Laser spectral characterization in multiphoton microscopy Franco Quercioli, Bruno Tiribilli, Massimo Vassalli, and Alessandro Ghirelli
Spectral and temporal characterization is a fundamental task when a tunable Ti:sapphire ultrafast laser system is operated for multiphoton microscopy applications. In the present paper simple procedures are reported that perform laser-peak-emission wavelength and bandwidth measurements without the need of any further instrumentation but a simple and inexpensive diffraction grating, by taking advantage of the confocal microscope imaging capabilities. © 2004 Optical Society of America OCIS codes: 180.1790, 120.6200, 140.4050, 170.0110, 190.4180.
1. Introduction
Multiphoton microscopy is becoming a widespread technique in biological applications.1– 4 The tunable Ti:sapphire mode-locked ultrafast laser system is the light source of choice for this kind of application. Laser central emission wavelength, bandwidth, and pulse duration are parameters that should be frequently measured,5–12 and a great choice of commercial testing tools is available for this purpose. However, a multiphoton microscope is a complex and huge device, and to purchase and manage further instrumentation, such as a spectrometer or an autocorrelator, which will be placed onto an already rather-crowded experimental table, is neither desirable nor necessary. The present paper will focus on source spectral characterization. Simple procedures will be described that make use of a diffraction grating placed onto the microscope specimen stage. The source spectral distribution will then be imaged simply by running the confocal acquisition system and will be displayed onto the instrument monitor. Peak wavelength and bandwidth are then trivially retrieved. The source temporal characterization is beyond the scope of the present paper. Bandwidth measurements will be used instead to compute a bandwidthlimited pulse width. Pulse duration widening due to
The authors are with the Istituto Nazionale di Ottica Applicata, Laboratorio di Biofotonica, Largo Enrico Fermi 6, 50125 Firenze, Italy. F. Quercioli’s e-mail address is
[email protected]. Received 3 October 2003; revised manuscript received 28 January 2004; accepted 18 February 2004. 0003-6935兾04兾153055-06$15.00兾0 © 2004 Optical Society of America
group velocity dispersion caused by the microscope optics will not be taken into account. Although this procedure will lead to a rather rough estimate of the source temporal behavior, it will nevertheless be useful in establishing the minimum pulse duration attainable. 2. Experimental Setup
The standard layout of a Confocal Multiphoton Microscope is shown in Fig. 1共a兲, and a picture of the instrument that has been set up in our laboratory is shown in Fig. 1共b兲. A PCM2000 Confocal Laser Scanning Microscope 共CLSM, Nikon, Tokyo, Japan兲, equipped with a Nikon TE2000-U inverted optical microscope has been modified to allow the use of an ultrafast laser source.13 For this purpose, one of the standard long-pass input excitation dichroic filters, required for one-photon excitation, has been replaced in the confocal head with a short-pass one, suitable for multiphoton operation. A cutoff at 650 nm, well beyond the source wavelength range, has been chosen 共650DCSPXR, Chroma Technology Corp, Brattelboro, Vermont兲 to assure maximum extinction of the excitation light into the fluorescence channels. The laser system is made up of a modelocked Ti:sapphire oscillator 共Mira 900 F兲 pumped by a frequencydoubled Nd:YVO4 laser system at 532 nm, 5 W 共Verdi V5兲 from Coherent, Inc. 共Santa Clara, California兲. The tunable wavelength range is: 700 –980 nm 共with X-Wave broadband optics兲, and the typical pulse duration is 130 fs, assuming a squared secant hyperbolic intensity pulse shape. The entire system was placed on an antivibration optical table, and the ultrafast laser source was directly coupled to the confocal scanning unit, avoiding 20 May 2004 兾 Vol. 43, No. 15 兾 APPLIED OPTICS
3055
Fig. 1. 共a兲 Standard layout of a Confocal Multiphoton Microscope; 共b兲 Picture of the instrument set up in our laboratory. A Nikon PCM2000 CLSM, equipped with a Nikon TE2000-U inverted optical microscope, has been directly coupled with a Mira 900 F Ti:sapphire oscillator pumped by a Verdi V5 frequency-doubled Nd:YVO4 laser 共Coherent, Inc.兲.
the use of the CLSM optical fiber delivery system to minimize group velocity dispersion. Alternative laser sources for one-photon excitation are also available in our laboratory, among them a single-harmonic-generation frequency doubler system at 350 – 490 nm. The wavelength and bandwidth measuring techniques described in this paper can also be used for this kind of tunable source. A.
