1. Introduction

Microfabricated pores in thin membranes have found a wide range of applications, such as cavities for flatter and shorter bonds in microwave circuits [1], vias in microelectronic packaging [2], interconnects in 3D microfluidic channels [3,4], and particularly, as patch holes in electrophysiological measurements [5–7]. The thin films can be made of non-transparent materials such as silicon [5], and teflon [7], or transparent materials such as glass [4,6], quartz [6,8], polydimethylsiloxane (PDMS) [3], and polymethylmethacrylate (PMMA) [9]. To make these pores, several techniques have been developed ranging from mechanical punching or drilling [10], micromolding and casting [7], optical lithography combined with reactive ion etching (RIE) [5], latent ion-track etching [6], and laser ablation [1–4,8,9]. Among these approaches, laser drilling has several advantages: first, it involves only one fabrication step and thus delivers results very quickly; second, as will be shown later, it can produce very small pores on a relatively thick substrate; third, it is inexpensive, i.e. does not require access to cleanroom or use of toxic agents, nor does it need large sophisticated facilities [6].

In this paper we report on a novel micromachining process using a high-power excimer laser to drill nano/micro-scale, smooth pores into borosilicate glass membranes. By sandwiching two glass slides with a thin layer of liquid in between (see Fig. 1 ), we not only obtain homogenous surfaces as the liquid media helps to melt the debris, but also are able to reduce the pore sizes much below the laser wave length of 193 nm. More specifically, we are able to make small pores ranging from several hundred micrometers to 90 nm in diameter, in glass slides with thickness from 150 µm to 300 µm. We also reduce the surface roughness to less than tens of nanometers, as compared to that of hundreds of nanometers if fabricated by direct drilling. Furthermore, we found a parameter range in which we observe a unique crater feature on the exit side with extremely smooth surface. This is of particular interest in biological applications where surface roughness is vital. For example, our technique can help ion channel screening [5–7,10,11], which has been widely adopted in the pharmaceutical industry. In this case high seal resistance, usually in the GΩ range, is needed for accurate transmembrane current recording. This requires a tight contact between the cell membrane and the pore, which has to have a smooth surface to prevent any possible leak. Placing the cell in a carefully machined crater of similar dimensions will naturally improve the measurement accuracy.

 

Fig. 1 Schematic sketch of the “sandwich” drilling method, where a thin layer of liquid is kept between two glass slides. Also shown are the tapered nature of the hole being drilled and the shockwave from laser ablation, which will be discussed in later sections.

Download Full Size | PDF

2. Experimental

For the work discussed here we use borosilicate glass slides (Fisher Scientific, 22 × 22 mm² square and 150 µm thick. RMS surface roughness is 10 to 30 nm). They are ideal for the ion channel recording due to their small dielectric constant and good mechanical properties. An ArF excimer laser (Lambda Physik LPX-200, 193 nm) is used for membrane fabrication. For borosilicate glass, the absorption coefficient at 193 nm wavelength is 64~105 cm⁻¹. At the output of the laser we operate with a 12 × 23 mm² beam. A stencil metal mask regulates the final shape of the beam. By choosing masks with different aperture sizes, the beam size can be changed (The results reported here all use a 170 μm aperture mask). A beam delivery system (JPSA Microtech) guides the laser beam onto the sample, which is placed on a servomotor-controlled stage. The accompanying software provides precise control of the stage movement. It also controls a variable attenuator, which can specify how much of the laser beam is to be transmitted. Other adjustable parameters are laser beam energy, pulse repetition rate, and the total number of pulses.

Individual glass slides can be drilled directly. During operation the laser beam is irradiated perpendicularly onto the front surface of the sample, causing vaporization and melt ejection of the sample material. Commonly, when the beam exits from the back of the slide, it also carries the molten debris, which after being cooled down, will pileup around the exit. This results in a rough surface. In order to prevent this, we developed a new fabrication process in which the target slide is placed on another slide with a thin layer of liquid sandwiched in between. We call this “sandwich” drilling method (Fig. 1). The liquid layer can be applied as a small droplet on the bottom slide, the target slide is then pressed on top; or we can put a PDMS fluidic channel between the two slides to hold the liquid, thus the thickness of the liquid layer can be adjusted.

