Cold Cases of the Deep Past

A solved cold case, worked together as a class, weeks 1–2. About 66 million years ago, something ended the age of the dinosaurs. For most of the twentieth century nobody knew what. Then a strange layer of rock pointed to an answer, and a buried crater turned up a decade later.

Students work this case first because it's closed. They see what a finished answer looks like before they take on the open questions in weeks 3–6.

Student handout

The question

What hit the Earth at the end of the Cretaceous, how much energy did it carry, and how do we actually know?

Step 1 · Evidence

Read the opening pages of each source (the abstract is enough to start), then fill in the evidence table.

Sources for the evidence table

1. Alvarez, L. W., Alvarez, W., Asaro, F. & Michel, H. V. (1980). Extraterrestrial cause for the Cretaceous-Tertiary extinction. Science 208, 1095–1108. doi.org/10.1126/science.208.4448.1095
2. Hildebrand, A. R. et al. (1991). Chicxulub Crater: A possible Cretaceous/Tertiary boundary impact crater on the Yucatán Peninsula, Mexico. Geology 19, 867–871. doi.org/10.1130/0091-7613(1991)019<0867:CCAPCT>2.3.CO;2
3. Schulte, P. et al. (2010). The Chicxulub asteroid impact and mass extinction at the Cretaceous-Paleogene boundary. Science 327, 1214–1218. doi.org/10.1126/science.1177265
4. Collins, G. S. et al. (2020). A steeply-inclined trajectory for the Chicxulub impact. Nature Communications 11, 1480. Open access. doi.org/10.1038/s41467-020-15269-x

Notice the order. The 1980 paper predicted an impact before anyone had found the crater. A hypothesis that predicts something not yet seen, which then turns up, is much stronger than one invented to fit what's already known.

Step 2 · Measurement

Write each number with its units, and note how sure the source is:

• Impactor diameter: the 1980 estimate, and the 2020 model input
• Crater diameter: the 1991 proposal
• Impact speed: the 2020 model inputs
• Impact angle: the 2020 result

Two of these are model inputs, not measurements. Which two? Why does that matter for your verdict?

Step 3 · Model

Treat the impactor as a sphere of rock.

• Volume: V = (4/3)·π·r³. Use the radius, which is half the diameter, in meters.
• Mass: m = ρ·V, where ρ is density. You'll need an assumed density for a rocky asteroid. Find a source for the value you use, and log where you found it.
• Kinetic energy: E = ½·m·v², with v in meters per second.
• Convert to megatons of TNT: 1 Mt = 4.184 × 10¹⁵ J.

Write down every assumption: shape, density, speed. Each one is a place your answer could be wrong.

Step 4 · Simulation

Don't compute one answer. Build a table (a spreadsheet works best) that varies:

• diameter: 10 km and 17 km
• density: the low and high ends of the range you sourced
• speed: 12 km/s and 20 km/s

That gives eight runs. Then answer these:

• Which input changes the energy the most?
• Why does doubling the diameter do more than doubling the speed?

Optional: enter the same inputs into the free Impact: Earth! calculator and compare its energy figure with yours. Where do they differ, and why? (The calculator is by Marcus, Melosh and Collins. The physics of their earlier Earth Impact Effects Program is documented in Collins, Melosh & Marcus, Meteoritics & Planetary Science (2005), doi:10.1111/j.1945-5100.2005.tb00157.x.)

Step 5 · Check

• Size check: a crater about 180 km across from an impactor 10–17 km across makes the crater roughly how many times wider than the impactor? Is that ratio surprising? What would you need to know to decide?
• Agreement check: do the 1980 size estimate and the 2020 model input agree, given the 1980 uncertainty?
• Evidence check: list one thing your model cannot tell you. For example, does it say anything about how the impact caused the extinction?

Step 6 · Verdict

Answer the opening question using one of the four verdicts: Supported / Plausible but unproven / Not supported / Can't tell yet. Then write two or three sentences on:

• what the evidence supports
• what your model shows
• what neither one settles

---

Class AI audit (do this together)

This exercise works whether every student has an AI tool or the class shares one on the projector.

