The present invention relates to a method for upscalable precipitation synthesis for the production of materials with tunable particle size distribution (PSD) and mainly consisting of spherical particles as precursor for the synthesis of battery cathode materials, more specifically by the precipitation of a metal carbonate material comprising nickel and manganese in an atomic ratio of 0 ≤ Ni:Mn ≤ 1/3, with tunable PSD.

Limited oil reserves together with an increasing effort to reduce the amount of CO ₂ exhausted not only cause a shift in electric energy production towards renewable forms of energy, but also result in increasing research for alternative engines in the automotive industry. Moreover, an increasing demand for durable high power storage media within consumer electronics and telecommunication promotes both improvement of existing energy storage systems and develop- ment of new energy storage systems with the main focus on efficiency, cost reduction and safety. Due to its high gravimetric energy storage capability, the lithium ion bat ^¬ tery (LiB) is a promising system, not only for today and future energy storage, but also for use in automobiles and in a large variety of other fields of application.

The first commercialized LiB was based on L1C0O ₂ as posi ^¬ tive electrode and graphite as negative electrode, deliver ^¬ ing a potential of 3.8 V vs. Li/Li ⁺ . However, since cobalt is both relatively expensive and toxic, the search for al ^¬ ternative cathode materials is ongoing, and an increasing number of other material combinations are being tested and partially commercialized. One possibility to increase the energy density of the LiB is to raise the operating voltage of the cells. This could be achieved by using cathode mate ^¬ rials with higher electrode potential like the LiNio.5Mn1.5O4 spinel, delivering a voltage of 4.7 V vs. Li/Li ⁺ .

A spinel means a crystal lattice where the oxide anions are arranged in a cubic close-packed lattice and the cations are occupying some or all of the octahedral and tetrahedral sites in this lattice. The symmetry of the spinel lattice is described by space groups of P4332 for the ordered phase and Fd-3m for the disordered phase. The spinel material may be a single disordered or ordered phase, or a mix of both (Adv. Mater. 24 (2012), pp 2109-2116).

Another important factor for the choice of materials in the LiB is the abundance of their components in the earth crust securing long term availability and cost reduction, due to which materials based on iron and manganese are of great interest. Especially manganese oxides constitute a promis ^¬ ing group of cathode materials, because manganese is a low priced and non-toxic element. In addition, manganese oxides have a rather high electronic conductivity together with a suitable electrode potential. Among the lithium manganese oxides, the layered LiMn0 ₂ and the spinel-type LiMn ₂ 0 ₄ (LMO) are the most prominent ternary phases. Advantages of the latter one in comparison to the layered phase are a higher potential of about 4.0 V against Li/Li ⁺ , whereas LiMn0 ₂ de ^¬ livers only 3.0 V on average. The LiMn ₂ 0 ₄ lattice offers a three-dimensional lithium diffusion, resulting in a faster uptake and release of this ion. The diffusion of Li ⁺ in doped LMO spinels is also equally fast in all three dimen ^¬ sions .

Among the transition metal doped LiM^C^ spinel electrodes, Li i _y Mn ₂ -y0 ₄ , where 0 ≤ y ≤ 0.5, is a very promising mate ^¬ rial: It has an additional capacity at a relatively high voltage of 4.7 V vs. Li/Li ⁺ due to the electrochemical ac ^¬ tivity of the Ni ²⁺ /Ni ⁴⁺ redox couple. LiNio. ₅ Mn ₁ . ₅ O ₄ has a theoretical specific capacity of 147 mAh/g and therefore an attractive theoretical energy density of 4.7 V-147 Ah/kg = 691 Wh/kg active material, referring to lithium metal. By replacing 25% of the manganese ions with nickel, there is in theory no Mn ³⁺ left in the structure. For charge balance reasons, the Ni ²⁺ incorporation forces all manganese ions to their tetravalent state. If the less than 25% of the manganese is replaced by Ni, some Mn ³⁺ will be left in the structure. For Li i _y Mn ₂ - _y 0 ₄ the theoretical Mn ³⁺ :Mn ⁴⁺ ratio is equal to (0.5 - y) . It should be noticed that in practice there will often be small deviations from the theoretical composition and average oxidation states when synthesizing a material. This can occur because of deviation from the exact stoichiometry due to either the existence of defects and inhomogeneity in the structure of Li i _y Mn ₂ - _y 0 ₄ or the ex ^¬ istence of impurity phases which alter the composition of the main phase. It is for instance well known that small amounts of a rock salt phase is present when synthesizing LiNio. ₅ Mn ₁ . ₅ O ₄ , which affects the stoichiometry of the spinel phase and renders the total material oxygen deficient (J. Cabana et al . , Chemistry of Materials 24 (2012), 2952).

