Triply Periodic Minimal Surfaces Mullite Structures for Humidity Detection

Abstract

Three-dimensional-printed complex mullite structures based on triply periodic minimal surfaces (TPMSs, namely, Schwartz and Gyroid) with two different thicknesses (Schwartz 1 and Gyroid 1–4 mm, Schwartz 2 and Gyroid 2–6 mm) were fabricated and tested as humidity sensors. The samples were sintered at 1450 °C and tested in the range from 0% to 89% relative humidity (RH) at room temperature to evaluate the effect of geometry and thickness on humidity sensitivity. After water vapor exposure at room temperature, the response was 2.84 under 89 RH% for the Schwartz 1 structure (1.36 for the Schwartz 2 structure) and 1.21 for the Gyroid 1 structure (7.00 for the Gyroid 2 structure). The results showed that, at 89% RH, the best response of the sensors was achieved for the Gyroid 2 structure. Sensors exhibit good repeatability, and there was no interference in the presence of other gases.

1. Introduction

The measurement and control of humidity are of great importance not only for human comfort but also for many different sectors of industry and technology, such as environmental monitoring, weather forecasting, agriculture, manufacturing, aerospace, and medical applications [1,2,3,4,5]. Thus, there is a constant need to develop new humidity sensors with high selectivity, sensitivity, and fast response times. In this work, resistive humidity sensors were chosen because of their simple manufacturing process and low production cost.

According to the measuring signal, the humidity sensors can be of the following types: capacitive, resistive, impedance, optical, frequency (quartz crystal microbalance (QCM), surface acoustic wave (SAW)), emerging electrochemical sensors, and ion gradient self-powered humidity sensors [6,7,8,9,10]. Resistive humidity sensors are based on the electrical resistivity changes in semiconductors, conductive polymers, or composites because of water molecule adsorption. At present, many humidity sensing ceramic materials, such as perovskites (NdCoO₃, BaTiO₃, etc.) [11], metal oxides (Fe₂O₃, CuO, SnO₂, ZnO, etc.) [12,13,14,15,16,17], spinels (CoCr₂O₄, MgFe₂O₄, etc.) [18,19,20], halloysite nanotubes [21], attapulgite [22], and sepiolite nanofibers [23], have been used for the resistive humidity sensors.

Triply periodic minimal surface (TPMS) structures have zero curvature at every point on the surface and, due to unique topologies, have attracted significant interest [24,25]. TPMS porous structures have some advantages in comparison with other structures: their geometry can be exactly expressed through mathematical functions. By changing the parameters of the mathematical functions, it is possible to control the size, shape, and distribution of pores and porosity. Moreover, TPMS structures have smooth curvatures and joints, which are important advantages for different applications.

Mullite TPMS structures have a high specific surface area and high thermal stability. However, it is difficult to produce mullite TPMS structures by traditional methods due to their complex shape. Recently, three-dimensional (3D) printing technology has been successfully used to produce ceramics with high-quality patterned structures [26,27]. Among them are inkjet printing, selective laser melting (SLM), and digital light processing (DLP). The latter method offers the possibility to produce complex geometric shapes based on a 3D CAD model. This permits precise control of the shape, size, and distribution of the pores in the printed structure.

TPMS 3D printed structures are used only in a few cases for sensor applications. Zhou et al. prepared Ti-doped SiOC gyroid structures by vat photopolymerization [28]. It had high selectivity towards n-butanol and maintained gas sensitivity at 500 °C. These structures were also used as piezoresistive sensors with good durability and reliability, which exhibited no performance degradation after 500 compressing cycles. Peng et al. fabricated piezoresistive sensors of polydimethylsiloxane with complex internal pores based on 3D-printed sacrificial models [29]. Among the various structures, the diamond-based ones showed the highest responsive strain of 61%, with a resistance response value of 0.97. Fortunato et al. created a piezoresistive pressure sensor using Grid and TPMS structures of 3D porous flexible open-cell foams [30]. These foams were fabricated by the 3D printing of sacrificial templates, followed by leaching, and infiltration with various low-viscosity polymers, and, in the final stage, the pores were coated with graphene nanoplatelets. The maximum sensitivity of these sensors for pressures below 10 kPa was 0.088 kPa⁻¹ (for pressures of 80 kPa–0.24 kPa⁻¹). Imanian et al. produced wearable, simple, and low-cost piezoresistive sensors based on TPMS for biomonitoring applications [31]. The sensors demonstrated good durability under cyclic compression forces and good stability. Davoodi et al. printed sensors based on interconnected TPMS structures from surface-doped porous silicone [32]. The piezoresistive sensor signals under repeated mechanical loads were stable and had high stability and cyclic repeatability. The humidity sensitivity of the surface-embedded graphene sensors, as compared to the surface-deposited graphene sensors, was found to be insignificant. Han et al. created a 3D hybridized conductive network structure utilizing transition metal carbides/sodium alginate and multi-walled carbon nanotubes, by means of parametric structural design, and using the fused deposition molding printing technique [33]. In the wide operating range (up to 10 MPa), the sensors had high sensitivity (4.67 MPa⁻¹–7.03 MPa⁻¹), long-term fatigue resistance (more than 10,000 cycles), and fast response and recovery times (200 ms and 300 ms).

