Scale-up approach for the preparation of magnetic ferrite nanocubes and other shapes with benchmark performance for magnetic hyperthermia applications

Magnetic nanoparticles are increasingly used in medical applications, including cancer treatment by magnetic hyperthermia. This protocol describes a solvothermal-based process to prepare, at the gram scale, ferrite nanoparticles with well-defined shape, i.e., nanocubes, nanostars and other faceted nanoparticles, and with fine control of structural/magnetic properties to achieve point-of-reference magnetic hyperthermia performance. This straightforward method comprises simple steps: (i) making a homogeneous alcoholic solution of a surfactant and an alkyl amine; (ii) adding an organometallic metal precursor together with an aldehyde molecule, which acts as the key shape directing agent; and (iii) reacting the mixture in an autoclave for solvothermal crystallization. The shape of the ferrite nanoparticles can be controlled by the structure of the aldehyde ligand. Benzaldehyde and its aromatic derivatives favor the formation of cubic ferrite nanoparticles while aliphatic aldehydes result in spherical nanoparticles. The replacement of the primary amine, used in the nanocubes synthesis, with a secondary/tertiary amine results in nanoparticles with star-like shape. The well-defined control in terms of shape, narrow size distribution (below 5%), compositional tuning and crystallinity guarantees the preparation, at the gram scale, of nanocubes/star-like nanoparticles that possess, under magnetic field conditions of clinical use, specific adsorption rates comparable to or even superior to those obtained through thermal decomposition methods, which are typically prepared at the milligram scale. Here, gram-scale nanoparticle products with benchmark features for magnetic hyperthermia applications can be prepared in ~10 h with an average level of expertise in chemistry. This protocol describes a solvothermal-based process to prepare gram-scale ferrite nanoparticles with well-defined shapes (nanocubes, nanostars, faceted and spherical) having heating properties appealing for clinical magnetic hyperthermia treatments.


Introduction
Magnetic nanoparticles (MNPs) are increasingly used in medicine, and developing chemical processes to prepare them is an active area of research because there are still many challenges that require technological solutions 1 . MNPs have been designed for a range of applications, with a precise control of the magnetic nanoparticle performance not only in colloidal dispersion, but also in the cellular environment 2 , i.e., in tumors.
These applications include magnetic targeting of drugs or genes 3 and the amplification of tumoral cell death by magnetic hyperthermia treatments (MHT) alone or in combination with heat-mediated immunotherapy or chemotherapy [4][5][6] . MHT requires a temperature increase in the range of 42-46°C to kill tumor cells mediated by MNPs under an alternating magnetic field (AMF). At clinically safe MHT conditions, proof-of-concept preclinical studies and clinical trials have successfully used MHT to treat primary tumors upon intratumoral deposition of MNPs directly at the solid tumor mass 7,8 .
In the past 20 years, there has been an increased interest in developing MNPs and correlating their magnetic structural features with their heating performance under MHT conditions for clinical use. Spinel ferrite nanoparticles (NPs) have the general molecular formula MFe 2 O 4 (where M represents various metal cations, including Fe, Mn, Co, Ni, Cu and Zn), and have gained attention owing to their unique properties 9 . Most ferrites exhibit superparamagnetic properties at the nanoscale dimensions of diameter below or~20 nm, and this property is ideal for clinical applications 10 . Magnetite (Fe 3 O 4 ) is one of the most commonly used ferrites in this field. In its inverse spinel structure, the A sites are occupied by Fe 3+ cations, while B sites are occupied with equal number of Fe 2+ and Fe 3+ cations and its properties can be modified by introducing doping transition metals such as Mn, Zn or Co ions 10 .
The most known routes to prepare MNPs such as iron oxide and other ferrites are based on the sol-gel process 11,12 , and include high-temperature thermal decomposition, solvothermal, polyol, chemical coprecipitation, precipitative oxidation and microwave-assisted sol-gel methods.
So far, thermal decomposition is one of the most popular methods for the production of shapecontrolled NPs. Inspired by the excellent heating properties of bacterial magnetosomes 13,14 , which naturally produce magnetite NPs of cuboctahedral shape 15 , thermal decomposition methods were adapted to produce cubic-shaped NPs (~18-30 nm cube edge), whose induction heat loss properties were proven to be similar to magnetosomes and definitely superior to spherical iron oxide NPs (IONPs).
