Fluorescence Rejection by Shifted Excitation Raman Difference Spectroscopy at MultipleWavelengths for the Investigation of Biological Samples

Shifted excitation Raman difference spectroscopy (SERDS) was applied for an effective fluorescence removal in the Raman spectra of meat, fat, connective tissue, and bone from pork and beef. As excitation light sources, microsystem diode lasers emitting at 783 nm, 671 nm, and 488 nm each incorporating two slightly shifted excitation wavelengths with a spectral difference of about 10 cm−1 necessary for SERDS operation were used. The moderate fluorescence interference for 783 nm excitation as well as the increased background level at 671 nm was efficiently rejected using SERDS resulting in a straight horizontal baseline. This allows for identification of all characteristic Raman signals including weak bands which are clearly visible and overlapping signals that are resolved in the SERDS spectra. At 488 nm excitation, the spectra contain an overwhelming fluorescence interference masking nearly all Raman signals of the probed tissue samples. However, the essentially background-free SERDS spectra enable determining the majority of characteristic Raman bands of the samples under investigation. Furthermore, 488 nm excitation reveals prominent carotenoid signals enhanced due to resonance Raman scattering which are present in the beef samples but absent in pork tissue enabling a rapid meat species differentiation.


Introduction
Due to its fingerprinting characteristics, Raman spectroscopy is well suited for the investigation of biological material, for example, for rapid and nondestructive identification purposes.Here, excitation wavelengths in the visible or nearinfrared range are preferable to avoid strong absorption of water leading to sample heating [1].For that reason, in situ Raman investigations are possible since no sample pretreatments, as for example, drying procedures, are necessary.According to the λ −4 -dependence of the Raman scattering intensity, the application of shorter excitation wavelengths can significantly improve the spectral quality.On the other hand, this leads to an increased fluorescence interference partly or completely obscuring the Raman signals and thus making the detection of useful spectra hardly possible.
There exist certain methods to remove the fluorescence background from the Raman spectra to overcome the fluorescence issue.In that way, mathematical approaches as polynomial [2] or advanced subtraction methods [3] as well as experimental techniques using a temporal discrimination of the slower fluorescence emission against the Raman photons [4,5] were successfully applied.As a technique additionally removing the fixed-pattern noise arising from CCD detectors, shifted excitation Raman difference spectroscopy (SERDS) [6] offers a promising alternative.SERDS application requires excitation laser sources with two narrow emission lines spectrally separated in the order of the full width at half maximum of the Raman signals under investigation [7].In that way, detection of two Raman spectra with these slightly different excitation wavelengths leads to a small shift in the Raman band positions, while the broadband fluorescence background remains essentially unchanged.Subsequent subtraction of the spectra eliminates the fluorescence contribution and the detector fixed pattern noise, whereas the characteristic Raman signals are preserved in the resulting derivative-like spectrum.A reconstruction by means of suitable algorithms, for example, using integration, produces a basically background-free SERDS spectrum revealing also weak Raman bands previously masked by fluorescence.
This has been demonstrated in the NIR spectral region by da Silva Martins et al. [8] applying a mechanically tunable laser in Littrow configuration for the investigation of human skin and tooth.However, to ensure highly reproducible and reliable operation, devices without moving parts would be preferable.Here, microsystem diode lasers with target-specific adjustable spectral properties according to the measurement task are well suited.Furthermore, their low power consumption as well as their compact size enables integration into portable systems suitable for SERDS application.
This paper presents SERDS investigations using microsystem diode lasers with two slightly shifted emission lines at 783 nm, 671 nm, and 488 nm as excitation light sources.The lasers emitting in the near-infrared and red spectral region were integrated into compact Raman measurement heads which are connected to laboratory (783 nm) as well as miniature spectrometers (671 nm), while the 488 nm light source is part of a laboratory setup.To test the performance of the SERDS technique, we chose meat, fat tissue, connective tissue, and bones from pork and beef as sample material to realize a variety of optical properties and fluorescence background levels.

