DissolvPCB: Recyclable Circuit Prototyping

• DissolvPCB is an electronic prototyping method that replaces conventional FR-4 boards with a water-soluble PVA substrate and eutectic gallium–indium channels for fully recyclable circuit assemblies.
• The fabrication workflow integrates FDM 3D printing, manual EGaIn injection, and non-soldered component insertion secured with PVA glue to enable intentional disassembly in water.
• This closed-loop process minimizes e-waste by allowing the recovery and reuse of substrate, conductive material, and components while maintaining practical electrical performance.

DissolvPCB is an electronic prototyping technique for fabricating fully recyclable printed circuit board assemblies using affordable FDM 3D printing, with polyvinyl alcohol as a water-soluble substrate and eutectic gallium-indium as the conductive material. In this approach, the conventional FR-4, copper, and solder stack is replaced by a 3D-printed PVA substrate, EGaIn-filled hollow channels as traces, and non-soldered component insertion secured with PVA glue. When obsolete, the assembly is recycled by immersion in water: the PVA dissolves, the liquid metal re-forms into recoverable droplets or beads, and the components separate for direct reuse. The method is presented as a locally executable, closed-loop alternative to conventional prototyping workflows, with explicit support for single-sided and double-sided PCBAs, 3D circuit topologies, and recovery of substrate, conductor, and components (Yan et al., 29 Jul 2025).

1. Definition, problem setting, and material system

DissolvPCB addresses the poor recoverability of conventional PCBAs built from FR-4 substrate, copper traces, and soldered components. That conventional stack is durable and suitable for mass production, but it is difficult to disassemble, laborious to desolder, and poorly suited to preserving functional components during recycling. The motivating context includes both industrial e-waste and short-lived, bespoke electronics produced in research labs, makerspaces, and personal fabrication settings. The paper cites around 62 million metric tons of e-waste generated globally each year, with less than 23% formally collected and recycled (Yan et al., 29 Jul 2025).

The material system consists of three principal substitutions. First, the substrate is FDM-printed PVA, used not as support material but as the main dielectric body. Second, conductive traces are implemented as interconnected hollow channels filled with EGaIn. Third, components are not soldered; they are inserted into printed sockets and secured with PVA glue. This combination is intended to preserve normal prototype functionality during use while enabling intentional disassembly in water at end of life (Yan et al., 29 Jul 2025).

The conductive alloy is eutectic gallium-indium composed of 75.5 wt% gallium and 24.5 wt% indium, described in the appendix as approximately a 3:1 ratio by weight. Gallium is warmed above its melting point of about $29.8^\circ\text{C}$, indium is gradually added, the mixture is stirred using a non-reactive stirrer or magnetic stirrer on a hot plate, and the finished EGaIn is stored in a sealed glass container to reduce oxidation. The paper states that EGaIn is considered non-toxic and has no vapor pressure at room temperature, while still recommending gloves and goggles (Yan et al., 29 Jul 2025).

This material architecture implies a shift from durability-oriented board construction to disassembly-oriented prototyping. A plausible implication is that DissolvPCB is not merely a change in substrate chemistry, but a redesign of the entire assembly logic: conductive geometry, package interfacing, and end-of-life handling are all co-optimized around water-triggered recovery.

2. Fabrication workflow and software pipeline

The fabrication workflow has four main stages: creating a printable substrate model, printing it in PVA, injecting EGaIn, and placing and securing components. The substrate model contains the board body, a conductive trace network implemented as hollow channels, and component sockets positioned so terminals align with channel openings. Geometry can be created manually in CAD for custom 3D routing or generated automatically from KiCad using the paper’s FreeCAD plugin (Yan et al., 29 Jul 2025).

Printing was performed, unless otherwise noted, on a BambuLab P1S using a 0.2 mm nozzle, 0.06 mm layer thickness, and 0.15 mm wall thickness. A 0.4 mm nozzle also worked with 0.12 mm layer thickness and 0.3 mm wall thickness. Because PVA stringing can block channels, the authors used print speed 30 mm/s, retraction distance 10 mm, bridge speed 15 mm/s, cooling fan 100%, and bridge infill angle $22.5^\circ$. With those settings, the minimum reliable hollow trace channel was $0.7\ \text{mm} \times 0.7\ \text{mm}$ for a 0.2 mm nozzle and $0.9\ \text{mm} \times 0.9\ \text{mm}$ for a 0.4 mm nozzle (Yan et al., 29 Jul 2025).

