A Microfluidic pH Measurement Device with a

2 downloads 0 Views 2MB Size Report
Jul 4, 2017 - Abstract: The pH values of aqueous solutions are conventionally measured .... not affected by the flow rate under the measurement condition.

sensors Article

A Microfluidic pH Measurement Device with a Flowing Liquid Junction Akira Yamada * and Miho Suzuki Department of Mechanical Engineering, Graduate School of Engineering, Aichi Institute of Technology, Toyota 470-0392, Japan; [email protected] * Correspondence: [email protected]; Tel.: +81-565-48-8121 (ext. 2326) Received: 25 April 2017; Accepted: 2 July 2017; Published: 4 July 2017

Abstract: The pH values of aqueous solutions are conventionally measured with pH-sensitive electrodes such as glass electrodes or ion-sensitive field-effect transistors (ISFETs) used in conjunction with Ag/AgCl reference electrodes and KCl solutions. The speed of pH measurement with these systems can be deficient, however, as the glass electrode responds slowly during measurements of sample solutions with low buffering capacities. Our group has constructed a new pH measurement system using a microfluidic device and ISFET sensors. The device has a channel with two inlets and one outlet, with a junction connected to a Y-shaped channel on the same plane. Two ISFET sensors and an Ag/AgCl pseudo reference electrode are fitted into the channel to construct a differential measurement device. A sample solution and baseline solution supplied into the inlets by gravity-driven pumps form a flowing liquid junction during measurement. The small size and fast response of the ISFET sensors enable measurement of about 2.0 mL of sample solution over a measurement period of 120 s. The 90% response time is within 2 s. The calibrated sensor signal exhibits a wide range (pH 1.68–10.0) of linearity with a correlation factor of 0.9997. The measurement error for all solutions tested, including diluted solutions, was 0.0343 ± 0.0974 pH (average error ± standard deviation (S.D.), n = 42). The new device developed in this research will serve as an innovative technology in the field of potentiometry. Keywords: pH; ISFET; pH-FET; microfluidic device; liquid junction; flowing junction

1. Introduction The pH scale is a numeric index used to specify the acidity and alkalinity of aqueous solutions. One of the domains in which it is frequently applied is environmental science. Precise pH values can be obtained by taking potentiometric measurements with devices such as glass electrodes [1] or ion-sensitive field-effect transistors (ISFETs) [2,3] used in conjunction with reference electrodes [4]. The most commonly applied pH electrode is a glass type used together with an Ag/AgCl reference electrode and KCl solution [4]. Though traditional and reliable, the glass electrode method still has shortcomings. The long response time of the glass electrode precludes a shortened measurement time and leads to large measurement error and low reproducibility [4–6]. These shortcomings become especially problematic in the measurement of sample solutions with low buffering capacities [4–6]. Environmental waters, such as natural water and water released from industrial plants, often have low buffering capacities. Precise measurements of their pH values using glass electrodes may therefore require long periods to complete. The ISFET sensors developed with semiconductor technology have many advantages over glass electrodes [2,3]. Glass electrodes and ISFETs differ in both size and responsiveness. The former are difficult to downsize without sacrificing the electrical characteristics. ISFET sensors, in turn, have smaller electrodes and perform with superior responsiveness.

Sensors 2017, 17, 1563; doi:10.3390/s17071563

www.mdpi.com/journal/sensors

Sensors 2017, 17, 1563

2 of 13

The liquid junction (LJ) is essential in potentiometric measurements. The KCl solution used in the reference electrode keeps the electromotive force of the Ag/AgCl electrode constant [4,7]. LJ devices based on several principles have been developed, including a ceramic type, double junction type, free diffusion junction type, and flowing junction type [4,6,8–10]. The first three types mentioned can be used in combination with electrodes [6,8,11]. While the fourth type, the flowing LJ, obtains stable measurement conditions due to the flow of the solution, the outer body is too large to fit into a practical measurement apparatus [4]. Our group has developed a flow-through-type differential sensor probe consisting of two ISFET sensors [12,13]. The sensor probe is integrated into an automated pH measurement system (Auto-pH) consisting of XYZ stages, a pump, electric circuit, and PC [14]. In trial measurements, the system performed highly accurate, high-speed measurements of 96 small-volume samples within a short time [14]. Special adjustments were necessary in some cases, however, as solutions with low buffering capacities required spatiotemporal control of the LJ to manage the microfluidic behavior in order to shorten the response time and improve the measurement accuracy [15]. Microfluidic technologies are used in applications in chemistry, biochemistry, molecular biology, and medicine [16,17]. A micro channel of at least one dimension smaller than a few mm has an extremely low Reynolds number (Re), which results in a laminar fluid flow without turbulence. If two flowing fluids meet in a Y-shaped channel with two inlets, an unmixed sharp boundary formed between the fluids acts as an LJ, especially under a flowing condition. Microfluidic devices for pH measurement have been reported in several papers. On the environmental front, a self-calibrating device fitted with an ISFET has been used to measure pH in situ at ocean depths [18]. In a study from Europe, a colorimetric microfluidic pH sensor was developed for autonomous seawater measurement [19]. In the field of biology, local pH values were measured in a microfluidic device using a resin gel-microbead material with a pH indicator [20]. In the present investigation we constructed and evaluated a new pH measurement system using a microfluidic device and ISFET sensors. A flowing LJ within the microfluidic device enabled stable measurements. The present device exhibited a rapid response in pH measurements of sample solutions with low buffering capacities. These features enabled superior measurement accuracy within a pH range of 1.68–10.0. The present pH measurement system can be applied in the environmental sciences. 2. Materials and Methods 2.1. Measurement Principle Figure 1 shows a schematic representation of the microfluidic device for pH measurement. The device consists of a channel with two-inlets and one outlet. Two pH-sensitive field-effect transistor (pH-FET) sensors and an Ag/AgCl electrode are installed into the channel. The pH-FET sensor for measurement (pH-FETm ) is set downstream of the junction, and the pH-FET for reference (pH-FETr ) and Ag/AgCl electrode are set near the inlet for the baseline solution (BLS). The measurements are performed under a flowing condition. Figure 2 indicates the measurement principle using a differential formula. The present system obtains the source potential for the measurement (Vm ) and reference (Vr ) pH-FETs with respect to the Ag/AgCl electrode under the two flow path conditions shown in Figure 2a,b. Figure 3 shows the time course of one measurement cycle. First, the two pH-FET sensors are immersed in the BLS and a differential sensor signal (∆V(0) = Vm – Vr ) is recorded in the final second of the washing period (Figure 2a). Next, the flow paths of the two solutions are switched to immerse the pH-FETm in the sample solution, and a differential sensor signal (∆V(1) = Vm – Vr ) is recorded in the final second of the measurement period (Figure 2b). The pH-FETm and Ag/AgCl electrode are electrically connected through an LJ of the two solutions. A double differential source potential is defined as ∆∆V = ∆V(1) – ∆V(0), and the pH values correspond to the ∆∆V values. A differential measurement system is thus constructed.

