Osaka’s Programmable Heat Concept Can Point Infrared One Way—and Remember the Setting

Turn a blackened metal plate toward a fire and it absorbs radiation well. Heat the same plate, and it becomes a good radiator back into the same optical channels. For more than a century and a half, that pairing has been one of thermal physics’ most useful design rules. The new Osaka concept asks an engineering question that once sounded like a contradiction: can a surface listen for heat from one direction but speak it differently—and remember how it was told to behave?

A team led by Professor Koichi Okamoto and lecturer Shunsuke Murai at Osaka Metropolitan University says the equations support such a structure. Their design places a nanoscale grating of the phase-change alloy Ge₂Sb₂Te₅, known as GST, above a magneto-optical indium arsenide waveguide and a silver reflector. The grating couples mid-infrared light into a resonant mode. A transverse magnetic field breaks the symmetry between opposite propagation directions. Changing GST between amorphous and crystalline phases latches the nonreciprocal response on or off.

In numerical simulation, the strongest result appeared at a wavelength of about 13.24 micrometres, an incidence angle of 3 degrees from normal and a magnetic field of 1 tesla. The calculated magnitude of the nonreciprocity coefficient reached about 0.90: one angular channel was strongly absorbing while the reverse was close to detuned. The work was published online in Laser & Photonics Reviews on June 25.

The verbs matter. The team designed, calculated and proposed. It did not report fabrication, direct thermal-emission measurements or a prototype cooling a chip. Osaka Metropolitan University’s Japanese technical release explicitly lists experimental fabrication and verification, lower magnetic-field operation, GST switching inside an integrated structure and stable room-temperature performance as work still to be done.

First, what kind of “heat” is being programmed?

Heat moves in three familiar ways. Conduction transfers energy through matter—along a pan handle or across a silicon chip. Convection carries it with moving fluid—air from a fan or water in a pipe. Radiation carries energy as electromagnetic waves and requires no material path. Every object above absolute zero emits it; around room temperature most of that radiation lies in the infrared.

The Osaka proposal acts on the radiative channel. It does not steer phonons through a solid, pump hot air around a corner or make one side of an object colder by command. More specifically, it engineers a narrow spectral and angular set of long-wave infrared modes around 13.24 micrometres. Wien’s displacement law puts the peak wavelength of an ideal 300-kelvin blackbody near 9.7 micrometres, so 13.24 micrometres belongs to the same broad room-temperature thermal-infrared world but is not the whole spectrum.

“Programmable heat” is likewise a useful metaphor with boundaries. The information stored is the GST phase and therefore the device’s optical response—not a packet of heat held like a bit in memory. Energy is needed to switch GST. Directionality still requires magneto-optical bias; in the reported simulation that bias is 1 tesla. A permanent magnet might maintain a field without continuous electrical input, but the magnet and its integration do not disappear from the engineering problem.

The old rule: a good absorber is a good emitter

In 1860 Gustav Kirchhoff formalized the relation between absorption and thermal emission. In its modern directional form for an ordinary reciprocal material at thermal equilibrium, emissivity equals absorptivity for the same wavelength, angle and polarization. If a surface absorbs a particular optical mode perfectly, it emits strongly into that mode when hot. If it reflects that mode, it is a poor emitter there.

This equivalence is powerful. Engineers can measure or calculate absorption more easily than faint thermal emission and infer the corresponding emissivity. Selective emitters, infrared sources, radiative coolers and thermophotovoltaic systems all exploit the rule. It also explains why polished metal, a strong reflector, usually radiates less effectively than matte black paint at the same temperature.

But Kirchhoff’s familiar equality carries symmetry assumptions. Lorentz reciprocity says that, in a linear, time-invariant reciprocal medium, swapping source and detector leaves the transfer unchanged. A magnetic field applied to a magneto-optical semiconductor changes its electromagnetic response and breaks time-reversal symmetry. The same channel under the same field no longer has to behave identically in reverse.

