Fabrication of Rugged and Reliable Fluidic Chips for Autonomous Environmental Analyzers Using Combined Thermal and Pressure Bonding of Polymethyl Methacrylate Layers.

The fabrication of highly reliable and rugged fluidic chips designed for use in autonomous analyses for nutrient monitoring is described. The chips are based on a two-layer configuration with the fluidic channels produced in one layer using precision micromilling. The second capping layer contains through holes for sample/standard and reagent addition and waste removal post-analysis. Two optically clear polymethyl methacrylate (PMMA) windows are integrated into the opaque PMMA chip, orthogonal to a 22.5 mm-long section of the channel downstream from a serpentine reagent and sample/standard mixing region. An LED source is coupled into the channel through one of the windows, and the light intensity is monitored with a photodiode located at the distal end of the channel outside the second optically clear window. Efficient coupling of the source through the channel to the detector is achieved using custom-designed alignment units produced using 3D printing. In contrast to fluidic chips produced using solvent adhesion, the thermal-/pressure-bonded simplified method presented removes the need for surface treatment. Optimization of the thermal/pressure conditions leads to very strong adhesion between the PMMA layers, requiring forces in the region of 2000 N to separate them, which is necessary for the use in long-term deployments. Profilometry imaging shows minimal evidence of channel distortion after bonding. Finally, we show the potential of these techniques for environmental applications. The fluidic chips were integrated into prototype nutrient analyzers that display no evidence of leakage in extensive lab tests involving 2500 phosphate measurements using the yellow (vanadomolybdophosphoric acid) method. Similarly, excellent analytical performance (LOD is 0.09 μM) is reported for a 28-day field trial comprising 188 in situ autonomous phosphate measurements (564 measurements) in total including calibration.

Introduction

The ability to monitor the chemical status of environmental waters requires sensing devices that can function in a wide variety of deployment scenarios, ranging from single-use disposable devices for on-the-spot measurements, to much more sophisticated analyzer platforms that provide continuous data from remote locations to databases.[1−4] The ability to reliably fabricate these microfluidic chips plays a critical role in an analyzer’s performance and function. Recent advances in microfluidic platforms have led to the to cost-effective detection of multiple analytes.[5−7] Although these devices provide many advantages, they are limited to either on-site detection or further analyses on return to the laboratory. Producing an inexpensive analyzer platform that can be left in deployment for several months unattended remains a significant challenge. Currently, autonomous analyzers are typically deployed by specialists from environmental agencies, the water industry, and university research teams due to unit costs and the need for considerable background experience and knowledge due to the complexity of these devices. The data generated from autonomous analyzers is increasingly augmented by satellite remote sensing data, opening the way toward much more comprehensive real-time and historical (trending) information that can be visualized and modeled via complex algorithms that feed into predictions of future water status.[8−10] Autonomous environmental analyzer platforms are expensive to buy and to manage in deployments due to the often-hostile conditions these devices experience when in use.[11−13] The high cost of these analyzers arises from the need to produce a device that incorporates reagents, at least one standard for checking calibration, pumps, and valves for flow control, waste storage, nonfouling sampling port, fluidic system, selective detection, electronics for data acquisition and wireless communications, and power source/management.[14]

Cost of ownership is primarily driven by short service intervals, the need for a specialist staff, maintenance/use of expensive facilities and boats/ribs, and replacement of devices/parts/reagents. This used model is obviously very expensive and has contributed to the slow uptake of autonomous sensors for tracking water chemistry.[15,16] The recent global nutrient challenge targeted a unit cost to buy of $5000 or less, compared to the typical $20,000 or more for such devices, as this lower price point would lead to a much larger uptake by state agencies and industries.[17,18] A key challenge is to make every aspect of the device and the analytical method very reliable in use, as failure of any single component renders the device faulty and requires a service visit to repair or replace the device. For the reagent-based colorimetric methods often used in these devices, reagent stability and compatibility within the fluidic system must be balanced with method sensitivity, as more stable reagents can provide for longer service intervals and lower maintenance costs.[19] Likewise, power supply and management strategy must be considered, although major international research efforts driven largely by the e-car industry will advance this technology rapidly.[20] In this paper, we focus on the development and reliability of the fluidic chip, due to its critical role in the autonomous platform operation. The proposed fabrication method utlizes micromilling, providing a fast tunable means for prototyping and testing various fluidic designs in small-scale batches during optimization of the fluidic layout.[21,22] Micromilling enables the design to be produced directly in the polymer layer without the need to reproduce expensive molds if design changes are required between batches.[23]