SERDS Setup for 783 nm Excitation.
To realize a SERDS measurement head for near-infrared excitation, an in-house developed and tested Raman optical bench [9] was combined with a 783 nm distributed feedback (DFB) diode laser [10,11].By variation of the laser injection current from 110 mA to 260 mA, two operation points necessary to perform SERDS at λ 1 = 782.65 nm and λ 2 = 783.15nm with a spectral shift of 0.5 nm (7 cm −1 ) were realized.The laser temperature of 25 • C as well as the injection current was controlled using a diode laser controller DC 100 (Toptica Photonics).Inside the Raman optical bench, the DFB diode laser beam is spectrally cleaned by a bandpass filter (AHF analysentechnik AG) and guided by dielectric mirrors and a Raman edge filter (Semrock, Inc.) to a lens with a focal length of 10 mm which focuses the excitation radiation through a quartz glass window onto the biological sample.All mirrors, lenses, as well as the quartz glass window were purchased from Thorlabs GmbH.The 180 • backscattered radiation from the sample is collected by the above-mentioned focusing lens and passes a Raman edge filter to reject the Rayleigh as well as anti-Stokes components.Only the Raman Stokes scattering is transmitted and subsequently focused into an optical fiber with a core diameter of 100 μm by means of a lens with a focal length of 16 mm.A more detailed description of the concept of the Raman optical bench can be found in [12].
The SERDS measurement head is coupled to a PI 320 spectrometer (Princeton instruments) equipped with a  Here, the excitation light source contains two separate laser cavities each comprising a volume Bragg grating as frequency selective element [42].Thus, two slightly different emission wavelengths at λ 1 = 671.0nm and λ 2 = 671.6 nm with a spectral shift of 0.6 nm (12 cm −1 ) are obtained.Two LDC 210C benchtop LD current controllers (Thorlabs GmbH) serve to control the injection current of 400 mA for both laser cavities, while the common heat sink temperature of 25  (11), Raman spectra of the meat component samples were recorded with laser powers of 20 mW (487.61nm) and 26 mW (487.91 nm) at the sample position.The integration times amount to 4 s except for bone from beef which exhibits an extremely high fluorescence enabling an integration time of only 1 s.

Meat Components and Data Analysis.
To obtain meat, fat, connective tissue, and bones from pork and beef, fresh meat slices including these components were bought in a local supermarket and investigated at the day of purchase.The biological tissue of interest was cut from the slices during sample preparation using a knife and transferred into Petri dishes to prevent it from drying.Until investigation, all sample material was stored in a laboratory refrigerator (Spezial-468, Philipp Kirsch, Germany) at 5 • C.During the Raman measurement, each type of tissue was probed at 15 different positions and 10 spectra were averaged each.From the two slightly shifted Raman spectra of each measurement spot, SERDS spectra were calculated applying an in-house developed reconstruction procedure [44].

Results and Discussion
3.1.Meat from Pork and Beef. Figure 2 displays a comparative overview of Raman and SERDS spectra of meat from pork and beef applying excitation wavelengths of 783 nm, 671 nm, and 488 nm.As expected, the fluorescence level increases with decreasing excitation wavelength as can be seen at the different scales for the Raman intensity.Thus, for 783 nm excitation (Figure 2(a)), there is only a moderate fluorescence contribution.However, also here the advantage of SERDS becomes obvious by effectively removing the background and leading to a straight horizontal baseline.In that way, normalization procedures on the net intensity of selected Raman bands often used for spectra comparison can be performed much easier.Numerous characteristic Raman signals of the meat protein structure can be identified  [16,18] in the 783 nm SERDS spectra.Furthermore, weak signals as the tryptophan band at 757 cm −1 are clearly visible in the SERDS spectra and broad overlapping signals in the Raman spectra as the CH bending mode at 1341 cm −1 and the amide III band at 1304 cm −1 are resolved applying SERDS.An overview of the identified meat protein signals and their corresponding vibrational assignment according to the literature [15][16][17][18][19]22] is presented in Table 1.The broad signal in the Raman spectra between 400 cm −1 and 500 cm −1 is due to the quartz window of the Raman measurement head.Here, it is remarkable that this quite broad band is reduced to a sharp peak around 490 cm −1 in the SERDS spectra which can be used for intensity normalization.
Applying an excitation wavelength of 671 nm results in a strongly increased fluorescence interference which becomes obvious in particular for beef as illustrated in Figure 2(b).However, the SERDS technique enables for an efficient background removal and all protein Raman bands which were present for 783 nm excitation can be identified also here.Furthermore, overlapping signals are resolved and weak bands can be clearly recognized.Figure 2(c) displays the spectra from meat from pork and beef acquired with an excitation wavelength of 488 nm revealing an overwhelming fluorescence background masking all Raman signals except the major carotenoid bands in the case of beef.Nevertheless, SERDS allows for an efficient background removal and enables the identification of various Raman bands.In contrast to an excitation at 785 nm or 671 nm, the 488 nm SERDS spectra of beef contain characteristic carotenoid signals, marked by asterisks in Figure 2(c), which are enhanced by resonance Raman scattering.This comprises strong bands at 1002 cm −1 , 1154 cm −1 , and 1520 cm −1 [21] as well as minor signals at 960 cm −1 , 1188 cm −1 , and 1210 cm −1 [20].In the pork meat spectra, these signals are completely absent enabling for a distinction between these two species.Furthermore, the tyrosine doublet at 826 cm −1 and 852 cm −1 , the 1002 cm −1 phenylalanine band (overlapping with a carotenoid signal for beef), and a CH bending vibration at 1447 cm −1 can be recognized in the SERDS spectra of beef and pork.The bands at 671 cm −1 and 1351 cm −1 which have higher signal intensities for beef than for pork might be attributed to myoglobin signals [13,14], which are enhanced by the resonance Raman effect.

