We present a prism-based spectrometer integrated into a multifocal, multiphoton microscope. The multifocal configuration facilitates interrogation of samples under different excitation conditions. Notably, the image plane of the microscope and the image plane of the spectrometer are coincident eliminating the need for an intermediate image plane containing an entrance slit. An EM-CCD detector provides sufficient gain for spectral interrogation of single-emitters. We employ this spectrometer to observe spectral shifts in the two-photon excitation fluorescence emission of single CdSe nanodots as a function of excitation polarization.
©2010 Optical Society of America
Spectroscopy plays a key role in multiphoton microscopy, augmenting the spatial information obtained via imaging with information about local chemical environments. As is often the case in biological systems, changes in the spectrum of an applied fluorophore reveal details regarding the chemical environment in which the fluorophore resides, or the object to which it is bound, if the marker is exogenous [1–3]. Spectral changes reflecting some physical, chemical, or electronic alteration are not restricted to the biological realm, however. Quantum dots – of increasing interest as laser sources, biological markers to replace dyes, and as additives to improve solar cell efficiency – exhibit changes in spectrum when a second carrier, known as a biexciton, is promoted to the excited state . These shifts and additional peaks are on the order of several nanometers and are very weak, requiring high spectral resolution, high sensitivity, and narrow emission bandwidth to observe.
More easily observed are the spectral shifts due to different degrees of quantum confinement in quantum dots. As confinement increases with a decrease in dot size, the bandgap of the dot will increase from that of the bulk [5–7].
1.1 Current spectrometer technologies
Morris, et. al write that spectral imaging, and thus spectrometry in general, “would benefit from an imaging spectrometer technology providing high spatial, spectral, and timing resolution.” They further expound upon the ideal spectrometer, seeking broad spectral range, a high acceptance angle, low loss, substantial out-of-band rejection, as well as polarization insensitive detection. Finally, this ideal spectrometer should be physically compact, mechanically rugged, and electronically controllable . To this list, we add one more characteristic: the system must be highly sensitive to low-light events, such as single molecule or quantum dot emissions.
Current spectrometer technologies and techniques address only portions of the above ideal. Common commercial grating-based spectrometers provide the simplest direct visualization of spectra. Most have a marked polarization dependence and a limited spectral range as a function of groove density or overlapping orders. Scanning spectrometers can use high-gain single element detectors to observe low-light events, but have moving parts that preclude high temporal resolution. Imaging spectrometers employing stationary gratings have gained ground with the recent introduction of high-gain electron-multiplying CCDs (EM-CCDs), improving acquisition times substantially. In order to maximize efficiency, gratings with different blaze angles must be substituted, especially if the feature of interest spans a sizable wavelength range. To optimize the system throughput for a grating-based spectrometer while maintaining spectral resolution, the f/numbers of the input beam and the spectrometer must be carefully matched. Small slit sizes (<25 μm) are required.
Alternative techniques for observing spectra are found in spectral imaging, a combination of spectroscopy and imaging. These methods construct a three or four dimensional data array by gathering the spectrum from each spatial location in a sample as the beam is moved across the sample. These techniques fall in to one of two categories: spatial scanning or wavelength scanning. Replacing the single slit in a grating-based spectrometer with multiple slits in some arrangement (known as a mask) improves throughput, but requires post-processing to remove artifacts .
A common technique in the latter category is Fourier spectroscopy, wherein the light produced from a fluorophore interferes with itself via a path length change [10,11] often produced by a Sagnac interferometer . This method boasts high throughput and sensitivity, with a spectral resolution of better than one-hundredth of the excitation wavelength, but requires high stability and algorithmic modifications for phase correction and apodization [10,11,13]. Alternative techniques include liquid-crystal tunable filters (LCTF) and acousto-optic tunable filters (AOTF). LCTFs provide fast wavelength selection, but limited spectral range, low transmission, and sensitivity to temperature and polarization . AOTFs provide electronically controllable wavelength tunability and high throughput, but have historically been subject to out-of-band rejection ratios of only 10□3 and require post-processing to correct the image quality [14,15]. A spatial resolution of 0.35 μm has been reported with post-processing .
