Project Reverb

Chapter One: The Sound Beneath

The signal arrived at 3:41 in the morning, and Suki Kawakami almost slept through it.

She was on the fourth night of a twelve-night rotation at the Japan Agency for Marine-Earth Science and Technology, in the low grey building above Yokosuka harbor where the deep-ocean seismic arrays came home to roost. Her job that month was ordinary: watch Sakurajima, watch Aso, watch the shallow swarms under Kirishima, and write a one-page bulletin at dawn for people who would skim it over coffee. Volcanology at four in the morning is mostly patience and instant miso.

The alert that woke her was not from a volcano.

It was a spectral flag from the long-period pipeline, the slow channel, the one that listened to Earth the way a doctor listens to a chest rather than a heartbeat. Volcanic harmonic tremor sings between half a hertz and five hertz — a fluttering, rasping sound, gas and magma scraping through a throat of rock. What the flag had caught sat at four point one millihertz. A period of just over four minutes. Below that, another line at two millihertz. Above it, a ladder of overtones, evenly spaced, patient, clean.

Suki sat up in the cot, put on her glasses, and looked at the ladder for a long time.

Earth rings. Every large earthquake sets the whole planet vibrating like a struck bell, and for weeks afterward you can find the tones in the data if you know their names. Every seismologist learns them. But those tones decay. They fade as the planet swallows the energy. And there had been no great earthquake — nothing above magnitude six anywhere on the globe for eleven days.

This ladder was not fading. Over the six hours the pipeline had been recording it, the lines had grown by three percent in amplitude.

She pulled the raw traces. Eight ocean-bottom stations in the Philippine Sea, four land stations on Kyushu, the borehole instrument at Matsushiro. She stacked them, aligned them, beamformed the array the way you would steer a radio telescope, hunting for the direction the energy came from.

The beam did not point at a volcano. It did not point at a subduction zone or a fault. It pointed down.

She ran the depth inversion three times because she did not believe it. Twenty-eight hundred and sixty kilometers. Twenty-eight ninety. Twenty-eight eighty-five, with an error bar of forty kilometers that comfortably swallowed the number every geophysicist on Earth knew by heart: 2,891.

The core-mantle boundary. The floor of the rock, the roof of the liquid iron. The place where the solid world ends and the molten dynamo begins.

Suki got up, walked to the window, and looked out at the harbor lights smeared in drizzle. She was thirty-four years old and had spent eleven of those years learning that the interesting result is almost always a broken cable.

"Instrument error," she said aloud, to be fair to herself. Then she went back to the desk and spent the next three hours trying to prove it.

She could not. A bad sensor gives you one bad station. She had twelve stations across four hundred kilometers of seafloor and island, on three different generations of hardware, from two manufacturers, and every one of them heard the same ladder with the correct arrival delays for a source directly beneath them. Tidal contamination was ruled out; the tides had their own lines and she knew them like old neighbors. Power-grid harmonics were fifty hertz, not four millihertz. The clocks were locked to GPS and agreed with each other to microseconds.

At 7:15 she wrote her bulletin. She put the volcanoes first because that was the format, and then, at the bottom, under a heading she typed and retyped four times, she described an ultra-long-period harmonic signal of unknown origin, apparently sourced at the core-mantle boundary, amplitude increasing, fundamental period 244 seconds and shortening.

Shortening. That was the part that had kept her at the desk. She had gone back through the archive and found the ladder faintly present three weeks earlier, buried under noise, at a period of 251 seconds. Whatever was making the sound, it was making it faster.

Her section head, Ryohei Inagaki, read the bulletin at nine and came to find her with the particular gentleness senior scientists reserve for junior ones who have been awake too long.

"Kawakami-san. When did you last sleep properly?"

"It's on twelve stations."

"Twelve stations sharing one telemetry backbone and one processing pipeline. If the pipeline has a phase error—"

"It doesn't."

"Then let the network tell us." He tapped the screen, not unkindly. "Push it to the GSN. If Chile hears it, you will have my full and grovelling attention."

She pushed it to the Global Seismographic Network at 10:02 that morning — the twenty-first of March, 2041 — with the raw traces, the beamforming code, and a request for independent confirmation.

The network's automated quality-control system flagged her upload as probable instrument malfunction within four hours. A polite human moderator in Albuquerque appended a note suggesting she check her instrument response files. The thread got nine views in a week.

Suki did not argue. She did something better. She wrote to twenty-two individual station operators, one by one, in careful, flattering, specific emails, asking each of them to look for a single narrow spectral line at four point one millihertz in their own long-period data, using their own tools, without looking at hers first.

The first reply came from Borgarnes, Iceland, on the twenty-ninth. It was two words and a plot: *What's this?*

The second came from a broadband vault outside Antofagasta, Chile, on the second of April, from an operator who had assumed his cryocooler was failing until he checked and found it was not.

The third came from Boshof, in the South African Karoo, on the fourth, from a woman who had been recording Earth's slow tones for nineteen years and wrote, simply, that she had never seen a normal mode that grew.

By the eighth of April there were eleven stations on four continents, and the fundamental period had fallen to 239 seconds.

Suki laid the plots out on the floor of her apartment in a row, because the desk was not big enough, and knelt beside them in her socks with a mug going cold in her hand. Somewhere far below the linoleum, below the crust and the whole vast slow-boiling mantle, something at the top of the outer core was ringing, and getting louder, and getting quicker, and none of the physics she had ever been taught explained why.

She took a photograph of the row of plots. Then she opened her laptop and searched the literature for anyone, anywhere, who had ever predicted a growing harmonic at the core-mantle boundary.

There was one preprint. It had been uploaded fourteen months earlier, never published in a journal, cited twice, both times dismissively. The title was dry to the point of camouflage: *Metastability and Slow-Phase Transition in the Terrestrial Outer Core: Predicted Standing-Wave Signatures.*

The author was a theoretical physicist at CERN's gravitational-wave annex in Geneva named Noor Kharazmi.

Suki opened Figure 4 of the preprint, a synthetic spectrum generated from a model, and held it beside the photograph of the plots on her floor.

The ladders matched. Line for line, spacing for spacing.

She wrote the email standing up, and sent it at 1:20 in the morning, Japan time.

*Dr. Kharazmi — I believe I am observing your Figure 4. Eleven stations, four continents, amplitude rising, period shortening from 251 to 239 seconds since February. Please tell me what your model says happens next.*