Chu
Li
a,
Guo-Qiang
Xia
b,
An-Qing
Jiang
a,
Jiu-Chun
Ren
a,
Er-Tao
Hu
c,
Jian-Ke
Chen
a,
Qiao-Chu
Zhang
g,
Lei
Yu
g,
Osamu
Yoshie
d,
X.-D.
Xiang
b,
Hai-Bin
Zhao
a,
Yu-Xiang
Zheng
a,
Song-You
Wang
a,
Yue-Mei
Yang
a,
Wei
Wei
c,
Young-Pak
Lee
e,
Junpeng
Guo
f,
Yun-Hai
Jia
g and
Liang-Yao
Chen
*a
aDepartment of Optical Science and Engineering, Fudan University, Shanghai, China. E-mail: lychen@fudan.ac.cn
bDepartment of Material Science and Engineering, SUSTC, Shenzhen, China
cSchool of Optoelectronic Engineering, NUPT, Nanjing, China
dGraduate School of IPS, Waseda University, Fukuoka, Japan
eDepartment of Physics, Hanyang University, Seoul, Korea
fDepartment of Electrical and Computer Engineering, University of Alabama in Huntsville, Huntsville, AL 35899, USA
gCentral Iron & Steel Research Institute, Beijing, China
First published on 16th January 2024
A coma-free superhigh spectral resolution optical spectrometer was successfully designed and constructed for practical application. The spectrometer consisted of 20 subgratings and a 2D BSI-CMOS array detector. The spectra were 20-fold with the physical size of the photoelectron detection along the diffraction direction and extended by a factor of 20 to achieve a spectral resolution of 0.01 nm per pixel without any moving parts. Ultrahigh dense emission lines of a hollow cathode Fe lamp were measured. Based on the data analysis and discussions, among the highly resolved spectral lines of Fe measured with an improved signal-to-noise ratio in the 170–600 nm spectral region, 2451 lines were identified in accordance with the reported ones in the literature; however, more than 1100 lines with lower intensity remained unassigned with an unknown mechanism of their origins. Our experimental data indicated the need to more fundamentally study and explore the spectral data of the Fe atoms with the potential for broad application prospects for Fe-based materials in the future.
Transition metals generally show very rich atomic emission lines in the spectral region below 600 nm. For the Fe-group elements, thousands of lines are observed in the spectral region from the vacuum-ultraviolet (VUV) to the infrared due to the characteristic of the partially filled d-orbitals of the Fe group elements.10,11 Studies on Fe spectral lines are also very important to thoroughly explore the properties of the absorption spectra of many astrophysical objects, including stars and galaxies.12–14
In previous works, both grating spectrometers and Fourier transform spectrometers (FTS) were generally used to measure the atomic emission spectra of the transition metals that possess multiple spectral lines.15–18 Research has mainly focused on FTS to measure the spectral lines of transition metals in recent years due to its high resolution in the ultraviolet (UV) to infrared (IR) spectral regions.19 The resolution of the FTS is highly dependent on the precise control of the mechanically step-moving parts with software interpretation of the data in the system.19 The step-moving parts have difficulty to achieve very high-precision control in the short wavelength UV and visible regions, and thus, FTS measurements are usually carried out in large research laboratories with better experimental conditions. In addition, as shown in Fig. 1, Clear showed the spectra of a Ni hollow cathode lamp measured at the 217.45–218.15 nm spectral region using the FTS and the grating spectrometer in his Ph. D. thesis.20 Compared to FTS, the grating spectrometer had a lower resolution power with a broader full width half maximum (FWHM) of the spectral lines, resulting in spectral lines in some regions overlapping and not being fully resolved. In contrast, the FTS spectra could sharply distinguish the spectral lines due to its high resolution. However, some spectral lines observed clearly in the spectra measured by the grating spectrometer were entirely missing in the spectra of FTS.20 Hinkle et al. also noted that the measurements using the FTS could not provide a high dispersion solution, and therefore, the lines in the noise of the atlas were barely visible.21
![]() | ||
Fig. 1 FTS and Grating spectra of a nickel-helium hollow cathode lamp. The inset is a partially enlarged schematic of the FTS spectrum.20 |
The studies of the electron orbital structures of the Fe atom will not only help to understand the mechanism of iron-based superconductors and other materials but are also very important to explore the properties of the absorption spectra of many astrophysical objects, including stars and galaxies.1–5,10–14 To date, the optical spectra with respect to the energy transitions between different electron orbits of the Fe atom have been most extensively measured and studied in broad spectral regions by using different experimental approaches.15–19 A total of approximately 37000 spectral lines are reported in the literature, but among them, the origins of approximately 11
