Tofu is soy milk that has been told to stop being a liquid. Heat unfolds the soy proteins; a pinch of salt with a doubly charged metal ion in it, magnesium or calcium, takes away the electrical charge that kept them apart; and within minutes the pot splits into soft white curds and clear yellow whey. In Japan the traditional salt is nigari, the bitter liquid left behind when sea salt is made. Its working ingredient is magnesium chloride.
This guide gives you a complete home recipe, a dose by weight rather than by guesswork, the chemistry of what happens in the pot, how nigari compares with gypsum, Epsom salt, glucono-delta-lactone (GDL) and lemon juice, and the fixes for tofu that will not set or comes out bitter. Where the published sources disagree, we show both sides.
We sell Magnesium Chloride Hexahydrate FCC, USP Grade, food-grade magnesium chloride in flake form, from a 2 lb pack up to full pallets. The buying section near the end explains what “FCC” and “USP” mean and why the grade matters for anything that goes into food.
Last reviewed 20 October 2026 against the sources cited below. This is a practical food chemistry guide, not medical, legal or regulatory advice. How we researched it is set out at the end.
What is nigari?
Nigari is the concentrated salt solution left over after sea salt crystallizes out of seawater, and its main component is magnesium chloride. Researchers at Tokyo University of Marine Science and Technology define it as “a concentrated solution remaining after the crystallization process of salt that has been used as a coagulant for tofu.” The Japanese word, にがり, comes from nigai, “bitter”, which tells you what it tastes like.
Natural nigari is not a single compound. A 1999 technical review for the USDA National Organic Program describes the traditional product as a mixture of magnesium chloride and magnesium sulfate, and notes that natural nigari carries about 4% magnesium sulfate. When the same team tested six commercial nigari products, the magnesium content ranged from 0.88 to 6.49 mol/L, a sevenfold spread. Shurtleff and Aoyagi, whose Book of Tofu introduced home tofu making to most English readers, distinguished “natural nigari”, which is beige, from “refined nigari”, which is white.
Refined nigari is, in practice, magnesium chloride. The flakes sold for tofu making are usually magnesium chloride hexahydrate (MgCl₂·6H₂O), the same compound that occurs as the mineral bischofite. That matters for one reason: a pure, specified salt doses the same way every time, while natural bittern varies from bottle to bottle. Andrea Nguyen, author of Asian Tofu, reported that one liquid nigari she tested “produced very yucky tofu.”
Name check: Magnesium chloride hexahydrate, MgCl₂·6H₂O, CAS 7791-18-6, molar mass 203.30 g/mol. International food additive number INS 511.
How do you make tofu with nigari?
You make tofu with nigari in four stages: make soy milk from soaked soybeans, cook it, stir dissolved magnesium chloride into the hot milk so it curdles, and press the curds in a mold. The whole process takes about an hour once the beans have soaked; Shurtleff and Aoyagi wrote that “the tofu will be ready 50 to 60 minutes after you start.”
Firm nigari tofu (one block, about 2 liters of soy milk)
- 250 g (about 1¼ cups) dried soybeans, soaked
- 2 liters of water for grinding, plus water for soaking
- 6 g magnesium chloride hexahydrate flakes (food-grade), weighed on a kitchen scale
- 120 mL (½ cup) water to dissolve the flakes
- Equipment: large heavy pot, blender, cotton straining bag or fine cloth, thermometer, wooden tofu mold or a perforated container lined with cloth, a 1–3 lb weight
The 6 g figure is our starting point of 3 g per liter of soy milk. Adjust by 1 g either way once you know how your beans behave. Steps below.
Step 1: Soak the soybeans
Cover the beans with plenty of cold water and soak them 8 to 12 hours at room temperature, longer in a cold kitchen. The Institute of Food Technologists gives 9 to 10 hours at 22 °C and 4 to 6 hours at 32 °C; Shurtleff and Aoyagi soak for 10 hours. Drain and rinse.
Step 2: Grind and cook the slurry
Blend the beans with the 2 liters of water in two or three batches, about three minutes each, until very smooth. The finer the grind, the more protein you get out; IFT notes that “the more thoroughly ground the slurry, the more protein extracted and the better the yield.” Pour the slurry into a large pot and bring it to a boil, stirring the bottom constantly, then simmer for 8 to 10 minutes. Cooking does two jobs: IFT says it “denatures the soy protein and removes volatile beany flavors.” Soy slurry foams up suddenly, so use a pot at least twice the volume and stay at the stove.