LNMO materials are lithium positive electrode active mate ^¬ rials dominated by a Ni-doped LiMn ₂ 0 ₄ spinel phase, which more specifically may be characterised by the general for ^¬ mula Li _x i _y Mn ₂ - _y 0 ₄ with typical x and y values of 0.9 ≤ x ≤ 1.1 and 0 ≤ y ≤ 0.5, respectively. The formula represents the composition of the spinel phase of the material. Such materials may be used for e.g. portable electric equipment (US 8,404,381 B2), electric vehicles, energy storage sys ^¬ tems, auxiliary power units (APU) and uninterruptible power supplies (UPS) . Lithium positive electrode active materials based on LNMO are seen as prospective successors to current lithium secondary battery cathode materials such as LiCo0 ₂ due to their high voltage and high energy density, coupled with lower materials costs.

Electrode active materials for lithium ion battery materi- als are described abundantly in the literature. Thus, US

5.631.104 describes insertion compounds having the formula Li _x+ iM _z Mn ₂ - _y - _z 0 ₄ wherein the crystal structure is spinel-like, that can reversibly insert significant amounts of Li at po ^¬ tentials greater than 4.5 V vs. Li/Li+. M is a transition metal, in particular Ni and Cr, 0 ≤x ≤1, 0 ≤y< 0.33, and 0 < z <1.

US 2015/0041710 Al relates to cathode active materials for lithium secondary batteries. More particularly it relates to a method for preparing 3V class spinel oxides with the composition Li _1+X [M _y Mn ₍₂ - _y) ] 0 ₄ - _z S _z (0 < x < 0.1; 0.01 < y < 0.5 and 0.01 ≤ z ≤ 0.5; M = Mn, Ni or Mg) by carbonate co-pre ^¬ cipitation. It describes that nickel and manganese sulfates are used in the preparation of Li ₁ . ₀₆ Nio. ₅ Mn ₁ .5O4.

In addition to the elemental composition and crystal struc ^¬ ture of the electrode materials, it is well known to those skilled in the art that control of the shape and micro- structure of the particles is vital, if batteries with good cycling behaviour and high performance are desired. For ex ^¬ ample, Zhu et al. (Solid State Ionics, 179 (2008) 1788) showed that powder morphology had a significant impact on the electrochemical performance of LMO spinel when tested in half cells where Li metal was used as the anode. In par ^¬ ticular, spherical LMO particles were found to have a sig ^¬ nificantly higher capacity retention after 100 charge-dis- charge cycles than irregularly shaped particles. According to the authors, increased performance was due to reduced stress on particles during cycling, because the local Jahn- Teller distortions on one side of the spherical particle were counteracted by identical distortions on the opposite side of the particle.

One way to quantify the size of particles in a slurry or a powder is to measure the size of a large number of parti ^¬ cles and calculate the characteristic particle size as a weighted mean of all measurements. Another way to charac ^¬ terize the size of particles is to plot the entire particle size distribution, i.e. the volume fraction of particles with a certain size as a function of the particle size. In such a distribution, D10 is defined as the particle size where 10% of the population lies below the value of D10,

