Humidity sensors are employed today in a wide range of applications, including environmental monitoring, automotive, industrial process, healthcare, agriculture, and increasing indoor air quality in smart buildings. Several kinds of humidity sensors are available based on different transduction principles, such as resistive, capacitive, optical, and surface acoustic wave [1
]. However, resistive-type sensors have the advantage to be cheaper and easier to read out over the other ones.
Typically, rigid substrates like ceramic, glass, or silicon are used as the fundamental building blocks of humidity sensors; but, recent advancements in the field of printed electronics show increased potential for the substitution of rigid substrates by flexible ones, since the latter potentially reduce the cost of sensors and offer good mechanical flexibility. Examples of flexible sensors integrating additional electronic functions like readout electronics [5
], thermal compensation systems [7
], and other sensors [8
] have opened a new route towards multi-functional sensors fabricated on flexible substrate. Despite that, silicon technology is still attractive for the fabrication of sensors due to its mass-production capability, its high degree of miniaturization resulting in high integration density, and, consequently its considerable cost reduction for sensor devices [10
]. Indeed, as given in the paper of Moore in 1965 [12
]: “With unit cost falling as the number of components per circuit rises”, the cost of one sensor must also decrease as more sensors are put on the substrate.
In addition, due to their very small size, silicon-based devices can be integrated with a flexible substrate. For example, a silicon strain sensor and multiplexed silicon non-volatile memory were transferred onto flexible substrate for wearable electronics [13
]. Miniaturization is then an important issue in printed electronics, which needs to be assessed to consider it a valuable alternative to silicon technology.
In the case of resistive-type sensors, the size of a device depends mainly on the surface area covered by the electrodes. Low-cost printing technologies such as ink-jet or screen-printing are forecasted to dominate the printed electronics era, since they allow high-volume production [15
]. However, for electrode designs, the line resolution/width achievable by printing technologies cannot reach micro-scale features generally higher than 100 μm, resulting in a large surface area. Thus, alternative technologies need to be considered to obtain miniaturized devices. Photolithography coming from expensive CMOS (Complementary Metal–Oxide–Semiconductor) technologies allows the design of micro-scale electrodes with high resolution in a large-scale manner, and this technology can be also employed for the mass production fabrication of sensors on flexible substrate [17
]. Nevertheless, chemical and baking steps are required in photolithography, which limits it to chemically resistant substrates such as polyimide.
Laser technology is gaining interest as another alternative micropatterning technique due to its high precision and the possibility to use it in open air without clean room facilities. This process was used for the fabrication of ozone sensors, and was compared with photolithography in [21
]. It was shown that this method can reach features of up to 60 μm. However, smaller features should be obtained to enhance its potential for sensor fabrication.
Furthermore, to take advantage of miniaturization, it is important to select an adequate sensing material. Among the various sensing materials, metal oxide nanomaterials possess good properties such as chemical and physical stability and high mechanical strength, and they have a high surface-to-volume ratio that makes them a perfect candidate for sensor applications. In recent years, TiO2
has received wide attention and has found applications in many promising areas, such as photovoltaics, photocatalysis, and sensors [22
]. Many examples of humidity sensors based on TiO2
can be found in the literature [27
]. The ability of the sensing material to be integrated in industrial production depends on the fabrication route used for its deposition and patterning. In the case of humidity sensors, the most standard methods to deposit TiO2
are the spin-coating, dip coating, or microdroper processes [27
], which are suitable for prototypes, but cannot be considered for large-scale process. In fabrication route selection, several factors must be considered, such as costs, throughput, and reproducibility, and the procedure ought to be compatible with the substrate, especially in terms of temperature. The screen-printing process meets all of the above, which has been demonstrated in [32
], where the screen-printing of TiO2
nanomaterials was used for dye-sensitized solar cells and electrodes.
In this context, the aim of this work was to introduce a cost-efficient and low-temperature procedure that allows the large scale fabrication of humidity sensors on flexible PET(Poly-Ethylene Terephthalate) substrate. The undertaken multidisciplinary approach combines expertise in materials science and chemistry, and fabrication processes and sensor characterization, aiming to present comprehensive bottom-up research in the field of flexible electronics and sensors. One of the main goals of the conducted research is not only to introduce an innovative technology process for the fabrication of sensor devices, but also to provide a proof-of-concept through extensive mechanical testing and a humidity response characterization of the fabricated miniaturized sensors.
