1. Introduction

Laser ablation has become an essential technique for micro processing in industry, e.g. for the manufacturing of highly efficient thin film solar cells. Especially, for the separation of panels in adjacent sub cells, laser ablation is used to selectively structure an absorber layer as well as a positive (p)- and a negative (n)-contact for achieving a so called monolithic serial interconnection. The p-contact of CIS (Copper-Indium-Diselenide) thin film solar cells consists of a 0.5 µm thin molybdenum (Mo) film on a 3 mm glass substrate. The patterning of such molybdenum films with pulsed laser beams has been investigated by several groups [1–5]. Currently in production lines, the molybdenum is patterned with pulse durations in the nanoseconds range. However, nanosecond pulses are connected with thermal effects creating burrs and micro cracks in the remaining molybdenum layer and in the underlying glass substrate, whereas ultra-short pulses in the picosecond and femtosecond regime significantly reduce the thermal influence and thus enable a selective removal of matter [6–8]. Moreover, several investigations demonstrated, that laser lift-off processes efficiently remove and transfer thin absorbing layers from transparent substrates free from thermal effects [1,2,5,9,10].

In recent studies, we demonstrated that picosecond laser irradiation of thin molybdenum [5] and other metal films [11] from the glass substrate side results in a lift-off effect also called “direct induced laser ablation”. As a result, ablated spots and grooves are free from thermal effects and show clean and regular edges. Furthermore, complete disks of the ablated metal film were found after irradiating with fluences above 0.5 J/cm², while blister formation in the layer was observed below this threshold. The direct induced laser ablation is assumed to be a general effect when irradiating absorbing metal layer of a few 100 nm thickness from the transparent substrate side [11].

In fact, the key to efficient laser process optimization is not only varying parameters to achieve best results but also understanding the process as a whole. Hence, ablation results of different thin film systems could be predicted. A rough energetic consideration for glass side ablation of Mo demonstrated that the energy per ablated volume is about 30 J/mm³, which is below the total evaporation enthalpy of 78 J/mm³ [5]. Thus, a simple thermo dynamical model was developed, assuming ultra-short laser pulse irradiation initiates only partial evaporation of the metal film. Then, gas expansion drives the bulging of the metal film, whereupon a metal disk shears and lifts-off, when the tensile stress limit of the metal blister is exceeded [11].

The best way to understand the ablation mechanisms is provided by a transient observation. However, fast cameras have frame rates in the order of 1 MHz, resulting in microsecond temporal resolution, which is clearly to slow to resolve reactions in the sub nanosecond area. A synchronization of multiple cameras in series enables nanosecond resolution, but the number of pictures depends on the number of cameras. However, transient ultrafast pump-probe microscopy enables the creation of stop-motion movies of ultrafast ablation process. This method is based on irradiating the sample with a pump pulse for initiating laser ablation, and stroboscopic illumination of the affected area with a temporarily delayed probe pulse for microscopic imaging with a camera. The stop-motion movie is created by capturing successive laser ablations on different sample positions with increasing delay times.

Pump-probe microscopy of ultra-short pulse laser ablation started in 1985, when Downer et al. [12] published photographs of bulk silicon ablation with a temporal resolution of 100 fs, while the probe pulse was optically delayed on a linear stage up to 600 ps. Further investigations of ultra-short pulse front side irradiation followed, showing the ablation of bulk material [13,14], thin metal films [15,16], and the lift-off of thin SiO₂ films on Si [17] with femtosecond resolution but only up to a maximum delay time of a 10-20 nanoseconds, due to the limited travel range of optical delay lines on linear stages. Except Mingareev et al. [15] used a Herriott cell to achieve a maximum delay time of 1.9 µs, showing melt ejections and shockwave propagation in air. In contrast, for observing slower processes such as the lift-off of silicone on print plates [18] and laser induced forward transfer (LIFT) processes, which are induced by ns-lasers [19,20] or by fs-laser [21,22], the probe-pulse is emitted by an electronically delayed second nanosecond probe laser. Thus, the temporal resolution depends on the jitter and pulse duration of the probe laser but the maximum delay time is unlimited.