D50 is defined as the particle size where 50% of the popu ^¬ lation lies below the value of D50 (i.e. the median), and D90 is defined as the particle size where 90% of the popu ^¬ lation lies below the value of D90. Commonly used methods for determining particle size distributions include laser diffraction measurements and scanning electron microscopy measurements, coupled with image analysis. Similarly, there are several ways to characterize and quan ^¬ tify the sphericity and shape of particles. Almeida-Prieto et al . in J. Pharmaceutical Sci . , 93 (2004) 621, lists a number of form factors that have been proposed in the lit- erature for the evaluation of sphericity: Heywood factors, aspect ratio, roughness, pellips, rectang, modelx, elonga ^¬ tion, circularity, roundness, and the Vp and Vr factors proposed in the paper. Circularity of a particle is defined as 4 ·π · (Area) /( Perimeter) ² . An ideal spherical particle will thus have a circularity of 1, while particles with other shapes will have circularity values between 0 and 1. Particle shape can further be characterized using aspect ratio, defined as the ratio of particle length to particle breadth, where length is the maximum distance between two points on the perimeter and breadth is the maximum distance between two perimeter points linked by a line perpendicular to length.

For most battery applications, space is at a premium, and high energy density is desired. Powders of the electrode material with a high tap density tend to result in elec ^¬ trodes with higher active material loading (and thus higher energy density) than powders with a low tap density. It can be shown using geometry-based arguments that materials com- posed of spherical particles have a higher theoretical tap density than particles with irregular shapes.

In addition to the shape of the spinel particles, the sur ^¬ face microstructure and roughness are thought to play a key role in determining the cycling behaviour and rate-capability of high-voltage electrode materials, such as LNMO, in batteries. The redox potential of the Ni ²⁺ /Ni ⁴⁺ redox couple in LNMO lies at approximately 4.75 V vs. Li/Li ⁺ , which is higher than the stability limit of common liquid carbonate- based electrolytes (typically between 4.0 and 4.3 V). This leads to the oxidation of the electrolyte on the positive electrode during battery operation, and to the formation of a resistive cathode/electrolyte interface layer on the electrode. The rougher the surface of the active powder in the electrode (i.e. the higher its specific surface area), the more of the electrolyte is lost due to oxidation and the formation of a resistive cathode/electrolyte layer at the interface between LNMO and the electrolyte. The rougher the surface of the active powder in the electrode (i.e. the higher its specific surface area) is, the more of the elec ^¬ trolyte is lost due to oxidation and the formation of the cathode/electrolyte interface layer. Therefore, positive electrode active materials with low specific surface area (i.e. smooth and non-porous surface morphology) are re ^¬ quired to avoid capacity fading during long-term battery operation .

Lithium positive electrode active materials may be prepared from precursors obtained by mechanically mixing starting materials to form a homogenous mixture, followed by calci ^¬ nation. For example, in US 8,404,381 B2, an intimate mix- ture of nickel carbonate, manganese carbonate and lithium carbonate is prepared by grinding in the presence of hex- ane . The mixture is dried and treated at 600°C (10 hours), and then at 900°C (15 hours) to form LNMO. Critically, LNMO materials prepared according to this method will not be spherical, which will have a significant negative effect on the cycling behaviour of the battery using such LNMO as the positive electrode, as well as on the tap density of the LNMO powder.

As used herein, a "precursor" is a composition prepared by mixing starting materials to obtain a homogeneous mixture

(see, for example, Journal of Power Sources, 238 (2013) pp. 245-250) or by mixing a lithium source with a composition prepared by precipitation of starting materials (see, for example, Electrochim . Acta, 115 (2014) pp. 290-296). Another way to prepare positive electrode active materials for LiB is to use the method of precipitation, followed by calcination/lithiation . In precipitation, two or more solutions are mixed under controlled conditions and react to form an insoluble product, which can then be isolated from the rest of the mixture. When precipitation is used for battery material synthesis, the process typically starts with the precipitation of the transition metal species in the desired elemental ratio. Examples of such transition metal species include but are not limited to Ni(OH) ₂ ,