500 lines, equal to approximately one-third of the total lines, are still unidentified.19
The grating spectrometer plays a significant role in acquiring rich photonic information of materials in modern science and technology, such as that in the material science, biotech and medical science fields.22–24 In the early stage of spectrometer development, mechanical control of the grating azimuthal rotation was used to change the grating diffraction angle to achieve wavelength selection with output of monochromatic light. Due to the relatively poorer long-term stability of the mechanically rotating optical elements to be operated in the broad wavelength region, it was difficult to meet the demand of high data acquisition speed in the spectral measurement. Therefore, instead of rotating planar gratings, in the development of modern optical analysis systems, integrated grating modules were used to simultaneously obtain spectral imaging consisting of multiple subwavelength regions without any mechanical moving parts in the application.25,26
These integrated modules greatly improved the measurement speed. However, as shown in Fig. 2, in typical Czerny–Turner (C–T) grating spectrometers, the off-axis optical configuration leads to image aberrations along the optical path of the system that seriously affect the imaging quality over the entire spectral range. In addition, as Gillian Nave et al. reported, to achieve high resolution in the experiment, the C–T configuration of the spectrometer required a very long focal length of 10.7 m in the system design.19 This also has led to the stagnation of research on the measurement of high-density spectral lines by grating spectrometers. To solve this problem, in this work, we designed and constructed a new type of coma-free grating spectrometer with superhigh resolution to measure the atomic emission spectra of a hollow cathode Fe lamp in the 170–600 nm spectral region.
![]() | (1) |
Different approaches have been applied to reduce coma aberrations. One approach uses a grating with a smaller size of W; however, this causes the side effect of decreasing the optical aperture, resulting in a lower number of grating lines and reducing the spectral resolution and sensitivity of the system. By analyzing eqn (1), we can use another approach that satisfies the condition in which the coma aberration Δ = 0; this results in a restricted condition as shown in eqn (2):
![]() | (2) |
The incidence angle θi and diffraction angle θd are all correlated to be functions of wavelength λ and grating constants d, as shown by eqn (3):
![]() | (3) |
In previous work, in terms of using a two-dimensional (2D) back-side-illumination (BSI) CCD camera, we fulfilled the goal of designing a 44-image-folded grating spectrometer in a prototype formation consisting of three groups of gratings with groove densities of 1200 lines per mm, 1800 lines per mm, and 3600 lines per mm working in spectral regions of 600–1000 nm, 300–600 nm, and 200–300 nm, respectively.28
In this study, by taking advantage of a coma-free spectrometer with ultrahigh resolution in a broad spectral region, a new spectrometer consisting of 20 integrated gratings with grating constants g of 1800 lines per mm and a 2D BSI-CMOS array detector was designed and constructed for practical application. The CMOS detector had a pixel size of 11 × 11 μm2 and a pixel density of 2048 × 2048 with high-dynamical 16 bit-AD conversion performed in situ on the chip to achieve a high data acquisition speed of approximately 25 spectra per second. A special MgF2 transparent window was used for the CMOS array detector to be suitable to measure the data down to the wavelength of 170 nm with a peak efficiency of approximately 90% at 400 nm.
In the system design, the entire spectral region of 170–600 nm was divided into 20 subspectral regions, corresponding to 20 subgratings that were vertically distributed along the direction of the incidence slit and set at specific diffraction angles to form an integrated grating module Gx. The diffraction image of the subspectral regions could then be precisely imaged on the focal plane of the 2D BSI-CMOS detector. The size of the subgrating was 70 mm × 5.7 mm along the diffraction and slit directions, respectively, with the specific blaze wavelength λb selected and arranged in accordance with the system design to achieve the best performance of diffraction efficiency in the spectral region.
This configuration without any mechanical moving parts achieved three key functions in one spectrometer: (1) wide spectral working range, (2) high spectral resolution, and (3) high data acquisition speed.
In the test, the spectra of a dual-element Hg–Ar lamp and a hollow cathode Mn lamp were used to calibrate the spectrometer in the 170–600 nm spectral region. After, the atomic radiation from a hollow cathode Fe lamp emitting the Fe-I and F-II spectral lines was measured at room-temperature and 20 mA operation current.
