1. Introduction

There is great merit in integrating sensitive analytical technologies onto microfluidic platforms for portability to perform testing at remote locations. For example this would be useful for applications such as water quality monitoring. Analytical detection techniques can broadly be classified as labeled and non-labeled. Labeled detection methods such as fluorescence detection [1] have already been adapted to microfluidic platforms. However, these techniques require that the samples of interest be tagged with particles such as organic dyes or quantum dots. For applications that require a large number of samples (and tests), such as elemental analysis for water quality assessment, labeling can be problematic requiring sample preparation steps that are expensive due to the tags and labor costs. Furthermore, although the sensitivity of these labeling methods is at the single-molecule level, they are typically limited by the lifetime of the fluorescent tags.

Laser-induced breakdown spectroscopy (LIBS) is an analytical technique which has drawn much attention in the last decade [2–4] due to its speed, flexibility and non contact nature. LIBS uses a short intense laser pulse to break down a sample, creating a plasma formed by the elemental constituents of the sample. As the plasma cools, the electrons and ions recombine and de-excite, emitting electromagnetic radiation with a spectrum that is characteristic of the elemental composition of the sample being tested. The key advantage of LIBS is the ability to perform a simple, rapid and minimally destructive analysis without any complicated sample preparation. Typical applications using the conventional LIBS technique utilize laser pulse energies in the range of 10–100 mJ and their typical laser focal spot sizes are on the order of hundreds of microns. These make use of high laser pulse energies since they are applied to large volume sample analysis, hence requiring large laser systems, which are not suitable for portable spectrometers and testing in the remote locations. However, more recently there have been studies [5–10] to scale the laser pulse energies down to the hundreds of microjoule to create the LIBS plasma while still realizing sensitivities comparable to the conventional LIBS technique, hence suggesting that portable µLIBS systems can potentially be realized. This regime of operation is termed as µLIBS. The pulse energies, on the order of a hundred of microjoules can be produced by compact fiber or microchip lasers [11–13]. The use of microchip lasers for µLIBS of low-alloy steel using 50 µJ pulses has already been demonstrated by Lopez-Moreno et al. [14]. We expect that µLIBS will assist in the development of portable LIBS systems and improve the utility of spectroscopy technique for field applications. Recently, there have been a few demonstrations of LIBS based detection (large volume and not on a microfludic chips) to various biomedical applications such as the detection of bacteria, spores and bioagents [7, 15, 16], and also cancer cells [17]. Another potential application of LIBS is the detection of low concentration elemental species in fluids either for simple chemical analysis or for analysis of cell contents in a fluidic stream which have thus far not been explored. The feasibility of detecting and quantifying sodium and potassium in single human red blood cells using LIBS has been reported earlier [18]. Though impressive, this demonstration makes use of a fluidic jet as a sample source, which we believe is suitable for use in a laboratory rather than a portable setting. The detection of contaminants in water using bench-top LIBS systems has been studied by several research groups with a variety of approaches for sampling specific volumes of fluid which include water jets [19–26] and breakdown within a bulk water sample [27, 28]. It has been found that in general, focusing laser pulses inside the bulk fluid generates more continuum emission and weak line emission with rapid recombination because of the high density plasma formed, leading to reduced detection sensitivities. For these reason it is a significant advantage to perform LIBS on minute sample droplets of the fluid.

Analysis of single isolated droplets by sampling a droplet stream has been previously reported using LIBS where an improvement over probing bulk liquids was reported [29]. Droplet dispensers using the microfluidic chip format have recently found applicability to several fields [29] but have not yet been applied to LIBS. Using µLIBS to analyze the content of a monodispersed droplet allows us to retrieve both qualitative and quantitative information of a sample while requiring low laser energies, the key aspect that needs to be dealt with to realize a portable system. An important implication of making use of µLIBS on monodispersed droplets emanating from microfluidic chips is the possibility of integration of several analytical techniques that have already been demonstrated within microfluidic platforms and eventually direct this work towards life-science technologies.