Sensors 2017, 2017, 17, 1563 Sensors 17, 1563 Sensors 2017, 17, 1563

Sensors 2017, 17, 1563

3 of 133 of 13 3 of 13

3 of 13

Figure 1. schematic A schematic representation of the microfluidic measurement device Figure 1. A representation of the pH pH measurement device mademade of of Figure 1. A schematic representation of microfluidic the microfluidic pH measurement device made of polydimethylsiloxane (PDMS) and a glass slide. 1: pH measurement device body; 2: Glass slide; polydimethylsiloxane (PDMS) and a glass slide. 1: pH measurement device body; 2: Glass slide; polydimethylsiloxane (PDMS) and a glass slide. 1: pH measurement device body; 2: Glass slide; 3: Channel; Measuring pH-FET (pH-FET );microfluidic 5: Reference pH-FET (pH-FET r) pseudo-reference and made pseudo-reference Figure 1. Measuring A4:schematic representation of pH measurement device of 3: Channel; 4: pH-FET (pH-FET 5:mmReference pH-FET (pH-FET m );the r ) and 3: Channel; 4: Measuring pH-FET (pH-FET ); 5: Reference pH-FET (pH-FET r) and pseudo-reference polydimethylsiloxane (PDMS) and a glass slide. 1: pH measurement device body; 2: Glass slide;(mm): electrode; 6: wires; Lead 7:wires; 7: solution Sampleinlet; solution inlet; solution 8: Baseline solution (BLS) inlet; 9: Outlet. electrode; 6: Lead Sample 8: Baseline (BLS) inlet; 9: Outlet. Sizes electrode; 6: Lead wires; 7: Sample solution inlet; 8: Baseline solution (BLS) inlet; 9: Outlet. 3: Channel; 4:aMeasuring pH-FET (pH-FET 5: Reference Sizes (mm): (1.0); b (0.5); c (17); d (12);me); (32); f (6–8). pH-FET (pH-FETr) and pseudo-reference a (1.0); b (0.5); c (17); d (12); e (32); f (6–8). Sizes (mm): a (1.0); b (0.5); c (17); d (12); e (32); f (6–8). electrode; 6: Lead wires; 7: Sample solution inlet; 8: Baseline solution (BLS) inlet; 9: Outlet. Sizes (mm): a (1.0); b (0.5); c (17); d (12); e (32); f (6–8).

Figure 2. Measurement principle usingusing a Y-shaped channel and pH-FET sensors. The flowing paths ofpaths Figure 2.Measurement Measurement principle a Y-shaped channel and pH-FET Thepaths flowing Figure 2. principle using a Y-shaped channel and pH-FET sensors.sensors. The flowing Figure 2. Measurement principle using a Y-shaped channel and pH-FET sensors. The flowing the two solutions at (a) theatinitialization stage before measurement and (b) during Thepaths of twosolutions solutions initialization measurement (b)measurement. during measurement. of the the two at (a)(a) thethe initialization stagestage beforebefore measurement and (b) and during measurement. of the two solutions at (a) the initialization stage before measurement and (b) during measurement. hatched shows the areathe of area contact between the two a region thatthat serves asas a liquid The region hatched of contact between the solutions, two a region serves a The hatchedregion regionshows shows the area of contact between thesolutions, two solutions, a region that serves as a Theduring hatched region shows the area of contact between the two solutions, a region that serves as a junction measurement. liquid junction measurement. liquid junctionduring during measurement. liquid junction during measurement.

Figure 3. Time course of oneofsample-measurement cycle: t1 , washing the pH-FET andand channel Figure 3. Time course one sample-measurement cycle: T1, washing the pH-FET m sensor m sensor inside; t2 , switching reservoir , measurement. 2, switching theheights; reservoirt3heights; t3, measurement. channel inside; Tthe Figure 3. Time course of one sample-measurement cycle: T1, washing the pH-FETm sensor and Figure 3. Time course of one sample-measurement cycle: T1, washing the pH-FETm sensor and channel inside; T2, switching the reservoir heights; t3, measurement. channel inside; T2, switching the reservoir heights; t3, measurement.