Polymethyl methacrylate (PMMA) is a useful material for producing prototype fluidic chips due to its broad chemical compatibility, thermal stability, and its availability in opaque and transparent forms.[24] PMMA is compatible with a range of manufacturing techniques, including direct machining or molding. Direct machining techniques suitable for use include laser ablation and micromilling, and applicable molding techniques include hot and cold embossing and thermoforming.[21,25] Micromachining provides the flexibility to rapidly produce and test various modifications to the fluidic design and thereby optimize the proposed method on-chip before moving to a higher volume method such as injection molding for low-cost production of the final chip design. In addition, opaque PMMA reduces the potential impact of stray light on the colorimetric detection approach used in this method.

In this paper, we present a simple method for bonding a two-layer PMMA chip that results in very strong adhesion between the PMMA layers. The method removes the need for surface treatment with no evidence of chip failure/leakage during lab and field trials.[21] These chips improve the reliability of the system fluidics, and in so doing, significantly enhance the overall platform ruggedness and reliability during testing cycles and in real deployments.

Materials and Methods

Analytical Method

Colorimetric detection of phosphate (PO43–) using the yellow method (vanadomolybdophosphoric acid method) was chosen to evaluate the performance of the chip, in view of the excellent reagent stability of 12 months or more,[23] and demonstrated compatibility with microfluidic analyzers.[26] The prototype analyzer used for testing and integration of the fluidic chip housed the method reagent, calibration standards (0 μM/blank and 50 μM PO43–), and sample taken either from the environment during in situ field testing or from a sampling tank during in-lab testing. Flow control was carried out using an in-house-produced fluidic board comprising two 2/2 (hit and hold) and two 3/2 (latching) solenoid valves and two piezo electric pumps to deliver reagent and calibration standard/sample to the fluidic chip in a 1:1 v/v ratio.

All reagents and standards were prepared using ultrahigh purity (UHP) water (MilliQ, Millipore, Burlington, MA, USA) and analytical grade chemicals (Sigma Aldrich, St. Louis, MO, USA), as described previously.[14] Briefly, the yellow reagent was prepared by dissolving 0.351 g of ammonium metavanadate (Sigma Aldrich) in ∼200 mL UHP water with an addition of 95 mL concentrated hydrochloric acid (Sigma Aldrich). Following this, 7.14 g of ammonium molybdate tetrahydrate was added and the resulting solution was made up to 1 L using UHP water. The phosphate standards were prepared by dissolving 0.408 g potassium dihydrogen phosphate (Sigma Aldrich) in UHP water in a 1 L volumetric flask to make a 3 mM stock phosphate solution. This stock solution was further diluted to make up the 50 μM high calibration standard used within the prototype analyzer. The blank solution contained equal volumes of yellow reagent and UHP. All reagent and standard solutions were tested and validated against reference spectra obtained with a Perkin Elmer L900 UV–Vis–NIR spectrometer (Perkin Elmer, Waltham, MA, USA) using a wavelength of 375 nm.

Figure shows the fluidic chip design and channel arrangement. It is a two-layer configuration comprising a lower “channel” layer and an upper “capping” layer. Fabrication of the two fluidic layers was carried out using a series of manufacturing techniques, including laser ablation and precision micromilling as illustrated in Figure . Plates of commercially available opaque 3 mm PMMA were initially cut to the dimensions of the micromill bed (145 mm × 145 mm) using a CO2 laser cutter (Epilog Zing 16). The features comprising the fluidic chip were micromilled (Datron CAT3D-M6 micromill) into the opaque PMMA to create the two-layer prototype design. This comprised ∼1 mm depth/width channels milled in the channel layer and three through holes (⌀ 6.5 mm) in the capping layer. The three through holes were configured to receive 6.5 mm threaded barbed liquid connectors used for connection of tubing (1/16th in. internal diameter Tygon) between the fluidic chip and the fluid board of the prototype analyzer. Two holes were drilled into both layers of the chip to ensure precise alignment during bonding of the chip. During use, the reagent and sample/calibrant enter via the inputs (7, Figure ), merge (1, Figure ), and mix in the serpentine channel (2, Figure ). The color developed via the reagent and analyte reaction is measured by means of a LED and photodiode arranged at opposite ends of a 22.5 mm optical detector channel (4, Figure ). Light is coupled into and out of the detector via transparent PMMA windows (5, Figure ). From the detection channel, the mixture is conveyed to the waste outlet (8, Figure ) via the outlet channel (3, Figure ) for external storage.