3. Results and Discussion

Under the same laser parameter settings, we compare two glass slides: one drilled directly and another using the “sandwich” drilling method. Specifically, the laser output power is fixed at 5 W and the variable attenuator is set to allow 73% of the total beam transmission. In both cases the laser is first operated for 800 pulses at a repetition rate of 50 Hz, and then followed by 1500 pulses at 100 Hz. Such a drilling scheme is helpful because at low pulse repetition rate the laser beam contains higher energy and tends to drill through the material faster but makes rougher surface, while at high pulse repetition rate the opposite is true. By combining the two we not only make a hole quickly through the substrate, but also with polished, smoother surface.

As mentioned before, molten debris tends to pile up at the exit pore in direct drilling. This is shown in Fig. 2(a) , where the re-solidified debris spills out from the pore. The biggest debris is shown to be larger than 200 nm. In comparison, our “sandwiching” technique avoids this problem. It is slightly different from other liquid-assisted drilling mechanisms, such as the laser induced backside wet etching (LIBWE) method [4,8], where the laser beam is focused at the back of the top slide such that the liquid medium in contact with it facilitates the drilling process from the very beginning. Instead, in our case during operation the laser beam is focused directly on the top surface of the slide. As it mills deeper, the laser beam becomes weaker due to focal change (The depth of field for the laser beam objective is 1 to 5 μm) and heat dissipation. Therefore the exit pore is smaller than the actual beam size, which is 12 µm. Once the laser beam drills through, the sandwiched layer of liquid helps to carry away the debris before they melt into the glass slide surface. As a result, the roughness is greatly reduced, as shown in Fig. 2(b).

 

Fig. 2 Comparison of (a) direct drilling and (b) “sandwich” drilling. In both cases the laser output power is 50 mJ with 73% transmission rate. A combination of 800 pulses at 50 Hz and 1500 pulses at 100 Hz is used. Scale bar is 200 nm.

Download Full Size | PDF

Comparing Figs. 2(a) and 2(b), it is obvious that aside from being smoother, the pore obtained with this sandwich drilling technique also shrunk further, from 1600 nm down to 400 nm. This allows us to make small pores not possible with conventional drilling. We have observed that immediately after the laser beam penetrates through the slide, the liquid underneath the slide gets drained into the pore. This influx of liquid will absorb the heat from the laser beam and protect the surrounding walls against being ablated. Hence the pore is not enlarged as much as direct drilling.

We also study the effect of laser beam attenuation on drilling. In this set of experiments the laser power is fixed at 6 W. De-ionized (DI) water is used as the liquid layer sandwiched between the two glass slides. The output laser beam can be attenuated by tilting a grating mirror, whose angle corresponds to the transmission rate in percentage. As a control study, the same experiments are performed with direct drilling. A summary of the results is shown in Fig. 3(a) . Again, the liquid layer helps to reduce the pore size. Moreover, the curve for the sandwich drilling technique (red squares) becomes steeper at low transmission rates. The trend indicates that for 6 W of laser power, direct drilling (black triangles) delivers pore sizes of only 2.2 µm, whereas with our technology sub-micrometer sized pores can be achieved.

 

Fig. 3 (a) Influence of laser attenuation on pore size for both direct drilling and “sandwich” drilling, (b) Contour map of pore sizes vs. laser power and transmission rate. Unit for scale bar on the right is micrometer.

Download Full Size | PDF

It will be interesting to replace the DI water with a liquid that absorbs the laser beam more strongly, as it might reduce the pore size even further. Repeating the above experiments for different laser powers, we obtain conditions to fabricate pores with a range of specific diameters, which is summarized in Fig. 3(b). The color coded scale gives the pore diameter in microns. It is clear that the slope of the contour lines are steeper at low power levels, which indicates the change in transmission rate is not as effective in these areas. For a specific power level or transmission rate, the achievable pore size cannot go below or above a certain range. Another observation is the pore sizes follow a distribution that peaks in the range between 0.9 µm to 2.5 µm. In other words, pore sizes in this range can be made over a relatively wide operation conditions.

