Czerny Turner instrument.
The Raman Spectrometer
Table 1 displays the specifications for the newly developed Raman spectrometer. The spectrometer is fiber coupled via a bespoke design to accommodate a >3 mm core fiber, or through a fiber SMA connector. Both of the collimating and imaging lenses have a 50 mm working aperture, and 80 mm and 100 mm focal lengths. An acylindrical lens is fixed before the detector to ensure all of the light is captured.
Table 1. Specifications of spectrometer.
| Parameter |
Value |
Notes |
| Laser |
|
|
| Operating wavelength |
785 nm |
|
| Power |
500 mW |
Laser fibre coupled |
| Gratings area |
2 × (52 × 52 mm) |
|
| Lines per mm |
60 |
|
| Working aperture |
34.4 mm |
|
| Detector |
Andor IVAC |
|
| Pixels |
1650 × 200 |
|
| Pitch |
16 µm |
|
| Resolution |
3.1 cm-1 |
Per pixel |
| Spectral range |
50 cm-1 – 2500 cm-1 |
|
| Maximum theoretical acceptance fibre aperture |
3.4 mm |
Diameter |
| Fibre NA |
0.22 |
|
Andor’s cooled IVAC CCD was used as a detector, as with other HES2000 models. The detector has 1650 × 200 pixels, and can be cooled to -60°C. A 785 nm diode laser, with an output laser power of 500 mW and a laser bandwidth of <100 pm, was utilized as an illumination source. To suppress the Rayleigh scattering, two long pass filters are employed. To limit the contaminant wavelengths, a narrow bandwidth filter (FWHM = 4 nm) was mounted just after the laser output fiber.
Table 2. displays the range of ‘receive’ fibers employed for the test program. The fibers were 2 m in length, and SMA connectors were used to terminate them. A custom ferrule was used to terminate a 3 mm bundle. A bundle of 550 µm core fibers was used to form 2 mm and 3 mm diameter fibers.
Table 2. List of fibers.
| Fibre Diameter |
Fibre Na |
Type / Notes |
| 3 mm |
0.22 |
Bundle 17 fibres with 0.55 mm core diameter NA of 0.22 |
| 2 mm |
0.22 |
Bundle 7 fibres with 0.55 mm core diameter NA of 0.22 |
| 1 mm |
0.39 |
Multimode |
| 0.91 mm |
0.22 |
Multimode |
| 0.6 mm |
0.22 |
Multimode |
| 0.55 mm |
0.22 |
Multimode |
| 0.4 mm |
0.39 |
Multimode |
| 0.2 mm |
0.39 |
Multimode |
| 0.05 mm |
0.22 |
Multimode |
A 3 mm thick paracetamol tablet was used to make the measurements against. A 1 mm diameter laser spot was used to illuminate the tablet at a distance of 785 nm and observed in a Transmission Raman configuration. Using this approach guarantees that the étendue of the target light is large, as the light emerges from an important fraction of the tablet surface area and performs the role of a Lambertian scatterer. Two 50 mm focal length lenses were used in the light capture assembly to capture a same-sized area on the tablet as the fiber core (so magnification is equal to 1), with an NA of approximately 0.22. If higher NA fibers are used there is no excitation of the higher modes, and the light exiting the fiber has an NA of approximately 0.22.
Quality of Diffraction Gratings
As the diffraction gratings have large illuminated aperture, the problems relating to the underlying plate flatness have to be analyzed. Figure 1 shows that the plates are under tension, and the laser’s emerging fringe pattern is examined with the system tuned just off a Littrow configuration. The shape of the fringes shows that the flatness of the plate is less than λ/2 (the path of a true fringe is indicated by a red line). The resolution will be lost when a cylindrical lens is used to compress the fringes.
.jpg)
Figure 1. Underlying Fizeau Fringe pattern for HES spectrometer but with plates placed under tension.
Image Credit: IS-Instruments
Figure 2 shows that the reduction in the fringes’ relative distortion, when the tension is relaxed, allows the cylinder lens to compress the fringes without significant blurring.