3. Results and Discussion
Using the process described above, the large-scale fabrication of humidity sensors has been successfully achieved. Figure 3
a shows matrices of 3 × 3 sensors printed on PET substrate. Figure 3
b depicts an SEM image of the ablated interdigitated electrodes on the PET substrate. This image indicates that the surfaces of the electrodes’ structures subjected to pulse ablation are highly consistent and spatially well-resolved. An individual digit of an electrode is 700 μm long and 55 μm wide, and it is separated by a gap of 40 μm to the next digit. Here, small electrode geometry was obtained compared to the standard printed sensors, which are generally above 100 μm.
c shows an optical picture of the interdigitated electrodes covered by the screen-printed TiO2
film. The resulting TiO2
film is well aligned with the film, covering perfectly the surface of the electrodes. This perfect alignment of the screen-printed film is also confirmed at larger scale, as shown in Figure 3
a. The screen-printed TiO2
nanoparticle-based film, defining the active area of the humidity sensor, forms a rectangle of 1 mm width and 1.5 mm length (Figure 3
The thickness of the TiO2
can be controlled by the number of printed layers. Indeed, Figure 4
a,b represents a cross-section of the TiO2
film after the printing of four layers and six layers, where the thicknesses were measured to be approximately 18 μm and 25 μm, respectively. The evolution of the thickness as function of the number of printed layers is summarized in Figure 4
The sensing properties are based on the change in the electrical conductance of the sensitive layer with the adsorbed water, which depends on the surface characteristics of the film. The surface morphology of the TiO2
film was investigated using SEM and AFM techniques. Figure 5
a shows the SEM image of a TiO2
layer with high magnification, where the porous structure of the TiO2
film can be observed, which is favorable for water vapor absorption due to the large surface area [39
]. Figure 5
b shows an AFM image of the TiO2
film, where the spherical structure of the TiO2
nanoparticles with a grain size of less than 100 nm, and the porosity of the film, can be clearly observed. Also, we can see in Figure 5
d that the film formed by TiO2
after screen-printing is quite uniform and homogeneous along the sensor. Next, an energy dispersive X-ray spectrometer (EDX) was employed to study the structural composition of the printed titanium dioxide film. Figure 5
c shows the EDX spectrum of the selected area shown in Figure 5
d, where the main peaks correspond to titanium and oxygen, indicating that the surface is well covered with TiO2
. The presence of carbon can be clearly observed, and it has been attributed to organic components of the functional paste. Note that, among them, an important component is the binder (hydroxypropylmethyl cellulose), since it assures a strong binding between TiO2
nanoparticles and a good adhesion of the TiO2
film with the substrate, improving the stability of the TiO2
film should possess good electrical performances to allow for precise and stable resistance measurements. Then, the electrical characteristics of the printed structures should be investigated as well. Current against voltage (I–V) measurements were obtained on sensors printed with one, two, four, and six layers by sweeping the applied voltage from −5 to 5 V. A typical current reading, as shown in Figure 6
a, clearly demonstrates that the TiO2
film provides a connecting Ohmic electrical contact between pairs of Au electrodes with constant resistance over the supply voltages. That means that a low voltage operation does not hinder the sensitivity, which is essential for low power operation. On the other hand, the conductance of the printed layer should be high enough to be measurable without a high-precision instrument. Figure 6
a highlights the influence of the number of printed layers on the electrical performance of the film. For one printed layer, the variation in current is about 4 nA at 5 V bias, which reveals a poor conductance of the TiO2
film. Generally, post-processing steps such as annealing are required to improve the conductivity of the material, leading to an increase in energy consumption and producing additional cost. In this work, in order to develop a low cost and low-temperature process adapted to flexible substrates, we have formulated a recipe for a functional paste that can be used for the printing of several TiO2
layers, and that can preserve the original material physical and transport properties. Indeed, in Figure 6
a, we can see that the sensor current increased as the number of successive printed layers is increased, due to the added TiO2
nanopartices (NPs). This leads to a drop of resistance from about 1 GΩ to 266 MΩ (Figure 6
b). With six printed layers, the resulting resistance (266 MΩ) is low enough to make the sensor compatible with a simple and low-powered electronic scheme, such as a Wheatstone bridge, for the signal read-out.
Afterwards, humidity sensing performance was evaluated using the following equation to define the sensors’ response:
are the resistances at a given humidity level, and Rini
is the resistance at zero humidity used as a baseline.