There are several microelectromechanical (MEMS)-based droplet dispensing technologies applied for various applications, e.g. resistive heaters [31, 32], piezoelectric [33, 34] and electrohydrodynamic (EHD) [35] methods. Of these, we believe that the resistive metal thin film heating is appropriate for use here due to its relatively low power consumption and ease in manufacturability. Furthermore, integrating thin film (platinum) resistive heaters within microfluidic chips is now a standard procedure using standard lab-on-a-chip (LOC) microfabrication technologies [36]. As a first step in this direction we demonstrate a droplet dispenser system which partially extrudes a droplet at a controlled exit orifice for µLIBS analysis.

There have been similar demonstrations of integration of droplets and analytical detection technologies, but none with LIBS-based techniques. Leonard et al. [29] demonstrated the use of a commercial thermal ink-jet based printer head to enable the rapid and reliable deposition of biological samples onto membranes for subsequent luminescence detection. Sung et al. [35] fabricated an EHD-based device for delivering ~4 pL volumes of analyte, which can potentially be connected to a biosensing unit, such as atomic mass spectrometry (AMS), mass spectrometry (MS) or protein microarray-type biochips. However, we believe that the integration of LIBS-based techniques with microfluidics will be highly suitable for a sensitive and a truly portable yet inexpensive analysis platform.

To demonstrate the applicability of this µLIBS-microfluidics integration, we perform sensitive detection of sodium by sampling hemispherical partially extruded monodispersed microdroplets generated by thermal actuation using a platinum resistive element within a microfluidic chip. This integration is a significant technological advancement for LIBS and one that will greatly broaden its range of applications.

2. Methods and materials

2.1 Microchip architecture and operating principle

The microchip consists of two layers (Fig. 1), the bottom layer is a 1.1 mm thick borofloat glass (Paragon Optical Company, PA, USA) on which platinum (with a titanium adhesion layer) is patterned as microheaters (details in [36]), while the top layer is made of pre-cast poly(dimethylesiloxane), PDMS (HT-6135, Bisco Silicons, Elk Grove, IL, USA).

 

Fig. 1. (a). Conceptual diagram showing the integration of µLIBS on a microfluidic platform that extrudes droplets and, (b). Assembled microchip architecture for continuous flow elemental analysis.

Download Full Size | PDF

The thickness of the PDMS layer is 0.254 mm and microchannels are patterned using a laser to form the fluidic channel, reservoir and an orifice. The PDMS and glass are irreversibly bonded by exposing the PDMS in a custom-built UV/Ozone cleaner for 6 min and then applying the PDMS on a clean glass surface [37].

To create a micro-droplet, the sample chamber is first loaded with sample solution. A microsecond pulse of tens of volts (section 3.1) is then applied across the platinum heater. The heater super-heats and vaporizes the surrounding solution instantaneously, forming a microbubble that partially extrudes a small volume of fluid in the form of a droplet through the orifice in the PDMS layer. Subsequently µLIBS probing was done while the droplet was well localized at the exit of the orifice.

2.1.1 Fabrication of microheaters

The microchip was designed in L-Edit v3.0 (MEMS Pro 8, MEMS CAP, CA, USA) and transferred to a mask wafer using a pattern generator (DWL 200, Heidelberg Instruments, CA, USA). The 4” ×4” borofloat glass substrate was cleaned in hot Piranha etch (3:1 of H2S04:H2O2). The glass substrate was sputter coated with 20 nm of Cr. The substrate was then spin-coated with AZ 4620 photoresist and soft baked (100°C for 90 s). The heater patterns were photolithographically transferred using a mask aligner (ABM Inc., CA, USA). The UV-exposed regions of photoresist on the substrate (14 s, 356 nm and with intensity of 19.2 mW/cm²) were then removed using an AZ developer. The exposed regions were then removed by a Cr etchant. The wafer subsequently was sputtered with Ti (20 nm)/Pt (200 nm). Next, a lift-off process was performed by immersing this wafer in an ultrasonic bath filled with acetone. The remaining Cr layer was then removed by immersing the wafer in chrome etchant, leaving the patterned resistive thin-films of Ti (20 nm)/Pt (200 nm). The substrate was then diced and bonded with laser patterned PDMS.

