Optimizing signal collection e ± ciency of the VIPA-based Brillouin spectrometer

Brillouin spectroscopy is an emerging tool for microscopic optical imaging as it allows for noninvasive and direct assessment of the viscoelastic properties of materials. Recent advances of background-free confocal Brillouin spectrometer allows investigators to acquire the Brillouin spectra for turbid samples as well as transparent ones. However, due to strong signal loss induced by the imperfect optical setup, the Brillouin photons are usually immersed in background noise. In this report, we proposed and experimentally demonstrated multiple approaches to enhance the signal collection e±ciency. A signal enhancement by > 4 times can be observed, enabling observation of ultra-weak signals.


Introduction
][7][8] Brillouin scattering originates from the inelastic interaction between the incident electromagnetic wave and the acoustic phonons within the material.As a result, the incident optical frequency experiences a shift proportional to the speed of sound in the medium.The medium's elastic modulus, which is directly related with the speed of sound in the medium, can be determined simply by measuring the Brillouin shift.Compared to other imaging techniques that are capable of quantifying material's mechanical properties (e.g., Refs.9 and 10), Brillouin spectroscopy o®ers a noninvasive and label-free contrast mechanism with microscopic spatial resolution.The microscopic elasticity of the living organisms enables investigations of cell mechanics, which is essential for understanding biological development and disease pathophysiology. 11Moreover, typical Brillouin spectrometers are fully based on optics, which makes it easy to couple with other optical imaging techniques (e.g., optical coherent tomography (OCT), 12,13 Raman spectroscopy [14][15][16][17] and optical sensors 18,19 ).However, despite being known for decades, Brillouin microspectroscopy is still considered as an emerging technology.There are two major hurdles: the weak Brillouin scattering cross section and the tiny Brillouin shift (usually less than 10 pm in wavelength).Recently, by taking advantage of single-or multi-stage virtually imaged phased array (VIPA, 16 ), Scarcelli and Yun developed Brillouin spectrometers to acquire 2D in situ images of biological tissues. 6,20It drastically saved the acquisition time and reduced the system complexity.Moreover, with the application of atomic/molecular absorption cells, Meng et al. developed backgroundfree Brillouin spectrometers, further reduced the contaminations from elastic scatterings induced by turbid samples. 21However, the weak signal is still highly possible to be immersed in the background.This becomes a signi¯cant issue when imaging turbid biological tissues. 8,22This problem is further compounded when paired with imperfect optical setups, which are usually polluted with unwanted background noises and signal losses.Scarcelli and Yun addressed these di±culties by introducing a multi-stage VIPA setup, which provides an additional 25 dB of the background suppression. 23owever, in most practical applications involving confocal imaging of highly scattering samples, even stronger background suppression is required.Moreover, the original signal was also weakened via this multi-stage setup.To induce higher signal-to-noise ratio, one possible strategy is to add additional VIPAs, and, while this would reduce the background, it would also further complicate the optical setup as well as signi¯cantly reduce the useful Brillouin signal.Another possibility, increasing the pumping power, is usually highly undesirable for biology-related applications, as the corresponding thermal e®ect may burn the samples.How to e®ectively utilize the limited number of Brillouin scattered photons becomes an important problem.
In this report, we speci¯cally investigated the signal loss induced by a single-mode (SM) ¯ber attached to the VIPA Brillouin spectrometer.We found that due to the insertion loss and the inherent attenuation, SM ¯ber becomes a signi¯cant source inducing signal loss.Further, we replaced the ¯berbased setup with an open space setup with a pinhole and found that the corresponding signal has been enhanced.We demonstrated that the employment of such a pinhole is helpful in enhancing the sensitivity of the VIPA-based Brillouin spectrometer.
Additionally, we also introduced Faraday rotators for the purpose of ensuring strict backscattering while maintaining considerable signal collection ef-¯ciency.In this study, we, for the ¯rst time, utilize a VIPA spectrometer in conjunction with a confocal pinhole setup, which allows the acquisition of in vivo viscoelasticity-speci¯c microscopic images.In this report, for the sake of simplicity, we only demonstrate spectroscopic applications to illustrate the proof of the principle.