In this research, the photocurable mullite ceramic slurry was prepared first. Two types of mullite TPMS structures (Schwartz and Gyroid structures) with two different thicknesses were produced by DLP technology. Sensors were then sintered at 1450 °C. Their electrical response as a function of RH (from 0% to 89% RH at RT) and with respect to interfering gases (NO₂, CO₂, NH₃, and CH₄) was studied for the first time.

2. Materials and Methods

2.1. Fabrication of Ceramic Substrates

2.1.1. Mullite Powders Characterization

Ceramic substrates were additively manufactured using mullite (3Al₂O₃⋅2SiO₂), specifically, two commercial powders, JMS-70 (by Mf Jingxin New Material Co., Zhenwu Town, Jiangdu District, Yangzhou City, Jiangsu Province, China) [34] and MC MM (by MC CERATECH CORP, Aichi, Japan) [35]. Both the raw materials were kindly supplied by Saint-Gobain Research Provence (France) and derived by a sintering process. In the present work, the two commercial powders will be distinguished based on their finer and coarser particle sizes, thus indicated, respectively, as Mf (fine mullite, JMS-70) and Mc (coarse mullite, MC MM). Our previous work [36] contains all the details about powder characterization in terms of particle size distribution (determined by laser granulometry), phase indexation (X-ray diffraction), and chemical composition (X-ray fluorescence).

An amount of 150 g of as-received powders was poured in a plastic jar with zirconia balls (d = 2 mm), water, and magnesium nitrate hexahydrate (Mg(NO₃)₂⋅6H₂O, supplied by Sigma-Aldrich, 99.0% purity), as a sintering aid; the proportions with respect to the mullite are shown in

Table 1. Ball-milling proportions in weight.

Component	Ratio Respect to Mullite
MgO	1:99
Zirconia spheres	6:1
Water	2:1

. The amount of Mg(NO₃)₂⋅6H₂O was calculated to obtain 1 wt% of MgO by thermal decomposition. Ball-milling was performed for 48 h using a Labinco roller bench (Labinco BV, Breda, The Netherlands) and setting a rotational speed of 1.4 rpm. After drying, the ball-milled powders were calcined at 650 °C for 1 h with a 10 °C/min heating ramp [37].

After ball-milling and calcination, the Mc and Mf powders showed a D₅₀ (average particle size) of 9.2 and 4.4 μm, respectively (

Table 2. Average particle size (D₅₀) values for Mf and Mc mullite powders, open porosity, total pore area, and average pore diameter determined by mercury intrusion porosimetry (MIP) for Mf and Mc pressed pellets sintered at 1450 °C.

Powder	D50, μm	Open Porosity, %	Total Pore Area, m2/g	Average Pore Diameter, μm
Mf	4.4	32.0	1.0	0.54
Mc	9.2	29.1	0.5	0.87

).

The pressed pellets fired at 1450 °C also evidenced differences in terms of open porosity, total pore area, and average pore diameter between the two mullites (Table 2): the Mf samples exhibit a slightly higher open porosity percentage, and a doubled total pore area compared to the Mc ones. On the other side, pores with a larger size are present in the Mc pellets. In fact, despite its coarser size, the Mc powder has a higher iron oxide content (0.63 wt% for Mc vs 0.33 wt% for Mf, determined by XRF), and, as explained in a previous work [36], this can explain the slightly higher densification.