Step 3: Strain off the okara
Pour the hot slurry through a cloth bag set over a second pot and squeeze out as much milk as you can (heat-resistant gloves help). The pulp left in the bag is okara. You should end up with roughly 1.8 to 2 liters of soy milk. Okara is good food in its own right; it goes into baking, patties and stir-fries.
Step 4: Bring the soy milk to coagulating temperature
Bring the strained milk to between 70 and 85 °C (158 to 185 °F). America’s Test Kitchen aims for 165 to 175 °F; several home recipes use 180 °F. A thermometer takes the guesswork out of this step.
Step 5: Add the nigari in portions
Dissolve the 6 g of flakes in the 120 mL of water. Stir the soy milk once around the pot, then pour in about a third of the solution over the back of a spoon or paddle. Give it a few slow strokes, stop, cover the pot and wait 3 minutes. Sprinkle another third over the surface, stir gently once or twice, cover for another 3 minutes, then add the rest the same way. Andrea Nguyen’s method uses this rhythm: “Clamp on a tight-fitting lid and let the mixture sit for 3 minutes.” Over-stirring breaks the forming curd into small grains and lowers the yield.
Step 6: Rest, then press
Leave the covered pot 15 to 20 minutes. You should see white curds floating in clear, yellowish whey. Line the mold with damp cloth, ladle in the curds, fold the cloth over, put the lid on and add weight. Pressing time decides firmness: Shurtleff and Aoyagi press with ½ to 1½ lb for 10 to 15 minutes, Nguyen with 3 lb for 15 to 20 minutes, and America’s Test Kitchen gives “about 30 minutes for medium-firm and 40 minutes for firm.”
Step 7: Set in cold water and store
Slide the block, still in its cloth, into a bowl of cold water, unwrap it underwater and leave it a few minutes to firm up. Keep it refrigerated, covered with water. Nguyen’s rule: “Tofu will keep a week in the refrigerator if the water is changed every day.”
How much nigari per liter of soy milk?
Start with about 3 g of magnesium chloride hexahydrate flakes per liter of soy milk, and stay within roughly 2 to 4 g per liter. That figure sits in the middle of every quantitative source we found:
| Source | Dose as published | What it means for flakes |
|---|---|---|
| Institute of Food Technologists (2016) | “Between 1.5 and 5.0 g/kg” for tofu coagulants generally | 1.5–5 g per liter of soy milk |
| Yang & James, CSIRO, J Sci Food Agric (2013) | Nigari or GDL at 1.5–5.0 g/kg; firmest silken tofu at 2.5–3.0 g/kg | 2.5–3 g per liter for maximum firmness |
| US Patent 9,848,617 | 36 g magnesium chloride hexahydrate in 12 L of soy milk at 80 °C | Exactly 3 g per liter |
| US Patent 4,732,774 | Magnesium chloride at 0.3–0.8% of soy milk weight | 3–8 g per liter, or less if the patent meant anhydrous salt (it does not say) |
| Shurtleff & Aoyagi, The Book of Tofu | “1-1/4 teaspoons magnesium chloride or calcium chloride” for the milk from 1 cup of soybeans | Teaspoon measure; flake size varies |
| Andrea Nguyen, Asian Tofu | 1½ teaspoons nigari for 8 cups (about 1.9 L) of soy milk | Teaspoon measure; form of nigari not stated |
We give the dose by weight on purpose. Recipes that say “2 teaspoons of nigari” may mean liquid bittern, damp natural granules or dry refined flakes, and those are not the same strength. A teaspoon of coarse flakes also holds a different mass than a teaspoon of fine crystals. A kitchen scale that reads to 0.1 g removes the variable.
Why the hexahydrate number matters
Magnesium chloride flakes are a hydrate: each MgCl₂ unit carries six water molecules locked into the crystal. PubChem gives the molar mass of anhydrous magnesium chloride as 95.21 g/mol and the hexahydrate as 203.30 g/mol, so only 46.8% of the flake weight is magnesium chloride itself.
