Spectrometry Problem Solutions

Problem 1: Spectrometer Design (1800–2000 cm⁻¹, resolution < 1

cm⁻¹)
• (a) Spectral resolution: For the grating spectrometer (600 lines/mm, blaze ~5 μm), we estimate
resolution from the linear dispersion and pixel width. With 2048×25 μm = 51.2 mm detector span
covering ~700 nm (5.556–5.000 μm = 1800–2000 cm⁻¹), the linear dispersion is ~0.032 mm/nm
(≈0.34 nm per 14 μm pixel). This corresponds to about 0.0667 cm⁻¹ per pixel near 2000 cm⁻¹
(≈0.133 cm⁻¹ for two-pixel FWHM), well below 1 cm⁻¹. Thus the grating system achieves ~0.1–0.2 cm⁻¹
resolution. For the FTIR interferometer, the resolution (in cm⁻¹) is given by Δσ ≈1/(max OPD). With
OPD_max = 2 cm, Δσ ≈0.5 cm⁻¹ 1 . In practice one can say ~0.5 cm⁻¹ resolution for the FTIR at full
OPD. Both systems therefore meet the <1 cm⁻¹ requirement.

• (b) Spectral range per scan: The grating + array covers a limited band in one shot. Numerically,
2048 pixels×0.34 nm/pixel ≈700 nm total, which in wavenumbers (at ~5 μm center) is roughly 135–
140 cm⁻¹ of coverage (in our setup ~1935–2070 cm⁻¹). In other words, one detector frame might cover
about 1930–2070 cm⁻¹. (If needed, one could adjust incidence to center on 1900 cm⁻¹.) By contrast, an
FTIR interferogram contains all wavelengths simultaneously (subject only to source and beamsplitter
cutoffs). In one FTIR scan the entire 1800–2000 cm⁻¹ band is obtained (and in fact the FTIR can cover
a broad mid-IR range with one scan). This “multiplex” capture of all wavelengths at once is a key FTIR
advantage 2 .

• (c) FTIR scan time: The mirror must move ~1 cm to achieve OPD = 2 cm (since OPD = 2×mirror
displacement). At 1 mm/s mirror speed, that takes ~10 s one-way. (A symmetric scan from –1 cm to
+1 cm would be 20 s total, but a single-direction 0→2 cm is ~10 s.) Hence each interferogram requires
on the order of 10 s of acquisition.

• (d) Suitability for real-time monitoring: The grating-based system can acquire a spectrum in a
single detector readout (subject to integration time), giving essentially “instantaneous” spectral
snapshots. In contrast, the FTIR must scan the moving mirror (~10 s) to record one full spectrum.
Moreover, dispersive (grating) spectrometers with arrays inherently measure all wavelengths at once
once the grating is set, whereas traditional dispersive systems had to tune to each wavelength. As
Bruker notes, dispersive IR requires scanning each λ, while FTIR measures all λ simultaneously 2 .
Here, however, the FTIR’s multiplex advantage is offset by its slow mechanical scan. For “real-
time” (sub-second) monitoring, the grating with a fast CCD readout is preferable: it can repeatedly
integrate and read out much faster than the ~10 s needed by the FTIR. (The grating system even
exceeds the required resolution, providing ~0.1 cm⁻¹, and can focus on just the NO/NO₂ lines.) In
summary, a dispersive grating + array spectrometer offers higher throughput and faster acquisition
(at the expense of covering a narrower band per shot), making it more suitable for rapid industrial
gas monitoring in this case 2 .

1
Summary (Problem 1): The FTIR achieves ~0.5 cm⁻¹ resolution (with OPD=2 cm), whereas the grating+array
design yields even higher resolving power (~0.13 cm⁻¹). A 2048×25 μm array covers on the order of 135–
140 cm⁻¹ per frame (sufficient to span 1800–2000 cm⁻¹). An FTIR scan (0→2 cm OPD) takes ≈10 s, so one
spectrum is obtained in ~10 s. In contrast, the grating spectrometer (once aligned) can record spectra in
much shorter times (determined by integration). Thus for real-time gas monitoring, the grating
spectrometer is generally preferred due to its faster effective acquisition, even though FTIR easily meets the
resolution requirement and covers broad bands in one scan 2 1 .