Methodology
The basic framework of the VIPA-based Brillouin spectrometer follows the description in Meng et al. 21igure 1(a) portrays the main experimental setup.A 532-nm single longitudinal mode solid-state laser (Lasermate Group Inc.; model: GMSL-532-100FHA) was employed as the source of incident radiation.The center wavelength is speci¯ed as 531:9587 nm AE 0:3 pm with its maximum output power at $100 mW.The nominal output linewidth for this pump source is $ 640 kHz.An optical isolator (Electro-Optics Technology, Inc., Model: BB-8-05-I-090) prevented unwanted feedback from the rest of the optical setup.A polarizing 50/50 beam splitter was placed in the beam path to re°ect the backscattered light toward a pinhole (Model: P25S, Thorlabs).In this study, for the purpose of making comparisons, we also tested di®erent entrance strategies of the VIPA spectrometer.A SM ¯ber (1-m long, Model: SM600, Fibercore Inc.) and a multi-mode ¯ber (1-m long, N:A: ¼ 0:22, Model: FG050LGA, Thorlabs Inc.) were taken into consideration.The corresponding ¯ber launching system (Model: NanoMAX-TS, Max313, Thorlabs Inc.) was adopted as well.An in¯nity-corrected microscope objective lens (Nikon Inc., CFI Plan Fluor 20x, N:A: ¼ 0:5) served to both focus the pump onto the sample and collected the backscattered light.A Faraday optical rotator (Electro-Optics Technology, Inc.) was placed prior to the objective lens.The power at the sample was less than 40 mW for all the measurements.The sample solution was placed in a quartz cuvette (Starna Cells Inc.).
Figure 1(b) depicts the VIPA spectrometer in detail.The pinhole/optical ¯ber output was collimated by a positive lens and sent to an iodine absorption cell (Opthos Instruments, Inc.).By tuning the cell's temperature, we are capable of controlling the vapor pressure.A 532-nm line ¯lter was placed behind the iodine cell to ¯lter the undesired molecular emission and Raman signals.The rest of the VIPA spectrometer followed the design set forth by Scarcelli and Yun. 20The VIPAs (Model: OP-5642, Light Machinery Inc.) were speci¯cally designed for 532-nm applications with a nominal free spectral range (FSR) as 33.3 GHz.A lens (focal length 1000 mm) was used after the VIPA so that Brillouin peaks and elastic peaks were well separated.A CCD camera (Model: Newton 971, Andor Technology, Inc.) was placed at the focal plane to collect the outputs.

