Mixed-bone demi-glace, batch 1

Bitterness in stock cannot be cooked out. Reduction concentrates it in proportion to everything else. Stop reducing the moment you notice it.

Taste and place it.

• Ashy, back of the palate, lingers after swallowing. Burnt bone or burnt fond scraped off the roasting tray.
• Sharp, mid-palate, fades quickly. Scorched allium. Garlic is the usual culprit. See the note on the garlic ingredient record.
• Bitter with a sour edge and tomato behind it. Tomato paste taken past pince into carbonisation.

1. Chill and lift the fat. Do this first, always. Many bitter compounds from scorching are lipophilic and partition into the fat phase. Removing the cap removes them with it. This is subtraction, not masking.

2. Add glutamate. The strongest available fix and the one that improves the sauce rather than compromising it. Umami suppresses bitterness centrally. Options: a piece of kombu steeped at 60 C for 20 minutes then removed; a teaspoon of white miso whisked in and strained; a dried shiitake. Kombu is what worked on batch 1 and it worked well.

3. Acid. Three drops of sherry vinegar or lemon per 100 ml, off the heat. Purely perceptual. Shifts attention forward on the palate and shortens the bitter finish.

4. Sugar. 0.25 g per 100 ml. Below the threshold where it reads as sweet. Works by central suppression, where sweet and bitter signals interfere downstream of the receptors. It also dulls savoury top notes, so use it last and sparingly.

5. Casein fining. 20 ml full-fat milk per 100 ml of stock, warmed to 70 C, held 10 minutes, strained through a coffee filter. Casein micelles bind bitter compounds through their proline-rich regions and carry them out. Real removal. Best against polyphenols, weaker against the organosulfur compounds that come from scorched garlic. Costs body and can cloud the result. Try it last.

6. Dilution with a clean stock. The only complete fix. Make a clean 5 hour stock with no tomato and no roasted garlic, and blend. Bitter perception is non-linear because TAS2R receptors have very low activation thresholds, so halving the concentration drops perceived bitterness by considerably more than half.

• A potato. It absorbs liquid at the same concentration as the surrounding liquid. Nothing in a cooked potato transports anything selectively. Removing it removes bitter compounds and water in identical proportion, so the concentration is unchanged. This is the same reason the potato does not fix oversalted soup. It also leaches starch and clouds the stock.
• More reduction. Makes it worse in direct proportion.
• More aromatics. Adds flavour without removing anything.
• A raft, at this stage. A raft needs thin liquid and a slow rise from cold, so it will not form properly in partly reduced stock. It also soaks up liquid and gelatin, which is a large loss at small volume. Egg protein binds tannins well but binds small sulphur compounds poorly. If you clarify at all, do it at full volume straight after straining, before any reduction.

Stock is a base, not a finished sauce. Salt and fat both suppress bitterness, and both are still to come. Make one test pan sauce, properly salted and mounted with butter, before deciding a batch is ruined.

Two separate mechanisms, both pointing the same way.

Clarity. At a rolling boil, vapour bubbles nucleate on the pan floor and collapse in the bulk liquid. That generates local shear high enough to break rendered fat into sub-micron droplets, which are then stabilised by denatured protein fragments and phospholipids acting as emulsifiers. Once that emulsion forms it is thermodynamically stable. Chilling will not break it. The result is a cloudy, greasy stock with a dull finish that will not mount cleanly with butter. Below the boil there is no bubble nucleation at all, only thermal convection, which gives the same mass transfer without the mechanical damage.

Gel strength. Collagen hydrolysis to gelatin and the subsequent degradation of gelatin into short peptides are both rate processes following Arrhenius kinetics, roughly doubling per 10 C. Gel strength depends on chain length. Holding 10 C below the boil roughly doubles the safe residence time, so 8 hours at 90 C does about as much chain damage as 4 hours at 100 C. This is why the usual six hour ceiling for poultry bones does not apply at sub-simmer temperatures.

Practical note. Albumin coagulation is essentially complete by the time the liquid passes 80-85 C. The skimming window is the first 30 to 60 minutes. After that there is very little left to skim.

At equal temperature the chemistry is identical. The oven is a control mechanism, not a technique.

What the oven buys you is that it cannot overshoot its setpoint, and that you can leave the house. Most domestic hobs on their lowest setting still sit above 90 C under a lidded pot, and gas in particular will creep up to a simmer over several hours unattended. If your hob holds 88-92 C with a probe in the pot, stay on the hob.

The oven also heats the pot on all faces rather than one, so there is no vertical gradient and no hot base. During extraction, with thin liquid, natural convection handles that fine either way. The difference only becomes material during the final reduction, when the liquid is viscous and convection has stalled, and by then you are standing at the pan anyway.

Practical points for an overnight oven run. 110 C is a very light duty cycle, well below a roast. Starting at 2700 ml with the lid cracked loses roughly 15 to 20 percent over eight hours, so it cannot run dry. Check the appliance for an automatic shutoff first, since many European models cut out around 12 hours. Running cost is roughly 3 kWh for the full run.

An immersion circulator in a covered stockpot is the better tool if you have one. It holds the target band exactly and is built for unattended overnight running.