Foundations · Article 25

Chemical Leavening

A reference article · 4,100 words · 47 cited sources · last reviewed October 2026 · fact-checked October 2026, 41 claims

Ingredient Science explains what soda and baking powder are. This article supplies what it leaves out: the arithmetic that lets you write a leavening line from scratch, balance it against the acids already in the formula, and read a soapy, yellowed, or collapsed result for what it is.

Every chemically leavened bake is a small acid-base reaction run inside a batter, and it either comes out balanced or it does not. Balanced, the bicarbonate is fully consumed, the crumb sits near neutral, and the gas arrives during the minutes in the oven between the batter loosening and the crumb setting. Unbalanced one way, unreacted soda decomposes in the heat and leaves sodium carbonate behind: the soapy taste, the yellow tint, the too-fast browning. Unbalanced the other way, the crumb tastes sour or metallic and bakes paler. The difference is measurable, and the measurements are not hard to use.

What a gram of soda can do

Sodium bicarbonate releases carbon dioxide two ways, and only one of them is what you want. With an acid and water present, one molecule of bicarbonate gives one molecule of CO₂ and leaves a neutral salt. Heated without acid, it decomposes on its own, yields half as much gas per gram, and leaves sodium carbonate, which is strongly alkaline, in the crumb (McGee, 2004). The theoretical yield with acid is the ratio of the two molar masses, 44 over 84, or 0.52 g of CO₂ per gram of soda, the figure BAKERpedia and the American Society of Baking both publish. Potassium bicarbonate, the sodium-free substitute, yields 0.44; ammonium bicarbonate, the baker’s ammonia of thin cookies, yields 0.56 (BAKERpedia, n.d.-a).

A double-acting baking powder packages the soda with a fast acid, a slow acid, and a starch buffer. Figoni’s from-scratch version is 30 g soda, 70 g cream of tartar, and 15 g cornstarch, which works out to 26 percent soda and, at cream of tartar’s neutralizing value, is almost exactly balanced (Figoni, 2011, via Lawandi, n.d.). That is why the substitution rules agree that a teaspoon of powder carries about a quarter teaspoon of soda (Lee, 2021). Run the other way, the same figures say a teaspoon of soda (4.6 g) carries the bicarbonate of about four teaspoons of powder, or 1 g of soda about 3.8 g of powder; the familiar kitchen swap of three teaspoons of powder per teaspoon of soda understates it (Lee, 2021). American household powders pair monocalcium phosphate as the fast acid with sodium aluminum sulfate or sodium aluminum phosphate as the slow one; European powders rely on sodium acid pyrophosphate, often as the only acid (BAKERpedia, n.d.-b).

Neutralizing value

The number that lets you balance acid against soda is the neutralizing value, defined as the parts by weight of sodium bicarbonate that 100 parts of the acid will neutralize (BAKERpedia, n.d.-a; Innophos, n.d.). The formula used identically by BAKERpedia and ICL Food Specialties is acid needed = soda × 100 ÷ NV. Ten grams of soda calls for 12.5 g of monocalcium phosphate (NV 80) or 22 g of cream of tartar (NV 45). An acid with NV 100 is used gram for gram with soda.

Several published values do not reconcile and are worth knowing about. Sodium acid pyrophosphate is 72 in every source, including BAKERpedia’s own formulation table, except BAKERpedia’s SAPP entry, which says 70; industry uses 72. Calcium acid pyrophosphate varies by grade, 60 to 73 in ICL’s table (ICL Food Specialties, 2025) and 67 in the USDA’s technical review (U.S. Department of Agriculture, Agricultural Marketing Service, 2018); the figure shows 67 as representative, and a CAPP and monocalcium phosphate blend is sold as a one-for-one replacement for SAPP 28 (Gorton, 2014). Coated anhydrous monocalcium phosphate is 80 in ICL’s table and 83 in BAKERpedia’s. And the trade table’s figures for citric (159), tartaric (116), and ascorbic acid (52) all exceed the stoichiometric ceilings for those molecules. The calculated maximum of 131 is used here for citric; tartaric and ascorbic are shown as published.

