1. Introduction

Ultrashort laser pulses are used to modify many kinds of transparent materials where the advantages of high-intensity nonlinear absorption are most advantageous to drive controllable processing in otherwise highly robust materials. Multi-photon interactions confined within a small focal volume at high intensities offer three-dimensional (3-D) structuring on micro- and nano-scale dimensions that enable the fabrication of novel components for diverse applications such as integrated optics, telecommunications, sensors, lab-on-a-chip, MEMs and many more devices in transparent media like glasses, crystals and polymers [1–7].

High-aspect ratio holes in diverse materials were created using various methods of laser processing: One method is the statically exposure with single femtosecond laser pulses focused at the front side [8] or the rear side of a glass sample which in the latter case was in contact with a liquid [9]. Additionally, scanning methods have been investigated like longitudinally scanning along the beam propagation direction from the rear surface of silica glass [10]. Another method for surface treatment is laser-induced back-side wet etching of fused silica which was realized using excimer laser radiation obtaining high quality etched surfaces with micropatterns [11]. However, disadvantages of this method are the long specified processing time of several minutes and the immersion technique with an organic solution. By applying the trepanning drilling technique through holes in 140-µm thick borosilicate glass have been achieved using Nd:YAG laser radiation [12]. Beside glass materials even metalloids like silicon wafers have been processed with a helical drilling method where smooth holes were fabricated with pulses of 10ps pulse duration [13]. In this case the processing time exceeds t = 100s for optimal drilling results.

With the described techniques holes with aspect ratios greater than 10 were obtained in diverse materials. Occurring problems are the formation of microcracks at the walls of the hole and on the surface which are created by single pulses due to the induced quickly rising temperatures in the focal volume forming stress especially in pure glass. An increase of the used laser power to reduce the processing time is limited by cracks forming in glass materials.

One solution creating improved surface morphology is the use of ultrashort pulses in the range of femtoseconds. The multiphoton absorption process becomes more efficient at pulse durations below 100fs [14]. Furthermore, in respect of the hole depth femtosecond pulses are more energetically efficient than nanosecond pulses because smaller amounts of energy per pulse are needed [8]. Due to the short time of laser interaction high aspect collateral damage and plume-shielding effects are reduced creating holes with great spatial precision [15].

Progress regarding crack-free holes has been achieved by using burst trains with pulses at high repetition rates instead of single pulses. With a pulse-train of 400 identical 1ps-pulses, separated by t = 7.5ns, shock-induced cracking effects have been eliminated [15]. Brittle materials like fused silica have been precisely ablated by high repetition rate (f = 133MHz) multi-pulses achieving high depth control. These holes were fabricated without debris or collateral damage. Burst trains enable a more gentle modification in the laser interaction volume due to the small time interval between subsequent pulses causing heat accumulation effects at high temperatures and therefore improving the ductility of the glass. These transient effects occur when the time between subsequent pulses is shorter than the thermal diffusion time of the material. Residual heat remains after the first pulse hit the material and following pulses further increase the heat. Thus a gentle heating over a period of several pulses take place. The glass becomes ductile due to heat accumulation effects and thus a crack-free modification of the material is possible [16]. The onset of heat accumulation effects depends on the material and laser conditions. In AF45 glass for example cumulative heating was noted at repetition rates f>200kHz with 450nJ-pulses at λ = 1045nm when the modified volume significantly exceeds the laser spot size [17].

In metals burst trains are also advantageous for hole drilling. A single burst train of 400 picosecond pulses with a pulse-to-pulse separation of t = 7.5ns created clean through-holes in 200-µm thick aluminum foil [18]. High repetition rates support the hydrodynamic expansion of the ablation plume away from the hole. At a repetition rate of f = 1kHz a shock wave front and a vapor front carrying ablated material have been observed during the processing of steel with ultrashort laser pulses [19]. The plume expanded linearly with 3km/s in the first 30ns after the pulse hit the surface.

