Third harmonic generation microscopy is used to make dynamical images of living systems for the first time. A 100 fs excitation pulse at 1.2 μm results in a 400 nm signal which is generated directly within the specimen. Chara plant rhizoids have been imaged, showing dynamic plant activity, and non-fading image characteristics even with continuous viewing, indicating prolonged viability under these THG-imaging conditions.
©1998 Optical Society of America
The phenomena of third harmonic generation (THG) under tight focusing conditions has recently been explored by Tsang , who demonstrated that, through the presence of an interface within the focal volume of the excitation beam, third harmonic production is allowed. Normally, tight focusing conditions inhibit the production of the third harmonic , requiring a positive wave vector (Δk > 0) mismatch between the fundamental and harmonic beam to achieve phasematching. In the case of an interface as considered by Tsang, no such mismatch was required, the resulting breakage in focal volume symmetry enabling efficient generation for Δk = 0. This can be understood by calculation of the third harmonic power as a function of the interface uniformity as done by Barad et al , who show that when there is either a change in refractive index or third-order nonlinear susceptibility, the third harmonic power is nonvanishing. Barad et al  demonstrate that as a result of this interface effect imaging with the third harmonic is possible and especially suitable for transparent specimens with low intrinsic contrast, and is sensitive to changes in the specimen’s nonlinear optical properties. Independently, we had also been exploring the imaging potential of the third harmonic and demonstrated that volumetric imaging was possible using this technique in both biological and non-biological specimens [4,5]. In these studies, another dependency of the third harmonic was revealed: the image contrast varied according to the relative orientation of the interface within the focus with respect to the excitation beam’s direction of propagation. In the present paper, the characteristics of third harmonic imaging are further developed. Specifically, we begin by examining the conversion efficiency, power scaling, and sectioning of the third harmonic signal within the microscope where different excitation and collection numerical apertures (NA’s) are employed. Next, we revisit image contrast as a function of surface orientation under point excitation (our previous studies employed line cursor excitation ). Finally, we demonstrate for the first time to our knowledge, dynamical imaging of live specimens, under moderate NA (0.6) and high NA (1.3) conditions. While significant work remains to fully develop THG microscopy into a useful tool as applied to biological systems, this first demonstration is critical in that it begins to help establish the relevance of the technique to a living specimen. We note that continuous viewing of the specimen was possible over hours without reduction in image intensity. This in contrast to three-dimensional fluorescence imaging techniques (either single, or multi-photon absorption), in which the fluorophores that label the specimen bleach in volume for single-photon excitation, or by section for two- or three- photon absorption.
The imaging system used in this study is a modified, inverted Leitz fluorescence microscope, depicted in Figure 1. The excitation beam is at 1.2 μm, 250 kHz, 100 fs, with an average power of 18 mW which can be variably attenuated by a ND filter wheel located in front of the beam scanners. The scanners are run asynchronously at variable rates resulting in a traveling Lissajou illumination pattern, which produces a near uniform object illumination. The excitation beam is delivered through the lower part of the microscope -- thus the inverted objective is the excitation objective. The 400 nm light emitted from the sample is captured by the collection objective in transmission. The transmitted beam is sent through a BG39 blocking filter and/or an interference filter (both of which serve to block the fundamental wavelength), after which it is imaged to the CCD camera (Hamamatsu, model C5985).
Tsang measured the efficiency of third harmonic production for a variety of interfaces [1,6]. The conversion efficiencies ranged from 10-7 to 10-10 at 100–300 GW/cm2. We measure conversion efficiencies in the 10-7–10-8 range in the aforementioned set-up for excitation in the 300–1000 GW/cm2 regime at the glass-air interface. These measurements include the transmission efficiencies of the excitation (0.6 NA) and collection (0.4 NA) objectives, the transmission of the BG39 filter at 400 nm, and the quantum efficiency of the detector. The collection efficiency is taken as ~100% which assumes that the NA of the emission is 1/3 of the excitation by virtue of the wavelength tripling, and that negligible scattering and Fresnel losses occur within the coverslip used to make these measurements. The power scaling law of the signal was continuously checked for a variety of imaging conditions and specimens. The power scaling curve and axial sectioning curve for a high NA excitation and collection objective combination is shown in Figure 2. The power law scales by the expected power of three, within the experimental error, as shown in the upper graph.
