I. Abstract of the Paper’s Core Research

This study formulated a PVP–DEG‑modified, co‑solvent EGaIn liquid‑metal ink and systematically optimized the entire process chain, including ultrasonic atomization and sheath‑gas focusing of aerosols. High‑resolution microscale liquid‑metal sensing circuits were directly written onto a variety of flexible substrates, such as polyimide (PI), 3M adhesive tape, silicone rubber, paper, and plant leaves. By tailoring strain distributions through diverse patterning strategies—including straight lines, S‑waves, L‑shaped sinusoids, and cross‑grid arrays—comprehensive validation was achieved across microstructural morphology, electrical performance, tensile sensing, and human physiological monitoring. The key findings are summarized below:

• A specialized nano‑ink system for stable aerosol‑based printing of liquid metals

PVP breaks down EGa into sub-micron particles, inhibiting sedimentation; the high-boiling DEG regulates volatility, preventing print‑out overflow. The ink remains stable for dozens of hours without layering or clogging. Compatible with ultrasonic atomization, it enables ultra‑fine conductive traces as narrow as 12 μm, representing the highest‑precision liquid‑metal printing solution currently available.

• Precise geometric control of the track geometry resolves the trade-off between sensitivity and measurement range.

S‑wave architecture: minimum detectable strain of 0.0033% (micron‑scale microdeformations), with a maximum tensile strain of 230%, designed to capture subtle physiological signals such as pulse and micro‑expressions.

L sinusoidal composite material: detection limit 0.1%, ultimate tensile strength 660%, suitable for joints undergoing large-amplitude motion;

60 μm linear device: Balances a low detection limit of 0.0067% with a wide measurement range of 360%, making it suitable for universal sensing applications.

• Ultrafast dynamic response and ultra-long cycle stability

At 0.5% strain, the response time is 53 ms, and the recovery time is 62 ms; after 350 cycles at 50% of maximum strain, the signal baseline shows no significant drift, and the resistance exhibits excellent linearity across both high and low deformation ranges.

• Multi‑substrate co‑formation, compatible with fully flexible applications.

Aerosol-based non-contact deposition enables direct printing of liquid-metal sensor arrays onto adhesive surfaces, silicone, nylon gloves, and plant‑like curved substrates, eliminating the need for transfer‑printing processes.

• Precise, all‑scenario acquisition of human physiological signals

When attached to the skin, it can clearly detect a variety of signals, including the three‑phase pulse waveform, facial micro‑expressions, vocal‑cord vibration during speech, large‑amplitude bending of the fingers and knees, and running gait, enabling integrated wearable health monitoring.

II. Complete Process Flow and Core Advantages of the Aerosol-Spraying Liquid Metal Sensor Technology

• Complete AJP Liquid Metal Fabrication Process

1) EGa liquid metal, anhydrous ethanol, and PVP are subjected to high-energy ultrasonic disruption to prepare a submicron LM particle dispersion.

2) Add 10% by volume of DEG to adjust volatility and rheology, yielding a stable, atomizable ink.

3) Plasma activation of the PI/viscose substrate enhances adhesion;

4) Aerosol ultrasonic atomization, with adjustable carrier and sheath gas flow rates, enables direct writing of various sensor patterns ranging from 12 to 1000 μm.

5) Low-temperature drying at 90°C, followed by roller pressing, breaks the liquid metal oxide layer and activates the conductive pathway.

6) After encapsulation, a tensile testing platform is set up, and human‑wearable experiments are conducted to calibrate the sensor’s performance.

• Five Core Advantages of Aerosol-Based Processes (Comparison with Inkjet, Screen Printing, and Lithography)

Micrometer-scale liquid-metal direct-writing scheme without masks

Inkjet and screen printing achieve minimum line widths exceeding 100 μm, whereas lithography cannot directly process liquid metals; AJP enables stable fabrication of ultra-fine 12-μm traces, enabling high‑stress‑concentration precision sensing structures.

Universal ink compatibility, expanding the nanomaterial printing system.

The same atomization platform is compatible with a variety of inks, including liquid metals, carbon nanotubes, graphene, and metal/ceramic particles, enabling rapid development of multifunctional flexible electronics.

All‑flexible / Non‑contact conformal deposition on irregular substrates

3–5 mm safe printing distance; does not compress or damage ultra-thin silicone, adhesives, or fabrics; enables direct patterning of sensing circuits on curved and irregular substrates.