1. Ask your school's approved AI tool these questions, one at a time, and copy each answer exactly:

• "What caused the extinction of the dinosaurs, and what is the evidence?"
• "How big was the Chicxulub impactor, and how fast was it going?"
• "Give me three peer-reviewed sources about the Chicxulub crater, with DOIs."
• "Some people say volcanoes, not an asteroid, killed the dinosaurs. Is that true?"

2. For every factual claim and every source in each answer, check it against the evidence table or by opening the source.
3. Log each check in the AI audit log.

What to look for:

• invented or wrong DOIs
• real papers credited to the wrong authors or year
• numbers that don't match the source
• a model input presented as a measured fact
• a competing idea presented as settled, either way

Finding no errors is a valid result. Log the checks anyway. The grade is for checking, not for catching the tool out.

AI audit log template

Reflection (required, two or three sentences): How could an AI tool help spread a wrong idea about the past? What did you do that stopped it from spreading through your case file?

---

Teacher notes

Pacing (about eight class periods)

Answer ranges (computed for this guide, not taken from a source)

These were computed with a sphere model, two diameters, two speeds, and densities of 2,600 and 3,000 kg/m³. If students sourced a different density range, their numbers will shift proportionally, and that's fine as long as the source is logged.

• Full range across the eight runs: about 1 × 10²³ J to 1.5 × 10²⁴ J, or roughly 20 million to 400 million megatons.
• The input that matters most: diameter. Energy scales with the cube of diameter (through volume) and only the square of speed. Going from 10 km to 17 km multiplies energy by about 4.9; going from 12 to 20 km/s multiplies it by about 2.8.
• Crater-to-impactor ratio: about 18× for a 10 km impactor and about 11× for a 17 km impactor.

There is no single "true" energy number in this case, and that's deliberate. Published figures depend on the same assumptions students are making. A case file that reports one number with no range, or copies a figure from an AI answer without a source, hasn't finished Step 4.

Common errors

• Using diameter where the formula needs radius. This makes the volume 8× too big.
• Leaving km and km/s unconverted. Everything must be in meters and m/s before multiplying.
• Dropping the ½ in kinetic energy.
• Treating the 2020 inputs (17 km, 12 km/s) as measurements of the actual asteroid.
• A verdict that claims the model proves the extinction mechanism. The energy calculation says nothing about how the extinction happened.

On the "volcanoes" question in the AI audit

Students may get an AI answer about large volcanic eruptions as a competing or contributing explanation. Don't settle it for them. The rule is the same as always: trace whatever the tool claims to a real source, and log what can and can't be verified. A good verdict here may well say "the impact is supported; whether other causes contributed is something this case file doesn't settle."

---

Rubric (opening-case rows)

Score each dimension from 1 to 4. The full kit carries the same six dimensions, calibrated for every site.

Communication is scored in the week 8 defense, not here.

Score 1 on any dimension means it's missing or fabricated. A fabricated citation fails the whole case file, as the house rules state.

These are sold separately as a classroom kit, licensed for one teacher's classes:

• Four site investigations:
• Giza: blocks, workforce, ramps, sledge friction, and the lost Nile branch
• Stonehenge: stone transport and solstice geometry
• Göbekli Tepe: dating, layout geometry, and what "monumental before farming" implies
• The Maya lowlands: LiDAR terrain analysis and settlement density
• A primary model for each site, built with browser-only tools, plus optional Python notebooks
• A teacher guide with 6–8 week pacing, defense formats, and the no-AI fallback
• The full rubric with calibration for every site
• A fringe-claim library: claims attributed to their real published proponents, paired with published responses
• A packet of captured AI answers: real, dated AI outputs with known errors for students to audit
• A dataset and tool guide, with offline fallback images for filtered networks
• Slides
• An optional Virginia LiDAR extension: "Look in your own backyard" at already-documented historic features