Iron oxide nanocubes (IONCs) produced by thermal decomposition have a magnetite composition, are generally monodisperse and have high saturation magnetization (M s ) values, even for sizes below 25 nm (ref. 16 ); with values that are close to those of bulk magnetite (92 emu/g Fe3O4 at room temperature) 17,18 . Such high M s values are due to the high crystallinity and low surface spin disorder at the same magnetic volume 19,20 , resulting in higher net NPs' magnetic moments than traditional spherical IONPs produced by coprecipitation methods and now under clinical trials. These magnetic particle features, in turn, contribute to the outstanding MHT performance measured in terms of specific adsorption rate (SAR) values [21][22][23] . Moreover, compositional tuning of the nanocubes (NCs), both as mixed ferrites or as core-shell systems, boosts the heat performance by affecting the inherent magnetic characteristics of the NCs 20, [24][25][26][27] .
IONCs have shown a faster biodegradation rate than spherical IONPs due to their lower coverage percentage of the coating ligands on the NC edges with respect to the more densely coated smooth curvature of spherical NPs 28 . The fast degradation of the NCs enables tumor progression to be followed and monitored by magnetic resonance imaging. Instead, the permanence of MNPs at the tumor site at the MHT dose hinders the proper acquisition of magnetic resonance images.
In preclinical murine studies, IONCs were shown to degrade through standard iron metabolism pathways (i.e., iron transfer and iron recycling chain into ferritin storage proteins) 29 . This biocompatible profile has pushed the investigation of NCs as heat mediators in efficacy in vivo studies on xenograft tumor models for MHT, with or without the presence of toxic agents (for instance, doxorubicin as an antiproliferating agent or cobalt leakage by cobalt-ferrite NCs) 29,30 . Furthermore, the controlled assembly of NCs into anisometric assemblies has been pursued as a strategy to further improve the heat produced under AMFs in comparison with the isolated NCs, and in some cases to preserve magnetic heat losses under intracellular environments 31,32 .

Information learned from thermal decomposition experiments
In high-temperature thermal decomposition, the way the IONCs grow can be controlled by fixing certain experimental conditions 33 . These parameters include the nature of the solvent, type of surfactants, the choice of surfactant/precursor molar ratio and the heating profile applied. Some compounds that can act as ligands-for example, sodium oleate 34 , decanoic acid 35 , trioctylphosphine oxide 36 or chloride ions 37 -have a preferential affinity for specific nanocrystal facets during the crystal growth. This preferential affinity lowers the surface energy of the growing nuclei and promotes their growth along a preferential direction (i.e., [111]) to form the IONCs 33,38 . It is remarkable that an excess of sodium oleate or a mixture of oleic acid (OA)/oleylamine have accounted for the formation of FeO (wüstite) NCs and deformed NCs (the so-called nano-octopods or star-like NPs) 39,40 .
In other work, researchers found that the solvents, usually considered inert, can also play a role in controlling the cubic shape in thermal decomposition 23 .
Dibenzyl ether (DBE), for instance, is a common solvent used in the synthesis of NCs 16,22,41 ; however, in presence of acid surfactants at high temperature (~300°C), it can undergo thermolysis producing volatile subproducts such as toluene, benzyl alcohol and benzaldehyde 42 . This solvent decomposition was linked to temperature instability during the growth stage and a lack of reproducibility in the NC synthesis. The problem of reproducibility could be overcome by adding another high boiling point solvent such as squalane 23 . Interestingly, among the subproducts generated by the DBE decomposition, benzaldehyde was indirectly hypothesized to be involved in NC shape control 23,43 . In this regard, recent studies have shown that the combination of organic solvents with aromatic molecules such as DBE with 4-biphenylcarboxylic acid 16 or 1-octadecene with benzyl benzoate 43 , was also connected to the successful synthesis of cubic-shaped IONPs.

Limitations of synthesis by thermal decomposition
However, thermal decomposition allows the production of only milligram-scaled NC amounts, which are often adequate for proof-of-concept studies, but far from clinical needs (a gram-scale dose per patient is often required in MHT clinical trials), limiting their real applicability. For instance, MHT treatment per glioblastoma multiforme patient employs 5-11 mL of a nanoparticle solution at an iron concentration of 120 g Fe /L, which corresponds to an iron dose of 0.6-1.32 g per patient [44][45][46][47][48] .
The scaling-up process frequently presents serious shortcomings related to the loss of control in size, size distribution and shape of NPs, in addition to the lack of reproducibility from batch to batch in the 'lab-scale' production 49 , these problems ultimately result in deterioration of magnetic performance of the product.