EGaIn is injected with a small syringe and tapered blunt needle. The main text specifies 25 gauge, while the appendix notes a 23-gauge tapered steel Luer-lock tip in the implementation. Injection proceeds from any channel opening until the network is filled, and good electrical contact is associated with formation of a convex meniscus at the opening. A micro-CT scan of a sample circuit showed no significant trapped air, which supports the practical viability of the filling step (Yan et al., 29 Jul 2025).

After filling, components are inserted into their sockets so that terminals contact exposed EGaIn at the channel openings. PVA glue is then applied to fix the parts and seal the openings. Commercial PVA glues are reported as usable, but the authors prepared a custom glue because excess water can soften the printed PVA substrate and cause shorts. Their formulation used PVA pellets to water at 3:5 by weight, with PVA molecular weight 31,000–50,000, stirred at $80^\circ\text{C}$ for 2–3 hours. After glue application, assemblies were dried at $60^\circ\text{C}$ for about 1 hour. The study reports no visible electrical failures, leakage, or shorts after drying across all produced examples and samples (Yan et al., 29 Jul 2025).

The software pipeline is an open-source FreeCAD plugin written in Python 3.13.2 and run through FreeCAD’s macro functionality. It parses .kicad_pcb files and extracts trace segment locations and dimensions, via locations and layer connectivity, component footprint positions and orientations, and board polygon dimensions. It then generates 3D traces or channels, vias, pads or openings, the substrate body, and component sockets. Trace segments are modeled as rectangular boxes, cylinders are added at ends as joints, vias are generated with 1.2 mm diameter, pads are modeled as thin cuboids, and the final printable model is produced by subtracting traces, vias, and pads from the board body while uniting socket geometry with the substrate (Yan et al., 29 Jul 2025).

The plugin also packages KiCad DRC settings: minimum trace width 0.7 mm, minimum trace spacing 0.15 mm, and minimum conductor-to-board-edge distance 0.15 mm. The paper states that model generation takes only a few minutes depending on complexity and hardware, and under 2 minutes for the sample circuit. This suggests that automation is important not only for convenience but also for making fabrication-aware 3D conversion routine rather than bespoke.

3. Recycling mechanism and closed-loop material recovery

The recycling mechanism is based on immersion in water. First, the PVA substrate absorbs water and dissolves into solution. At $22^\circ\text{C}$ with no stirring and just enough water to cover the sample, a small sample board dissolved in about 36 hours. With stirring and heating to $90^\circ\text{C}$, most of the sample dissolved in less than 1 hour. Dissolution rate is therefore strongly condition-dependent and can be traded between low-energy waiting and accelerated recovery (Yan et al., 29 Jul 2025).

As the PVA walls disappear, EGaIn loses geometric confinement. Because it is a liquid metal with high surface tension, it contracts and coalesces into droplets or beads. Oxidation can inhibit clean beading, so the recovered EGaIn is treated using NaOH. The paper specifies a 2 mol/L NaOH solution applied at droplet scale to restore surface tension and permit dispersed liquid metal to reform into a single bead, which can then be collected using a syringe or pipette. The resulting small amount of high-pH liquid can be neutralized with dilute citric acid until pH 6–8 or left exposed to air until conversion to a carbonate or bicarbonate mixture (Yan et al., 29 Jul 2025).

Components separate because they are not metallurgically bonded by solder. As surrounding PVA dissolves and the glue loosens, the components detach and settle. The paper states that components and EGaIn generally settle at the bottom and naturally separate from each other. Components are then retrieved, dried, and reused directly (Yan et al., 29 Jul 2025).