Wavelength Measurement
The basic setup for wavelength measurements is shown in Fig. 2. The main characteristic of the present arrangement is the absence of any objective, which has been removed from the optical path to allow the imaging of the far-field diffraction pattern produced by a grating placed on the microscope specimen stage. The CLSM coordinate system is defined in the following manner: The x and y axes lay along the fast and slow scanning directions, respectively, with the origin placed at the upper left corner of the field; the z axis coincides with the microscope optical axis. Figure 2 shows the plane y ⫽ 0. For the sake of simplicity, the grating orientation is chosen so that the groove’s direction is perpendicular to the x axis, and its normal is set parallel to the z axis. Without the presence of any objective, the collimated laser input beam is let to impinge directly onto 3056
APPLIED OPTICS 兾 Vol. 43, No. 15 兾 20 May 2004
Fig. 2. Basic setup for wavelength measurements. A grating is placed onto the microscope specimen stage. The absence of any objective allows the far-field diffraction pattern to be imaged by the CLSM.
the grating. The image of the back-diffracted laser light is then recorded. During acquisition, the laser beam scans the field angularly along the x and y directions and hits the grating surface at different angles. In the plane y ⫽ 0, ␣ is the scan angle between the projection of the incident laser beam direction and the z axis, whereas the corresponding angle in the perpendicular plane x ⫽ 0 共not shown兲, is . The scan angles ␣ and  are proportional to the x and y galvanometric mirror tilt angles, respectively. The proportionality factor is twice the angular magnification of the interfacing optics placed between the galvanometers and the objective 共not present in this setup兲. This is an afocal system made up of the scanning lens, inside the CLSM scanner module, and the microscope tube lens. In the detection optical path the diffracted light retraces back the incoming route, passes through the confocal pinhole and a small fraction 共but still well detectable兲 is transmitted through the 650-nm shortpass dichroic and, taking care to remove any barrier filter, it reaches the photomultiplier. The detected intensity versus the ␣ and  scan angles values is used by the instrument software to build a twodimensional image. For a generic scan angle value, none of the diffraction orders backward retraces the ongoing laser beam direction, as depicted in the example of Fig. 2. No
Fig. 3. Three scanning angles at which the Littrow condition is fulfilled for the ⫹1, 0, ⫺1 diffracted orders, respectively.
diffracted light is then detected by the CLSM photomultiplier, and the corresponding image point is dark. When, on the other hand, the Littrow condition is satisfied14 共that is, when the incident and diffracted rays are in autocollimation兲, the detected intensity value is high. Littrow condition holds when the scan angles ␣ and  satisfy the following equations: 2 sin ␣ ⫽ ng;
 ⫽ 0,
(1)
where n is the diffraction order integer and g is the grating spatial frequency. Figure 3 shows the three cases for which the Littrow condition is fulfilled, for ⫹1, 0, ⫺1 diffracted orders, respectively. The image produced by the instrument is made up of three bright spots on a dark background. The central spot, corresponding to the zero order, points out the scanning angles’ origin 共 zaxis direction兲, which will be right at the center of the image field. The angular coordinates for the three diffracted orders are ␣ 0 ⫽ 0, ␣ ⫾1 ⫽ ⫾g兾2,
 0 ⫽ 0,  ⫾1 ⫽ 0,
(2)
where the trigonometric functions are replaced by their arguments since, as we will see later on, the total scanning range allows for paraxial approximation to be applicable. Image data are displayed by the instrument using the linear coordinate system x and y, whose values are computed by the CLSM software with the following formulas: x ⫽ ␣f ob ⫹ x c,
Fig. 4. Experimental setup for bandwidth measurements. A blazed grating is used here and is again placed onto the microscope specimen stage. The grating normal is rotated of an angle ␣n with respect to the z axis to allow the central wavelength at the nth order to be diffracted back along the z axis. The diffraction angles at which Littrow condition is fulfilled at wavelengths and ⫾ ⌬ are shown.
y ⫽ f ob ⫹ y c.
