1. Introduction
In extreme environments, such as in aerospace applications [
1] or wind turbines [
2], the importance of accurate ice detection cannot be overstated. Ice sensors serve as the first line of defense against potentially catastrophic events or major damage caused by ice accretion on critical surfaces. The reliable and accurate detection of ice is also needed in standard environments, such as roads and bridges [
3] or power lines [
4], to ensure the desired level of safety. Several techniques for ice detection have been assessed in recent years, based on pneumatic, magnetostrictive, piezostrictive, ultrasonic, dielectric, acoustic, thermographic, microwave, and photonic sensing principles [
5,
6,
7,
8,
9,
10].
This large variety in ice sensor technologies provides the possibility of choosing the best solution for any given application, considering the specific requirements. However, each of the proposed technologies has advantages and drawbacks, and thus research is still needed both to optimize existing solutions and to find novel ones. Sensors based on physical contact are based on the influence of the ice on the surface acoustic waves [
11], on the electrical parameters [
3], on the latent energy [
5], on the heat transfer on temperature sensors made by thermocouples [
12], or by fiber Bragg gratings [
9,
13]. These sensors have the advantage of easier embedding but can lead to an underestimation or overestimation of the actual ice accumulation, being most highly sensitive to surface or near-surface contact. In addition, they are usually made of a sensing material that is different from the surrounding material, and this fact can lead to a thermal island effect, as indicated in [
3] with reference to road ice detection. Methods based on indirect measurements, such as, for instance, those based on the microwave response of planar split rings [
7] or on tomography imaging using ultrasonic waves [
14], have been shown to be effective but usually require complex hardware and software components, thus imposing severe constraints on implementation and significant maintenance costs.
In all the techniques cited here, the sensitivity of the sensors is strongly affected by the type and thickness of ice and by environmental conditions such as temperature fluctuations and humidity levels [
15,
16]. Finally, it must be pointed out that, in many applications, the chosen ice sensor technology is not necessarily the best one, being selected from among those compliant with the given standards. This is the case, for instance, for ice-sensors installed in aircraft which must fulfill the strict requirements imposed by aeronautical standards such as RTCA-DO-160G [
17].
For all the above reasons, the current research on ice sensors not only addresses the optimization of current technologies but also the creation of new solutions based on the use of novel materials, such as nanomaterials. Indeed, due to their specific features that may lead to unprecedented performance, these materials are definitely interesting candidates to realize novel sensing elements. Specifically, materials derived from graphene have undergone substantial research [
18,
19] that also suggested their use in sensing technologies [
20,
21,
22,
23,
24,
25], and more recently in ice-sensing applications [
26].
In addition, the same materials have been proposed to realize novel thermoelectrical actuators [
27,
28], with several interesting implications. For instance, the use of graphene strips as heating elements opens the route to novel approaches to de-icing based on the Joule effect. This would allow the use of such techniques for aircraft applications, where they are not currently used, due to the impact of the conventional heaters, in terms of weight for instance. The authors have investigated the electrothermal properties of graphene strips which may be proposed as heating elements for de-icing systems [
29,
30,
31]. Similar results have recently been assessed in the literature [
32,
33].
Starting from these considerations, the main motivation of this paper is to study the possibility of realizing an ice sensor with the same graphene strips that are used for de-icing. This possibility would open a route to the realization of an integrated system capable of sensing and removing ice.
With the scope of studying the feasibility of this idea, this paper analyzes industrial-grade graphene, which is not optimized for the specific purpose of sensing, but for that of heating. Indeed, the graphene and graphene-related materials that can be synthesized in a laboratory provide much better physical properties [
34,
35], but are still unfeasible for mass production due to the high costs associated with the fabrication technologies which are currently available. Alternative and less expensive materials are therefore being researched, such as nanocomposites with carbon-based reinforcements like graphene flakes, carbon nanotubes, buckyballs, and so forth. The work carried out in [
36] provides a thorough analysis of such inexpensive graphene variants, comparing prices and effectiveness. It is demonstrated that materials based on Graphene Nano-Platelets (GNPs) offer a suitable trade-off between good physical qualities, mass production, and affordable costs. The GNPs are examples of industrial graphene that exhibit good properties [
37] and that can be produced by using a number of industrially scalable methods, such as wet-jet milling [
38], microwave irradiation [
39], and liquid exfoliation [
40]. The latter technique is used by Nanesa [
41], who created the commercial GNP-based strips used here as an element of a capacitive sensor for ice detection.