Beside heat accumulation other transient effects like filamentation and plasma effects influence the laser-material interaction and therefore the morphology of the ablated and modified material. Powerful femtosecond laser pulses propagate in a nonlinear way through optical media. Due to their high power nonlinear effects like the optical Kerr effect and thus self-focusing and the formation of filaments can occur [3,20]. The Kerr effect and therefore self-focusing are generated because of a refractive index in the material which depends on the intensity of the incoming radiation: $\Delta n=n_{2}I$. The third order nonlinear susceptibility χ⁽³⁾ gives rise to the nonlinear refractive index and is positive for the most nonresonant optical media. In a Gaussian beam the center of the pulse has the highest intensity and therefore creates the highest refractive index in the material compared to the wings of the spatial intensity distribution. This area acts like a focusing lens and reduces the diameter of the laser beam. The critical power for self-focusing depends on the laser wavelength (λ), the linear (n) and the non-linear refractive index (n₂) of the medium [21]:

The effect of filamentation can occur in all media which are transparent for the used laser radiation. Filaments created by powerful ultrashort laser pulses in air can reach a length in the order of meters or even kilometers [23,24]. Filaments in condensed matter are shorter because of the way how free electrons are generated. Firstly, the central slice of the pulse is self-focused due to the highest intensity. The electrons in the matter are excited from the valence band to the conduction band and receive energy through inverse Bremsstrahlung [21]. Further electrons are generated due to partial cascade ionization. The electron density reaches N_e≈10¹⁸cm^-3 which is three orders of magnitude lower than the atomic density of condensed matter. A plasma is generated which defocuses the central part of the pulse back to the background reservoir [25]. The energy of the slice is reduced due to the loss in ionization. Now the next slices of the pulse self-focus but at later positions in the propagation direction due to the lower peak power [21]. The process of Kerr lens self-focusing and defocusing in the plasma repeats itself. Thus an extended series of hot spots appear along the propagation axis, the so-called filament.

Future work should identify the role of filaments during the formation of holes. Laser-generated filaments may guide the direction of the forming hole in the material. Consideration is given in this paper regarding a hole generated by a burst train of four pulses. Further investigations should answer this question.

In this paper focused femtosecond laser burst trains create precisely shaped high-aspect ratio holes in BK7 glass. The morphology of the holes together with time-resolved imaging during hole drilling prove the possibility of a gentle and more effective hole drilling without cracks due to heat accumulation effects by using burst trains instead of single laser pulses. Studying transient and time-dynamic phenomena such as plume-shielding effects with burst trains in glasses temporally resolved optical images were captured during the hole drilling process. Dynamical mechanisms like photoluminescence, laser-generated filaments and plume formation in the range from 2ns to 400ms were studied to understand and profit from high repetition rate processing. The occurring time-dependent mechanisms could be investigated based on the captured temporally resolved images.

An accurate imaging of interactions between the burst pulses was achieved using a microscope objective and a time-gated intensified CCD camera. This setup was implemented to observe the processes from the side of the sample, in the plane parallel to the beam propagation direction. The presented method enable a better description and further physical insight into the laser machining process with femtosecond burst pulses in glass materials. Benefiting from heat accumulation effects in a burst train energetically efficient deep and crack-free holes were fabricated.

2. Experimental setup

Uncompressed Ti:Sapphire laser pulses (Spectra Physics, Spitfire) with a wavelength of λ = 800nm and a repetition rate of f = 500Hz were used to form burst trains. Each pulse was divided by a resonator forming one burst with a certain number of pulses. The resonator provided bursts with pulses at a repetition rate of f = 38.5MHz due to the length of the cavity. A Pockels cell inside the cavity partially rotated the polarization of the incoming pulse. A tunable delay generator was used for time shifting the high voltage bias pulse driving the crystal of the Pockels cell. Depending on the applied voltage the polarization of the incoming pulse is partially changed. The vertically polarized portion of the pulse left the cavity whereas the horizontally polarized portion circulated inside the resonator. The latter portion could be further energetically divided by the next pass through the cavity and the Pockels cell. The group velocity distortion introduced by the Pockels cell was compensated by double passing a pair of prisms. By multiple passes through the resonator the pulse energy of one inserted pulse was divided forming a burst. With the burst resonator the energy of each pulse within a burst could be varied independently so that any desirable pulse profile was available. The bursts were then sent to the Spitfire system for compression to a pulse duration of τ = 100fs. A detailed description of the burst train generator will be published elsewhere.