Grating no. | Filter wavelength range (nm) | Design wavelength range (nm) | Center wavelength (nm) | Working wavelength range (nm) | Redundant wavelength range (nm) | Diffraction angle (Deg) |
---|---|---|---|---|---|---|
1 | 350 | 579–600 | 589.5 | 21 | 22.81 | 22.60 |
2 | 350 | 558–579 | 568.5 | 21 | 23.02 | 21.30 |
3 | 350 | 537–558 | 547.5 | 21 | 23.21 | 20.02 |
4 | 350 | 516–537 | 526.5 | 21 | 23.40 | 18.76 |
5 | 350 | 495–516 | 505.5 | 21 | 23.56 | 17.81 |
6 | — | 258–280 | 269 | 22 | 24.64 | 4.23 |
7 | — | 236–258 | 247 | 22 | 24.67 | 3.05 |
8 | — | 214–236 | 225 | 22 | 24.70 | 1.87 |
9 | — | 192–214 | 203 | 22 | 24.71 | 0.69 |
10 | — | 170–192 | 181 | 22 | 24.71 | −0.48 |
11 | — | 280–302 | 291 | 22 | 24.60 | 5.42 |
12 | — | 302–324 | 313 | 22 | 24.54 | 6.62 |
13 | — | 324–346 | 335 | 22 | 24.48 | 7.83 |
14 | — | 346–368 | 357 | 22 | 24.40 | 9.04 |
15 | — | 368–390 | 379 | 22 | 24.31 | 10.27 |
16 | 350 | 390–411 | 400.5 | 21 | 24.21 | 11.47 |
17 | 350 | 411–432 | 421.5 | 21 | 24.11 | 12.66 |
18 | 350 | 432–452 | 442.5 | 21 | 23.99 | 13.85 |
19 | 350 | 452–474 | 463.5 | 21 | 23.86 | 15.06 |
20 | 350 | 474–495 | 484.5 | 21 | 23.72 | 16.28 |
Compared to the former work in designing the coma-free grating spectrometer, some improvements were made. Instead of using the CCD array detector, the 2D BSI-CMOS array detector was used to achieve high-speed data acquisition in our experiment. A set of filters was placed in front of the grating module Gx to filter the high-order (m ≥ 2) diffracted light above the wavelength of 350 nm in the 170–600 nm wavelength region. As shown in Fig. 4, one end of the lined-fiber array with the slit (10 μm) was extruded out of the small hole located at the center of the grating module Gx to avoid blocking of the light propagating along the optical path. Instead of using the toroidal mirror configuration, a new spherical mirror module M2 consisting of 20 submirrors with the same focal length of 500 mm was designed and constructed. This will keep the surface shape of mirror M2 and add extra flexibility in optical path alignment related to the grating module Gx consisting of 20 subgratings, and thus to make all 20 sub-spectra be precisely imaged on the focal plane of the 2D-CMOS array detector. In the spectral data reduction, the image Ip_background of the background noise of each pixel was measured in advance and then subtracted from the image Ip_image of the normal signal data of each pixel to obtain the real image Ir_p_image of the spectrum by suppressing the background noise more effectively in the entire spectral region according to eqn (4).
Ir_p_image = Ip_image − Ip_background | (4) |
![]() | ||
Fig. 4 Internal structure of the input light coupling through the lined fiber array in the system: (a) 3D view and (b) Front view. |
Therefore, without any mechanical movement parts, the new spectrometer had the advantages of compact size, good data repeatability, high reliability and long working life. It could be widely used in the fields of real-time study of spectral properties of high-performance optoelectronic materials and devices, as well as in the field of optical communication in convenience.
The 20 gratings were arranged to form a grating module. The incidence angle θλ of each grating with respect to the wavelength and system constant q (q = 10°) was fixed and determined by eqn (5) with the specific relevant parameters shown in Table 1.
![]() | (5) |
The maximum spectral resolution corresponding to the two adjacent pixels could be resolved according to eqn (6).
![]() | (6) |
To reduce the temperature fluctuation that will affect the resolution in the calibration and working conditions, the temperature was monitored and fixed at 24 ± 0.5 °C in the normal steel-wall-plate-sealed lab clean room condition. Thus, the reduced effect of the refractive index of air changing with temperature and humidity in the room was omitted.29
When the grating density of 1800 lines per mm was applied in the 170–600 nm wavelength region, the calculated Δλmax values were approximately equal to 0.012 nm, which was equivalent to Δλmax per pixel = 0.012 nm per pixel for the spectral resolution with respect to the theoretical prediction of the system.