Figure 1

Two-layer design showing (A) the bottom channel layer containing milled fluidic channels incorporating (1) a merging point for sample and reagent inputs, (2) a serpentine mixing region for mixing of the sample and reagent, (3) an outlet line for flushing of waste after analysis, (4) an optical channel for detection on the chip, and (5) two optically clear windows for coupling light into and out of the detection channel. (B) Upper capping layer incorporating (6) two holes to house alignment pins for the two layers during bonding and three 6.5 mm circular through holes: (7) two inlets for the sample and reagent intake and (8) one outlet for the waste.

Figure 2

Manufacturing scheme for prototyping of two-layer PMMA fluidic chips. (1) Micromilling is used to produce a batch of two layer fluidic chips (×3) from a 145 mm × 145 mm laser cut PMMA milled plate; (2) cleaned capping, channel layer, and optically clear windows are aligned to the fluidic chip design; (3) bonding is carried out using heat and pressure with the aligned fluidic chip positioned between two glass compression plates and a brass weight on top; and (4) barbed liquid connectors are integrated and the fluidic chips are leak tested prior to use.

Bonding Procedure

A simple bonding method employing an external weight in combination with thermal and vacuum conditions was used to effectively seal the two-layer fluidic chip. Prior to bonding, the two layers were cleaned using an ultrasonic bath to remove debris and residues left over from the micromilling and swarf build-up within the channels or through holes of the chip. Thermal/vacuum bonding of the two-layer PMMA chips was performed under optimized conditions (see Table S1, Supporting Information) using a vacuum oven (Gallenkamp vacuum oven) set at the glass transition temperature of this particular PMMA (∼165 °C)[27] and a vacuum of 200 mbar. A 220 g brass weight (A) was placed on top of two glass compression plates (B) with the two aligned layers of the PMMA fluidic chip (C) positioned between these plates (see Figure S1, Supporting Information). The optical windows clamped in place and the adhesion of the two layers and optical windows occurs under these conditions for a period of 20 min.

PMMA Fluidic Chip Characterization

Characterization was carried out to assess the reproducibility of the fabricated fluidic chips using a series of techniques including optical profilometry, stylus profilometry, and optical microscopy. Inspection and visualization of micromilled channels was carried out using a Keyence 3D microscope equipped with a composite 50–500× lens. Depths and profile measurements were carried out using complimentary profilometers, a Bruker DektakXT stylus profilometer and a Bruker ContourGT optical profilometer. The stylus profilometry was carried out at a range of 65.5 μm, tracking surface features with a 2.5 μm stylus tip. Both approaches were used to create surface profiles of both fluidic layers at the same targeted areas, covering the merging point, serpentine mixing region, waste outlet, and optical channel (see Figure ). These measurements were performed before and after thermal bonding to assess the accuracy and precision of the fabrication techniques, and the extent of any channel deformation associated with the thermal bonding procedure.

Figure 3

Channel layout and target areas used for characterization of the channels in the channel layer of the fluidic chip. (1) Sample and the reagent inlet where the sample and reagent merge after intake onto the chip, (2) initial portion of the serpentine mixing region where the sample and reagent mix on chip prior to measurement, (3) waste outlet channel through which the sample and reagent are flushed after analysis, and (4) optical channel for colorimetric detection of the sample/reagent mixture. These regions are discussed further (see Figure ). The red line in the waste outlet channel (3) is the location used to obtain the section shown in Figure .

