Butterscotch, Toffee & Hop Creep

Diacetyl is the butter, butterscotch and slick-mouthfeel fault, and it has an unusual property among beer faults: you almost never catch it in the fermenter. The beer smells clean, you crash it, you keg it, and two weeks later it tastes of toffee.

That is not bad luck and it is not a fermentation failure. It is a timing failure, and it follows directly from how the compound is made and removed. Once the sequence makes sense the fix is obvious, and it is not the one most people reach for.

Three steps, not one

Removing diacetyl is three separate processes running at wildly different speeds, and the slow one has no yeast in it at all.

	What happens	Driven by	How long
1. Formation	Yeast leaks alpha-acetolactate into the beer as a side-branch of making the amino acid valine	Yeast growth	While the yeast is growing
2. Conversion	Alpha-acetolactate turns into diacetyl, outside the cell	Plain chemistry — temperature, pH, time. No enzyme, no yeast	Hours to days
3. Reduction	Yeast absorbs the diacetyl and reduces it to compounds you cannot taste	Yeast enzymes	Hours

That last step is dramatic. Diacetyl is detectable at somewhere around 0.1 mg/L. The compounds the yeast turns it into have thresholds of roughly 50 mg/L and 4,500 mg/L respectively. Healthy yeast in suspension does not struggle with this.

Why the beer smells clean anyway

Reduction is fast and conversion is slow, so diacetyl never gets a chance to build up while there is yeast around. It is eaten as fast as it appears. Your beer smells clean because the yeast is keeping up — not because the problem is gone.

A beer can smell perfectly clean and still be full of precursor.

That alpha-acetolactate does not care what you do next. Crash the beer, drop the yeast out, rack it to a keg, and the conversion carries on for weeks in the cold with nothing left in suspension to clean up after it. The butterscotch you taste in week three was sitting in the fermenter invisibly on the day you packaged.

So the question is never “can I taste diacetyl?” It is “is there precursor left?” — and your nose cannot answer that.

The “diacetyl rest” is misnamed

If you have absorbed the standard advice, this is the part that sharpens it.

The usual version — warm it up so the yeast can clean up the diacetyl — is not wrong. The yeast very much does the cleaning up. It is incomplete, and the missing half is what determines how long the rest needs to be.

The warmth is doing two jobs at once:

Speeding up the slow step. Step 2 is plain chemistry and it is the bottleneck. The rest exists largely to let it finish, so the diacetyl appears while there is still yeast in the beer to eat it.
Keeping the yeast healthy and in suspension to do step 3. Reduction takes hours, but only if there are active cells still up in the beer. Cold, flocculated yeast on the bottom cannot clean up anything.

Miss either one and the rest fails.

Two common mistakes fall out of that immediately:

A rest on a calendar is meaningless. “Days five to seven” is a typical timeline, not a criterion. The rest is over when the beer says so.

Cooling to protect the beer is exactly backwards. Cold slows the conversion and drops the yeast out. It does not remove the problem, it postpones it until the only thing that could have fixed it is gone. Cold-crashing a beer that still has precursor in it is precisely how you get a buttery keg.

The test that actually answers it

The forced diacetyl test drives the slow step to completion artificially, so you can smell what the beer will become rather than what it currently is. It takes twenty minutes and needs no equipment.

Draw two samples of the same beer into clean glass you can seal.
Leave one at room temperature as your control.
Hold the other at 60°C for 20 minutes, sealed so nothing escapes.
Cool the heated sample back to the control’s temperature. Comparing a warm sample to a cool one tells you nothing.
Smell and taste both.

Reading it

They smell the same — no meaningful precursor left. Safe to crash.
Heated shows butter, control does not — precursor is still there. Hold warm and retest in a day. This is the test working, not a failure.
Both show it — the yeast is not keeping up. Check whether it has flocculated out early, and whether the beer is warm enough.

Twenty minutes at 60°C converts most of the precursor rather than all of it, so treat a borderline result as a positive and give it longer.

Hop creep restarts the whole thing

Here is the part that catches people out on modern hoppy beer, and it is the reason a well-run IPA can end up buttery when a clean lager does not.

Dry hops carry active starch-degrading enzymes — most importantly glucoamylase. These survive drying and pelletising, and in the fermenter they break down dextrins that the mash deliberately left behind. Your yeast cannot touch those dextrins; that is the entire reason mashing at 67°C gives a fuller beer than 63°C. The hops undo it.

The extra attenuation is the least of it. What matters is that a restarted fermentation means restarted yeast growth, and growth is step 1. Fresh precursor, days after primary looked finished.

