1 Introduction

Microfluidics have a great potential for biomedical applications (Yager et al. 2006) and in 2007, patterned paper has been proposed as a substrate material resulting in a low-cost and portable bioassay (Martinez et al. 2007). Since their establishment as a medical diagnostic tool (Martinez et al. 2010), microfluidic paper-based analytical devices (\(\mu\)PADs) initiated a new class of inexpensive and easy-to-use point-of-care diagnostics for the developing world (Yetisen et al. 2013; Cate et al. 2015; Liu et al. 2016; Fu 2017).

The functionality of \(\mu\)PADs is based on the ability of paper to transport liquid passively due to capillary action in regions where the imbibed fluid undergoes a kind of chemical analysis. Hence, the majority of current research focuses on: (1) advanced fabrication techniques for accurate control of imbibition dynamics, which include effects of paper geometry (Elizalde et al. 2015; Walji and MacDonald 2016), repeated wettings (Elizalde et al. 2016), humidity (Castro et al. 2017), fibre nature (Böhm et al. 2014), hydrophobic patterning and wax boundaries (Hong and Kim 2015; Koivunen et al. 2016), use of programmable devices that include 3D paper channels (Li and Liu 2014), as well as on (2) the principal functionality of \(\mu\)PADs, which is the ability to quantify the analyte.

The most common detection method is based on colorimetric approaches, e.g. Dungchai et al. (2010); Evans et al. (2014), where the colour intensity of the wetted paper is proportional to analyte concentration. This is based on enzymatic or chemical colour change reactions, while the compatibility of such devices with modern smartphone cameras (Il Hong and Chang 2014; Wu et al. 2017) facilitates a portable diagnostic procedure. Colorimetric methods have been used for glucose detection in urinalysis (Gabriel et al. 2016; Tarso et al. 2014), tear electrolyte analysis (Yetisen et al. 2017), paraquat poisoning in human serum (Kuan et al. 2016) as well as blood separation (Songjaroen et al. 2012). However, different light conditions provide inconsistent image intensities limiting the usage in real-world applications (Yamada et al. 2017; Morbioli et al. 2017). Increased sensitivity has been achieved by coupling \(\mu\)PADs with electrochemical readers (Noiphung et al. 2013; Lankelma et al. 2012; Nie et al. 2010), while distance-based detection methods have been also reported (Cate et al. 2013; Li et al. 2014; Elizalde et al. 2016), where the concentration of the analyte is related to the travelled distance of the imbibed liquid.

The diagnostic method we are presenting here is based on the temporal heat release at a wetting front during spontaneous imbibition. Such temperature spikes have been initially suggested as a detector for solid–liquid chromatography in late 1950s (Claxton 1959) and they have been also observed during infiltration of dry soils (Anderson and Linville 1960). However, Terzis et al. (2017) indicated that the released heat can be enormous for the given scale of \(\mu\)PADs and related the level of temperature rise to the energetics of imbibition compounds. The objective of this work is to show that the temperature rise can be associated with a prospective liquid concentration demonstrating a great potential for quantitative paper-based microfluidics.

Fig. 1
figure 1

Schematic illustration of experimental setup and characterisation of paper substrate by scanning electron microscopy (SEM) and X-ray diffraction (XRD) analysis. In the XRD patterns, the characteristic diffraction peaks of cellulose are observed for 16.5\(^\circ\), 22.7\(^\circ\) and 34.2\(^\circ\). SEM images were obtained on a FEI Nova Nano 450 operating at 5 kV. XRD patterns obtained using a Bruker D8 Advance (CuKa, 40 kV, 40 mA)

Full size image

2 Results

2.1 Introductory remarks

The test setup is described in Terzis et al. (2017); thus, only a brief overview will be given here. All experiments were carried out using an uncoated and chemically untreated paper (Löschblätter Art. 10-41 546, BRUNNEN), as shown in Fig. 1. The scanning electron images and X-ray diffraction analysis of the material revealed an average fibre size of 12 \(\upmu \mathrm{m}\) and its purely cellulosic character. The paper strips have 231 \(\upmu \mathrm{m}\) thickness, 140 g/\(\mathrm{m}^2\) weight per area and a porosity of 0.59. In order to eliminate liquid evaporation from the wetted paper, its front surface was covered with a clear polypropylene tape that is also transparent to infrared radiation. The liquid bath was set at room temperature, and once isothermal conditions were obtained, it was vertically transversed to force imbibition against gravity. Optical and thermal images were recorded at 4 Hz to quantitatively visualise the liquid invasion process.