Although size and size distribution can be improved by post-synthesis size-sorting protocols, by sorting out the size fractions with optimal MHT performance 50 , the isolation of a samples fraction with the desired shape is not trivial. The successful isolation of cubic shaped NPs from a mixture of [110]-faceted gold nanostructures with a size range below 100 nm has been reported recently 51 . However, shape sorting has not been reported for iron oxide yet, and it may not be an effective strategy because, so far, shape control in iron oxides and ferrites has been so dependent on the exact parameters used in the synthetic protocol.
There has been an attempt to scale up nanoparticle synthesis by the thermal decomposition method for the production of quasi-spherical IONPs of 9-22 nm (ref. 52 ). Unfortunately, the NPs produced presented limited values of M s (below 40 emu/g at 5 K), making them a lot less suitable for MHT. Such low values of M s could be related to the metal precursor used in the thermal decomposition protocol (iron oleate) and to the formation of undesired iron oxide phases, as reported by Gonzales-Weimuller et al., who determined that the air-free thermolysis of iron(III) oleate favors the formation of wüstite (FeO) (ref. 53 ). While the authors did not report the heating efficiency of the NPs synthesized using this scaled-up protocol, NPs obtained using a very similar approach reveal that FeO hinders MHT performance 54 owing to its paramagnetic nature at room temperature and antiferromagnetic behavior at temperatures below 195 K (ref. 17 ). In fact, the improvement in MHT performance for FeO@Fe 3 O 4 core-shell NCs was recently reported by Lak et al. after a stepwise chemical phase transformation of the FeO phase into Fe 3 O 4 NCs 55 . There is, therefore, still room for further development toward the scalable production of MNPs with benchmark heat performance for MHT treatment of cancer.
Well-defined shapes and sizes with a solvothermal method We here developed a method to achieve gram-scale production of ferrite NPs with well-defined shape, i.e., NCs, star-like and other faceted NPs, showing optimal structural and magnetic properties for MHT applications. We made use of the shape-directing agents that worked well with the thermal decomposition method 22,23,33,43 to optimize the solvothermal method. The use of such directing agents allowed the preferential growth of nuclei and crystals in specific crystallographic directions. The solvothermal method has greater potential for scale up, as it provides low-energy consumption in synthesizing most of the nanomaterials and it is feasible to conduct multiple synthesis in parallel in the same oven using multiple autoclaves 56 . In addition to these advantages, the control of the morphology and size of the as-synthesized products can be achieved by exercising the thermodynamic variables such as the temperature, reaction time, pressure, reactant concentration and solvent polarity 56 . In particular, the solvent used in this method, typically an alcohol, is a versatile compound of great industrial value 57 , since it is oxygenated and provides unique combinations of vapor pressure and solubility. In view of this, the solvothermal method is more suitable than thermal decomposition methods to scale up the synthesis of ferrite NPs with well-defined shape because (i) we achieve high crystallinity of the NPs at reaction temperatures lower than thermal decomposition, and (ii) alcohols dissolve a wide range of polar and nonpolar compounds, i.e., the shape directing used in thermal decomposition. In particular, we investigated the use of aldehydes as shape-directing agents. Indeed, here the use of a known amount of aldehyde derivatives, either aromatic or aliphatic, as shape-directing agents, is the key element to achieve preferential shape control for different mixed ferrite compositions. We found that benzaldehyde and its derivatives favor the formation of cubicshaped NPs. The used of nonconjugated aromatic aldehyde (but still having a phenyl ring as substitute of the alkyl aldehydes) favors the formation of faceted NPs with preferential hexagonal shape, while the use of pure aliphatic aldehyde promotes the formation of spherical NPs.
The protocol described here is robust and ensures a high level of control of the size and shape of the NPs, as it occurs at lower temperatures (usually 200°C) than those needed in thermal decomposition methods (thus not being affected by solvent thermolysis and temperature instabilities) it does not require temperature ramp profile during the growth, thus saving time and energy, and occurs in the presence of known amounts of molecules that have been found to be linked to specific shapes, i.e., cubic. Moreover, the approach reported here is straightforward as none of the reaction steps requires an inert atmosphere or vacuum, simplifying the equipment setup required for the synthesis. As such, the protocol is simple, reproducible and scalable, and is based on simple steps (Figs. 1 and 2a), following both good manufacturing and laboratory practices. Controlling the growth and crystallization obtains MNPs with ideal heating properties for MHT, which are comparable or even superior to those found for MNPs obtained through thermal decomposition methods or through standard solvothermal methods.
