Anal. Chem. 2009, 81, 7309–7313
Chromatically Resolved Optical Microscope (CROMoscope): A Grating-Based Instrument for Spectral Imaging Michael R. Webb,† Christopher N. LaFratta, and David R. Walt* Department of Chemistry, Tufts University, 62 Talbot Ave., Medford, Massachusetts 02155 The chromatically resolved optical microscope (CROMoscope) is capable of spectral imaging with tunable spectral and spatial resolutions. Because of its remarkably simple design, the CROMoscope can be easily assembled and aligned. Spectral resolution as low as 2.5 nm full width at half maximum (fwhm) was measured using an atomic emission line of Hg. Absorption spectra of different parts of a micrograph can readily be compiled using white-light illumination. Chloroplast absorption from an Elodea plant leaf was used to demonstrate this capability. Spectral imaging is widely applicable to many areas of science, and the CROMoscope is particularly simple to adapt to conventional microscopes and should enable detailed spectroscopic information to be obtained from microscopy. Analytical science is often advanced by combining techniques. Such mergers have been called tandem, hyphenated, or multidimensional. A common example of this combination is the coupling of separations with spectrometry (e.g., gas chromatography-mass spectroscopy (GC-MS)), which allows analyte quantifications and identifications in complex mixtures and with a specificity that is not achievable by a single method. Another multidimensional method is spectral imaging (also called multispectral or hyperspectral imaging) and is a combination of spectrometry and imaging. Often, the data are described as a cube where two dimensions are spatial and one is spectral. Within the field of microscopy, spectral imaging has been applied to a variety of objects, such as biological cells,1-3 karyotypes,4 and pharmaceutical tablets.5 Spectral imaging can be accomplished through spatial scanning, spectral scanning, or multiplexing. In spatial scanning, spectra are acquired at all wavelengths simultaneously but at a limited number of spatial positions, because of time constraints. * Author to whom correspondence should be addressed. E-mail address:
[email protected]. † Current address: Department of Chemistry and Biochemistry, University of North Carolina Wilmington, 601 S. College Road, Wilmington, North Carolina 28403. (1) Frank, H. H.; Elder, A. D.; Swartling, J.; Venkitaraman, A. R.; Jeyasekharan, A. D.; Kaminski, C. F. J. Microsc. 2007, 227, 203–215. (2) Hanley, Q. S.; Verveer, P. J.; Jovin, T. M. Appl. Spectrosc. 1998, 52, 783– 789. (3) Wachman, E. S.; Niu, W.-h.; Farkas, D. L. Biophys. J. 1997, 73, 1215– 1222. (4) Schro ¨ck, E.; du Manoir, S.; Veldman, T.; Schoell, B.; Wienberg, J.; FergusonSmith, M. A.; Ning, Y.; Ledbetter, D. H.; Bar-Am, I.; Soenksen, D.; Garini, Y.; Ried, T. Science 1996, 273, 495–497. (5) Sˇasˇic´, S. Appl. Spectrosc. 2007, 61, 239–250. 10.1021/ac9011655 CCC: $40.75 2009 American Chemical Society Published on Web 08/04/2009
The positions are scanned between consecutive measurements, and a series of measurements are appended to form the data cube. In spectral scanning, all spatial positions are observed simultaneously, but for only a single spectral position (wavelength) at a time. The wavelength is scanned between consecutive measurements and a series of measurements is appended to form the data cube. In multiplexing, techniques such as Fourier4 or Hadamard2 transformations are applied to at least one of the spectral or spatial dimensions. A series of measurements is required to reconstruct the transformed dimension. None of these approaches is clearly superior for all applications, but spectral scanning allows a flexibility that can be advantageous for many analytical microscopy applications. For example, quantifying n components in a mixture that contains only those components requires only n spectral windows if there is adequate spectral separation between each component. Limiting the analysis to n wavelengths, which is easily done in spectral scanning, has advantages. For example, time can be saved by not observing other wavelengths. Alternatively, the total analysis time can be kept the same but the integration time for each individual wavelength can be lengthened, thereby improving sensitivity. Furthermore, spectral scanning allows the integration time at each wavelength to be individually optimized, which is advantageous when the intensities are vastly different at the different wavelengths. Spatial scanning and transform techniques do not allow this flexibility. In fact, the multiplex disadvantage introduces noise from large signals to measurements of smaller signals. The multiplex disadvantage is an effect where noise from a large signal at one wavelength or position interferes with the measurement of small signals at other wavelengths or