Digital pattern free design, rapid iteration of sensing structures

With CAD, you can switch between straight lines, wavy patterns, and composite grids with a single click, enabling you to complete comparative tests of multiple‑range sensors in a single day—eliminating the need for costly photolithographic mask fabrication.

Low-loss additive manufacturing, with liquid metal deposited only in the sensing region.

No full‑area coating is required, significantly reducing the consumption of precious‑metal liquid‑metal feedstock and substantially lowering the cost of small‑batch research prototypes.

III. Five Core Technological Innovations of This Study

Innovation 1: PVP-DEG binary liquid-metal aerosol-specific ink

By leveraging the dual mechanisms of PVP particle dispersion and DEG‑induced evaporation control, we have overcome the challenges of atomization and print‑outflow inherent in pure EGaIn, enabling long‑term, stable aerosol jetting.

Innovative 2-micron-level high-precision forming process for liquid metals

Breaking through the bottleneck in liquid-metal printing precision, with conductive lines as narrow as 12 μm, and leveraging microscale patterning to concentrate stress, it enables the detection of minute deformations at the ten‑thousandth‑of‑a‑millimeter level.

Innovation 3: Geometric Regulation Simultaneously Balances Sensitivity and Stretch Range

A composite sensing architecture integrating S‑waves and L‑mode Lamb waves is designed, enabling bidirectional regulation of stress concentration and dispersion to overcome the inherent trade-offs in the performance of flexible sensors.

Innovation 4: Low-Temperature Roll-Pressing for In-Situ Activation of Conductive Liquid Metal Pathways

After printing, the thin-film particles are isolated by an oxide layer; simple roll‑pressing then breaks this oxide film, forming a continuous liquid‑phase conductive network—eliminating the need for high‑temperature sintering and making it compatible with all thermally sensitive flexible substrates.

Innovative 5-in-1 Integrated Wearable Physiological Monitoring Platform for Multiple Scenarios

A single liquid-metal sensing patch can simultaneously capture micro‑pulses, facial expressions, and large‑amplitude limb movements, enabling continuous multi‑dimensional monitoring of human health.

IV. Existing Process Challenges and Corresponding Solutions for Overcoming Them

Challenge 1: Pure liquid metals have high viscosity and tend to agglomerate, leading to nozzle clogging in conventional printing equipment.

✅PVP ultrasonic fragmentation generates nanoparticles, while DEG prevents droplets from evaporating too rapidly, ensuring the ink remains stably atomized for dozens of hours without clogging the print head.

Challenge 2: Liquid-metal printed lines are thick, making it impossible to capture minute deformations at the micrometer scale.

✅ The sheath‑gas‑focused beam achieves 12‑μm ultra‑fine lines, with microstructures that concentrate stress and a detection limit as low as 0.0033% strain.

Challenge 3: High-sensitivity devices have a narrow strain range, and wide-range sensors cannot detect micro‑vibrations.

✅ Differentiated wave‑pattern composite design, tailored to meet both micro‑sensing and high‑stretch dual‑application needs, balancing two core requirements.

Challenge 4: After printing, the oxide layer on the surface of the liquid metal particles is insulating and lacks electrical conductivity.

✅ The room-temperature roll‑pressing process fractures the oxide layer, enabling interconnection of the internal liquid metal to form a highly conductive pathway, all while maintaining low temperatures that avoid damaging the flexible substrate.

Challenge 5: Conventional microfabrication is limited to rigid, planar substrates; fabrics and silicone elastomers cannot be fabricated in large quantities.

✅ Aerosol-based non-contact printing enables direct fabrication on adhesive, nylon, and curved substrates, eliminating the need for transfer or complex lamination processes.

V. Future Diversified Industrial Application Scenarios

1 All-day, non-invasive wearable health-monitoring patch

Wrist and neck pulse monitoring, facial emotion recognition, and voice‑vibration sensing enable continuous home‑based screening for chronic diseases.

2 Soft-Robotics Biomimetic Electronic Skin

A conformal-printed liquid-metal sensor array with a curved silicone outer layer simultaneously detects subtle touches and large-scale limb deformations.

3 Smart Glove Human-Computer Interaction Device

High-precision strain sensing at the fingertips and finger joints, VR/AR gesture recognition, and remote medical‑procedure terminals.

4 Sports Rehabilitation Gait Monitoring Equipment

Flexible sensing straps for the knees and ankles capture real-time data on running, jumping, and joint flexion, supporting quantitative assessment in rehabilitation training.