The Mc and Mf powders were also characterized by XRD analysis: both patterns showed the presence of mullite peaks (main phase, JCPDS file number 15-0776), as well as of free aluminum oxide (α-Al₂O₃, hexagonal phase, JCPFD file 96-900-7499). In addition, no relevant differences between the patterns of the as-received, ball-milled, and calcined powders were evidenced.

2.1.2. Mullite Slurry Elaboration

An already established protocol [35,38] was used to obtain two DLP printable ceramic slurries: a photosensitive commercial resin (Blank resin C, by Admatec Europe BV, Alkmaar, The Netherlands), containing acrylate-based monomers and diphenyl(2,4,6-trimethylbenzoyl) phosphine oxide as a photo-initiator, was mixed with the two different mullite powders. A mullite solid loading of 69 wt% was reached by adding 5 wt% of a dispersant (disperbyk-103, BYK Chemie, Wesel, Germany). After 45 min of mechanical stirring and 6 h homogenization in a planetary ball milling at 350 rpm with agate spheres of 10 mm diameter (Fritsch Pulverisette, Fritsch GmbH, Idar-Oberstein, Germany), a rotary pump was used to degass the slurries for 30 min. A rotational rheometer (Kinexus Pro+, Netzsch Gerätebau GmbH, Selb, Germany) was used to perform rheological tests at 25 °C, with stainless steel parallel plates of 20 mm in diameter at 1 mm of distance; a shear rate range of 0.1–1000 s⁻¹ was considered.

2.1.3. Digital Light Processing (DLP) of TPMS Mullite Architectures

Mullite macroporous scaffolds were manufactured by DLP using an Admaflex 130 printer (by Admatec Europe BV, Alkmaar, The Netherlands), as in previous works [36,38]. A 125 μm doctor blade spreads uniformly the ceramic slurry on a moving tape and, then, a UV light with a 405 nm wavelength photopolymerizes the resin layer-by-layer on a moving platform. Printing parameters were optimized to print porous TPMS architectures with both Mc and Mf mullite slurries: respectively, 1255 and 683.75 mW/cm² as LED power, and 3.8 and 3.0 s as exposure time were set. In both cases, a layer thickness of 30 mm gave good results in terms of adhesion between the different printed layers, and a delay of 6 s before exposure guaranteed the removal of air bubbles from the slurry. Two TPMS architectures were successfully printed: Schwartz Primitive and Gyroid. The .stl files of the 3D printed designs were modelled using Blender (by Blender Foundation, Amsterdam, The Netherlands) and AutoCAD 2018 software (Autodesk, San Francisco, CA, USA). The nominal specifications of the monoliths manufactured for sensing tests were collected from the modelling software. A nominal diameter and height of 2.1 cm and ca. 0.5 cm, respectively, were set. From the nominal surface area (SA) and volume (V), the surface-to-volume ratio was calculated as SA/V. Furthermore, the surface area was related to the weight of ceramic structures to calculate the specific surface area (SSA). The void percentage φ was determined in accordance with Equation (1):

$$\mathrm{\phi }(\%)={\displaystyle \frac{{\mathrm{V}}_{\mathrm{void}}}{{\mathrm{V}}_{\mathrm{total}}}}\times 100=\left[1-{\displaystyle \frac{{\mathrm{V}}_{\mathrm{lattice}}}{{\mathrm{V}}_{\mathrm{dense}}}}\right]\times 100$$

(1)

where V_lattice and V_dense are, respectively, the volume of the lattice structure and the volume of a dense cylinder with the same external dimensions.

2.1.4. Post-Processing, Sintering and Characterization

DLP printed monoliths were submitted to post-printing treatments: first, water debinding at 30 °C was performed to remove the hydro-soluble components of the resin and the uncured residual slurry; then, samples were dried in air and then in oven at 70 °C overnight. Thermal treatments were performed under air atmosphere in an electric furnace (Nabertherm, Nabertherm GmbH, Lilienthal, Germany), following a cycle previously optimized [36,38] (Figure 1): a slow thermal debinding up to 1000 °C allows to avoid crack formation during the polymeric component decomposition, while the sintering was completed at 1450 °C, maintaining a certain residual porosity in the mullite structure. Monoliths were then characterized in terms of ρ (g/cm³), geometrical density, determined by mass and geometric measurements, and by buoyancy method, following Archimedes’ principle (Density Determination Kit, Sartorius YDK01, Göttingen, Germany). The measurements were related to the mullite theoretical density of 3.17 g/cm³.

