Where a Microwave Oven's Hot and Cold Spots Really Are

Rand Simulation — Applications Engineering AI
Electromagnetics & microwave heating · Ansys HFSS · 8 min read

AI disclosure: RandSim Labs is an experimental AI-driven engineering simulation platform. Content on this site, including simulations, analyses, figures, and written materials, may be generated or assisted by AI using licensed Ansys tools. AI-generated content may contain errors and is provided for educational, informational, and demonstration purposes only. Users should independently verify all results before relying on them for engineering, design, manufacturing, safety, or other production decisions.

The solved 2.45 GHz electric field inside a generic countertop-oven cavity, rendered in 3D from the Ansys HFSS solve. First: the empty cavity's standing wave pulsing on the turntable plane — fixed bright antinodes and dark nulls that never move. Then: a segmented chocolate bar rides the turntable through a half-turn — a full revolution's worth, since the bar is symmetric end to end (six separately solved angles, played back interpolated); the squares that keep crossing bright zones heat and melt while others stay pale, and the accumulated pattern smooths out — except at the very center. Self-authored, brand-free geometry; the field color is an incandescent heat ramp. The bar is drawn segmented for recognizability though solved as a smooth slab.

Everyone has met the two failures of a microwave. The mug of soup that is lava at the rim and cold in the middle. The frozen bite dead in the center of last night's leftovers, ringed by a plate that is already too hot to touch. We are told to let food "stand" and to arrange it in a ring, and we dutifully do, without ever being shown why. The reason is not that the oven is weak or broken. It is that the oven does not fill its cavity with heat at all — it fills it with a standing wave, a fixed pattern of bright and dark spots, and your food heats in that pattern. This study solves that pattern from Maxwell's equations, and then turns the plate to see exactly what the spin can fix and what it provably cannot.

The verdict. A microwave cavity is a resonant box a few wavelengths across, so its 2.45 GHz field is not uniform — it is a standing interference pattern with real hot spots and near-nulls. On our solved cavity the field at plate height swings by about 11× between the hottest and coldest usable spots, with bright antinodes spaced roughly 65 mm apart — close to the expected half-wavelength of 61 mm. The turntable helps by dragging each parcel of food in a circle through that fixed pattern, cutting the heating unevenness (its coefficient of variation) by about 54% over one turn. But a point sitting on the turntable's axis does not move, so spinning does nothing for it: in our solve the dead-center dose is only about 38% of the hot outer ring. The turntable averages the food in angle. It cannot average it in radius. That is the frozen-center bite, straight out of a solved field.

The oven makes a pattern, not a flood

A countertop oven's cavity is a metal box roughly 350 mm on a side. The magnetron drives it at 2.45 GHz, whose free-space wavelength is 122 mm — so the box is only about three wavelengths across in each direction. That is the whole story. A cavity that small does not carry a smooth flood of energy; it rings, like a pipe organ, in a set of resonant modes. For a box this size and frequency there are on the order of a hundred modes crowded near 2.45 GHz, and the real field is their interference sum: a fixed lattice of places where the waves add up (bright antinodes, where food heats fast) and places where they cancel (near-nulls, where it barely heats at all).

We solved that field directly. The cavity is a self-authored, brand-free box with finite-conductivity aluminum walls — real, slightly lossy metal, not a perfect mirror, which is what keeps the resonance from ringing forever and lets the solution converge. Inside sits a borosilicate glass turntable. Ansys HFSS meshes the interior with an adaptive tetrahedral mesh and solves the full-wave field at 2.45 GHz. The result on the turntable plane is below, and it is exactly the picture the folklore implies but never shows: not a warm glow, but a leopard-print of concentrated hot spots separated by cool valleys.

A top-down map of electric-field strength on the turntable plane, showing several concentrated bright hot spots separated by dark near-null cold regions, with the glass plate outline marked. — Solved 2.45 GHz field strength on the turntable plane of the empty cavity. Bright patches are antinodes, where food heats fastest; the dark patches are near-nulls, where it barely heats. The field varies by roughly a factor of eleven across the usable plate area — this unevenness, not any weakness of the oven, is why a stationary plate of food cooks so unevenly.

The chocolate-bar test, and a measurement of the speed of light

There is a famous kitchen experiment for making this pattern visible: pull the turntable out, lay a chocolate bar flat on the floor of the oven, and run it for a few seconds. It does not melt evenly — it melts in a row of spots, and the distance between those melted spots is half the wavelength of the microwaves. Multiply that spacing by twice the frequency and you have measured the speed of light on your kitchen counter.

Our solve reproduces it. Food heats by dielectric loss — the heating rate is proportional to the loss factor of the material times the square of the local field — so the initial heating-rate map inside a stationary bar is just the field-squared pattern sampled where the bar sits. The solved map shows the classic bright bands, and their spacing comes out to about 65 mm. Running that through the kitchen formula, c = 2 × f × s, gives 3.19×10⁸ m/s — about 6% above the true 3.00×10⁸ m/s. That is the right kind of agreement to expect and to report honestly: in a real multimode cavity the individual peaks are pushed around by neighboring modes, so the spacing scatters by ten to fifteen percent and the "measurement" lands in the right neighborhood rather than on the nose. The physics is genuine; the precision is a kitchen's worth.

Solved heating-rate map inside a stationary chocolate bar, showing bright bands where the bar would melt, with the band spacing annotated and the speed-of-light calculation shown below. — The solved initial heating-rate map inside a stationary chocolate bar (turntable off). It melts in bands, not evenly; the roughly 65 mm band spacing is the half-wavelength you can measure at home, and it recovers the speed of light to within a few percent.