Sensors 2017, 17, 1563

4 of 13

The flow velocity (v (cm/s)) and Re were calculated as follows: v = Q/wh and Re = wv/n, where The velocity (cm/s)) and Re were as Q follows: v = Q/wh and(ml/s), Re = wv/n, w and h flow are the width(vand height (cm) of thecalculated flow path, is the volume flow and n where is the w and h are the width and height (cm) of the flow path, Q is the volume flow (ml/s), and n is the kinetic 2 kinetic viscosity (cm /s) at 25 °C. The v of the flowing fluid and Re during measurement were 2 ◦ C. The v of the flowing fluid and Re during measurement were estimated viscosity at 25cm/s estimated(cm to be/s) 0.067 and 0.75, respectively. The source potential of the two pH-FETs, Vm and tor,be 0.067 and by 0.75, The source potential ofcondition. the two pH-FETs, andflow Vr , were V were notcm/s affected therespectively. flow rate under the measurement While theVm total rates not affected by the flow rate under the measurement condition. While the total flow rates of the of the sample solution and BLS remained almost constant during measurement, the flow velocity sample the solution andr sensor BLS remained almostelectrode constantdiffered during between measurement, the flow velocity around around pH-FET and Ag/AgCl the washing and measurement the pH-FET sensor and Ag/AgCl electrode differed between the washing and measurement periods r periods (Figures 2 and 3). (Figures 2 and 3). 2.2. Fabrication of the Microfluidic pH Measurement Device 2.2. Fabrication of the Microfluidic pH Measurement Device The microfluidic device for pH measurement was composed of a concave Y-shaped channel microfluidic device for pH(PDMS) measurement was composed a concave channel using made madeThe from polydimethylsiloxane and a glass slide. The of device bodyY-shaped was constructed from lithography polydimethylsiloxane a glass The device body was by constructed using soft soft [16]. A (PDMS) convex and mold withslide. a bank was designed three-dimensional lithography [16]. A convex mold with a bank was designed by three-dimensional computer-aided computer-aided design (3D-CAD) and fabricated using a rapid prototyping system (Objet 500, design (3D-CAD) and fabricated using(Figure a rapid4a). prototyping system (Objet 500,was Connex, Connex, Stratasys Ltd., Rehovot, Israel) A transparent liquid PDMS pouredStratasys into the Ltd., Rehovot, (Figure 4a). A transparent liquid was poured into hardened mold, hardenedIsrael) overnight, and removed from the mold.PDMS As shown in Figure 4b,the themold, channel side of overnight, removed theslide mold.were As shown in Figure by 4b, plasma the channel side of the PDMS structure the PDMS and structure andfrom glass hydrophilized irradiation (PDC-32G, Harric and glass slide were hydrophilized by plasma irradiation (PDC-32G, Harric Plasma, NY, USA) and Plasma, NY, USA) and adhered together. The holes for the inlets and outlet of the device body were adhered together. The holes for the inlets and outlet of the device body were punched by a hole punched by a hole puncher (ϕ = 3 mm). The pH-FET sensors [21,22] and Ag/AgCl electrode were puncher (ϕ = the 3 mm). The through pH-FET the sensors and gap Ag/AgCl electrode were inserted intowas the inserted into channel tiny [21,22] holes. The between each hole and lead wire channel through the tiny holes.toThe gap between eachfrom hole leaking. and leadEach wirepH-FET was filled with measured a silicone filled with a silicone adhesive prevent the solution sensor adhesive to prevent the solution from leaking. Each pH-FET sensor measured 5.5 × 0.45 × 0.2 mm. 5.5 × 0.45 × 0.2 mm.

Figure 4. (a) A mold fabricated using a rapid prototyping system. External form: 36 × 63 × 12 mm. Figure 4. (a) A mold fabricated using a rapid prototyping system. External form: 36 × 63 × 12 mm. (b) Overview of a microfluidic device with pH-FET sensors. External form of the main body: (b) Overview of a microfluidic device with pH-FET sensors. External form of the main body: 32 × 76 × 7 mm. 32 × 76 × 7 mm.

2.3. Measurement System 2.3. Measurement System Figure 5 shows schematics of the measurement system used in this work. The system was Figureof5 ashows schematics of the in for thispH-FET work. The system was composed microfluidic device, twomeasurement reservoirs, ansystem electricused circuit sensors, a 16-bit composed of a microfluidic device, two reservoirs, an electric circuit for pH-FET sensors, a 16-bit analog-to-digital converter (CONTEC, Osaka, Japan), and a PC (DOS/V). In our trial measurements, analog-to-digital converter Osaka, Japan), and a PC (DOS/V). In our trial measurements, the source potentials of the(CONTEC, pH-FETs were taken at 1.0 s intervals. The reservoir was connected to the source potentials of the pH-FETs were taken at 1.0 s intervals. The reservoir was connected the the inlet of the microfluidic device by a tube. The solutions were fed from the reservoirs to the to inlets inlethydrostatic of the microfluidic byby a tube. The and solutions wererates fed from reservoirs the inletsthe by by pressuredevice caused gravity, the flow weretheadjusted bytochanging hydrostatic pressure caused by gravity, and the flow rates were adjusted by changing the heights of the heights of the reservoirs. Just before measurement, the reservoirs were set above the measurement reservoirs. Just before were with set above the solution measurement device at heights device at heights of h1measurement, = 25 and h2 =the 55 reservoirs mm and filled sample and BLS, respectively of h1 = 253 and 55 mm and filled with sample solution BLS, respectively (Figures and 5). 2 = (Figures and h 5). Each reservoir consisted of a narrow (ϕ = and 4 mm) pipe kept parallel to the3floor to Each reservoir consisted of a narrow (ϕ = 4 mm) pipe kept parallel to the floor to keep the hydrostatic keep the hydrostatic pressure constant during measurement. We derived values for the control pressure constant during valuesand for the control parameters to shorten the parameters to shorten themeasurement. time required We for derived measurement improve the measurement accuracy. timepresent requireddevice for measurement andmL improve the measurement accuracy. The present device consumed The consumed 2.0 of sample solution per measurement and completed the cycle in 120 s: 59 s for washing, 1 s for switching the solutions, and 60 s for measurement (Figure 3).

Sensors 2017, 17, 1563

5 of 13

2.0 mL of sample solution per measurement and completed the cycle in 120 s: 59 s for washing, 1 s for and 60 s for measurement (Figure 3). 5 of 13

switching solutions, Sensors 2017, the 17, 1563

Figure 5. Schematics Schematics of ofthe themeasurement measurementsystem. system.Reservoir Reservoir 1, sample solution; Reservoir 2, BLS. Figure 5. 1, sample solution; Reservoir 2, BLS. The 1 and h 2 are the heights of the reservoirs. The h h1 and h2 are the heights of the reservoirs.