The material production and characterization are briefly summarized in
Section 2. This section also contains the design and implementation of the capacitive sensor and its modeling in terms of an electrostatic model and a simplified circuital one. The experimental characterization of the sensor performance in the presence of ice is investigated and discussed in
Section 3, along with the identification of the model parameters. Conclusions are given in
Section 4.
2. Materials and Methods
2.1. GNP Strips Fabrication and Characterization
The films of industrial graphene analyzed in this paper are strips with a lateral dimension of 1 cm, lengths between 10 and 18 cm, and thicknesses ranging from 55 to 75 µm. These films were created by Nanesa [
41], using the following steps of a proprietary industrial process:
- (i)
Graphene nano-platelets (GNPs) are produced through the thermal expansion and liquid exfoliation of an affordable graphitic precursor, namely intercalated expandable graphite.
- (ii)
A blend is created by dispersing GNPs in either acetone or an aqueous solution, employing magnetic stirring, and concluding with a sonication step. If a polymeric binder is included, it is introduced during the sonication phase. In terms of mechanical properties, polyurethane (utilized in this study) or epoxy are identified as suitable binders for the objectives of this research.
- (iii)
The mixture is then sprayed at a controlled pressure (using the semiautomatic 3-axes pantograph Computer Numeric Control plotter EXTREMA, model Basic), to realize the GNP strips.
- (iv)
The GNP strips undergo a final treatment of calendaring (optionally followed by annealing), that compacts them and provides an optimized thickness/alignment ratio.
The size and thickness of GNPs are known to significantly influence the physical properties of the resulting composite material. For example, the impact of GNP thickness has been discussed in [
42], indicating that thinner GNPs are recommended to enhance overall mechanical and thermal performance. On the other hand, the analysis in [
43] focused on the effect of size/thickness aspect ratio on electrical conduction, revealing that the percolation threshold in the composite increased with higher aspect ratio values. Additionally, [
44] demonstrated that the electrical conductivity of the composite improved with increasing GNP size and surface area. Consequently, to ensure stability in the behavior of GNP films, it is crucial to evaluate a fabrication process capable of controlling GNP dimensions. As shown in the Scanning Electron Microscope picture displayed in
Figure 1a, the typical surface dimensions of a GNP are of the order of some tens of µm, with an average thickness of around 12 nm. Similar characteristics can be obtained by using alternative techniques suitable for industrial-scale production, such as wet-jet milling [
38] and microwave irradiation [
39].
Table 1 summarizes the characteristics of the industrial graphene strips analyzed in this paper (
Figure 1b).
The electro-thermal properties of these GNP strips are analyzed in [
29,
31].
The electrical resistivity exhibited by these strips is suitable for their use as heater elements based on the Joule effect, as shown in [
29]. Therefore, they are investigated with great interest in view of their potential use in industrial applications (e.g., to replace conventional ovens) and in aeronautical ones (e.g., in novel de-icing and anti-icing systems for aircraft wings). Note that a decreasing value of resistivity was observed with increasing temperatures, and thus these films qualify as Negative Temperature Coefficient (NTC) materials. This is an unusual behavior for electrical conducting materials, usually being found in non-conducting or semiconducting materials (such as silicon or germanium). As pointed out in [
30], this behavior is associated with different mechanisms related to the charge transport within a single GNP flake and between two adjacent flakes. As the temperature increases, the carrier mobility within a single GNP is reduced by a shorter mean free path, but is enhanced by a higher number of conducting channels [
45,
46]; when the second mechanism is dominant, the temperature increase improves the electrical transport. In addition, the transport between two adjacent flakes is associated with classical contact conduction (that worsens with higher temperature) and to quantum mechanisms such as tunneling and hopping effects, which provide higher mobility at higher temperature. For instance, in [
47] the electrical conductivity associated to the hopping and tunneling is reported to increase with the temperature
T as
.
The relationship between resistivity and temperature is well described by the standard linear law adopted for conventional materials like copper:
with
being the reference temperature, and
the Temperature Coefficient of the Resistance, TCR. The fitting parameters for the GNP strips and for copper are given in
Table 2: compared to copper, the GNP strips exhibit a higher conductivity (more than one order of magnitude), but a negative TCR.