positions. It is an inherent consequence of Fourier and Hadamard transform techniques. Filters are often used for spectral scanning in microscopy. A filter is used to select a spectral band, and a two-dimensional image is obtained through it. Multiple spectral bands are interrogated using multiple filters. At some expense, custom filters can be acquired for any desired wavelength, but there are practical limits to the number of filters that can be installed on a microscope at one time. Greater flexibility can be attained through the use of an acousto-optic tunable filter (AOTF) or liquid-crystal tunable filter (LCTF). In AOTFs, light is diffracted by a propagating acoustic wave in a suitable crystal. An advantage of AOTFs is that the wavelength can be switched within less than a millisecond;3 however, they do have some limitations. AOTFs are expensive and complex. The bandpass of the diffracted light is strongly Analytical Chemistry, Vol. 81, No. 17, September 1, 2009
7309
dependent on the wavelength.6 In imaging applications, two aberrations are observed with AOTFs: image blur and image shift.7 A well-designed AOTF and the addition of an appropriate prism can compensate for most of these aberrations. Finally, AOTFs also can transmit out-of-band light3 and sidebands,1 which can be detrimental in many fluorescence and scattering experiments. Several designs of LCTFs exist, all with similar working principles. A LCTF uses birefringence to separate light into a fast component and a slow component. Constructive and destructive interference are used to pass one wavelength while blocking others. The central passed wavelength is varied by tuning the birefringence. The time required to switch wavelengths is slower than that of an AOTF (typically 50 ms or more) but is still quite fast.8 With LCTFs, the bandpass is not readily adjustable and is inversely proportional to wavelength,8 the passed wavelength can be shifted by thermal effects,8 and out-of-band light can present problems.9 An alternative approach that addresses these problems is to use a modified diffraction-grating-based monochromator to act as a filter, so that sequential monochromatic images can be acquired. Monochromators have high rejection of out-of-band light, have high and easily variable spectral resolution, and can be fairly inexpensive. Normally, they are not used for two-dimensional imaging but it is possible to do so. Previously described instruments of this type include slitless spectrographs,10-13 the monochromatic imaging spectrometer,10 and Delhaye and Dhamelincourt’s Raman microscope.14 Only the last of these instruments was explicitly designed for microscopy. In a slitless spectrograph, the entrance slit is opened very wide and an image is passed through it. The resulting image on the spectrograph’s focal plane contains a convolution of spectral and spatial information. Because of their design, slitless spectrographs are most useful for looking at spectra of objects that appear effectively as point sources11,12 or for looking at spatial information from sources of narrow-band emission.10,13 The monochromatic imaging spectrometer (MIS) consists of a Czerny-Turner monochromator with two external lenses.10 The image of the object is not strongly affected by the monochromator’s slits, because it is totally out of focus (collimated) as it passes through them. The Raman microscope of the type described by Delhaye and Dhamelincourt14 operates similarly to a MIS but uses a concave-grating monochromator. In this case, defocused light passing through the entrance slit will not be wholly defocused at the exit slit, so the arrangement of external lenses has more stringent requirements10 and some convolution may occur between spectral and spatial information. All three of these designs suffer from an aberration with the same root cause. In most monochromator designs, light diffracts off a grating at an angle different from the incident angle. In all (6) Bei, L.; Duffin, K. L.; Carnahan, J. W. J. Anal. At. Spectrom. 2004, 19, 1151–1157. (7) Voloshinov, V. B.; Yushkov, K. B.; Linde, B. B. J. J. Opt. A: Pure Appl. Opt. 2007, 9, 341–347. (8) Gat, N. In Wavelet Applications VII; Szu, H. H., Vetterli, M., Campbell, W. J., Buss, J. R., Eds.; Proceedings of the SPIE, Vol. 4056, Orlando, FL, April 26-28, 2000; SPIEsInternational Society for Optical Engineering: Bellingham, WA, 2000; pp 50-64. (9) Stratis, D. N.; Eland, K. L.; Carter, J. C.; Tomlinson, S. J.; Angel, S. M. Appl. Spectrosc. 2001, 55, 999–1004. (10) Olesik, J. W.; Hieftje, G. M. Anal. Chem. 1985, 57, 2049–2055. (11) Orville, R. E. Science 1966, 151, 451–452. (12) Isailovic, D.; Li, H. W.; Phillips, G. J.; Yeung, E. S. Appl. Spectrosc. 2005, 59, 221–226. (13) Horlick, G.; Furuta, N. Spectrochim. Acta, Part B 1982, 37, 999–1008. (14) Delhaye, M.; Dhamelincourt, P. J. Raman Spectrosc. 1975, 3, 33–43.