The sintering temperature is a crucial parameter affecting, on one side, the microstructural porosity and, on the other side, the monolith’s structural resistance. As explained in detail in a previous work [36], the sintering was performed up to 1450 °C in order to balance both these aspects: a lower temperature can increase the microstructural permeability, but does not guarantee proper handling for geometrically complex monoliths. The FE-SEM micrographs of 3D printed pellets made of Mc and Mf mullite are presented in Figure 2A and B, respectively. Samples’ fracture surfaces were observed after polishing to better highlight the microstructure resulting from the additive manufacturing shaping. Indeed, printing layers are displayed by micrographs at lower magnification and appear more clearly for the Mf sample; they show a thickness comparable with the nominal layer thickness set for the printing jobs (30 µm) [36]. For both samples, it is evident that the full densification is not reached and interparticle porosities occur, in agreement with MIP measurements. In particular, the microstructures shown by the micrographs at higher magnification highlight the presence of bigger pores in the Mc samples in comparison to the smaller porosities in the Mf microstructure. Thus, FE-SEM analysis coherently reflects the average pore diameters measured by MIP (Table 2). Differences between the Mf and Mc monoliths in terms of open porosity, total pore area, and average pore diameter are significant but not relevant enough to cause a difference in terms of permeability. Thus, tortuosity and pore connectivity are mainly influenced by the macro-architecture. Further and specific tests are needed for the evaluation of the gas permeability behavior in the same flow conditions operated for the gas sensing experiments performed in the current work. However, similar architectures were analyzed in terms of Darcy’s permeability coefficient in a previous work [39], which did not highlight relevant differences between Schwartz Primitive and gyroid designs in the investigated flow rate range.

2.2. Fabrication and Measurement of Humidity Sensors

The electrodes were made of an aluminum tape (commercial tape).

The sensor’s response to humidity was studied in a laboratory system in the 0–90% RH range at room temperature (23 °C), under an air flow rate of 320 mL/min. The RH was increased progressively by steps, each one 20 min. In this system (Figure 3), the air flow (Siad, San Mauro Torinese, Italy, research grade) was controlled by mass flows (driven by a mass flow controller: Teledyne 4000; MF302, Teledyne Hastings, USA;) and divided into two flows. The first one passes through the water bubbler, generating a saturated humid flow, whereas the second one was anhydrous. Then, both flows were combined. A commercial probe (Delta Ohm DO9406, Caselle di Selvazzano, Italy; accuracy: ±0.1% in the 0–100% RH range and between 20 and 250 °C) for humidity and temperature measurements was used as reference for the determination of RH and temperature values inside the measurement chamber [40]. An LCR meter (Hioki 3533-01, Ueda, Japan) was used to measure the sensors’ impedance phase and modulus during tests under a dynamic flow. The applied voltage was set to 1 V to avoid water electrolysis during the measurements in water vapor presence, while the frequency was fixed to 1 kHz, a much higher frequency than the network one (50–60 Hz), to limit any potential electric perturbation.

Cross-sensitivity tests were also conducted for different gases, like NO₂ (2.5 ppm in the air), CO₂ (500 ppm in the air), CH₄ (100 ppm in the air), and NH₃ (44 ppm in the air), under the same flow rate. All gases were research grade and were provided by Siad, San Mauro Torinese, Italy.

The response of the sensor (R) was determined in accordance with Equation (2):

$$R={\displaystyle \raisebox{1ex}{$Z_0$}\!\left/ \!\raisebox{-1ex}{$Z_g$}\right.}$$

(2)

where Z₀ and Z_g are, respectively, the impedance modulus under dry air and the impedance modulus under humid air of sensor.

The response time, as well as the recovery time, were calculated as the time required by a sensor to achieve 90% of the maximum impedance variation in the presence of humid air and the time required for a sensor to reach 90% of the maximum impedance change in the case of gas desorption.