The thermal properties of the GNP material analyzed in this paper are instead investigated in [
31];
Table 3 summarizes the estimated emissivity (ε) and thermal conductivity (
k). As expected, the thermal conductivity value of this industrial graphene is lower than the values expected for pure graphene (that can rise by up to thousands of W/mK, e.g., [
48]), but is still of interest, being comparable to those exhibited by conducting materials usually adopted for industrial thermal and electro-thermal applications (like copper, as reported in
Table 3). In [
29], an example of the use of these GNP strips as heater elements based on the Joule effect is provided, exploiting their good electrical and thermal properties.
2.2. ICE Sensor Operating Principle and Design
Keeping in mind the possibility of realizing heaters based on the GNP strips previously introduced, in the following, we investigate the idea of using the same strips to also realize an ice sensor, so that a potential device can be envisaged that is able to detect the ice and remove it.
Therefore, following the geometry suggested by the strips, the ice sensor here is designed as a planar capacitor with the upper face exposed to air, water, or ice, whose capacitance values vary as a consequence of the huge difference between their relative permittivity (dielectric constant), as reported in
Figure 2 [
49]. The relative permittivity of water and ice is significantly higher than air, and depends, in a different way, on the frequency at which the capacitor is working.
The idea developed here is of using each GNP strip as the central arm of a planar capacitor, whose cross section is depicted in
Figure 3 (“Graphene strip”). The external arm is provided by a conventional conductor, here chosen in copper (“External arm”, in orange). A guard ring of conventional conductor is also designed (“Guard ring”, in orange), that provides an additional degree of freedom, since it can be connected either to arm 1 or arm 2, thus changing the topology of the capacitor. The top surface is in contact with air or ice, depending on the operating conditions.
The design in
Figure 3 has been implemented in the multi-layered printed circuit board (PCB) in
Figure 4, where the horizontal tracts of the conductors 2 and 3 are realized by metallization in different layers, and vertical tracts are made by using vias. The green regions are instead made by FR4 dielectric. The PCB, with surface dimensions of 108 × 15 mm, has been designed for enabling the 4-probe techniques, with two amperometric cables (in orange and brown in
Figure 4) separated by the two voltmetric ones (in black and red in
Figure 4).
2.3. Electromagnetic and Circuital Models
In order to theoretically investigate the behavior of the sensor, the circuit model depicted in
Figure 5 was proposed. In this circuit, the GNP arm is modelled as a parallel R-C element, taking into account the high sensitivity of the complex permittivity of graphene to environmental conditions. Here, the resistance
also includes the contact term. The other arm, made of copper, is instead modelled only by the equivalent resistance of the arm itself,
, that also includes any contact term. The core of the sensor is given by two parallel capacitances:
, which is associated with the inner part inside the PCB (FR4 dielectric), and
, which is related to the displacement filed lines developing outside the capacitor. The element that is supposed to be mostly affected by the presence of air/ice is
, but an effect can be also observed
as a consequence of the varying temperature values, see (1).
The estimation of the bulk capacitance
+
in presence and absence of ice has been carried out by using a simple 2D model developed in COMSOL Multiphysics (Ver. 6.2), as shown in
Figure 6. As detailed in the Results section, the best option for the sensor sensitivity was found to be that of putting the guard to the same potential as the central graphene arm. Therefore, the sensor cross section described in the previous section has been further simplified by considering the graphene strip and the guard ring as a single conductor made of copper, with
(show in
Figure 6). In this way, this model does not contain the capacitance
The second arm is also made of conventional conductor with
, while the dielectric parts in the FR4 substrate have been assigned a value of
. The element “Air/Ice” in
Figure 6 is the area upon the sensor, and it represent both air (
) or ice (
) according to the operational conditions considered. Finally, between area 4 and the sensor itself, a thin “Plastic layer” (25 µm thickness) has been interposed (with
) to adhere to the real conditions in which the experiments have been carried out. Indeed, this plastic was needed to avoid the GNP strip being in direct contact with the ice.
The results of the solution to the electrostatic problem are displayed in
Figure 7, which shows the distribution of the electrical field (black arrows) and the electrical charge surface density (colored map). As shown in
Figure 7a, the field (and the charge density) are confined within the PCB structure when it is surrounded by the air. When the air is replaced by ice (
Figure 7b), the steep increase in the dielectric constant is responsible for the field spreading externally through the PCB (fringe fields), hence modifying the capacitance value. The electrical capacitance is then extracted from the solution to the electrostatic problem.
