Article pubs.acs.org/ac
Pillar Cuvettes: Capillary-Filled, Microliter Quartz Cuvettes with Microscale Path Lengths for Optical Spectroscopy Gregor Holzner, Frederik Hermanus Kriel, and Craig Priest* Ian Wark Research Institute, University of South Australia, Adelaide, SA, Australia ABSTRACT: The goal of most analytical techniques is to reduce the lower limit of detection; however, it is sometimes necessary to do the opposite. High sample concentrations or samples with high molar absorptivity (e.g., dyes and metal complexes) often require multiple dilution steps or laborious sample preparation prior to spectroscopic analysis. Here, we demonstrate dilution-free, one-step UV−vis spectroscopic analysis of high concentrations of platinum(IV) hexachloride in a micropillar array, that is, “pillar cuvette”. The cuvette is spontaneously filled by wicking of the liquid sample into the micropillar array. The pillar height (thus, the film thickness) defines the optical path length, which was reduced to between 10 and 20 μm in this study (3 orders of magnitude smaller than in a typical cuvette). Only one small droplet (∼2 μL) of sample is required, and the dispensed volume need not be precise or even known to the analyst for accurate spectroscopy measurements. For opaque pillars, we show that absorbance is linearly related to platinum concentration (the Beer−Lambert Law). For fully transparent or semitransparent pillars, the measured absorbance was successfully corrected for the fractional surface coverage of the pillars and the transmittance of the pillars and reference. Thus, both opaque and transparent pillars can be applied to absorbance spectroscopy of high absorptivity, microliter samples. It is also shown here that the pillar array has a useful secondary function as an integrated (in-cuvette) filter for particulates. For pillar cuvette measurements of platinum solutions spiked with 6 μm diameter polystyrene spheres, filtered and unfiltered samples gave identical spectra. cells and, for wavelengths less than ∼300 nm, the cells must be UV-transparent, that is, usually quartz. However, the prefix “micro” typically refers to the volume of sample and not the path length. In addition, these cuvettes are difficult to rinse between measurements and aliquots of several hundred microliters is often required. Independent of the cuvette design, the measured UV−vis absorbance, A, increases linearly with analyte concentration, [X], over a specific concentration range according to the Beer− Lambert law:
T
he demand for rapid, simple, and inexpensive analytical methods for unskilled users is growing rapidly, driven by the convergence of innovation in portable electronics, data analysis, and microfluidics/lab-on-a-chip technology. Taking the laboratory to the sample in a compact and inexpensive device has major advantages, including short analysis-to-action turn-around times. Applications in security,1 health monitoring,2 environmental detection,3 and process control in industry4 are but a few examples where rapid on-site analysis is highly desirable. In these examples, sample preparation must not be laborious or time-consuming and conventional laboratory tasks, such as dilution, are a source of error and often impractical. In this paper, we present a micro path length cuvette design that permits straightforward dilution-free spectroscopic analysis of ∼2 μL samples with high molar absorptivity. The cuvette is filled spontaneously and precisely by capillarity and, as it is an open cuvette, can be quickly and easily rinsed between measurements. Spectroscopy offers fast and reliable analysis and can now be achieved using compact and relatively inexpensive instruments in remote locations.5−8 The focus of this paper will be UV−vis spectroscopy; however, preparing liquid films with a precisely defined and reproducible thickness at the microscale is also very relevant to other spectroscopic techniques. Typical path lengths in commercially available UV−vis cuvettes, or “cells”, are between 1 and 100 mm, with a few exceptions. Different cell types include square cells, semimicro cells, micro cells and flow © 2015 American Chemical Society
A = −log Ts = ε ·h·[X ]
(1)
The sample transmittance is defined as (Ts = (I/I0)), where I0 and I are the intensities of incident and transmitted light, respectively. In eq 1, ε is the molar absorptivity and h is the path length. At high concentration, the relationship A ∝ [X] will depart from linearity until the absorbance reaches a plateau. This limit is influenced by the magnitude of ε, which varies greatly for different analytes. In some cases, ε is very large, for example, for blood,9 chlorophyll,10,11 inks,12 and sensitizers.13 Wherever the magnitude of ε and [X] is large, it is necessary to reduce h and dilute the sample before measurement. The latter Received: December 23, 2014 Accepted: April 6, 2015 Published: April 6, 2015 4757
DOI: 10.1021/acs.analchem.5b00860 Anal. Chem. 2015, 87, 4757−4764
Article
Analytical Chemistry is highly undesirable, as it is laborious, time-consuming, and a possible source of error, particularly where sample preparation is remotely carried out by unskilled personnel. For dilution-free spectroscopic analysis of high ε or [X] samples, we developed a quartz “pillar-cuvette” that is spontaneously and precisely filled by capillarity to yield a reproducible path-length of tens of μm, Figure 1. Only one
Figure 1. Schematic of absorbance spectroscopy using a pillar cuvette. Lower right: An image showing a droplet of aqueous sample in contact with a pillar array. A thin film has spontaneously formed in the pillar array in contact with the droplet.
