Crystals and Collisions
EMGT 5061 · Introduction to Space Industry · Fall 2026

Crystals and Collisions

Space medicine meets space debris. A company that makes cancer-drug crystals in orbit has to share that orbit with debris, and every dodge costs it a batch. In class you run its operations. For homework you are the engineer advising an investment firm.

October 28, 2026Dr. Bidushi BhattacharyaRecord your answers in this week's student deck

This week's companies

Three fictional companies. Stillwater Bio makes drug crystals on a platform in low Earth orbit. Halden Therapeutics, a drug maker, buys those crystals. Keelwatch Tracking watches debris from radars on the ground and sells collision warnings to satellite operators. Stillwater isn't a customer yet, but Keelwatch's sharper warnings could help it decide when to dodge.

Stillwater BioDrug crystals made in microgravity

PlatformAn uncrewed platform at 525 km, near the Starlink shells. Operating for two years. Worth $150M, with an 8-year design life.
ProductAntibody crystals for Halden Therapeutics, under a 5-year contract. The crystals let a cancer drug be given as a short injection instead of an IV drip.
The catchEach batch needs 3 days of undisturbed microgravity. Any engine burn to dodge debris ruins the batch that's growing.
Tracking nowFree public tracking data only. Stillwater does not subscribe to Keelwatch.

Halden TherapeuticsStillwater's customer

WhoA drug company. It turns Stillwater's antibody crystals into a cancer drug given as a short injection instead of an IV drip.
Contract5 years. Halden pays $1.0M for each good batch.
Why it mattersAs the customer, Halden sets the prices. You won't evaluate Halden itself.

Keelwatch TrackingDebris-tracking data, sold by subscription

Network4 ground radar sites. Each cost $30M to build and costs $8M a year to run.
ProductCollision warnings for satellite operators. Its radars track debris more precisely than public data, so its odds settle sooner and are clearer either way.
Customers25 commercial satellite operators paying about $4M a year each, plus one government contract worth $25M a year.
NextWants to sell to orbital factories like Stillwater, a new kind of customer. Stillwater isn't a customer yet. A subscription would cost it $4M a year.

How a debris warning plays out

  1. 1

    A warning arrives

    Three days before a close pass, tracking shows a piece of debris heading near the platform. Its exact position isn't known yet: it could be anywhere along a stretch of its path that can be kilometres long. The collision odds are the chance that the platform will be inside that stretch when the debris passes.

  2. 2

    Each day, the picture gets clearer

    Every time a radar sees the debris again, its predicted position gets sharper and the stretch shrinks. Usually it shrinks to a spot away from the platform, and the warning clears. Sometimes it shrinks around the platform's path, the odds go up, and Stillwater has to burn. That's why waiting a day can be worth it, and why only 18 of about 60 warnings a year end in a burn.

  3. 3

    Better tracking clears it sooner

    Stillwater uses free public data and does not subscribe to Keelwatch. Keelwatch's radars see the debris more often and more precisely, so the picture sharpens faster and the odds come out clearer: either very low or clearly high. In class you decide on three warnings with free data first, then again with Keelwatch's data.

Stillwater's numbers

Batch cycle3 days growing + 1 day reset, so 90 batch slots a year
Halden pays per good batch$1.0M
Consumables per batch$0.3M, so a good batch earns Stillwater $0.7M
Keelwatch cost per batch (if subscription is in place)$4M a year ÷ 90 batches ≈ $0.044M
Cost of a ruined batch$1.0M: $0.3M of wasted consumables + the $0.7M it would have earned
Debris warningsAbout 60 a year. The first alert comes 3 days before the closest approach. Most warnings fade as tracking improves, so only 18 a year end in a burn.
Dodge lineBurn if the collision odds are above 1 in 10,000, the same threshold NASA uses for the ISS.
Burns a year with public data18
InsuranceThe policy is void if Stillwater sets its dodge line looser than 1 in 10,000.

