Investigating the Reliability and Dynamic Response of Fully 3D-Printed Thermistors

Abstract

This paper investigates the measurement capability, dynamic response, and mechanical reliability of all 3D-printed multi-material thermistors. The thermistor design consisted of three main components: a polycarbonate (PC) substrate, a silver (Ag) electrode pair, and a poly(3,4-ethylenedioxythophene):poly(4-styrenesulfonate) (PEDOT:PSS) thermosensitive layer. The thermistors were fabricated using two manufacturing techniques: fused deposition modeling (FDM) for the substrate and micro-dispensing for the Ag and PEDOT:PSS films. Two designs with different sensing areas, D1 (90 mm²) and D2 (54 mm²), were fabricated. As the indicator of measurement capability, the highest thermal indexes were recorded as 905.64 and 813.03 K for D1 and D2 thermistors, respectively. Thermistors exhibited comparable dynamic performance, with normalized resistance variations ranging from 0.96 to 1 for temperature changes between 25 and 45 °C. The sensing area influenced both measurement capability and dynamic performance, where larger sensing areas enhanced measurement capability but extended the time required to complete dynamic cycles, around 400 s for D1 versus 350 s for D2. Adhesion tests revealed a strong bonding between the PEDOT:PSS and Ag layer with less than 5% material removal. However, the adhesion of the PEDOT:PSS to the PC substrate was weak, with over 65% material removal. Morphological analysis indicated that the poor adhesion was caused by suboptimal surface properties of the 3D-printed substrate, even resulting in gaps between these two surfaces. This study demonstrates that our all 3D-printed multi-material thermistors can match reported measurement performance with an acceptable dynamic performance while highlighting the need to improve 3D-printed substrate surface properties to enhance the performance of such multi-material structures.

1. Introduction

Additive manufacturing (AM), commonly known as 3D printing, is defined as the process of joining materials to make parts from 3D model data, usually layer upon layer, as opposed to subtractive and formative manufacturing technologies [1]. Its ability to fabricate complex geometries, minimize material waste, and enable customization has positioned AM as a key technology in various areas such as aerospace [2] and medical [3]. Moreover, with the growing interest in the utilization of AM technologies, new approaches have emerged for the development of multi-material 3D-printed functional parts [4]. These approaches often combine AM and Printed Electronics (PE) techniques within a hybrid machine or across two distinct platforms [5,6,7]. The concept involves the use of AM technologies to fabricate polymer-based substrates while employing PE methods to precisely deposit functional materials onto these substrates. This, therefore, has opened new opportunities for manufacturing smart multi-material structures such as interconnects, sensors, and circuits fabricated and integrated within AM-manufactured housings [8,9]. Among different AM technologies, fused deposition modeling (FDM) is widely preferred for the fabrication of the substrate or base material for such multi-material 3D-printed structures due to its advantages, including material availability, ease of use, and seamless integration with different PE platforms [10,11,12].

Such an approach has been demonstrated by different researchers. FDM-fabricated polyetherimide substrate was employed in CubeSat subsystems for aerospace applications [13]. A multi-material circuitry was created by directly writing functional inks onto FDM-manufactured acrylonitrile butadiene styrene substrate [14]. In another study, a 3D electric circuit was developed using an automated hybrid direct-write 3D printing process that employs a dual-FDM system [15]. A free-form functional structure was created by embedding electrical components within FDM-fabricated housing by Goh et al. [16]. Nassar and Dahiya [17] explored the feasibility of embedding electrical parts using copper-based conductive polymer composites into FDM-fabricated polylactic acid parts. Hong et al. [18] used vacuum-forming technology to shape FDM-manufactured sheets to be used as substrates. In another study, the implementation of fusible alloys for creating conductive paths inside 3D-printed polymer structures was also demonstrated [19]. Cicek et al. [9] comprehensively evaluated the effect of substrate performance on FDM-fabricated parts made from various materials, in terms of the mechanical and electrical performance of printed films. Similarly, the influence of process parameters on the electrical conductivity of the printed paths was also assessed [20].