droplet (∼2 μL) of sample is required and the “pillar-cuvette” is open, allowing for quick and easy rinsing between samples. In the experiments presented here, a 100-fold dilution is avoided and the analysis takes only a few seconds without loss of analytical precision. Furthermore, the micropillars play an additional role in removing particulates, which is demonstrated by spiking the solution with 6 μm polystyrene spheres. The latter avoids the need for prefiltration of samples and, therefore, sample handling, equipment, and the general complexity of the analysis.
■
THEORY Sample Loading via Capillarity. The pillar-cuvette investigated in this study takes advantage of “wicking”, which is the spontaneous imbibition of liquid into a small porous structure via capillarity, see Figure 2a.14 Both micro15−17 and nano18−24 structures can drive wicking on surfaces but our focus here will be on highly ordered micropillar arrays. In this study, the pillars are arranged in a square lattice with lattice constant, d, pillar height, h, and pillar width (diameter), w. Taking into account the surface energy and geometry of the solid−liquid, solid−vapor, and liquid−vapor interfaces involved and ignoring gravity (assuming that the parent droplet is smaller than the capillary length), wicking can be thermodynamically predicted for regular arrays of pillars, as shown by Bico et al.:14 1−ϕ cos θ > r−ϕ (2)
Figure 2. (a) Illustration of wicking and nonwicking liquids in pillar arrays and the relevant dimensions of the pillar arrays. (b) Critical contact angle (see discussion) plotted against projected area fraction of pillars with heights from 5 to 50 μm. (c, d) Experimental wicking results for (c) h = 12 ± 2 μm and (d) h = 23 ± 3 μm (the height varies slightly with ϕ due to the sample preparation method, see Experimental Section). The crosses and circles represent wicking and nonwicking events. The solid lines in (c) and (d) represent the critical contact angle for h = 10 and 20 μm, respectively.
prediction is observed due to contact line pinning on the pillars that opposes the wicking effect.26 For pillars with square and circular cross sections, the effect of pinning shifts the wicking criterion to larger (h/w) and (w/d), and, unexpectedly, wicking is prevented again for square cross-section pillars at very high area fraction.26 Clean quartz is completely wetted by water and most other solvents in air, such that the material contact angle is intrinsically small. However, high-energy surfaces are readily contaminated in air, which raises the material contact angle and could prevent wicking altogether in practical applications. Figure 2b plots the predicted critical contact angle against the area fraction of the pillars for different pillar heights. Note that wicking occurs below the curves, that is, wicking is favored for
In eq 2, θ is the contact angle measured on a flat substrate of the same material, or “material contact angle”, r is the Wenzel25 roughness factor (r = 1 + (h/w)4ϕ), and ϕ is the projected surface area fraction of the circular pillars (ϕ = ((π/4)(w2/ d2))); dimensions w, d, and h are defined in Figure 2. When cos θ = (1 − ϕ)/(r − ϕ), the contact angle is at the boundary between wicking and nonwicking, that is, θ is a “critical contact angle”. In practice, departure from this thermodynamic 4758
DOI: 10.1021/acs.analchem.5b00860 Anal. Chem. 2015, 87, 4757−4764
Article
Analytical Chemistry
Equation 7 is unusual for absorbance spectroscopy in that Tp and Tr are relevant to the measurement, despite being included in the spectrum of the blank. This is due to the different transmitted intensities in the two light paths (pillars and reference solution), coupled with only one path being affected when the sample is loaded. Furthermore, light that travels to the detector without passing through the sample is generally termed “stray light” and is undesirable for analysis of high absorbance samples. Here the stray light is that which passes through the pillars. The stray light effect can be avoided, as shown in this paper, by preparing opaque pillars, that is, Tp = 0, which recovers Tm = Ts and, thus, the Beer−Lambert law (eq 1). For pillar cuvettes where Tp ≠ 0, the transmittance of the pillars and reference solution are both required to calculate Ts from Tm. Given that the optical properties of the pillar material and the reference solution are typically known or easily determined, this is not a major barrier to the use of semitransparent cuvettes, as shown later.
