Additive Manufacturing of Composite Structures with Transverse Thermoelectricity

Abstract

This study investigates the application of additive manufacturing (AM) in fabricating transverse thermoelectric (TTE) composites, demonstrating the feasibility of this methodology for TTE material synthesis. Zinc oxide (ZnO), a wide-bandgap semiconductor with moderate thermoelectric performance, and copper (Cu), a highly conductive metal, were selected as base materials. These were formulated into stable paste-like feedstocks for direct ink writing (DIW). A custom dual-nozzle 3D printer was developed to precisely deposit these materials in pre-designed architectures. The resulting structures exhibited measurable transverse Seebeck effects. Unlike prior TE research primarily focused on longitudinal configurations, this work demonstrates a novel AM-enabled strategy that integrates directional compositional anisotropy, embedded metal–semiconductor interfaces, and scalable multi-material printing to realize TTE behavior. The approach offers a cost-effective and programmable pathway toward next-generation energy harvesting and thermal management systems.

1. Introduction

Thermoelectric (TE) materials have attracted considerable interest due to their unique capability to achieve direct bidirectional conversion between thermal and electrical energy [1,2]. Recent advancements have been driven by increasing demands for sustainable energy technologies and high-efficiency thermal regulation systems [3]. Emerging applications highlight their transformative potential across diverse fields, including industrial waste heat recovery [4], decentralized power generation in off-grid regions [5], self-powered remote sensing networks [6], and extraterrestrial thermal control systems [7]. These developments collectively position thermoelectric materials as critical enablers of next-generation energy conversion and management technologies.

Unlike conventional TE devices that operate along a single axis, transverse thermoelectric (TTE) systems exploit the transverse Seebeck effect, where the generated voltage is orthogonal to the direction of heat flow [8,9]. This geometric decoupling enables enhanced design flexibility and material utilization, particularly through the strategic exploitation of anisotropic properties [10,11]; while these advancements show great promise, existing fabrication methodologies for TTE materials face challenges in complexity and scalability, which limit their broader implementation.

Figure 1 schematically illustrates a transverse thermoelectric (TTE) composite structure featuring engineered rod-like inclusions [8]. In this design, high-conductivity fibrous elements are embedded within a thermoelectric matrix at a predefined orientation angle $\theta$, creating a deliberately anisotropic architecture. A temperature gradient $\mathsf{\Delta }{T}_y$ is applied along the y-axis (red arrow), while the resulting transverse voltage $\mathsf{\Delta }{V}_x$ is measured along the orthogonal x-axis (blue arrow). This configuration demonstrates the transverse Seebeck effect, where anisotropic properties give rise to an electric field perpendicular to the heat flow. The alignment angle $\theta$ modulates this effect through its influence on directional transport behavior. The transverse Seebeck coefficient is defined as

$$S_{xy}={\displaystyle \frac{\mathsf{\Delta }{V}_x}{\mathsf{\Delta }{T}_y}}$$

where $S_{xy}$ quantifies the material’s ability to convert a lateral temperature gradient into a transverse voltage. Its magnitude depends on directional differences in electrical conductivity, thermal conductivity, and the intrinsic Seebeck coefficients [12,13].

In composite systems like ZnO–Cu, metal–semiconductor interfaces can form between the Cu inclusions and the ZnO matrix. These interfaces may introduce Schottky-type barriers that influence charge carrier transport across phases [14,15]. Such interfacial effects, combined with engineered anisotropy, are key to enhancing TTE performance.

Despite notable advancements in transverse thermoelectric (TTE) material development, significant manufacturing challenges remain, particularly in terms of structural complexity and scalability. Conventional mechanical machining, although commonly used for shaping thermoelectric components, lacks the precision needed for intricate geometries and often introduces stress-related performance degradation [16]. Likewise, bulk processing techniques such as hot pressing and spark plasma sintering (SPS), though effective for densification, offer limited control over microstructural features essential for anisotropic behavior [17]. Thin-film deposition methods, including sputtering, face inherent limitations due to expensive equipment requirements and poor scalability [18]. Additionally, during TTE device assembly, misalignment of thermal and electrical interfaces—especially between oxide and metal phases—can lead to increased contact resistance and reduced energy conversion efficiency [19]. These combined limitations underscore the critical need for fabrication strategies that unify structural precision with scalable, low-cost manufacturing.

