Microscopic crystals pulled from ancient lakebed rock in northeastern Arizona carry an age signature that pins the Colorado River inside a landlocked basin 6.6 million years ago, more than a million years before the river broke through to the sea, and that timing closes the oldest standing argument in American geology.
Findings published in Science in April 2026 record the mineral fingerprint of Colorado River sediment in the upper Bidahochi Formation in northeastern Arizona, dated by uranium-lead analysis of 3,596 individual zircon grains across 19 sandstone samples, placing the river’s arrival at the Bidahochi basin no later than 6.62 million years ago.
The Grand Canyon is 277 miles long, up to 18 miles wide, and over a mile deep. Six million people stand at its rim every year. Most of them assume someone already knows how it formed. Nobody did, until now. Geologists had confirmed the Colorado River existed in western Colorado 11 million years ago and had already reached the Gulf of California somewhere between 5.6 and 4.8 million years ago. That left a five-million-year gap with no verified answer: where was the river, and how did it punch through the rock wall that stands between northeastern Arizona and the canyon corridor? Four competing theories had been fighting over that question since the 1930s without any of them producing enough physical evidence to win.
The rock wall at the centre of that argument is called the Kaibab arch, a natural ridge of ancient limestone running north to south across the Colorado Plateau, with its lowest crossing point sitting at roughly 2,300 metres above sea level along the south rim of the Grand Canyon. East of that ridge sat the Bidahochi basin, a broad low-lying area that collected sediment for millions of years and, during the period between 7 and 6 million years ago, held a large inland lake. West of the ridge sat the canyon corridor, dropping away toward what would eventually become the river’s outlet to the sea. Getting the Colorado River from one side of that ridge to the other is the entire problem. Two of the four theories involve the river cutting its way through: one by eating backward from below through a process called headward erosion, one by dissolving underground limestone passageways until the rock collapsed and captured the river. The other two theories involve water rather than cutting: groundwater seeping through cliff faces and undermining them from beneath, or the lake east of the ridge simply filling until it overtopped the wall and spilled downhill into the canyon corridor by gravity alone.
Zircon is a mineral that forms inside cooling volcanic rock and locks uranium atoms inside its crystal structure the moment it solidifies. That uranium decays into lead at a fixed rate, turning every grain into a dated receipt for its birthplace. Volcanic regions produce zircon grains at specific ages, so a geologist reading the age distribution of zircon in a sediment sample can trace where that sediment came from, the same way a customs officer reads a passport. The Colorado and Green River system carries a recognisable passport: grains that crystallised between 40 and 25 million years ago, eroded from ancient ash deposits left by the San Juan volcanic field in Colorado and transported downstream. Those grains are so specific to that river system that their presence in any deposit confirms Colorado River sourcing, and their absence rules it out. The lower Bidahochi Formation, deposited before 7 million years ago, contains almost none of them, meaning the Colorado River was not feeding the basin during that earlier period.
The upper Bidahochi Formation, deposited between 7 and 6 million years ago, is full of them. The zircon age distribution in the upper unit matches known downstream Colorado River deposits, including Pliocene sediments from the Bouse Formation and the deposits of Hualapai Wash dated to between 5.2 and 4.8 million years ago, at shared age peaks of roughly 95, 170, 1,100, 1,450, and 1,700 million years. The lower unit matches none of those peaks and clusters statistically in a completely separate group. Uranium-lead dating of volcanic ash beds within the upper Bidahochi Formation at Roberts Mesa and Greasewood Mesa in northeastern Arizona pins the unit’s age to between 6.69 and 6.42 million years ago, with individual date uncertainties of roughly 90,000 years. The switch between the two units is not gradual. It is sharp, and it marks the moment the Colorado River arrived.
When the river arrived, it did not arrive quietly. Sediment accumulation in the Bidahochi basin before 6.6 million years ago ran at 10 to 20 metres per million years, the pace of a quiet lake fed by local streams. After 6.6 million years ago it jumped to between 100 and 400 metres per million years, a twenty-fold acceleration consistent with a large river suddenly dumping material into a previously underfed basin. The chemistry of freshwater carbonate minerals in the upper Bidahochi unit shifted at the same time: strontium isotope ratios climbed to match the elevated ratios recorded in downstream Colorado River deposits in the Bouse Formation, because the Colorado passes through ancient Precambrian basement rocks in its upper watershed that stamp river water with a chemically distinct signature. Fossil fish recovered from the upper unit include species reaching 75 centimetres in length, with enlarged fins and narrow tail fin bases, body shapes built for fast-moving rivers, not still lakes, and those species match fish recorded from the Snake River and Sacramento River systems, not any local Arizona drainage. Three completely independent signals, mineral chemistry, sediment physics, and fossil biology, all change at the same stratigraphic boundary, which is the same boundary where the Colorado River’s zircon passport appears.