1.2 Prism-based spectrometry
Prisms are an alternative to gratings, though not nearly as dispersive. However, prisms provide advantages over gratings – lacking higher orders, prisms possess greater efficiency and spectral range. The glass type and geometry of a prism dictates the amount of dispersion, and thus the spectral spread of detected light. Prism throughput is less sensitive to polarization and wavelength than grating throughput. A SF-10 prism at Brewster's angle passes 99.9% of p-polarized light and 75.4% of s-polarized light. The transmission of a prism will only vary by a few percent over a wavelength range of 500 nm to 1 μm. Reflective diffraction gratings may lose 50% transmission or more (from peak transmission) over the same range, often with increasing disparity between orthogonal polarization states.
With the advantages of prisms over gratings in mind, Lightwave, Inc. has fabricated a prism-based spectrometer with a specialized curved prism . This system, known as PARISS®, relies on the same principles as an imaging grating-based spectrometer: an entrance slit is ultimately imaged to a detector plane, while the curved prism provides the dispersion necessary to capture a spectrum. The curved prism and another spherical mirror compensate for the non-linear dispersion glass applies to passing light.
In the present work, we have implemented a simpler prism-based spectrometer fabricated with “off the shelf” optics specifically for multiphoton microscopes that has no slit, requires no pre-filtering of the beam to the spectrometer, and yields a spectral resolution that varies between 0.3 nm and 2.2 nm across the detector. The variation in resolution as a function of position is a direct result of the nonlinear dispersive characteristics of the prism. In multiphoton spectroscopy, with the source wavelength far removed from the emission wavelength, there is no truncation of the observed spectrum. It achieves a calculated throughput of 83% (for unpolarized light, from input to detector) via minimal optical elements. In addition, it possesses a tunable, broad spectral range of ~310 nm determined by both the index of the prism glass and the optical power of the focusing lens. Unlike traditional grating-based spectrometers, high spectral resolution is not dictated by a slit, but rather by the dispersion of the prism and the beam size. Thus, our spectrometer employs the full acceptance of a high-numerical aperture objective without the loss associated with focusing through a slit. A schematic comparison between traditional spectrometers and ours is shown in Fig. 1 . The high NA and throughput yield spectrum acquisition times from single emitters of 30 seconds or less. Further, it is largely polarization insensitive, with 88.57% throughput for p-polarized light and 78.03% for s-polarized light and has no moving parts when in operation. Lastly, and most significantly, it is sensitive to single-emitters through use of an EM-CCD.
We characterize the performance of this spectrometer by examining single CdSe quantum dots. Detecting emissions from single emitters tests the sensitivity limit – a function of detector gain and system throughput. Small spectral shifts are observed by imaging the same single dot with different excitation polarizations. Additionally, geometric optical models are used to examine how the resolution varies across the detector as a result of using a prism as the dispersing element.
2. Experimental apparatus & sample preparation
The source laser and the multifocal, multiphoton microscope are described in detail in Ref. 17, and are only briefly summarized here. Our Yb:KGW laser outputs ~50 nJ, 250 fs FWHM pulses at 56 MHz, centered at 1038 nm. This pulse train feeds into our “optical multiplexer,” a Michelson interferometer-like construct that outputs two interlaced pulse trains of 56 MHz, as shown in Fig. 2 . These overlapped pulse trains are orthogonally polarized and time-delayed by half of the round trip cavity time, resulting in a pulse train of 112 MHz with sequential orthogonally polarized pulses. The pulse train is directed into an Olympus IX-71 inverted microscope equipped with a piezoelectric stage (Mad City Labs, Nano-view 200M/3) for sample placement or scanning. A dichroic (Chroma, DCXR1000) placed before the objective passes the excitation light while reflecting wavelengths from 300 to 900 nm. An Olympus UPlan FLN 40x/0.75 NA objective both images and collects in the epi-direction. At 0.75 NA, we are imaging at the diffraction limit – 0.9 μm. A pair of 170 mm lenses relays the plane mid-way between the close-coupled GSI Lumonics scan mirrors to the back of the objective, enabling beam rastering at the sample plane. We can examine a given sample for polarization sensitivity quickly by imaging with multiple foci matched at the same plane, then further explore spectral variations as a function of polarization via the prism-based spectrometer.
2.1 Imaging and spectral detection
We modified our detection apparatus from Ref. 17 to include the prism-based spectrometer, as shown in Fig. 3 . When scanning the sample, a Hamamatsu H7422P-40 PMT equipped with the appropriate filters detects two-photon excitation fluorescence (TPEF) or the second- or third-harmonic (SHG, THG) of the Yb:KGW excitation laser. A silver mirror oriented at 45° to the outgoing light directs the signal into the PMT. This mirror is removed when acquiring spectra.