Figure 7

Optical profile and microscope images of the channel layer before and after bonding of the (1) merging point, (2) mixing channel, (3) waste outlet, and (4) optical channel (see Figure )

Figure 6

Section across the lower layer channel (blue) and the equivalent section of the upper layer imprint (red) at the location marked in red in Figure . Data obtained using the Bruker ContourGT optical profilometer.

Chip Leak Testing

In order to optimize bonding of the chip layers, temperature, vacuum, and time were varied systematically (see Table S1, Supporting Information). The so-produced chips were tested for leakage following bonding and cooling prior to integration into the prototype analyzer. A 60 mL syringe was used to pump UHP water through each chip for 1 h, refilling the syringe as needed. Flow rates of 3.0, 6.0, and 12.0 mL/min were applied while the chips were visually checked for any signs of leakage throughout testing. Particular attention was paid to potential failure points including seams, optical windows, and fluidic connectors. In order to minimize potential distortion of the fluidic channels, the selected conditions were the least extreme that exhibited no evidence of leakage (temperature = 165 °C, oven pressure = 200 mbar, time = 20 min). After this preliminary testing, the fluidic chips were integrated into a prototype water nutrient analyzer for further lab assessment and testing, during which the reagent and sample were delivered to the fluidic chips at a flow rate of 3.0 mL/min for 1 h, mimicking the operational flow rate of the analyzer in field deployments. The analyzer was then setup to run in fully autonomous mode for further in-lab and in situ field testing (see Section ).

Interlayer Tensile Strength Measurements

Quantitative measurements of fluidic chip interlayer bond strength were carried out using a Zwick tensile strength test system. Figure S2 shows the experimental setup designed to assess the strength of the chip layer bonding wherein each face of the thermally bonded chip was fixed to an aluminum plate using a two-part epoxy adhesive (araldite rapid). The aluminum plates were attached to the instrument, and the force and the stress applied to attempt to separate the fluidic layers was measured.

Results and Discussion

Chip Interlayer Bond Strength Analysis

Three methods were carried out to assess the fluidic chip bond strength (see Figure and Table ).

Figure 4

Bond strength analysis carried for methods. (1) Fluidic chip adhered directly to metal plates using epoxy, failure was seen at the epoxy interface; (2) fluidic chip with roughened surface adhered to metal plates using epoxy, failure was seen at the epoxy interface; and (3) fluidic chip with roughened surface adhered to the metal plate with roughened and scored surface using epoxy, failure occurred at the bond between the fluidic chip layers.

Table 1

Summary of Bond Strength Testing Carried out with Surface Modifications and Method Optimization Carried out for Methods 1–3

method	mean Fmax (N)	%RSD	failure point
1: fluidic chip adhered to the metal plate using epoxy	237.29	24.92	epoxy interface between the fluidic chip and plate
2: fluidic chip adhered to the metal plate using epoxy with the surface of the fluidic chip roughened	1485.66	1.92	epoxy interface between the fluidic chip and plate
3: fluidic chip adhered to the metal plate using epoxy with the surface of the fluidic chip roughened and the metal plate scored	1977.68	7.06	bond between the two layers of the bonded fluidic chip

Method 1

The fluidic chips were directly adhered to the aluminum base plates using a two-part epoxy. During the testing of this method, forces in excess of 300 N were applied. The fluidic chips failed at the interface between the epoxy adhesive and the fluidic chip, not at the interface between the chip layers. A number of epoxies (Araldite Instant, Standard and Rapid) were tested using the above method, all giving similar results.

Method 2

Modification of the fluidic chip surface was carried out by manually roughening the surface in contact with the epoxy using a 1000 grit sandpaper, to further improve adhesion of the epoxy adhesive between the chip and the metal support. Bond strength testing was again carried out using greater forces in excess of 1500 N. Once again, failure occurred at the epoxy interface between the chip and the metal plate, not at the interface between the chip layers.

Method 3

In addition to the chip surface roughening as per method 2, the surface of the metal plate was scored and roughened manually using a hand file to further increase the strength of the bond between the epoxy, the metal plate, and the fluidic chip. The tests were repeated for a further set of chips, and under these conditions, the interlayer bond in the fluidic chips failed at forces in the range 1878.88–2076.47 N, indicating the high degree of interlayer adhesion generated by the simple bonding process.