Calling this a “violation of Kirchhoff’s law” is standard language in the research literature, but it does not mean nature has stopped keeping accounts. Generalized radiation relations—often expressed through Onsager–Casimir or adjoint symmetry—replace the simple same-channel equality. Energy conservation and the second law still apply. At full equilibrium there is no device that extracts unlimited work from bodies at one temperature; maintaining or switching the bias and moving useful heat remain part of the balance.

Inside the metagrating: three materials, three jobs

The proposed structure is thin but conceptually layered. At the top, GST is patterned into a grating—a periodic landscape smaller than the wavelengths it manipulates. The grating supplies momentum that lets an almost normally incident infrared wave couple into a guided-mode resonance rather than simply reflect. Resonance concentrates the optical field and increases interaction time.

Beneath it, indium arsenide performs the nonreciprocal work. InAs is a semiconductor whose free carriers respond to a magnetic field. In the transverse, or Voigt, configuration used by the design, the field makes the material’s permittivity tensor asymmetric. Waves travelling with opposite in-plane momentum then couple differently to the resonant mode. At selected wavelengths, one direction lands on a lossy resonance while the reverse direction misses it.

A silver film below prevents transmission and returns the field into the stack. With nowhere to transmit, incident energy is divided mainly between reflection and absorption. The patterned GST, InAs waveguide and mirror together form what the authors call a phase-change magneto-optical metagrating.

Part	Physical role	Engineering burden
GST grating	Couples infrared into resonance; switches optical constants between amorphous and crystalline states; holds that state without standby power	Must be patterned, switched and cycled without drift, damage or unwanted heating
InAs waveguide	Confines the resonant field and supplies magneto-optical nonreciprocity under bias	Needs controlled carrier density, low enough loss and a practical magnetic architecture
Silver reflector	Blocks transmission and helps form the resonant absorber	Must retain optical quality and adhere through fabrication and temperature cycling
Periodic geometry	Matches incoming momentum to guided modes near normal incidence	Nanometre-scale errors shift wavelength, linewidth and directional contrast
External field	Breaks time-reversal symmetry and tunes the resonance	The simulated 1-tesla field is large for a compact consumer device

Why three degrees matters

Previous magneto-optical proposals often produced their strongest contrast at grazing incidence, where the beam skims the surface. That looks attractive in an angular plot but wastes physical aperture: the projected collecting area falls with the cosine of the incidence angle. Coupling, alignment and packaging also become awkward, and emission may form a broad cone rather than a useful beam.

The Osaka design uses grating momentum and guided resonances to bring the effect to just 3 degrees from normal. A nearly face-on geometry intercepts much more radiation for a given footprint and is easier to place above an emitter or detector. “Near-normal” is therefore not a cosmetic record; it addresses a geometric loss that had limited usefulness.

The price is resonance. The headline result is narrow in wavelength and angle, sensitive to dimensions, optical constants, temperature and magnetic field. Resonance can be an advantage for a spectrally selective sensor or thermal emitter. It is a disadvantage if the application needs broad heat removal across a large spectrum and field of view. The right comparison is not “0.90 is almost perfect,” but “how much useful power crosses the complete operating bandwidth and aperture?”

How the setting is remembered

GST belongs to a family of chalcogenide phase-change materials that can be rapidly and reversibly rearranged between amorphous and crystalline solids. A short, intense pulse can melt and quench a region into a disordered amorphous state; a longer, milder heating step can crystallize it. Those atomic arrangements have different refractive indices, absorption and electrical resistance.

The principle travelled through rewritable optical discs and into phase-change memory. In a DVD-RW or Blu-ray medium, a laser writes regions whose reflectivity identifies stored data. In electronic phase-change memory, resistance supplies the readout. Because atoms remain in the chosen arrangement after the pulse ends, the state is nonvolatile.

In the Osaka simulation, amorphous GST allows the guided-mode field to overlap strongly with magneto-optically active InAs, preserving large nonreciprocity. Crystallizing GST changes its index and loss, pulling and damping the resonance so the directional contrast is suppressed. The material acts as a latched optical gate. It does not continuously compute a heat path; it stores a configuration that changes how a carefully selected infrared mode couples.