5 Structural Health Monitoring of Curved-Surface Industrial Structures

Sensing films are adhered to the irregular curved surfaces of pipelines and fans to enable real-time detection of minute cracks and tensile deformations in structural components.

6 Flexible Electronics R&D Platform for Universities

Desktop aerosol‑based equipment enables the rapid fabrication of liquid metals and a variety of carbon‑based strain sensors, without requiring a cleanroom lithography production line.

Image and Text Guide

Figure 1. Integrated patterning strategy for liquid-metal AJP:

• a)   Schematic of the AJP process and sensor fabrication: The bulk liquid metal (EGaIn) is ultrasonically fragmented into nano- to submicron-sized particles within an ethanol solution containing PVP (with a Ga₂O₃ shell encapsulating the LM core and surface‑adsorbed, degraded PVP molecules). A high-boiling-point solvent, DEG, is added to suppress volatilization and splattering. The resulting mixture undergoes micro‑porous ultrasonic atomization and is focused into a micro‑beam by N₂ gas, which is precisely deposited onto the substrate by the print head. On the right, the printed device is shown attached to an arm, exhibiting a resistance–time response.
• b)   LM patterns printed via AJP on various substrates (scale bar: 2 mm): glass, silicon wafer, PI film, paper, leaves, nitrile gloves, PDMS film, and 3M adhesive tape, demonstrating the process’s broad substrate compatibility.
• c)   This work addresses the trade-off between pattern resolution and the sensor range/detection limit ratio for various stretchable strain sensors (ε _emax /ε _emin ) "Comparison: This study (★This Work) achieves the highest ε with a minimum resolution of approximately 12 μm." _emax /ε _emin   Compared with the LM, Ag, CNT, graphene, and other approaches reported in the literature, this method demonstrates significantly superior performance.

Figure 2. Systematic optimization of the physicochemical properties of liquid metal ink (LMI) as a function of PVP concentration:

• a)   SEM images of LMI after ultrasonic disruption for 2 h (left: pure ethanol; right: ethanol + 10 mg/mL PVP, scale bar 1 μm): the addition of PVP results in smaller particles with a narrower size distribution.
• b)   Variation of average particle size with PVP concentration at different ultrasonic treatment times (30–120 min): the particle size first decreases and then increases, with LMI@10P exhibiting the smallest average particle size.  291 nm (A PVP layer that is too thick will hinder further comminution);
• c)   Relative absorbance ΔA/A₀ as a function of time: LMI@0P begins to stratify after approximately 46 minutes and achieves complete sedimentation within 2.2 hours; PVP at concentrations ≥ 10 mg/mL remains stable for up to 20 hours, while at ≥ 15 mg/mL it remains stable for up to 35 hours.
• d)   Shear viscosity curve: EGaIn exhibits shear-thinning behavior, while LMI behaves as a Newtonian fluid.
• e)   Viscosity increases linearly with PVP concentration (y = 0.10x + 1.81, R² = 97.77%), and the EGaIn value is 18.99 mPa·s, with the maximum ultrasonic atomization viscosity limit also indicated.
• f)   As viscosity increases, both the atomization rate and the aerosol concentration decrease.
• g)   The aerosol number‑average particle diameter is approximately 1.00 μm, while the mass‑average diameter is about 1.33 μm; both are essentially unaffected by viscosity (approximately three times the average particle diameter).
• h)   Surface tension increases slightly with PVP, but its maximum value is only about 4% of that of EGaIn (624 mN/m).
• i)   The contact angles of LMI@0P and LMI@30P on both glass and PDMS are less than 90° (20.2°–53.8°), whereas those of EGaIn reach as high as 121.1°–131.5°, demonstrating that LMI can wet and deposit well on a variety of substrates.

Figure 3. Regulation of AJP printing accuracy by ink properties:

• a)   Optical micrographs of printed lines at a sheath gas flow rate of 60 sccm for different DEG concentrations (scale bar: 100 μm): 0% — severe spattering; 5% — noticeable improvement; 10% — smooth lines with no scattering (optimal); 15% — increased line width and reduced precision.
• b)   3D morphology and   c)   Cross-sectional profile (LMI@30P, sheath gas 100 sccm): the line height is approximately 1.659 μm; based on this, the line width Lw and the equivalent thickness are extracted.
• d)–g)   Maximum spreading diameter D at different sheath gas flow rates (20–100 sccm) _max , line width Lw, normalized line width Lw/Lw₁, and normalized D _max /D_max1  As viscosity varies: increased viscosity suppresses droplet impact and spreading, leading to a narrower line width (at the highest viscosity, the line width decreases to 71.13%–81.31% of the baseline, consistent with theoretical predictions, with a linear fit R² = 98.40%); increasing the printing speed to 13.5 mm/s enables the formation of the thinnest lines.   12 μm   Continuous lines.