Additive manufacturing (AM) has emerged as a promising solution, enabling the construction of complex thermoelectric architectures through layer-by-layer material deposition. AM offers unprecedented geometric freedom and cost-effectiveness, overcoming many of the constraints associated with traditional processing methods [20,21]. Among various AM approaches, direct ink writing (DIW) has shown particular promise in producing ceramic-based structures with engineered anisotropy.

Several studies have demonstrated the applicability of AM in thermoelectric (TE) systems. Kim et al. formulated DIW-compatible Bi₂Te₃ inks to fabricate conformable TE lattices with enhanced impedance matching [22]. Burton et al. utilized a pseudo-3D printing method to construct SnSe-based devices with embedded copper current collectors, achieving over 200% improvement in power density [23]. These works illustrate the potential of AM to tailor geometry, interfaces, and architecture for high-performance TE devices.

Nevertheless, most prior research has centered on longitudinal Seebeck transport. In contrast, this study focuses on extending AM capabilities to transverse thermoelectric applications by integrating directional anisotropy and functional phase alignment within a co-extruded composite design. We propose a cost-effective, FDM-derived AM strategy for fabricating TTE structures. This method incorporates two major innovations: (i) the adaptation of standard FDM hardware with modular paste extruders for metal/ceramic co-printing, and (ii) the implementation of layer-wise deposition protocols compatible with conventional FDM workflows, removing the need for specialized operator training. A custom dual-nozzle 3D printer was developed to precisely control composition ratios and spatial orientation, enabling fabrication of composite structures with tunable anisotropy. Zinc oxide (ZnO) was selected as the thermoelectric matrix due to its favorable TE properties, while copper (Cu) served as the conductive phase, consistent with its successful integration in prior multilayer ceramics [14,24]. To ensure stable and reproducible printing, we optimized ZnO and Cu formulations to develop AM-compatible paste feedstocks, enabling consistent long-duration operation [25].

2. Materials and Methods

2.1. Material Synthesis

The experimental materials comprised copper powder (Sigma-Aldrich, St. Louis, MO, USA, 98%) and zinc oxide powder (Sigma-Aldrich, St. Louis, MO, USA, 99.9%). A solvent-based paste system was formulated using isopropyl alcohol (Florida Laboratories, Fort Lauderdale, FL, USA, 95% IPA) as the solvent, B72 resin (TALAS, Brooklyn, NY, USA, Paraloid B72) as the polymeric binder, and glycerol (Eisen-Golden Laboratories, Irvine, CA, USA, Glycerol 99.7%) as a dual-function rheological modifier to enhance paste stability and modulate drying kinetics.

For the copper paste, 15 g of copper powder was blended with 5 mL IPA, 1 g B72 binder, 0.5 g PVP40 dispersant (Sigma-Aldrich, St. Louis, MO, USA, Polyvinylpyrrolidone, mol wt 40), and 1 mL glycerol to reduce cracking caused by rapid drying. The zinc oxide paste contained 60 g ZnO powder mixed with 34 mL IPA, 2 g B72 binder, and 6 mL glycerol, utilizing its hygroscopic nature to maintain consistent viscosity during extended printing sessions.

According to product specifications, both ZnO and Cu powders possess stable polycrystalline structures under ambient conditions. ZnO typically exhibits a hexagonal wurtzite phase, which is thermodynamically favored and widely observed in functional ceramic applications [14]. Copper adopts a face-centered cubic (FCC) crystal structure, which contributes to its high electrical conductivity and mechanical formability [26]. The paste preparation process was designed to preserve these crystalline features, ensuring uniform dispersion and rheological compatibility for direct ink writing. Thermal behavior and microstructural evolution during formulation and drying have been extensively studied for similar oxide–metal systems, further supporting the phase stability throughout processing [27].

Each paste was prepared by sequentially dissolving B72 resin in preheated IPA (60 °C) under continuous stirring. For the copper paste, PVP40 was added after dissolution to ensure particle dispersion. Glycerol was subsequently incorporated as a plasticizer and drying-rate regulator, extending the open time to 45 to 60 min before gradually introducing the metal or oxide powders. This approach minimized viscosity fluctuations and structural defects caused by premature solvent evaporation. The mixtures were stirred until homogeneous and then cooled to room temperature.

Final pastes were loaded into polypropylene syringes for printing, with ZnO and copper stored separately to prevent cross-contamination. The Cu paste required over 0.5 MPa air pressure to counteract its higher viscosity and density, whereas the ZnO paste operated at approximately 0.3 MPa. This pressure adaptation ensured synchronized extrusion stability and minimized disruptions from external environmental variations.