.jpg)
Figure 2. Underlying Fizeau Fringe pattern for HES spectrometer with tension in the plates reduced.
Image Credit: IS-Instruments
This is a critical step to achieve the best resolution of the system. The relative distance separating the fringes is vital in the Fizeau fringe configuration. Figure 3 shows the final spectrometer when the gratings are correctly tuned, with the 3 mm fiber ferule mount fixed.
.jpg)
Figure 3. HES 2003 spectrometer.
Image Credit: IS-Instruments
Transmission Raman Experiments
The Baseline Spectra
The arrangement for collecting the transmission Raman light, and coupling it into the spectrometer is displayed in Figure 4. In all of the experiments, the fiber (bundle) was the only part that was replaced. All of the tests were carried out on the same 3 mm thick paracetamol tablet. To ensure fluorescence-free clean spectra, the external coating of the tablet was removed.
.jpg)
Figure 4. Transmission Raman configuration.
Image Credit: IS-Instruments
A base spectrum was generated to compare all other data sets. To do this a 2 mm fiber bundle with a 2.5 second integration time was used to take a spectrum to provide an average count number per pixel of 3970. Figure 5 displays the resulting spectrum. A Hanning window was used to apodise the data prior to its Fourier transformation.
A high signal-to-noise ratio was realized, whilst ensuring that practical exposure lengths could be used to collect the same signal levels when using smaller core fibers. The read noise was 6 counts s-1 pixel-1 though the detector dark count was negligible.
.jpg)
Figure 5. Paracetamol spectra, light coupled via a 2 mm fiber bundle.
Image Credit: IS-Instruments
In addition, when the laser was switched off a background count rate (of approximately 0.3 counts s-1 pixel-1) was also determined. This occurred because of an imperfect light seal with the Raman system to accommodate the various fiber mounting mechanisms. The fiber was systematically replaced for all of the fibers given in Table 2.
The exposure duration was adjusted to yield 3970 counts per pixel, and the spectra was recorded. Figure 6 shows a selection of these spectra. The two fiber bundles were configured in a round-to-round arrangement. When the 0.39 NA fibers were used, the number of bends were minimized to reduce the higher mode excitations. This was validated by visual observation.
.jpg)
Figure 6. Paracetamol Raman spectra measured in transmission, light is coupled by fiber of different diameters: Black Line= 2 mm fiber bundle; orange line = 0.91 mm multimode fiber; blue line = 0.4 mm multimode fiber; red dotted line = 0.2 mm multimode fiber; grey line = 3 mm fiber bundle.
Image Credit: IS-Instruments
The 3 mm fiber bundle is the final fiber to be tested. The coupling assembly had to be adjusted to utilize the 3 mm ferrule when this fiber bundle was used. Figure 7 shows the resulting spectrum, and the spectra from the 1 mm diameter 0.39 NA fiber for comparison. The exposure time was 1.09 seconds.
There was no loss of resolution, and the peaks around 1630 cm-1 seem to be better resolved, compared to those when smaller fibers were used. The reason for this effect is the improved light divergence when exiting the bundle, which illuminates more lines at the grating surface. A small change in the calibration also occurred after a 3 mm aperture fiber was used. This change took place when the 3 mm ferrule mount replaced the SMA connector, and the relative positions not being identical.
.jpg)
Figure 7. Paracetamol spectra, light coupled via a 3 mm fibre bundle (Black line), light coupled via a 1 mm diameter fibre with a NA of 0.39 (red line).
Image Credit: IS-Instruments
The resolving system power is given by:
R = 2Dg
Where D denotes the beam diameter (34.4 mm in this instance) and g denotes the number of lines per mm for the grating.
This clearly shows the spectrometer’s ability to resolve all of the necessary features if a 3 mm diameter fibre is used, enabling light to be collected from a major portion of the tablet’s exit side.