tall pillars at high area fractions. Experimental results are shown for water on quartz pillar cuvettes exposed to a laboratory environment or, in selected cases, hydrophobization (Figure 2c,d). The results are in qualitative agreement with Semprebon et al.26 Pinning lowers the critical contact angle below the thermodynamic prediction (eq 2). Experimental results in Figure 2c,d shows that filling of cuvettes with ϕ = 0.05 and h ∼ 12 μm would be less reliable than the other arrays studied because the material contact angle must be less than ∼20° for wicking to occur. Clearly, the height, lattice spacing, and material contact angle of the pillars are critically important for reliable operation of the pillar cuvette. Beer−Lambert Law Derivation: Pillar Cuvette. The linear dependence of absorbance with concentration (Beer− Lambert law) is very useful. Here, we derive the measured transmittance, Tm, for a sample that has pillars embedded in a thin liquid film. To consider the absorbance measured against a reference sample (or “blank”), we must first consider the pillar cuvette filled with the reference solution, as shown in Figure 3
■
EXPERIMENTAL SECTION Solution Preparation. The platinum solution was prepared by dissolving sodium hexacholoroplatinate(IV) from Johnson Matthey in 0.5 M HCl and diluted as required to make standard solutions over the concentration range studied. Pillar Cuvette Fabrication. The pillar cuvettes were prepared using UV-photolithography and plasma etching on optical grade quartz (4″ diameter, 700 μm thick, Shin-Etsu). For opaque pillars, the wafer was first sputter-coated with ∼100 nm thick layer of Cr and ∼50 nm layer of Au in a vacuum (2 × 10−5 mbar). The wafer was dehydrated at 200 °C and cooled to room temperature to spin-coat (Suss Microtec, Delta 80RC) SU8−10 photoresist. The sample was then prebaked, exposed to UV at 365 nm (EVGroup, EVG 620) through a chrome-onglass photomask (Optofab, ANFF). The photoresist was developed and hard-baked (200 °C). The backside of the substrate was sputter coated (HHV/Edwards TF500) with a chromium layer to enable adhesion to the electrostatic chuck in the plasma etching tool (ULVAC 570NLD). Anisotropic plasma (C3F8) etching was used to replicate the SU8 structure in the quartz wafer. The photoresist was removed by etching the Au layer using 4:1 of KI/I2 in water, then washed using Piranha solution (7:3 ratio of 98% H2SO4 and 30% H2O2) at room temperature (Caution: mixing Piranha solution is exothermic and strongly oxidizing). The pillars chosen for this study were cylindrical because pinning effects are reduced compared to other geometries with vertical edges.26 Figure 4 shows scanning electron microscopy
Figure 3. Cross-section of the pillar cuvette (green) loaded with the reference solution (blue), that is, 0.5 M HCl(aq), and the sample (orange) shown separate for clarity.
(reference solution is shaded blue). The incident light, I0, is effectively divided into two light paths: one through the pillars and one between the pillars through the reference solution. The portion of light in each path is related the area fraction of pillars, so that the incident intensity for the pillars and the reference solution are given by ϕI0 and (1 − ϕ)I0, respectively. The quartz substrate (excluding the protruding pillars) can be ignored, since it is optically homogeneous (equivalent to the optical window of a conventional cuvette). It follows that the transmitted intensities for the pillars, Ip, and the reference solution, Ir, are given by Ip = ϕI0Tp
(4)
Ir = (1 − ϕ)I0Tr
(5)
When the sample is loaded, it attenuates the light that is able to pass through the reference sample, Ir, according to the transmittance of the sample, Ts: Is = (1 − ϕ)I0TrTs
(6)
and, therefore, the overall measured transmittance, Tm, in our experiments is Tm =
Ip + Is I p + Ir
=
ϕTp + (1 − ϕ)TrTs ϕTp + (1 − ϕ)Tr
Figure 4. Scanning electron microscopy images of pillar arrays for ϕ = 0.04 (a), 0.08 (b), 0.16 (c), 0.23 (d), 0.31 (e), and 0.39 (f).