2.4. 2.4. Measurements, Measurements, Calibration, Calibration, and and Evaluation Evaluation of of Linearity Linearity The The procedures procedures for for calibration calibration and and measurement measurement were were the the same same (Figure (Figure 3). 3). To To start start the the cycle, cycle, 2.0 mL of sample solution and 2.0 mL of BLS were poured into the respective reservoirs. Pinch 2.0 mL of sample solution and 2.0 mL of BLS were poured into the respective reservoirs. Pinch cocks cocks wereopened then opened to commence the solution flows. V was three for standard each pH were then to commence the solution flows. ∆∆V was read outread threeout times fortimes each pH standard solution and each sample solution. solution and each sample solution. Six Six standard standard solutions solutions (described (described later) later) were were measured measured for for calibration calibration and and the the pH pH values values were were derived from the following equation. derived from the following equation. pH = pH0 + V/k, (1) pH = pH0 + ∆∆V/k, (1) The pH0 and k values were determined by linear approximation and the linearity of the The pH were determined by linearfactor. approximation and the linearity of the calibration calibration function was evaluated by a correlation 0 and k values function was evaluated by a correlation factor. 2.5. Evaluation of Response Times and Measurement Accuracy 2.5. Evaluation of Response Times and Measurement Accuracy The response times and measurement error of the present device were analyzed using nine Thesolutions, response six times measurement errorsolutions of the present were analyzed nine sample sample of and which were standard and device three of which were using diluted solutions solutions, six of which were standard solutions and three of which were diluted solutions (described (described later). The response curves were recorded for all nine solutions. The response times of later). The response curves recorded all nine solutions. The response of the present the present system and glasswere electrode werefor compared using a diluted solution times (6.86/100; described system and glass electrode were compared using a diluted solution (6.86/100; described later). The later). The V measurement in the present system was performed for 50 s after the solution was ∆∆V measurement in the present system was performed for 50 s after the solution was switched switched (Figure 3). In the glass electrode, the pH values were read for 10 min and plotted at 1 min (Figure 3).The In the glass electrode, the pH values were for 10 min and plotted 1 min intervals. intervals. measured values were compared withread the values provided by theatmanufacturer of The measured values were compared with the values provided by the manufacturer the standard the standard solutions and the values measured by Auto-pH [14] for the diluted of solutions. The solutions and the values measured by Auto-pH [14] for diluted solutions.byThe measured as by values measured by Auto-pH agreed well with thethe values measured a values glass electrode, Auto-pH agreed well with the values measured by a glass electrode, as previously reported [14]. previously reported [14]. The measurement measurement by by the the glass glass electrode electrode was was performed performed using using aa pH MeterF-53, F-53, The pH meter meter (pH/Ion (pH/Ion Meter Horiba, Japan) Japan) and and glass glass electrode electrode (Type (Type 9611, 9611, Horiba, Horiba, Japan) Japan) under under aa stirring stirring condition condition inside inside aa Horiba, bottle. The electrode was washed with distilled water and wiped clean just before immersion intointo the bottle. The electrode was washed with distilled water and wiped clean just before immersion sample solution. the sample solution. 2.6. Preparation of the Solutions 2.6. Preparation of the Solutions The pH values of the standard solutions were 1.68, 4.01, 6.86, 7.41, 9.18, and 10.01. To prepare The pH values of the standard solutions were 1.68, 4.01, 6.86, 7.41, 9.18, and 10.01. To prepare the diluted solutions, the standard solutions with pH values of 4.01, 6.86, and 9.18 were diluted with the diluted solutions, the standard solutions with pH values of 4.01, 6.86, and 9.18 were diluted 100-fold volumes of distilled water (pH 4.01/100, 6.86/100, and 9.18/100) [13,14]. A BLS containing with 100-fold volumes of distilled water (pH 4.01/100, 6.86/100, and 9.18/100) [13,14]. A BLS containing 60.86 M KH2PO4, 17.39 M Na2HPO4, and 200 mM KCl was prepared by diluting the pH 7.41 standard solution with a 500-fold volume of distilled water and then adding KCl solution. Air was bubbled into the three diluted solutions and BLS in sufficient quantity to prevent the CO2 in the atmosphere from changing the pH values of the solutions. The buffering capacities of the

Sensors 2017, 17, 1563

6 of 13

60.86 µM KH2 PO4 , 17.39 µM Na2 HPO4 , and 200 mM KCl was prepared by diluting the pH 7.41 standard solution Sensors 2017, 17, 1563with a 500-fold volume of distilled water and then adding KCl solution. Air was 6 of 13 bubbled into the three diluted solutions and BLS in sufficient quantity to prevent the CO2 in the solutions have reported previous [13,14]. of the reagents were obtained from atmosphere frombeen changing the in pHour values of thepapers solutions. TheAll buffering capacities of the solutions Wako Pure Chemical Industries, Ltd. (Osaka, Japan) all of the measurements were performed have been reported in our previous papers [13,14]. Alland of the reagents were obtained from Wako at room temperature (23–25 Pure Chemical Industries, Ltd.°C). (Osaka, Japan) and all of the measurements were performed at room temperature (23–25 ◦ C). 3. Results 3. Results 3.1. Calibration and the Evaluation Of Linearity 3.1. Calibration and the Evaluation Of Linearity Figure 6 shows a calibration function for the present device. The calibration line and Figure shows a calibration forpH the present device. The calibration line and calibration calibration6Equation (1) yieldedfunction values of 0 = 6.316 and k = 53.5. The correlation factor between Equation (1) yielded values of pH = 6.316 and k = 53.5. The correlation and 0 pH and V was 0.9997. The maximum standard deviation (S.D.) valuefactor of thebetween source pH potential, ∆∆V was 0.9997. The maximum standard deviation (S.D.) value of the source potential, 2.04 mV, was 2.04 mV, was recorded for the 10.01 standard solution. recorded for the 10.01 standard solution.

250 200 150 V (mV)

100 50 0 -50 -100 -150 -200 -250 1

2

3

4

5

6

7

8

9

10

pH Figure 6. Calibration function of a microfluidic pH measurement device. Average error ± S.D. The S.D. Figure 6. Calibration function of a microfluidic pH measurement device. Average error ± S.D. The values were 0.18, 0.58, 0.61, 0.15, 0.83, and 2.04 (mV) from the left to right. S.D. values were 0.18, 0.58, 0.61, 0.15, 0.83, and 2.04 (mV) from the left to right.

The mV/pH, Thek kvalue, value,53.5 53.5 mV/pH,was wasapproximately approximatelyequal equaltotothe thepH pHsensitivity sensitivityofofthe thepH-FET pH-FETused usedinin our ourexperiments experiments(58 (58mV/pH). mV/pH).We Weare arestill stillunable unabletotoexplain explainthe thesmall smalldisagreements disagreementsininthese thesevalues. values. The correlation factor, 0.9997, demonstrates the the highhigh linearity of the with the in a The correlation factor, 0.9997, demonstrates linearity ofmeasurements the measurements withdevice the device range of pH of 1.68 10.0. in a range pHto1.68 to 10.0. 3.2. 3.2.Response ResponseTime Time Figure 7 shows the response curvecurve of the of present device using pH-FETs. The raw source potentials Figure 7 shows the response the present device using pH-FETs. The raw source were converted to pH values based on the results obtained from the calibration. The 90% response potentials were converted to pH values based on the results obtained from the calibration. The 90% times were times aboutwere 2 s inabout solutions high buffering solutions low buffering response 2 s inwith solutions with highcapacities bufferingand capacities andwith solutions with low capacities. shown in As Figure 7b, the had the an approximate of 0.12 pHovershoot (9.18/100)of buffering As capacities. shown in pH-FET Figure 7b, pH-FET hadovershoot an approximate immediately after solutions with low switched. The longest settling time 0.12 pH (9.18/100) immediately afterbuffering solutionscapacities with lowwere buffering capacities were switched. The (±longest 1%) was 11 s (9.18/100). For reasons not known, the response of the 6.86 solution wasofslightly settling time (±1%) was 11 s (9.18/100). For reasons not known, the response the 6.86 delayed compared to those of the other solutions. solution was slightly delayed compared to those of the other solutions.