In the case of direct induced laser ablation of thin metal films, the relatively slow and delayed mechanical deformation and lift-off proceeds on a 10-100 nanosecond time scale, while ultra-short laser pulses initiate ultrafast electron excitation, electron-phonon heating, phase transitions, and heat transfer, all proceeding in the femto and picosecond range [23]. Thus, a combination of both probe methods is necessary to investigate the direct induced ablation as well as other ablation processes as a whole.

We already presented pump-probe microscopy investigations of a molybdenum lift-off initiated by a 10 ps laser up to delay times of 4 ns [24], showing the start of mechanical movement but not the actual shearing and ablation of a Mo disk. Thus, we enhanced the setup by an ultra-fast pump pulse (660 fs FWHM) combined with an optical delay for the first 4 ns and an electronic delay (600 ps FWHM) for longer delay times. Furthermore, we also demonstrate the calibration of the delay time zero point, the synchronization of both probe sources, and a method for image quality enhancing. As a result, a pump-probe setup is presented in this paper that enables the complete observation of laser induced processes, especially of thin film ablation processes, with a high temporal dynamic range. Moreover, we present for the first time to our knowledge a complete pump-probe microscopy investigation of the directly induced laser ablation of Mo from femtoseconds to microseconds.

2. Results and discussion

2.1 Pump-probe microscopy setup

The enhanced pump-probe microscopy setup is sketched in Fig. 1 . An ultrafast laser source (Nd:Glass chirped pulse regenerative amplifier) emits pulses with a duration of about 660 fs (FWHM) at a centre wavelength of 1053 nm. The emitted laser pulse is split up by a half wave plate and beam splitter combination (beam splitter 1, Fig. 2 ) into pump and probe pulse. Here, the energy of the probe pulse is set for a sufficient illumination of sample.

 

Fig. 1 Pump-probe microscopy setup. A femtosecond laser pulse is split by beam splitter 1 into pump- and probe pulse. The pump pulse (red branch) is focused on the sample. The probe pulse (green branch) is frequency doubled by a LBO crystal, and passes an optical delay line for time delays up to 4 ns. A picosecond laser emits a second probe pulse, which is triggered by a digital delay generator for time delays above 4 ns. Both probe beams are combined in beam splitter 3, and illuminate the sample, imaged by a microscope with CCD camera.

Download Full Size | PDF

 

Fig. 2 A streak related to the pump pulse propagation in the glass can be traced from the bottom of the image to the ablation centre at fluences above 1 J/cm². The delay time zero point is defined when the streak reaches the ablation centre.

Download Full Size | PDF

The pump pulse passes a mechanical shutter, which is synchronized with a digital delay generator to pick a single pulse, while the laser permanently operates at a repetition rate of 500 Hz. The next half wave plate and beam splitter combination (beam splitter 2, Fig. 1) is used to adjust the pump pulse energy for initiating the ablation. Then, the beam is expanded and focused by a lens (f = 150 mm) on the sample for ablation with a spot diameter of about 40 µm at 1/e² intensity, measured by an in focus beam profiler. The angle of incidence is about 35 degrees with respect to the microscope objective to avoid clipping. The microscope itself is oriented perpendicular to the sample for optimal imaging in the focal plane.

The probe pulse is frequency doubled by a lithium triborate (LBO) crystal to a wavelength of 527 nm with a pulse duration of about 510 fs (FWHM), enabling spectral separation of pump and probe beam by a band pass in the microscope and higher sensitivity in a CCD camera. Subsequently, the time delay between pump and probe pulse is varied by an optical delay line on a 300 mm translation stage. Hence, an optical delay of up to 4 ns is achieved by folding the beam path twice using 3 retro reflectors, which simplify adjustment.

For probing delay times above 4 ns, a second probe pulse is emitted by an actively q-switched 600 ps (FWHM) Nd:YVO₄ laser source at a wavelength of 532 nm. The time delay between the ultra-fast pulse from the regenerative amplifier and the q-switched laser is set electronically by the digital delay generator and can be used for probing up to 2000 s, while the jitter, mainly generated from the q-switched laser, is approximately 200 ps. Thus, the whole time frame of the ablation process can be covered with both probe sources maintaining the ultra-fast pump pulse.