The output burst pulses were focused with a 10x microscope objective with a numerical aperture of NA = 0.16 (Fig. 1 ). The radiation was focused onto the surface of BK7 glass which was statically exposed. The focus spot was located x≈100µm from the edge of the sample. Ablation was performed at atmospheric pressure.

 

Fig. 1 (Color online) Scheme of the experimental setup. The intensified CCD-camera was triggered by a digital delay generator which also triggered the mechanical shutter for getting time-resolved images of the laser interaction zone. The shortpass filter λ = 750nm avoided the imaging of the laser wavelength λ = 800nm. At the sample the hole and the ablation plume are schematically drawn. The inset on the right shows a typical burst train with four pulses separated by t = 26ns at f = 500Hz repetition rate.

Download Full Size | PDF

A 20x microscope objective with the numerical aperture NA = 0.4 was used to image the laser glass interaction area onto the chip of the intensified CCD-camera Andor iStar DH734 (Fig. 1). Time-resolved images were captured during the ablation process. A short pass filter with λ = 750nm avoided the detection of the laser light. The gate width for the imaging was chosen to be the minimum available of t = 2ns. Only one image was taken at exactly the same position of the sample. After taking one image the sample was moved linearly perpendicular to the beam propagation direction in the plane of the imaging with a two-axis air-bearing motion stage (Aerotech ABL1000). At the same time the delay time of the ICCD was changed gradually by Δt = 2ns to get the next image. The camera was activated by a trigger signal from a digital delay generator (DDG) which itself was triggered by a signal from the timing and delay generator of the Spitfire that was synchronized with the 500Hz laser pulses (Fig. 1). The DDG also triggered the mechanical shutter which mechanically opened for a defined time to control the number of bursts. The opening time of the shutter remained the same for one series of images with different time delays.

The experiments were performed in two different structuring modes: the burst and the pulse mode. In the burst mode pulses of the same pulse energy with a repetition rate of f = 38.5MHz within a burst train irradiated the sample. The burst trains had a repetition rate of f = 500Hz. In this work we present the results regarding a burst train with four or five pulses of the same pulse energy. In the pulse mode only the first pulse in a burst train with a repetition rate of f = 500Hz was used.

3. Results and discussion

Micrographs and time-resolved images were examined on a nanosecond time scale from t = 2ns to t = 400ms. In detail the pulse mode and the burst mode with four equal pulses are discussed in this paper up to an accumulation of 200 burst trains. Intensity information during the hole drilling process delivered by time-resolved optical images show details of transient effects like laser-generated filaments, plume formation and photoluminescence which are discussed in the following sections.

3.1 Filamentation

A self-focusing filament was created by using two bursts with five equal pulses in a burst (Fig. 2 ). The used pulse energy was E_p = 32µJ and the laser radiation with a wavelength of λ = 800nm was focused onto the surface with a 10x objective (NA = 0.16).

 

Fig. 2 A self-focusing filament in BK7 glass created with the presented burst resonator (two bursts with five pulses, respectively). The laser radiation was focused onto the surface and propagated from left to right.

Download Full Size | PDF

After two burst trains hit the sample a small pit with a depth of approximately d = 10µm has been created. Below the pit following the beam propagation direction a temporally stable filament with a length of l≈140µm was formed. The circular shaped deformation at the half of the length is assumed to be a defocusing area affected by plasma which defocused the laser radiation. The shape of the filament itself and also the circular area are reproducible. Experiments with two to six pulses in a burst, with a total number of bursts up to 100 and E = 124µJ per burst, have shown that the total depth of the filament measured from the sample surface remained almost constant (l≈150µm). After the 100th burst the forming hole exceeded the depth of the filament and therefore it was not observable any more. In all cases regarding a varying number of pulses in the burst train a defocusing circular area was seen.