NiC0 ₃ ,Ni ₂ C0 ₃ (OH) 2, Ni(HC0 ₃ ) ₂ , Ni ₃ C0 ₃ (OH) ₄ , Mn(OH) ₂ , MnC0 ₃ ,

Nio.25Mno.75 (OH) 2, Ni ₀ . ₂₅ Mn ₀ . ₇₅ C0 ₃ , Nio. ₅ Mn ₀ .5 (OH) ₂ , Ni ₀ . ₅ Mn ₀ . ₅ C0 ₃ , Nii _/3 Mni _/3 Coi _/3 (OH) 2 and Nii _/3 Mni _/3 Coi _/3 C0 ₃ . Such transition metal hydroxides, carbonates, or hydroxycarbonates are of ^¬ ten referred to as precursor materials. After precursor ma- terials are obtained, they are typically mixed with a Li- containing material and calcined to elevated temperatures to ensure the formation of the correct crystallographic phase. The Li-containing materials include, but are not limited to lithium hydroxide, lithium carbonate and lithium nitrate. It is generally accepted that the shape and size of the powder particles of the final active material are to a great extent determined by the morphology and size of the precursor particles. Importantly, when reaction parameters are properly controlled, spherical precipitate particles can be obtained. Precursor materials for LNMO-type LiB positive electrode materials can be prepared by precipitating the mixture of Ni and Mn as hydroxides, i.e. Ni _y Mn ₂ - _y (OH) ₄ , where 0 ≤ y ≤ 0.5. Such precipitation commonly uses the corresponding transition metal sulfates and NaOH as starting materials and ammonia (N¾, NH ₄ OH or N¾ ·¾0) as chelating agent.

Other commonly used chelating agents include but are not limited to NH ₄ CO ₃ , citric acid, glycolic acid, oxalic acid, polyacrylic acid, malonic acid, and EDTA. Another way of preparing precursor materials for LNMO-type LiB positive electrode materials is to precipitate the mix ^¬ ture of Ni and Mn as carbonates, i.e. Ni _y Mn ₂ - _y (CO ₃ ) 2 , where 0 ≤ y ≤ 0.5. Such precipitation commonly uses the correspond ^¬ ing transition metal sulfates and Na ₂ C03 as starting mate- rials and ammonia (N¾, NH ₄ OH or N¾ ·¾0) as chelating agent. For example, in US 2014/0341797 Al, NiS0 ₄ , MnS0 ₄ , Na ₂ C03, and NH ₄ OH are fed into a special reactor system at a molar flow rate of 3.25 mol/hr, 6.5 mol/hr, 10.8 mol/hr and 1.1 mol/hr. Spherical nickel manganese carbonate particles could be obtained, but only when the continuously stirred tank reactor was combined with centrifugal dispensers and particle size separators. The first disadvantage with the above process is that a complex reactor design is required to obtain spherical carbonate particles. The second disad- vantage with the above process is that a chelating agent, in this case ammonia, is used. The intention of the present invention is to use a method for upscalable precipitation synthesis to prepare a high quality LNMO precursor product with a tunable particle size distribution (PSD) .

Continuous precipitation in one stirred tank reactor is generally used as a synthesis method for large scale pro ^¬ duction of spherical precursors for battery materials. In such a continuous precipitation set-up, there is a continu- ous feed of raw materials and a continuous extraction of product from the stirred tank reactor. There is a steady state between the formation of new seeds, the growth of particles and removal of a fraction of all particles. The synthesis could follow the reaction scheme:

NiS0 ₄ (aq) + 1.5 MnS0 ₄ (aq) + 2 Na ₂ C0 ₃ (aq) 2 Na ₂ S0 ₄ (aq) + Ni ₀ . ₅ Mni. ₅ (C0 ₃ ) 2 (s)

If the process is carefully designed, it will result in a product dominated by spherical particles, but with a broad PSD. The PSD is broad because particles in all development stages (seeds, aggregated seeds, agglomerates that have grown a little, and agglomerates that have grown a lot) are present in the product (see Fig.l) .

Batch precipitation in a stirred tank reactor is also well known in the art. Here it is possible to make spherical precursor particles with a very narrow PSD, since seed formation can be limited to the start of the precipitation process, if the process is carefully designed. A disad ^¬ vantage of such a process is that the same reactor is used both in the early stage and the late state of the process, where very different reactor designs will be advantageous.