95.21 ÷ 203.30 = 0.468 → 1 g anhydrous MgCl₂ = 2.14 g hexahydrate flakes
That factor of 2.14 explains why research papers seem to disagree with recipes. Most studies write “0.4% MgCl₂” without saying whether they mean the anhydrous salt or the hydrate, so their numbers can be off by up to 2.14 times when you try to use them in a kitchen. The 3 g per liter figure above is stated as hexahydrate, which is what you are weighing.
Thicker milk needs more
The richer your soy milk, the more coagulant it can take. The patent dose was for soy milk at 12.5 °Brix, a fairly rich milk. If you grind with more water than our recipe, start lower (2 to 2.5 g per liter); if you grind thicker, you can go to 4. Researchers at Jiangnan University found a minimum protein level below which magnesium could not curdle the milk at all, so very thin soy milk will disappoint no matter how much nigari you add.
What does nigari do to soy milk? The chemistry of the curd
Nigari curdles soy milk because its magnesium ions cancel the negative electrical charge on cooked soy proteins, letting the proteins stick together into a gel. Two proteins do most of the work. A 2021 review in Foods calls 7S β-conglycinin and 11S glycinin “the key components of tofu curd, accounting for more than 70% of the total soy protein content.”
Heat comes first. In raw soy milk the proteins are tightly folded. Boiling unfolds them and, in the same review’s words, exposes “hydrophobic groups such as sulfhydryl groups, disulfide bonds, and hydrophobic amino acid side chains,” which makes them want to cling to each other. Glycinin needs more heat than conglycinin: its denaturation temperature is about 20 °C higher, at 85 to 95 °C, which is why a full simmer matters even though you add the coagulant cooler.
Then the charge goes. At the natural pH of soy milk, about 6.5, the unfolded proteins carry negative charges that push them apart. Magnesium ions, each with a double positive charge, neutralize those charges. Japanese researchers describe this as the formation of “salt bridges by MgCl₂ addition”, and found that magnesium chloride also nudges the pH down slightly, from 6.5 to 6.1, a drop they concluded “is essential for bringing proteins closer together.” Once the repulsion is gone, the exposed hydrophobic patches lock the proteins into a three-dimensional network that traps water. That network is tofu.
The short version: heat unfolds the soy proteins, magnesium switches off the charge that keeps them apart, and they knit into a gel. Too little magnesium and the charge stays on; too much and they knit too fast, into a coarse, bitter curd.
The precise mechanism is still argued over. A 2024 review concluded that “the underlying mechanisms, remain unclear,” and a 2025 study from Jiangnan University found that a direct “protein-Mg2+-protein bridge was hardly formed”, pointing to charge neutralization as the main effect. For the cook, the practical result is the same either way.
Acid coagulants work differently. GDL slowly turns into gluconic acid in water and lowers the pH toward soy protein’s isoelectric point, about pH 4.5, where the protein carries no net charge. In practice curds start well before that: one study measured the midpoint of curd formation at pH 6.3 with magnesium chloride and pH 6.0 with GDL.
Water treatment plants use the same principle, on a different scale. Aluminum and iron salts neutralize the charge on suspended particles so they clump and settle; we explain that version in our guide to aluminum sulfate in coagulation. Tofu is the edible version of the same physics.
Nigari vs gypsum vs GDL: which tofu coagulant should you use?
Use nigari (magnesium chloride) for pressed block tofu with a slightly sweet, mineral edge, gypsum (calcium sulfate) for a more forgiving set and a milder taste, and GDL for silken tofu set in its container. The Institute of Food Technologists sorts coagulants into “salts, acids, and enzymes”; home cooks mostly use the first two.