Comparison with the coprecipitation method Standardized methods to produce high-quality MNPs are of essential importance 43 since the progress of nanomedicine critically depends on these materials 58 . Currently, MNPs used in clinics for MHT are produced by the coprecipitation method, which is already a gram-scaled-up procedure, but the magnetic properties of the obtained NPs are far from being optimal due to physicochemical and structural limitations. With respect to actual NPs used in clinics (spherical IONPs), our cubic-shape MNPs have at least a 10-20-fold increase in SAR values, which would require a lower MNP dose per patient and treatment to be administered intratumorally, to reach the same tumor therapeutic temperature. Thus, this protocol represents the key to the translation of new-based nanotechnology from the laboratory scale to the clinic.  9 A/ms, where H is the magnetic field amplitude and f is the frequency) and 49 W/g (Hf = 9.9 × 10 9 A/ms), respectively, under field conditions that are above the biosafety limit for MHT application. Only in one paper were NCs obtained through an autoclave reaction but such NCs were the intermediate product of the synthesis described and had a wüstite FeO composition so they were paramagnetic 61 . Indeed, when we reproduced this synthesis (experimental details are found in Supplementary Methods and transmission electron microscopy (TEM) and X-ray powder diffraction (XRD) characterizations of the sample are found in Supplementary Fig. 1) and recorded the temperature of the sample while applying an alternating magnetic field, no substantial temperature increase was recorded, given the low crystallinity of the obtained NCs. In comparison, our IONCs have SAR values very similar in magnitude to NCs of same size obtained through high-temperature thermal decomposition (for instance, for IONCs of 19 nm, the reported SAR values are 550 W/g at 220 kHz and 20 kA/m) (ref. 23 ) and exceed by far SAR values of the MNPs prepared by the above-mentioned coprecipitation or solvothermal methods.
Comparison with previous method for star-shaped NPs Nemati et al. previously synthesized MNPs with similar star shapes by a high-temperature thermal decomposition method using a mixture of OA and oleylamine at a well-defined molar ratio to control the star shape 62 . However, the as-synthesized nanostars had a wüstite antiferromagnetic phase, and not magnetite as in our case. To convert wüstite to magnetite, the authors needed to apply a postsynthesis annealing treatment. Even on the annealed NPs, the SAR values measured for any of the sizes produced (between 17 and 47 nm), post-annealed samples had SAR values below 100 W/g Fe (below Hf = 4.96 × 10 9 A/ms), much lower than the ones measured for our samples. Intensity (a.u.) 30 40 50 60 70 2θ CuKα (°)

Experimental design
A variety of parameters directly affect the size, shape and chemical composition of the NPs, as well as the mass of NPs produced per batch. To achieve NPs with the desired properties and in the desired quantity, the choice of the following parameters must be considered: 1 Choose the shape-directing agent, the aldehyde derivate molecule, to decide which nanoparticle shape will be produced (aromatic, i.e., benzaldehyde versus non-aromatic, i.e., heptanal). This parameter affects the shape of the NPs (see for instance Fig. 2b for the benzaldehyde example to obtain IONCs and Tables 1 and 2). Aldehydes in solid state are added in Step 1 of either Procedure 1 or 2, while aldehydes in liquid state are added in Step 3. 2 Choose the autoclave vessel size (25, 50 or 100 mL). 3 Choose the filling percentage (20%, 40%, 46%, 50% or 67%). 4 Choose the oven temperature (160, 180, 200 or 220°C). 5 Choose the oven reaction time (3, 6, 9 or 12 h).
These parameters affect the size of the NCs (Table 3). 6 Choose the metal precursor(s) (Fe(CO) 5 or a mixture of Fe(CO) 5

and M(Acac) 2 )
This parameter affects the final chemical composition of the NPs (Table 4) 7 Choose the amine's branching (primary, secondary or tertiary) and length (12 < C x ≤ 16) These affect the final shape of the NPs (Table 5). Amines are always added at Step 1, regardless of their state. 8 Choose the scale of the reaction.
The gram-scale production of NPs in autoclave vessels requires the use of multiple reactors with a capacity of 50-100 mL to be used in parallel for the same oven (with eight reactors for the 50 mL capacity or six reactors for the 100 mL). Set the oven temperature at 220°C (Supplementary Table 1) and the filling percentage at 20%. Alternatively, choose the Parr reactor synthetic procedure to improve the yield, achieving over 1 g of NCs per batch. 9 Choose a ligand to solubilize the NPs in aqueous media.