Problem 2: FTIR of Atmospheric CO (2143 cm⁻¹, 16 scans, boxcar)

• (a) Effective resolution: Each single scan has OPD up to ~1.6 cm (mirror travel 0.8 cm ×2). Thus
nominal resolution is Δσ = 1/(OPD) ≈ 1/1.6 = 0.625 cm⁻¹. (Boxcar apodization imposes essentially no
additional broadening, so the effective resolution remains ≈0.63 cm⁻¹.)

• (b) Noise power in 400 cm⁻¹ band: The detector NEP is 5×10⁻¹² W/√Hz. The optical bandwidth Δν =
400 cm⁻¹ corresponds to Δf ≈ 400·(3×10¹⁰ Hz/cm⁻¹) = 1.2×10¹³ Hz. Integrated noise (RMS) = NEP ×
√(Δf) = 5×10⁻¹² W/√Hz × √(1.2×10¹³ Hz) ≈ 1.7×10⁻⁵ W. So the total noise power over the 400 cm⁻¹
band is on the order of 1×10⁻⁵ W (RMS).

• (c) Norton–Beer apodization: Boxcar (rectangular) apodization gives minimal peak broadening but
larger sidelobes and noise. Switching to a medium Norton–Beer function smooths the interferogram
edges, which reduces noise and sidelobes but at the expense of some resolution. As Thermo notes,
“strong apodization reduces more noise but can also reduce the resolution… and broaden
peaks” 3 . Norton–Beer (medium) is intermediate: compared to boxcar it will broaden each line
slightly (worse resolution) but significantly suppress sidelobes and lower high-frequency noise. Thus
SNR improves (since random noise and ringing are reduced) while the FWHM of lines increases
(resolution degrades).

• (d) More scans (64 vs 16): SNR scales as the square root of the number of averages. Going from 16
to 64 scans is a 4× increase in scans, so SNR improves by √4 = 2. In other words, at the same
integration time per scan, 64 scans will double the SNR compared to 16 scans.

Summary (Problem 2): With a 0.8 s scan (OPD≈1.6 cm), the un-apodized resolution is about 0.625 cm⁻¹. The
total detector noise power over the 400 cm⁻¹ band is ~1.7×10⁻⁵ W (RMS). Applying a medium Norton–Beer
apodization will broaden the CO line shapes but reduce noise and sidelobes 3 , thus improving SNR at the
cost of a slightly lower resolution. Finally, quadrupling the number of averaged scans (16→64) yields a 2×
increase in SNR (SNR ∝√N).

Problem 3: Compact Grating Spectrometer (400–1100 nm, ~2 nm

res)
• (a) Grating choice and dispersion: A moderate-density grating is suitable. For example, a 600 lines/
mm grating (blazed ~700–800 nm) provides good efficiency across 400–1100 nm. Its angular
dispersion is $$\frac{d\theta}{d\lambda} = \frac{m}{d\cos\theta}\approx\frac{1}{d} \approx
6\times10^5\ {\rm m^{-1}} \,,$$ or about 0.0006 rad/nm (at small angles). More precisely at

2
 λ≈750 nm one finds ≈0.00031 rad/nm. For a focal length f (say ≈100 mm), the linear dispersion is
≈0.030–0.032 mm/nm. That is ~2.2 pixels/nm (14 µm pixel), or ~0.45 nm per pixel. Thus in 2 pixels
one spans ~0.9 nm, meeting the ~2 nm requirement comfortably. (For reference, a 300 lines/mm
grating under similar optics would give ~0.57 nm/pixel 4 .)

• (b) Focal length matching: To span 400–1100 nm (Δλ=700 nm) onto 2048 pixels (28.672 mm), we
need Δλ/Δx ≈700 nm/28.672 mm ≈24.4 nm/mm. Using d/(f) = Δλ/Δx, we solve $f\approx d\,\Delta x/
\Delta\lambda$. For d = 1/600 mm and Δx = 28.672 mm, one finds $f\approx(0.001667\times28.672)/
0.0007\approx68$ mm. In practice a focal length on the order of 70–100 mm will cover the full 400–
1100 nm range on the 2048-pixel array. (Avantes notes that 900 nm range on 2048 px yields
~0.44 nm/pix 5 , consistent with our calculation of ~0.34–0.45 nm/pix for 700 nm range.)