| Coagulant | Type | Typical tofu | Notes from the sources |
|---|---|---|---|
| Nigari / magnesium chloride | Salt | Firm block tofu | Nguyen: “Nigari coagulates quicker”, firmer tofu, “a touch bitter”; fast setting makes it harder to control |
| Gypsum / calcium sulfate | Salt | Soft to firm block tofu | IFT: the “most traditional and widely used” coagulant commercially, chosen because it “does not mask the taste of the soybeans” |
| Calcium chloride | Salt | Block tofu | IFT: chlorides are more soluble than calcium sulfate |
| Epsom salt / magnesium sulfate | Salt | Softer block tofu | Shurtleff & Aoyagi: “somewhat higher bulk yields and a softer end product” |
| Glucono-delta-lactone (GDL) | Acid | Silken tofu | IFT: lets silken tofu be made “inside the container”; “leaves a slight sour taste” |
| Lemon juice, vinegar | Acid | Block tofu, lower yield | Shurtleff & Aoyagi: “rather small yields”; IFT: may “leave undesirable flavors” |
On texture, the sources differ. IFT says magnesium and calcium chlorides “produce smoother-textured tofu” than calcium sulfate; a 2024 review reports the opposite, that calcium sulfate gives a “finer and smoother surface finish”; the USDA review cites work finding calcium sulfate tofu “both harder and more brittle than nigari tofu.” The honest reading is that dose, temperature and stirring matter as much as which salt you pick. Many tofu makers end up with a favorite and stick with it.
Nigari and gypsum also leave different minerals behind. USDA FoodData Central lists these values per 100 g:
| USDA entry (FDC ID) | Water | Protein | Calcium | Magnesium |
|---|---|---|---|---|
| Tofu, extra firm, prepared with nigari (174290) | 82.3 g | 10.0 g | 282 mg | 35 mg |
| Tofu, firm, prepared with calcium sulfate and magnesium chloride (172448) | 82.9 g | 9.0 g | 201 mg | 37 mg |
| Tofu, raw, regular, prepared with calcium sulfate (172476) | 84.6 g | 8.1 g | 350 mg | 30 mg |
| Tofu, raw, firm, prepared with calcium sulfate (172475) | 69.8 g | 17.3 g | 683 mg | 58 mg |
Compare rows with similar water content: firmer tofu holds less water, so every nutrient is more concentrated. Composition data only.
Why won’t my tofu curdle?
Tofu that will not curdle almost always has soy milk that is too cool, too thin, or not given enough coagulant. Work through them in that order.
- Too cool. Check the milk is at least 70 °C (158 °F). If it has cooled below that, reheat gently and wait.
- Too thin. Soy milk ground with a lot of water has less protein to form a curd. Next time grind with less water or squeeze the okara harder.
- Not enough coagulant. If the milk is hot and the whey still looks milky after 10 minutes, dissolve another 0.5 to 1 g of flakes in a little water and sprinkle it over the surface. Stir once, cover and wait.
- Weak nigari. Liquid bittern and natural granules vary a lot in strength. A specified salt removes that variable.
The sign of a good set is clear whey. If the liquid between the curds is pale yellow and see-through, the coagulation is complete; if it is still white and cloudy, there is protein left to catch.
Why does my homemade tofu taste bitter?
Bitter tofu usually means too much nigari, or nigari added too fast. Magnesium chloride is bitter on its own (the name says so), and any excess that does not bind protein stays in the whey and the curd. A 2021 review notes that tofu made with bittern tends to hold water poorly, and a 2024 review found that higher magnesium chloride concentrations produce rough rather than smooth curds. Both show up as a dry, crumbly block.
The fixes are straightforward. Cut the dose back by 1 g per liter. Add the solution in portions with rests between them instead of all at once. Press a little longer to drive out more whey, and let the block sit in cold water for a few minutes, changing the water once. One recipe developer who tested coagulants side by side noted that “the fresh unpressed curds were bitter,” one more reason to press well and rinse the block.
Is magnesium chloride safe to use in food?
Yes, when it is food-grade. The U.S. Food and Drug Administration lists magnesium chloride as Generally Recognized as Safe (GRAS) at 21 CFR 184.1426, used in food with “no limitation other than current good manufacturing practice.” The same regulation requires that the ingredient “meets the specifications of the Food Chemicals Codex.” The listed functional uses are as a “flavoring agent and adjuvant” and a nutrient supplement. Internationally, the FAO/WHO Joint Expert Committee on Food Additives classes magnesium chloride (INS 511) as a firming agent and set its acceptable daily intake as “not limited.” The USDA’s 1999 review notes that the material “has long been used in organic tofu production.”