2.2 Capillary thermodynamics of aqueous urea solutions

Figure 2 shows optical and thermal images of the imbibition process for two aqueous urea solutions with urea concentration of 16.6 and 33.3% (w/w). The 16.6% solution propagates faster, and this is directly reflected in Washburn’s equation, z(t) \(\sim\) \(\sqrt{\gamma _l/\mu }\) (Washburn 1921). Although both the dynamic viscosity (\(\mu\)) and surface tension (\(\gamma _l\)) increase with urea mass fraction, the influence on \(\mu\) is more pronounced (Halonen et al. 2016) decelerating the wicking rates.

Fig. 2
figure 2

Optical (black and white) and thermal (colour) images of urea–water solution spontaneous imbibition. Consecutive frames are displayed in a logarithmic time scale. The paper strips are cut in 5 \(\times\) 50(mm)\(^2\)

Full size image
Fig. 3
figure 3

Temperature spikes at the wetting front for two urea–water solutions (UWS) at \(z=20\) mm and contact angles (CA) on a PDMS surface. std = ± 0.08 \(^{\circ }\)C

Full size image
Fig. 4
figure 4

a Capillary dynamics of various urea–water solutions (UWS), b local temperature rise at the wetting front as a function of urea concentration at various imbibition heights and c average temperature rise for the complete wetting of the paper

Full size image

Regarding the released heat, for both solutions the temperature is temporarily increased at the interface between the wetting and the non-wetting phases, while the level of heat release is gradually reduced at later imbibition stages where the liquid front becomes more diffusive and hence less saturated (Helmig 1997). However, the temperature rise is higher for \(x_i\)=16.6% over the complete length of the paper. Figure 3 shows that \(\Delta T\) at \(z=20\) mm is about 0.5 K higher for \(x_i\) =16.6%, while thermal spike is similar to that observed in solidification fronts inside sessile drops (Schmid et al. 2018). The figure also shows the similar wettability of the solutions on a low energy and nearly apolar PDMS (polydimethylsiloxane) surface. The contact angle is just slightly reduced for \(x_i\) = 33.3% indicating the more apolar nature of the solution at higher mass fractions, in agreement with Terzis et al. (2018). Therefore, the released energy cannot only be attributed to the heat of wetting on the slightly polar cellulosic surface. Terzis et al. (2017) related the level of temperature rise to the energetics of imbibition compounds, as an acid–base neutralisation reaction.

Table 1 Surface energy components of urea–water solutions (Terzis et al. 2018) and cellulose (Van Oss 2006) at 25 \(^{\circ }\)C
Full size table

Cellulose is known to be characterised by a significant electron-donor capacity (Dourado et al. 2012), while Terzis et al. (2018) indicated that as urea concentration increases, the electron donating behaviour of the solution is also enhanced. Table 1 summarises typical energetic characteristics for cellulose and urea–water solutions, including their acid-based contributions. As urea concentration increases, the paper and the imbibed liquid become energetically more similar. Hence, any charge transfer or neutralisation reaction at the interface is less intense. Since the interfacial energy between two energetically similar surfaces is zero, this means that the released heat at the wetting front can be related to the energetic dissimilarity between the compounds involved.

The capillary dynamics of urea–water solutions is shown in Fig. 4a where the imbibition rate is significantly reduced as urea concentration increases. This is due to the considerable increase in dynamic viscosity, which is twice the value of pure water for \(x_i\) = 50%. Figure 4b shows for various travelled distances, how the local \(\Delta T\) varies with urea mass fraction. Pure water provides the maximum thermal peak due to the strong hydrogen bonding with the available hydroxyls (OH\(^{-}\)) of the fibre surfaces. However, Terzis et al. (2018) showed that the polarity of the solution decreases with urea concentration, and thus, the hydrogen bond opportunities with the fibre surfaces should be also reduced. Indeed, for a 50% (w/w) solution, the local \(\Delta T\) is reduced from 2.35 to 1.26 K at \(z=10\) mm (− 46.3%) and from 1.25 to 0.41 K for \(z=\) 25 mm (− 67.2%). This shows also that the influence of urea on the released energy is more pronounced at higher imbibition lengths. The latter can be associated with the stronger adsorption capacity of water molecules on the cellulose that results in higher urea concentrations at later imbibition stages. A direct relation of temperature rise with the imbibition speed is indicated in Fig. 4c which shows five individual experiments using distilled water and a 16.6% aqueous solution. For a given wetting front speed, water provides higher temperature rise which indicates that the differences in the released heat are due to the presence of the analyte. Furthermore, the influence of urea is more intense when the wicking velocity is slow (higher imbibition heights). In particular, the obtained temperature rise for water is about 38 and 8% higher when the imbibition velocity is 0.1 and 0.8 mm/s, respectively.