Among various applications, printed thermistors are particularly promising due to their simple structure, fast and accurate thermal response, and wide sensing range [21,22,23]. A typical printed thermistor consists of three main layers: a substrate, a pair of electrodes, and a thermosensitive (active) layer [24,25]. The substrate serves as the mechanical support for the device and is expected to have high thermal stability and compatibility with printing processes [26]. The electrodes, commonly made from conductive materials such as silver or carbon-based inks, are patterned onto the substrate to create conductive paths, forming an electrical interface that enables current to pass through the thermosensitive layer [27]. The active layer, also known as the thermosensitive layer, is the key functional component, whose electrical resistance varies with temperature thereby allowing the device to detect and quantify thermal changes based on resistance measurements across the electrode pair [28]. Among various thermosensitive materials, poly(3,4-ethylenedioxythophene):poly(4styrenesulfonate) (PEDOT:PSS) is commonly preferred due to its advantages such as low-cost processing, high conductivity, improved wettability, and compatibility with PE technology [29,30,31].

In the current literature, PEDOT:PSS is demonstrated to be printed generally on traditional substrates such as polyethylene terephthalate [32,33], and polyimide [34,35]. However, only a limited number of studies have reported the utilization of 3D-printed substrates, thereby enabling a fully printed thermistor structure encompassing the substrate, electrode pair, and thermosensitive layer. For example, a fully 3D-printed multi-material thermistor was recently developed by micro-dispensing silver (Ag) and PEDOT:PSS onto an FDM-manufactured polycarbonate (PC) substrate [8] due to the increasing use of PC material in the FDM process [36].

Despite these advancements, there remains a lack of comprehensive understanding regarding the measurement capability, dynamic performance, and reliability of such fully 3D-printed multi-material smart structures, especially given the use of unconventional substrates, i.e., 3D printed substrates. Addressing these knowledge gaps is pivotal for unlocking new frontiers in functional 3D-printed technologies, revealing their full potential in the multi-material 3D printing domain. This study, therefore, investigates the measurement capability, dynamic behavior, and mechanical reliability of 3D-printed multi-material thermistors fabricated via FDM and micro-dispensing technologies. By examining the performance and reliability of these devices, this research contributes to the expanding knowledge base on multi-material 3D-printed smart structures and provides valuable insights into their practical applications in the areas of 3D printing, such as smart textiles, health monitoring, and aerospace engineering.

2. Materials and Methods

The multi-material thermistor was designed with three core components: a polymer substrate, a pair of conductive electrodes, and a thin sensing layer. The overall device dimensions were adjusted based on the sensing area, which was determined from values reported in the literature [22,33,37]. During the thermistor design process, fabrication flexibility and printing resolution limits of the equipment used such as deposition width and layer thickness were also taken into account to ensure manufacturability and performance. The dimensions of the thermistor are depicted in Figure 1a. Two distinct designs featuring different sensing areas, referred to as D1 and D2, were developed using SolidWorks^® (Dassault Systemes, Waltham, MA, USA). Figure 1b,c demonstrate the designed structure and a manufactured thermistor.

The thermistors were constructed using three distinct materials: Ultimaker’s PC filament for the substrate, an Ag paste (C2180423D2, Sun Chemical, Parsippany, NJ, USA) for the contact electrodes, and a PEDOT:PSS (Sigma-Aldrich, Burlington, MA, USA, 768650) for the sensing layer. Both the Ag paste and PEDOT:PSS were used in their original form as supplied by the manufacturers, without further synthesis or purification. Although primarily formulated for screen printing, based on the information gained from the literature, Ag and PEDOT:PSS materials can also be suitable for the dispensing method. Further technical details on the thermistor materials are provided in

Table 1. Technical details of PC, Ag, and PEDOT:PSS materials.

PC Substrate
Property	Specific gravity	Melt mass-flow rate	Glass transition	Heat deflection
Unit	g/cm3	g/min	°C	°C
Value	1.18–1.20	2.30–2.60	107.70	104.50 ± 0.70
Ag electrode
Property	Viscosity	Sheet resistance	Specific gravity	Solid content
Unit	Pa.s	mΩ/sq	g/cm3	%
Value	7.00–10.00	25.00–85.00	1.996	55.69–57.69
PEDOT:PSS layer
Property	Viscosity	Resistance	Concentration	Appearance
Unit	Pa.s	Ω/sq	%	Color
Value	≥50.00	≤130.00	5.00 wt	Dark to very dark blue and black

.