Neutralizing value is measured by titration and tracks the stoichiometry closely, though not exactly: monocalcium phosphate’s 80 is above its simple two-proton value because it reacts on in the dough. It tells you how much acid is eventually consumed, not when. Two acids with the same value can behave completely differently in the oven, which is the subject of the next section, and a balanced formula finishes near neutral pH, which is the design goal of every commercial powder (Innophos, n.d.). An acid in excess carries its own flavor: McGee notes that sulfates and pyrophosphates are distinctly astringent (McGee, 2004).

Rate of reaction

The industry measure is the dough rate of reaction: the percentage of an acid’s available gas released during mixing, during a bench rest at 27 °C, and in the oven (Innophos, n.d.; ICL Food Specialties, 2025). Innophos publishes the profile for its own products, and the spread is the whole story of leavening design. Monocalcium phosphate monohydrate releases roughly 60 percent of its gas in the mixer, a few percent more on the bench, and the rest in the oven. Dicalcium phosphate dihydrate releases nothing until heated. Sodium aluminum phosphate releases roughly 22 percent in the mixer, 9 on the bench, and 69 in the oven; BAKERpedia’s profile for it puts even less in the mixer. Sodium acid pyrophosphate grades are named for their early release: SAPP 28 gives about 28 percent of its gas in the first minutes of the test, SAPP 40 to 45 about 40 to 45 percent, and the rest comes in the oven, with higher numbers reacting faster (Gorton, 2014; ICL Food Specialties, 2025). Glucono-delta-lactone is the slow, temperature-driven acid of the set, about 10 percent at two minutes, 20 at ten, and 70 in the oven (American Society of Baking, n.d.-a).

Bellido, Scanlon, Sapirstein, and Page (2008) put the rate test on a pressuremeter and found that GDL and adipic acid follow first-order kinetics while potassium acid tartrate and SAPP 40 fit a two-fraction model, one fast and one slow, which is the kinetic signature of double-acting behavior inside a single acid.

Why timing matters more than quantity was shown directly by Godefroidt, Ooms, Bosmans, Brijs, and Delcour (2021), who baked cream cakes in an electrical-resistance oven while monitoring the headspace gas. Early-acting organic acids released much of their CO₂ in the mixer, where it could not be held, and gave low-volume cakes. SAPP and the late-acting acids gave high-quality cakes. Gas produced after the crumb had set was simply wasted. The window that counts opens when the batter goes in and closes when starch gelatinization and protein coagulation fix the structure, and a leavening acid is chosen for how much of its gas lands inside it. This is also why a from-scratch soda-and-cream-of-tartar blend, which is fast and single-acting, has to be baked at once, while a SAPP or SALP powder tolerates a rest.

How much

For converting the rules that follow, a teaspoon of baking powder is taken at 4 g (King Arthur Baking, n.d.) and a teaspoon of baking soda at 4.6 g (U.S. Department of Agriculture, n.d.), and a US cup of all-purpose flour at 125 g, the figure the per-cup rules assume. The sources differ: the USDA lists powder at 4.6 g a teaspoon, and King Arthur lists soda at 6 g and flour at 120 g a cup, so treat a teaspoon conversion as good to about 15 percent for powder and about 30 percent for soda.

The home rule, stated by Corriher, is 1 to 1¼ teaspoons of baking powder per cup of flour and, where soda is used, ¼ teaspoon per cup with enough acid to consume it (Corriher, 2008). In baker’s percentage that is about 3.2 to 4 percent powder and about 1 percent soda. Pastry Arts sets a slightly higher ceiling, no more than five to six grams of powder, 4 to 4.8 percent, or one gram of soda per 125 g of flour (Kohout, n.d.).

Professional formulas run higher. Gisslen’s muffins carry 6 percent powder, his biscuits 6, his scones about 5, his pancakes nearly 7, and his creaming-method yellow cake 5; where buttermilk enters, the powder drops and soda appears, to 2.5 percent powder and 1.25 percent soda in the buttermilk cake, or 5 and 1.25 in buttermilk biscuits (Gisslen, 2005). These are the fourth-edition print figures. BAKERpedia’s industrial guidance is 3 to 5 percent powder on flour for cakes and over 6 for scones (BAKERpedia, n.d.-b). The home rule is the safer starting point for a hand-mixed all-purpose cake.

