Portfolio sections
Portfolio Alexandre Kiss

↑↓ Navigate · Enter Open · Esc Close

research · Sept 2026 – present

Time-resolved switching in two-dimensional ferroelectrics

My master’s thesis at MIT studies how polarisation switches in sliding ferroelectrics. Their fast, low-energy switching makes them candidates for the efficient memory that AI and neuromorphic computing need.

Schematic van der Waals device stack with graphite gates, boron nitride dielectrics, a graphene sensor and a ferroelectric boron nitride bilayer.
Reversing the electric field shifts the boron nitride bilayer between AB and BA stacking and reverses its polarisation. The animation shows this change propagating as a domain wall. My thesis aims to measure that motion during switching. Layer thicknesses are not to scale. Concept after Yasuda et al., Science 372, 1458 (2021).

Summary view

Project details

  1. AB BA slide < 1 Å field P boron nitrogen

    Sliding ferroelectricity

    A parallel bilayer of boron nitride can be ferroelectric. Sliding one layer relative to the other by less than an ångström reverses its out-of-plane polarisation. Switching proceeds through the motion of a domain wall: as the wall moves across the device, the region behind it changes polarisation.

  2. read before read after read during? 10⁻⁴10⁻²110² wall speed, m/s hundreds of µm/s operando TEM 1000 m/s BN switching times > 10⁶ difference

    Domain-wall motion and measurement timescales

    Published measurements apply a pulse and read the device afterwards, so they show where the wall stopped, not how it got there. The speed estimates also disagree: a few hundred micrometres per second from operando electron microscopy of a twisted bilayer, and about a thousand metres per second inferred from switching times in boron nitride.

  3. 1 µm 20 µm bar 6 samples over 100 samples 05101520 ns a sample every 160 ps

    Device dimensions for time-resolved measurements

    I am exploring whether pinning helps explain the different reported speeds: a wall may move quickly between sites but spend longer waiting at them. With the oscilloscope’s 160 ps sampling, a wall moving one micrometre per nanosecond gives about six samples per micrometre. A twenty-micrometre bar would give over a hundred samples during one crossing.

  4. exfoliatepick up flakesrelease onto bottom gateassembled stack graphite gate BN dielectric graphene BN bilayer BN dielectric graphite gate after JoVE 221, e68230 (2025)

    Assembling the two-dimensional material stack

    The device follows a published design: graphite gates, boron nitride dielectrics, a graphene sensor and the parallel bilayer. It is built by the published method: flakes are exfoliated with tape onto silicon chips and chosen under a microscope, then a polymer stamp picks up each layer in turn and lays the stack on the bottom gate.

  5. e-beam lithographyCHF₃/O₂ etchCr/Au evaporationdefine the bar edge contact

    Graphene edge contacts and device patterning

    The published fabrication process defines the contacts by electron-beam lithography. A CHF₃/O₂ plasma exposes the graphene edge, followed immediately by chromium and gold deposition to form edge contacts. A second lithography and etching step patterns the long bar intended for the time-resolved measurement.

  6. device buffer, at the sample oscilloscope slow readout: broadened response fast readout: crossing resolved illustrative traces

    Reading the device during switching

    The current readout measures the device seconds after a pulse, while the domain wall may cross it in nanoseconds. I am considering a fast buffer amplifier placed millimetres from the sample. It would allow measurement during the pulse and help preserve the signal on its way to the oscilloscope.

AB stacking P BA stacking P slide < 1 Å boron nitrogen schematic, not to scale
Two sheets of boron nitride stacked in parallel. Sliding the top sheet by less than an ångström turns AB stacking into BA, and the out-of-plane polarisation flips with it.

My contribution

I started this thesis in September 2026. I am developing the concept for a fast readout and the devices needed to resolve a single switching event. The work is at the concept stage; there are no results yet.