Soft, miniaturized, wireless olfactory interface for virtual reality

The study addresses the need for robust, wearable olfactory feedback to complement visual, auditory, and haptic interfaces in VR/AR. Existing smell-generation systems rely on bulky, tethered atomizers or room-scale emitters with slow switching, limited channels, and poor programmability, which hinders integration with immersive systems. The authors aim to develop a soft, skin-interfaced or mask-based olfactory interface that is miniaturized, lightweight, wireless, and programmable, capable of precise and rapid odor intensity control, long-duration operation, and safe, comfortable wear. The central hypothesis is that integrating miniaturized odor generators with precise thermal control and active cooling into flexible electronics will enable fast, localized, multi-odor delivery suitable for real-time VR/AR and assistive applications.

The paper reviews advances in human–machine interfaces emphasizing sensory feedback and notes olfaction’s importance alongside vision and audio. Prior work has focused on odor detection with rigid and flexible gas sensors capable of decoding complex mixtures. In contrast, odor generation remains nascent: existing systems are room-scale or built into bulky VR headsets, often wired, with slow response, limited odor types, large liquid reservoirs, and maintenance issues. Wearable OGs based on commercial atomizers are clumsy, rigid, and unsuitable for high-channel, lightweight, flexible arrays. The authors outline desired features for new systems: soft, skin-integrated formats; many odors with adjustable concentration and long operation; wireless, programmable control; biocompatible, easy-access sources; and rapid response with accurate concentration control, citing multiple prior attempts and their limitations (e.g., Supplementary Table 1 and refs. 13–26).

System architecture: Two wearable formats were designed. Device 1 is a skin-integrated unit above the upper lip with 2 odor generators (OGs). Device 2 is a flexible face mask integrating a 3×3 array (9 OGs). Both include a flexible control panel with MCU, Bluetooth, power management (12 V and 3.7 V batteries, LDO, DC–DC boost to 3.3 V, 5 V, 16 V), MOSFET-switched heaters, H-bridge-driven electromagnetic actuators, and thermistor feedback. Wireless ranges reached 2.8 m (Device 1) and 5.9 m (Device 2). The mask shell is 3D-printed TPU; Device 1 uses PDMS encapsulation and an adhesive layer.

Odor generator (OG) design: Each OG comprises (1) odor source: food-grade paraffin mixed with safe liquid perfumes; (2) thermal actuator: PI-supported Au heater with embedded high-B thermistor for precise temperature control; (3) mechanical actuator: a cantilever with a Cu electromagnetic coil interacting with a permanent magnet to actively lift/push the heater for rapid heat dissipation (active cooling). A soft silicone (PDMS) frame forms the OG chamber, providing travel space and conformability. Miniaturized OGs allow arrays with distinguishable frame colors per scent.

Electronics and control: The MCU reads thermistor voltages via ADC, converts to temperature, and modulates heater power via MOSFET switching (up to 50 Hz). The H-bridge reverses coil polarity to move the cantilever for cooling/heating transitions. Device 2 uses four 8-bit shift registers to expand GPIO control for 9 heaters and 9 H-bridges, plus an 8-channel multiplexer for multi-thermistor sensing. For stability, simultaneous OG operation is limited to 5 to avoid ripple voltage and electromagnetic coupling affecting the MCU.

Performance optimization: Parameters optimized include thermistor B-value (3435, 3950, 4250 K), heater power (0.12–0.37 W), and actuator lift height via coil power (0.85–1.20 mm corresponding to 73–188 mW). Airflow effects on temperature stability were characterized at 45 °C setpoint with perpendicular wind up to 6.6 m/s. Active cooling efficacy was quantified by comparing cool-down times with and without cantilever actuation.

Odor generation and sensing: Ethanol was used as a model volatile for quantitative tests. A commercial ethanol sensor positioned 1 cm above a working OG in a controlled enclosure measured concentration versus OG temperature steps (45→50→55→60 °C). Distance and airflow effects, and in-mask measurements (Device 2) under two conditions (post-release wearing vs stationary) were recorded. Human sensory tests assessed odor recognition across heating temperatures and recovery times for nine odors; additional tests evaluated recognition of the presence/absence of odors at room temperature.

