Solvent-free fabrication of biodegradable hot-film flow sensor for noninvasive respiratory monitoring

Figure 2(a) shows the scanning electron microscope (SEM) image of the printing paper, which indicates its porous nature. It is also evident that a network of cellulose fibers is randomly oriented, which assists in rubbing off particles from the pencil lead, and retaining them on the paper during the drawing process. Figure 2(b) displays the SEM image of the pencil trace deposited on the paper substrate. To investigate the properties of pencil trace drawn on paper, we performed Raman measurements and observed three sharp peaks at wavenumbers of 1350, 1580 and 2725 cm⁻¹ (figure 2(c) corresponding to the D, G and 2D bands of graphite material [29, 34], respectively. The good quality of graphite material is confirmed because the intensity of G-band (showing a sp² bonded carbon network) is approximately three times higher than that of D-band (showing the information about defects). As such, the average size L of graphite crystallites is calculated to be approximately 45 nm using the following estimation [34], $L=k\times {{\lambda}^{4}}\times \left({{I}_{\text{G}}}/{{I}_{\text{D}}}\right)$, where k is a constant and λ is the laser wavelength; $\left({{I}_{\text{G}}}/{{I}_{\text{D}}}\right)=3$ is the density ratio of the G-band and D-band. In addition, the main components of pencil lead (5B type) used in this study are graphite crystallites (82 wt$\%$ [35]). These graphite crystallites contribute to the good electrical conductivity of the as-drawn graphite trace. Evidently, figure 2(d) shows a relatively low resistance (10 k$\Omega$) for the 10 mm-length graphite trace. It is also evident that a linear current-voltage relationship was obtained for the pencil trace, indicating an Ohmic characteristic of the resistive film.

Zoom In Zoom Out Reset image size

Self-heated hot films have previously been proven to be an effective platform for various sensing applications, including Pirani gauges [36], gas sensors [37], and flow sensors [13, 14, 38]. To realize a high sensitivity for such diverse applications, good thermal isolation and temperature uniformity are desired for the hot films. This can be achieved with free-standing and isolated thin films such as bridge or cantilever configurations. Therefore, we constructed the graphite sensing layer on cellulose fiber paper that has a low thermal conductivity and lower heat conduction through mechanical links. Moreover, the nominal resistance of the hot film sensor is approximately 2.1 k$\Omega$, which is suitable for the heating purpose as a hot film. In addition, pencil graphite as a sensing layer would offer high sensitivity, owing to the high temperature coefficient of resistance of graphite [29].

Figure 3(a) shows the temperature distribution around the GOP flow sensor at a supply power of 120 mW taken by an infrared camera (GOBI-1513, Xenics Infrared Solutions) at an ambient temperature of 23 °C. The temperature distributions in horizontal position AA' and vertical position BB' were measured and a high temperature gradient was observed for the hot film (figures 3(b) and (c)). This high temperature gradient indicates that the temperature rise is well-localized on the GOP heater. This is attributed to the fact that power loss due to conduction is small because of the low thermal conductivity of paper [39]. It is also evident that a high temperature of more than 100 °C was achieved for a power consumption of 120 mW. This property confirms that the graphite film can be used as a self-heated element at low power consumption and low supply voltage.

Zoom In Zoom Out Reset image size

Figure 4(a) shows the working principle of the hot film flow sensor. As such, when a large constant current is applied to the sensor, temperature rise occurs around the graphite film due to the resulting Ohmic loss, and reaches a steady state when the heat loss is equal to the supplied power [40]. Under constant current operation (I  =  const), the following equation describes the the relationship between the Joule heating power and the power loss due to convective heat transfer in the steady state regime:

$$\begin{eqnarray}&&{{I}^{2}}R=hA\left(T-{{T}_{\text{e}}}\right)\end{eqnarray} \tag{ 1 }$$

where R, A and T are the resistance, the surface area and the temperature of the graphite hot film, respectively; T_e is the air temperature and h is the convective heat transfer coefficient. According to the King's law, the convective heat transfer coefficient can be expressed in the following form:

$$\begin{eqnarray}&&h=a+b{{\upsilon}^{n}}\end{eqnarray} \tag{ 2 }$$

where a, b and n are constants and $\upsilon$ is the air velocity. When a flow passes around the film, power loss increases and the temperature of the film decreases. This leads to the increase in electrical resistance of the graphite film, owing to the negative temperature coefficient of resistance of the pencil graphite material [29]. This resistance variation can be obtained by measuring the change of output voltage from the sensor. Making the substitution of equation (2) in equation (1), the output voltage V  =  IR from the sensor can be determined from the following expression:

$$\begin{eqnarray}&&{{V}^{2}}=\left(a+b{{\upsilon}^{n}}\right)RA\left(T-{{T}_{\text{e}}}\right)\quad or\quad {{V}^{2}}=A+B{{\upsilon}^{n}}\end{eqnarray} \tag{ 3 }$$

where A and B are temperature-dependent coefficients. The relationship between the output voltage difference $\Delta V$ and the air velocity can be empirically described as follows [13, 14]:

$$\begin{eqnarray} &&\Delta V=\alpha +\beta {{\upsilon}^{n}}\end{eqnarray} \tag{ 4 }$$

where α and β are empirical constants. To test the response of the flow sensor to the air flow, we employed a four-point measurement to monitor the resistance change of the sensor (figure S2, supporting information). The flow measurement setup is shown in figure S3 (supporting information). Figure 4(b) shows the response of the sensor to different air flow velocities, under differently applied constant currents. The flow responsivity of the sensor is estimated using differential output voltage [$\Delta V=V-{{V}_{o}}$], where V and V_o are the output voltages from the sensor at flow velocities of $\upsilon$ and 0, respectively. Based on the different $\Delta$V values, the empiric constants α, β and n of equation (4) were estimated to be 226.5, 53.7 and 0.8, respectively. Therefore, the sensitivity of the sensor was calculated to 53.7 mV (m s⁻¹)^−0.8 (see table 1).