2.2. Printer Setup

The 3D printer used in this study is based on a standard Cartesian motion platform. Two servo motors are independently mounted on the X-axis to control the dual extruders, sharing a common linear rail while maintaining independent operation (Figure 2). The control system includes a Sovol SV04 32-bit Silent Mainboard (V5.2.1) equipped with TMC2209 drivers. Since the materials used in this study do not require heating, the printer is not equipped with a heated bed or heating block. The build platform consists of a smooth glass plate. Each print head integrates a servo-driven custom screw extruder with interchangeable nozzles of 1.2 mm and 0.4 mm diameters, optimized for the extrusion of ZnO and Cu pastes, respectively.

The printing material is stored in syringes mounted adjacent to the extruder, where external compressed gas is applied to drive the syringe plunger and feed material into the screw extruder. The slicing software used was SOVOL3D Cura (Version 1.3.0).

2.3. Printing Process

The test model, a $10\times 10\times 20$ mm zinc oxide (ZnO) rectangular prism with embedded copper (Cu) wires, was fabricated using a dual-extruder 3D printer. Copper lines were integrated into the ZnO matrix at vertical positions of 2 mm, 4 mm, and 6 mm, with each layer comprising three evenly spaced traces oriented at a 45° angle. A structured arrangement of three layers and three traces per layer was selected to ensure high fabrication success and structural consistency while validating the feasibility of the multi-material DIW process. Using three traces per layer provided a balanced compromise between geometric complexity and mechanical reliability. Configurations with a higher trace density were excluded to reduce the likelihood of printing failure and thermal short-circuiting.

For ZnO, the printing layer thickness was set to 0.6 mm with 100% infill density, three outer shells, and a grid infill pattern. The infill speed was 15 mm/s. For Cu, which was printed as discrete fiber traces, a finer layer thickness of 0.1 mm and a reduced print speed of 10 mm/s were used. No infill or shell parameters were applied to the Cu phase, as it consisted of individual line segments. These differentiated printing parameters were designed to accommodate the distinct rheological behaviors of ZnO and Cu pastes, ensuring synchronized deposition and structural stability.

The Cu traces were deposited directly onto fully infilled ZnO layers, followed by additional ZnO layers that enclosed and sealed the copper elements. The resulting structure incorporated embedded Cu features with an overall copper volume fraction of approximately 1.94% and a mass fraction of about 3.0%. This composition was primarily selected to demonstrate printability and structural integration. Prior studies have reported that excessive metallic content in TTE composites can diminish performance by promoting parallel thermal conduction pathways [13].

The layer-by-layer direct-overlapping strategy, made possible by the tuned paste formulations and controlled printing resolution, facilitated continuous integration during drying and sintering (Figure 3). The printing method exploits the contrast in paste viscosity and extrusion resolution to enable precise material placement and robust interfacial adhesion. This sequential fabrication protocol ensures structural continuity and interfacial integrity between the ZnO and Cu phases throughout the multi-material DIW process.

2.4. Post-Processing

After printing, samples were dried at room temperature for 12 h before sintering in a GSL-1500x-50 tube furnace under argon (0.5 L/min flow). A stepwise thermal profile was applied: gradual heating to 500 °C with a 1 h hold for binder removal, followed by sintering at 600 °C, 700 °C, 800 °C, 900 °C, and 1000 °C, each maintained for 6 h to investigate microstructural evolution. Samples were then cooled inside the furnace to room temperature (Figure 4).

The sintering condition of 1000 °C for 6 h was chosen based on previous optimization studies of similar paste-based systems, aiming to achieve sufficient densification while preserving the functional integrity of both ceramic and metallic phases. For ZnO-based ceramics, Lim et al. reported significant grain coarsening and densification above 800 °C, consistent with conventional solid-state sintering behavior [28]. Similarly, Chelluri et al. demonstrated that high-temperature sintering facilitates particle necking and the development of stable microstructures in Cu-based pastes [25]. These findings collectively support the assumption that both ZnO and Cu retain their crystalline phases and undergo beneficial microstructural transformation under the selected conditions.

No evidence of chemical reaction or interdiffusion between the Cu and ZnO phases was observed during fabrication. Post-sintering SEM imaging confirmed distinct phase boundaries between the two components.