Stillwater's work is modeled on real research: Merck grew crystals of the cancer drug Keytruda on the ISS, and NASA credits that work with informing the injectable form the FDA approved in 2025. Debris is the same problem you modeled in Orbit Commons on October 7; this week you see what it costs one company. All three companies and all figures are fictional, for teaching.

Pairs · about 30 min

The conjunction log

You are Stillwater's operations manager. Three debris warnings came in over a recent 12-day stretch. For each one, decide whether to start the next batch now or hold it for a day, and log your decision and reason.

How to decide

Each warning arrives 3 days before the closest approach, just as the next batch is due to start. You can start the batch now, or hold it a day while the debris's position becomes clearer. The tables below tell you how often a warning ends in a burn, using free public data or Keelwatch's data.

Risk cost

Risk cost = probability of the loss × size of the loss. You compare that number with the cost of the alternative, usually the cost of preventing or avoiding the loss. If prevention costs less than the risk cost, you prevent. If it costs more, you accept the risk. NASA's risk matrices rest on the same idea, rating each risk by likelihood and consequence.

With no warning, there's nothing to weigh. The calculation only starts when a specific risk shows up.

Free public data

Odds at the first alert
(free public data)
Chance it ends in a burn*Risk cost of starting the batch now**
1 in 100,0002%$0.02M
1 in 20,00010%$0.10M
1 in 5,00030%$0.30M
1 in 1,00070%$0.70M

With a Keelwatch subscription

Odds at the first alert
(Keelwatch)
Chance it ends in a burn*Risk cost of starting the batch now**
1 in 100,0002%$0.02M
1 in 20,00010%$0.10M
1 in 5,00030%$0.30M
1 in 2,00050%$0.50M

* Chances are for illustrative purposes only.
** The chance of a burn × $1.0M. See step 2 of the rule below for why.

Start the batch now

If the warning ends in a burn, the batch is ruined. That costs $1.0M.

Hold one day

Wait a day for clearer tracking, then start. In this exercise, one day is enough for the warning to settle: by then you know whether Stillwater must burn, and any burn happens before the batch begins. So a held batch is never ruined. You lose one day of a 4-day cycle: $0.7M ÷ 4 = $0.175M.

The dodge line never changes. Stillwater always burns when the collision odds are above 1 in 10,000, because its insurance is void with any looser line. Your decision is about something else: whether a batch is growing when a burn might happen.

The rule: compare the risk cost of starting the batch now with the cost of holding it a day

  1. Cost of holding one day. A batch cycle is 4 days, and a good batch earns $0.7M, so each day of the cycle is worth $0.7M ÷ 4 = $0.175M. Holding loses one day, so it always costs $0.175M. In return, the batch is safe.
  2. Risk cost of starting the batch now. If the warning ends in a burn, the batch is ruined, which costs $1.0M. If the warning clears, the batch finishes normally and costs nothing extra. To compare this with holding, weigh the $1.0M by how likely it is: risk cost = chance of a burn × $1.0M. A 30% chance of losing $1.0M, for example, has a risk cost of $0.30M.
  3. Compare the two. Hold if the risk cost of starting the batch now is more than the cost of holding it for one day; otherwise start. The tables above give the risk cost for each kind of warning.

Example: with a 25% chance of a burn, the risk cost of starting the batch now is 25% × $1.0M = $0.25M, more than the $0.175M cost of holding, so hold. With a 5% chance, the risk cost is $0.05M, less than $0.175M, so start.

Card 1

A quiet morning

One warning: 1 in 20,000 at first alert. Nothing else is scheduled.

Card 2

A closer call

One warning: 1 in 5,000 at first alert. Nothing else is scheduled.

Card 3

A close one

One warning: 1 in 1,000 at first alert. Nothing else is scheduled.

