Low-Temperature Continuous Flow Synthesis of Metal Ammonium Phosphates

The synthesis of the high performance inorganic materials essential to the quality of modern day life is hindered by traditionalist attitudes and reliance on outdated methods such as batch syntheses. While continuous flow methods have been extensively adopted in pharmaceutical circles, they remain largely unexplored for the preparation of inorganic compounds, despite higher efficiency, safety and versatility. In this publication, we demonstrate a step-change for the synthesis of metal ammonium phosphates through conversion of the extant batch process to a low-temperature continuous regime, exhibiting a tenfold increase in throughput combined with a significant decrease in particle size.

Although inorganic materials are essential to life in the 21 st century, advancements in their preparation are underutilised as traditionalist attitudes in academia and particularly industry mean batch processes predominate. Continuous flow synthesis (CFS) is one such alternative technique: it is prevalent in pharmaceutical research 1,2 , yet has received lesser attention for inorganic synthesis outside of nanomaterials research 3 . This is largely due to the typically high capital costs associated with flow systems 4 ; despite the technique's inherent flexibility 1,2,4 , safety 1,2 , scalability 1,3,5,6 , and controllability [1][2][3]7,8 . Low temperature (<100 °C), ambient pressure flow syntheses presents an affordable preparative route to a plethora of inorganic materials [8][9][10][11][12][13][14][15] , without the high capital cost of proprietary systems 4 . In this communication, we demonstrate a step-change, low-cost continuous preparation of versatile metal ammonium phosphates and derivatives. We compare CFS with the commonly used batch method 16 and demonstrate a tenfold increase in throughput combined with a significant reduction in particle size for continuously synthesised particles.
Based upon a novel low temperature (<100 °C) ambient pressure continuous flow synthesis of the biomaterials brushite and hydroxyapatite [8][9][10][11] , we adapted the synthesis of MAPs to a continuous mode. Bassett noted that an additional source of NH 4 + or performing the precipitation at an elevated pH, can serve to increase the rate of crystallisation of some MAPs 16 . Through a combination of these factors and the rapid mixing of a continuous flow regime, we are able to reduce the reaction time for preparation of MAPs from a typical 180 minutes down to 7 minutes. An excess of phosphate relative to the metal is required to prevent formation of the undesired M 3 (PO 4 ) 2 phase 16,34 .
We prepared six MAPs (M = Mg, Mn, Fe, Co, Ni and Zn) and one related hydrogen phosphate (M = Sn) using this technique in a variety of hydration states. Our CFS process provides two feeds: one containing the metal (M 2+ , 0.05 M) typically as the nitrate salt, and the other a mixture of triammonium phosphate (TAP, 0.25 M) and ammonium nitrate (AN, 0.5 M) [8][9][10][11] . These feeds mix continuously before heating the flow, which serves to crystallise the initially amorphous product 8 . For some metals (notably Fe and Co), this crystallisation results in a clear colour change. The characterisation data of our MAPs are presented in Table 1, with comparative particle sizes, yields, and throughputs for material produced using the common batch process 16 . Reaction yields are expressed as a percentage of the theoretical maximum. Throughputs are expressed in grams per litre of liquor per hour, accounting for the reaction time.
MgAP forms a mixed-phase mono-and hexa-hydrate in CFS, and a monohydrate in batch ( MgAP can interchange between mono-and hexahydrated phases through hydrolysis and dehydration respectively 35,38 , though our continuous flow process forms a mix of the two phases, though a longer or hotter reaction time fully dehydrates the product. NH 3 was added to the Mg 2+ feed to prevent formation of MgHPO 4 8 . Mn, Fe and CoAP all favour the monohydrated phase and were prepared in this form in both batch and flow 16,19,36,37 . NiAP can, like MgAP, interconvert between the two phases, though the monohydrate phase is metastable and the synthesis is sensitive to the reaction conditions used 34 . The NiAP batch sample was of the monohydrate phase as the crystals have sufficient time to dehydrate. No stable ZnAP hydrate has been prepared as the system favours the anhydrous form 16 . We were unable to synthesise SnAP as hydrolysis of the acidic Sn 2+ ion takes precedence over the formation of the desired phase. SnHP is prepared by exchanging the triammonium phosphate for phosphoric acid and omitting the ammonium nitrate. The phases of the batch produced materials match those of the flow produced materials except for NiAP 34 , which is synthesised as the monohydrate in batch 16 . Neither CaAP or CuAP can be synthesised via our continuous flow route, as hydroxyapatite 39 (for Ca) or a hydroxyphosphate 40 (for Cu) form preferentially instead. Bassett noted the difficulty in synthesising CuAP relative to other divalent metals 16 .
The yields of all our flow process were excellent (>94%) except for NiAP (83%) and SnHP (79%) due to the marginal solubilities of the metals in aqueous ammonia and acid respectively [41][42][43] . The yields can be increased through process optimisation. Yields for the batch produced material were quantitative, except for SnHP, where the yield matched the flow produced example. All of our flow produced materials, except MnAP, were at least half the size than the batch produced materials with narrow size distributions. Flow-produced FeAP is less than one fifth the size of equivalent batch material. The reduction in particle size is due to the increased rates of mixing and turbulence within the flow process and our adapted reaction conditions which promote a significant reduction in crystallisation time. For MnAP, the larger sized particles produced in flow are due to increased Ostwald ripening during the crystallisation process 44 . The observed surface areas of our flow-produced compounds vary with the nature of the crystalline product. FeAP shows a BET high surface area as the plate-like crystals are significantly  Table 1. Characterisation data for MAPs: reaction yield (%), particle size for flow (CFS) and batch (Bx) produced material as 10 th , 50 th and 90 th percentiles (D 10 , D 50 and D 90 , µm), BET surface are (m 2 /g) and observed XRD crystal hydration phase. MgAP (*) required the addition NH 3 to the Mg 2+ feed in flow to precipitate the desired product 8 .  (Fig. 1a) than those of the similarly plate-like MnAP and CoAP. Throughputs for the CFS process are at least five times that of the batch process, and up to sixteen times for ZnAP due to the lengthy reaction times required in batch 16 . The efficiency of the CFS process can be further improved by the feasible regeneration of the effluent liquor. Concentration, replacement of precipitated phosphate and addition of the required NH 3 allows for use in further syntheses. Recovery of process heat by pre-heating precursor streams would further increase efficiency. Changing reagent stoichiometry, overall concentrations, and process parameters such as temperature, flow rate and mixing regime would allow for control of particle sizes produced within the CFR 45 . Use of specialist mixers such as confined impinging jet mixers may also allow for preparation of MAPs on the nanoscale due to more turbulent mixing during the precipitation step 45 . The flows within our demonstrated system are all laminar.