Optimization of the third harmonic signal as a function of the laser pulse duration, energy, and repetition rate differs from that of two photon absorption due to the instantaneous response time of the signal, and the cubic as opposed to quadratic intensity dependence. Specifically, the two photon fluorescence signal will scale linearly with pulse duration, the third harmonic scales quadratically (assuming a square temporal pulse profile and non-saturating conditions). In terms of energy dependence the two photon fluorescence signal shows a quadratic gain, the third harmonic a cubic gain. Both signals scale linearly with repetition rate, with the two photon signal eventually rolling over due to the fluorescence lifetime of the material. However, due to the instantaneous nature of the third harmonic, the repetition rate can be scaled to an essentially arbitrarily high value. In the imaging studies that follow, the pulse duration and repetition rate were fixed by the laser source used to generate the infrared wavelengths necessary for THG imaging and were not necessarily optimal. Energies were kept as high as possible without causing visible specimen damage. In general in THG imaging it is only necessary to consider the dispersion of the excitation objective and not the collection objective. Fortunately, at these wavelengths dispersion is not expected to significantly broaden the pulse. For instance, a 1.2 μm, 100 fs pulse only broadens by ~10% for 2 cm of SF10 glass, while a 800 nm, 100 fs pulse broadens by slightly more than 30%.
In  a line cursor excitation was employed whereas in this study, point excitation is used to achieve maximum sectioning capability. In order to verify the same surface orientation dependent contrast for this system, a specimen containing 250 μm glass spheres in immersion oil was used. Note that there is only a small change in refractive index between the glass spheres and the immersion oil. This image series is shown in Figure 3. Each section represents an axial step of 2.5 μm. The excitation objective was a Zeiss Plan Apochromat 20x, 0.6 NA air objective, and the collection objective was an Olympus 20x, 0.4 NA air objective. The input power to the microscope for this image series was 15 mW. The measured transmission efficiency to the sample was 50%, resulting in an average power of 7.5 mW (i.e., 30 nJ per pulse) at the sample. For these parameters, the excitation intensity is then on the order of 1012 W/cm2, which is very comparable to, for instance, the typical 5 1011 W/cm2 used for two-photon absorption microscopy (1.3 NA, 100 fs, 0.1 nJ/pulse). The first image shows the back surface of the front coverslip, which recedes and gives way to the surface contours of the sphere as the specimen is stepped axially through the image plane. Note that as the sectioning proceeds through to the center of the sphere, the contrast of the image decreases. The excitation power, image integration times, and camera gain are all held constant in this image series. As the sectioning passes through the center of the sphere, the image intensity visibly increases. The series concludes at the inner surface of the second cover slip. The second series of rings is not as sharp as the first series -- the excitation light in this case must be imaged through the sphere itself, which results in visible defocusing and aberration of the imaged contour. By replacing the 400 nm interference filter with a 600 nm interference filter, we checked for possible second harmonic generation. None was observed within the signal-to-noise of the detection system. This simple image series serves to illustrate the rather complex contrast behavior of THG imaging. Clearly, further work remains to fully quantify and identify the contrast mechanisms in THG microscopy, especially for use with more complex specimens.
The utility of THG imaging with live specimens was tested by imaging rhizoids from Chara plants. The rhizoids - which are tubular, single cells forming the roots of the green alga Chara - have been studied widely, especially with respect to their response to gravity . Within the cell is a strong cytoplasmic streaming to and from the rhizoid tip. The tip contains so-called statoliths, which are vesicles containing BaSO4 crystals. Without the statoliths no graviresponse is observed. The statoliths are linked to an actin filament network, preventing them from precipitating on the lower cell wall. The various features of the Chara rhizoids have been checked using conventional phase contrast microscopy. Once again the excitation objective was the 20x, 0.6 NA Zeiss Plan-Apochromat and the collection objective was the 20x, 0.4 NA Olympus. In general these images where made with only the BG39 filter in place. For all images featuring the Chara rhizoid there was approximately 1.2 mW average power at the sample. The emission wavelength was periodically checked by adding an interference filter centered at 401.2 nm, with a full width half maximum (FWHM) of 16 nm, and net transmission efficiency of 40%. The removal of the filter simply resulted in a decreased camera integration time, and enabled more rapid image acquisition. The frame integration times were 9 seconds for the sectioned image series, each axial step being 1 μm (Figure 4). Notably the sectioned image is generated without the use of any labeling molecules as would be required in traditional laser fluorescence confocal microscopy. Any section could be viewed for extended periods (hours) without any reduction in image intensity, demonstrating the non-fading nature of this imaging technique. Additionally, the third-order power dependence of the rhizoid images was verified, using the imaged signal. For instance, an introduction of 0.3 ND into the excitation beam requires an increase in image integration time of 8 times to produce the identical image intensity. Finally, the polarization dependence of the signal was verified by introduction of a polarizer within the collimated beam path after the collection objective. Rotation of the polarizer results in significant amplitude modulation, indicating the highly polarized nature of the generated signal. Thus, the combination of wavelength (no Stokes-shift), third-order power dependence, and strong polarization dependence are strong indicators that the signal is indeed third harmonic and is not the result of autofluorescence.