(3)
The constant of proportionality fob represents the focal length of the objective, which is declared in the CLSM software settings. In this arrangement there is no objective, and neither fob nor the x, y coordinate values have any physical meaning here; only the scan angles ␣ and  are meaningful, xc, yc take into account the relative shift between the x, y coordinate system origin 共upper left corner of the image兲 and the ␣,  one 共center of the image兲. Any pair of the three diffraction orders could be used to calculate . Taking for example the first and zeroth order coordinates we obtain, from Eqs. 共2兲 and 共3兲, ⫽ 2共 x ⫹1 ⫺ x 0兲兾共 gf ob兲 ⫽ 共⌬x兾f ob兲共2兾g兲.
(4)
The difference ⌬x兾fob ⫽ 共x⫹1 ⫺ x0兲兾fob represents the angular distance between the two diffraction spots, which are located at the center of the image field 共the zeroth order兲 and are lined up along the x axis. If the grating alignment procedure previously described has not been correctly done 共that is, the grating normal is not parallel to the z axis or the grooves are not parallel to the y axis兲, the position and the orientation of the diffraction pattern would be different. This is not a drawback, however, since only distances between the diffraction orders are involved in the wavelength computing. B.
Bandwidth Measurement
Figure 4 shows the experimental set up for bandwidth measurements. The use of a high spatial frequency grating g or a higher diffraction order n is better suited for this kind of application in order to enlarge the diffraction extent and enhance the spectral resolution. As for all spectrometric applications, a blazed grating will be the better choice, although the abundance of input light available from the laser source make the selection of a highefficiency grating not mandatory. The arrangement is similar to the one previously described for wavelength measurements. The grat20 May 2004 兾 Vol. 43, No. 15 兾 APPLIED OPTICS
3057
Fig. 5. Typical image of the far-field diffraction pattern as it appears on the CLSM monitor. A 19.7-mm⫺1 spatial frequency Ronchi ruling has been used, a Ti:sapphire emission wavelength of 740 nm has been tuned, and the smallest 20-m diameter pinhole has been selected. With a suitable choice of the display software setting parameters, wavelength values can be directly read as distances between diffraction order pairs. The angular spread of the diffracted order spots is about 0.3 mrad FWHM, corresponding to a spectral broadening of ⬃30 nm.
ing is again placed onto the microscope specimen stage with its groove direction along the y axis. This time, however, the grating normal is rotated in the plane 共of the figure兲 y ⫽ 0 of an angle ␣n with respect to the z axis to allow the central wavelength at the nth order to be diffracted back along the z axis. Following this procedure, the diffraction pattern results again are located at the center of the image and are lined up along the x axis. Figure 4 shows three diffraction angles at which Littrow condition is fulfilled at the nth order for wavelengths and ⫾ ⌬. The bias angle ␣n can be calculated from Eq. 共1兲 while, differentiating Eqs. 共1兲 and 共3兲, after some trivial manipulations we obtain: ⌬ ⫽
冋
⌬x 4 ⫺ 2 f ob 共ng兲 2
册
1兾2
,
⌬y ⫽ 0.