The grade is the important word. Lab-reagent magnesium chloride is made to a chemical purity standard, not a food one, and one major lab supplier’s safety data sheet lists food use under “uses advised against.” De-icing and dust-control magnesium chloride is not food-grade either. For tofu, buy a product made to the Food Chemicals Codex (FCC) or United States Pharmacopeia (USP).
Handling and storage
- Keep it sealed. Magnesium chloride hexahydrate pulls water from the air. The FAO/WHO specification says “it is very deliquescent”: left open in a humid kitchen, dry flakes turn into a puddle. Store in a tightly closed container in a cool, dry place.
- Not a hazardous chemical, but not a drink. Our safety data sheet states that it is not classified as hazardous under GHS, and the large majority of company notifications to the European Chemicals Agency agree. Swallowed in large amounts, it causes stomach upset and diarrhea. Never drink concentrated nigari or undiluted coagulant solution; drinking concentrated bittern has caused serious poisoning.
- Label it. Dissolved coagulant looks like water. Mix it in a marked cup, use it the same day, and keep the flakes out of reach of children.
Which magnesium chloride should you buy for tofu?
Buy magnesium chloride hexahydrate that is labeled food-grade, ideally to the FCC or USP specification, in a resealable pack. Our Magnesium Chloride Hexahydrate FCC, USP Grade is that product: white flakes of MgCl₂·6H₂O, CAS 7791-18-6.
FCC means the material meets the Food Chemicals Codex monograph, the purity standard FDA’s GRAS listing for magnesium chloride cites. USP means it meets the United States Pharmacopeia monograph; federal food and drug law names the USP as an “official compendium.” The USDA review puts the point simply: there is “no distinction between refined and unrefined magnesium chloride, only magnesium chloride that meets the standard of identity set in the Food Chemicals Codex.” We explain the grade names in more depth in our guide to chemical grades.
At 6 g per batch, our smallest 2 lb pack covers well over a hundred blocks of tofu. Larger packs (5 lb, 55 lb bags and multi-bag pallets) suit small tofu shops and food producers. A certificate of analysis for a specific lot is available on request.
Magnesium chloride does a lot more than make tofu: road de-icing, dust control, Sorel cement, and battery research. Our magnesium chloride guide covers the industrial side, and magnesium chloride in energy storage and Sorel cement goes deeper on the engineering.
How we researched this
We read the sources below on 20 October 2026: the Institute of Food Technologists’ overview of tofu processing; peer-reviewed studies and reviews on soy protein coagulation in Foods, Heliyon, Food Chemistry: X, Gels and the Journal of the Science of Food and Agriculture; the USDA National Organic Program’s technical review of magnesium chloride; 21 CFR 184.1426 and related food additive regulations on eCFR; the FAO/WHO JECFA specification; USDA FoodData Central; PubChem; two tofu patents; and published recipes by William Shurtleff and Akiko Aoyagi, Andrea Nguyen and America’s Test Kitchen. Where sources disagree, as on texture and on dose units, we report both. The hexahydrate conversion and the per-liter doses are our own arithmetic, shown in full. Product names and pack sizes come from our live catalog. The photographs in this article are illustrations, not test results.
