Ultrasound and the Right Coagulant Combine to Make Better Tofu

Tofu has been made for centuries from little more than soybeans, water, heat, and a coagulating agent, yet the process remains surprisingly hard to perfect. The final block’s texture, yield, protein content, and even its microbial safety depend on a web of interacting variables that manufacturers have mostly tuned by trial and error. A new study published in Frontiers in Soft Matter brings statistical rigor to that craft, showing that the combination of high-intensity ultrasound treatment of soymilk and the choice of coagulant can be modeled, predicted, and optimized to produce tofu that is firmer, safer, and better liked by consumers.

Researchers at the Universidad de las Américas Puebla in Mexico set out to answer a deceptively simple question: what matters more, how the soymilk is pretreated or which coagulant is used to turn it into curd? To find out, they prepared soymilk in three ways—untreated, conventionally pasteurized at 60 °C for 30 minutes, and treated with thermal high-intensity ultrasound, or thermosonication, at the same temperature and duration. Each of the three soymilk types was then coagulated with one of three agents: calcium chloride (CaCl2), magnesium chloride (MgCl2), the main component of traditional nigari, or acetic acid. The result was a nine-tofu experimental matrix, each combination replicated three times and analyzed in triplicate.

The physical basis for expecting ultrasound to matter lies in acoustic cavitation. When high-intensity, low-frequency ultrasound propagates through a liquid, microscopic bubbles form and collapse violently, generating localized hotspots of extreme pressure and temperature, along with micro-streaming and shear forces. In soymilk, these forces disrupt the compact tertiary and quaternary structures of globular soy proteins, unfolding them and exposing functional groups—hydrophobic amino acid residues and free sulfhydryl groups—normally buried inside. Once exposed, these reactive sites are free to form covalent disulfide bonds and divalent ionic salt bridges during coagulation, producing a denser, more homogeneous protein network. The same process also inactivates lipoxygenase, the enzyme responsible for generating hexanal and other volatile carbonyl compounds behind soy’s notorious “beany” off-flavor.

On the microbial side, the team found that thermosonication outperformed pasteurization alone. Ultrasound reduced aerobic mesophilic bacteria by roughly 2.5 log cycles compared with about 2 log cycles for pasteurization, and it cut total coliforms further as well. Modeling the inactivation kinetics with the nonlinear Weibull equation—whose shape parameter exceeded 1 in every case, indicating an initial shoulder of resistance before cells rapidly succumbed—showed that decimal reduction times dropped from 21.1 to 19.4 minutes for mesophiles and from 25.9 to 24.6 minutes for coliforms when ultrasound was added. The researchers attribute the effect to cavitation physically weakening bacterial cell walls, raising membrane permeability, and generating reactive oxygen species that damage DNA and lipids, all of which sensitize microbes to concurrent heat.

When the nine tofu formulations were finally pressed and analyzed, the coagulant emerged as the dominant factor shaping macroscopic quality. Calcium chloride produced the highest yield at 14.6 percent, followed by magnesium chloride at 12.8 percent and acetic acid at 11.8 percent, with protein content following the same ranking at 10.2, 8.5, and 6.6 percent respectively. The chemistry explains the hierarchy: divalent calcium and magnesium cations neutralize the negatively charged carboxyl groups on soy globulins, bridging them into tight, stable networks. Calcium, in particular, forms more effective bridges with glycinin, yielding a compact gel that traps fewer soluble solids in the whey. Acetic acid, by contrast, drops the pH to around the proteins’ isoelectric point of roughly 4.5, causing disordered electrostatic collapse into a porous, weak gel that sheds soluble protein and retains little water.

Pretreatment acted as a powerful secondary modifier, and its fingerprint appeared most clearly in texture. For every coagulant, ultrasonicated soymilk produced the firmest tofu: hardness reached 15.0 newtons with calcium chloride, 11.5 newtons with magnesium chloride, and 6.8 newtons with acetic acid, each surpassing the pasteurized and untreated versions of the same combination. The consistent ordering—ultrasonicated greater than pasteurized greater than untreated—tracks directly with the molecular picture of cavitation-driven protein unfolding and enhanced cross-linking. Interestingly, pasteurization boosted yield without hardening the curd, because heat denaturation exposes hydrophilic sites that hold whey during pressing, whereas ultrasound built mechanical strength instead.