Fat from Pork and Beef.
The recorded spectra for fat from pork and beef are presented in Figure 3. Excitation at 783 nm results in a moderate fluorescence interference, whereas the 671 nm spectra exhibit a significantly increased background for beef.However, application of SERDS leads to essentially background-free spectra in both cases clearly displaying all characteristic bands of fat tissue [23][24][25][26][27][28][29], which, are summarized in Table 2. Additionally, overlapping signals in the Raman spectra, for example, in the range between 800 cm −1 and 900 cm −1 , are clearly separated and resolved in the SERDS spectra.
Compared to meat, fat has a higher Raman scattering intensity and thus despite of the high fluorescence level even in the Raman spectra excited at 488 nm some strong lipid signals can be recognized (see Figure 3(c)).However, the SERDS technique allows for efficient background removal and reveals nearly all fat signals that were also present at excitation at 785 nm and 671 nm.Additionally, the carotenoid bands mentioned in Section 3.1 are present in fat from beef but absent in the pork samples.In that context, the improved resolution in the SERDS spectra can exemplarily be seen at the carotenoid signals located at 1153 cm −1 , 1188 cm −1 , and 1211 cm −1 , which are overlapping in the Raman spectra but are clearly separated using SERDS.In combination with the straight horizontal baseline this furthermore offers the potential to simplify quantitative carotenoid analysis in biological tissue.

Connective Tissue from Pork and Beef.
Connective tissue possesses a complex Raman spectrum comprising a large number of collagen signals [30][31][32][33][34]36] displayed in Figure 4  and summarized in Table 3.Here, the excitation wavelengths 783 nm and 671 nm cause only moderate fluorescence interference which allows revealing at least the major bands in the Raman spectra.Nevertheless, this type of biological sample is well suited to demonstrate the potential of SERDS because the Raman signature contains several weak and overlapping bands.In the essentially background-free SERDS spectra, this overlap issue is overcome since these signals are well separated and resolved as can be seen exemplarily in the ranges of 800-1000 cm −1 and 1200-1400 cm −1 .In that way, SERDS has a great potential for the investigation of complex biological samples with numerous and partly overlapping signals.
As in the case of meat, the Raman spectra of connective tissue excited at 488 nm presented in Figure 4(c) exhibit a very strong fluorescence influence making it impossible to determine any Raman signals except the resonance-enhanced carotenoid bands at 1003 cm −1 , 1154 cm −1 , and 1521 cm −1 for beef.The SERDS spectra allow for identification of various prominent collagen bands, for example, the 1002 cm −1 phenylalanine signal (overlapping with a carotenoid band for beef), the amide III band at 1243 cm −1 and 1270 cm −1 as well as CH 2 bending vibrations at 1374 cm −1 and 1448 cm −1 .

Bone from Pork and
Beef.The Raman and SERDS spectra of bone from pork and beef are displayed in Figure 5.It is remarkable that even for excitation at 783 nm the bone from beef exhibits a significantly increased fluorescence background compared to meat, fat, and connective tissue.
Here, the SERDS spectra allow for identification also of weak Raman bands like the phenylalanine signal at 1004 cm −1 .
In addition, the SERDS technique demonstrates its ability to resolve overlapping signals exemplarily shown for the symmetric bending vibration of the phosphate group at 432 cm −1 .This band is overlapping with the broad quartz signal between 400 cm −1 and 500 cm −1 in the Raman spectra but can be clearly recognized in the SERDS spectra.The determined characteristic bone signals and their corresponding vibrational assignment using the literature [35,[37][38][39][40][41] are given in Table 4.However, several weak signals in the SERDS spectra were not mentioned in the previous works but could be attributed to the collagen matrix using the references mentioned in Section 3.3.Using an excitation wavelength of 671 nm, the fluorescence interference is even more pronounced (see Figure 5(b)).However, the application of SERDS serves for an efficient background removal and enables obtaining essentially the same spectral pattern as for 783 nm excitation exhibiting all major and minor Raman signals.
The bone Raman and SERDS spectra excited at 488 nm are presented in Figure 5(c).Due to the overwhelming fluorescence background only the intense symmetric stretching vibration of the phosphate group at 957 cm −1 can be identified in the Raman spectra.Nevertheless, minor characteristic bone signals as the asymmetric bending (581 cm −1 ) and stretching modes (1067 cm −1 ) of the phosphate group are present in the SERDS spectra.Additionally, the amide III band at 1241 cm −1 as well as the CH 2 bending vibration at 1447 cm −1 of the collagen component in bone can be recognized.Resonance-enhanced carotenoid signals which were already found in meat, fat, and connective tissue from beef are also contained in bone from beef.However, only the major bands at 1001 cm −1 , 1155 cm −1 , and 1518 cm −1 are clearly visible.