Print parameters were set using Ultimaker Cura software (version 4.9) for the substrate layer made from PC, while Slic3r software (version 1.2.9) and FullControl GCode Designer [38] were used to generate the gcodes for the Ag and PEDOT:PSS layers. PC substrates were fabricated using an Ultimaker³ 3D printer (Ultimaker B.V., Utrecht, The Netherlands) based on FDM technology. The substrates were cleaned with isopropanol prior to further processing. The contact electrodes were then dispensed and cured inside a temperature incubator (GenLab–INC/50/DIG, Wolflabs, York, UK) at 100 °C for 2 h. Subsequently, the PEDOT:PSS sensing layer was dispensed onto the structure and cured at 80 °C for 1 h, as per the manufacturer’s recommendation. Both the Ag and PEDOT:PSS layers were dispensed using a syringe-based micro-dispenser printer, Hyrel System 30M (Hyrel3D, Norcross, Atlanta, GA, USA). For each design, a set of five thermistors was fabricated. All printing parameters utilized for the thermistor layers are given in

Table 2. Manufacturing parameters of thermistor layers.

Parameter	Unit	PC Substrate	Ag Electrode	PEDOT:PSS Layer
Thickness	[µm]	500	50	50
Layer thickness	[mm]	0.05	0.05	0.05
Nozzle diameter	[mm]	0.4	0.2	0.2
Dispensing width	[mm]	0.04	0.15	0.15
Print speed	[mm/s]	20	30	25
Nozzle temperature	[°C]	275	-	-
Curing temperature	[°C]	-	100 (for 2 h)	80 (for 1 h)

.

The thermistors were characterized by exposing them to a temperature range of 25–45 °C while measuring their corresponding resistance. This process was conducted within an environmental control chamber, HygroGen (Rotronic Instrument Ltd., Crawley, West Sussex, UK), which was used to acclimatize the thermistors. The chamber’s temperature and humidity were precisely controlled using the integrated HygroClip S control probe.

The 3D-printed thermistor operates based on the temperature-dependent resistance behavior of its PEDOT:PSS sensing layer. As temperature changes, the electrical resistance of PEDOT:PSS alters. In other words, the resistance varies as a function of temperature. This variation is detected through the silver contact electrodes, which capture the corresponding electrical response. Therefore, resistance measurements, used to evaluate thermal measurement capability, were performed using the two-wire resistance measurement method with a Fluke 179 True RMS multimeter. After an initial stabilization period of 1 h at 25 °C, the temperature was adjusted in 2 °C increments up to 45 °C and then gradually reduced back to 25 °C for each measurement cycle. The thermal index (β), a critical parameter for evaluating the temperature sensor’s measurement capability, was considered to evaluate the measurement capability of the manufactured thermistors [39]. Therefore, data collected during both the heating and cooling phases were employed to determine the β values using Equation (1):

$$\mathrm{\beta }={\displaystyle \frac{\ln \left(\frac{{\mathrm{R}}_1}{{\mathrm{R}}_0}\right)}{\left(\frac{1}{{\mathrm{T}}_1}-\frac{1}{{\mathrm{T}}_0}\right)}},$$

(1)

where

R₁: the resistance at the peak temperature,

R₀: the resistance at the base temperature,

T₁: the peak temperature (K),

T₀: the base temperature (K).

To reflect the variation in β values across consecutive temperature points and provide insight into the stability of the thermistor’s performance over the measured temperature range, the calculated β values were used to evaluate the consistency of the thermistor’s response to temperature changes. For this, the percentage error between the β value measured at a given temperature and the β value measured at the previous temperature was calculated using Equation (2):

$$\mathrm{\%}\mathrm{E}\mathrm{r}\mathrm{r}\mathrm{o}\mathrm{r}={\displaystyle \frac{\left|{\mathrm{\beta }}_{\mathrm{i}}-{\mathrm{\beta }}_{\mathrm{i}-1}\right|}{{\mathrm{\beta }}_{\mathrm{i}-1}}}\times 100\%,$$

(2)

where β_i is the β value calculated at the current temperature step, and β_i−1 is the β value calculated at the previous temperature step.