(5)
Similar consideration holds to the ones leading to Eq. 共4兲; that is, neither ⌬x nor fob alone play any physical relevance, only their ratio ⌬x兾fob is meaningful and represents the angular distance between the image points corresponding to the diffracted wavelengths ⫹ ⌬ and . A misalignment of the grating will again change the position and the orientation of the diffraction pattern, but Eq. 共5兲 still holds to take distances along the spectrum. 3. Experimental Results
Figure 5 shows a typical image of the far-field diffraction pattern as it appears on the CLSM monitor when the measuring setup described in the wavelength measurement paragraph is followed. A simple Ronchi ruling has been chosen for all of the wavelength measurements presented here. The grating spatial frequency is g ⫽ 19.7 mm⫺1, and the reflection from its glass substrate is more than adequate to obtain a detectable diffraction pattern. The first step in calculating from Eq. 共4兲 is to retrieve angular coordinates from the linear ones used by the software to display the image data. The proportionality factor fob used by the instrument during the measurement can be obtained from the confocal software settings. From the declared objective magnification value Mob we can calculate fob with use of the 3058
APPLIED OPTICS 兾 Vol. 43, No. 15 兾 20 May 2004
Fig. 6. Profile plot along a line section passing through the centers of the diffraction spots. The left part of the diffraction pattern, formally corresponding to negative wavelengths, bears no additional information.
formula fob ⫽ ftl兾Mob, where ftl is the tube lens focal length; for a Nikon microscope its value is 200 mm.15 A useful trick is to define a fake objective in the software settings with fob ⫽ 1; that is, Mob ⫽ ftl 共equal to 200 for our microscope兲 to directly display angular coordinates. An even better solution is to set fob ⫽ 2兾g 共that is, Mob ⫽ gftl兾2 ⫽ 1970兲 to instantly read values onto the monitor confocal image. A more direct and accurate method, which will also remove all possible instrumental calibration errors, is to perform a test measurement with a reference wavelength. With use of a He–Ne laser at 612 nm, the angular distance 共⌬x兾fob兲 between the first and zeroth diffraction order spots can be calculated from Eq. 共4兲, and results in 6.03 mrad; this value can then be used to calibrate the linear coordinates on the image display. The measured total scan angle field of our CLSM is 56.7 mrad, which is a sufficiently low value to justify the application of paraxial approximation in the previous analysis. The diffraction pattern shown in Fig. 5 has been obtained with an emission wavelength of 740 nm from the Ti:sapphire oscillator. The tuning has been done according to the factory-provided calibration curve. The smallest pinhole, with a diameter of 20 m, has been chosen inside the confocal scanning head. This finite size causes the angular spread of the diffracted order spots in Fig. 5, with a doubled 共convolution兲 effect, in both the illumination and detection paths 共the pinhole plays the role of both the input and output slits in a spectrometer兲.16 Moreover, diffraction effects due to the finite pupil size of the interfacing optics further increase the spreading to reach a final angular dimension of ⬃0.3 mrad FWHM. In Fig. 6 the intensity values along a line section passing through the centers of the diffraction spots are plotted versus . The previously described angular spread corresponds to a wavelength broadening of ⬃30 nm FWHM. The peak abscissa can be located with a far better precision, and we can estimate
Fig. 7. The entire Ti:sapphire emission tuning curve has been explored, and the differences between measured and nominal wavelength values are plotted. All data lay around zero within the experimental uncertainty.
an a priori maximum error of ⫾5 nm, mainly owing to pattern asymmetries. As we can see in Fig. 6, the left part of the diffraction pattern, formally corresponding to negative wavelengths, bears no additional information. To exploit the whole field of view, the distance between the two symmetrically positioned first orders could be used to calculate . Moreover, only a sixth of the entire range of 56.7 mrad will be filled by the diffraction pattern at the maximum Ti:sapphire emission wavelength of 1 m with the present grating. A higher spatial frequency grating will then be a more appropriate choice, or a pair of higher diffraction orders could be used. This will reduce measurement uncertainty to a minimum of ⫾1 nm, which is a quite good resolution for this kind of application. The entire Ti:sapphire emission tuning curve has been explored, and the differences between measured and nominal wavelength values are plotted in Fig. 7.