2.5. Testing Instruments

The transverse Seebeck coefficient ($S_{xy}$) of the sintered samples was measured near room temperature using a custom-built measurement system (Figure 5). A thermoelectric cooler (TEC) module served as the cold end, while the hot end consisted of an electric heating block integrated with a copper base plate. Prior to testing, the sintered samples were polished to ensure surface flatness, and 1 $\mu$m thick copper electrodes were sputtered onto both ends along the length of the rectangular sample to facilitate reliable electrical contact. Voltage was measured using an HP 34420A nanovoltmeter under a stable thermal gradient ($\mathsf{\Delta }{T}_y$) applied across the sample’s width (y-direction), after temperature stabilization.

3. Results and Discussion

3.1. Printability

The printability of ZnO and Cu composite pastes was tested, with both showing smooth extrusion through the nozzle and precise layer-by-layer deposition without any structural failure. The ZnO paste, modified with glycerol to improve rheological properties, displayed strong interlayer adhesion and self-supporting characteristics. In addition to controlling extrusion rates using servo-driven screw mechanisms and independent pneumatic pressure regulation, the material flow was ensured by managing the servo’s rotation speed and air pressure values, maintaining consistent material extrusion. The Cu paste required over 0.5 MPa of air pressure to overcome its higher viscosity and density, while the ZnO paste functioned at around 0.3 MPa. The integrated wipe tower helped prevent nozzle clogging by clearing leftover paste; it led to material waste and required periodic syringe replacement. After printing, solvent evaporation quickly hardened the surface (within 10 min), allowing safe removal from the glass substrate without damaging the structure (Figure 6).

3.2. Shrinkage

To quantify shrinkage behavior, reference samples of pure ZnO and pure Cu were fabricated and sintered under the same thermal profile (1000 °C, 6 h). Dimensions before and after sintering were measured using digital calipers (±0.1 mm resolution) to calculate volumetric changes. For composite structures, both overall sample dimensions and the protruding length of embedded Cu wires were recorded to estimate the shrinkage of the metallic phase.

Figure 6a compares the dried green body (unsintered) and the 1000 °C sintered sample, showing a 6% shrinkage in the ZnO matrix during drying (24 h) and a total 23% post-sintering shrinkage relative to the wet green body, yielding a final ZnO density of 5.14 g/cm³ (91.6% of ZnO’s theoretical density, 5.61 g/cm³). The Cu wires exhibited lower shrinkage rates (4% drying, 16% sintering), with a sintered density of 8.55 g/cm³ (96.1% of Cu’s theoretical 8.9 g/cm³). These consistent shrinkage trends and final densities demonstrate effective co-sintering compatibility and validate the measurement methodology.

3.3. SEM Analysis

3.3.1. Oxide Zinc Sintering

Unlike conventional oxide ceramic processing that typically relies on mechanical pressing to maintain shape prior to sintering [27], this study employed organic binders to preserve the composite’s structural integrity during the pre-sintering stage. This approach initially introduced porosity between zinc oxide (ZnO) particles, as observed in the loosely packed pre-sintered state (Figure 7a).

After sintering at 600 °C (Figure 7b), ZnO particles remained loosely connected, with porosity resembling that of the pre-sintered state (Figure 7a). At 700 °C (Figure 7c), particle growth became evident, with the formation of initial grain-to-grain contact. At 800 °C (Figure 7d), distinct necking between particles appeared, indicating the onset of densification.

When sintered at 900 °C (Figure 7e), ZnO grains coalesced into a nearly continuous matrix. By 1000 °C (Figure 7f), only isolated pores remained amid well-developed grain boundaries. This progressive densification was primarily driven by particle bonding, grain growth, and pore elimination [28]. The grain size distribution followed a unimodal profile centered around $2–4\mathsf{\mu }\mathrm{m}$, with an average grain diameter of $2.61\mathsf{\mu }\mathrm{m}$ (Figure 7f).

3.3.2. Copper Sintering

For the Cu phase, the pre-sintered structure shown in Figure 8a exhibited a loosely packed particle arrangement with visible interstitial porosity, typical of binder-containing green bodies. Compared to ZnO, the Cu particles were significantly larger, resulting in more pronounced interparticle voids.

Following binder removal and sintering at 600 °C (Figure 8b), SEM imaging revealed discrete Cu particles with minimal neck formation. As the sintering temperature increased to 700–800 °C (Figure 8c,d), interparticle connections progressively developed via surface diffusion, producing necking features characteristic of intermediate-stage sintering.