Post-Bonding Fluidic Chip Channel Characterization

The imprint of the lower fluidic layer channel layout appearing in the upper layer shown in Figure is an effect arising from the temperature and vacuum conditions combined with the external weight used during the bonding process. Sections across the channel and equivalent imprint locations on the capping and channel layers (Figure ) show that the capping layer protrudes slightly into the channel (∼3.0 μm), forming a very tight seal at the channel edges in the process.

Figure 5

Images of chips after layer separation. (A) Photo of the bottom fluidic layer showing channels; close up photos of the (B) bottom layer and (C) top layer showing the channel layout imprinted on the top layer; the Keyence microscope images the lower layer showing a section of (D) the serpentine mixing channel and (E) its imprint on the same region of the top layer.

Figure shows characterization carried out on the target areas (see Figure ) of the fluidic chip’s channel layer using optical profilometry (Bruker ContourGT) and microscopic imaging (Keyence 3D microscope). Optical profilometry was used to generate 3D maps of the surfaces from which channel widths and depths could be measured before and after bonding (see Figures S3 and S4, Supporting Information).

Table summarizes typical measurements of average depth, average width, depth standard deviation, and surface roughness (Sz). These techniques were used to assess if residues or swarf build-up from the milling process could be seen and to determine the extent of any channel deformation caused by the layer bonding process. Table before bonding shows low standard deviation for channel depth at each of the target areas measured and low channel roughness. These results show that micromilling is a suitable method of fabrication and can produce accurate and precise channels in the PMMA chips at the required dimensions with acceptable tolerances. Comparisons before and after bonding show a small increase in channel width and decrease in channel depth due to the impact of the weight and vacuum used during the bonding method to ensure strong adhesion between the two PMMA layers.

Table 2

Optical Profilometry Measurements of Fluidic Chip at Target Areas (Figure ) (A) before Combined Thermal and Pressure Bonding and (B) after Combined Thermal and Pressure Bonding

	merging point (mm)		mixing channel (mm)		waste outlet (mm)		optical channel (mm)		total chip profile (channel layer) (mm)		total chip profile (capping layer) (mm)
parameter	(A)	(B)	(A)	(B)	(A)	(B)	(A)	(B)	(A)	(B)	(A)	(B)
channel depth average	1.244	1.130	1.157	1.149	1.207	1.137	1.187	1.089	N/A	N/A	N/A	N/A
channel depth standard deviation (N = 3)	4 × 10–4	5.8 × 10–4	4 × 10–4	2.5 × 10–4	3.1 × 10–3	3.8 × 10–3	3.1 × 10–3	1.3 × 10–3	N/A	N/A	N/A	N/A
surface roughness(Sz)	0.012	0.0239	0.016	0.0351	5.1 × 10–3	5.8 × 10–3	5.5 × 10–3	0.0167	0.549	0.110	0.061	0.248
channel width	1.045	1.105	N/A	N/A	1.045	1.087	1.045	1.109	N/A	N/A	N/A	N/A

Figure also shows that after thermal bonding, the channels remained clear with no blockages or damage occurring and the mill marks remain present on the surface of the channels both before and after bonding (compare for example the microscope images of the waste channel before and after bonding). The inclusions at the interface between the layer after bonding are most likely due to small amounts of air trapped during the bonding process in this interfacial region. However, these do not give rise to any leakages, apparently due to the formation of a tight seal between the channel and the capping layer (Figure ).

Fluidic Chip Characterization: Leak Testing

Leak testing showed no evidence of any leakages or failure points from 10 replicate chips assessed at flow rates of 3.0, 6.0, and 12.0 mL/min over a period of 1 h. When integrated into the prototype analyzer, no evidence of failure or leakage occurred over 1 h of testing using a flow rate of 3.0 mL/min. The chips were further tested with the analyzer running in a fully autonomous mode during which 2500 in-lab phosphate measurements were taken using the yellow method without chip failure (see ref (14)).