Zoom In Zoom Out Reset image size

Table 1. Comparison of performance and sustainability of the flow sensor in this work and literature.

Material	Sensitivity	Power	Biode.	Solvent	Cleanroom	Cost	References
Platinum	7.98 mV(m s−1)−0.5	45.10 mW	No	Yes (KOH)	Yes	High	[13]
Platinum	14.1 mV(m s−1)−0.8	20 mW	No	Yes (KOH)	Yes	High	[14]
Platinum	0.7479 $\Omega$(m s−1)−1	—	No	Yes (KOH)	Yes	High	[15]
Polysilicon	—	—	No	Yes (SF6)	Yes	High	[16]
Pt and Au	—	—	No	Yes (SF6)	Yes	High	[17]
W/Ti/Pt	105 μV/(μl/min)	23.5 mW	No	Yes (HCl)	Yes	High	[18]
Silicon	—	600 mW	No	Yes (KOH)	Yes	High	[19]
CNT	1.208 mV(m s−1)−0.8	4.4 mW	Yes	No	Yes	High	[30]
Nickel	50 mV/(l min−1)	—	No	Yes	Yes	High	[20]
Platinum	0.385 V2(m s−1)−0.52	—	No	Yes	Yes	High	[21]
Graphite	53.7 mV(m s−1)−0.8	78.4 mW	Yes	No	No	Low	This work

It is noteworthy that the differential output voltage increases with increasing air flow velocities due to the increase of heat losses. It is also evident that the higher the supply power or supply current, the higher the differential output voltage at the same air flow velocity. For example, at a flow rate of 1 m s⁻¹, $\Delta V$ is approximately 240 mV at a constant current of 8 mA (78.4 mW power consumption), which is at least twice the value for $\Delta V$ at 6 mA (45.3 mW power consumption). Therefore, we could improve the sensitivity of the flow sensor by increasing the supply power.

As a high SNR is desired for the flow sensor, we measured the noise of the sensor. As shown in figure 4(c), the noise density increases with increasing applied power, and decreases with increasing working frequencies. However, the maximum noise level was estimated to be lower than  −50 dBV. Since the relationship between the noise power (P_dBV) and the noise voltage (V_noise) can be expressed as ${{P}_{\text{noise}}}\left(\text{dBV}\right)=20{{\log}_{10}}\left(\frac{{{V}_{\text{noise}}}(V)}{1(V)}\right)$, the corresponding noise voltage was found to be 0.00316 V. Therefore, the minimum SNR of 40 was estimated, which is at least one order of magnitude higher than a standard SNR of a true signal (SNR  =  3). In addition, the 63.2$\%$ thermal response time of the sensor was estimated to be 1.4 s (figure S4, supporting information).

When the heating power compensates for the heat loss at the paper-based hot film interfaces, a relatively large temperature of more than 100°C could occur and be maintained for a while. This may lead to the degradation of the hot-film performance. Therefore, it is important to evaluate the long-term repeatability of the sensor response. We measured the response of the flow sensor on different days, showing a good repeatability of the sensor signal (figure 4(d)). This confirms the good long-term stability of the sensor. We also measured the resistance of the sensor under the exposure of a high humidity environment, indicating that drifts in the sensor signal need to be eliminated for accurate monitoring of flow (figure S5, supporting information).

The performance and properties of this thermal flow sensor is compared with the other thermal flow sensors reported on various rigid and plastic substrates as shown in table 1. Being a low cost, solvent-free and cleanroom-free approach, this can be employed in developing disposable flexible and wearable thermal flow sensors.

Measuring respiration is an effective approach that has been employed to support therapy for respiratory diseases. However, existing systems have normally exploited nasal cannulas, which include two small pipes to be invasively inserted into the nostrils, leading to uncomfortable breathing for patients. Therefore, there is a great demand to develop wearable flow sensors for noninvasive respiratory monitoring. We demonstrated the feasibility of using the as-fabricated hot film sensor for human respiration by affixing it to the upper lip (philtrum). Figure 5(a) shows that the sensor is flexible and hence wearable, which is casually affixed to the philtrum of a patient without causing any uncomfortable breathing. All experiments regarding measurement of human respiration were performed in compliance with the relevant laws and institutional guidelines and approved by the Human Research Ethics Committee (HREC) of Griffith University.

Zoom In Zoom Out Reset image size

Figure 5(b) shows periodic resistance change from the hot film sensor induced by the exhalation and inhalation of breath, with a high SNR. It is evident that exhalation brings a larger resistance change in comparison to inhalation, indicating a larger air flow rate passing over the sensor. The results suggest a strong feasibility of using the lightweight and biodegradable sensor for monitoring respiratory disease, sleep quality and other healthcare applications.