• (c) Slit width (resolution vs. throughput): A narrower slit (or small entrance aperture) improves
spectral resolution but reduces light throughput and SNR; a wider slit increases signal but broadens
lines. The design goal is ~2 nm resolution. Given ~0.45 nm/pix dispersion (from part a), 2 nm
corresponds to ~4.4 pixels. Thus an image of the slit of ~4–6 pixels FWHM is appropriate. If the optics
are roughly 1:1, this suggests a physical slit ~60–100 µm. As a concrete choice, a slit around 50–
100 μm gives ~4–7 pixels width, yielding ~1–2 nm resolution. (For example, 84 μm = 6×14 µm would
give ~6 pixels = 2.7 nm; 100 μm →7px ≈3.2 nm; 50 μm→3.5px≈1.6 nm.) The optimal compromise is
often chosen to just meet the 2 nm spec while maximizing throughput. In general “larger apertures
yield better SNR, while smaller apertures give better resolution” 6 , so one would pick the smallest
slit that still provides acceptable light.

• (d) Light source: A broadband, stable visible/NIR source is needed. Typically a stabilized tungsten–
halogen lamp is used (e.g. 360–2600 nm output 7 ) because it produces strong, continuous
spectrum from the visible into the NIR. Alternatively, a xenon lamp could be used for more UV–blue
output, but a quartz-halogen lamp (e.g. 5–20 W) is common for reflection spectroscopy and matches
the 400–1100 nm range. Such lamps often include feedback stabilization to maintain constant
intensity 7 .

• (e) Reducing stray light / improving SNR: In a factory environment, stray light and noise are critical
issues. Optically, one should incorporate beam stops and baffles: for example, block the 0th-order
(undispersed) beam and use order-sorting filters to suppress higher/different-order light 8 . All
internal surfaces should be blackened or baffled to absorb scatter. High-quality optics (mirrors/
gratings with good coatings) also minimize diffuse scatter 8 . Mechanically, the spectrometer
should be light-tight and shielded from ambient illumination. Electronic measures (chopping the
light source and lock-in detection) can reject uncorrelated noise. Cooling the CCD lowers dark noise.
In short, careful optical design (e.g. blocking unwanted orders, use of filters, internal baffles)
combined with stable electronics and source modulation will maximize SNR and suppress stray-light
artifacts 8 .

Summary (Problem 3): A ~600 lines/mm grating (blazed for visible/NIR) and a focal length of order 70–
100 mm will map 400–1100 nm onto a 2048×14 µm CCD, yielding roughly 0.4–0.5 nm/pixel dispersion
(≈0.9 nm over 2 pixels). This exceeds the ~2 nm resolution target even with a few-pixel slit image. A slit of
order 50–100 µm width is a good compromise: smaller slit yields finer resolution but less light, as noted in
[19] 6 . A stabilized tungsten-halogen lamp covers 360–2600 nm 7 , making it ideal for broadband VIS/
NIR reflectance. To reduce stray light, one should carefully design the optics (block 0th order, remove stray

3
orders, use black baffles and anti-reflection coatings) 8 , and shield the system from ambient light. These
strategies will enhance SNR and ensure reliable tablet coating spectra in the factory setting.

Sources: Standard spectrometer theory (resolution = 1/OPD 1 , Fellgett/throughput advantages 2 ),

manufacturer/application notes 3 6 7 8 5 .

1 Fourier-transform infrared spectroscopy - Wikipedia

https://en.wikipedia.org/wiki/Fourier-transform_infrared_spectroscopy

2 Difference IR vs FTIR | Bruker

https://www.bruker.com/en/products-and-solutions/infrared-and-raman/ft-ir-routine-spectrometer/what-is-ft-ir-spectroscopy/
difference-ir-vs-ftir.html

3 6 3358_OMNIC Next_UG_Adv Meas Options.fm

https://assets.thermofisher.com/TFS-Assets/MSD/Product-Guides/335800-A-Adv-Meas-Options-touchscreen.pdf

4 5 Understanding Spectrometer Resolution Specifications - Avantes

https://avantesusa.com/understanding-spectrometer-resolution-specifications/

7 Tungsten-Halogen Light Source

https://www.thorlabs.com/newgrouppage9.cfm?objectgroup_ID=7269

8 Stray Light Causes and Reduction Gigahertz-Optik

https://www.gigahertz-optik.com/en-us/service-and-support/knowledge-base/stray-light-suppression-correction-in-
spectroradiometers/