Fig. 5
figure 5

Optical and thermal images of water imbibition for various paper sizes. The images are displayed 40 s after the initiation of imbibition

Full size image

2.3 Influence of paper size

Figure 5 shows image-pair snapshots of pure water imbibition at 40 s and various paper sizes. The first three image pairs show a single paper layer of 4, 6 and 8 mm width. In general, the travelled distance of water is observed to be reduced for narrower channels in agreement with Castro et al. (2017). Although an opposite trend in the literature has been also observed (Böhm et al. 2014; Walji and MacDonald 2016), the imbibition dynamics in this study is additionally influenced by the transparent polypropylene tape on the front and side surfaces of the paper. This generates hydrophobic boundaries which are able to retard imbibition dynamics since the interfacial tension in these regions acts in an inverse direction to the flow (Hong and Kim 2015). Interestingly, the influence of paper width is significantly reflected in the thermal images since the temperature rise increases for wider channels. This is probably related to the higher influence of the hydrophobic boundaries that reduce the saturation of the wetting front. The last two image pairs show a double and a quadruple layer of a 6 mm paper width. The imbibition speed increases with the overall thickness of the substrate; however, the effect is less pronounced for a quadruple layer. This could be attributed to an imperfect packing of the individual paper strips. In order to truly evaluate paper thickness effects, individual strips made of different thicknesses, instead of multi-layer packing, should be used. Nevertheless, there is a substantial increase in the released heat since the thermal peak for the quadruple paper exceeds 4.5 K. For multi-layer paper samples, similar findings were also reported by Aslannejad et al. (2017).

Figure 6 summarises all experiments performed with water (including paper strips that are not present in Fig. 5) and shows that the average temperature rise for the complete wetting of the paper increases almost linearly with the substrate cross section (width \(\times\) thickness). This is due to the larger interfacial area between the fluid and the porous platform for thicker papers and due to the smaller influence of hydrophobicity in the boundaries for wider channels. The latter has a direct influence on the saturation of the wetting front as well as the heat losses in the side boundaries.

Fig. 6
figure 6

Average temperature rise over the complete length of paper substrates as a function of imbibition paper cross section

Full size image

2.4 Artificial urine samples

The applicability of biological fluids to produce a temperature rise at the wetting front has been also examined using artificial urine solutions produced according to DIN EN 1616:1999, while glucose was selected as the specific analyte for identification. Similar to cellulose, glucose is also characterised by a significant electron-donor capacity (Van Oss 2006), which means that the higher the glucose concentration is, the more energetically compatible are the cellulosic platform and the imbibed liquid. Hence, the temperature spikes should be reduced. Indeed, Fig. 7c shows that the average temperature rise for the complete wetting of the paper is reduced with increasing glucose concentration. However, the detectable level of glucose does not represent realistic applications, since typical values in human urine are less than 30 mM. Nevertheless, Fig. 7 demonstrates a very clear monotonic trend where proper patterning of the cellulosic platform with appropriate reagents might enhance the thermal sensitivity in various concentrations.

Fig. 7
figure 7

Glucose level detection based on the average temperature rise during the complete wetting of 5 \(\times\) 35(mm)\(^2\) paper strip. The error bars correspond to the uncertainty of five individual experiments

Full size image

3 Concluding remarks

In this brief communication, we presented the potential of a temperature-based quantitative diagnostic approach for paper-based microfluidics, inspired by the thermal spikes observed at the wetting front during capillary filling of cellulosic materials (Terzis et al. 2017). The applicability of biological fluids was examined, and the obtained temperature rise at the liquid front was successfully related to the concentration of specific analytes, such as urea (in aqueous solutions) and glucose (in artificial urine). Although the detectable concentrations do not represent realistic levels, we believe that the monotonic trend that can be achieved as well as proper patterning of the cellulosic platform with appropriate reagents, will result in a more sensitive response of a temperature sensor according to different analyte concentrations. The temperature rises can be enormous for the given length scale of \(\mu\)PADs exceeding locally 3 K for a purely cellulosic substrate. This demonstrates a great potential for future, rapid and quantitative diagnostics for paper-based microfluidics.