7310
Analytical Chemistry, Vol. 81, No. 17, September 1, 2009
Figure 1. Diagram of the spectroscopic imaging microscope (approximately to scale). Dashed lines show several light paths, and the red line shows the principal ray. The object plane (A), diffraction grating surface (A′), and detector surface (A′′) are conjugate planes. The monochromator entrance (B) and exit (B′) apertures are conjugate planes.
three of these instruments, this diffraction results in a final image that is either stretched or compressed in the horizontal dimension in a wavelength-dependent fashion. In all three designs, it is possible to add a second monochromator to reverse the distortion;15-17 however, this correction method complicates the design, reduces light throughput, introduces additional opportunities for distortions to be introduced into the image, and increases the cost of the system. In this paper, we present a chromatically resolved optical microscope (CROMoscope), using a single plane grating. Our design uses an out-of-plane Littrow configuration to avoid this distortion. In the diffracting axis (the axis parallel to the grating’s grooves), the light is retroreflected. The grating is tilted in the nondiffracting axis (that is, the axis perpendicular to the grooves), so that the incident and reflected light are separated in this axis. Because it is designed specifically for imaging, other modifications not found in a nonimaging spectrometer have been made. For example, the internal optics are arranged such that collimated light entering the monochromator portion of the instrument remain collimated when it exits. This design is simple and escapes some of the shortcomings of previous optical imaging spectrometers. The CROMoscope is also capable of acquiring absorption spectra using bright-field imaging and has been used to collect such spectra from colored samples, including individual organelles in living cells. CROMOSCOPE DESIGN A diagram of the chromatically resolved optical microscope (CROMoscope) is shown in Figure 1, and a photograph of it is shown in Figure 2. The instrument can be viewed as a microscope with a monochromator inserted in the “infinity space” where filters would often be placed. The optics are designed such that the image of the sample is completely defocused (focused at infinity) both at the entrance and at the exit of the monochromator. Because of this design, the apertures have little effect on the image. However, small apertures will reduce the numerical aperturesand, thus, the spatial resolutionsof the system. After the monochromator portion of the instrument, a lens forms a focused image of the sample on the detector. A more-detailed description follows. (15) Allemand, C. D.; Brewer, D. L. U.S. Patent 4,455,087, 1984. (16) Webb, M. R.; Hieftje, G. M. Appl. Spectrosc. 2006, 60, 57–60. (17) Dhamelincourt, P.; Wallart, F.; Leclercq, M.; Nguyen, A. T.; Landon, D. O. Anal. Chem. 1979, 51, A414–A421.
Figure 2. Photograph of the spectroscopic imaging microscope. Red line shows the approximate path of the principal ray. A, object; B, microscope objective; C, aperture; D, mirror; E, lens; F, diffraction grating; G, lens; H, aperture; I, camera optics; and J, camera.