Performance of Fluidic Chip in the Prototype Nutrient Analyzer

In-Laboratory Testing

In-lab testing was carried out on the integrated fluidic chip on the prototype analyzer using a series of phosphate calibration standards (0–25 μM PO43–) that were autonomously loaded onto the chip using the fluidic handling of the analyzer. Standards of known concentration were measured optically on the chip using the integrated 375 nm LED coupled with a photodiode-based detector, and the results compared against a UV–vis spectrophotometer employing the same method.

The fluidic chip had two inlets, the first taking in the phosphate standard and the second taking in the yellow reagent at a 1:1 v/v ratio of the sample to the reagent. As the two solutions were pumped through the chip, mixing occurred within the serpentine mixing region, see Figure . Once the chip had been fully filled and the reaction mixture was present in the optical channel, pumping was paused for 20 min to allow the reaction to come to completion at room temperature.[28] Optical measurements were then taken using the LED and photodiode. Figure and Table present the linear calibrations generated and the analytical performance summaries for the phosphate sample detection on chip versus the UV–vis spectrometer.

Figure 8

Optical detection for phosphate (PO43–) using the yellow method comparing the prototype analyzer incorporating the fluidic chip and the UV–vis spectrometer (375 nm).

Table 3

Analytical Performance of the Prototype Analyzer versus UV–Vis Spectrometer (N = 10 per Measurement)

	UV–vis spectrometer		prototype analyzer
PO43– (μM)	mean (μM) ± %RSD	%RE	mean (μM) ± %RSD	%RE
1	1.07 ± 3.93	6.65	0.99 ± 4.01	–0.70
2	2.00 ± 3.34	0.01	2.06 ± 2.03	2.76
4	3.91 ± 1.60	–2.33	3.93 ± 1.04	–1.74
6	6.23 ± 2.04	3.85	5.89 ± 0.77	–1.82
8	8.07 ± 1.50	0.84	7.84 ± 0.59	–1.94
10	10.05 ± 1.00	0.52	9.93 ± 0.52	–0.65
15	15.05 ± 1.18	0.35	15.08 ± 0.05	0.50
20	20.01 ± 0.61	0.07	20.11 ± 0.24	0.57
25	25.02 ± 0.45	0.07	24.86 ± 0.21	–0.56
mean (N = 90)	1.74	1.63	1.05	1.25

Both methods have excellent linearity (0.9999 UV–vis reference method, 0.9999 prototype analyzer) as seen in Figure . The UV–vis reference method had a mean %RSD of 1.74 and a mean %RE of 1.63 (N = 90), compared to the prototype analyzer, which had a slightly better performance with a mean %RSD of 1.05 and mean %RE of 1.25 (N = 90). The results summarized in Table and shown in Figure also show an increased sensitivity for the analyzer (y = 0.043x) in comparison to the UV–vis spectrometer ( = 0.026x). This can be accounted for by the longer 22.5 mm detector pathlength of the fluidic chip in comparison to the 10 mm cuvette used in the UV–vis spectrometer.

In Situ Field Testing

A further field study was arranged to test the prototype analyzer, incorporating the fluidic chip, under real deployment conditions. The analyzer was deployed at the River Liffey, Co. Dublin (Figure ), and its field performance assessed over a period of 28 days (21/02/2018 to 20/03/2018). During this time the analyzer functioned in a completely autonomous manner, obtaining and analyzing samples, as well as performing 2-point calibrations between each sample measurement.

Figure 9

(A) Satellite image of the sampling location at Palmerstown, Co. Dublin, Ireland. (B) Picture of the prototype sensing platform deployment at Liffey Deployment location, Co. Dublin, Ireland (coordinates 53.356246, −6.360325999999986).

Throughout the period of the deployment, samples were measured every 3 h, comprising a total of 564 measurements, including the environmental sample and internal sensor calibration measurements. Occasional grab samples were taken to validate the analyzer performance via the UV–vis lab-based reference method. The internal temperature of the analyzer was also monitored, along with the external environmental temperature and water level throughout the deployment, see Figure .

Figure 10

In situ environmental monitoring. (A) Water Level (cm) from 21/02/2018 to 20/03/2018. (B) External and Internal Temperatures (°C) from 21/02/2018 to 11/03/2018.