The present invention relates to a method for upscalable precipitation synthesis giving a high quality product with a tunable PSD (both in terms of D50 and D90/D10 of the dis ^¬ tribution) . In this synthesis method, the seed formation has been separated in time and space from the final growth and polishing of the particles.

More specifically, the invention relates to a method of up- scalable precipitation synthesis as a production step in the production of battery materials wherein:

(1) seed particles are produced continuously, agglomerated and grown to a certain extent in a first reactor by a continuous feed and mixing of raw materials, i.e. a metal ion solution and a solution comprising carbonate ions, in a fixed ratio,

(2) a specific amount of the product suspension produced in step (1) is transferred batch-wise or continuously to a stirred second reactor, and a further continuous feed of raw materials in a fixed ratio is added to this second re ^¬ actor to grow the particles,

(3) step (2) is optionally repeated one or more times by transferring batch-wise or continuously a specific amount of the product of the previous step to a subsequent stirred tank reactor where a new continuous feed of raw materials in a fixed ratio is added to grow the particles further, and

(4) the final product is collected batch-wise or continu ^¬ ously from the last reactor for further processing, wherein the desired size characteristics of the agglomer ^¬ ated particles in the final product is obtained without particle size separation and/or particle size selection.

The optimal rate of seed formation in step 1 is achieved by tuning process parameters like temperature, concentrations and feed rates of the metal ion and carbonate solutions (pH) , reactor design and optionally stirring speed. The particles from the previous step should grow and only very small amounts of new seeds should be formed in step 2 and any subsequent steps. This is also achieved by tuning pro ^¬ cess parameters like temperature, the acid and base feed rates, reactor design and stirring speed.

The desired size characteristics of the agglomerated parti ^¬ cles in the final product is obtained without particle size separation and/or particle size selection. In particular, no return flow of product suspension, e.g. subsequent to particle size separation, takes place in the method of the invention. Thus, there is no need for a return flow, e.g. of a part of the product suspension obtained in the first, second or third reactor back into to the first, second or third reactor. Hereby, the method of the invention provides a simple way of obtaining a metal carbonate material with specific size characteristics.

A preferable synthesis reaction is carbonate precipitation y NiS0 ₄ (aq) + (2-y) MnS0 ₄ (aq) + (2+x) A ₂ C0 ₃ (aq) ->

2A ₂ S0 ₄ (aq) + x A ₂ C0 ₃ (aq) + Ni _y Mn ₂ - _y (C0 ₃ ) 2 (s) with N1SO4 (aq) and MnSC^ (aq) as metal ion solution,

A ₂ C0 ₃ (aq) as carbonate solution, A = Na or K, 0 < y ≤ 0.5 and x ≥ 0, and optionally x and y can be different in the different reactors. The principle of the invention when using three reaction steps is illustrated in Fig.2, where A and B are the metal ion solution stream and the carbonate solution stream, respectively, and P stands for the final precipitated prod- uct . Note that Reactor 1 does not need to be a stirred re ^¬ actor, and it can differ substantially from the other two reactors in its mechanical design, e.g. be a tee-piece re ^¬ actor, for example similar to the reactor described in US 2013/0136687 Al .

Generally, as illustrated in Fig.l, the steps of precipita ^¬ tion and growth of a i _y Mn2- _y (CO ₃ ) 2 precursor material are the following:

1. nucleation

2. aggregation

3. agglomeration

4. growth and polishing.