The prism-based spectrometer pathway begins with a baffle to remove on-axis, non-collimated light reflecting off of the objective. Achromatic 50 mm lenses (ThorLabs, AC254-050-A) focus and recollimate the light, while a 150 μm pinhole removes stray light. The presence of the baffle and lack of descanning in our detection system require sample movement to directly compare spectra from different points within the sample, which is performed with the piezoelectric sample stage. A ½” tall equilateral SF10 prism set at Brewster's angle disperses the incoming light onto a 200 mm focal length achromatic lens (ThorLabs, AC254-200-B) located past the prism. This lens, placed 100 mm from the prism, results in a spectral range of 320 nm. The detector (Photometrics, Cascade II:1024) is placed at the focal point of the lens. Just prior to the camera, a small arc of BG-39 glass is used to block the IR source light without attenuating the remainder of the signal light. The baffle does not completely remove all back-reflected non-collimated IR source light, so an additional 3 mm thick KG3 filter is placed just prior to the BG-39 filter. The design is particularly well-suited to confocal imaging systems, given the detection is descanned and the existing pinhole removes out of plane and stray light.
While the system is passive, minor substitutions or rotations greatly expand the versatility of the spectrometer. With a 200 mm lens the spectral range spans 320 nm. Lens substitutions will vary the observed spectral range. Substituting any lens with a focal length less than 100 mm will focus a monochromatic collimated beam of radius 7 mm to a spot smaller than the pixel size of our detector, which is 13.3 μm. Lastly, the prism is mounted on a rotation stage to enable convenient wavelength tuning.
2.2 Spectrometer calibration
The spectrometer is calibrated by using a gas lamp source. Imaging the slit in front of a neon atomic gas lamp to the focal plane between tube lenses – shown schematically in Fig. 4 – provides sufficient spectral resolution to resolve nearby peaks. The incoming light reflects off of a mirror placed at the focus of the imaging objective, then reflects off of the dichroic into the camera. Only infrequent calibrations are necessary, given the mechanical stability of the spectrometer. Figure 5 contains a lineout from the image, with selected peaks marked with their appropriate wavelengths, as obtained from the neon gas standard. This particular calibration applies to a configuration where the spectral range was ~320 nm. A second-order polynomial fit is applied to the selected peaks to generate a wavelength vs. pixel calibration.
2.3 Geometric optical modeling
From Fig. 5, it is immediately apparent that the spectral resolution varies across the detector width, especially upon observation of the wide peak at 703 nm. Geometric optical modeling of the system over a 200 nm spectral range provides straightforward quantification of the effects of both axial color and non-linear dispersion on resolution, as shown in Fig. 6 . The inset in Fig. 6 illustrates how the spatial resolution at the central wavelengths is diffraction-limited, while the longer wavelengths near the bottom of Fig. 6 expand rapidly as a result of aberrations, reducing spectral resolution as a function of position on the detector surface. Table 1 lists the Airy disk size, RMS spot size (for spots larger than the Airy disk radius), and spectral resolution at selected wavelengths across the detector range. The spectral resolution listed is determined by the Rayleigh criterion, and does not take detector pixel size into account. The measured calibration from a neon lamp and the theoretical calibration obtained through geometric optical modeling agree within 6%.
By shifting the detector 2.9 mm further from the 200 mm lens, the longer wavelengths can be brought into focus while defocusing the shorter wavelengths. Because the shorter wavelengths are dispersed more than the longer wavelengths, the increased distance between the shorter wavelengths on the detector surface actually helps to compensate for the increased spot size caused by axial color in this configuration, maintaining an improved spectral resolution over the entire detector range.
As mentioned previously, only the dispersion of the prism and the beam size affect the spectral resolution of the spectrometer. The geometric optical model indicates the positioning of the final lens relative to the prism leaves spectral range and resolution unchanged. Additionally, the optical power of the final lens only adjusts the spectral range and leaves the resolution unchanged so long as the spot size is larger than the detector pixel size. The decrease in spot size with increasing optical power is offset by the reduced angular spread of the different wavelengths.