Gisslen’s governing rule is the one to carry: “The amount of soda used in a formula is generally the amount needed to balance the acid. If more leavening power is needed, baking powder, not more soda, is used” (Gisslen, 2005). And on the ceiling: “Do not include more baking powder than necessary in a formula because undesirable flavors may be created. Also, excess leavening may create an undesirably light, crumbly texture. Cakes may rise too much and then fall before they become set.”

Two experiments bracket the band. Book and Brill (2015), at a leavening-acid manufacturer, ran yellow cake at soda levels of 1.4, 3.0, and 4.6 percent of flour and found maximum volume at 3.0 percent, under-leavening at 1.4, and collapse at 4.6. At 26 to 30 percent soda in a powder, 3.0 percent soda is about 10 to 11.5 percent of a finished powder, a commercial high-ratio level. Asamoah and colleagues varied powder from about 1 to 4.5 percent of flour in pound cake and found volume and porosity rising and then falling near 4.5 percent; cakes at 2.75 percent, the design’s center point, stood out (Asamoah et al., 2023). That sits just below the home rule; the yellow-cake figure shows how far a commercial layer cake pushes past it.

What the recipe’s own acids can carry

When a formula leavens with soda alone, the acid is an ingredient, and the soda must not exceed what that ingredient can neutralize. McGee’s rule of thumb is that half a teaspoon, 2 g, of baking soda is neutralized by a cup of fermented milk or 1¼ teaspoons of cream of tartar (McGee, 2004); both check out. His companion figure, a teaspoon of lemon juice or vinegar, does not: lemon juice carries about 48 g of citric acid per liter (Penniston et al., 2008), so a teaspoon neutralizes only about 0.3 g of soda, and a teaspoon of 5 percent vinegar about 0.35 g. Neutralizing 2 g of soda takes about two tablespoons of either.

The buttermilk figure can be checked. Cultured buttermilk is fermented to about pH 4.5 (Gettys & Davidson, 1985), typically reaching a titratable acidity of about 0.75 to 0.9 percent as lactic acid; the US standard for cultured buttermilk sets the floor at 0.5 percent (21 CFR 131.112). A cup of buttermilk at 0.9 percent carries 2.2 g of lactic acid, and neutralizing it takes 2.05 g of bicarbonate, which is McGee’s 2 g and about half a teaspoon. At the 0.5 percent floor the figure falls to about 1.1 g, a quarter teaspoon. At 0.8 percent the figure is about 1.8 g. The rule sits at the acidic end of real buttermilk, so a milder product leaves some soda unreacted. Sour cream is defined at not less than 0.5 percent titratable acidity (21 CFR 131.160), about 1.1 g of soda per cup at the minimum, and yogurt only by a pH of 4.6 or below (21 CFR 131.200).

Natural cocoa runs about pH 5.3 to 5.8, and alkalized cocoas about 6.5 to 7.2 when lightly treated, 7.2 to 7.6 medium, and above 7.6 heavy (Miller et al., 2008); BAKERpedia’s bands run about half a unit lower (BAKERpedia, n.d.-c). Gisslen’s balancing table is the only published quantity: natural cocoa carries 80 g of soda per kilogram, Dutch-process none, unsweetened chocolate 50 g per kilogram, sweet chocolate 25, and “if you are substituting dutched for natural cocoa, you must increase the baking powder by 1 oz (30 g) for each ½ oz (15 g) soda omitted” (Gisslen, 2005). King Arthur’s home version of the same swap, for recipes with more than three tablespoons of cocoa: replacing Dutch with natural, use half the amount of baking soda in place of the powder; replacing natural with Dutch, use twice the amount of powder in place of the soda (Hamel, 2020). Dutch cocoa in a soda-only recipe leaves the soda largely unneutralized: little lift and a soapy taste (Hamel, 2020).

Honey averages pH 3.9 but buffers weakly. For honey and molasses only rules of thumb exist, such as a quarter teaspoon of soda per cup of honey (University of Missouri Extension, n.d.), and none for brown sugar; all three are used as the acid in Gisslen’s molasses-muffin and gingerbread formulas at 0.75 to 1.4 percent soda. The calculator treats them as acid sources without a quantity and says so.