Dynamic response measurements were performed in triplicate using a Model DI-145 data acquisition kit (DATAQ^® Instruments Inc., Akron, OH, USA) to capture the instantaneous resistance changes in the thermistors in response to temperature fluctuations. For this purpose, an electrical circuit, as illustrated in Figure 2, was created. The voltage across the thermistors, reflecting resistance variations with temperature, was recorded using the DATAQ hardware manager software. The recorded data were then exported to an Excel spreadsheet, where resistance changes were calculated based on Ohm’s Law.

As the PEDOT:PSS sensing layer directly contacts both the PC substrate and Ag contact electrode pair, its mechanical reliability was evaluated by dispensing a square PEDOT:PSS (20 mm × 20 mm, 50 µm thick) film on the PC substrate and on a square Ag film (25 mm × 25 mm, 50 µm thick) that was deposited onto a different PC substrate. A total of ten specimens, five for each design, were fabricated for this evaluation. The layout of the adhesion test specimens is illustrated in Figure 3.

Adhesion levels of such structures, as an indicator of mechanical reliability, were evaluated using test method B of ASTM D3359-17: standard test methods for rating adhesion by tape test [40]. For the peel test, a pressure-sensitive tape (3M 8981) with a peel strength of 7.10 N/cm was utilized. The adhesion grade of the PEDOT:PSS layer was determined by visually inspecting the pattern of film detachment on both the PC substrate and the Ag film after the test. The detachment percentage was estimated by comparing the observed detached amount of the material to reference patterns described in the ASTM D3359-17 standard classification table. Adhesion was then rated on a scale from 0B to 5B, where 5B indicates the highest adhesion and 0B represents the lowest adhesion. The ratings correspond to detachment percentages of 0%, <5%, 5–15%, 15–35%, 35–65%, and >65%, respectively. All tests were performed at room temperature and 50% relative humidity to prevent the influence of extreme environmental conditions on the adhesion of the tape or conductive films.

Morphological analysis of multi-material thermistors was conducted using the scanning electron microscopy (SEM) method. Imaging was conducted on a JSM-7800F Schottky field emission scanning electron microscope (JEOL, Tokyo, Japan). Thermistors were cross sectioned into 1 mm × 1 mm samples from their back face using a surgical scalpel. The specimens were coated with a thin layer of gold (80%) and palladium (20%) blend in an argon environment for 90 s using a rotary-pumped coater (Q150R S, Quorum, East Sussex, UK). SEM imaging was conducted at an accelerating voltage of 5 kV to capture high-resolution magnified images.

3. Results and Discussion

The thermal index is a key parameter that reflects the measuring capability of thermistors with higher values providing better measurement capability and better temperature stability. The calculated thermal indexes of manufactured thermistors at different temperatures are demonstrated in

Table 3. Calculated thermal index values with their standard deviation.

Temperature Range	D1		D2
	Heating	Cooling	Heating	Cooling
25–27	905.64 ± 16.20	719.48 ± 14.75	813.03 ± 44.00	691.60 ± 17.40
27–29	852.28 ± 26.04	588.02 ± 18.42	681.69 ± 37.00	621.96 ± 24.90
29–31	805.49 ± 16.87	474.26 ± 22.19	651.05 ± 30.50	557.60 ± 35.70
31–33	771.36 ± 15.42	485.90 ± 26.03	606.15 ± 14.64	463.43 ± 29.27
33–35	747.56 ± 25.57	476.26 ± 35.54	570.75 ± 17.80	433.44 ± 15.13
35–37	741.44 ± 39.96	472.04 ± 44.59	557.22 ± 18.26	425.12 ± 31.35
37–39	730.16 ± 42.46	467.65 ± 48.75	551.69 ± 22.27	422.16 ± 26.12
39–41	725.61 ± 44.76	453.59 ± 49.27	540.33 ± 21.62	398.49 ± 16.15
41–43	724.61 ± 54.76	440.08 ± 59.49	533.45 ± 17.58	387.80 ± 7.64
43–45	723.74 ± 65.08	435.89 ± 62.24	532.44 ± 22.57	361.67 ± 8.70

