Struvite recovery fails in predictable ways: scale in the pipework upstream of the reactor, fines that will not settle, and crystals too small to dry. All three are the same fault — operating at the wrong supersaturation. This guide sets the chemistry that fixes the operating window.
Solubility product and supersaturation
Magnesium ammonium phosphate hexahydrate dissolves according to:
MgNH4PO4·6H2O ⇌ Mg2+ + NH4+ + PO43− + 6H2O
The conditional solubility product varies with ionic strength and temperature. At 25 °C and I ≈ 0.1 M, which is representative of digester supernatant, pKsp′ is about 13.26 ± 0.15.
Supersaturation is then the log ratio of the ion activity product to that solubility product:
SI = log10 ( {Mg2+}{NH4+}{PO43−} / Ksp′ )SI > 0 means precipitation is thermodynamically favoured; braces denote activities, not concentrations
Why the pH window is narrow
Only the fully deprotonated orthophosphate ion participates in the solubility product, and its fraction is strongly pH dependent:
α(PO43−) = Ka1Ka2Ka3 / ( [H+]3 + Ka1[H+]2 + Ka1Ka2[H+] + Ka1Ka2Ka3 )At 25 °C: pKa1 = 2.15, pKa2 = 7.20, pKa3 = 12.35
That fraction is negligible below pH 8 and only approaches unity above pH 12. Struvite is therefore squeezed between three competing constraints:
- Below about pH 7.5 there is too little PO43− for SI to exceed zero at realistic magnesium doses.
- Above about pH 9.5 ammonium (pKa 9.25) converts to free ammonia and leaves the reaction, while magnesium starts to precipitate as the hydroxide and as Mg3(PO4)2.
- The working window is pH 8.0–8.5 — high enough for workable kinetics, low enough to keep the ammonium and the magnesium in play.
Supersaturation decides crystal quality
Where you sit in the supersaturation range determines which nucleation mechanism dominates, and that determines whether you get a product or a nuisance.
| SI range | State | Dominant nucleation | Product |
|---|---|---|---|
| < 0 | Undersaturated | None | No precipitation |
| 0 – 1 | Metastable | Heterogeneous only | Large, well-formed crystals |
| 1 – 2 | Metastable | Heterogeneous plus some homogeneous | Medium crystals with fines |
| > 2 | Labile | Homogeneous dominates | Amorphous solids, poor settling |
This is why dosing point and mixing matter more than dose. A correctly sized magnesium dose delivered into a poorly mixed line produces a local SI far above the vessel average, and the crystal forms on the pipe wall instead of in the bed.
Magnesium dosing and alkalinity
Precipitation is one-to-one in molar terms, but practical dosing runs 10–30 per cent over stoichiometric to cover magnesium complexation with carbonate, adsorption to organic matter, and the kinetic penalty of operating in the metastable zone.
- Molar phosphorus: 120 / 30.97 = 3.87 mmol/L.
- Magnesium required: 1.2 × 3.87 = 4.65 mmol/L.
- As MgCl2·6H2O (203.3 g/mol): 4.65 × 203.3 = 945 mg/L, i.e. 0.945 kg/m³.
Prefer chloride or hydroxide to sulphate. Magnesium sulphate recycles sulphate to the digester, where it feeds sulphate reduction and puts hydrogen sulphide into the biogas — see odour control for hydrogen sulphide.
Alkalinity is rarely the constraint
Each mole of struvite consumes roughly a mole of protons, through ammonium deprotonation and the conversion of HPO42− to PO43−. For 100 mg/L of phosphorus removed, alkalinity consumption is of the order of 160 mg/L as CaCO3 — small against the 2,000–3,000 mg/L typical of digester supernatant. Caustic is normally needed to raise pH into the window, not to hold it there.
Reactor geometry
Fluidised bed
A fluidised bed grades crystals by size: large product settles to the bottom for withdrawal, fines are carried upward and retained. Bed expansion follows Richardson–Zaki:
ε = ( v / vt )1/nε = bed voidage; v = superficial velocity; vt = terminal settling velocity; n ≈ 2.39 for struvite at sphericity 0.85
Crystal residence time follows from the linear growth rate:
Stirred tank
A CSTR holds a uniform supersaturation throughout, which is precisely the wrong condition: every element of fluid sits at the same SI, so nucleation competes with growth everywhere. The mixed-suspension mixed-product-removal model gives the size distribution:
n(L) = n0 exp( −L / Gτ )n0 = nuclei population density; L = crystal size; G = growth rate; τ = residence time
In practice that yields a median size below 0.3 mm — too fine to separate cleanly without a hydrocyclone and seed recycle.
| Parameter | Fluidised bed | Stirred tank | Air agitated |
|---|---|---|---|
| Crystal size | 1–3 mm | 0.1–0.3 mm | 0.5–1 mm |
| Settling velocity | 50–150 m/h | <10 m/h | 20–50 m/h |
| Phosphorus recovery | 80–90% | 60–75% | 70–85% |
| Energy | 0.05–0.15 kWh/m³ | 0.2–0.5 kWh/m³ | 0.1–0.3 kWh/m³ |
| Scaling risk | Low | High | Medium |
For UK works the fluidised bed is the default on crystal quality and maintenance grounds. A stirred tank with hydrocyclone classification and seed recycle is a reasonable fallback where headroom is limited. Downstream handling is covered in sludge thickening design and dewatering equipment.
Design rules
- Target SI 0.8–1.0 at the inlet and let it decay along the reactor. Control the dosing point and mixing, not just the dose.
- Hold pH 8.0–8.5. Higher and you lose ammonium to stripping and magnesium to hydroxide; lower and the kinetics stall.
- Dose 1.2 times stoichiometric magnesium as chloride or hydroxide. Avoid sulphate.
- Allow 8–12 hours crystal residence in a fluidised bed for a product above 1 mm.
- Track conductivity as an ionic-strength proxy and correct Ksp′ seasonally. Supernatant conductivity swings between roughly 10 and 25 mS/cm with co-digestion feedstock, and that moves the whole SI scale under you.
Frequently asked questions
Why is struvite scaling my pipework rather than forming in the reactor?
Because local supersaturation at the dosing point is far above the vessel average. If magnesium or caustic enters a poorly mixed line, SI locally exceeds 2 and nucleates homogeneously on the nearest surface. Move the dosing point into a well-mixed zone, or use a static mixer, before increasing anything else.
Can I calculate SI from concentrations instead of activities?
Not reliably at digester-liquor ionic strengths. The activity coefficient for phosphate can be below 0.1, so a concentration-based figure can overstate SI by a log unit — enough to put you in the labile zone while the calculation says metastable.
Is a stirred tank ever the right choice?
Where headroom rules out a fluidised bed. It needs hydrocyclone classification and seed recycle to compensate for the uniform supersaturation, and it will still produce a finer, lower-value product with a higher scaling risk.
Does struvite recovery consume significant alkalinity?
No. Roughly 160 mg/L as CaCO3 per 100 mg/L of phosphorus removed, against typical supernatant alkalinity of 2,000–3,000 mg/L. Caustic demand is dominated by raising pH into the operating window, not by the precipitation itself.