Sensors 2017, 17, 17, 1563 Sensors Sensors 2017, 2017, 17, 1563 1563

of 13 777 of of 13 13

10 10 9 9 8 8

1.68 1.68 4.01 4.01 6.86 6.86 7.41 7.41 9.18 9.18 10.01 10.01 4.01/100 4.01/100 6.86/100 6.86/100 9.18/100 9.18/100

pH pH

7 7 6 6 5 5 4 4 3 3 2 2 1 1 0 0

10 10

20 20

30 30

Time (s) Time (s)

(a) (a)

40 40

50 50

60 60

(b) (b)

Figure Response curves of nine solutions measured by the system; Figure 7.7. 7.(a)(a) (a) Response curves of sample nine sample sample by system; the present present system; Response curves of nine solutionssolutions measuredmeasured by the present (b) Expansion (b) Expansion of the Y-axis of four sample solutions of (a). The data are plotted from 10 s before the (b) Expansion the Y-axis four sample solutions of (a). The data are plotted from 10the s before the of the y-axis ofoffour sampleofsolutions of (a). The data are plotted from 10 s before solutions solutions were switched. solutions were switched. were switched.

Figure 88 compares the response curves measured by the present system and aa conventional Figure comparesthe theresponse responsecurves curves measured by the present system and conventional Figure 8 compares measured by the present system and a conventional glass glass electrode using the diluted solution (6.86/100). The pH values taken from the present system glass electrode using the diluted solution (6.86/100). The pH values taken from the present system electrode using the dilutedof solution (6.86/100). The pH valuessignal takenof from the present systemwhich were were plotted at 11 s. contrast to sensor the electrode, were plotted at intervals intervals of contrast s. In In to contrast to the the sensor signal of the glass glass electrode, which plotted at intervals of 1 s. In the sensor signal of the glass electrode, which plateaued in plateaued in 77 min, the signal the system plateaued within 20 s. plateaued min, the present signal of ofsystem the present present system plateaued 7 min, the in signal of the plateaued within 20 s. within 20 s.

Figure 8. Response curves of the diluted solution (6.86/100) measured by the present system and a conventional glass electrode. data at Time = 0 (6.86/100) for the microfluidic plottedsystem immediately Figure 8. curves the solution measured by and Figure 8. Response Response curves of of The the diluted diluted solution (6.86/100) measureddevice by the theispresent present system and aa before the solution for the microfluidic device was switched. The data at Time = 0 for the glass electrode conventional glass electrode. The data at Time = 0 for the microfluidic device is plotted immediately conventional glass electrode. The data at Time = 0 for the microfluidic device is plotted immediately is plotted the electrode device was immersed in the solution forat theTime electrode. before the solution the was The == 00 for before theimmediately solution for forafter the microfluidic microfluidic device was switched. switched. The data data at Time for the the glass glass electrode is plotted immediately after the electrode was immersed in the solution for the electrode. electrode is plotted immediately after the electrode was immersed in the solution for the electrode.

3.3. pH Measurement and Evaluation of Measurement Accuracy 3.3. Measurement and Evaluation Measurement Accuracy 3.3. pH pH Measurement andmeasurement Evaluation of oferror Measurement Accuracy Table 1 shows the for principal parameters in the results for the six undiluted

Table measurement error for in the for six solutions 5) andthe three diluted solutions 4). Figureparameters 9 shows the distribution in the Table(n11 =shows shows the measurement error (n for=principal principal parameters in error the results results for the the six undiluted solutions (n = 5) and three diluted solutions (n = 4). Figure 9 shows the error distribution measurements. The(nmargin of three error diluted was 0.0343 ± 0.0974 (average error the ± S.D., = 42). The undiluted solutions = 5) and solutions (n = pH 4). Figure 9 shows errorndistribution in the of was ±± 0.0974 (average ±± S.D., == 42). measurement error forThe themargin six undiluted was 0.0158 pH ± 0.0849 pH error (n = 30) andn forThe the in the measurements. measurements. The margin of error errorsolutions was 0.0343 0.0343 0.0974 pH (average error S.D., nthat 42). The measurement error for the six undiluted solutions was 0.0158 ± 0.0849 pH (n = 30) and that for the three diluted solutions lowundiluted buffering capacities = 12). measurement error forwith the six solutions was 0.0806 0.0158 ± ± 0.115 0.0849pH pH(n(n = 30)The andmeasurement that for the three diluted solutions with low buffering capacities was 0.0806 ± 0.115 pH (n error and S.D. were larger for the diluted solutions than for the undiluted solutions. The measurement three diluted solutions with low buffering capacities was 0.0806 ± 0.115 pH (n == 12). 12). The The measurement error and S.D. were larger for the diluted solutions than for the undiluted solutions. error was especially large for the 9.18 and 9.18/100 solutions. measurement error and S.D. were larger for the diluted solutions than for the undiluted solutions. The The measurement measurement error error was was especially especially large large for for the the 9.18 9.18 and and 9.18/100 9.18/100 solutions. solutions.