PVA recovery is also part of the closed loop. The dissolved PVA can be dried at room temperature or on a hot plate below water’s boiling point; the dried material peels off as sheets. These sheets can be redissolved into glue, shredded, and re-extruded into filament. The paper demonstrates re-extrusion using a Filabot EX2 at $185^\circ\text{C}$. For one spool, 16 pieces of dried PVA sheets were processed, diameter was measured every 2.5 m, and the average filament diameter was 1.792 mm with standard deviation 0.057 mm. The recycled filament was then used both as support material and to fabricate new DissolvPCB circuits, including a motor driver module (Yan et al., 29 Jul 2025).

This recovery sequence is central to the system’s definition. Unlike partial repair or conventional board recycling, it is designed to preserve all major material classes in reusable form: functional components, conductive medium, and substrate polymer. The resulting workflow aligns board fabrication and board unmaking as parts of one technical process rather than separate industrial domains.

4. Design rules, package support, and electrical performance

DissolvPCB establishes design rules experimentally rather than through formal equations. The plugin and DRC encode the key geometric constraints derived from fabrication testing. Minimum printable hollow channel size is $0.7\ \text{mm} \times 0.7\ \text{mm}$ with a 0.2 mm nozzle and $22.5^\circ$0 with a 0.4 mm nozzle. Minimum X/Y insulation wall thickness tested ranged from 0.15 mm to 0.35 mm in 0.05 mm steps, with no shorts observed even at 0.15 mm; accordingly, the minimum insulation distance between traces and the minimum conductor-to-board-edge distance are both 0.15 mm. For Z-direction insulation, tested thicknesses ranged from 0.18 mm to 0.66 mm in 0.06 mm increments, and 0.18 mm was sufficient to prevent shorts (Yan et al., 29 Jul 2025).

Layer thickness follows directly from these constraints. With trace height 0.7 mm and minimum Z insulation 0.18 mm, the paper concludes that the minimum single-layer PCB thickness is 1.06 mm and the minimum double-layer PCB thickness is 1.94 mm. The plugin example used a 2.3 mm board height consisting of three 0.3 mm insulation layers and two 0.7 mm trace-height layers. The method therefore supports single-sided and double-sided designs, and more than two layers in principle (Yan et al., 29 Jul 2025).

Component support is mediated through socket geometries rather than pads and solder joints. The paper defines sockets for THT components, SMD components with extended leads or pins, and two-terminal SMD components. Supported package constraints include components with extended pins at pitch 0.85 mm or greater and two-terminal SMD components 0603 or larger. Explicitly mentioned packages include DIP, SIP, PGA, SOIC, SSOP, QFP, SOT-23, and PLCC. The software library includes custom socket models for SOIC-8, SOIC-14, SOT-23, 0603, 0805, and 1206 (Yan et al., 29 Jul 2025).

Electrical performance is characterized empirically. Prior EGaIn resistivity is cited as $22.5^\circ$1, about 10× higher than copper, but the traces are thicker than standard copper foil traces. For traces of length 30 mm and cross-section $22.5^\circ$2, with $22.5^\circ$3, the measured average resistance was 0.03 $22.5^\circ$4 per 30 mm with standard deviation 0.0012 $22.5^\circ$5. A conventional 1 oz copper PCB with width 0.7 mm is given as approximately 0.02 $22.5^\circ$6 per 30 mm. The paper therefore positions DissolvPCB traces as somewhat more resistive but still in the same general range for low-voltage, low-current electronics (Yan et al., 29 Jul 2025).

Current capacity was tested on the thinnest viable $22.5^\circ$7 traces at 1 A, 3 A, and 5 A, with three samples per condition and duration capped at 5 minutes. These traces had 0.3 mm insulating layer on both top and bottom. The maximum stabilized surface temperature under 5 A was 51°C, below the PVA glass transition range of 85–95°C, with no overheating and no visible EGaIn displacement or overflow. Resistance stability was high, with average standard deviations $22.5^\circ$8 at 1 A, $22.5^\circ$9 at 3 A, and $0.7\ \text{mm} \times 0.7\ \text{mm}$0 at 5 A. The paper concludes that continuous current load up to 5 A appears safe for these traces under the tested conditions (Yan et al., 29 Jul 2025).