One important advantage of the current invention is that in each reactor, the precipitation conditions can be tailored towards different steps of the above process. For example, conducting the precipitation at pH < 8 by adjusting the ratio between the acid and base feed suppresses the agglomer ^¬ ation step, which is preferable in the first precipitation reactor. High pH (surplus of base) is preferable in the later steps because this favors growth and polishing over aggregation of new small seed particles and ensures that most of the metal ions are precipitated. Another advantage is related to the uniformity and parti ^¬ cle-size distribution of the formed particles or agglomer- ates. Continuous precipitation using a single stirred reac ^¬ tor, which has so far been the preferred method in the art, results in a relatively broad particle size distribution. A narrower particle size distribution will minimize the stresses that a particle will experience during the charg ^¬ ing and discharging of the battery, thus increasing the lifetime of the battery. The method of the invention re ^¬ sults in a narrower PSD than previously obtainable in a continuous precipitation. This is illustrated in Fig.l.

Continuous preparation is a production method by which a dynamic equilibrium between formation of seeds and the growth of precipitates is obtained. A continuous feed of chemicals is delivered into the reactor, and the product is continuously removed.

In contrast, the method of the invention comprises a semi- batch reaction. The only seeds in the reaction are added as the starting solution, and these seeds are prepared in a controlled continuous process in which a special reactor is used. If the batch reaction is starting from a number of identical seeds at a low pH (pH < 8 for carbonate precipi ^¬ tation) to limit agglomeration, then ideally the particles will grow equally as more raw material feed is added. The product is finished when the reactor is full. Then the product should be collected, and a new batch reaction can start .

Yet another advantage of the method according to the inven- tion is that it results in production of spherical parti ^¬ cles with a narrower PSD (i.e. a lower ratio of D90/D10) of the product compared to that of the standard continuous precipitation method using one stirred reactor.

Batch precipitation as such is a known technique, and it will lead to a narrow PSD. The advantage characterizing the method of the invention is the reproducibility that has been increased by controlling the seed formation in a separate reactor and stabilizing the growth by adjusting pH to the size of the particles. The smaller particles require a lower pH to limit agglomeration.

The invention also allows obtaining a narrower PSD by a continuous process by mainly limiting the seed formation to the first reactor.

A further advantage of the multi-sequential reactor concept is that it gives the possibility to alter the radial chemi ^¬ cal composition of the agglomerates by changing the compo ^¬ sition of the metal ion solution.

The precipitated materials comprising i _y Mn ₂ - _y (CO ₃ ) 2 , where 0 ≤ y ≤ 0.5 can be different from the exact chemical formula and still be used to produce LNMO material. One difference can occur because not all the Ni and Mn will be in oxida- tion state +2 which might be due to oxidation by oxygen originating from the atmosphere. Another difference might be due to contamination of the raw materials. Yet another difference might be due to intentional doping of the pre ^¬ cursor material by adding other elements to the raw materi- als in the whole or part of the precipitation process that ends up in the precipitated precursor. Possible elements include but are not limited to Mg, Co, Cr, Fe, Al, Ti. Such differences are included in the present invention.

The invention is described further in the examples which follow .

Example 1 (comparative example)

A metal ion solution of NiS0 ₄ and MnS0 ₄ with a Ni:Mn atomic ratio of 1:3 was prepared by dissolving 281 g of NiS0 ₄ -7H ₂ 0 and 507 g of MnS0 ₄ ·¾0 in 1757 g water. In a separate flask, a carbonate solution was prepared by dissolving 424 g of Na ₂ CC>3 in 1939 g water. The metal ion solution and the carbonate solution were added separately into a reactor provided with vigorous stirring (450 rpm) and a temperature of 35°C. The flow rates of the two solutions were chosen such that the molar ratio of the metal ions to the car ^¬ bonate ions was 1:1.1, resulting in a constant pH through ^¬ out the precipitation between 7.5 and 7.6. The volume of the reactor was 1 liter.