2.4 Quantum dot preparation
We obtained CdSe quantum dots from Invitrogen (QDot ITK 625 carboxyl quantum dot) that fluoresce at a central wavelength of 620 nm with a FWHM of 20 nm, as extracted from provided linear excitation fluorescence curves. These quantum dots possess a core diameter of approximately 5 nm and are roughly 20 nm in diameter when including the outer functional group. Following a procedure adapted from Ref. 18, the quantum dots were covalently bonded to aminopropylsilane microarray slides (Pierce Protein Research Products). 11.23 mg MES buffer (2-(N-morpholino)ethane sulfonic acid, Sigma) was added to 10 mL deionized water, then 8.5 μL of this mixture was added to 8.5 mL deionized water to achieve the proper pH for the reaction that follows. The pH of this solution was not measured. A separate solution of sulfo-NHS (N-hydroxysuccinimide, Sigma) and EDC (1-ethyl-3-[3-(dimethylamino)propyl]carbodiimide hydrochloride, Sigma) was prepared by adding 10.44 mg NHS and 9.69 mg EDC to 2 mL deionized water and stirring. 500 μL of this solution was added to the 8.5 mL pre-buffered solution. 1 μL of 8 μM CdSe quantum dots was diluted in 99 μL deionized water. 1 μL of this diluted solution was added to the 8.5 mL mixture of MES, NHS, and EDC. A small stir bar was used to agitate the solution, and 1” square pieces of the aminated slides were placed into the solution for varying amounts of time. The sample used in this experiment was exposed for 16 minutes in solution before removal. All samples were washed three times with deionized water, and once with acetone and then toluene, and finally dried with a nitrogen jet.
3. Data and results
Initial tests with the prism-based spectrometer were performed with a vastly more dense batch of CdSe quantum dots than the covalently bonded sample. This second sample was made by simply placing a single drop of 8 μM CdSe quantum dots on a 170 μm glass coverslip, then covering the sample with another coverslip. This drop-cast sample was allowed to dry, then was illuminated with 5 mW average power at the focus of the objective by the Yb:KGW laser source.
An algorithm was written in Matlab to extract the spectra in a few steps. First, the algorithm imports the image and trims along the spatial direction (vertical) until only ~10 extra rows remain on either side of the spectrum. No columns are removed, so the entire wavelength range of the spectrum is maintained. The algorithm computes the average value in each column from these extra rows – rows preserved to provide a measure of background counts – then subtracts background noise from the entire image. This background is a combination of stray light, thermally generated electrons, shot noise and read noise from the camera. Any values less than zero are set to zero. The extra rows are removed, and columns summed to generate a 1D array. A running average filter is applied to the 1D array, with a window size of 3 pixels. The results of this procedure are shown in Fig. 7 , the spectrum of a 2 ms exposure of the drop-cast quantum dots.
In order to examine single emitters, the sparsely populated covalently bonded sample was imaged with the PMT detecting TPEF. The stage translated in a stepwise fashion, with a pixel dwell time of 10 ms, and 5 mW average power at the focus of the objective. The widefield image is shown in Fig. 8 – as with all images acquired with the PMT in our system, the image is scaled directly in photon counts. One 3 mm thick KG3 and one ThorLabs FMF005 filter were used to filter out the IR source and pass the TPEF. The maximum count scale of the image has been reduced in order to make the single emitters more easily observable.
The bright spot circled in Fig. 8 was illuminated for 30 s with the “short arm” of the optical multiplexer blocked and a spectrum acquired. Another 30 s exposure was taken after blocking the “long arm.” The results of each exposure are contained in Fig. 9 . A marked spectral shift and broadening are observed between the two excitation polarizations. The magnitude of the spectral shift is consistent with observed values for CdSe quantum dots .
The turning mirror was replaced to direct the signal into the PMT, and two sequential 100 s time-series were taken, with a bin width of 10 ms. Total (100%) amplitude modulation was observed – as shown in Fig. 10 – suggesting this bright spot was a single emitter.
We have constructed a prism-based spectrometer with single-emitter sensitivity and up to 0.3 nm spectral resolution. Additionally, this passive spectrometer provides straightforward adjustment of central wavelength and spectral range. The simple design allows for straight-forward implementation on multifocal, multiphoton microscope systems. Further refinements could include electronic controls for prism rotation and prism-lens/camera pair distances, providing a complete single computer driven interface. Additionally, a water-cooled EM-CCD capable of reaching −90°C would reduce the thermal background sufficiently to count single photons. Modeling of the system has yielded further insight into the optimum camera position to improve spectral resolution. Lastly, we have observed spectral shifts as a function of excitation polarization from single CdSe quantum dots.