In the studies described here, sample illumination is provided by a halogen lamp (Cole-Parmer Model 41723), with light brought to the sample using a fiber-optic bundle and a single 25mm diameter, 30-mm focal length lens. This configuration equates to a fixed numerical aperture of NA 0.4. Future versions of the instrument will have illumination with an adjustable NA, as is the case for traditional microscopes. Light from the sample is collected by a microscope objective. As with a typical microscope, this objective can be replaced to achieve different levels of magnification. The CROMoscope is designed to be used with infinity-corrected objectives. The defocused light passes through an adjustable iris. The iris serves as the entrance aperture of the monochromator portion of the instrument. A mirror is used to direct the light to lens 1. This lens has a focal length of 300 mm and is positioned that distance from the entrance aperture. On the far side of lens 1, also at a distance of 300 mm, is a diffraction grating with 1200 grooves/ mm. An enlarged image of the sample is formed on or near the grating. The grating is tilted in two axes. One axis, which runs in and out of the page as the instrument is depicted in Figure 1, is set at a fixed tilt of ∼6.3°. The tilt of the other axis is variable and is adjusted to select the wavelength passed by the instrument. Lens 2 is one focal length (300 mm) from the diffraction grating, so the sample image is once again focused at infinity after this lens. Another adjustable iris is one focal length away from the lens, on the opposite side from the grating. This iris serves as the exit aperture of the monochromator portion of the system, and a one-to-one image of the entrance aperture is projected onto it. A lens after the aperture focuses the monochromatic image of the sample onto a CCD camera. We used three lens-camera combinations in this work: a Unibrain Fire-i400 camera with a 70 mm focal length lens (Schneider Tele-Xenar 2.2/70 mm Compact C-Mount), a Hamamatsu ORCA-II BT 512G camera with a 160-mm focal length lens (Olympus U-TLUIR), and a Canon PowerShot A530 camera with incorporated optics. Most monochromators use off-axis concave mirrors, which result in astigmatism (a distortion). Because the CROMoscope uses on-axis lenses instead, this astigmatism is not present. Mirrors are vital to avoid chromatic aberrations in the UV region, but achromatic lenses are available in the visible to near-infrared regions. Because we used such lenses, chromatic aberrations are negligible in this system.
CROMOSCOPE SPECTRAL AND SPATIAL RESOLUTION CHARACTERIZATION The wavelength calibration and background rejection were measured using a miniature fiber-optic spectrometer (Ocean Optics Model USB4000). The variable tilt that determines the passed wavelength is controlled by a sine bar mechanism, which converts the linear motion of a micrometer into a sinusoidal motion of the grating. This motion leads to an almost-linear relationship between the micrometer position and the wavelength passed by the instrument. To calibrate the wavelength scale, a fiber optic connected to a spectrometer was placed at the position normally occupied by the CCD. The rest of the instrument, including the 70-mm focal length lens, was left intact. The wavelength was measured for 26 grating positions corresponding to wavelengths between 372 and 1007 nm. In principle, the CROMoscope is capable of a wider wavelength range, but low efficiency of both the spectrometer and the halogen lamp prevented calibration outside of this range. The relationship is well-fit (R2 ) 0.9999) by a second-order polynomial having the equation y ) -0.181x2 + 31.1x + 273 where x is the micrometer reading (given in millimeters) and y is the wavelength (given in nanometers). Background rejection or stray light was measured using the same arrangement, with the CROMoscope set to 598 nm, and the entrance and exit apertures at 4 mm. Ten signal (lamp on) and ten background (lamp off) spectra were acquired, and the average background was subtracted from the average signal. Stray light was quantified as the ratio of off-peak intensity to peak intensity. This ratio actually measures a combination of the stray light of the CROMoscope and the stray light of the Ocean Optics spectrometer. At 25 nm offpeak (573 and 623 nm), the stray light was 0.06% ± 0.03%. At 50 nm offpeak (548 and 648 nm), the stray light was 0.03% ± 0.03%. Per the manufacturer, the stray light of the Ocean Optics spectrometer is “less than 0.05% at 600 nm”, so the contribution of the CROMoscope may be significantly less than the measured values. The spectral resolution of the CROMoscope is proportional to the iris aperture and was measured using the 546-nm line from a mercury lamp (UVP, Inc., Model UVG-11). Images of the lamp were acquired at wavelengths in the range from 500-600 nm, using the Fire-i400 and 70 mm lens. The intensities of all the pixels in an image were summed. Plots of these summed intensities values versus wavelength produced peaks that indicated the bandpass of the system (see Figure 3, inset). The bandpass of the CROMoscope for a given iris aperture was determined from the full width at half maximum (fwhm) of the intensity-versus-wavelength plots. The process was repeated for several different aperture diameters. As Figure 3 shows, the bandpass is linearly dependent on the aperture diameters and the highest spectral resolution (2.4 nm) was measured for 1-mmdiameter apertures. This spectral resolution is slightly better than would be expected from an otherwise identical monochromator with a 1-mm-wide rectangular slit. As with traditional microscopes, the spatial resolution of the CROMoscope is highly dependent on the NA of the microscope objective-iris combination. For large irises, the NA of the system will be the same as the NA of the microscope objective. For Analytical Chemistry, Vol. 81, No. 17, September 1, 2009
7311
Figure 3. Dependence of the CROMoscope bandpass at 546 nm on aperture diameters. Bandpass is defined as the full width at half maximum (fwhm) of the integrated image intensity-versus-wavelength plot; the entrance and exit apertures were equal. Inset shows the spectral profile used to determine the bandpass with 8-mm-diameter apertures.