The synthesis proceeds according to the following simple reaction between N1SO ₄ (aq) , MnS0 ₄ (aq) and Na ₂ CC>3 (aq) :

0.5 NiS0 ₄ (aq) + 1.5 MnS0 ₄ (aq) + 2.2 Na ₂ C0 ₃ (aq) -> 4.4 Na ⁺ (aq) + 2 S0 ₄ ^" (aq) + 0.2 C0 ₃ ^2" (aq) + Ni ₀ . ₅ Mni. ₅ (C0 ₃ ) 2 (s)

The product was continuously removed from the reactor, so that the residence time of the reactants in the reactor was 60 minutes. In Fig.3, a scanning electron microscope (SEM) image of io.sMni. ₅ (CO ₃ ) ₂ particles, collected after 3 resi ^¬ dence times, is shown. Both aggregated seed particles and larger agglomerates were clearly present in the precipita ^¬ tion product. The corresponding PSD is shown in Fig.4, and the PSD data confirm the trend seen in Fig.3. The particle size distribution was wide and bi-modal, with a D10 of 6.3 rn, D50 of 10 μιτι, and D90 of 14.1 μιη. The ratio of D90/D10 was thus 2.24.

Example 2

First, a metal ion solution of iS0 ₄ and MnS0 ₄ with a Ni:Mn atomic ratio of 1:3 and a carbonate solution of a ₂ C0 ₃ , both essentially identical to these used in Example 1, were pre- pared. Before starting the precipitation process, the car ^¬ bonate solution was diluted lOx using de-ionized water. The precipitation was carried out in a tee-piece reactor, consisting of a Swagelok 1/8" stainless steel union tee-piece with 1/16" stainless steel inlet for the metal ion solution and a 1/8" inlet for the carbonate solution. The metal ion solution stream entered the reactor from the top, the carbonate solution stream from the side, and the product left the reactor from the bottom. The flow rate was chosen such that the residence time of the product in the reactor was approximately 0.2 seconds. The precipitation was carried out at room temperature. The product was collected in a stirred beaker.

The molar ratio of the metal ions to the carbonate ions was varied to investigate the effect of seed stability or ag ^¬ glomeration on pH. Agglomeration was investigated by storing the samples in closed glass bottles without stirring for up to 14 days at room temperature. The PSD evolution for the 5 different samples is summarized in Fig. 5. Parti ^¬ cles precipitated and stored at relatively low pH (at pH ≤ 8.17, Samples D and E) are substantially more stable to- wards agglomeration than particles precipitated at higher pH (Samples A, B, and D) .

Example 3 (batch-wise precipitation) This example describes a semi-batch process with 3 reactors for preparing LNMO precursor particles with narrow PSD. First, seed and aggregate particles were prepared in a tee- piece reactor using a procedure identical to that used for preparing Sample A particles in Example 2. The flow rate was chosen such that the residence time of the product in the reactor was approximately 0.2 seconds. The product was collected in a stirred beaker. To avoid agglomeration, the pH of the product was adjusted to approximately 7.9 with dilute H ₂ S0 ₄ . A 300 milliliter sample of the product from the first reac ^¬ tor was transferred to a second reactor and used as the starting solution for the next step, which is a batch precipitation. The second reactor was mechanically substantially different than the first reactor, being a 1-liter stirred reactor with a heating jacket. The 300 milliliter starting solution from the first reactor was mixed with 200 milliliters of water and heated to 36°C, followed by the introduction of additional continuous streams of the metal ion solution and the carbonate solution, identical to solu- tions used in Example 1. The flow of the solutions was mon ^¬ itored and controlled to keep the pH between 8.1 and 8.6 during precipitation. Throughout the 1-hour batch precipitation, the mixture in the reactor was stirred at 450 rpm. Subsequently, the above process was repeated 3 times to re ^¬ sult in 3 liters of product from the second reactor. These 3 liters were transferred from the second reactor into a third reactor and used as a starting solution for the next step, another batch precipitation. The third reactor was a stirred reactor with a heating jacket with an inner volume of 8 liters. The starting solution was heated to above 35°C, followed by the introduction of additional con ^¬ tinuous streams of metal ion solution and carbonate solu ^¬ tion, which in this example were identical to solutions used in Example 1. The flow of the solutions was monitored and controlled to keep the pH between 8.79 and 9.42 during precipitation. After 2.5 hours of batch precipitation, the product in the reactor was collected. The corresponding PSD is shown in Fig.6. The particle size distribution was sig ^¬ nificantly narrower than in Example 1 and unimodal, with a D10 of 10.00 rn, D50 of 14.15 μπι, and D90 of 19.96 μπι. The ratio of D90/D10 was thus 1.996. The final product was thoroughly washed, dried in a drying oven and stored for further processing ( lithiation/calcination) .