and summarized in Figure 4. Both thermistors exhibited changes in their β values with temperature changes during both heating and cooling, with higher β values recorded at lower temperatures. The highest calculated thermal indexes were 905.64 K and 719.48 K for D1 thermistors and 813.03 K and 691.60 K for D2 sensors, for heating and cooling, respectively. Given that high β values indicate improved temperature stability and measurement capability [39], the proposed thermistors fabricated on substrates fabricated from PC demonstrated a satisfactory level of measurement capability. This is evident when compared to commercial thermistors, which typically achieve β values exceeding 1000 K [39,41]. Furthermore, with an increase in temperature, there was a decrease in thermal indexes, which was a result of the semiconductor-like nature of PEDOT:PSS that exhibited negative temperature coefficient (NTC) behavior, as previously shown by Cicek et al. [8].

Further interpretation of Figure 4 shows that D1 thermistors exhibited slightly better measurement capability compared to D2 thermistors. This can be attributed to their larger sensing area, which enhanced measurement capability. Furthermore, the consistency analysis of β values across temperature intervals reveals notable differences in thermistor behavior between D1 and D2, as well as between heating and cooling cycles, as demonstrated by the percentage errors presented in

Table 4. %Error of thermistors depending on their β values.

Temperature Range	D1		D2
	Heating	Cooling	Heating	Cooling
27–29	6.26	24.47	19.27	11.20
29–31	5.81	21.88	4.71	13.08
31–33	4.43	2.39	7.41	16.28
33–35	3.18	2.02	6.20	9.13
35–37	0.83	0.89	2.43	1.96
37–39	1.54	0.94	1.00	0.70
39–41	0.63	3.10	2.10	5.94
41–43	0.14	3.07	1.29	2.76
43–45	0.12	0.96	0.19	4.22

. Overall, thermistors had low inconsistency in the measurement capability during heating compared to cooling. Regardless of the thermal cycle, thermistors were more inconsistent at lower temperatures. For instance, D1 thermistors demonstrated improved consistency after 31 °C, while D2 thermistors stabilized after approximately 33 °C. This behavior can be attributed to the initial measurement drift caused by transient conditions as the thermistors equilibrated with the environment [8]. As the temperature increased, these inconsistencies diminished, reflecting a reduction in the influence of environmental and material-related fluctuations. The improved stability at higher temperatures suggests that both thermistors perform more reliably under steady thermal conditions. Similar results have been reported in previous research [42,43].

The dynamic measurement performance of the thermistors was evaluated over three consecutive heating and cooling cycles by analyzing changes in normalized resistance. The results are presented in Figure 5, depicting the relationship between chamber temperature and normalized resistance variation over time.

According to Figure 5, both thermistors exhibited NTC behavior as evidenced by their inverse and parallel responses to the temperature changes within the environmental chamber. The dynamic performance of D1 and D2 thermistors was comparable, with normalized resistance variations ranging from 0.96 to 1. The consistent correlation between chamber temperature and resistance variations over time highlights the effectiveness of the proposed sensors for sensing applications. For D1 thermistors, the time required to complete a dynamic cycle was approximately 400 s, while the D2 thermistors completed the cycle slightly faster, in under 350 s. This variation is likely due to the shorter contact electrodes and smaller sensing area of the D2 thermistors, which facilitated faster response times and more rapid relaxation of the conductive particles during each cycle [8].

Notably, the thermistors demonstrated distinct behaviors during the dynamic cycles. Their response was relatively linear and rapid during heating but displayed significant fluctuations and slower recovery during cooling, which is likely due to the high heat capacitance of the substrate manufactured from PC material [44] since cooling is affected by how efficiently heat is dissipated. Similar observations have been reported in the literature for alternative substrates, including textiles [45].

Additionally, negligible deviations from the initial normalized resistance values were observed across the dynamic cycles, with the initial resistance value of subsequent cycles being slightly lower than that of the first. This behavior is likely attributed to the thermal degradation of the thermistors and their exposure to environmental humidity, as no insulation layer was incorporated during fabrication. Even so, the temperature data collected from the thermistors during both heating and cooling cycles showed good agreement with the environmental chamber results.