Substituting one system for another

Powder from soda and acid: per teaspoon of baking powder, ¼ teaspoon soda plus ½ teaspoon cream of tartar plus ¼ teaspoon cornstarch, or ¼ teaspoon soda plus ½ teaspoon lemon juice or vinegar (Lee, 2021). By neutralizing value both fall well short of balanced: the cream of tartar neutralizes only about three-fifths of the soda, and the lemon juice or vinegar about a seventh, so expect some soda flavor. The essential caveat is that both are single-acting: the gas is released in the bowl, and the batter must be baked at once.

Soda from powder: King Arthur’s kitchen rule is three times the volume of baking powder; by bicarbonate content it is closer to four, and Gisslen’s professional swaps use about twice the weight (Gisslen, 2005). Expect a slightly bitter note from the extra acid salts, and accept that whatever acidic ingredient the soda was meant to neutralize will now stay acidic (Lee, 2021).

Buttermilk for milk: half a teaspoon of soda per cup of buttermilk, and two teaspoons of baking powder removed for each half-teaspoon of soda added, is a common home conversion consistent with the acidity data (For Love of the Table, 2011); Gisslen’s professional version is in the callout above.

Too much and too little

Any bicarbonate the acids cannot consume decomposes in the oven, and the crumb finishes alkaline. McGee: “a bitter, soapy, or ‘chemical’ flavor results. Colors are also affected in even slightly alkaline conditions: browning reactions are enhanced, chocolate turns reddish, and blueberries turn green” (McGee, 2004). The yellow tint of an alkaline crumb comes from flavonoids in the flour that are colorless at normal pH and yellow at high pH; in alkaline noodles the compounds have been identified as apigenin-C-diglycosides and their esters (Asenstorfer, Wang, & Mares, 2006), and the same flour flavonoids are the likely cause in an over-sodaed crumb. BAKERpedia adds coarse grain and low volume in muffins (BAKERpedia, n.d.-d). Pop (2007) baked a cake series at crumb pH 5.0, 6.5, and 8.5 and found the high-soda cake darker, coarser, and tasting “like a soap film on the tongue”; Stahl and colleagues (2009) measured commercial chocolate cake mixes finishing above pH 8.3 and recommended a finished pH of about 7.25 or less for good leavening and retention of cocoa flavanols.

Excess leavening of any kind gives Gisslen’s “undesirably light, crumbly texture,” and a cake that rises too far and falls before it sets; Book and Brill’s 4.6 percent cake collapsed. Over-expansion is also the characteristic high-altitude failure, which is why the leavening cut is the first correction in Environmental Variables & Troubleshooting.

Too little leavening gives a dense, heavy crumb. Soda without enough acid shows the other failure: in America’s Test Kitchen’s comparison, soda-only cornbread and cookies were deep brown, the cornbread dense and the cookies spread instead of rising, while powder-only cornbread tasted sour and metallic (America’s Test Kitchen, n.d.-a); the sourness is the recipe’s acid left un-neutralized, and excess leavening acid can also leave a bitter aftertaste (Pop, 2007).

Storage, potency, and baker’s ammonia

Soda, powder, and ammonia “must always be kept tightly closed when not in use. If left open, they can absorb moisture from the air and lose part of their leavening power” (Gisslen, 2005). The starch in baking powder is there to absorb that moisture and prevent premature reaction (McGee, 2004). Producers claim about a year once opened; America’s Test Kitchen found biscuits made with ten-month-old powder rose to half the height of fresh and recommends discarding an opened tin after six months (America’s Test Kitchen, n.d.-b). The test: half a teaspoon of powder in two tablespoons of warm water should foam at once, and half a teaspoon of soda in a tablespoon of vinegar should fizz immediately and vigorously (Hamel, 2015), though America’s Test Kitchen found even year-old powder passes the water test, so the opening date is the better guide (America’s Test Kitchen, n.d.-b).