Filaments have been observed not only in the burst mode but also in the pulse mode. A filament at the bottom of the hole is observable after the 50th pulse (Fig. 3a ). It seems that the filament guides the direction of the hole in the beam propagation direction.

 

Fig. 3 Optical micrographs of holes in BK7 drilled with four pulses (E_p = 45µJ) in a burst in the pulse mode (left hole) and in the burst mode (right hole) for the 50th and the 200th burst, respectively (a) and with five pulses (E_p = 25µJ) in a burst for the 500th burst in the burst mode (b).

Download Full Size | PDF

3.2 Pulse and burst mode drilling

Holes in BK7 were drilled with a pulse energy of E_p = 42µJ for both the pulse and the burst mode (Fig. 3a). The hole created after the 50th burst in the pulse mode has a depth of d = 36µm in contrast to the hole created in the burst mode with a depth of d = 72µm. Thus, using the burst instead of the pulse mode at an exposure time of t = 100ms increases the depth by a factor of 2. After the 200th burst the hole depths differ by a factor of 1.6 when the hole created in the pulse mode has a depth of d = 115µm and the hole created in the burst mode is d = 189µm deep. It is clearly seen that in both cases the burst mode achieved a much larger hole depth than the pulse mode.

Additionally, the holes drilled with the burst mode are smooth and straight whereas the holes created in the pulse mode are non symmetrical and show microcracks at the sides of the holes. The observed difference in the morphology shows a strong dependency regarding the used structuring mode. The smooth shape of the holes in the burst mode is explained by the formation of heat accumulation due to the high repetition rate within a burst train. In the pulse mode absorbed energy could not be transferred into the surrounding material fast enough so that stress had been build up and finally released by cracks visible in the micrograph (Fig. 3a).

With five pulses in a burst train recast is visible at the walls of the hole after the 500th burst (Fig. 3b). The hole has a depth of d = 403µm. Due to the large depth the ablated material from the bottom of the hole could not reach the surface anymore and therefore is confined in the hole. The material was deposited at the walls and resolidified there. In this way ablated material shapes the hole.

3.3 Time-resolved ICCD images

Time-resolved optical images were taken with the ICCD during laser machining with a burst consisting of four pulses with equal pulse energy. Images from t = 0ns up to t = 178ns show the occurring transient processes after the 1st burst and after the 200th burst, respectively (Fig. 4 ). For the experiment with the 1st burst the pulse energy of each pulse in the burst was E_p = 48μJ; for the 200th burst the pulse energy was E_p = 43μJ. The dashed white line shows the surface of the glass sample (Fig. 4). On the left side back lighting microscopy images show the generated structures whereas the radiation was focused onto the surface of the glass sample. The magnification for the micrograph of the 1st burst is 2x compared to the other images for making the filament clearly visible.

 

Fig. 4 Time-resolved intensity images during the ablation process for the 1st and the 200th burst in the burst mode. The white line indicates the surface of the glass sample. The images on the far right side show the detected intensity values multiplied by a factor of 5. The beam propagation direction is top down. (Online available: Media 1 contains the pulse and the burst mode for the 1st burst, whereas Media 2 shows the pulse and the burst mode for the 50th burst, respectively.)

Download Full Size | PDF

For the 1st burst experiment (Fig. 4, first row) the first pulse of the absolutely 1st burst hit the surface of the sample at t = 0ns. Photoluminescence of the ablated material was observable in an ablation plume which departed from the surface as can be seen on the next few images. The expansion was impaired by atmospheric pressure. At t = 26ns the second pulse of the 1st burst hit the sample. By then a thin filament with a width of δ≈6µm in the volume of the sample was clearly visible. It survived about a long time of tens of nanoseconds whereas its length remained nearly the same at l = 75µm. The third and the fourth pulse of the burst hit the sample at t = 52ns and t = 78ns, respectively. At t = 78…178ns the intensity of the ablation plume got weaker and weaker. In order to make the ablation plume still observable the intensity values for the image at t = 178ns have been multiplied by a factor of 5.