Reproducibility is one of the first requirements for a sensor’s application. Typically, it is defined as a condition wherein the sensors exhibit multiple vapor adsorptions/desorption behaviors under cyclic operating conditions. In order to examine this, the humidity environments of the sensor were sequentially changed from 0 to 70% in periods of 30 min for several sorption and desorption processes. Figure 7
reveals that during the fourth response/recovery cycles, the sensor response shows a good sensing repeatability during cycling tests, which represents another advantage for its potential application. However, a drift of about 8% in the initial value of the response can be observed in Figure 7
. This was attributed to residual moisture that had accumulated in the TiO2
film after several sorption and desorption processes. Indeed, the highly porous structure of the TiO2
film highlighted in Figure 5
a,b can easily trap moisture, producing the observed drift in the measurements.
Next, in order to study further the characteristics of our humidity sensors, it is important to investigate the sensors’ response at different humidity levels.
a shows the sensors’ response for several dynamic cycles of absorption/desorption at humidity levels varying from 0 to 70%. It is important to mention that low relative humidity levels were detected with designed miniaturized sensors, introducing a significant improvement in comparison to the other flexible humidity sensors found in the literature [40
]. This can be attributed to the highly porous surface of the printed TiO2
film, which results in a large surface area providing more surface active sites and paths for water molecule adsorption and diffusion [39
b presents the sensors’ response as a function of humidity level, where it can be observed that sensor response is linearly proportional to the relative humidity level, implying a more precise measurement at a low humidity level and simple calibration, which are important parameters for potential sensor application.
The response and recovery times are also very important factors to determine the performance of humidity sensors, and they also need to be evaluated. The response time is the time taken by a sensor to achieve 90% of the maximum response, and the recovery time is the time needed for the senor to drop to 10% of its initial response. Both parameters were calculated from a long cycle time (30 min), which was used to ensure that the device response reached its saturated limit without any noticeable drift.
c shows the response and recovery time as function of the relative humidity level including the equilibration time of water vapor inside the test chamber. In this Figure, it can be seen that the response and the recovery times are fast in a range from 5 to 40 % RH, varying between 40 s and 3 min for the response times, and about 50 s concerning the recovery times. However, the response and recovery times become much slower at higher RH levels (>50% RH). This could be attributed to the humidity sensing mechanism. In fact, at low RH, the decrease of resistance is mainly due to the chemisorbtion of water molecules by the active sites available on the TiO2
surface. In that case, the dominant charge transport mechanism is electronic transport, which is much faster than proton conduction. On the other hand, the subsequent layer of the water molecule is generally physisorbed by double hydrogen bonding with the hydroxyl groups formed on the previous water layer [30
]. Afterwards, successive physisorbed water layers are accumulated on the surface of the TiO2
film as the humidity level increases. In that case, the proton conduction mechanism becomes dominant, which could explain the slower response times for high humidity levels.
Mechanical stability is essential to flexible electronic devices, especially for applications where high stability over the mechanical deformation is required, such as wearable electronic and smart food packaging. Therefore, the influence of the mechanical strain on the electrical behavior of the devices has to be explored. To do so, bending experiments were performed by attaching the flexible sensors to a cylinder (Figure 9
a) and the curvature angle was calculated to be approximately 100°, as depicted in Figure 9
c shows the resistance change during several bending and return to flat position cycles in periods of 5 min. It can be seen that the resistance decreased during the bending experiments, but it retrieved its initial value quickly after the mechanical excitation, i.e., after a relaxation time of about 1 min.
c exhibits the resistance as a function of the number of bending cycles. The device showed only a slight decrease in resistance (2.3% of the initial value) after five cycles. Moreover, a scanning electron microscopy (SEM) analysis revealed no morphology change of the film caused by mechanical bending (Figure 9
To validate the stable sensing operation under mechanical deformation, humidity measurements were performed when the sensor was in a flat position and bended at 100°.
At each indicated position, the sensor was exposed to RH varying from 0 to 35% RH in periods of 10 min. Note that the measurements in a bended position were performed 2 min after bending the sensor in order to leave it enough time to recover its initial resistance value. Figure 9
e shows that the response of the sensor when it was bended increased by less than 3% from that measured when it was in a flat position. It can be concluded that the sensors’ response showed negligible effect over the mechanical strain.
The obtained results indicate that the TiO2 paste formulation offers high mechanical stability for a TiO2-sensitive layer when it is printed on a plastic substrate, which consequently allows the devices to be used for flexible sensor applications.