Conclusions
We have developed an efficient continuous flow synthesis of metal ammonium phosphates and compared this to traditional batch methods. We have demonstrated a step change in the preparation of these compounds as synthesised particles are smaller, more evenly sized and produced with far greater efficiency than previous syntheses. In comparison to the traditional batch methods, we observe up to a tenfold increase in throughput (in terms of mass per unit volume per unit time) with our flow process. This value can be increased further with process optimisation. For many applications, smaller particles are preferably as higher surface areas provide higher activities. Continuous flow synthesis provides this benefit with the option of greater control over particle size and morphology. Through the application of appropriate engineering and chemistry, the syntheses of other key inorganic materials could be improved in a similar manner as demonstrated here.
Our continuous flow reactor (CFR) was constructed based on previous literature references [8][9][10][11] . Simply, this consists of two peristaltic pumps (Watson Marlow SciQ 323), a simple Y-mixer and a coil of PVC tubing (16 m length, 4 mm internal diameter, wall thickness 1 mm, total volume 210 cm 3 ) contained within a thermostatically-controlled water bath (Grant JB Aqua 18 Plus). This is illustrated in Fig. 2.
The rotational rate and thus flow rate of each pump can be varied from 5-40 ml/min, corresponding to a total flow rate of 10 to 80 ml/min and thus a residence time of between 3 and 20 minutes, though a total flow rate of 30 ml/min and thus a residence time of 7 minutes was used in this work. Both pumps were run at the same flow rate in all cases, with one providing the metal salt feed, and the other the phosphorus source and any ancillary reagents. The temperature of the water bath could be varied between ambient and 100 °C. All reactions presented here were performed at 80 °C. A narrow aperture (c.a. 1 mm) at the outlet of the CFR served to even the flow within the system, and check valves after each pump ensured a continuous unidirectional flow. Products were purified by repeated washing with deionised water, and were then dried overnight at 80 °C. The batch syntheses of MAPs were adapted from the work of Bassett and Bedwell 16 , using the same procedure but scaled to give the same volume of product as in our flow process.
The rate of heating within the process is rapid, as a total flow rate of greater than 120 ml/min emerges from the reactor at 80 °C with the same tubing setup utilised for these syntheses, suggesting that temperature equilibrium is attained within 2 minutes and thus 5 minutes are spent at temperature to allow completion of reaction. Further modelling of this factor is beyond the scope of the research presented here.
Powder X-Ray Diffraction (PXRD) analysis was conducted using a Brucker D2 Diffractometer with a copper kα radiation source with data collected between 5 and 80 degrees 2Θ. XRD patterns were converted from proprietary formats using PowDLL 46 . Scanning Electron Microscopy (SEM) and X-Ray Energy Dispersive elemental analysis was conducted under high vacuum using a FEI Quanta 200 scanning electron microscope equipped with an EDAX Sapphire Energy Dispersive X-Ray detector. Surface areas were determined using the Brunauer-Emmett-Teller (BET) model at 77 K using nitrogen absorption (Micrometrics ASAP2010), with accuracy checked against an alumina standard. Particle size distribution analysis (PSD) was conducted using a Malvern Mastersizer 2000. Samples were sonicated in deionised water before analysis to disperse any agglomerates. Results are expressed as D 50 , the 50 th percentile of particle size.

Data Availability Statement
The data underpinning this publication can be found on https://clok.uclan.ac.uk/.