All nine formulations met international microbiological standards. Aerobic mesophilic counts ranged from 2.10 to 2.98 log CFU/g, yeasts and molds stayed below 2 log CFU/g, and coliforms and Staphylococcus aureus were each under 10 CFU/g in every sample. The ultrasonicated batches carried the lowest mesophilic loads, and the acetic-acid varieties showed a slight extra edge attributable to their low pH. Given that commercial tofu has been reported with mesophilic counts above 5 log CFU/g, the laboratory process demonstrated a comfortable safety margin, though the authors caution that industrial-scale validation would still be required.

The sensory panel delivered perhaps the most commercially telling verdict. Thirty untrained tasters rated each tofu on a 9-point hedonic scale, and a multiple linear regression distilled their judgments into a Single Sensory Quality Index. Acidity dominated the model, accounting for 49.38 percent of the weight in predicting acceptability, followed by consistency at 25.68 percent and flavor at 24.94 percent. Tofu made from untreated soymilk coagulated with either salt or acid fell into the rejection zone, scoring around 3.2 to 3.3, while acid-coagulated samples generally languished below 4.3. In stark contrast, ultrasonicated soymilk coagulated with divalent salts earned scores of about 6.5—”moderately liked”—the highest acceptance in the study. The panel also found ultrasound improved perceived consistency dramatically, lifting MgCl2-coagulated tofu from a disliked 2.35 to a near-neutral 4.71, and even improved the taste of acid-coagulated tofu, likely because cavitation helps volatilize the compounds responsible for beany and vinegary notes.

To untangle these effects rigorously, the researchers deployed chemometrics rather than relying on one-variable-at-a-time comparisons. Principal component analysis separated the nine samples into three clusters along the first two axes—explaining 63.6 percent of variance—grouped almost entirely by coagulant, with the pretreatments acting as modulators within each cluster. Calcium chloride samples loaded strongly with yield, protein, ash, and hardness; magnesium chloride samples sat in a low-moisture, fat-rich region; acetic acid samples shifted sharply toward poor structural quality. A partial least squares regression with two latent components, achieving a cross-validated predictive R² of 91 percent, quantified the drivers: tofu hardness was most influenced by ultrasound pretreatment and near-neutral pH, while yield depended chiefly on moisture retention, protein and ash content, pasteurization, and—most strongly—hardness itself, since a robust gel network limits syneresis and curd breakage during pressing.

The takeaway for producers is a clear recipe grounded in mechanism rather than tradition: combine thermosonication with divalent salt coagulation, ideally calcium chloride, to simultaneously maximize firmness, protein retention, yield, microbial safety, and consumer acceptance, while avoiding the acidic, low-yield products that acid coagulation tends to produce. The study also demonstrates how multivariate analysis can convert a multi-factorial food process from folklore into engineering. The authors note that their findings should be re-evaluated at industrial scale, but the framework they offer—pairing an emerging physical technology with classical coagulation chemistry and modern statistical modeling—points toward tofu that finally matches its nutritional reputation with the texture and flavor consumers actually want.

Subject of Research: Effects of thermosonication pretreatment and coagulant type on the quality, yield, and sensory properties of tofu

Article Title: Multivariate analysis of the interactive effects of thermosonication and coagulant type on tofu quality characteristics and yield

Article References: Multivariate analysis of the interactive effects of thermosonication and coagulant type on tofu quality characteristics and yield. (n.d.). https://doi.org/10.3389/frsfm.2026.1957457

Image Credits: AI Generated

DOI: 10.3389/frsfm.2026.1957457

Keywords: tofu, thermosonication, ultrasound, soymilk, coagulant, calcium chloride, magnesium chloride, acetic acid, principal component analysis, partial least squares regression, food texture, sensory evaluation