Sources of uncertainty
If it is assumed that the tablet surface that is examined is a true Lambertian scatterer, with a diameter of less than 3 mm and with no other illuminating sources, the quantity of light collected will be in accordance to the fibre’s surface area. When a change of the fibers takes place, for a particular signal-to-noise ratio, the resolution should be unchanged as larger apertures are utilized. To validate this a high signal photon level was selected to compare all of the measurements. This level was set at 3970 counts per pixel, or an exposure time of 2.50 seconds if the 2 mm bundle is used. The ratio of the fiber area to the active area of the 2 mm fiber bundle was used to calculate the exposure time utilized for each individual fiber to achieve the set level. Loss in collection area (47% for 2 mm bundle and 43% for 3 mm bundle) occurs as a result of the geometry of the fiber bundle.
Several sources of uncertainty were noted:
- The laser power fluctuated for several minutes by up to 5%.
- The ambient room temperature caused the DC offset on the detector counts by up to 10 counts (490 counts per pixel is the nominal value).
- The estimated background light sources were 0.3 per second.
To reduce these uncertainties, all observations were taken three times to determine the mean value. The standard deviation in the measurement was estimated at 0.15 seconds for the integration times less than 10 seconds. This value is increased to approximately 0.5 seconds for measurements in the range of 20 - 60 seconds. The background light source affects the results when small fiber cores are used, especially when the core diameter is 50 µm. In this instance, the standard deviation was 25 seconds and the integration time was 1500 seconds.
Throughput Results
Figure 8 shows the necessary integration time to observe 3970 counts per pixel for each fiber corresponding to its diameter. The black line shows the results that were observed, and the red line indicates the calculated levels.
.jpg)
Figure 8. Required integration time to observe 3970 counts per pixel as a function of fiber diameter (Red line = simulated data; Black line = Observed data).
Image Credit: IS-Instruments
The measured data is in full agreement with the calculated response. Figure 9 illustrates this further by showing the exposure time plotted in relation to the area displayed in a log/log plot. The 400 µm fiber always showed a slightly poorer performance than the expected levels. Post-inspection it was found that there was some damage to an end face of the fiber, which results in increased losses. The overall result consistency was quite good, since the observed coefficient of power was - 0.97 against the targeted -1.
.jpg)
Figure 9. Required integration time to observe 3970 counts per pixel as a function of fiber area (Red line = simulated data; Black line = Observed data).
Image Credit: IS-Instruments
This shows that during transmission Raman arrangement the tablet does behave as a Lambertian scatterer with an aperture of more than 3 mm. It also demonstrates that the spectrometer’s étendue acceptance is not less than 7.065 mm2 with an NA of 0.22 to maintain a resolution of 4 cm-1. The actual collection area is 4.027 mm2, which is the result of gaps in the fiber bundle configuration. However, the key criteria is the total synthetic area if the acceptance area is to be determined.
Stand-off Raman Observations
Figure 10 shows the experimental setup, which has been altered into a stand-off backscatter arrangement. The tablet is placed at a distance of about 70 mm from a single axis LIDAR arrangement. The diameter of the laser spot is about 2.5 mm. When the telescope is focused at the surface of the tablet, the scatter from the tablet’s surface is mostly Lambertian from the 2.5 mm diameter (approximately 5 mm2 area). However, the laser’s intensity profile can cause some amount of Gaussian bias. This configuration has a higher signal strength compared to the transmission arrangement.
Stand-off Raman Observations
Figure 10 shows the experimental setup, which has been altered into a stand-off backscatter arrangement. The tablet is placed at a distance of about 70 mm from a single axis LIDAR arrangement. The diameter of the laser spot is about 2.5 mm. When the telescope is focused at the surface of the tablet, the scatter from the tablet’s surface is mostly Lambertian from the 2.5 mm diameter (approximately 5 mm2 area). However, the laser’s intensity profile can cause some amount of Gaussian bias. This configuration has a higher signal strength compared to the transmission arrangement.
.jpg)
Figure 10. Stand-off Raman setup.
Image Credit: IS-Instruments
To ensure that the exposure times are similar to those of the transmission arrangement, the target count levels were fixed at 20000. Figure 11 shows the resulting exposure for various fibers.