At 900–1000 °C (Figure 8e,f), sintering produced a dense metallic network with significantly reduced porosity, consistent with final-stage sintering behavior and comparable to the morphologies observed in our previous study [25]. The copper phase exhibited a broader grain size distribution, ranging from 6 to $10\mathsf{\mu }\mathrm{m}$, averaging $8.1\mathsf{\mu }\mathrm{m}$.

3.3.3. Sintering at the Zinc Oxide–Copper Interface

The images showing the interfacial region between ZnO and Cu are presented in Figure 9. A distinct interface with a visible gap was observed in the dried green body (Figure 9a). As the sintering temperature increased, the interfacial gap gradually diminished (Figure 9b–d). Previous studies have suggested that joining between ZnO and Cu layers may be facilitated by a combination of external pressure and sintering [24].

Although press-less sintering was employed in this study, it is hypothesized that interfacial pressure may arise from differential shrinkage between ZnO and Cu. During sintering, ZnO exhibited greater shrinkage than Cu, which could have caused the ceramic matrix to progressively tighten around the embedded metallic phase. This shrink-wrapping effect appears to reduce interfacial gaps and potentially apply compressive stress to the Cu phase. The combined influence of intrinsic Cu sintering and the surrounding ZnO contraction may contribute to enhanced Cu densification. Moreover, this mechanism could help mitigate warping or cracking—issues commonly reported in co-sintered systems—by accommodating differential shrinkage and constrained sintering behavior [27].

Although this explanation is based on empirical observations and remains to be fully validated, it aligns with the expected material behavior and offers a plausible interpretation of the observed densification improvement.

3.4. Seebeck Coefficient Measurement

To evaluate the transverse behavior of the composite structure, Seebeck voltage was measured as a function of temperature difference when the cold side was maintained near room temperature. Figure 10 illustrated the temperature distribution obtained from infrared imaging during a typical measurement. As shown in Figure 11, when the temperature difference increased, the Seebeck voltage also increased steadily. The slope obtained from a linear fitting is $141\mathsf{\mu }\mathrm{V}/\mathrm{K}$, which corresponds to the average transverse Seebeck coefficient $S_{xy}$.

In contrast, a pure ZnO sample fabricated using the same printing parameters and processing conditions exhibited a much lower $S_{xy}=36\mu \mathrm{V}/\mathrm{K}$ in the same temperature range. Although it is expected that powder-based polycrystalline ZnO possesses negligible transverse Seebeck coefficient due to its isotropic nature, the printed ZnO sample in this work still showed a small but measurable transverse voltage. This result might be explained by the anisotropy introduced during printing, as the material deposition was conducted in a linear pattern within each layer.

Notably, the composite structure exhibited a transverse Seebeck coefficient approaching the longitudinal Seebeck coefficient of ZnO, which was measured to be 220 $\mu$V/K in this study using an identically printed sample. This value is consistent with previously reported Seebeck coefficients for ZnO-based materials [29]. The strong transverse response confirms the effectiveness of the fabrication strategy and highlights the critical role of engineered metal inclusions in enhancing transverse thermoelectric performance.

The 45° rod-like inclusion geometry used in this study was selected primarily to ensure printing reliability, rather than to achieve optimal thermoelectric performance; while this configuration effectively demonstrates the feasibility of embedding conductive paths to induce transverse Seebeck response, further improvements are expected through tuning the inclusion orientation, density, and interfacial design. In particular, the influence of printing direction on Seebeck behavior remains an important factor. A systematic investigation of angular alignment and its impact on $S_{xy}$ will be conducted in future work to optimize device performance.

4. Conclusions

This work developed a custom dual-head 3D printer and optimized direct ink writing (DIW) processes for precise multi-material patterning, achieving a transverse Seebeck coefficient of $S_{xy}=141\mu$V/K for Cu–ZnO composites. Sintering protocols leveraging differential shrinkage between ZnO (23%) and Cu (16%) enhanced interfacial bonding and mechanical stability, with final densities of 5.14 g/cm³ for ZnO and 8.55 g/cm³ for Cu, confirming successful co-densification.

This study demonstrated several advantages of additive manufacturing, including batch-to-sample consistency, reduced post-processing requirements, and programmable anisotropy. These encouraging results underscore the viability of dual-nozzle additive manufacturing (AM) for fabricating transverse thermoelectric (TTE) composites, and highlight significant advancements in geometric control, material integration, and performance scalability. Collectively, these advancements position AM as a transformative platform for the fabrication of next-generation energy harvesting and thermal management systems—bridging the gap between lab-scale innovation and industrial-scale manufacturability of high-efficiency TTE devices.