2.1.2 Micromachining of the PDMS layer

The microchannel, sample chamber, reservoir and orifice on the PDMS were laser micro-machined. Details about the laser drilling setup and hole characteristics can be found in [35]. Briefly, the laser source used was a Nd:YAG laser (Big Sky, Ultra CFR, Montana, USA) at its 4^th harmonic at 700 µJ per pulse running at 20 Hz. The laser was focused down to a 50 µm spot size on the PDMS surface using a 12.5 cm focal length aberration corrected triplet lens. The dimension of the reservoir in the PDMS that is formed by laser ablation is 100×100 µm with a depth of 40 µm (PDMS membrane thickness 254 µm). The width of the channel is 50 µm and the depth is 20 µm. To form the PDMS channel, the laser fluence was ~4 J/cm² and the PDMS membrane was moved using an automated stage at a speed of 10 µm/s to pattern the microchannels in PDMS. The orifice with a diameter of ~25 µm on the PDMS was drilled by 500 laser shots on a given spot using the same fluence. The debris deposited on the orifice and the channel surfaces were removed by a high pressure air jet. The PDMS membrane used here has a protective layer (as supplied by the manufacturer) that was retained during micromachining, and thus, debris landing on this protective layer was removed prior to PDMS/glass bonding.

2.2 LIBS setup

2.2.1 Bubble imaging setup

The experimental setup for imaging the formation of a microbubble is based on a stroboscopic system as the nucleation process occrs on the microsecond time scale. A delay pulse generator (Stanford DG535, Rohde & Schwarz, Munich, Germany) was used to control the delay between the heater temperature pulse (for droplet extrusion), the illumination system, and the imaging system. A Nd:YAG laser (Big Sky, Ultra CFR, Montana, USA) at 532 nm with a pulse width of 10 ns was used as the illumination source. This pulse of 10 ns was sufficiently short to capture the nucleation process. The imaging system equipped with a 10 X objective lens gave a resolution in the micron range, thus assisting in estimating the size of the droplet. By adjusting the delay in the probe pulse, the image of the nucleation process could be observed.

2.2.2 Laser-induced breakdown spectroscopy setup

The experimental setup used for LIBS analysis is shown in Fig. 2. The ablation pulse used was generated from a frequency-quadrupled Q-switch Nd:YAG laser (Big Sky, Ultra CFR, Montana, USA) at 266 nm with a pulse width of 10 ns. The pulse energy was 200 µJ and was monitored using a calibrated photodiode (Hamamatsu, R1193U-52, Bridgewater, NJ, USA). Pulse energy was controlled using a combination of a half-wave plate and a Glan-Taylor prism. The beam was focused by a 10X objective lens (Optics For Research, 10x, 15 mm working distance, NA=0.25) to form an ~10 µm spot on the target (the hemispherical microdroplet). By adjusting the delay between the heater pulse and the laser, the droplet was ablated when it was half exposed on top of the orifice. The plasma created was imaged in the x-y plane through the same 10 X objective lens (parallel to the microchip surface) and projected onto the entrance slit of a 1/4 m, f/3.9 imaging spectrometer (Oriel MS260i, Newport, California, USA) equipped with an intensified charge coupled device (ICCD) camera (Andor iStar DH720, Andor, Belfast, Northern Ireland). A grating with 2400 lines mm-1 was employed for spectral dispersion in the spectrometer. A slit width of 300 µm and effective spectral resolution of 0.45 nm were used for this measurement.