The fish carry a detail that stretches this story much further north than Arizona. The zircon signature of the upper Bidahochi Formation matches the Browns Park Formation specifically, a sequence of sedimentary rock from the Green River drainage in Utah and Colorado deposited between 25 and 8 million years ago. The Green River is a major Colorado tributary, joining it in present-day Utah. During the late Miocene, the volcanic centres that preceded the Yellowstone hotspot system formed a topographic ridge across what is now southern Idaho, funnelling water southward into the ancestral Green River and from there into the Colorado. Zircon grains in the upper Bidahochi unit carry isotopic properties matching samples from a borehole drilled through the Twin Falls Caldera in southern Idaho, a transport distance exceeding 800 kilometres. Fish species in the upper Bidahochi unit link the same network: Snake River lineages reaching northeastern Arizona through a continental-scale drainage that pulled from Idaho in the north and delivered water to a landlocked basin in the high desert southwest.
Once that drainage network filled the Bidahochi lake, the rest is physics. Remnants of the upper Bidahochi Formation are preserved today at elevations between 1,805 and 2,250 metres above sea level. The highest deposits include beach sand, beachrock, and capping tufa, a mineral crust that forms only at a waterline, placing the lake surface close to the 2,300-metre sill elevation of the Kaibab arch. Stratigraphic mapping across the basin shows that sediment contacts remain flat toward the west, exactly the geometry expected when lake deposits push against a wall they have not yet cleared. A 2024 Arizona Geological Survey field report separately documented beach sand and beachrock at Balakai Mesa, placing the paleowater surface within reach of the crossing point. When the lake surface hit the sill, water began flowing west over the arch, cutting into the rock below and establishing the channel that became the Grand Canyon corridor.
The zircon record places Colorado River sediment flowing continuously into the Bidahochi basin from 6.6 million years ago through approximately 6 million years ago. Colorado River sediment does not appear in the Lake Mead area or the lower Colorado corridor until after 5.6 million years ago, and possibly as late as 4.8 million years ago. That gap of 400,000 to 1.2 million years is the time it took the lake to fill, reach the sill, spill, and cut enough of a channel to pass sediment all the way through. Modern sedimentation rates at Lakes Powell and Mead place equivalent basin-filling at 100,000 to 200,000 years. Late Miocene conditions were drier and more stable, meaning accumulation was slower and the gap is geologically consistent with a filling-and-spillover sequence rather than any mechanism that would move sediment through the canyon faster.
The other three theories each fail a specific physical test. Headward erosion predicts Colorado Plateau detritus in Miocene deposits at the canyon’s western mouth, and those deposits contain none. Karst piping, the underground dissolution pathway, is implausible across 277 miles of canyon because the main karst unit, the Redwall Limestone, sits more than 600 metres below the surface with no linear collapse features of the required scale anywhere in the canyon system. Groundwater sapping also predicts Colorado Plateau detritus at the canyon’s mouth through the same headward logic, and that material is absent. Only the lake spillover mechanism is supported by the zircon provenance data, the strontium isotope shift, the sediment accumulation jump, and the fossil fish fauna simultaneously, four independent data streams pointing to the same event at the same time in the same place.
Collected samples from the Bidahochi Formation are archived at the US Geological Survey office in Flagstaff, Arizona. Detrital zircon mounts are stored at the University of California Los Angeles Department of Earth, Planetary, and Space Sciences, with data deposited at UCLA Dataverse under identifier 10.25346/S6IXERBM. USGS uranium-lead data are separately archived under data release identifier 10.5066/P13WZQBW.
Source:
He, J. J. Y., Crow, R. S., Douglass, J., Holm-Denoma, C. S., Vazquez, J. A., Gootee, B. F., Lidzbarski, M. I., Pianowski, L. S., Gray, H., Heitmann, E., Pearthree, P. A., House, P. K., & Dulin, S. (2026). Late Miocene Colorado River arrival in the Bidahochi basin supports spillover origin of Grand Canyon. Science, 392(6795). https://doi.org/10.1126/science.adz6826