Sources
- McHugh T., How Tofu Is Processed, Food Technology, Institute of Food Technologists (2016)
- Kuda T., Yano T., Mineral composition of seawater bittern nigari products and their effects on changing of browning and antioxidant activity in the glucose/lysine Maillard reaction, Applied Biochemistry and Biotechnology (2014)
- Magnesium Chloride (Processing), Technical Advisory Panel Review, USDA Agricultural Marketing Service, National Organic Program (1999)
- Guan X. et al., Changes of Soybean Protein during Tofu Processing, Foods (2021)
- Geng Y. et al., Research Progress on Tofu Coagulants and Their Coagulation Mechanisms, Foods (2024)
- Arii Y., Sano Y., Nishizawa K., Direct comparison of the tofu-like precipitate formation by adding different coagulants: magnesium chloride and glucono-δ-lactone, Heliyon (2021)
- Arii Y., The role of pH decrease in tofu-like precipitate formation using magnesium chloride, Heliyon (2024)
- Li Q. et al., Metal ions binding and colloidal destabilization in the model systems: implication on the magnesium chloride coagulation mechanism in tofu making, Food Chemistry: X (2025)
- Li Z. et al., Compound salt-based coagulants for tofu gel production: balancing quality and protein digestibility, Gels (2025)
- Yang A., James A.T., Effects of soybean protein composition and processing conditions on silken tofu properties, Journal of the Science of Food and Agriculture (2013)
- Wang R. et al., Preparation and characterization of soy protein isolate-collagen self-assembled nanomicelles, RSC Advances (2018)
- Darmajana D.A., Afifah N., Indriati A., The effect of extraction method and types of coagulants on the results and physicochemical properties of tofu, Food Science and Technology, Campinas (2020)
- 21 CFR 184.1426, Magnesium chloride, Electronic Code of Federal Regulations
- 21 CFR 184.1230, Calcium sulfate, Electronic Code of Federal Regulations
- Magnesium chloride hexahydrate (INS 511), JECFA evaluation, IPCS INCHEM
- Magnesium chloride specification, FAO JECFA Monograph 1
- 21 U.S.C. 321, Definitions, U.S. Government Publishing Office
- FoodData Central, SR Legacy tofu entries (FDC IDs 174290, 172448, 172476, 172475), USDA Agricultural Research Service
- Magnesium chloride hexahydrate (CID 24644), PubChem, National Library of Medicine
- US 9,848,617 B1, Coagulant composition for tofu, Google Patents
- US 4,732,774 A, Process for preparing tofu charged into a container, Google Patents
- Shurtleff W., Aoyagi A., How to Make Tofu (from The Book of Tofu), Mother Earth News (1976)
- Nguyen A., Tofu Coagulant Buying Guide, Viet World Kitchen (2012)
- Homemade tofu, adapted from Andrea Nguyen’s Asian Tofu, Portland Press Herald (2017)
- Homemade Tofu (Zizhi Doufu), America’s Test Kitchen
- Which Tofu Coagulant is BEST?, Mary’s Test Kitchen
Frequently asked questions
Is nigari the same as magnesium chloride?
Mostly. Natural nigari is seawater bittern, a mixture whose main component is magnesium chloride with some magnesium sulfate, potassium chloride and other salts. Refined nigari and most nigari flakes are magnesium chloride. A specified food-grade magnesium chloride doses more consistently than natural bittern.
How much nigari do I need for 1 liter of soy milk?
About 3 g of magnesium chloride hexahydrate flakes, dissolved in a little water. Use 2 to 2.5 g for thin soy milk and up to 4 g for rich soy milk. Weigh it rather than using spoons, because flake size changes how much fits in a teaspoon.
What temperature should soy milk be when adding nigari?
Between 70 and 85 °C (158 and 185 °F). Research studies typically coagulate at 80 to 85 °C, and America's Test Kitchen aims for 165 to 175 °F.
Can I use Epsom salt instead of nigari?
Yes. Epsom salt (magnesium sulfate) also coagulates soy milk. Shurtleff and Aoyagi found it gives somewhat higher bulk yields and a softer tofu than nigari. Use food-grade Epsom salt only, not a bath product with added fragrance.
Can I use lemon juice or vinegar to make tofu?
Yes, but expect a lower yield and a slightly sour, crumblier tofu. Shurtleff and Aoyagi reported that lemon juice and vinegar give rather small yields, and IFT notes that acids like acetic and citric acid can leave off-flavors.
Do I need a tofu press?
No. Any container with drainage holes, lined with cloth, works as a mold, and a jar of water or a couple of cans makes a 1 to 3 lb weight. A wooden mold with a fitted lid presses more evenly.
How long should I press homemade tofu?
About 15 to 20 minutes for soft to medium tofu and 30 to 40 minutes for firm tofu, with 1 to 3 lb of weight. Longer pressing gives a firmer, denser block.
Why is my tofu crumbly?
Usually too much coagulant or too much stirring after adding it. Both make the curd form fast and in small, coarse grains. Reduce the dose, add it in portions, and stop stirring once the curds appear.
How long does homemade tofu last?
About a week in the refrigerator, submerged in water that you change every day.
How should I store magnesium chloride flakes?
In a tightly sealed container in a cool, dry place. The flakes are very deliquescent, which means they absorb moisture from the air until they dissolve, so an open bag will turn to liquid in a humid kitchen.
About the author
Andre Taki
This article is for informational purposes only.