A more dynamical application of THG imaging is demonstrated in the next several image series. Figure 5 is centered at the middle of the rhizoid (axially), and below the tip. The cytoplasmic streaming is clearly visible. In Figure 6 the rhizoid tip is imaged, and the movement of statoliths located within the tip is seen. The combination of statolith movement and continual cytoplasmic streaming are indicative of specimen vitality throughout the imaging process. An image series of the Chara rhizoid was also performed under high NA conditions. Using a Zeiss Plan Neofluar 100x/1.3 oil immersion excitation and Zeiss Plan Neofluar 63x/1.25 oil immersion collection objective, Figure 7 shows an image of the outer membrane at the rhizoid tip. As shown in Figure 2, the axial sectioning capability for these high NA conditions is ~.8 μm, FWHM. Further experiments are presently underway to fully quantify the imaging and resolution properties of THG microscopy.
A final image series (Figure 8) captures small, single cellular organisms swimming in and out of the excitation plane in a sample of pond water. The cross sectional image of a small root is apparent in these images and provides a fixed reference. The image integration time of one second was too slow to capture the rapidly moving organisms, and consequently they appear as elongated blurs as they move across and through the excitation plane.
In conclusion, third harmonic microscopy has been performed with living, dynamical specimens for the first time. Importantly, these systems continued to function throughout the imaging process. Even with hours of continual exposure, no loss of image contrast was noted. Many tasks remain at hand, however, to fully develop this technique into a reliable and useful tool for imaging both biological and non-biological specimens. For instance, much more work needs to be done to fully quantify the mechanisms responsible for image contrast. Further, no attempt was made in this study to vary the laser parameters to truly optimize the efficiency of the THG signal. For instance, recent work has shown that extremely short pulse durations can be produced at the focus of high NA systems . Clearly, the THG efficiency would benefit greatly from the use of shorter pulses. It is unclear however, how shorter pulses may affect the viability of a living system. Thus, a systematic study that varies pulse repetition rate, energy, and pulse duration in an attempt to optimize THG efficiency and cell viability is necessary.
We gratefully acknowledge the technical assistance in sample preparation and microscope modification by J. A. Grimbergen and J. A. W. Kalwij.
References and links
2. R. W. Boyd, Nonlinear Optics. (Academic Press, Boston, 1992).
3. Y. Barad, H. Eisenberg, M. Horowitz, and Y. Silberberg, “Nonlinear scanning laser microscopy by third-harmonic generation,” Appl. Phys. Lett. 70, 922–924 (1997). [CrossRef]
5. J. Squier, M. Müller, K. R. Wilson, and G. J. Brakenhoff, “3D-Microscopy using third harmonic generation at interfaces in biological and nonbiological specimens,” to appear in proceedings of Ultrafast Phenomena XI, Garmisch-PartenKirchen/Germany, July 12–17, 1998.
7. A. Sievers, B. Buchen, and D. Hodick, “Gravity sensing in tip-growing cells,” Trends in Plant Sci. 1, 273–279 (1996). [CrossRef]
8. M. Müller, J. Squier, R. Wolleschensky, U. Simon, and G. J. Brakenhoff, “Dispersion pre-compensation of 15 femtosecond optical pulses for high-numerical-aperture objectives,” J. Microsc. 191, 141–150 (1998). [CrossRef] [PubMed]