Sensors 2017, 17, 1563

8 of 13

Table 1. pH measurement error using the present system. Number 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42

Solution (True Value)

∆∆V (mV)

pH

Error of pH

Calibration *

1.68 (1.68)

236.03 233.22 234.44 236.03 235.72 117.86 116.27 115.24 115.48 117.49 −33.45 −33.94 −35.71 −32.53 −34.97 −62.81 −61.89 −63.78 −59.39 −60.98 −139.89 −138.98 −139.47 −142.21 −141.66 −193.30 −190.98 −194.28 −192.69 −191.53 75.99 77.27 82.52 76.30 −35.10 −31.01 −33.26 −35.28 −44.92 −45.59 −43.70 −44.13

1.644 1.658 1.634 1.602 1.608 3.960 3.960 3.982 3.977 3.937 6.925 6.921 6.955 6.893 6.941 7.501 7.471 7.509 7.422 7.453 9.012 8.990 9.000 9.054 9.043 10.06 10.01 10.08 10.05 10.03 4.696 4.670 4.562 4.690 6.984 6.900 6.947 6.988 7.187 7.201 7.162 7.171

0.036 0.022 0.046 0.078 0.072 0.050 0.050 0.028 0.033 0.073 −0.065 −0.061 −0.095 −0.033 −0.081 −0.091 −0.061 −0.099 −0.012 −0.043 0.168 0.190 0.180 0.126 0.137 −0.049 −0.004 −0.069 −0.038 −0.015 −0.097 −0.071 0.037 −0.091 0.075 0.159 0.112 0.071 0.186 0.172 0.211 0.202

1 2 2 2 2 1 2 2 2 2 1 2 2 2 2 1 2 2 2 2 1 2 2 2 2 1 2 2 2 2 3 3 3 3 3 3 3 3 3 3 3 3

4.01 (4.01)

6.86 (6.86)

7.41 (7.41)

9.18 (9.18)

10.01 (10.01)

4.01/100 (4.599)

6.86/100 (7.059)

9.18/100 (7.373)

Averaged Error of pH ± S.D. 1–42 (all) 1–30 (Non-diluted) 31–42 (Diluted)

0.0343 ± 0.0974 0.0158 ± 0.0849 0.0806 ± 0.115

* (1/k, pH0 ) = 1: (−0.0196, 6.270); 2: (−0.0197, 6.252); 3: (−0.0206, 6.262).

Sensors 2017, 17, 1563

89 of 13

Figure 9. The error distribution of six standard solutions and three diluted solutions (with arrow) measured by the present system. The measurement error was 0.0343 ± 0.0974 pH (average error ± S.D., Figure 9. The error distribution of six standard solutions and three diluted solutions (with arrow) n = 42). The dotted lines show the average error and S.D. measured by the present system. The measurement error was 0.0343 ± 0.0974 pH (average error ± S.D., n = 42). The dotted lines show the average error and S.D.

4. Discussion

Table 1. pH measurement error using the present system

4.1. Response Time and Mechanism of Reduction Solution Figure 7 demonstrates the quick device Number Vresponse (mV) of the pHpresentError of even pH in low-buffering-capacity Calibration * (True Value) solutions. Figure 8 shows the dramatically reduced response time of the present device compared to 1 1.68 236.03 1.644 0.036 1 that of the glass electrode. We conjecture that the rapid response was accomplished not only by the 2 (1.68) 233.22 1.658 0.022 2 fast response of the pH-FET sensor itself, but also the positive effect of the flowing junction [4,8,10]. 3 234.44 1.634 0.046 2 The pH-FET used in this work is a catheter-tip type [21] with a gate region covered with a pH-sensitive 4 236.03 1.602 0.078 2 layer of Ta2 O5 and a material enabling a response time of less than 100 ms [22]. The rapid response of 5 235.72 1.608 0.072 2 the present system with the flowing junction is assumed to be achieved by the following mechanism: 6 4.01 117.86 3.960 0.050 1 (i) When the sample solution and BLS—two solutions with different compositions—meet, a liquid 7 (4.01) 116.27 3.960 0.050 2 junction potential (LJP) forms due to the difference between the transport number of the electrolytes 8 115.24 3.982 0.028 2 for each solution. The LJP causes measurement error, and the degree of the LJP varies in each sample 9 115.48 3.977 0.033 2 solution. (ii) In a usual case, the LJP requires time to stabilize because the speed with which it forms 10 117.49 3.937 0.073 2 depends on the speed with which the molecules diffuse. (iii) In the present device, the pH-FET sensor 11 6.86 −33.45 6.925 −0.065 1 and Ag/AgCl pseudo reference electrode set near the junction picks up the sensor signal before the LJP 12 (6.86) −33.94 6.921 −0.061 2 forms. The flow of the two solutions breaks the LJP as soon as it forms, thereby maintaining a stable LJ 13 −35.71 6.955 −0.095 2 condition during the measurement. A previous paper describes a similar mechanism in another LJ 14 −32.53 6.893 −0.033 2 device bed as a “fresh junction” [8]. 15 −34.97 6.941 −0.081 2 The overshoot of the present device was clearly lower than that of a flow-through-type sensor 16 7.41 −62.81 7.501 −0.091 1 probe studied earlier, which overshot the readings by 0.2 pH when similar sample solutions were fed 17 (7.41) −61.89 7.471 −0.061 2 in [14]. We presume that the flowing state of the LJ is critical for this reduction in the overshoot during 18 −63.78 7.509 −0.099 2 measurement. The area of contact between the two solutions in the channel may also be a decisive 19 −59.39 7.422 −0.012 2 factor. This area of contact was remarkably larger in the present device than in the flow-through-type 20 −60.98 7.453 −0.043 2 sensor probe just mentioned (about 10 mm2 vs. 0.23 mm2 ) [14]. For reasons we have yet to determine, 21 9.18 −139.89 9.012 0.168 1 the sensor signals in some of the sample solutions became unstable for periods of about 20 s after the 22 (9.18) −138.98 8.990 0.190 2 solutions were switched. 23 −139.47 9.000 0.180 2 The potential to shorten the measurement times of the present device looks promising. Judging 24 −142.21 9.054 0.126 2 from the stability of the response curves 20 s after the solutions were switched (Figure 7), we surmise 25 −141.66 9.043 0.137 2 that it may be possible to shorten the measuring period by another 20 s per sample even when leaving 26 10.01 −193.30 10.06 −0.049 1 27 (10.01) −190.98 10.01 −0.004 2