Example 4 (continuous precipitation) This example describes a continuous process with 2 reactors for preparing LNMO precursor particles with narrow PSD. A first metal ion solution of iS0 ₄ and MnS0 ₄ with a Ni:Mn atomic ratio of 1:3 was prepared by dissolving 480 g of NiS0 ₄ -6.3H ₂ 0 and 1051 g of MnS0 ₄ · 1.14H ₂ 0 in 3577 g water. In a separate flask, a first carbonate solution was prepared by dissolving 1106 g of K ₂ CO ₃ in 3762 g water. A second metal ion solution of N1SO ₄ and MnSC^ was prepared by dilut ^¬ ing the first metal ion solution by a factor of two with demineralized water. A second carbonate solution was pre ^¬ pared by diluting the first carbonate solution by a factor of two with demineralized water.

The second metal ion solution and the second carbonate so ^¬ lution were added separately into a reactor provided with vigorous stirring (2000 rpm) at room temperature (approxi- mately 20°C) . The flow rates of the two solutions were cho ^¬ sen such that the molar ratio of the metal ions to the car ^¬ bonate ions was 1:1.05, resulting in a steady state pH of around 7.5. The volume of the reactor was 750 milliliters. The product was continuously removed from the reactor, so that the residence time of the reactants in the reactor was 7.5 minutes. The PSD after 7 hours of precipitation is shown in Fig. 7a, and the evolution of D10, D50, and D90 during precipitation is shown in Fig. 7b. According to Fig. 7b, steady state in the first reactor was obtained after 4 hours of precipitation at the latest, while other experi ^¬ ments (not included for brevity) showed that steady state could be obtained much faster (after around 30 minutes of precipitation) . The particle size distribution of the product from first reactor was bi-modal, with a D10 of 0.6 μιτι, D50 of 3.5 μιτι, and D90 of 7.5 μιη. The ratio of D90/D10 was thus 12.5.

Part of the product from first reactor was continuously pumped over to the second reactor using a peristaltic pump. The second reactor was a stirred reactor with a heating jacket with an inner volume of 8 liters that was held at a temperature above 35°C. Simultaneously with the introduc- tion of the stream of product from first reactor, addi ^¬ tional continuous streams of first metal ion solution and first carbonate solution were separately added to the sec ^¬ ond reactor. The flow rates of the metal ion and carbonate solutions were chosen such that the molar ratio of the metal ions to the carbonate ions was 1:1.16, resulting in a steady state pH of around 9.3. The stirring rate in the second reactor was 500 rpm. The product was continuously removed from the second reactor, so that the residence time of the reactants in the reactor was 43 minutes. The PSD af ^¬ ter 6.5 hours of precipitation is shown in Fig. 8a, and the evolution of D10, D50, and D90 during precipitation is shown in Fig. 8b. According to Fig. 8b, steady state in the second reactor was obtained after around 5 hours of precip- itation. The particle size distribution of the product from the second reactor was bi-modal, with a D10 of 1.5 μιτι, D50 of 8.6 rn, and D90 of 14.9 μιη. The ratio of D90/D10 of the product from the second reactor was 9.93, which is lower than D90/D10 from first reactor, as desired. Importantly, the volume fraction of smallest particles (below 1 μιη in size) in the product from the first and second reactor, re ^¬ spectively, has decreased from 16 vol% to 8 vol%. It is ex ^¬ pected that narrower product particle size distributions are obtainable, if more reactors are connected in series. The final product was thoroughly washed, dried in a drying oven and stored for further processing ( lithiation/calcina- tion) .

It will be appreciated by those skilled in the art that changes could be made to the embodiments described above without departing from the broad inventive concept thereof. It is understood, therefore, that this invention is not limited to the particular embodiments disclosed, and that it is intended to cover modifications within the spirit and scope of the present invention as defined by the appended claims .