An open-space pinhole setup
enhances the Brillouin signal strength while maintains strict confocality When collecting data in Fig. 2(a), we followed the ¯ber launching system described by Meng et al. 21A spherical lens (f ¼ 8 mm, N:A: ¼ 0:5) was applied in order to launch the backscattered beam into the ¯ber.The iodine cell was heated to 140 C.The result shown in Fig. 2(a) proves that the SM ¯ber setup is capable of collecting Brillouin signals as well as elastically scattered photons, though the overall signal level is low (< 30,000 photons per Brillouin line).Compared with Scacelli and Yun, 20 our result is $ 5 times weaker, which may due to the absorption of the iodine cell.
When collecting the data shown in Fig. 2(b), we employed an N:A: ¼ 0:22 multi-mode ¯ber manufactured by Thorlabs (Model: FG050LGA).There was no observable Brillouin peak when using multimode ¯bers.Therefore, we chose to replace the sample with a re°ective mirror, and reduced the CCD integration time (50 ms).Compared with SM ¯bers, the multi-mode ¯ber gave a loosely focused stripe.This is mainly caused by the imperfect input of the VIPA.When using multi-mode ¯bers, their large core size and high numerical aperture make its output di±cult to be tightly re-focused.Therefore, the input may be sent to cover a broad area around the optimal entrance.Aspherical and acylindrical lens may squeeze the focusing spot tighter by eliminating optical abbreviations.However, this strategy also increases the system cost and complexity.This loosely focused input also forbids signal to be e±ciently coupled into the VIPA.For example, in order to ensure 90% signal to be coupled into the VIPA using a 50 m wide stripe, the corresponding minimal tilting angle of the VIPA would be greater than 2.6 (calculations follow Shirasaki 24 and Xiao et al. 25 ).As a result, when focusing the VIPA output at 1-m away, the separation between the peaks of di®erent orders will be $ 2 mm, or 125 pixels in the Newton 971 camera (pixel size: 16 Â 16 m).Considering the FSR of the VIPA (33.3 GHz), each pixel will represent at least 260 MHz, which results in a very poor spectral resolution.The peak separation could be expanded with smaller tilting angle.However, the re°ective coating will block a signi¯cant portion of the signal photons.Moreover, instead of the desired fundamental mode (LP01) given by SM ¯bers, the output of a multi-mode ¯ber is not simply a Gaussian distribution.In most cases, the output of multi-mode ¯bers will include a composition of many fundamental modes, many of which are hollow at the center.As a result, the image received by a CCD camera will be a®ected by the ¯ber output pattern and contaminate the acquired Brillouin spectrum.
Figure 2(c) presents the typical output of a VIPA spectrometer with an open-space entrance without any spatial ¯lters.By removing the ¯ber setup, the corresponding insertion loss and the ¯ber's attenuation e®ect were avoided, and hence the CCD reading was drastically increased.However, when the equivalent pinhole is not presented, the inevitable scatterings/re°ections induced by the optical components and the surface of the sample container/holder would be sent to the CCD detector, which induces unwanted contaminations.Moreover, when employing high N.A. objective lenses, Brillouin peaks broadening may happen due to the dependence of the frequency shift on the scattering angle. 26ithout an equivalent confocal pinhole (e.g., without an SM ¯ber), this broadening e®ect cannot be e±ciently suppressed. 8,27In the speci¯c measurements shown in Figs.2(a) and 2(c), the Brillouin line was broadened by 5 pixels (i.e., $400 MHz in Brillouin shift).The comparison is also given in Fig. 2(e).
Figure 2(d) depicts the CCD reading of a VIPA spectrometer with an open-space entrance plus a pinhole.Compared with Fig. 2(c), the Brillouin peaks' linewidth becomes narrower, though the signal strength was weakened as well.Nevertheless, the signal was still stronger than ¯ber-based setups.
Figure 2(e) compares the obtained Brillouin spectra by aforementioned di®erent strategies.Note that the SM ¯ber's result has been zoomed-in by 3 times for better visibility.Due to the laser wavelength drift, the Stokes peak has drifted as well, and may have passed through an absorption band of the iodine cell.Therefore, the Stokes peak in the pinhole experiment did not match its counterparts in the \no pinhole" and \SM ¯ber" measurements.Nevertheless, the anti-Stokes peak was not drastically a®ected by the iodine cell.The comparison shows that the anti-Stokes peak in the \pinhole" setup was $ 10.70 times stronger than its counterpart in the \SM ¯ber" setup.The inset of Fig. 2(e) depicts the linewidth di®erence between various setups.The peaks have been normalized.Unlike the \no pinhole" result, the Brillouin linewidth in the pinhole setup was similar with the \SM ¯ber" setup, indicating that the pinhole setup's confocality was maintained.