The River Liffey is a dam controlled with regular releases upstream of the deployment site being carried out by the Electricity Supply Board at Golden Falls and Leixlip. Figure A shows the environmental monitoring of the river level using a depth gauge deployed alongside the analyzer. During the 28-day deployment river levels ranged between 63 and 187 cm in depth. Dam releases can be seen during the period 25/02/19 to 02/03/18. Heavy snow fall occurred from 24/02/19, with a subsequent melt beginning on 04/03/18. As a result, a rapid rise in the water level was recorded (70 to 170 cm) before reaching a maximum level of 187 cm on 11/03/18. The average water level was 115 cm during the period of the deployment.

The prototype analyzer was insulated to minimize heat loss and to prevent internal freezing within the platform. Figure B shows the elevated temperature internally versus external environmental temperatures measured due to the insulation of the analyzer and small amounts of heat generated internally by the electronics. The internal temperature dropped to a low of 5.1 °C in comparison to a low of −4.8 °C being recorded externally (night of 28/02/18). Due to this insulation, the internal temperature changes are also dampened compared to the more rapid external fluctuations (e.g., see the changes from 05/03/18). From 27/03/18 to 03/03/18, a major subzero temperature event occurred with the external temperature dropping to −5.0 °C and thereafter rising rapidly to 11.7 °C. A series of much shorter subzero sharp temperature drops can be seen at 24/02/18, 07/03/18, and 09/03/18.

Figure presents the results of the 28-day deployment to track variations in phosphate concentration during the deployment (grey squares), along with lab validation measurements (red circles) and water level (blue line). A total of 188 environmental phosphate measurements were recorded over the 28-day deployment (i.e., 564 measurements in total including two-point calibration), using the fluidic chip in the analyzer. During the deployment, the phosphate water concentrations together with the calibration data were transmitted to a cloud database for remote access.

Figure 11

Prototype analyzer with integrated fluidic chip, phosphate (PO43–) detection at the deployment site over a 28-day period.

There appears to be a general correlation between the phosphate concentration and the water level, which could in turn be linked to wash out of nutrients from fields into the river. The effect of run-off caused by the snow melt into the river from the surrounding area can be seen through the increased water levels from 05/03/18, which is coincident with the gradual increase in phosphate concentration. A number of grab sample values correlate reasonably well with measured levels recorded from the in situ analyzer. The maximum detected phosphate (PO43–) level during this deployment was 26.26 μM with an average of 13.0 μM detected over the 28-day period.

During the deployment, the analyzer did not generate valid sample measurements during the period 01/03/2018 to 03/03/2018, highlighted in red in Figure . This was due to a significant freezing event during which the external sampling line froze at the intake point, preventing the prototype analyzer from acquiring samples. However, despite this, the internal functions were unaffected (reagents, pumps, electronics, power), and the platform continued to perform valid calibrations and transmit data. After 03/03/2018, the external sampling line thawed and recommenced, acquiring samples as before, with no servicing or external intervention needed.

Conclusions

A cost-effective simplified reproducible methodology for combined thermal and vacuum bonding has been realized for the fabrication of fluidic chips for environmental applications. These manufactured fluidic chips were characterized and performance-assessed through extensive laboratory testing and autonomous in situ field trials in real-world senarios. Careful optimization of temperature and pressure during bonding of the two PMMA layers and optical window layers resulted in a very strong adhesion between the layers with little or no evidence of distortion to the fluidic channel or optical window with no observed leakage or failures between batches. Once integrated into the prototype analyzer, the analyzer produced accurate and reproducible optical measurements for the detection of nutrients using colorimetric chemistries. The fluidic chip design allows for sample and reagent intake, mixing, and detection to be carried out on-chip. In field trials the protoype analyzer incorporating one of these fluidic chips functioned successfully over a 28-day period during which time it was exposed to subzero fluctuating temperatures. The ruggedness and reliability of these fluidic chips produced by the method reported have enhanced the functionality of the prototype analyzer and in so doing, has significantly improved the overall analyzers analytical performance, reliability, and ruggedness in long-term field deployments. The simplified bonding method has the potential to be transferred to varying fluidic designs ulilised by research group and industries where the application utilizes higher pressures.

The manuscript was written through contributions of all authors.