Graduate Seminar: Hanseup Kim, Sept. 2, 2026

Main ideas, concepts, and lessons

• Backup seminar + research focus

  • Professor Hanseup Kim (graduate seminar) had planned a talk by Professor Florian Schulze, but Schulze became sick, so Kim gave the seminar instead.
  • The talk explains electrical and computer engineering research applied to unusual domains:
    • Agriculture (with ocean/space mentioned as part of the group’s broad context)
    • Microsystems

• Big “where is microsystems going?” vision

  • Kim argues that modern devices increasingly rely on small components assembled into systems, enabled by microtechnology.
  • His personal “dream” is a “Tinkerbell”-like system that is:
    • Intelligent
    • Autonomous
    • Able to create societally impactful micro-systems

• Five key components (high-level system requirements)

  • The talk claims there are “mainly five” components, emphasizing:
    • Power as the biggest issue
    • Energy harvesting / replacing batteries (self-powered operation)
    • Sensor-generated communication
    • Micro-actuation
    • Communication / electronics wake-up strategy for ultra-low-power operation

• Grand challenge framing (United Nations)

  • Kim links the work to the UN’s “grand challenges,” focusing on:
    • Environment
    • Food
    • Energy

• Nature-inspired agriculture defense mechanism

  • Plants defend against herbivores by emitting gas molecules (described as “screaming,” without sound).
  • These gases form a detectable chemical signal that helps locate herbivores and improve survival.
  • Engineering goal: build sensing systems that can “smell” these gases to enable:
    • Localized treatment instead of widespread spraying
    • Higher yields, healthier soil/crops, and lower costs

• Core technical problem

  • Farms/forests lack power grids, making powered sensors impractical.
  • Typical gas sensors consume power continuously, even when they do little most of the time.
  • Kim’s ambition: continuous sensing without continuously powered sleeping sensors, so sensor lifetime can increase dramatically.

---

Method / technical approach (detailed bullet points)

A) System-level concept: “wake only when gas arrives”

• Target behavior
  • Enable continuous sensing capability while keeping sensor circuits nearly off most of the time.
• Key idea
  • Keep the sensor in a sleep state (very low power / near-zero current) until the target gas triggers a switch.
• Analogy
  • Like waking only when a mosquito appears—the system should “wake up” only on relevant chemical stimuli.
• Practical implication
  • Pair with energy harvesting so the system can be self-powered, reducing or eliminating battery replacement.

B) Ultra-low-power target metric

• Reference
  • 10 nW is cited as leakage current in CMOS devices even when “off” but still connected.
• Target
  • Aim for sensor consumption well below 10 nW, ideally ~1 nW or pW-range for “near-zero power” operation.

C) Device mechanism: molecular bridge acting as an electrical switch

• Why straightforward sensing is hard
  • Gas molecule arrival is random.
  • A single fixed “gap” may yield inconsistent sensing at low concentration.
• Baseline approach
  • Use a nano-gap whose size is matched to the target gas molecule.
• Molecular bridging
  • When a target molecule enters and is captured by designed chemistry on both sides:
    • It forms a molecular bridge
    • The bridge enables electrical conduction through the previously non-conducting nano-gap
    • Conduction “switches” the circuit on, powering the rest of the electronics

D) Statistical robustness: percolation concept using arrays

• Percolation idea
  • At low concentration: connectivity is effectively absent; the chance of forming a conducting path is near zero.
  • Above a threshold: probability of a conducting path rises toward certainty.
• Implementation
  • Use many nano-gaps in an array, rather than relying on a single gap.
  • The ensemble behavior produces a more uniform turn-on threshold.
• Result
  • Reduced device-to-device variation in switching concentration (goal: devices turn on around a similar ppm level rather than widely scattering).

E) Threshold tuning and selectivity control

• Sensitivity / threshold
  • Kim claims device geometry and arrangement can tune sensitivity (discussed conceptually: thresholds like 70 ppm, 7 ppm, 4.2 ppm).
• Selectivity
  • Improved by choosing:
    • Target-molecule-matching gap size
    • Specific chemistry/linkers that preferentially bind the intended gas
• Chemistry collaboration
  • The work credits collaboration with chemistry professor Ryan Rupert for linker/functionalization chemistry.

F) Microfabrication / nanofabrication execution (key specs)

• Manufacturing
  • Fabricated in a cleanroom using micro/nanofabrication capabilities (named as the Utah nano fabrication facility).
• Gap sizing
  • A featured gap size is 5.2 nm, described as matching the target gas molecule’s length scale.
• Device architecture
  • An “open mouth”-like electrode structure defines an active sensing area between electrodes.

G) Scaling strategy to reduce required bias voltage

• Trade-off
  • Arrays improve threshold reliability but can demand higher bias voltage.
• Solution
  • Use 3D stacked arrays:
    • Stack multiple layers (example: five layers)
    • Parallel pathways reduce overall resistance
• Outcome
  • Bias voltage reduced to below what’s supported by a lithium-ion battery, while maintaining improved threshold and accuracy.

H) Packaging + real-environment testing (humidity robustness)

• Packaging
  • Built packaging to improve performance in humidity.
• Demonstration method
  • Gas jar testing:
    • Introduce target gas
    • Device resistance drops
    • A diode lights (showing switching)
    • Repeatable on/off behavior is demonstrated

I) Agriculture field deployment demonstration

• Natural “gas warning” simulation
  • Insects couldn’t be used (restriction).
  • Instead, simulate herbivore attack by cutting plants, which releases the relevant gas molecules.
• Deployment location
  • Tested in a sorghum research area in Puerto Rico (winter prevented Nebraska testing).
  • A U.S. government land farm conducting sorghum research is mentioned.
• System behavior in the field
  • Electronics start asleep.
  • When cutting releases target gas:
    • Sensor forms a molecular bridge
    • Power flows to the electronics
    • Electronics “wake up”
    • Sensor signals a computer in Salt Lake City
  • Kim reports observing:
    • Signals propagating/spreading as cutting occurs
    • Other plants potentially emitting related signals after detecting the first (not fully established science, but presented as “actual measurement”)

J) Intended downstream use with arrays + AI

• Array-based localization
  • Example discussed: a small array (e.g., 2 × 3 × 6) to infer where attack is occurring.
• AI inference
  • Use sensor activation order/timing to infer where herbivore attack likely occurs.

---

Speakers / sources featured (as named in the subtitles)

• Hanseup Kim (speaker; presenter)
• Professor Florian Schulze (originally scheduled speaker; became sick)
• Ryan Rupert (chemistry collaborator)
• John Deere (mentioned regarding device size constraints)
• United Nations (cited for the “grand challenges” framing)
• DARPA (cited as approving/encouraging the idea after submission)
• Nebraska / Lincoln, Nebraska (test location mentioned)
• Puerto Rico (field deployment location mentioned)