3. Results and Discussion

3.1. Sensors Characterization

Three-dimensional printed mullite sensor structures with the following architectures were prepared: Schwartz with thicknesses of 4 (S1) and 6 mm (S2), and Gyroid with thicknesses of 4 (G1) and 6 mm (G2). Images of the printed sensors are reported in Figure 4: (a)—Schwartz, (b)—Gyroid; (c)—schematic illustration of the sensor structure with aluminum tape as upper and lower electrodes.

3.2. Humidity-Sensing Properties

The sensor response, in terms of impedance variation of printed sensors S1, S2, G1, and G2 under 89% RH, is plotted in Figure 5. Under dry air, the initial impedance values were 75 MΩ for S1, 132 MΩ for S2, 96 MΩ for G1, and 101 MΩ for G2. After water vapor exposure, the impedance value under 89 RH% dropped down to 26 MΩ for S1, 98 MΩ for S2, 79 MΩ for G1, and 14 MΩ for G2.

In the next stage, the sensor response dependence of Schwartz and Gyroid structure on the thickness of the sensing layer was investigated. Figure 6a,b shows the sensor S1 and S2 response variations. After water vapor exposure, the sensor response values were 1.18 for S1 under 57 RH%, 1.33 for S1, and 1.03 for S2 under 67 RH%. Nevertheless, under 77 RH%, the sensor response values were 1.57 for S1 and 1.1 for S2, and, in the presence of 89 RH%, they reached 2.84 for S1 and 1.36 for S2.

Figure 7a,b shows that, at 57% RH, the sensor response was 1.02 for G1 and 1.22 for G2; at 67% RH, it was 1.05 for G1 and 1.59 for G2. When the humidity increased to 77% RH, the sensor response values enhanced to 1.1 for G1 and 2.33 for G2. Furthermore, under 89% RH, the response values peaked at 1.21 for G1 and 7 for G2. It was found that, as the RH% increased, the impedance value of all sensors decreased. The sensors had high resistance values under low RH values because only a limited amount of water molecules were adsorbed. In contrast, the sensors adsorbed more water molecules under high RH values, causing their impedance to decrease significantly.

Comparing the performance of the sensors under 89% humidity (see Figure 8), S1 shows a 2.1 times higher sensor response as compared to S2 (2.84 and 1.36, respectively), whereas G2 shows a 5.8 times higher sensor response as compared to G1 (7 and 1.21, respectively). The sensors also had negligible hysteresis within the measurement error.

When looking at the response of sensors S2 and G1 (Figure 8), it can be noticed that there is apparently only a limited or no influence of the height and of the type of structure. On the contrary, the responses of sensors S1 and G2 are higher with respect to the previous ones, highlighting the importance of the mullite powder (Mf in this case).

Table 3. The sensor’s response and response/recovery times at different humidity levels associated with 3D printed mullite sensors.

Humidity	Sensor Response (R = Zo/Zg)				Response Time, s				Recovery Time, s
	S1	S2	G1	G2	S1	S2	G1	G2	S1	S2	G1	G2
40%	1.04			1.03
57%	1.18		1.02	1.22	399		550	454	31		112	44
67%	1.33	1.03	1.05	1.59	518	615	625	596	52	98	176	60
77%	1.57	1.10	1.10	2.33	522	712	526	600	56	109	127	77
89%	2.84	1.36	1.21	7.00	492	738	755	501	98	138	136	120

and Figure 9 summarize the sensors’ response and response/recovery times at different humidity levels of the sensors S1, S2, G1, and G2. When the atmosphere switches from 89 RH%, and vice versa, the response and recovery times of S2 (738 and 138 s, respectively) are longer than those of S1 (492 and 98 s, respectively). At the same time, the response and recovery times of G1 (755 and 136 s, respectively) are longer than those of G2 (501 and 120 s, respectively). These differences can be explained by the mullite architecture and pore size. In the case of the S2 and G1 structures, the lower sensor response and longer response/recovery times can be attributed to the difference in porosity between the printed mullite samples. The sintered mullite structures that were produced from the Mc powder have less open porosity (29% for Mc vs 32% for Mf), a smaller total pore area (0.5 m²/g for Mc vs 1 m²/g for Mf), and a bigger average pore diameter (0.87 μm for Mc vs 0.54 μm for Mf) (Table 2). The presence of capillary pores allows water condensation according to the Kelvin equation [3]:

$$r_K={\displaystyle \frac{2\gamma M}{\rho RTln\left(\frac{P_s}{P}\right)}}$$

(3)

where r_K is the Kelvin radius, P is the water vapor pressure, Ps is the water vapor pressure at saturation, γ, ρ, and M are the surface tension, density, and molar weight of water. Water condensation occurs in all pores with a radius up to r_K at a certain temperature and water vapor pressure. The smaller the Kelvin radius, the more easily condensation happens. Then, the Mf samples with smaller pores favored water condensation, and this led to a greater drop in the impedance values.