The pit created after the 1st burst had a diameter of δ = 26µm and a depth of d = 10µm. At this time material could easily be removed from the surface. Each pulse in the burst effectively removed new material. A forming filament was not observed until t = 10ns after the first pulse hit the sample (Fig. 5 ).

 

Fig. 5 (Color online) Intensity decay of the filament brightness during the 1st burst in the burst mode (black square) and the pulse mode (red triangle). A double exponential decay was fitted to the data, respectively. The red arrows indicate the four hitting laser pulses in the burst mode.

Download Full Size | PDF

For the 200th burst experiment the first pulse hit the bottom of the already existing hole. No ablation plume was observable above the sample surface during the next tens of nanoseconds. The detected intensity values in the hole show that ablated material was confined inside the hole which strongly lighted up over the whole observation time. When the following pulses had struck the material, they interacted with the confined material causing plume-shielding effects. Thus, the depth of the hole after the 200th burst is near the saturation depth. At this time only a small amount of ablated material was seen escaping from the hole and enlarging the depth.

The hole created after the 200th burst shows a smooth and narrow shape with a hole diameter of δ = 32µm and a depth of d = 187µm. The aspect ratio is approximately 6. At the 200th burst a filament was not observable anymore. On the one hand the hole depth of d = 187µm would have exceeded the length l = 75µm of the filament if it had been created. On the other hand the optical propagation of the beam was distorted by the already existing hole shape so that the filament could not form. However, the formation of the hole might be guided by the filament as discussed above.

Furthermore, microcracks are not observable in the cross section images of the holes. Ablated material from the bottom and the walls of the hole which could not depart was deposited inside the hole again. Thus deposited material smoothened the walls of the hole.

3.4 Filament brightness

To understand and interpret the observed intensity data time-dependently, areas in the filament, the plume emission confined in the channel and the plume emission above the surface were selected. In this way photoluminescence as well as transient defect emissions were visualized in the time range of t = 0…200ns. The relative brightness values were captured for calculating the lifetime of the emission and compare them between the pulse and the burst mode. For getting the lifetime of the filament brightness intensity values were taken from a region of 3x3µm² around the brightest spot in the forming filament (Fig. 4, first row). The values were added and afterwards divided by the number of pixels in order to get an integrated value for the filament brightness in the burst mode. The same method was applied to appropriate ICCD images of the pulse mode which were taken with the same laser and focusing conditions. The values for the burst and the pulse mode show a strong time-dependency (Fig. 5). The red arrows in the diagram indicate the four hitting laser pulses in the burst mode. In the pulse mode only the first pulse hit the sample.

Concerning the burst and the pulse mode an increase in the filament brightness was observable after the first pulse hit the sample. A local maximum was achieved in both investigated modes at t = 10ns after the first pulse hit. It is remarkable that the observed photoluminescence from the filament tracks was delayed. Concerning the arrival time of the pulse at t = 0ns the maximum filament brightness was delayed by t = 10ns, the physics of which needs to be further investigated.

After the maximum filament brightness has been achieved in the pulse mode it decreased over the time. A double exponential decay was fitted to the data and resulted in a long lifetime of t_p1 = (13.8 ± 3.2)ns and a short lifetime of t_p2 = (4.1 ± 2.9)ns.

In the burst mode the local maximum brightness increased with increasing number of pulses and stayed nearly constant for the third and the fourth pulse. Between the pulses the filament brightness did not decrease to zero but remained at a certain value. This effect is due to heat accumulation which takes place at high repetition rates (here f = 38.5MHz). Compared to the pulse mode also here a delayed photoluminescence from the filament was observed after each pulse. The brightness decreased double exponentially after the fourth pulse. The long and the short lifetime for the burst mode were t_b1 = (80.7 ± 5.9)ns and t_b2 = (4.0 ± 0.6)ns, respectively.