Signal transmission was tested through two collinear EGaIn traces, each 30 mm long and $0.7\ \text{mm} \times 0.7\ \text{mm}$1, connected by a 0 $0.7\ \text{mm} \times 0.7\ \text{mm}$2 resistor at the center. A Keysight 33210A function generator and SIGLENT SDS 1104X-E oscilloscope were used. Frequencies from 100 Hz to 10 MHz were evaluated with sinusoidal and quasi-square signals. The reported result was no noticeable attenuation, with input and output waveforms showing similar amplitudes and noise levels and no significant signal loss or distortion up to 10 MHz (Yan et al., 29 Jul 2025).

During use, the study reports no visible leakage or short circuits after glue drying, proper operation of all circuits in the study, and continued functionality for over 60 days under indoor humidity fluctuating between 19% and 65%, with no visible deformation. The paper does not report a formal mechanical stress, bending, or fatigue dataset for flat boards, so the strongest claims concern normal handling and indoor survivability rather than quantified structural endurance (Yan et al., 29 Jul 2025).

5. Lifecycle assessment and environmental profile

A major contribution of DissolvPCB is a cradle-to-grave lifecycle assessment comparing a DissolvPCB magnetic field detector circuit with a conventional FR-4 PCB fabricated by CNC milling. The study used ecoinvent v3.10 Cutoff and CML v4.8 LCIA implemented in OpenLCA. The functional unit was defined as “fabrication of one PCB using the PVA/FR-4 substrate.” The system boundary included material synthesis, manufacturing processes, transportation, and end-of-life management (Yan et al., 29 Jul 2025).

Because PVA was unavailable in the database, PVC was used as a substitute in the LCA model. The paper explicitly presents this as an approximation that should be considered when interpreting absolute values. This caveat is significant: it limits strict interpretation of exact category magnitudes, while leaving the comparative structure of the modeled workflow intact (Yan et al., 29 Jul 2025).

The DissolvPCB inventory for the sample circuit included 1.17 g PVA filament for printing, 0.013 g PVA pellets, 0.026 g water for glue preparation, 0.651 g gallium, and 0.217 g indium, yielding about 0.868 g EGaIn. Energy inputs were 4.37e-3 kWh for filament drying, 1.52e-4 kWh for glue preparation, 5.208e-3 kWh for EGaIn synthesis, and 2.7e-2 kWh for 3D printing. Recycling involved 135.46 g water for dissolution and 20 g of 2 wt% NaOH solution for liquid-metal recovery. Recovered materials were about 0.848 g EGaIn and 1.16 g PVA, with reprocessing energy of 2.556e-5 kWh for grinding and 4.243e-4 kWh for filament extrusion (Yan et al., 29 Jul 2025).

The conventional FR-4 baseline inventory included 2.228 g FR-4 and 0.628 g solder paste, with 7.0e-4 kWh for soldering and 6.68e-3 kWh for CNC milling. Eight environmental indicators were evaluated: Acidification Potential, Eutrophication Potential, Freshwater Aquatic Ecotoxicity Potential, Global Warming Potential, Human Toxic Potential, Photochemical Ozone Creation Potential, Abiotic Depletion Potential (fossil), and Ozone Layer Depletion Potential (Yan et al., 29 Jul 2025).

For DissolvPCB with in-lab recycling, the reported impact values were: $0.7\ \text{mm} \times 0.7\ \text{mm}$3 kg CO$0.7\ \text{mm} \times 0.7\ \text{mm}$4 eq for GWP, $0.7\ \text{mm} \times 0.7\ \text{mm}$5 kg CFC-11 eq for ODP, $0.7\ \text{mm} \times 0.7\ \text{mm}$6 kg NO$0.7\ \text{mm} \times 0.7\ \text{mm}$7 eq for POCP, $0.7\ \text{mm} \times 0.7\ \text{mm}$8 kg SO$0.7\ \text{mm} \times 0.7\ \text{mm}$9 eq for AP, $0.9\ \text{mm} \times 0.9\ \text{mm}$0 kg P eq for EP, $0.9\ \text{mm} \times 0.9\ \text{mm}$1 kg 1,4-DCB eq for FAETP, $0.9\ \text{mm} \times 0.9\ \text{mm}$2 kg 1,4-DCB eq for HTP, and $0.9\ \text{mm} \times 0.9\ \text{mm}$3 kg oil eq for ADP. The paper states that ODP, HTP, EP, and FAETP were reduced to around 40–70% of conventional values, while ADP, AP, and GWP were reduced by about an order of magnitude. Under the study assumptions, DissolvPCB outperformed the CNC-milled FR-4 baseline across all eight metrics (Yan et al., 29 Jul 2025).