The three warnings: free data vs. Keelwatch

Odds at the first alert for each warning, as free public data reported them and as Keelwatch would have reported them. Keelwatch tracks the debris more precisely, so its odds can differ, sometimes lower and sometimes higher.

CardOdds at the first alert
(free public data)
Odds at the first alert
(Keelwatch)
11 in 20,0001 in 100,000
21 in 5,0001 in 2,000
31 in 1,0001 in 5,000

Your log

Copy both tables into the student deck and fill in one row per card.

Table 1: free public data

CardOdds at the first alertChance of a burnRisk cost of starting the batch nowStart or hold?Reason, for the log
11 in 20,000
21 in 5,000
31 in 1,000

Table 2: with a Keelwatch subscription

Now suppose Stillwater subscribed to Keelwatch ($4M a year, about $0.044M per batch). Here are the same three warnings as Keelwatch would have reported them. Use the How to decide table to fill in the rest.

CardOdds at the first alert
(Keelwatch)
Chance of a burnRisk cost of starting the batch nowStart or hold?Reason, for the log
11 in 100,000
21 in 2,000
31 in 5,000

Due before November 4, 3 pm

You are now the engineer advising the investment firm. Answer in this week's student deck and upload it to Canvas.

Homework 1

Fix the orbit or fix the process?

Stillwater has asked your firm to fund one fix for its debris problem. Compare the options using profit per year.

What "profit per year" means

It's what Stillwater makes from its batches in a normal year: what Halden pays for good batches, minus consumables, minus any yearly fee the option adds. It leaves out the platform's own cost, staff and overhead, which are the same under every option.

Profit per year = (good batches × $0.7M) − (ruined batches × $0.3M) − any yearly fee

Each burn ruins one batch, so ruined batches = burns a year, and the rest of the 90 slots are good batches. One-time costs, such as moving the platform, are listed in their own column, not taken out of profit per year.

Example, today's baseline: 72 good × $0.7M − 18 ruined × $0.3M = $45.0M.

OptionBurns a yearProfit per yearOne-time costsCatch
Baseline: no change18$45.0MNone
A. Subscribe to Keelwatch9$50.0MNone$4M a year, already taken out of profit
B. Move down to 450 km6$57.0M$7.5M to move, including 20 days of downtime; $15M refueling visit in year 5More drag means 8 reboosts a year, which ops must fit into reset days
A + B together3$56.0MSame as B
C. Shorter 36-hour batches18$36.9MHalden must re-qualify its product180 slots a year; Halden pays $0.45M per short batch, consumables $0.2M, so a good short batch earns $0.25M
D. The CFO's idea: loosen the dodge line to 1 in 1,0005$58.0MNoneVoids the insurance on the $150M platform and adds about a 0.4% yearly chance of losing it in a collision
H1.1

Check one row yourself using the formula. Then explain why A + B earns less than B alone.

H1.2

Option B earns $12M a year more than the baseline. How long does it take to pay back the $7.5M move? Does the year-5 refueling visit change your view?

H1.3

About how many burns a year would it take for shorter batches (C) to beat the 3-day batches?

H1.4

Option D has the best number in the table. What do you tell the investment committee about it, and why?

H1.5

Which option do you recommend your firm fund, and what risk does it leave Stillwater carrying?

TRL and MRL

For worked examples of both ladders, see Ready for the Moon? Use it and the definitions below for Homework 2.

Ladder definitions (TRL 1–9 and MRL 0–9)
TRLTechnology Readiness (NASA)
MRLMarket Readiness (Mealing, Starbridge Venture Capital)
Homework 2

Rate both companies

For each item, read the evidence, pick a level and defend it in the box. Then open the suggested rating and compare. A level either side of the suggestion is fine if your reasons are good.

H2.1

Which of Stillwater's ratings should worry your firm most if it funds Option B in Homework 1?

H2.2

Keelwatch's radars are its most proven part and its sharper odds are its newest. Why does that matter for what your Keelwatch log really showed?