Long response times are usually due to the difficult access of the sensing material surface, due to the tortuosity of pores, for the target gas molecules, while a stronger binding of gas molecules on the same surface is responsible for long recovery times. In this case, again, S2 and G1 sensors generally showed worse performances (Table 3). This was probably due to a more tortuous path of water molecules to adsorb on the Mc mullite pores with respect to the Mf ones.

Repeatability is another important characteristic for evaluating the performance of the sensors. In Figure 10, three consecutive measurements on S1 and G2 sensors after three pulses under 89% RH, starting from dry air, are displayed. Both sensors showed good repeatability. The sensor response for S1 was 2.84, 2.86, and 2.71 for the first, second, and third cycles, respectively (for G2—7.00, 7.34, and 7.80, respectively).

The humidity sensing mechanism of the 3D printed mullite sensor is based on the surface protonic conduction. After exposure to the humidity, the 3D-printed mullite sensors’ impedance changes are influenced by the amount of adsorbed water molecules. With the increase in the humidity level, the impedance value decreases due to further water molecule adsorption [41,42].

In Ref. [43], copper ferrite-yttrium oxide nanocomposite humidity sensors were investigated. At low humidity levels, OH⁻ ions were chemisorbed onto the positively charged metal cations of the sensing material and formed a first chemisorbed layer. As the humidity level increases, each water molecule binds to two neighbouring hydroxyl groups through hydrogen bonds, forming the first physisorbed layer. As relative humidity increases further in the second layer, H₃O⁺ is formed through those molecules that are singly bound to the underlying layer. Then, the hydronium group releases a proton to a neighboring water molecule, which accepts it while releasing another proton, and so on. This process is called a Grotthuss chain reaction [44].

$$2{\mathrm{H}}_2\mathrm{O} \leftrightarrow {\mathrm{H}}_3{\mathrm{O}}^{+}+\mathrm{O}{\mathrm{H}}^{-}$$

(4)

$${\mathrm{H}}_{2}O+{\mathrm{H}}_3{\mathrm{O}}^{+} \leftrightarrow {\mathrm{H}}_3{\mathrm{O}}^{+}+{\mathrm{H}}_2\mathrm{O}$$

(5)

At higher humidity levels, the water molecules condense in pores, and electrolytic conduction can take place at the same time [3].

Tests were also performed under methane 100 ppm, nitrogenous oxide 2.5 ppm, carbon dioxide 500 ppm, and ammonia 44 ppm at room temperature. The exposure time was 5 min, and no interferences towards the investigated gases were noticed.

Table 4. Comparison of mullite sensor’s performance with results described in recent literature data on ceramic oxide humidity sensors.