The reliability of Ag films on PC substrates has been previously evaluated, demonstrating a high adhesion strength [9]. It is also well known that the adhesion of PEDOT:PSS films can pose challenges depending on the substrate [46]. Therefore, in this study, the mechanical reliability of PEDOT:PSS films on PC substrates and Ag films was of particular interest due to its direct contact within the thermistor structure. Consistent with this, the study findings indicated slight variations in the adhesion performance of PEDOT:PSS films based on the underlying substrate. The tested adhesion specimens are demonstrated in Figure 6.

According to visual evaluations in line with the reference pattern described in the ASTM D3359-17 standard classification table, PEDOT:PSS film deposited on the PC substrate exhibited very weak mechanical reliability, with an adhesion strength of 0B, as more than 65% of the film detached, as seen in Figure 6a. This is a likely result of the disparity between the surface energy of the substrate and the surface tension of the PEDOT:PSS solution [47]. On the other hand, the adhesion performance of PEDOT:PSS film on Ag film was very robust compared to the PC substrate, having an adhesion strength of approximately 4B with almost no material removal, as demonstrated in Figure 6b. Similar outcomes have been reported by different researchers [48,49]. The reason for the better adhesion is related to the possible electrostatic attraction between these layers, where the negatively charged sulfonate groups in PSS formed electrostatic interactions with the positively charged Ag silver film. Another possible reason could be the print orientation of the PEDOT:PSS film relative to the underlying substrate. Although PEDOT:PSS was dispensed at 0°, the actual orientation varied depending on the substrate material. For example, on the PC substrate, the relative orientation was 45°, whereas on the Ag film, it was 0° with respect to the surface of each material. Therefore, the effect of print orientation on mechanical reliability should be further investigated.

To further investigate the variation in the adhesion depending on the underlying feature, SEM images were analyzed. As shown in the SEM image in Figure 7, a distinct gap line between the PEDOT:PSS and PC substrate was observed. This gap likely weakened the adhesion mechanism between the two layers, resulting in the PEDOT:PSS films detaching more easily. This is attributed to the coarse surface of the PC substrate, which contributed to weak adhesion and needs surface treatments.

The proposed method involves a multi-step manufacturing process that requires manual handling of the substrate between printing stages, which can introduce variability and alignment challenges. Additionally, unlike traditional substrates with smooth surfaces that offer better surface quality and interfacial adhesion, this study utilizes 3D-printed substrates without further surface treatment. The inherent roughness of these substrates leads to irregularities such as porosity, resulting in poor uniformity and inadequate interfacial adhesion between the PC and PEDOT:PSS layers. This is evidenced by the relatively weak adhesion observed between these layers. Therefore, further optimization to improve interfacial stability between layers is essential. Addressing these challenges through enhanced substrate preparation, process automation, and adhesion promotion strategies will be crucial for improving the reliability and scalability of fully 3D-printed thermistors.

4. Conclusions

This study presents a novel 3D-printing approach for fabricating multi-material thermistors, integrating the substrate, Ag electrodes, and PEDOT:PSS thermosensitive layer using FDM and micro-dispensing technologies. The practical applications of this work include but are not limited to the following: smart textiles, health monitoring, and aerospace engineering, where fully 3D-printed sensors can offer new capabilities, such as integration into complex geometries.

Key findings from this work include the following:

• The thermistors were characterized in terms of measurement capability, dynamic performance, and mechanical reliability over the temperature range of 25 to 45 °C.
• The highest recorded thermal indexes were 905.64 K for D1 and 813.03 K for D2 thermistors, comparable to values reported in the literature.
• Both designs showed similar dynamic performance, with the smaller sensing area of D2 providing quicker response times and faster relaxation.
• Mechanical reliability was material-dependent: strong bonding was observed between PEDOT:PSS and the Ag layer, while weak adhesion with the PC substrate resulted in over 65% material removal.

Although this study demonstrated measurement capability and dynamic performance comparable to the existing literature, future work should focus on improving the surface properties of FDM-fabricated substrates to improve the adhesion performance towards achieving highly reliable, fully 3D-printed thermistors. The remaining challenges include material compatibility and interface engineering, with a focus on incorporating additional functionalities such as on-board heating and closed-loop thermal regulation. These advancements will help the realization of fully printed smart systems for wearable and biomedical applications.