All data lay around zero within the experimental uncertainty. Bandwidth measurements have been carried out following the setup described in Subsection 2.B. The Ti:sapphire oscillator has again been tuned around 740 nm. A blazed diffraction grating with a spatial frequency of 600 mm⫺1 has been used at the second order. Once again, the choice has been exclusively dictated by practical reasons of an immediate availability. The grating normal has been tilted around the optical z axis by a bias angle ␣n ⫽ 460 mrad, as computed from Eq. 共1兲. Figure 8 shows the diffraction pattern CLSM image when 共a兲 a cw emission is generated and 共b兲 when mode locking is active. In Fig. 9, the corresponding intensity profiles, 共a兲 and 共b兲 of Fig. 8, along a central line section, are plotted versus . Equation 共5兲 has been used this time to relate the spatial image coordinate x with wavelength values. It would be again convenient to define, from Eq. 共5兲, a fake objective in the confocal software settings to directly display angular 共Mob ⫽ 200兲 or wavelength coordinates 共Mob ⫽ 133924; this grating dependent magnification value could be an impractically large one for the software to be accepted兲 on the monitor image. The cw narrow laser linewidth yields a diffraction pattern 关Fig. 8共a兲 and 9共a兲兴, which is the instrumental spread function previously described, and with the present setup its angular dimension corresponds this time to a wavelength broadening of ⬃0.4 nm FWHM. When the laser linewidth broadens due to pulsed emission, the spectrum of Fig. 8共b兲 is generated. A bandwidth B of 4.7 nm can be estimated from the
Fig. 8. Typical diffraction patterns when bandwidth measurements are carried out. The Ti:sapphire oscillator has again been tuned around 740 nm. A blazed diffraction grating with a spatial frequency of 600 lp兾mm has been used at the second order. The grating normal has been tilted around the optical z axis by an angle of 460 mrad. Shown are a diffraction pattern when 共a兲 a continuous wave emission is generated and 共b兲, when mode locking is active. 20 May 2004 兾 Vol. 43, No. 15 兾 APPLIED OPTICS
3059
function is, in our case, 0.3 mrad FWHM, corresponding to spectral broadening of approximately 30 nm and 0.4 nm in our wavelength and linewidth measurement setups, respectively. These values affect the ultimate spectral resolution attainable, which 共though quite adequate for this kind of application兲 can be greatly improved, making use of smaller pinhole sizes 共when available兲 and higher spatial frequency gratings.
References Fig. 9. Intensity profiles versus ⌬ along a central line section, corresponding to Figs. 8共a兲 and 共b兲, respectively. The cw narrow laser linewidth yields a diffraction pattern 关curve 共a兲兴, which is the instrumental spread function. Its angular dimension corresponds, in the present setup, to a spectrum width of ⬃0.4 nm FWHM. When the laser linewidth broadens, owing to pulsed emission 关curve 共b兲兴, a bandwidth of 4.7 nm results.
curve 共b兲 of Fig. 9. The duration-bandwidth product formula17 is: cB兾 2 ⱖ 0.315,
(6)
where c is the velocity of light, is the FWHM pulse duration, and the numerical constant 0.315, whose value depends on the actual pulse shape, refers to an assumed sech2 pulse shape. From Eq. 共6兲 a pulse duration ⱖ 122 fs results. This bandwidth-limited value is consistent with the factory-specified typical pulse duration of 130 fs. 4. Conclusions
