Aperture-Sensitive Distal Depth Scanning Raman Spectroscopy

Depth sensitive Raman spectroscopy has shown its advantages in the detection of Raman spectra for layered tissues. However, till now the depth scanning mode in Raman spectroscopy is the proximal scanning mode, which has disadvantages in significant variation of probe-sample contact and the size of the probe. Here, we developed a new Raman system with an axicon lens and a graduated ring-activated zero aperture. The focal range of the system was about 35 mm and nearly two orders of magnitude higher than that of the traditional Raman system. The Raman spectrum obtained from different materials at a fixed depth was demodulated with two Raman spectra from two adjacent depth ranges. Two and three-layer models were used to demonstrate the effectiveness of the proposed Raman system. The Raman spectra of polyvinyl chloride, polyphenylene ether and silica gel were demodulated successfully from overlapped Raman spectra. The new Raman system used the distal scanning mode and offered the following advantages. First, the Raman spectra of full range and a fixed depth can be obtained simultaneously. Second, depth scanning is performed far from the sample. Third, the method can be used in endoscopic applications to reduce the size of the probe.


I. INTRODUCTION
C OMPARED with other optical spectroscopy, Raman spectroscopy has advantages in chemical discrimination, nondestructive and label free characteristics [1], [2], [3]. The method provides the fingerprints of functional groups, and these fingerprints can be used to identify intrinsic molecular constituents [4], [5], for example, Olivo et al. developed a portable ultrawideband CRS system which used dual-wavelength excitation with a dual passband laser cleaning filter and high speed fiber array multiplexer to improve system robustness. Corneum thickness was quickly and roughly simultaneously obtained with its Raman spectra in both the fingerprint and high wavenumber regions [6]. Moreover, since Raman scattering background from water in Raman spectroscopy is weak, it is suitable for tissue measurements [7], [8], [9], [10]. For this application, especially for endoscopic measurements, step motor is not suitable for depth scanning [11], Manuscript [12], [13]. Therefore, novel depth scanning techniques attract much attention. Depth scanning techniques have improved significantly in the past decade, and spatial offset Raman spectroscopy (SORS) [14], [15], [16] was proposed for depth-sensitive Raman spectroscopy. SORS uses the divergent excitation light from an optical fiber [17], a collimated beam [18], or a focused beam for excitation but a ring of fibers spatially apart from the excitation spot for subsurface detection. However, these methods based on SORS rely on moving the area of collection of scattered light away from the laser-illuminated zone. Accordingly, SORS uses a proximal depth scanning mode, which is the same depth scanning mode used in the motor stage-based Raman system. This scanning mode increases the complexity of the system and the size of the probe because of the existence of a moving part in the system that is used to reposition the collection probe [17], [19].
In this paper, we developed a distal scanning Raman system. In this system, a graduated ring-activated zero aperture was used to select the Raman spectrum from the wide focal range of an axicon lens. The main difference between our system and the traditional Raman system was the lens and increased aperture. Instead of an objective lens, an axicon lens was used to modulate incident light over the depth direction to obtain the full range of Raman information without depth scanning [20], [21]. Depth scanning was realized by changing the sizes of the graduated ring activated zero aperture to select the Raman spectrum from a fixed depth range.

A. Experimental Setup
The full range optical setup of Raman spectroscopy was illustrated in Fig. 1. The near infrared excitation light source with a central wavelength of 785 nm was collimated by lens L1. The collimated light was filtered by a Rayleigh filter to reject the undesired wavelength and subsequently reflected. The excitation light was focused onto the sample by an axicon lens to obtain an extensive focus depth. The backscattering light from different depths was collimated by the axicon lens. The Raman signal was selected with a long-wavelength pass filter. An aperture whose size can be adjusted was inserted in the front of the pinhole to select the Raman signal with a different depth. Finally, the Raman signal was collected by a spectrometer. Different from traditional Raman spectroscopy, this Raman spectroscopy used an axicon lens instead of an objective lens to extend the focus depth and obtain the full-range Raman signal. The Raman signal with a different depth was selected by adjusting the aperture sizes.