An optical Faraday rotator helps further enhance the collected signal and reduce the setup complexity
Due to the tiny wavelength shift in Brillouin scattering, there are no available dichroic mirrors that could separate the Brillouin-scattered photons from elastically scattered photons.In previous studies, investigators usually introduce a small o®set in the incident angle, so that the Brillouin scattered photons will not follow the original incident path (e.g., Refs.4, 5, 20 and 23).In this way, scattered photons could be spatially separated from the incident beam.However, the scattering angle of the collected photons would be slightly away from 180 with an uncertainty (resulted from manual tuning).The corresponding Brillouin shift, due to its dependency on the scattering geometry, will be a®ected.Another approach is to utilize a 45 non-polarizing beam-splitter (usually a re°ective neutral-density ¯lter) to split the incident and the backscattered light (e.g., Refs.8 and 21).This approach ensures a 180 scattering angle.However, even in the best case, only 50% of the signal could be induced to the detector.Considering the beam splitter also blocks another 50% of the incident energy, 75% of the total system total potential was wasted.
On the other hand, in most of the biomedical sensing and imaging applications, only C 11 in the sti®ness tensor is taken into consideration.In this case, we only need to measure the backscattered Brillouin photons which keep the original polarization state of the incident beam. 28In this experiment, we inserted a Faraday optical rotator prior to the objective lens.In this way, the polarization axis of the incident beam would be rotated by 45 when passing the rotator.The Brillouin scattered photons originated from the sample would pass the rotator again from the opposite direction.Consequently, the polarization status of the Brillouin photons would be rotated by an additional 45 , which makes the incident and the backscattered beams to be polarized in perpendicular directions.Therefore, the two beams can be simply separated by a polarizingbeam splitter (PBS).In practice, we adopted a broadband PBS (Model: PBS051, Thorlabs) and a broadband Faraday rotator (Electro-Optics Technology, Inc.).A typical output is shown in Fig. 3(a).The signal was collected without ampli¯cation setup (for example, electron multiplication).A fully vertical binning (FVB) option was selected, so the readings on each column of the CCD array were added up.The strongest Brillouin peak reaches $ 20,000 photons in amplitude.Considering the linewidth and multiple replicas of the Brillouin peaks, there were over 2,000,000 photons contributing to a speci¯c Brillouin shift.Figure 3(b) shows a comparison among all the strategies mentioned in this study.The SM ¯ber's result was again zoomed-in by 3 times for better visibility.Note that the \SM ¯ber" and the \pinhole" measurements in this ¯gure were derived without Faraday rotators.Nevertheless, the results shown in Fig. 2 were collected with assistance of Faraday rotators.The anti-Stokes peak in \pinhole þ FR" was 2.85 times stronger than the standalone \pinhole" setup, and was 30.21 times stronger than \SM ¯ber" setup.Again, due to the iodine absorption and laser wavelength drifting, the Stokes peaks were not in a consistent position, and the absorption e±ciency of the elastic peaks were distinct during separate measurements.

Conclusion
We have demonstrated a simple and e±cient Brillouin microspectroscopy setup, which allows for an accurate and fast assessment of Brillouin spectra of weak scattering samples.For the purpose of enhancing the signal collecting e±ciency, we replaced the ¯ber-based entrance by an open-space setup with a pinhole.To ensure the scattering angle to be 180 while maintaining 100% signal collection e±ciency, we inserted a Faraday optical rotator prior to the focusing objective lens.The optimizations of the setup are simple but e±cient.We anticipate the wide use of this experimental arrangement in future applications of Brillouin microspectroscopy.

Figure 2
Figure 2 shows the typical outputs of the VIPA spectrometer using various entrance strategies, including single-and multi-mode ¯bers (Figs.2(a) and 2(b), respectively), open-space without and

Fig. 2 . 3 J
Fig. 2. CCD images acquired with di®erent setups utilizing (a) a SM ¯ber, (b) a multi-mode ¯ber, (c) open space without pinhole and (d) open space with a pinhole.(e) the retrieved Brillouin spectra for each setups.The \SM ¯ber" result has been zoomed-in by 3 times.S: Stokes peak; AS: anti-Stokes Peak; E: Elastic peak.We used acetone as the sample in all tests except (b).Incident power: 35 mW.Integration time: 4 s.

Fig. 3 .
Fig. 3. Brillouin spectrum acquired with the optimized system.The exposure time was 2 s, and the incident power was 35 mW.(b) Comparison between di®erent VIPA spectrometer entrances.The \SM ¯ber" result has been zoomed-in by 3 times.The sample was acetone.S: Stokes peak; AS: anti-Stokes peak; E: elastic peak; FR: Faraday rotator.