The interpretation offered in the paper is not that DissolvPCB eliminates impact. It still requires PVA, gallium, indium, printing energy, dissolution water, NaOH, and re-extrusion energy. The claim is instead that local disassembly, direct separation, and reuse of components and materials can lower total life-cycle impact relative to a prototyping workflow built around effectively non-recyclable FR-4 boards (Yan et al., 29 Jul 2025).

6. Demonstrations, limitations, and relation to adjacent reuse approaches

Three prototype demonstrations establish the breadth of the method. The first is a recyclable Bluetooth speaker with an ESP32-WROOM-32E (8MB) breakout board, UART interface IC, audio amplifier, DAC, voltage regulator, and 31 two-terminal SMD components. It used a double-layer PCB generated from KiCad through the software pipeline. Recycling recovered 99.4% of 10.13 g PVA and 98.6% of 3.64 g liquid metal (Yan et al., 29 Jul 2025).

The second prototype is an electronic fidget: a cubic gadget with a joystick on top, one SMD LED on each of four vertical faces, an ATtiny microcontroller, and nine peripheral components. It was fabricated as a single-piece 3D-printed cubic substrate with channels routed across five surfaces and sockets integrated into a non-planar form. Recovery yielded 99.1% of 29.92 g PVA and 97.7% of 2.09 g liquid metal (Yan et al., 29 Jul 2025).

The third prototype is a 4D printed three-finger gripper that deforms under electrical activation and can grip a cup. Here EGaIn channels function not only as conductors but also as embedded heaters for Joule-heat-driven morphology change in the surrounding PVA matrix. Recycling recovered 98.7% of 6.68 g PVA and 99.0% of 6.12 g liquid metal (Yan et al., 29 Jul 2025).

The principal limitations identified by the paper are geometric, environmental, and procedural. Because the minimum trace width is 0.7 mm, boards may need larger footprints than FR-4 boards for equivalent netlists; the paper notes that for a 0.5 A application, a 1 oz/ft² copper board may need a recommended trace width of 15 mil, about half the width required by DissolvPCB. Water sensitivity limits direct use in wearables, outdoor devices, or humid and wet environments without protection. Component compatibility requires deeper study beyond the demonstrated packages, especially for moisture-sensitive parts. Software support does not yet include native 3D EDA or automatic routing over arbitrary 3D surfaces. Manual EGaIn injection and manual glue sealing remain part of the workflow, and the authors identify automation via syringe extruders as a future direction (Yan et al., 29 Jul 2025).

A useful contrast is provided by “PCB Renewal: Iterative Reuse of PCB Substrates for Sustainable Electronic Making” (Yan et al., 18 Feb 2025). PCB Renewal also targets sustainable electronics making, but it reuses existing FR-4 boards by selectively depositing conductive epoxy into isolation grooves and then re-engraving new traces, rather than dissolving the substrate. DissolvPCB instead redesigns the assembly stack around PVA and EGaIn so that the board itself can be intentionally disassembled in water and major material classes recovered directly. This suggests two distinct but adjacent paradigms in circular electronics: remanufacture of existing FR-4 substrates through conductive restoration (Yan et al., 18 Feb 2025), and fully recyclable additive prototyping through soluble substrates and liquid-metal conductors (Yan et al., 29 Jul 2025).

Taken together, these features place DissolvPCB at the intersection of additive manufacturing, liquid-metal electronics, sustainable HCI, and prototyping infrastructure. Its central technical claim is not only that a PCB-like assembly can be printed and used, but that its materials, conductor geometry, component attachment strategy, and software flow can be coordinated so that fabrication and recycling become parts of one reversible prototyping system (Yan et al., 29 Jul 2025).