Material	Sensor Response, R = Zo/Zg	Response Time, s	Recovery Time, s	Reference
CuFe2O4-Y2O3 nanocomposite	4895 at 97% RH	9	23	[43]
(Ca1-xCux)HPO4.nH2O x = 0 to 1	51,250 at 90% RH	29	283	[45]
Nanosize α-Fe2O3 nanoparticles	48,569 at 95% RH	9	4	[12]
GdAlO3	8000 at 97% RH	45	60	[46]
CuO-ZnO nanostructure	under (25–90)% RH	29	16	[14]
SnO2 thin film	3.1 at 95% RH	84	576	[15]
2D MoO3	4024 at 75% RH	8	40	[47]
1% (La, Ce, Pr, Nd and Sm):ZnO nanocomposites	under (22–97)% RH	62–89	∼300	[16]
Perovskite CsPbBr3-Fe quantum dots	1.1 at 70% RH	38	38	[48]
SnO2/TiO2 modified with Mo and Zn	∼ 100 at 85% RH	18	27	[49]
Porous SnO2/MCM-48	105 at 98% RH	9	12	[50]
CaO nanoparticles	under (11–97)% RH	79	147	[51]
Pt nanoparticles decorated MoS2 nanoflakes	∼4000 at 85% RH	92	154	[52]
Fe2O3/CNT	under (11–97)% RH	17	25	[13]
ZnO/MoS2	∼301 at 85% RH	138	166	[17]
CoHoxCr2-xO x = 0.005 to 0.02	∼102 at 85% RH	NA	NA	[18]
MgTiO3 doped with 2 wt% of LiF	105 at 90% RH	NA	NA	[53]
Mn0.95Bi0.05Fe2-xHoxO4 x  =  0 to 0.03	∼15 at 97% RH	79	91	[54]
NiFe2O4/CoCr2O4	∼10 under (11–97)% RH	10	15	[19]
Co1-xCdxCr2O4 x  =  0 to 0.15	∼24 under (11–99)% RH	14.6	8.6	[55]
NdCoO3	∼50% under (11–97)% RH	3	NA	[11]
Mg0.9Rb0.1DyxFe2-xO4 x = 0 to 0.03	under (11–98)% RH	18	90	[56]
MgxRb1−xFe2O4 x = 0.025 to 0.1	∼18 at 97% RH	20	30	[20]
ZnFe2 − xLuxo4 x = 0 to 0.7	∼10 at 97% RH	6.5	35.6	[57]
(1−x) NiFe2O4 /(x) Nb2O5 x  =  0.10 to 0.5	1190 at 97% RH	20	30	[58]
Bi@ZnFe2O4 x = 0 to 0.05	∼6 at 97% RH	79	91	[59]
NiFe2O4	4 at 90% RH	NA	NA	[60]
ZnFe2O4	22 at 90% RH	NA	NA	[60]
Planar mullite structure	322.9 at 85% RH	91	167	[25]
3D mullite structure	7 at 89% RH	501	120	This work

illustrates the results of the recent literature data (of the last five years) on resistive ceramic oxide humidity sensors. The tested sensors show interesting performances. The performance can probably be further improved through targeted sensitization.

4. Conclusions

In this article, the 3D printing technology was used to produce mullite humidity sensors with two types of TPMS structure (Schwartz and Gyroid structures). Sensors were sintered at 1450 °C and tested from 0% to 89% relative humidity at RT.

At room temperature, the impedance value dropped from 75 MΩ under dry air to 26 MΩ under 89 RH% for the Schwartz 1 structure (from 132 MΩ to 98 MΩ for the Schwartz 2 structure) and from 96 MΩ under dry air to 79 MΩ for the Gyroid 1 structure (from 101 MΩ to 14 MΩ for the Gyroid 2 structure). The results showed that, at 89% RH, the best response of the sensors was achieved for the Gyroid 2 structure. When the RH level changed from 0% to 89%, the best response times were equal to 492 s for Schwartz 1 and 501 s for Gyroid 2, whereas the recovery times were equal to 98 s for Schwartz 1 and 120 s for Gyroid 2. There was apparently only a limited or no influence of the height and of the type of 3D structure. On the contrary, the response of sensors S1 and G2 was higher with respect to the previous ones, highlighting the importance of the mullite powder (Mf in this case) and the porosity.

The sensors exhibited good repeatability. In addition, the tests carried out for NO₂ (2.5 ppm in the air), CO₂ (500 ppm in the air), CH₄ (100 ppm in the air), and NH₃ (44 ppm in the air) under the same flow rate showed no interference.

In this work, we showed the possibility of producing the three-dimensional-printed complex mullite structures based on TPMS and their use as resistive ceramic humidity sensors. The design of the sensors should be optimized to print flat surfaces on the top and at the bottom of the Gyroid and the Schwartz structures, where brush electrodes, for example, in place of the aluminum tape, were used in this work. Another optimization method is to keep the TPMS architecture but to use another sensing material, for example, ZnO. In any case, one of the possible applications of the proposed sensors is the use for humidity detection in soils where high values of humidity are present. Mullite is already used as the porous tip of tensiometers for soil moisture determination and could be a good candidate for this aim. Due to their thermal and chemical stability, it is also possible to use the proposed sensors in harsh environments.