Both probe pulses are expanded to match their diameters, and both are combined in a beam splitter (beam splitter 3, Fig. 1), resulting in a s-polarized 510 fs probe pulse and a p-polarized 600 ps probe pulse. Consequently, a half wave plate mounted in a motorized rotation stage adjusts both probe pulses to p-polarisation, maximizing transmission through the next beam splitter (beam splitter 4, Fig. 1). A lens (f = 125 mm) in front of beam splitter 4 focuses the probe beam to the parfocal plane of a microscope objective (20x magnification, NA = 0.29, and 30.8 mm working distance), resulting in a collimated illumination of the sample. In between, a quarter-wave plate turns the polarisation from linear to circular, whereupon backscattered and reflected light from the sample is turned by 90 degrees in total and now is reflected in beam splitter 4 to the CCD camera. Next, the 527 nm (fs-laser) and 532 nm (ps-laser) probe pulses pass a 531 nm band pass filter (40 nm FWHM), that rejects reflections and scatters of the pump beam as well as ambient and plasma light. Finally a tube lens (TL) focuses the light to a CCD camera for imaging. The 8 bit CCD camera with 1280x960 pixels provides a field of view of 285 µm x 223 µm and a pixel resolution of 0.23 µm, while the calculated optical resolution is R = 0.61 λ/NA = 1.22 µm The exact timing of the camera exposure is also controlled by the digital delay generator.

Since each time delayed image affords a new ablation spot, a second motorized stage is used to move the sample. Thus, a LabView program automatically takes a series of pictures at different time delays on different locations on the sample. These pictures can be combined to a stop-motion movie

2.2 Image processing

To optimize the image quality, two essential editing steps are performed. At first, the images are normalized to compensate brightness fluctuations generated by the movement of the delay line or by pulse fluctuations. The standard deviation of the pulse fluctuations is about 1% for the fs probe-pulse, and about 6% for the ps probe-pulses. Since the affected area is relatively small compared to the captured image, the corresponding grey value of the histogram peak is proportional to the applied probe pulse energy and enables the normalization. The second step is a difference picture creation by subtracting a reference picture, captured before the onset of reaction (∆t < 0), from the picture at the adjusted delay time (∆t = t_n). Thus, the difference pictures contain information about the relative change of reflectivity and enable subtraction of sample impurities. Measurements of the background noise of a single shot revealed a pixel reflectivity resolution ∆R/R of about 2%, limited by the colour depth of the 8 bit CCD camera.

2.3 Temporal resolution and visualization of the delay time zero point

The temporal resolution or minimum instrumental response time of a pump-probe setup can be estimated with the cross correlation width of pump and probe pulse, in our case approximately 840 fs (FWHM), assuming a Gaussian pulse form. The delay time zero point was calibrated by shifting the probe pulse in 300 fs steps from negative to positive delay times with respect to the pump pulse. The pictures in Fig. 2 show a black streak propagating with the speed of light in glass from the bottom of the image to the centre of the ablation area in 600 fs time steps. The black streak is related to the pump pulse propagation in the 3 mm thick glass, since the molybdenum layer is observed perpendicular through the glass substrate and the pump pulse enters the glass substrate under an angle of 35 degrees with respect to the normal of the sample surface. At +600 fs a dark spot of about 5 µm in diameter is observed (Fig. 2, right), resulting in glass damage in the centre of the ablation area in the following time steps. In contrast, the actual ablation area in the molybdenum layer has a significantly larger diameter of about 25 µm (see Fig. 4 for comparison). In fact, the propagating streak in Fig. 2 is only observed at fluences above 1 J/cm², why its appearance is related to nonlinear glass absorption. Since the nonlinear reflectivity change originates from the response of the electronic system it is considered as quasi instantaneous. Thus, the propagation of the pump pulse is resolved in 600 fs steps in Fig. 2, demonstrating the temporal resolution of the setup is in the range of the calculated instrumental response time of 840 fs. In our work the delay time zero point is defined, when the streak reaches the point of ablation just before the appearance of the dark spot. At this point in time, we assume the pump and the probe pulse reach the substrate/layer interface. Hence, the different optical path lengths in the glass substrate do not influence the calibration.