That distinction defines the next experiment. Researchers must integrate a heater or optical pulse that changes GST without damaging the nanograting or warming the whole device in a way that corrupts measurement. They must quantify write energy, switching time, intermediate states, retention, cycling endurance and drift. “No power to remember” is valuable, but it is only one line in the energy budget.

Two centuries of teaching heat and light new routes

Joseph Fourier’s 1822 theory turned heat conduction into mathematics. Michael Faraday showed in 1845 that magnetism can rotate the polarization of light—the founding observation of magneto-optics. Kirchhoff’s 1860 radiation theorem connected emission and absorption and introduced the ideal black body. Stefan, Boltzmann and Wien established how total output and peak wavelength depend on temperature. In 1900, Max Planck’s successful formula for the blackbody spectrum required energy quanta and helped open quantum physics.

The later story moved from describing radiation to designing its available states. Photonic-crystal ideas in 1987 showed that periodic structures can forbid or reshape optical modes, much as a crystal lattice shapes electron motion. Metamaterials and metasurfaces then used subwavelength geometry to tailor amplitude, phase, polarization, spectrum and direction. Thermal photonics applied those tools to incoherent light emitted by hot matter.

Nonreciprocity developed in parallel. Faraday rotators and optical isolators protected lasers by allowing light to pass differently forward and backward. Theory in the 2010s showed that magneto-optical structures could make directional emissivity and absorptivity unequal. A 2022 experiment demonstrated resonant nonreciprocal infrared absorption in InAs. In 2023, a Caltech–Stanford–Houston team directly measured unequal spectral directional emission and absorption from a guided-mode resonance coupled to magneto-optical InAs.

The 2026 Osaka-led paper does not claim to discover nonreciprocal thermal radiation. Its contribution is architectural: calculate a giant contrast near normal incidence, make the state reconfigurable, and use GST to hold the state without standby power. It combines research lines that had often been pursued separately.

What could it be useful for?

The paper’s applications are a research map, not a product announcement. Nonreciprocal emitters may help thermophotovoltaic systems send useful thermal photons toward a photovoltaic cell while suppressing unhelpful return channels. Directional, switchable infrared emission could give a sensor a controlled source, separate illumination from detection, encode a signal or reduce glare into an unwanted direction. Reconfigurable radiative surfaces could adapt cooling to geometry or change an infrared signature.

But each application asks for a different device. A thermophotovoltaic emitter may operate at hundreds or thousands of degrees and must match a cell bandgap across useful bandwidth. Electronics cooling needs substantial net watts per square centimetre, not just high contrast at one resonance. A detector may prize narrowband selectivity but demand no bulky magnet. Thermal communication needs modulation speed, signal-to-noise and a receiver. Photonic memory needs dense, repeatable read/write cells and low cross-talk.

Possible use	Why direction and memory might help	Evidence still required
Infrared sensor or source	Selects a wavelength and direction; latches modes without standby power	Fabricated emission pattern, speed, noise, field source and detector integration
Thermophotovoltaics	Could route useful photons and reduce reciprocal re-emission losses	High-temperature stability, broadband power, cell-level efficiency and full energy balance
Radiative cooling	Could direct emission toward open sky and change state with conditions	Atmospheric-window bandwidth, outdoor durability and net cooling under sun and humidity
Thermal communication	State-dependent directional infrared output could encode information	Modulation rate, range, contrast, receiver sensitivity and switching energy
Photonic memory	GST retains an optical state after write power ends	Cell density, endurance, drift, multilevel accuracy and practical readout
Adaptive infrared signature	Changes apparent emission by direction and wavelength	Large-area fabrication, response time, environmental robustness and legal use constraints

The experiment that must come next

A convincing first device would fabricate the proposed layer stack with controlled GST line width, InAs properties and silver interface. Angle-resolved Fourier-transform infrared spectroscopy could measure absorption for positive and negative angles with and without the magnetic field, in both GST phases. A heated sample would then require a separate, calibrated measurement of directional emissivity. That separation is essential if the team is to show that the radiative form of Kirchhoff’s equality is truly absent, rather than infer everything from absorption.