B. Principle
The focal length of the axicon lens is given by [22] (1) where D is the diameter of incident light, α is the base angle of the axicon lens, and n is the refractive index of the axicon lens. The equation indicates that focal length varies with the diameter of incident light when the axicon lens is used. The diameter of incident light in our system is about 12 mm. The base angle and refractive index of the axicon lens are 20°and 1.45, respectively. The calculated focal length is about 36 mm. When only the backscatter Raman signal is considered, the relationship between the selected depth and the change of aperture sizes is [22], [23] ΔL= 2.97 * ΔD According to (2), depth resolution is related to the minimum aperture sizes that the Raman spectra can be detected when the optical power and performance of the spectrometer are ensured. However, in practical application, when the sizes of the aperture are selected, the measured depth may be larger than the theoretical value. The reason is that part of the Raman signal from other direction, except for backscatter light can also be detected.
To demodulate the overlapping Raman spectra of different materials, two Raman spectra are required. One is the Raman spectrum S(L 1 ) on the front surface of the desired material (located at L 1 corresponding to the aperture size D 1 ), and the other is the Raman spectrum S(L 2 ) on the back surface of the desired material (located at L 2 corresponding to the aperture size D 2 ) The Raman signal of this material can be obtained by The Raman characteristic of the material can be demodulated from the overlapping spectrum when the material's thickness is larger than the depth resolution of the system.
As previously stated, only the backscatter Raman signal can be detected as a portion of the Raman signal. That means even  though the aperture size is corresponding to the depth L 1 , the Raman signal from deeper region can still be detected. The Raman signal of the desired material can be reconstructed by improving (3). The detail reconstruction process can be performed as follows.
To identify the Raman spectrum of a single layer from the multi-layer samples, different Raman spectra with various aperture sizes was measured. For example, the Raman spectra when the aperture sizes are 5 mm and 3 mm are selected to reconstruct the Raman spectra of PVC and PPE. The Raman spectrum of PVC can be obtained by whereS Ramanshif t (D 1 ) is the unique Raman intensity for material 1 when the aperture size is D 1 and S(D 1 ) is the overlapping Raman spectrum obtained when the aperture size is D 1 .
For the data presented in Fig. 4, the Raman spectrum of PVC derived from the multi-layer sample can be expressed as Focal range is measured by Si which is located close to the apex of the axicon lens. Raman spectra are acquired with a depth step of 5 mm. The power of light source is set to 300 mw and the imaging time of one Raman spectrum is 10 s. The Raman spectra are obtained 10 times. The depth range of Si when the size of the aperture is 3 mm is obtained with a depth step of 6 mm. The power of the light source is 300 mw and the imaging time of one Raman spectrum is 10 s. The Raman spectra are collected only once. The aperture is positioned on one side of the optical axis to avoid the error induced by the position of the aperture.

C. Axial Resolution
The axial resolution of the Raman spectroscopy between the objective lens and axion lens didn't measure in this manuscript because of the following two factors.
1) The focal length of the objective lens and axion lens was different which means axial resolution for both system should be different. As a result, the comparison between them is meaningless. 2) In reference (6), Olivo et al. has demonstrated that the axial resolution of the Raman spectroscopy is related to the value of NA of the objective lens and the core size of the fiber. Considering the NA of the axion lens for different diameters of incident light keeps the same [23] and the core size of the fiber is also the same, the axial resolution for different depth of the axion lens keeps the same. And if the focal length of the axion lens is the same as the objective lens, the theoretical axial length should be same.

D. Preparation of PVC and PPE
The PVC and PPE are bought from Alibaba. The thickness of PVC and PPE are about 4 and 1 mm, respectively. PPE is positioned above PVC. The distance between PPE and the apex of the axicon lens is about 7 mm.