2.4 Temporal overlap of optical and electronic delay

The temporal overlap from optical to electronic delay is either calibrated by a fast photodiode or directly by the sample answer. In both cases, the intensity peaks of both pulses must be synchronized temporally. In this work, the trigger signal of the ps probe-pulse was shifted until the observations from both probe sources match in the temporal transition area at the delay times 1 ns, 2 ns, and 3 ns (Compare Fig. 3 , upper and lower row). In the case of glass side ablation, a transition from homogeneous reflectivity to a formation of one Newton ring in the ablation region is observed using the femtosecond probe source. The corresponding images taken with the picosecond probe laser reveal the same dynamic behavior, although the exposure time was increased by a factor of 1000 (Fig. 3, lower row). As a result, we expect the temporal overlap accuracy is below 1 ns, taking into account that the jitter of the ps-probe laser is about 200 ps.

 

Fig. 3 Comparison of pump-probe images using the optical delay line with a pulse duration of about 510 fs (upper row) and the electronic delay with a pulse duration of 600 ps (lower row), each at delay time of 1 ns, 2 ns, and 3 ns (columns). The delay time of the ps-probe pulse is calibrated in respect to the optical delay time by means of matching images.

Download Full Size | PDF

2.5 Transient microscopy

Figure 4 shows pump-probe microscopy images of Mo films irradiated and observed from the glass substrate side at fluences of 0.5 and 0.7 J/cm² (beam diameter 40 µm at 1/e² intensity), resulting in a blister formation with a horizontal diameter of about 22 µm and a sheared-off Mo disk with a diameter of about 25 µm. Images at delay times below 4 ns were illuminated with the optically delayed 510 fs probe pulse, while images at delay times above 4 ns were illuminated with the electronically triggered 600 ps probe pulse.

 

Fig. 4 Pump-probe microscopy images of Mo film on glass observed and irradiated from the glass substrate side. The two upper and two lower rows show the temporal evolution of a bulging and Mo lift-off process in reading direction for fluences of 0.5 J/cm² and 0.7 J/cm², respectively. The black dotted horizontal lines indicate the boarders of the blister and the ablated hole and also separate Newton’s rings from diffraction rings.

Download Full Size | PDF

At 1 ps delay, a weak increase of reflectivity in the centre is observed at both fluence levels. A similar increase of surface reflectivity for that time domain was observed by von der Linde et al. in Si bulk material and was related to the generation of photoexcited electron-hole plasma, changing over into a liquid metallic phase after 2 ps by very strong excitation [13]. In metals, the pulse energy is absorbed by free electrons followed by an energy transfer to the lattice, leading to ultrafast melting after a few picoseconds [23]. In addition we observed the reflectivity in the melting zone of film side irradiation. Here, the reflectivity also increased at about 1 ps and remained at the same level until the final state. Thus, the observed increase of reflectivity at 1 ps is related to ultrafast melting.

Up to 10 ps delay time, a low reflective circular area forms in the middle of the high reflective region. Here, the horizontal diameter of the dark region is about 17 µm, while the diameter of the bright rim is about 35 µm. Bonse et al. and von der Linde et al. made similar observations in bulk laser ablation and explained this as the onset of ablation [14].

At 100 ps, ablation of bulk material from the front side leads to further reflectivity decrease, as a consequence of scattering and absorption in an expanding two-phase regime containing gas and liquid [13,16]. In our case of a substrate side ablation, the reflectivity in the centre region passes a minimum at 10 ps and subsequently increases with time, while the bright rim shrinks and both regions merge into a single bright circular area with a diameter of 18 µm at about 400 ps. Thus, we assume the expansion of a gas liquid mixture is suppressed due to its confinement and the increase of reflectivity is related to cooling and condensation.

Between 400 ps and 4 ns, a dark rim appears around the bright circular area, while the contrast increases with time. The outer diameters of the dark rims are about 22 µm (0.5 J/cm²) and 25 µm (0.7 J/cm²). Both match the blister and the hole diameter, which are created at later delay times, as indicated by the black dotted lines in Fig. 4. Regarding the pump-probe pictures at 10 and 100 ps, a further shrinking of the outer bright rim would be expected due to further cooling, but the reflectivity increases with time and a significant border in form of the dark rim appears. Thus, we assume both effects originate from the onset of the film delamination, probably starting between 100 and 400 ps.