User demonstrations: Applications included 4D movie scenes (dynamic temperature schedules to simulate approach of floral source), smell-message communication for vision/hearing-impaired users (training at 10 vs 60 minutes), emotion modulation with 30 odors, memory recall concepts, and VR/AR interactions (virtual garden). A soft motion capture system (cloth-integrated accelerometers with Bluetooth) enabled real-time avatar control in VR and synchronized olfactory cues; it was validated for joint angle tracking and endurance (>13,500 cycles), with >100 min battery life (80 mAh).

Fabrication: Heaters were fabricated on 25 µm PI laminated to glass. Au/Cr (200/40 nm) was sputtered, patterned by photolithography (AZ5214) and wet etch; a PI encapsulation layer (~2 µm) was spin-cast and patterned by RIE to expose contacts. A 0402 NTC thermistor (B=4250 K) was silver-paste bonded at the heater center; a Cu coil was glued atop. Assemblies were mounted to a PET platform with a central magnet and PDMS ring chamber; 2 mg of odorous paraffin was added and melted at 60 °C for sealing. Device 1 and Device 2 control boards were fabricated as flexible PCBs (Cu 10 µm, Au 50 nm), components soldered with low-temp paste, and encapsulated in PDMS; Device 2 panel was mounted inside a TPU mask. Weights: OG 1.8 g; Device 1 7.4 g; Device 2 control panel 26.3 g; power system mass 107.6 g.

Modeling and characterization: Finite element analyses (Abaqus) assessed mechanical strains under bending/twisting and thermal behavior (heat transfer elements, natural convection h=10 W/m²K). IR thermography validated temperature fields. Stability tests included >7000 thermal cycles (45↔50 °C), vibration up to 10 Hz, and Device 1 bending for 2000 cycles at 0.33 Hz and 40° while monitoring electrical stability.

• Miniaturized, flexible OGs with integrated thermistor control and electromagnetic active cooling achieved rapid, precise odor delivery near the nose in two wearable formats (skin-mounted 2-OG Device 1 and 9-OG mask Device 2).
• Wireless control distances: 2.8 m (Device 1) and 5.9 m (Device 2). OG array density up to 0.88/cm³ in Device 2; response time as short as 1.44 s at 50 °C due to minimal source–nose distance.
• Temperature control: Using a high-B thermistor (B=4250 K) reduced temperature fluctuation at 60 °C from 3.8 °C (B=3435 K) to 0.5 °C, and at 45 °C from 2.74 °C to 0.3 °C. At 45 °C setpoint, airflow up to 6.6 m/s increased fluctuation to ~2 °C; 45 °C was selected as the optimal standby temperature.
• Heater power optimization: 0.25 W balanced fast response and low overshoot; higher power (0.37 W) yielded limited response improvement but larger overshoot (e.g., to 53.3 °C when targeting 50 °C). Cool-down response was largely independent of heater power.
• Mechanical actuator optimization: Coil input of 108 mW produced a 0.99 mm lift height and short response times, balancing performance and power. Lift heights 0.85–1.20 mm gave heating response times 4.72→3.62 s (45→60 °C steps). Sub-stable overshoot remained ~1.2–2.1 °C across heights.
• Active cooling efficacy: Cool-down times improved dramatically with the cantilever actuator versus passive cooling: 50→45 °C in 0.8 s vs 3.9 s; 55→45 °C in 1.37 s vs 6.4 s; 60→45 °C in 1.5 s vs 8 s.
• Odor output (ethanol model): At 1 cm from OG, concentrations reached ~531 ppm (50 °C), 2821 ppm (55 °C), and 4531 ppm (60 °C), with rise times 4–8 s and recovery times 40–129 s; human smell threshold ~80 ppm. Greater distance increased detection delay (1 cm: 1.2 s; 3 cm: 9.1 s; 5 cm: 15.6 s). Airflow accelerated recovery (129 s→9 s at 6.61 m/s).
• Human sensory performance: With 9 odors, recognition rates averaged 0.93 (range 0.73–1) across 45–60 °C heating. For smell-message delivery, average recognition of 9 messages was 55.6% after 10 min training versus 87.7% after 60 min.
• Recovery and accumulation: In-mask ethanol tests showed human breathing shortened recovery (1.2 min wearing vs 3.1 min stationary) at 60 °C. In volunteer tests with 9 odors, average recovery was 92.1 s (wearing) vs 77.1 s (stationary), with minty as an exception.
• Longevity and capacity: Odor duration increased with lower temperature, stronger scents, higher perfume ratio, and larger mass; up to 52 h for 30 mg paraffin/perfume mixture (10:3) at 60 °C. Larger wax mass increased response times (2→30 mg increased 45→60 °C heating time 3.9→12.1 s). Wax overflow time depended on mass, placement angle, and temperature (e.g., 30 mg at 60 °C, 90° placement: 40 s; 2 mg: >8 h).
• Stability and safety: OGs endured >7000 thermal cycles (45↔50 °C) over ~39 h with stable performance; vibration up to 10 Hz caused negligible impact; Device 1 maintained stable operation over 2000 bending cycles (voltage 3±0.02 V). Thermal imaging showed ambient air at 1.5 mm from a 60 °C OG remained near room temperature; user nose distances exceeded 1.5 mm in all 11 volunteers. Device 1 skin interface temperature was ~32.2 °C with a 23 mm nose gap.
• System constraints: In Device 2, to ensure circuit stability, simultaneous operation was limited to 5 OGs due to ripple voltage and coupled EM fields.
• Demonstrations: Effective 4D movie scent modulation, VR garden interaction with synchronized odors, assistive smell-messaging for sensory-impaired users, potential emotion modulation (e.g., joy probabilities up to 65% for certain scents) and memory recall concepts. A soft motion capture system provided accurate joint tracking (>13,500 cycles) and >100 min battery life to enable VR interactions.