The observed difference in the long lifetimes for the pulse mode t_p1 and for the burst mode t_b1 can be explained with thermally activated photoluminescence. Due to heat accumulation effects in the burst mode the filament brightness decreased over a longer time period than in the pulse mode.

3.5 Channel brightness and ablation plume

Not only the filament brightness was investigated time-dependently but also the brightness of the ablation plume above the surface and the brightness of the channel which forms the hole. The goal is the understanding of the channel and ablation plume physics regarding high repetition rate laser machining. For this purpose integrated ablation plume brightness data were taken in a region of 13x13µm² 65µm above the surface. Additionally, data for the channel brightness were taken in a rectangle consistent with the hole. For the 1st, the 25th, the 50th and the 200th burst brightness data for the channel and the ablation plume were captured every t = 2ns over a time period of t = 200ns (Fig. 6 ).

 

Fig. 6 (Color online) Brightness of the channel (black line) and of the ablation plume (red line) after the 1st, the 25th, the 50th and the 200th burst in the burst mode. The brightness is given in arbitrary values and the red arrows indicate the four hitting laser pulses in the burst mode.

Download Full Size | PDF

Corresponding to the arrival time of the four pulses in the burst (with a time delay which is also observed here) the channel and plume brightness exhibited four local maxima in the 1st burst. The channel brightness was increasing with each pulse due to heat accumulation effects. The ablation plume brightness slightly increased with each pulse and showed the smallest local maximum for the fourth pulse. Material was removed from the surface of the glass and expanded into the air as it can be seen in the brightness of the ablation plume (Fig. 6).

At the 25th burst the channel lighted up again and material was ablated. It is remarkable that the maximum of the ablation plume brightness was delayed. It did not start until the second pulse hit the sample. Actually, material was still removed but it was more difficult to escape from the already existing hole. Two pulses were needed to really remove the material out of the hole.

The data for the 50th burst showed a similar behaviour. Even three pulses were needed to ablate the material and let it escape from the inside of the hole. The etch rate slowed down even more. The channel brightness still showed a strong emission which resulted from the excitation of the solid walls in the hole.

After the 200th burst the hole was that deep that almost no material could be removed from the hole. The etch rate was almost zero and near the saturation effect. The channel brightness did not indicate the arrival of the individual pulses in the burst train anymore. Luminescence from the walls of the hole through reflection and excitation and from the pulses in the burst overlapped. Thermally activated photoluminescence might lead to a longer decay of the signal.

If the data of the plume brightness are compared to the etch rate of the drilled hole a strong correlation can be seen. Reaching the saturation depth for the hole at which no material was removed any more the plume brightness decreased and became almost zero. This correlation leads to an in situ monitor and control for the hole drilling of transparent materials.

The appropriate data for the pulse mode with the same pulse energy showed that no effective machining is possible. At the 1st burst (with only one pulse) the ablation plume brightness reached only a third compared to the data of the burst mode (not shown here). At the following bursts no significant emission was observable. In conclusion, an effective machining of transparent materials is only possible using the burst instead of the pulse mode.

4. Summary and outlook

For the first time to our knowledge, time-resolved optical images were taken with an intensified CCD-camera during burst laser machining of deep holes in BK7 glass. Transient effects like the formation of filaments and the expansion of the ablation plume were observed on a nanosecond time scale from 2ns to 400ms. Plume-shielding effects caused by multiple pulses in a burst train which interacted with ablated material confined in the created hole have been temporally investigated. The shown correlation between the ablation plume brightness and the etch rate of holes in BK7 leads to in situ process monitoring and the controlling of hole drilling. The investigations demonstrate that high repetition rates within a burst train improve the glass ductility during laser machining compared to single pulses causing precisely shaped and crack-free high-aspect ratio holes. Moreover, a particular role of laser-generated filaments regarding hole guiding was considered in this work.

In conclusion, a much better and more effective hole drilling in glass has been achieved using the burst instead of the pulse mode furthering the control of ultrafast laser-micromachining.