Sensors 2017, 17, 1563

10 of 13

a margin of 10 s. Our results also indicate that the present device using pH-FET sensors responds more rapidly than the conventional method with glass electrodes even in solutions with low buffering capacities (β < 0.5 mM/pH) [13,14]. As previously mentioned, more than 7 min is required to reach a plateau during measurement of such solutions by the glass electrode method [13,14]. 4.2. Measurement Accuracy We know that the flowing junction worked to reduce the measurement error, because no remarkable overshoots or delays of the response curve were observed when the solution fed to the pH-FETm was switched from the BLS to a low-buffering-capacity sample solution (pH 4.01/100, 6.86/100, or 9.18/100). As shown in Figure 9, the measurement errors were large for the 9.18 and 9.18/100 solutions. These solutions are composed of tetraborate, and the mechanism explaining the large measurement error is still unexplained. As calculated in Table 1, the measurement error and S.D. were both larger for the diluted solutions than for the undiluted solutions. This difference in measurement accuracy between the two types of solutions could stem from the pH changes caused by CO2 , as solutions with low buffering capacities easily absorb CO2 in the atmosphere during measurement. The measurement error was large for the solutions with low buffering capacities and was slightly larger overall compared to the data obtained by the flow-through-type sensor probe [13,14]. 4.3. Advantages of the Present Device While the present system consumes BLS, the concentration of the BLS is not altered during the measurement. The concentration of the internal filling solution inside the glass electrode is increased by evaporation. This advantage presumably contributed to the reduced measurement error. The linearity of the sensor signal of the present system remained in the range of 1.68–10.0, indicating that calibration within a narrow range becomes unnecessary even for precise measurements. With a glass electrode, two-point calibration within a narrow range is indispensable for precise measurements due to the insufficient linearity of the signal. The present device has advantages compared to the glass electrodes. Thanks to its small dimensions, the pH-FET sensor used in this work can be easily packed into an outer casing made of plastic to effectively insulate the sensor body against shock. The pH-FET sensor is also made by the same mass production process used for semiconductors, assuring a reasonable manufacturing cost should the present system come onto the market in the future. While the total size of the present system may be larger than a conventional glass electrode apparatus at this trial stage, the system will be downsized by the integration of peripheral constitution elements such as a reservoir, amplifier, and microprocessor. 4.4. Considerable Application of the Present Device The pH values of environmental water are evaluated to protect the environment from adverse effects. Environmental water—such as river water, pond water, tap water, or water discharged from an industrial plant (e.g., boiler water)—often has a low buffering capacity, a condition that considerably extends the time required for precise pH measurements using a glass electrode. The superior response of the present system to solutions with low buffering capacities probably helps to shorten the measurement time and improve measurement accuracy. The present system could be potentially applied, for example, as an in-line pH measurement apparatus for small-scale manufacturing plants (Figure 10). Solutions in such a plant could be measured using small amounts of sample solution without any risk of contamination, as sample solution in a main pipe could be branched to the inlet by a tube, requiring no insertion of an electrode directly into the solution.

Sensors 2017, 2017, 17, 17, 1563 1563 Sensors

11 of of 13 13 11

Branch

Flow

Transport Pipe

Sample solution

Reservoir 2 (BLS)

Figure 10. AAschematic schematic representation representation ofof application application ofof the the present present system system as as an an in-line in-line pH measurement apparatus.

4.5. Prospects Prospects for Improving the Device The concentration concentration of of the the KCl KCl solution solution needs needs to to be more fully considered. considered. For For ease ease of of handling, handling, we prepared prepared the the BLS BLSfrom fromaa200 200mM mMKCl KClsolution solution rather than a saturated (3.3 M) KCl solution, as rather than a saturated (3.3 M) KCl solution, as the the latter would reduced theremarkably LJP remarkably [23]. LJP occurring thesolution sample latter would havehave reduced the LJP [23]. The LJPThe occurring betweenbetween the sample solution KCl solution to cause measurement errors due to the variation the LJP and KCl and solution is knownistoknown cause measurement errors due to the variation in the LJPinresulting resulting from the changing compositions of thesolution sample [4–8]. solution [4–8]. In this research, we obtained from the changing compositions of the sample In this research, we obtained superior superior measurement accuracy, presumably ofsolution the sample solution and BLS measurement accuracy, presumably because thebecause flows ofthe theflows sample and BLS decreased the decreased thethe LJP betweenThe theLJP solutions. The LJP be could probably be by further reduced increasing LJP between solutions. could probably further reduced increasing the by concentration the concentration of KCl in the BLS. of KCl in the BLS. Many factors can improve the measurement performance of the present device. The dimension of the thechannel, channel, example, is presumed the flow characteristics. The surface for for example, is presumed to affectto theaffect flow characteristics. The surface characteristics characteristics the channel, such as the roughness and surface energy, are to also the fluid of the channel,ofsuch as the roughness and surface energy, are also critical thecritical fluid to flows. The flows. Thetime, response time, measurement stability, and measurement accuracy may beby improved by response measurement stability, and measurement accuracy may be improved optimizing optimizing these parameters. Direct measurement of the LJP is essential, but the very narrow width these parameters. Direct measurement of the LJP is essential, but the very narrow width of the channel of the channel hinders thereliability. measurement pH-FET is not crucial for hinders the measurement Whilereliability. pH-FETr isWhile not crucial forr measurement, the measurement, present device the present device composed of a differential measurement circuit with two pH-FETs and an composed of a differential measurement circuit with two pH-FETs and an Ag/AgCl exhibits superior Ag/AgCl exhibits superior performance against noise and changes of temperature. performance against noise and changes of temperature. We expect expect that that rearrangements rearrangementsof ofthe thepH-FETs pH-FETsand andchannel channel will allow further miniaturization will allow further miniaturization of of the LJ device. The integration of reservoirs into the main body of the measurement device also the LJ device. The integration of reservoirs into the main body of the measurement device also seems seems The required be reduced by narrowing the channel likely. likely. The required samplesample volumevolume can be can reduced by narrowing the channel further.further. Conclusions 5. Conclusions

We have combined a microfluidic device with pH-FETs to construct a new pH measurement device. The The present present device device had had a shorter response time than the glass system with a flowing-type LJ device. electrode in the pH measurement for the low-buffering-capacity solution. The method presented in this paper will be useful for the measurement of pH in fields of environmental environmental science. science. Acknowledgments: The grateful to Tetsushi Sekiguchi of Waseda University, and Tetsuya Acknowledgments: The authors authorsare arevery very grateful to Tetsushi Sekiguchi of Waseda University, and Makino, and Toshiro Sugiyama of Nihon Kohden Corp. for kindly providing the ISFET sensors used in the Tetsuya Makino, Sugiyama of Nihon Kohden Corp. kindly providing and the ISFET in present work. Weand are Toshiro also grateful to Michihiro Nakamura for hisfor helpful discussions criticalsensors readingused of this the present work. We are also Michihiro Nakamura for his helpful discussions and critical of paper. The authors would likegrateful to thanktoAkihiko Wakuda, Masashi Nobukuni, and Tsuyoshi Miura forreading their kind technical Thiswould study was partly supported by a Grant-in-Aid for Young and Scientists (B) (No. this paper.support. The authors like to thank Akihiko Wakuda, Masashi Nobukuni, Tsuyoshi Miura25820093) for their from the Japan Society for the Promotion of Science. kind technical support. This study was partly supported by a Grant-in-Aid for Young Scientists (B) (No. 25820093) from the Japan Society for the Promotion of Science.