In this paper simple, inexpensive, and effective procedures are reported to be able to perform spectral characterization of a tunable Ti:sapphire ultrafast laser system for multiphoton microscopy applications without the need of any further instrumentation but the confocal microscope itself. Although cost effectiveness is not a main issue for multiphoton microscopists, since the high price of the instruments is highly superior to that of any spectrometer, the choice, setting, and handling of further instrumentation is time consuming and unnecessary. To measure the peak emission wavelength, a low spatial frequency Ronchi ruling, a device easily available in any laboratory, is placed onto the microscope specimen stage. Once any objective and barrier filter is removed, the diffraction pattern can be imaged onto the confocal monitor. With a suitable configuration of the display software settings, wavelength values can be directly read as distances between diffraction order pairs. For bandwidth measurements, a higher spatial frequency grating is required. In this paper a blazed grating for spectroscopic applications was used. However, the abundance of input light available from the laser source make the choice of a high-efficiency grating not mandatory. Again, any easily available high-frequency grating can be effectively employed. The angular dimension of the instrumental spread 3060
APPLIED OPTICS 兾 Vol. 43, No. 15 兾 20 May 2004
1. W. Denk, J. H. Strickler, and W. W. Webb, “Two-photon laser scanning fluorescence microscopy,” Science 248, 73–76 共1990兲. 2. W. Denk, D. W. Piston, and W. W. Webb, “Two-photon molecular excitation in laser-scanning microscopy,” in Handbook of Biological Confocal Microscopy, J. B. Pawley, ed. 共Plenum, New York, 1995兲, pp. 445– 458. 3. K. Konig, “Multiphoton microscopy in life sciences,” J. Microsc. 200, 83–104 共2000兲. 4. A. Diaspro, ed. Confocal and Two-Photon Microscopy: Foundations, Applications, and Advances 共Wiley-Liss, New York, 2002兲. 5. P. E. Ha¨ nninen and S. W. Hell, “Femtosecond pulse broadening in the focal region of a two-photon fluorescence microscope,” Bioimaging 2, 117–121 共1994兲. 6. G. J. Brakenhoff, M. Mu¨ ller, and J. Squier, “Femtosecond pulse width control in microscopy by two-photon absorption autocorrelation,” J. Microsc. 179, 253–260 共1995兲. 7. M. Mu¨ ller, J. Squier, and G. J. Brakenhoff, “Measurement of femtosecond pulses in the focal point of a high-numericalaperture lens by two-photon absorption,” Opt. Lett. 20, 1038 – 1040 共1995兲. 8. J. B. Guild, C. Xu, and W. W. Webb, “Measurement of group delay dispersion of high numerical aperture lenses using twophoton excited fluorescence,” Appl. Opt. 36, 397– 401 共1997兲. 9. M. Mu¨ ller, J. Squier, R. Wolleschensky, U. Simon, and G. J. Brakenhoff, “Dispersion precompensation of 15 femtosecond optical pulses for high numerical aperture objectives,” J. Microsc. 191, 141–150 共1998兲. 10. R. Wolleschensky, T. Feurer, R. Sauerbrey, and U. Simon, “Characterization and optimization of a laser-scanning microscope in the femtosecond regime,” Appl. Phys. B 67, 87–94 共1998兲. 11. A. C. Millard, D. N. Fittinghoff, J. A. Squier, M. Mu¨ ller, and A. L. Gaeta, “Using GaAsP photodiodes to characterize ultrashort pulses under high numerical aperture focusing in microscopy,” J. Microsc. 193, 179 –181 共1999兲. 12. F. Cannone, G. Chirico, G. Baldini, and A. Diaspro, “Measurement of the laser pulse width on the microscope objective plane by modulated autocorrelation method,” J. Microsc. 210, 149 – 157 共2003兲. 13. A. Diaspro, M. Corosu, P. Ramoino, and M. Robello, “Adapting a compact confocal microscope system to a two-photon excitation fluorescence imaging architecture,” Microsc. Res. Tech. 47, 196 –205 共1999兲. 14. M. Born and E. Wolf, Principles of Optics, 6th ed. 共Pergamon, Oxford, 1980兲, p. 412. 15. M. Abramowitz and M. W. Davidson, “Microscope ObjectivesNikon CFI60 200兾60兾25 Specification,” 共2003兲, http:兾兾micro. magnet.fsu.edu兾primer兾anatomy兾nikoninfinity.html. 16. J. M. Lerner and A. Thevenon, The Optics of Spectroscopy, paragraph 2.12.1 共Jobin Yvon, Edison, N.J., 1988兲, http:兾兾www.jobinyvon.com兾usadivisions兾oos兾index.htm. 17. J. C. Diels and W. Rudolph, Ultrashort Laser Pulse Phenomena: Fundamentals, Techniques, and Applications on a Femtosecond Time Scale 共Academic, San Diego, 1996兲, p. 7.