In our case, the thin film is irradiated and observed from the glass side, thus reflected light from the bulging concave Mo cap interferes with Fresnel reflections from the glass substrate, resulting in Newton ring formation. Newton rings were detected for bulk material laser ablation, but at much earlier delay times of about 100 ps, due to interference of reflected light from the remaining bulk material and the ablation front [13,14]. Moreover, Newton rings were also monitored during indirect induced ablation of 300 nm SiO₂ on Si at about 1 ns [17]. Consequently, the dark rim observed between 400 ps and 4 ns is related to destructive interference starting at the borders, indicated by the dotted line in Fig. 4, and the bright centre region to constructive interference. The number of newton rings increases with the delay time due to further bulging of the Mo cap. These rings enable time resolved measurements of the blister bulging height and velocity for our future work.

Furthermore, at delay times above 1 ns a second ring pattern starts appearing around the blister region outside the dotted lines in Fig. 4, that is still observable in the final bulging state and even after the removal of a Mo disk. The ring spacing is about 2 µm, which is consistent with calculations of the diffraction pattern for imaging a 25 µm pinhole with a NA of 0.29. Thus, the appearance of the outer ring pattern is a further indicator for the onset of mechanical deformation. Moreover, the contrast of the pattern increases with the bulging height and thus also with the curvature at the edges to a maximum at about 20 ns.

For the irradiation with 0.5 J/cm² the maximum number of rings is counted between 18 and 25 ns. Then, the ring number decreases again, possibly due to cooling of the blister, and the final stage of the bulging process is reached at about 10 µs.

In contrast, an irradiation with 0.7 J/cm² reveals a further smaller dark region with a diameter of about 7 µm within the bright region at about 400 ps. In the following, the dark region shows dynamic behavior and fringes at the border, that is probably a gas bubble. Between 15 and 25 ns, cracking at the edges of the hole is observed indicating the shearing of the Mo cap. At about 53 ns, the diffraction rings lose their concentric form, which is also observed for partly connected Mo films in the final state. After 250 ns, reflections of the Mo disk fail to appear because the Mo disk has moved out of sight.

In summary, at early delay times up to approximately 400 ps we observe reflectivity changes, at medium delay times from 400 ps to 15 ns mechanical deformations, which for higher fluences result in a Mo cap lift-off.

3. Conclusion

In this paper for the first time to our knowledge, a pump-probe microscopy setup is introduced that combines an optical and an electronic delay line for the complete observation of a directly induced laser ablation up to delay times of 10 µs. The pump-probe microscope provides a high temporal dynamic range of 10⁷ s, when comparing the pump pulse duration with the maximum observation time used here. In principle a maximum delay time of 2000 s can be set. Up to 4 ns, the optical delay of a 510 fs probe pulse at 523 nm is accomplished with a translation stage. Above 4 ns, an actively q-switched laser pulse with a duration of 600 ps at 532 nm is electronically delayed for probing. The delay time zero point of a substrate side ablation of molybdenum films is calibrated at a fluence above 1 J/cm² with an accuracy of about 600 fs, since a nonlinear effect in the glass substrate enables tracing of the pulse propagation through the glass. Furthermore, the electronic delay can be calibrated relative to the optical with an accuracy below 1 ns by matching images taken with each probe source at delay times of 1, 2 and 3 ns. It was noticed that both pump-probe images series contain the same temporal information and it is justified to use the about 1000 times longer q-switched probe pulse for delay times longer than 4 ns, without losing physical information. Furthermore, a method based on difference image generation is presented that increases the image quality and corrects probe pulse fluctuation.

First experiments on directly induced laser ablation of molybdenum films from the glass substrate side were performed. An irradiation with a fluence of 0.7 J/cm² revealed that mechanical deformation of the film starts at about 400 ps and continues until approximately 15 ns. Then, a Mo disk is sheared and lifted-off free from thermal effects. Furthermore, a molybdenum blister was formed at a lower fluence of 0.5 J/cm². Here, the molybdenum film bulges to its maximum height at about 100 ns and thereafter shrinks due to cooling until it reaches the final state at 10 µs.

The newly developed pump-probe microscopy setup enables further time resolved investigations of laser ablation processes in general from the femtosecond domain to the equilibrium state. Our future work will focus on induced laser ablation processes of thin film systems and on quantitative evaluation.