 

Fig. 2. LIBS experimental set-up consisting of a frequency quadrupled Nd:YAG laser producing pulses up to ~260 µJ to ablates the microdroplet target. LIBS emission is observed along the laser axis.

Download Full Size | PDF

3. Results and discussions

3.1 Nucleation process

There are several factors that contribute to the formation of the microbubbles, we find in the literature that there are two main types of microbubble generation mechanisms [28]. We believe that the bubbles here are generated due to a combination of mechanisms. One mechanism is the spontaneous nucleation that can occur due to fluctuations in the temperature as the sample is heated, which causes a vapor film to grow due to superheating of the liquid. The other mechanism is homogeneous nucleation wherein the gas trapped on the surface of the heater material may expand and grow.

In our set-up a square wave pulse of 14 V amplitude and 15 µs duration was applied across a 90 µm wide folded stripline heater (50 Ω) with geometry as shown in Fig. 3(a). A series of images was captured to study the nucleation process is shown in Fig. 3. After 5 µs, nucleation was first observed in the close vicinity of the heater regions. They grew rapidly for the next 5 µs and merged together into a bigger bubble. The bubble expanded until the electrical pulse was turned off and collapsed within the next 10 µs.

 

Fig. 3. Microbubble nucleation in action during the application of short electrical pulses. Nucleation process of a heater activated with a 14V 15µs pulse at different times, from left to right (a) t=-5 µs (before the start of the heating pulse); t=(b) 0; (c) 5 µs; (d) 10 µs; (e) 15 µs; (f) 20 µs and (g) 25 µs.

Download Full Size | PDF

3.2 Microdroplet extrusion and operation

To create a microdroplet, the sample liquid was first introduced into the reservoir by a syringe through a 0.2 µm pore size particle filter. Then, the liquid was directed to the sample chamber through the microfludic channel by applying external positive pressure. Inside the chamber, the sample was heated by the microheater and liquid was extruded from the orifice on the top of the chamber. The initial nucleation and growth of the bubble pushes the liquid out from the orifice, creating a droplet. (Fig. 4) As another heater with a resistance of 60 Ω and a line width of 50 µm was used in this experiment, the amplitude and the width of the heater pulse were readjusted to 13 V and 5 µs, (established empirically). In a particular example presented here liquid was extruded from the chamber of the microfluidic chip in the form of a hemispherical droplet during the first few microseconds and grew up to ~28 µm in diameter while staying attached to the surface of the orifice, which corresponds to a volume of ~6 pL for the hemispherical droplet. It was found that the size and extrusion speed of the droplet could vary by 15%, making it difficult to adjust the spatial and temporal settings for ablation. Thus, the ablation beam was aligned normal to the device surface and focused right above the orifice where the sample is extruded. The laser beam was focused such that there was no damage to the orifice of the microfluidic device. It could be that this variation is due to details of the nucleation of the vapor bubble by the heater which could vary from pulse to pulse.

 

Fig. 4. Droplet formation captured over a period of a 5 µs. Sequence of droplet formation using a 13V 5 µs pulse across a 60 Ω heater. The images were captured using the setup described in Fig. 2. Images from left to right show the bubble at (a) 1 µs; (b) 2 µs; (c) 3 µs; (d) 4 µs; and (e) 5 µs. A reflection of the partially extruded microdoplet from the PDMS surface is seen in the bottom of each image.

Download Full Size | PDF

3.3 LIBS study

The spectral emission lines of interest for Na are the 588/589 nm doublet lines, which involve transitions from 3 ²P_1/2 upper states and 3 ²P_3/2 to the 3 ²S_1/2 lower states [38]. The sample tested here contained 200 ppm by weight of Na as an analyte in the form of NaCl dissolved in distilled water. LIBS was applied on the hemispherical microdroplet which was generated as described in the methods section.