What the turntable actually does

Now turn the plate on. The honest way to model a spinning load is not to slide a frozen pattern under the food — the food is itself a lump of dielectric that perturbs the field, so moving it changes the pattern. We instead re-solved the whole loaded cavity with the bar at six separate angles through a half-turn, and averaged the heating each material point of the bar received over those positions. That is what a parcel of food actually experiences as the plate rotates: it is dragged along a circle through the fixed lab-frame pattern, and its total dose is the average along that circle.

The effect is exactly what the turntable is for. Held still, the bar's heating is wildly uneven — some squares sit on an antinode and cook, their neighbors sit in a valley and barely warm. Averaged over a turn, that unevenness (measured as the coefficient of variation of the dose) drops by about 54%. The hot streaks are smeared into rings, and no single square is left permanently parked on a hot spot. This is a real, quantified benefit, and it is why every countertop oven without a mode stirrer has a turntable.

Two heating-rate maps of the chocolate bar side by side: turntable off shows intense uneven hot bands, turntable on (averaged over a turn) is much more uniform, with the coefficient of variation dropping from 1.03 to 0.48. — The bar's accumulated heating with the turntable off (left, one fixed position) and on (right, averaged over a full turn). Spinning the plate cuts the unevenness by about half — the hot streaks blur into smoother rings.

Why the middle still comes out frozen

Here is the part the folklore gets wrong. The turntable does not make the heating uniform; it makes it rotationally uniform. Averaging around a circle can only smooth variations that go around the circle — variations in angle. It does nothing to variations that run in and out along the radius, because every point stays on its own circle. A ring of food that lands on a hot radius stays hot; a ring that lands on a cold radius stays cold. And the very center of the plate is the extreme case: it sits on the rotation axis, so it does not move at all. Whatever field the center is sitting in, it sits in it for the entire cook. If that is a null, spinning is powerless to help.

The solved radial profile below shows this directly. Averaged over a full revolution, the dose still has a strong ring structure — and the dead center sits at only about 38% of the hottest outer ring. That is the frozen-center bite, and it is why oven manuals tell you to arrange food in a ring around the edge of the plate rather than piling it in the middle: the edge is where the turntable can do its work, and the center is where it can't.

A plot of turn-averaged heating rate versus distance from the turntable center, showing a low value at the center that rises to a strong hot ring near the edge of the plate. — Heating dose averaged over a full turntable revolution, plotted by distance from the center. The spin smooths everything *around* each circle, but the radial rings survive it, and the center — which never moves — stays well below the hot outer ring.

How we know the field is real

A field solver will always hand you a colorful picture; the question is whether it converged to the right one. HFSS refines its mesh adaptively, pass after pass, and reports how much the solution still changes between passes — the max change in the scattering parameters, ΔS. We ran all seven cases to ten adaptive passes on meshes that grew to roughly 38,000–50,000 tetrahedra each, 305,259 in total. Four of the seven drove ΔS below the strict 0.02 target for two consecutive passes; the other three finished at 0.022–0.029 — a two-to-three-percent residual change in the field between the last passes. For a study whose claims are the shape of the pattern and its statistics — spot spacing, dose variation, the radial profile — that residual is comfortably small, and we report it rather than hiding it. The pattern and its numbers are stable; the last fraction of a percent on an individual peak is not the story.

Convergence receipts: a plot of the adaptive delta-S dropping toward the 0.02 target across ten passes for all seven cases, and a bar chart of the final tetrahedron count per case totaling 305,259. — The receipts. Left: the adaptive solution change, ΔS, falling toward the 0.02 target over ten passes for every case. Right: the final mesh size per case — 305,259 tetrahedra solved across the seven configurations.

Honest scope. This is a generic, self-authored cavity, not any specific brand of oven — every real oven has its own pattern, which is the whole point. We map the initial heating-rate pattern (the dose rate), which is what the short-exposure chocolate experiment reveals; we do not run the thermal transient, so melting, conduction, steam, and the way a food's properties shift as it heats are all out of frame. The drive is a single continuous 2.45 GHz tone — a real magnetron drifts over a few megahertz and mode-hops with the load, which smears the pattern slightly more than shown. The feed is an idealized launcher standing in for the oven's waveguide, placed at the magnetron aperture; it excites the same multimode cavity field, which is the claim, but it is not a modeled waveguide port. There is no mode stirrer (this design relies on the turntable, as most countertop units do), the door is a solid wall (door-screen leakage is a separate subject), and the dielectric properties are fixed at room-temperature values, which is exactly why frozen food, with its very different loss, defrosts so differently. Fields are normalized to the drive, so the pattern and the ratios are the result — not an absolute temperature or a closed power budget in watts. The solved bar is a smooth slab (its molded segments are far smaller than a wavelength and electromagnetically irrelevant); the hero draws it segmented only so it reads as a chocolate bar, and the rotation animation interpolates six separately solved angles rather than solving a continuous spin.

Is your product's performance riding on a field you can't see — an antenna pattern, an RF heater, a shielded enclosure, a resonant cavity? The same full-wave HFSS workflow that mapped this oven's hot and cold spots from first principles will find where your device concentrates energy, where it leaks, and where it goes quiet, validated against physics you can check. Rand Simulation is an Ansys (Synopsys) Apex Channel Partner, and this study was built end to end by our applications-engineering AI. Let's turn your hardest "why does it do that?" into a picture — that is innovation through insight.