Figure 5. Images of three inks, taken with the CROMoscope: (A) white light image taken at zero order, (B) chromatically resolved image at 505 nm, (C) chromatically resolved image at 575 nm, and (D) chromatically resolved image at 650 nm, all obtained with a 7.3-nm bandpass.
Figure 4. Demonstration of the effect of aperture size on spatial resolution. Traces show the cross sections of bars on a USAF 1951 resolution target. Each trace shows the same five sets of bars. Each set consists of three equally wide bars separated by spaces of the same widths. Bar sizes, from left to right, are 3.47, 3.11, 2.76, 2.46, and 2.19 µm.
smaller irises, the system NA and spatial resolution will be somewhat reduced. A USAF 1951 target was used to demonstrate this effect with a 0.5-NA 50× objective. The target consists of sets of three bars. The bars within each set are equally sized and are separated by gaps equal to their widths. Figure 4 shows cross sections of several groups of bars using three iris diameters. For 3- and 5-mm irises, the smallest bars (2.2 µm) are baselineresolved. For 1-mm irises, which correspond to a spectral resolution of 2.4 nm, the intensity in the valley between bars is ∼60% of the intensity at the centers of the bars. As expected, better spatial resolution is possible with a higher-NA objective. RESULTS AND DISCUSSION The CROMoscope enables the collection of monochromatic images with adjustable spatial and spectral resolution. Spectroscopic imaging can be performed by capturing multiple images at different wavelengths. Absorbance spectra for each pixel in the collection of images can then be compiled. To demonstrate the 7312
Analytical Chemistry, Vol. 81, No. 17, September 1, 2009
ability to collect localized absorption spectra, we examined two samples: the first sample was of colored inks on a piece of glass, and the second sample was of chloroplast-containing cells in an Elodea plant leaf. The micrographs in Figure 5 show the coverslip sample with red, green, and blue ink drawn on it. The CROMoscope collected monochromatic images at 505, 575, and 650 nm, and the diffraction grating was rotated to zero-order to view the polychromatic image. Entrance and exit apertures were 3 mm for all four of the images shown in Figure 5. The monochromatic images were taken with a black-and-white camera (Hamamatsu ORCA-II BT 512G), and the color image was taken with a Canon PowerShot A530. To obtain spectra of the sample, 46 monochromatic images were taken, and each image was analyzed in three locations to obtain the absorption spectra for each of the three inks. Approximately 10 min were required to obtain the images required for the spectra manually; however, this length of time likely could be reduced to seconds by automating the rotation of the grating and image acquisition. The central region without ink was used as a “blank” for the correction of the lamp output and the grating dispersion efficiency. A clean glass slide could also have been used as a blank, but the built-in control was more convenient. Separate dark- and flat-field images were also acquired. The 3-mm-diameter apertures correspond to a spectral resolution of 7.3 nm. Figure 6A shows the absorption spectra obtained with the CROMoscope. These data closely match the ultravioletvisible light (UV-vis) absorption spectra of coverslips coated with only one color of ink obtained using conventional instrumentation (Beckman Model DU 530) (see Figure 6B). Another minor point worth noting is that the images show a very slight twisting (0.15 ± 0.02 mrad/nm), as a function of wavelength. This shift is caused by the small tilt in the grating and the change in optical path length as the grating is rotated. Over the entire visible spectrum, the image rotation is