The gate width of the ICCD detector was set to 1 µs at full gain while the gate delay varied. The single shot Na 589 nm doublet spectra with different gate delay times were shown in Fig. 5. The signal is defined as the normalized intensity under the Na doublet with the average background subtracted, and the noise as the standard deviation of the background scaled to a bandwidth equivalent to that used to measure the line emission taking neighboring channel pixel correlations into account. Initially, at 100 ns gate delay in addition to the sodium lines many photons are detected across the full spectrum range viewed due to continuum emission form the plasma. When the plasma cooled down, the doublet could still be resolved but the line intensity also dropped significantly. The Na lines completely disappeared by 800 ns. The SNR of the spectrum with a gate delay of 100 ns was 10 resulting in a 3 σ limit of detection (LOD) of 60 ppm.

 

Fig. 5. Single shot LIBS spectra on a micro-droplet. The pulse energy for ablation was 200 µJ, and the spectra were obtained at a gate width of 1 µs with gate delay varied: (a) 100 ns; (b) 200 ns; (c) 500 ns; and (d) 800 ns. The corresponding SNR are 11, 4, 2, and 0 respectively.

Download Full Size | PDF

However, certain applications may require a better LOD. To improve the SNR, an average of 100 shots were accumulated resulting in SNR of 100 with the same gate delay of 100 ns (Fig. 6) and corresponding to a 3σ LOD of 6 ppm corresponding to a mass within the hemispherical droplet of 360 fg. Janzen et al. [39] using 100 shot average detection with much higher laser energies found a 3σ LOD of 0.2 ppm for Na using the same emission lines. We attribute this increased sensitivity to the fact that they used a higher energy 25 mJ, 266 nm, laser pulse and a larger 80 µm droplet. However, using larger laser pulse energies limits the portability of their system.

 

Fig. 6. Multiple shots accumulation LIBS spectra on microdroplets. 100 shot accumulation LIBS spectra on micro-droplet. The gate delay was set at 100 ns.

Download Full Size | PDF

The SNR distribution for 100 single shot data shows that majority of the spectra have a signal to noise ratio (SNR) between 10 and 20, and the standard deviation for the 100 data set is 28.6. The fluctuations in the SNR were due primarily to the variations in droplet size (±15%) and in laser pulse energy (±10%), leading to a change in the focusing condition of the ablation beam, the sampling volume, and the plasma conditions. Overall, we demonstrated 3σ LOD’s of 60 ppm and 6 ppm for single or 100 shot detection of sodium in water with 200 µJ laser pulses under the current conditions.

4. Conclusions and future directions

A microfluidic-based LIBS system suitable for inexpensive and portable elemental detection/analysis devices has been demonstrated. We believe it will eventually have important applications in elemental analysis for monitoring of water quality for public health and biomolecule detection in point-of-care diagnostics. Our results indicate that µLIBS can be integrated with a microfluidic platform for such non-labeled analysis. This extension of LIBS could be implemented by integrating compact laser sources and filtered photomultiplier tubes (PMTs) or avalanche photodiode arrays (APDAs) – these are more compact and hence appropriate for portable platforms. Such fully integrated systems will require further miniaturization of lasers and detector systems that is occurring naturally due to increasing demand to use spectroscopic techniques in numerous applications. The statistical spectral analysis results for 100 microdroplets suggest that the reproducibility of the platform needs to be addressed if high precision single droplet measurements are required. The fluctuation of the SNR in the data in part is due to the instability in the microdroplet size and the microdroplet generation that resulted in varying degrees of ablation from shot to shot. Further, we are also currently characterizing bubble retracting post ablation and the effect of debris. Even though we show the feasibility of using this technique with Na, this technique can potentially be used for less sensitive elements such as Ca by integrating more number of laser shots while keeping the detection time short by making use of microchip laser which are capable of multi-KHz. repetition rates and also by making use of larger collection apertures coupled to large aperture sensitive detectors such as PMTs.