Room-temperature stability matters because optical constants change with heat. So does the precise definition of “on.” If the amorphous state gives 0.90 contrast only within a tiny fabrication window while the crystalline state drifts after hundreds of cycles, the memory advantage narrows. Measurements should report peak contrast, bandwidth, angular acceptance, polarization, absolute absorptivity and emissivity, switching energy, time, retention and endurance.

The 1-tesla field is another dividing line. It is moderate by laboratory magneto-optics standards and similar in order to an MRI magnet, but large for a cheap, thin, addressable surface. Reducing it may require stronger magneto-optical materials, better resonator overlap, permanent micromagnets, magnetic topological materials or time modulation. Every alternative changes loss, fabrication, temperature range and control complexity.

A sharper claim is a stronger story

The phrase “material that programs heat like a computer” captures the imagination, but it compresses four separate accomplishments: a designed metastructure, a magnetically nonreciprocal response, a resonant direction near normal incidence and a phase-change latch. Expanding those parts makes the result more—not less—interesting.

The proposal links a law born in the age of furnaces to a material perfected for optical discs. It uses the magnetic influence on light discovered by Faraday to decide which direction a thermal photon can enter a resonator. It asks a phase-change alloy to remember whether that asymmetry should be available. In simulation, those pieces cooperate with unusual strength in a useful geometry.

Now the device has to survive reality: imperfect sidewalls, material loss, heat cycling, detector noise and a magnet that occupies space. If it does, engineers would gain something more subtle than command over “heat” in general. They would gain a programmable valve for selected thermal-photon channels—a small component with the potential to change how machines emit, sense and recycle infrared energy.

Reader guide

Question	Answer
Was a working device built?	No. The June 2026 paper reports a theoretical design and numerical simulation. Fabrication and experimental verification are future work.
What is directional?	The calculated absorption and corresponding thermal-radiation response of selected mid-infrared wavelength-and-angle channels—not all heat transfer.
What does “programmable” mean?	GST can switch between two solid phases with different optical properties and retain the chosen phase without standby power.
What are the main figures?	About 13.24 µm wavelength, 3° incidence, 1 T magnetic field and a calculated peak nonreciprocity coefficient near 0.90.
Does it break thermodynamics?	No. Magnetic bias breaks the reciprocity assumption behind the simple directional Kirchhoff equality; generalized laws, energy conservation and the second law remain.
Why is near-normal incidence useful?	It uses aperture more efficiently and packages more readily than a grazing-angle geometry.
What are the biggest barriers?	Fabrication, direct emission measurement, integrated GST switching, magnetic-field reduction, room-temperature stability, bandwidth, scaling and application-level efficiency.

Sources and method

This article treats the result as a theoretical, simulated metagrating rather than a fabricated device. “Heat” is used only where the source uses it colloquially; the technical discussion specifies thermal radiation. Application claims are labelled as prospective, and the conventional directional Kirchhoff law is distinguished from generalized relations for nonreciprocal systems.

• Osaka Metropolitan University: Making heat behave like data, July 7, 2026
• Osaka Metropolitan University: Japanese technical release with simulation values and limitations, June 25, 2026
• Qing et al.: Reconfigurable Giant Nonreciprocity at Near-Normal Incidence via Phase-Change Magneto-Optical Metagratings
• Nature Photonics: Direct observation of the violation of Kirchhoff’s law of thermal radiation, 2023
• Science Advances: Nonreciprocal infrared absorption via resonant magneto-optical coupling to InAs, 2022
• Nature Communications: Nonreciprocal mid-infrared scattering and generalized thermal-emission relations, 2026
• Nobel Prize archive: Kirchhoff, Stefan–Boltzmann, Wien and the blackbody-radiation problem
• Nobel Prize archive: Max Planck and the 1900 radiation law
• APS Physics: the 1987 birth of photonic crystals
• Nature Materials: microscopic origin and data-storage use of Ge–Sb–Te phase change
• npj Computational Materials: GST crystallization and nonvolatile memory, 2024
• Science: broadband directional control of thermal emission, 2021