It is straightforward to extend the functionality of this spectrometer to spectral imaging without descanning the detection pathway in two ways. First, the movement of the sample can be synchronized to the camera exposure, so that a spectrum is acquired at each sample position. A second, high-speed approach involves involves using an input scan mirror to sweep the excitation beam along a direction aligned to the vertical axis of the camera during each exposure. Such a line scan would provide a series of spectra for each column of the image. The sample is then stepped in the horizontal direction for each subsequent column.
This material is based upon work supported by the National Science Foundation through the Renewable Energy Materials Research Science and Engineering Center under Grant No. DMR-0820518. This work was partially funded by the National Institute of Biomedical Imaging and Bioengineering (EB-003832 J.A.S.).
References and links
2. G. Patterson, R. N. Day, and D. Piston, “Fluorescent protein spectra,” J. Cell Sci. 114(5), 837–838 (2000).
3. B. Barstow, N. Ando, C. U. Kim, and S. M. Gruner, “Alteration of citrine structure by hydrostatic pressure explains the accompanying spectral shift,” Proc. Natl. Acad. Sci. U.S.A. 105(36), 13362–13366 (2008). [CrossRef] [PubMed]
5. C. B. Murray, D. J. Norris, and M. G. Bawendi, “Synthesis and Characterization of Nearly Monodisperse CdE (E = S, Se, Te) Semiconductor Nanocrystallites,” J. Am. Chem. Soc. 155, 8706–8715 (1993). [CrossRef]
6. B. O. Dabbousi, J. Rodriguez-Viejo, F. V. Mikulec, J. R. Heine, H. Mattoussi, R. Ober, K. F. Jensen, and M. G. Bawendi, “(CdSe)ZnS Core-Shell Quantum Dots: Synthesis and Characterization of a Size Series of Highly Luminescent Nanocrystallites,” J. Phys. Chem. B 101(46), 9463–9475 (1997). [CrossRef]
7. R. D. Schaller, M. A. Petruska, and V. I. Klimov, “Tunable Near-Infrared Optical Gain and Amplified Spontaneous Emission Using PbSe Nanocrystals,” J. Phys. Chem. B 107(50), 13765–13768 (2003). [CrossRef]
8. H. R. Morris, C. C. Hoyt, and P. J. Treado, “Imaging Spectrometers for Fluorescence and Raman Microscopy: Acousto-Optic and Liquid Crystal Tunable Filters,” Appl. Spec. 48(7), 857–866 (1994). [CrossRef]
9. M. T. E. Golay, “Multi-Slit Spectroscopy,” J. Opt. Soc. Am. A. 39(6), 437–444 (1949). [CrossRef]
10. R. J. Bell, Introductory Fourier Transform Spectroscopy. London: Academic Press, 1972.
11. J. E. Chamberlain, The principles of Interferometric Spectroscopy. New York:Wiley, 1979.
12. Z. Malik, D. Cabib, R. A. Buckwald, A. Talmi, Y. Garini, and S. G. Lipson, “Fourier transform multipixel spectroscopy for quantitative cytology,” J. Microsc. 182(2), 133–140 (1996). [CrossRef]
13. Y. Garini, M. Macville, S. Manoir, R. A. Buckwald, M. Lavi, N. Katzir, D. Wine, I. Bar-Am, E. Schröck, D. Cabib, and T. Ried, “Spectral karyotyping,” Bioimaging 4(2), 65–72 (1996). [CrossRef]
16. D. W. Warren, and J. A. Hackwell, “Compact prism spectrograph suitable for broadband spectral surveys with array detectors,” U.S. Patent No. 5127728 (1992).
17. E. Chandler, E. Hoover, J. Field, K. Sheetz, W. Amir, R. Carriles, S. Ding, and J. Squier, “High-resolution mosaic imaging with multifocal, multiphoton photon-counting microscopy,” Appl. Opt. 48(11), 2067–2077 (2009). [CrossRef] [PubMed]
19. K. T. Early, P. K. Sudeep, T. Emrick, and M. D. Barnes, “Polarization-driven Stark shifts in quantum dot luminescence from single CdSe/oligo-PPV nanoparticles,” Nano Lett. 10(5), 1754–1758 (2010). [CrossRef] [PubMed]