III. RESULTS
We demonstrated the wide focal range of our system by positioning Si at different depths. As shown in Fig. 2, the Raman peak of Si was at 520 cm −1 . The measured depth varied from 0 mm to 35 mm (distance between Si and the apex of the axicon lens). The imaging depth of our system was ∼35 mm when the diameter of the incident beam was 12 mm. This imaging depth was almost one order of magnitude larger than the Raman system using a low NA objective lens. Therefore, the full range of Raman information was obtained without depth scanning in our system. Fig. 3 showed the imaging depth when the aperture was set to 3 mm. The aperture was placed at the left side of the optical axis of the Raman system to avoid aperture-induced errors. The Raman spectra of Si, whose depth varies from 10 mm to 30 mm, were obtained. The Raman peak was evident when the depth varies from 11 mm to 29 mm. When the depth was 10 or 30 mm, the Raman peak was almost equal to the peak of noise. The imaging depth was about 18 mm which was a little bigger than the theoretical value 17.82 mm. The depth at which the Raman spectrum can be detected varies from 11 mm to 29 mm instead of 0 mm to 18 mm because the center of the aperture was far from the optical axis of the Raman system. Fig. 4 presented the relationship between the change of aperture size and the change of depth which was introduced in (2). The aperture sizes used here were 2.5 mm, 3 mm, 3.5 mm, 4 mm, 4.5 mm, 5 mm and 5.5 mm. The depth scanning was realized using a stepper motor. The step between two adjacent sampling points was 0.5mm. The imaging depth was defined as the depth where the Raman intensity dives to two times of the noise. Fig. 4(a) showed the fitting curve of the intensity when the aperture sizes were 3.5 mm, 4.5 mm and 5.5 mm. Fig. 4(b) showed the fitting curve of the relationship between aperture size and depth. The slope of the line in Fig. 4(b) was 2.585±0.088 which was close to the theoretical value 2.97 in (2). The main error was induced by the scattering light from different direction, the error of depth adjusting by the stepper motor and noise.    The main difference in the Raman peaks of PVC and PPE can be observed at 636 and 809 cm −1 , where the Raman peak was not overlapped, and the intensity value is at its maximum. Fig. 6 showed the Raman spectra obtained from the multilayer sample of PVC and PPE. The shapes of the spectra were the same, except when the size of the aperture was 1 mm. The intensity ratios at 636 and 809 cm −1 varied with the size of the aperture shown in Table I. This result indicated that the Raman signal of PVC decreased when the size of the aperture decreased.
To demonstrate that this phenomenon was not induced by the signal-to-noise ratio of the system, we calculate the maximum peak ratio of PVC, which was at 636 cm −1 . The results were shown in Table I. When the aperture size decreases from 12 mm to 5 mm, the ratio is almost equal; when the aperture size decreases from 5 mm to 3 mm, the ratio changes. When the aperture size was changed to 2 mm, the theoretical imaging depth became 5.94 mm. This value was larger than the thickness of PVC. The reason was that the imaging depth calculated by (2) was in air, and the refractive index of the material was not considered. The refractive index of PVC was about 1.52 and its optical thickness was about 6.08 mm, which was close to the measured value of 5.94. The main error arose from the assumption that only the backscattered Raman signal was collected. In practical application, the exits of other Raman signal were also collected to extend the imaging depth. Fig. 7(a) and (b) showed a comparison of the separated PVC and PPE from the multi-layer sample with single PVC and PPE. The results indicated that using the axicon lens and graduated ring-activated zero aperture was effective in obtaining the Raman spectra from different depths. The Raman peak of PPE obtained from the multi-layer sample was similar to the Raman peak obtained from the pure PPE. In Fig. 7(a) and (b), the spectra were smoothed and the baseline was removed with the algorithm developed by Renishaw. After these pre-processing procedures, the Raman peak separated from the multi-layer sample and that obtained from pure PVC or PPE became similar. Compared with   A three-layer model included PET, silica gel and PC was used to analyze the influence of the number of layers. The full range Raman spectra at different aperture's sizes were shown in Fig. 8. And the demodulated Raman spectra of PET, silica gel and PC were shown in Fig. 9. Compared with that of pure PET, pure silica gel and pure PC obtained by traditional Raman spectroscopy, the main Raman peaks were the same, especially for the upper layers. For the bottom layer PC, there was only one Raman peak for the signal demodulated from full range Raman spectra. One of the reason was the light power incident on PC which is the bottom layer is relatively low and the Raman signal collected need penetrate through PET and silica gel which would decrease the Raman signal. The other reason was the demodulated algorithm used here was the subtraction of two Raman spectra. Since the subtraction would induce that the Raman intensity that we want to demodulate would decrease when subtraction is performed, while the noise would increase. The signal-to-noise of the demodulated signal would decrease. So the demodulated algorithm need to be improved to obtain high signal-to-noise. Fig. 10 was the Raman spectra of pork in a range of 1000 cm −1 to1700 cm −1 . The skin in the pork is thinner than 1 mm. It's mostly fat below the skin. The Raman spectra for skin and fat are nearly the same except 1299 cm −1 (attributed to CH2 twist) for fat and 1253 cm −1 (attributed to amide III band) for skin. As shown in Fig. 10, when the aperture size is 3 mm the intensity at 1253 cm −1 was higher than that at 1299 cm −1 . As the aperture size increased, the ratio of the intensity at 1253 cm −1 and 1299 cm −1 decreased. Especially when the aperture size was 12 mm, the intensity at 1299 cm −1 was higher than that at 1253 cm −1 . This demonstrated that it was effective to modulate the Raman intensity from different depth by adjusting the sizes of the aperture.

V. CONCLUSION
In this study, we described a new kind of Raman spectroscopy. It can obtain the comprehensive Raman spectra for full range and demodulate the Raman spectra of the desired material over the full range. Two and three-layer models were used to demonstrate the new system. The advantage of this depth-sensitive Raman system is that it uses distal depth scanning mode which allows for the simplification and size reduction of the endoscopic probe. This design is convenient for clinical measurements especially, especially in vivo ones, because precise control of the lens-sample distance is unnecessary. Considering that the Raman signal can be obtained synchronously, future improvements are likely to use a spatial light modulator to select the Raman signal from different depth ranges synchronously and improve the demodulated algorithm to avoid the decrease of the signal-to-noise ratio.