.jpg)
Figure 11. Required integration time to observe 20000 counts per pixel as a function of fiber area when in a stand-off configuration (Red dots = observed data; Blue dots = simulated data).
Image Credit: IS-Instruments
As predicted the theoretical and experimental results match with each other to a large extent. However the transmission arrangement has a stronger correlation, the determined coefficient of power across the range is- 0.946. If the fiber diameter is more than 1 mm, the gains observed are a little less than those from the fiber area. This is to be expected giventhe nature of scattering surface and the light profile (Gaussian) striking the tablet.
In the standoff arrangement the light is Lambertian, and it is scattered from a diameter of approximately 1.5 mm, which is less than the 2.5 mm diameter predicted by eye, due to the spot’s intensity profile. The target’s étendue is the same or slightly less than the spectrometer. So for this set up, a standard HES spectrometer with a 1 mm fiber input is acceptable for the target examination. However, it is difficult to keep a small spot size when there is an increase in the distance to the target. Since the signal strength is α 1/R2 (R – distance to the target), any signal loss is a cause for concern.
Comparison with a Dispersive Spectrometer
The Czerny-Turner spectrometer uses an uncooled CCD, and the increased signal return in the stand-off arrangement allows the spectrometer’s performance to be analyzed and compared with a HES 2003 device. The slit in the spectrometer measures 0.2 mm x 1 mm. Figure 12 displays the spectrum obtained with a 1.55 seconds integration time, when the 0.91 mm fiber is utilized to couple the light to the system. For comparison, Figure 12 also shows the return from the HES spectrometer with the data normalized for the display.
.jpg)
Figure 12. Czerny Turner return with 0.91 mm fiber, when operating in a stand-off configuration (blue line), the equivalent HES 2003 response when using a 0.91 mm diameter fiber (red line).
Image Credit: IS-Instruments
Figure 12 shows that the detailed information on the paracetamol spectrum is not available, although the spectrum’s general shape can be observed. The loss of information is caused by the reduced resolution if a 200 µm slit is used. A 0.05 mm fiber was utilized to increase the resolution of the Czerny Turner spectrometer. Figure 13 shows the spectrum obtained, overlaid with the corresponding HES measurements using a 50 µm core fiber. The spectra is clear in approximately 10 seconds when the HES instrument is used, and to realize 20000 counts per pixel, a time period of 349 seconds was needed.
.jpg)
Figure 13. Paracetamol spectra from stand-off configuration: Blue line = HES spectrometer return with 0.05 mm fiber; grey line = Czerny Turner return with 0.05 mm fiber.
Image Credit: IS-Instruments
Despite the improved resolution of the Czerny Turner spectrometer, the signal-to-noise ratio is low to resolve the feature in a clear manner. Only a maximum integration time of 20 seconds is allowed by the detector, restricting the quantity of the collected light.
Summary and Conclusions
The étendue benefit of static Fourier transform spectrometer has been experimentally shown. In particular, the benefits of utilizing this type of spectrometer to conduct transmission Raman observations has been demonstrated by measurements using a 3 mm diameter fiber bundle with no resolution loss. As far as the authors know this is the biggest system of this type that is available, and signifies a positive forward technological step.
The experiment also demonstrated that the light exiting a tablet has a homogeneous nature when examined in transmission, proving that Raman light originates from the entire tablet volume and not from a spot on the surface. The stand-off results demonstrate that the gains are similar when using this type of configuration to make observations. However, it is vital that target étendue is more than that of the instrument to realize the full benefits of the system.
The demonstration proves that the HES range has benefits regarding measurement speed or accuracy when carrying out QA measurements in the pharmaceutical field. It is perfect to carry out out bulk sample measurements (transmission Raman), and for stand-off measurements when accessibility is an issue with respect to the target sample. This could include process sector applications. In terms of throughput, the total gains realized by the instrument when transmission Raman experiments are performed is more than 500 when compared to a classic dispersive system.
This information has been sourced, reviewed and adapted from materials provided by IS-Instruments Ltd.
For more information on this source, please visit IS-Instruments Ltd.