To evaluate the spectral resolution of the spectrometer, the spectral lines of the Hg–Ar lamp with relatively fine and clean characteristics at 253.653 nm, 312.564 nm, 435.833 nm and 546.074 nm are shown in Fig. 5(b–e), respectively. The full width at half maximum (FWHM) ΔλFWHM of a spectral line was measured to describe the spectral resolution. To measure the ΔλFWHM data, a spectral line generally needs to be represented by at least three pixels. The experimentally measured data of ΔλFWHM were approximately 0.028 nm, 0.026 nm, 0.026 nm and 0.026 nm for the lines of 253.658 nm, 312.564 nm, 435.833 nm and 546.074 nm, respectively; thus, the pixel resolution (ΔλFWHM/3) was uniformly better than approximately 0.01 nm per pixel in the entire spectral region and was in good agreement with the design expectation.
A spectral data density k factor can be used to characterize the high spectral performance of the spectrometer. Without any moving part, the k factor can be defined in approximation as the ratio of the full working wavelength range of the spectrometer to the total data points presented by the pixel resolution; specifically, k = (total working wavelength range Δλ)/(pixel resolution). Therefore, the k factor was approximately 4.3 × 104 in the wavelength range of 170–600 nm (Δλ = 430 nm) for the spectral data measured at one time without any moving optical elements.
In previous studies, some of the unidentified spectral lines of the Fe atoms in the measurement were also attributed to the thermal energy transformation that occurred in the Fe hollow cathode lamp, in which some inert gas, such as Ne, was partially filled to cause possible collisions between ionized Ne and Fe atoms in a process called energy transformation:16,33–36
Ne+ + Fe → Ne + Fe+* + ΔE | (7) |
In the experiment, by using a coma-free grating spectrometer with ultrahigh spectral resolution and sensitivity, we measured both the atomic spectral lines of a cathode Fe lamp partially filled by inert Ne gas and those of a pure Ne lamp in the 170–600 nm spectral range. After, the measured spectral lines of Ne as the reference were subtracted to remove the Ne spectral lines from the cathode Fe lamp from the data reduction process.
The spectral lines of the Fe atom are shown in Fig. 7(a). The Fe lamp possessed a very large number and high density of spectral lines, and a local magnification of the Fe spectrum showed that the spectrometer could clearly distinguish the multiple spectral lines in the 170–180 nm wavelength region, as seen in Fig. 7(b).
Rolf et al. reported that the final spectral data were composed of eight scans with a total scanning time of approximately one hour.37 By comparison, the entire spectral data was acquired in only 1 s with our designed coma-free grating spectrometer.
A partial enlargement of the Fe spectrum is shown in Fig. 8. The newly designed coma-free grating spectrometer greatly improved the signal-to-noise (S/N) ratio of the spectrum. All peaks with S/N ratios higher than 15 could clearly be detected by using the peak-find software.
![]() | ||
Fig. 8 Weak spectral lines of the Fe atoms with S/N ratios higher than 15 and intensities over 30 clearly detected by using peak-find software. |
The spectra of the Fe atom measured in this study were compared with the data published in the NIST database and CRC manual with an uncertainty of less than 0.015 nm.30,38 No characteristic peaks of other impurities were found in the spectra. The specific statistics of the spectral line numbers are shown in Table 2.
Atom species | Number of lines |
---|---|
Fe | 2451 |
Ne | 279 |
Unassigned | 1109 |
To date, there are approximately 37000 spectral lines of the Fe atom reported in the literature, but among them, the origins of approximately 11
500 lines, equal to approximately one-third of the total lines, are still unidentified.19 According to the data analysis in this study, approximately 2451 spectral lines were identified in accordance with those published in the literature, while more than 1100 unassigned spectral lines were present, shown in the ESI 1–3.† For the results of unassigned lines reported in the previous studies of Fe, error sources, such as the residual coma aberration not being completely eliminated in the entire spectral region, were partly the cause for the extra weak lines observed in the experiment. The spectral resolution of the spectrometer with ultrahigh precision could still be improved to resolve those unassigned fine lines with weaker intensity that could potentially be multiply overlapped in some very narrow spectral regions.
The unassigned spectral lines all had intensities below 500 with S/N ratios higher than 15; this result indicated that more theoretical studies were needed to understand the origin of the weakly excited spectral lines of Fe atoms. The unassigned spectral lines observed with higher precision due to the elimination of coma aberrations could provide strong experimental support for further theoretical studies and explorations in the future.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d3ja00342f |
This journal is © The Royal Society of Chemistry 2024 |