Ammonium bicarbonate decomposes on heating into carbon dioxide, ammonia gas, and water and leaves no salt residue, which is why it gives exceptionally crisp, light thin cookies and crackers and none of soda’s aftertaste (McGee, 2004). The ammonia has to escape, so it belongs only in small products baked until dry, under about 5 percent moisture: Leckerli, Spekulatius, cream puffs, crackers (BAKERpedia, n.d.-e). It decomposes completely and leaves no residue (Gisslen, 2005). Sources differ on when it starts: BAKERpedia gives both above 34 °C (93 °F) and above 40 °C (104 °F) (BAKERpedia, n.d.-e; BAKERpedia, n.d.-f). The practical reading is that it begins gently in the mid-30s to 40 °C and is spent in the oven. Van der Sman (2021), modeling a biscuit dough leavened mostly with soda, found that almost all the gas before the bubbles rupture is carbon dioxide from bicarbonate, with little ammonia in the gas phase, which questions how much ammonium bicarbonate actually contributes; the work is a model, not a baking measurement.

Aluminum, sodium, and the newer acids

Sodium aluminum sulfate and sodium aluminum phosphate are the cheap, potent, heat-triggered acids of American powders and mixes. Sodium aluminum sulfate can leave a slight metallic aftertaste, though most tasters in America’s Test Kitchen’s comparison could not tell (America’s Test Kitchen, n.d.-c); sodium aluminum phosphate is described as bland (American Society of Baking, n.d.-b). The aluminum-free market exists for taste and for the aluminum question, and the question is real: the aluminum in SALP is bioavailable, at about 0.11 to 0.13 percent from a baked biscuit in rats (Yokel & Florence, 2006), and a week of SALP-leavened pancakes more than doubled urinary aluminum excretion in healthy adults (Glynn & Lignell, 2019). The European Food Safety Authority re-evaluated the aluminum sulfates and sodium aluminum phosphate in 2018 and found them of no safety concern at authorized uses, because they are permitted in only a few specific products and exposure from them is “probably near zero” (EFSA ANS Panel, 2018). In the United States, where SALP remains common in pancake mixes and powders, Saiyed and Yokel (2005) measured up to 180 mg of aluminum per serving in some baking powders, pancake and waffle mixes, and ready-to-eat pancakes. Aluminum-free powders substitute SAPP, or MCP with calcium acid phosphate, for the aluminum salt.

On the sodium side, potassium bicarbonate substituted for sodium bicarbonate in cookies matched spread and baking performance without harming sensory scores (Chen, Hu, & Li, 2020), and a calcium acid pyrophosphate blend replaces SAPP 28 one for one (Gorton, 2014). Gélinas’s reviews of the patent record since 1833 show that more than half of all leavening-acid inventions concern phosphates and that the current wave is sodium-free, aluminum-free, and phosphate-free (Gélinas, 2021; Gélinas, 2022a; Gélinas, 2022b). Fat-coated, encapsulated bicarbonate is used to hold leavening in refrigerated batters until the coating melts; in one study batter pH rose in the first week, suggesting some early release, and in microwave-baked muffins the coating’s melt point affected volume after storage, but not in a simple way (Dorko & Penfield, 1993).

A worked leavening line

A buttermilk cocoa layer cake, flour 1,000 g, buttermilk 700 g, natural cocoa 60 g. The buttermilk at 0.9 percent lactic acid carries 6.3 g of acid and can consume 5.9 g of soda; McGee’s rule gives 5.7. The cocoa at Gisslen’s 80 g per kilogram carries 4.8 g. The acids can absorb about 10.5 g of soda, just over 1 percent of flour, at the ceiling of the home band. How much powder that soda replaces depends on the equivalence. By bicarbonate content, at a powder’s 26 percent soda, it is about 40 g, or 4 percent; by Gisslen’s own swap of about 2 g of powder per gram of soda, about 21 g. Against a creaming-method target of 3.2 to 5 percent powder-equivalent, the balance from baking powder is none to about 1 percent on the first basis and about 1 to 3 percent on the second, per Gisslen’s rule that extra lift comes from powder and never from more soda. The calculator uses the bicarbonate basis. If any of the acid is fast, bake promptly; a SAPP or SALP powder tolerates a rest. The Leavening Calculator runs exactly this arithmetic.