(7) 4759
DOI: 10.1021/acs.analchem.5b00860 Anal. Chem. 2015, 87, 4757−4764
Article
Analytical Chemistry
of the liquid, pillar array geometry, and the distance traveled through the array will affect the filling velocity.27 We used optical profilometry (VEECO, WYKO NT9100) to characterize the morphology of the liquid−vapor interface. Figure 5
(SEM) images for pillar arrays with different area fractions. For most coverages, that is, ϕ = 0.04 to 0.23, the pillars are resolved very well; however, defects appear as the lattice spacing of the pillars reduces. The images in Figure 4 are captured at the outer row of pillars in the array, where the defects were more pronounced. Further inspection revealed that only ϕ = 0.37 exhibited significant defects and the latter was used in this study to determine how sensitive the analysis is to the quality of the pillar structure. The actual pillar heights in the cuvettes varied depending on the aspect ratio of the etched region, (h/(d − w)). While the different pillar arrays had slightly different heights (as noted in the Results and Discussion section), the fabrication of the pillar cuvettes was reproducible for a fixed ϕ, w, and d. Height measurements on 10 different locations on a single pillar array gives a representative uncertainty in the pillar height of ±0.33 μm or 3%. Spectroscopy. For all the spectroscopic measurements, the Ocean Optics DT-Mini-2 light source and the Ocean Optics QE65000 detector have been used. The light source and detector are compact and easily coupled with lab-on-a-chip platforms for application in remote locations. A detailed description of sample loading into the pillar cuvette, data analysis method, and other phenomena is given below.
■
RESULTS AND DISCUSSION A conventional 2 mm path length quartz cuvette was used for comparison with the pillar-cuvette. The molar absorptivity, ε, at 259 nm for PtCl62− complex in the aqueous phase was found to be very high (∼24000 L·mol−1·cm−1) so that the measurable concentration range is limited to values less than ∼50 ppm in this cuvette. Industrial processing of precious metals, including platinum, is carried out at much higher concentrations. For example, platinum concentrations in excess of 20 g/L might be expected. In this case, analysis by UV−vis in a 2 mm cuvette requires predilution factors of up to 1000. These dilutions are time-consuming and require large volumes of solvent (here, a 0.5 M HCl aqueous solution), even for very small aliquots of sample; 100 μL sample requires 100 mL of solvent for a single measurement. The additional sample handling may also lead to greater analytical uncertainty and, for hazardous materials, an occupational or environmental hazard. In the following sections, we first address the operation of the pillar cuvettes. Then, absorbance spectroscopy is demonstrated for opaque pillars and semitransparent pillars, with the latter addressing the effect of stray light. Finally, the unique advantage of simultaneous particle filtration and sample analysis will be shown. Cuvette Filling by Capillarity. The UV−vis spectroscopy analysis using the pillar-cuvette proceeds as follows. A single droplet of undiluted sample was placed immediately adjacent to the pillar array, see Figure 1. The droplet volume is unimportant; partially dipping a pillar-cuvette into a bulk liquid will give the same results. Here, the sample volume is ∼2 μL. This is 50 times less than that required for accurate dilution for a conventional cuvette, no additional solvent is required, no special dispensing equipment is required (e.g., a micropipette), and loading and rinsing the pillar cuvette takes only a few seconds. After placement, the droplet spontaneously spreads to meet the edge of the pillar array region, which triggers wicking. The film spreads rapidly through the pillar array (typically between 1 and 4 mm/s, based on high-speed microscopy). The viscosity
Figure 5. Optical profilometry of the liquid (aqueous sample)-vapor interface at the tops of the pillars: (a) ϕ = 0.05 and (b) ϕ = 0.20. (c) Profiles of pillar arrays before and after filling with the liquid sample. The scale bar is 20 μm.
shows the meniscus profiles for ϕ = 0.05 and 0.20, with the empty pillar profiles shown for comparison. The depth of the meniscus profile is small, that is, ∼300 nm (within the uncertainty of the etch depth, see Experimental Section), leading to a negligible change in the actual path length between measurements. It is important to note that the liquid neither engulfs the tops of the pillars nor fills part way up the pillars, ensuring that capillary filling of the cuvette is reproducible between measurements. The meniscus morphology was found to be stable for long periods of time, provided the volume of the sample droplet was in excess of the interstitial volume of the pillar array (typically