Figure 4. Electrical and electromechanical sensing properties of AJP‑fabricated liquid metal patterns:

• a)   SEM images of the iron rod before and after roll‑press activation (scale bar: 1 μm): After drying, it consists of an agglomeration of insulating particles; rolling disrupts the oxide layer, allowing the underlying lithium metal to extrude and coalesce into a continuous conductive film.
• b)   Electrical conductivity σ and relative conductivity decrease with increasing PVP concentration (as the insulating layer thickens). After balancing print accuracy and electrical performance, LMI@15P was selected, yielding σ = 4.84×10⁵ S/m (with the resistance of a 30 mm trace dropping to 73.64 Ω after nine prints).
• c)   Reversal and   d)   The bending-induced ΔR/R₀ is essentially stable;
• e)   Tensile sensing curves: three segments—0–100%, 100–250%, and 250–360%—with GF values of   0.62 / 1.33 / 2.08 , electrical conductivity fails after strain exceeds 370%;
• f)   Microstrain (0.0067%–1%) cyclic signals are clearly discernible, with the minimum detectable displacement as low as   2 μm (0.0067%) ；
• g)   Response time at 0.5% strain  53 ms , Recovery Time  62 ms；
• h)   At 50% strain, signals at different tensile rates of 1.5–19.5 mm/s (0.05–0.65 Hz) are stable;
• i)   At 50% strain, the signal remains stable over 350 cycles (with waveform amplification consistent between 209–241 s and 2540–2577 s), and the I–V characteristics exhibit ohmic linearity throughout.

Figure 5. Influence of line width and pattern design on sensing performance:

• a)–c)   Five line-width (60–1000 μm) linear sensors: the narrower the line width, the higher the sensitivity and the lower the detection limit (60 μm: ε _emin   0.0067% / ε _emax   360%), the wider the line width, the larger the measurement range (1 mm: ε _emin   0.2% / ε _emax   660%) — The linewidth can only be adjusted independently for a single parameter;
• d)–f)   Eight patterns (including five basic types—straight line, square wave, sawtooth, sine wave, and S‑wave—and three composite patterns—L‑sinusoid, S‑sinusoid, and cross‑hatch—with a minimum line width of 60 μm): The S‑wave achieves the lowest detection limit of 0.0033% through stress concentration, while sine‑type patterns attain a wider measurement range via stress dispersion (ε _emax   Up to 530%);
• g)   L-sinusoid (straight line + sine) combination: ε _emin   0.1%, ε _emax   Upgrade to  660%； h)   S-sinusoid: ε _emin   0.005%, ε _emax   290%; i)   Cross-grid: In the high-strain region, sensitivity surpasses that of the linear and triangular waveforms, ε _emax   Up to 620%. The composite pattern achieves a balanced optimization of “low detection limit + wide dynamic range” under narrow line widths.

Figure 6. Human motion detection application (60 μm linear LMSS attached to various body parts):

• a)–b)   Radial artery pulse: single cycle duration 0.84 s (approximately 71.4 beats/min), with three distinct peaks—P (main wave), T (tidal wave), and D (dicrotic wave)—clearly identifiable (ε _max   ≈ 0.049%);
• c)–d)   Carotid pulse: Signal intensity is approximately eight times that of the radial artery (ε _max   ≈ 0.3%);
• e)–f)   Jaw movement during the pronunciation of “Liquid Metal”: Multimodal signals can distinguish between different syllables (ε _max   ≈ 9.03%);
• g)–i)   Facial microexpressions: blinking, raising the eyebrows, and smiling all elicit distinct responses (ε _max   ≈ 13.44%–21.34%);
• j)–l)   Joint flexion: Flexion of the index finger (ε _max   ≈ 36.37%), the elbow (≈ 117.99%), and the knee (≈ 148.76%), with the response increasing as the bending angle grows;
• m)   Limb movements at the knee—such as jumping in place, squat jumps, stepping, high‑knee running, and walking—generate distinct waveform signals, enabling the recognition of movement states.