Figure 4 provides examples of nano-pores drilled with diameters in the range from below 100 nm to 300 nm. We note again that the pore diameter shown in Fig. 4(a) is a factor of two below the wavelength of the laser used. As noted before the pores show a very smooth surface with a clear circular opening. In addition, the scanning electron beam pictures reveal a bright outer ring, which appears as a corona of the pore. Further investigation showed that this corona is a crater-like opening formed around the pore during the sandwich drilling. The width and depth of the craters depend on the laser power, repetition rates, and sandwich “filling” used.

 

Fig. 4 Nano-pores with diameters in (a) 100 nm and (b) 300 nm. Scale bar is 200 nm.

Download Full Size | PDF

In order to elucidate this crater formation further, we ran another set of experiments shown in Fig. 5 . Here two glass slides are tightly bound together as it can be achieved using wafer bonding techniques. The resulting crater has a diameter of roughly 10 microns, a factor of ten above the actual pore diameter. Furthermore, the crater consists of an inner and an outer region, separated by a clearly defined ring. While the inside region is extremely smooth with a surface roughness below 50 nm, the outer region appears to be rougher. The cross sectional view in Fig. 5(b) indicates that the two regions of the crater are well defined into the depth of the material. This cut through the glass chip was obtained by using a Focused Ion Beam (FIB) machine, which enables direct milling of the micron sized samples. The cone-like shape is clearly identified with a smooth transition narrowing down from the crater’s outer ring to the micron-sized pore. In contrast to the result shown in Fig. 2(b), here we apply very high pressure in sandwiching the two glass slides. During the subsequent drilling of such tightly bound materials the molten glass cannot escape from the pore’s vicinity. Furthermore, the intensity profile of the laser beam generally resembles a Gaussian distribution, which induces a temperature gradient of the same shape. This temperature gradient in turn creates a surface tension difference between the center and the edge that results in a thermal capillary force. According to Kingery [12], for borosilicate glass, the surface tension γ increases with increased temperature T (dγ/dT>0). This means the thermal capillary flow of the molten glass is driven from the cold periphery to the hot center, which explains why the pore takes a crater shape. It is also well known that laser ablation of solid materials generates a shock wave due to the explosive evaporation of the material, which can be studied by shadowgraph imaging [13–15]. It was found that the shock wave produced by a similar laser propagates at 47 km/s and reaches pressures as high as 800 atm. The temperature behind the shock wave front is still at around 2000 °C after 1 µs [16]. Under such tremendous heat and pressure, the molten materials are squeezed against a limited surface area and become very dense and fine, hence makes a very smooth surface.

 

Fig. 5 Scanning electron micrograph (SEM) images of (a) a crater at the beam exit and (b) the side view of a crater after being milled by focused ion beam (FIB). Both images are viewed from an angle. The pores are drilled at 8 W laser power with 87% transmission, for 2500 pulses at 100 Hz. Scale bar is 2 µm.

Download Full Size | PDF

4. Conclusion

We demonstrate a one-step sandwich method, where a thin layer of liquid is kept between two glass membranes, for drilling small pores from several hundred micrometers in diameter down to only 90 nm by using an excimer laser. Our method produces conical holes, tapering from the top to the bottom, thus enabling us to make pores smaller than the beam spot size. The liquid serves to absorb the energy of the laser shockwave and to remove the debris from the drilling region once the shock wave front reaches the back of the slide. The resulting pores are very small and very smooth. In addition, we have observed the creation of extremely smooth craters at the beam exit side when two glass slides are tightly bound together. Such a feature, which is uniquely obtained with the ’sandwich‘ drilling method, is beneficial for making good seals between substrates and cell membranes in whole cell recording, where high resistance from the tight contact is often crucial in measuring small electrical signals. Moreover, compared to the pores obtained through chemical etching, we are able to make them with comparable quality on thicker membranes. This provides another advantage in ion channel recording because thicker slides means smaller slide capacitance, and hence smaller background noise. In conclusion, slides made with this technology are ideal for transmembrane conductance measurements and cell detection.