So the clock runs from the dry hop, not from the end of primary. A beer that had genuinely completed its diacetyl cycle gets handed a second one, and crashing on your original schedule puts the new precursor straight into the keg.

What to watch for

Gravity flattens, then starts falling again after the dry hop. A normal primary tail slows smoothly and levels off; a creep shows a distinct second shoulder
Attenuation overshooting what the grist was built for
On a pressure ferment, watch the gauge rather than the hydrometer. Roughly a tenth of a gravity point produces about 1 psi in a typical homebrew fermenter, where a glass hydrometer resolves maybe half a point. “Gravity flat but pressure climbing” is not a contradiction — it is what a small creep looks like on an instrument fine enough to see it. This only works while your spunding valve is set above the current pressure; once it is venting, the gauge stops telling you anything

Three things worth knowing

Warmth is the test. These enzymes work far faster warm. A flat gravity trace on a cool hold is not evidence of anything — the creep may simply be waiting for you to raise the temperature.

Cryo is not creep-free. Lupulin-concentrated pellets carry a lower enzyme load than whole T90 pellets, because the enzymes travel with the vegetal matter cryo removes. Lower is not negligible. On a dextrinous wort a large cryo charge still creeps.

Cold slows these enzymes, it does not destroy them. A keg that warms up in a garage or in transit can quietly resume attenuating months later. Keep heavily dry-hopped kegs genuinely cold.

If you bottle

This is the one part of the topic that is a safety issue rather than a quality one. Priming sugar calculations assume no further fermentation. If a creep is still live, the dextrin breakdown adds CO₂ on top of your priming sugar, and glass has a hard limit.

If you are bottling a heavily dry-hopped beer, confirm gravity is genuinely stable for several days after the dry hop — not after primary — or dry hop cold and short, or keg instead.

When to crash

Crashing is a one-way door. It slows the conversion and removes the yeast that performs the reduction, in a single action. Three conditions, all of them:

Gravity has genuinely levelled off, not merely slowed. If a creep is running, wait for it to finish.
On a pressure ferment, the gauge has flattened too. It sees the tail before gravity does.
A forced test comes back clean.

The competing pressure is usually the hops rather than the yeast. Sitting warm costs little biologically — esters and fusels need active growth, which is behind you, and autolysis is a matter of weeks not days — but polyphenol extraction off a big dry hop accrues the whole time. That is a real trade rather than “wait as long as possible”.

Where the two conflict, though, the call is not close. Hop harshness is a defect of degree. Diacetyl in the package is a defect of kind.

Other ways it gets in

Not every buttery beer is a timing problem.

A long lag phase. Under-pitching, tired yeast or poor oxygenation stretches the window in which precursor accumulates, so the batch simply makes more than the yeast can later clear. No length of rest fixes that. The fix is next batch: pitch to rate, check the date on the packet, aerate properly.

Under-nourished wort. Alpha-acetolactate is not a waste product — it is an intermediate on the route to the amino acid valine. Short the yeast of free amino nitrogen and it has to build its own valine, runs that pathway hard, and leaks correspondingly more precursor. Lean worts make more diacetyl, and the leanest worts are the high-adjunct ones: sugar, rice, corn and heavy flaked additions all contribute fermentable extract without the nitrogen that comes with malt. On those grists a yeast nutrient stops being insurance.

Poor aeration, twice over. Oxygen at pitch does not only prevent a long lag. The yeast’s reduction of diacetyl is energy-hungry and NADH-dependent, and the cell’s capacity to generate that energy depends on the membranes it built with the oxygen it got at the start. Under-aerated wort therefore makes more precursor and clears it more slowly — which is the strongest single argument for taking aeration seriously.

Yeast flocculating out early. A very flocculant strain can drop clear while the conversion is still running, which ends the cleanup prematurely. A gentle rouse or a warm hold keeps cells up.

Infection. Pediococcus and some Lactobacillus produce diacetyl directly, with no acetolactate involved and no yeast required. The tell is diacetyl that rises after packaging in a beer that passed a forced test, usually with sourness, haze or ropiness alongside. That is a cleaning problem, not a technique problem.

The short version

Your nose in the fermenter tells you almost nothing
Warm and wait; do not cool and hope
Dry hopping restarts the clock
Run the forced test before every crash that matters

See also: Dry Hopping & Whirlpool for charge rates and contact times, Pressure Fermentation & Closed Transfer for reading the gauge and setting a spunding valve, and Kegging & Carbonation for what a live creep does to your carbonation figures.