Conclusion
In this paper, we applied shifted excitation Raman difference spectroscopy (SERDS) at different excitation wavelengths to demonstrate the potential for an efficient rejection of the fluorescence background in the Raman spectra of selected meat components.Microsystem diode lasers emitting at 783 nm, 671 nm, and 488 nm were used as excitation light sources.Each laser device was able to emit at two slightly different wavelengths with a spectral separation of about  10 cm −1 necessary to perform SERDS.To realize a variety of biological samples, meat, fat, connective tissue, and bone from pork and beef were selected for our investigations.
With an excitation wavelength of 783 nm, the fluorescence interference was moderate.However, the SERDS technique effectively removed the signal background resulting in a straight horizontal baseline.Furthermore, weak bands in the Raman spectra become clearly visible in the SERDS spectra, and also overlapping signals are separated and resolved applying the SERDS technique.
An excitation at 671 nm results in an increased fluorescence background which becomes obvious in particular for the beef samples.Despite of this fluorescence issue, SERDS allows for efficient background removal enabling the identification of essentially the same spectral patterns of the samples under investigation as for 783 nm excitation.
As expected, the Raman spectra excited at 488 nm exhibit an overwhelming signal background masking nearly all Raman signals of the probed tissue except in the case of fat.Application of SERDS allows determining not all, but a majority of the signals which were present in the spectra excited at 783 nm and 671 nm.In addition, the tissue samples from beef reveal prominent signals of carotenoids which are enhanced by means of the resonance Raman effect.These bands are not present in samples from pork and thus allow for a rapid meat species distinction.
The presented study demonstrated the ability of the SERDS technique for an effective fluorescence background removal of selected biological material.Furthermore, weak bands are enhanced, and overlapping signals are resolved allowing for improvement of spectral quality.In that way, SERDS has a great potential for numerous other analytical Raman applications which were limited by the fluorescence issue up to now, for example, investigation of various natural compounds, medical diagnostics, and forensic usage.

Figure 2 :
Figure 2: Comparison of Raman (upper traces) and SERDS spectra (lower traces) of meat from pork (red curves) and beef (green curves) excited at (a) 783 nm, (b) 671 nm, and (c) 488 nm.Each curve represents the average from 15 measurement positions.Resonance-enhanced carotenoid signals in (c) are indicated by asterisks.

Figure 3 :
Figure 3: Comparison of Raman (upper traces) and SERDS spectra (lower traces) of fat from pork (red curves) and beef (green curves) excited at (a) 783 nm, (b) 671 nm, and (c) 488 nm.Each curve represents the average from 15 measurement positions.

Figure 4 :
Figure 4: Comparison of Raman (upper traces) and SERDS spectra (lower traces) of connective tissue from pork (red curves) and beef (green curves) excited at (a) 783 nm, (b) 671 nm, and (c) 488 nm.Each curve represents the average from 15 measurement positions.

Figure 5 :
Figure 5: Comparison of Raman (upper traces) and SERDS spectra (lower traces) of bone from pork (red curves) and beef (green curves) excited at (a) 783 nm, (b) 671 nm, and (c) 488 nm.Each curve represents the average from 15 measurement positions.
Raman edge filter only transmitting the Stokes scattering while removing the Rayleigh as well as the anti- (9)kes components.A subsequent spatial filter comprising two lenses(6)with focal lengths of 50 mm and 25 mm as well as a pinhole(8)with a diameter of 600 μm is applied to reject unwanted stray light.By means of a lens with a focal length of 50 mm, the Raman Stokes radiation is then focused onto an iHR320 spectrometer (Horiba Jobin Yvon)(9)and detected by a back-illuminated CCD detector (CCD-2048x512-BIUV-1LS) (10) cooled down to −130 • C by liquid nitrogen.Applying a computer running Labspec software (Horiba Jobin Yvon)

Table 1 :
Raman bands identified in the SERDS spectra of meat from beef and pork and their vibrational assignment, Phe: Phenylalanine, Trp: Tryptophan, Tyr: Tyrosine.

Table 2 :
Raman bands identified in the SERDS spectra of fat tissue from beef and pork and their vibrational assignment.

Table 3 :
Raman bands identified in the SERDS spectra of connective tissue from beef and pork and their vibrational assignment, Hyp: Hydroxyproline, Pro: Proline, Phe: Phenylalanine, Trp: Tryptophan, Tyr: Tyrosine.

Table 4 :
Raman bands identified in the SERDS spectra of bone from beef and pork and their vibrational assignment, Hyp: Hydroxyproline, Pro: Proline, Phe: Phenylalanine, Tyr: Tyrosine.