Sensors 2017, 17, 1563

12 of 13

Author Contributions: A.Y. conceived and designed the experiments; A.Y. and M.S. performed the experiments; A.Y. analyzed the data and wrote the paper. Conflicts of Interest: The authors declare no conflict of interest.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.

11. 12.

13. 14. 15.

16. 17. 18.

19.

20.

Haber, F.; Klemensiewicz, Z. Über Elektrische Phasengrenzkräfte. Z. Phys. Chem. 1909, 67, 385–431. [CrossRef] Bergveld, P. Development of an ion-sensitive solid-state device for neurophysiological measurements. IEEE Trans. Biomed. Eng. 1970, BME-17, 70–71. [CrossRef] Bergveld, P. Thirty years of ISFETOLOGY what happened in the past 30 years and what may happen in the next 30 years. Sens. Actuators B Chem. 2003, 88, 1–20. [CrossRef] Kakiuchi, T. Salt bridge in electroanalytical chemistry: past, present, and future. J. Solid State Electrochem. 2011, 15, 1661–1671. [CrossRef] Covington, A.K.; Whalley, P.D.; Davison, W. Procedures for the measurement of pH in low ionic strength solutions including freshwater. Analyst 1983, 108, 1528–1532. [CrossRef] Covington, A.K.; Rebelo, M.J.F. Reference electrodes and liquid junction effects in ion-selective electrode potentiometry. Ion-Sel. Electrode Rev. 1983, 5, 93–128. Tower, O.F. Ueber potentialdifferenzen an den berührungsflächen verdünnter lösungen. Z. Phys. Chem. 1896, 20, 198–206. [CrossRef] Lamb, A.B.; Larson, A.T. Reproducible liquid junction potentials: The flowing junction. J. Am. Chem. Soc. 1920, 42, 229–237. [CrossRef] MacInnes, D.A.; Yeh, Y.L. The potentials at the junctions of monovalent chloride solutions. J. Am. Chem. Soc. 1921, 43, 2563–2573. [CrossRef] Scatchard, G. The activities of strong electrolytes. III. The use of the flowing junction to study the liquid junction potential between dilute hydrochloric acid and saturated potassium chloride solutions; and the revision of some single-electrode potentials. J. Am. Chem. Soc. 1925, 47, 696–709. [CrossRef] Metcalf, R.C. Accuracy of ross pH combination electrodes in dilute sulphuric acid standards. Analyst 1987, 112, 1573–1577. [CrossRef] Yamada, A.; Mohri, S.; Nakamura, M.; Naruse, K. Optimizing the conditions for pH measurement with an automated pH measurement system using a flow-through-type differential sensor probe consisting of pH-FETs. J. Robot. Mechatron. 2010, 22, 197–203. [CrossRef] Mohri, S.; Nakamura, M.; Naruse, K. Automation of pH measurement using a flow-through type differential pH sensor system based on pH-FET. IEEJ Trans. Sens. Micromach. 2007, 127, 367–370. [CrossRef] Yamada, A.; Mohri, S.; Nakamura, M.; Naruse, K. A fully automated pH measurement system for 96-well microplates using a semiconductor-based pH sensor. Sens. Actuators B Chem. 2010, 143, 464–469. [CrossRef] Yamada, A.; Mohri, S.; Nakamura, M.; Naruse, K. A Simple method for decreasing the liquid junction potential in a flow-through-type differential pH sensor probe consisting of pH-FETs by exerting spatiotemporal control of the liquid junction. Sensors 2015, 15, 7898–7912. [CrossRef] [PubMed] Sia, S.K.; Whitesides, G.M. Microfluidic devices fabricated in poly (dimethylsiloxane) for biological studies. Electrophoresis 2003, 24, 3563–3576. [CrossRef] [PubMed] Yager, P.; Edwards, T.; Fu, E.; Helton, K.; Nelson, K.; Tam, M.R.; Weigl, B.H. Microfluidic diagnostic technologies for global public health. Nature 2006, 442, 412–418. [CrossRef] [PubMed] Fukuba, T.; Tamai, Y.; Kyo, M.; Shitashima, K.; Koike, Y.; Fujii, T. Development and field evaluation of ISFET pH sensor integrated with self-calibration device for deep-sea oceanography applications. In Proceedings of the 12th International Conference on Miniaturized Systems for Life Sciences, San Diego, CA, USA, 12–16 October 2008; pp. 1983–1985. Rérolle, V.M.; Floquet, C.F.; Harris, A.J.; Mowlem, M.C.; Bellerby, R.R.; Achterberg, E.P. Development of a colorimetric microfluidic pH sensor for autonomous seawater measurements. Anal. Chim. Acta 2013, 786, 124–131. [CrossRef] [PubMed] Maruyama, H.; Arai, F.; Fukuda, T. On-Chip pH Measurement Using Functionalized Gel-Microbeads Positioned by Optical Tweezers. Lab Chip 2008, 8, 346–351. [CrossRef] [PubMed]

Sensors 2017, 17, 1563

21. 22. 23.

13 of 13

Shimada, K.; Yano, M.; Shibatani, K.; Komoto, Y.; Esashi, M.; Matsuo, T. Application of catheter-tip ISFET for continuous in vivo measurement. Med. Biol. Eng. Comput. 1980, 18, 741–745. [CrossRef] [PubMed] Nakatsuka, T.; Nakamura, M.; Sano, T. Applications of hydrogen ion sensitive FET to the kinetic study of fast reaction in solution. Bull. Chem. Soc. Jpn. 1988, 61, 799–802. [CrossRef] Guggenheim, E.A. A study of cells with liquid-liquid junctions. J. Am. Chem. Soc. 1930, 52, 1315–1337. [CrossRef] © 2017 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

Suggest Documents