2.4. Experimental Setup
This section discusses the methodologies and the measurement process adopted in the experimental characterization of the ice sensor. The capacitance of the sensor has been evaluated by measuring its electrical impedance with the set-up conceptually described in
Figure 8: the sensor is connected to a
GW-INSTEK LCR-8101G (Gw Instek: Taipei, Taiwan) impedance analyzer, and a PC is used to drive the instrument via an RS-232C interface (sanwa supply: Okayama, Japan). Regarding the measurement process adopted, Impedance Spectroscopy has been used to investigate the response of the sensor to changes in frequency. In detail, the 20 Hz–1 MHz range was explored, sampling three points per decade in logarithmic steps (e.g., 1-2-5-10...). As shown in the set-up diagram, the connection between the instrument and the sensor has been realized by using a 4-probe configuration, consistent with the features of the PCB sensor (
Figure 4). The sensor can possibly be inserted in a climatic chamber (ACS DY110, Angelantoni: Massa Martana (PG), Italy) in order to control temperature and humidity.
The impedance was measured between the external arm of the planar capacitor (
Figure 3) and the graphene strip was set at the same potential as the guard ring. This configuration was the best one in terms of sensitivity, being associated with the smallest gap between the arms (equal to 0.1 mm). In line with the sensor model adopted in
Figure 5, an R-C parallel model has been chosen for the impedance analyzer to directly detect the capacitance value.
Prior to the measurement process, a calibration of the impedance analyzer was performed to exclude any contributions from the parasitic, capacitive, or inductive effects of the set-up. This phase is fundamental for this type of instrument and must be performed each time the instrument is used in a new environment, or if the test set is changed. The calibration consists of two phases, the so-called open circuit and short circuit ones: in the first, the instrument clips are spaced an equal distance apart from the normal test position; in the second, the instrument clips are connected by a short circuit.
In the measurement campaign performed, the operation of the sensor under different working and environmental conditions has been characterized. In detail: (i) at fixed or varying temperature and humidity; and (ii) in the absence or in the presence of different types of ice. To this end, to assess sensitivity to the environment, the sensor has been characterized inside the cited climatic chamber as capable of controlling the temperature and the percentage of humidity in the environment. The temperature values considered are: −40, −20, 0, 20 °C. While the humidity values selected are: 0, 10, 25, 50%.
Finally, three types of ice samples (clean ice) are considered, differing in their dimensions, as listed in
Table 4. The ice thickness values have been suggested for the specific use of the sensor in aerospace applications, where the maximum ice thickness is limited to 30 mm. Specifically, the sensor should be able to provide an alert signal to the control unit at the beginning of the ice formation during flight (few mm thickness). Instead, the values of the ice width are chosen to investigate different cases of the coverage of the planar capacitance: minimum (10 mm), medium (12 mm), and very large (28 mm).
In order to realize a metrologically robust characterization of the sensor, a stability analysis was performed. Specifically, in order to assess and estimate the ability of the tested sensor to respond more or less consistently to the same stimulus, 30 measurement tests were conducted for each considered working condition. In this way, it was possible to obtain an indication of sensor stability and repeatability.
In particular, since the causes of possible variations in terms of response can be many and varied (instrumentation, environmental perturbations, real variations in the measurand, etc.), the tests were not only repeated for each operating condition in the presence and absence of ice but also considering variations in environmental conditions. As shown in
Section 3, for each test, the response at each frequency was given in terms of the average value over the repeated measurements and the corresponding standard deviation, which is therefore representative of the stability of the sensor.
4. Conclusions
In this work, the potential use of a graphene-based planar capacitor as an ice detector has been successfully proven. A printed circuit board which hosts the graphene-nanoplatelets strip, realizing the planar capacitor, has been designed and developed, whose electrical impedance is measured in terms of the presence or absence of ice. The same design can be used to feed the graphene strip with an electrical current high enough to produce heat via the Joule effect; hence, an integrated system could be envisaged that uses the same element (a graphene strip) to detect and remove the ice.
The experimental results highlight the possibility of using the sensor as an ice detector based on the variation in its electrical capacitance in a low frequency range (up to 300 Hz), where the difference in terms of relative dielectric constant between air and ice is higher (1 to about 80). The repeatability and stability of the sensor’s response to environmental changes are analyzed, assessing a relationship between the working frequency range and the desired level of confidence. The detector is shown to work up to 200 Hz with a confidence level of 95.4%, and up to 50 Hz with a confidence level of 99.7%.
The results can be interpreted by using a circuit model that considers the effect of the internal (inside the PCB) and external capacitances, as well as the electrical resistance of the graphene strip. Despite its simplicity, the model is able to reproduce the experimental results with a good agreement.
Future work will address the sensitivity of the sensors to ice thickness and to different types of ice.