The work demonstrates that precise thermal control with high-sensitivity thermistors, combined with an active electromagnetic cooling mechanism, enables rapid, programmable, and localized odor delivery from miniaturized flexible OGs. By optimizing heater power (0.25 W) and actuator lift (0.99 mm at 108 mW), the system achieves fast rise and cool-down times with minimal overshoot and low power consumption, addressing the historical limitations of slow switching and poor concentration control in olfactory interfaces. Quantitative ethanol measurements and human sensory tests confirm that operating temperatures between 45–60 °C produce odor concentrations well above human thresholds, with short response times and tunable intensity. Stability under prolonged cycling, bending, and vibration, along with thermal safety margins at typical nose distances, support real-world wearability. The two wearable formats (skin-integrated and mask-based) offer complementary capabilities: ultra-short diffusion distance for fast response and multi-channel arrays for complex scent portfolios. Demonstrated applications in VR/AR, assistive communication, and emotion-related use-cases show the interface’s versatility. Constraints such as odor lingering and multi-OG interference are identified, motivating design refinements (e.g., better sealing, miniaturization) for faster odor switching and higher channel counts.

This study introduces soft, miniaturized, wireless olfactory interfaces that integrate arrays of flexible odor generators with precise thermal control and active cooling into wearable formats (skin-mounted and mask-based). The systems deliver fast, programmable, multi-odor feedback with high stability and safety, enabling immersive VR/AR experiences, assistive smell-based communication, and emotion-related applications. Key contributions include device/materials/electronics co-design, optimization of thermistor selection, heater power, and actuator lift height, and validation through quantitative sensing, user studies, and long-term stability tests. Future work should focus on miniaturizing OGs, improving airtightness to reduce odor persistence and cross-odor interference, expanding channel counts while mitigating electrical and electromagnetic coupling, and further personalizing odor portfolios for users with diverse thresholds and emotional responses.

• Odor persistence leads to delayed switching between scents; recovery depends on airflow, wearing conditions, and odor type.
• Larger paraffin masses extend odor duration but slow response and can shorten overflow times at certain orientations/temperatures.
• In the 9-OG mask, stable simultaneous operation is currently limited to 5 OGs due to ripple voltage and electromagnetic coupling affecting the microcontroller.
• Mask breathing holes can cause minor odor leakage; user-specific thresholds and recognition variability complicate standardized concentration coding for single-OG messaging.
• Recovery time results differ across ethanol sensor measurements and human tests for certain conditions, indicating environment- and method-dependent dynamics.