Among these, TMAOH, PEG-GA, ND-PEG-COOH or a amphiphilic polymer shell of PMAO. The procedures are found in Supplementary Boxes 1-4 in Supplementary Information.

Setting up
The chemical purity of the reagents and the cleanliness of the equipment is essential to the success of this procedure. Take specific care over the purity of iron pentacarbonyl, the freshness of OA (this reagent tends to age depending on the storage conditions) and the cleaning of the autoclave reactors      (see Troubleshooting), which affect the size distribution of the IONCs and their aggregation state. The use of high purity (99.99%) iron pentacarbonyl, fresh OA (the defrosting must be made before its usage) and clean autoclaves ensure monodispersed and highly stable IONCs are obtained in chloroform (CHCl 3 ) solution.
For the preparation of the stock solution (Steps 1-3 of Procedures 1 and 2) used to fill the autoclave vessel, no condenser unit nor vacuum line is needed as the dissolution of the chemicals take place at atmospheric pressure and requires a heating step at 60°C under air. The stock solution after Step 3 appears clear and homogeneous ( Supplementary Fig. 2). To make the protocol as simple as possible and to scale it up, Steps 1-2 can be carried out in an Erlenmeyer flask using a heating plate at the desired temperature, as in Fig. 2a. No magnetic stirring is applied during the oven reaction to avoid any possible aggregation of magnetic materials on the magnetic stirring bars. In addition, the oven temperature is preset at the reaction temperature, and it is fixed for the whole reaction duration while the formed autogenous pressure of the autoclave reactor ensures the mixing of the reaction elements and the high crystallinity of the final NPs obtained 63 . The typical reaction time in the oven is 6 h (overall timing of the protocol in this case is~10 h) and the reaction time in the oven can vary from 3 to 12 h.
The produced NCs can be characterized using different techniques (inductively coupled plasma (ICP) optical emission spectroscopy, TEM, XRD, superconducting quantum interference device (SQUID) and alternating current (AC) magnetometers, calorimetric measurements, dynamic light scattering (DLS) and Fourier transform infrared spectroscopy-attenuated total reflectance (ATR-FTIR)); detailed steps for sample preparation and measurement conditions are described in Procedure 3, with further information and example data available in Supplementary Information.

Adding an aldehyde
The addition of aldehyde is the key to control the shape of the NPs and to obtain stable colloids. Further discussion regarding the choice of aldehyde is included in the section 'Shape control with other aromatic/conjugated aldehydes and comparison with aliphatic aldehydes'. The use of an aromatic aldehyde such as benzaldehyde leads to the formation of highly monodispersed (21 ± 2 nm) and crystalline NCs (Fig. 2b,c), as evidenced by the XRD diffractogram that presents narrow peaks, which can be attributed to ferrite (inverse spinel) structure of the IONCs (Fig. 2d). To obtain a cubic shape, it is critical to use benzaldehyde: its absence leads to a drastic loss of shape and size control of NPs, as seen in a control synthesis carried out under the same reaction parameters, with the only difference being the lack of benzaldehyde ( Supplementary Fig. 3).
The presence of benzaldehyde also changes the solution stability of the NCs. Indeed, the as-synthesized NCs in presence of benzaldehyde had a very high dispersability in organic media, while in contrast, NPs synthesized under the same synthesis conditions but in the absence of benzaldehyde, all precipitated in the autoclave flask and it was very hard to disperse them in CHCl 3 after the synthesis (data not shown).
Interestingly, the replacement of 1-octanol with benzyl alcohol as solvent, which is the other subproduct generated during the thermolysis of DBE (used in thermal decomposition) 42 , along with the use of benzaldehyde, leads to the formation of well-defined NCs of 27 ± 7 nm ( Supplementary  Fig. 4b). However, the use of only benzyl alcohol in absence of benzaldehyde leads to a loss of the cubic shape control and the formation of truncated octahedron morphology, with a size distribution of 20 ± 7 nm ( Supplementary Fig. 4a). This experiment highlights again the key role of benzaldehyde as shape directing, which is in good agreement with what has been hypothesized for thermal decomposition synthesis of NCs 23 .
Optimization of the SAR values by NC size control Besides crystallinity and shape, the NC size is a critical parameter to tune, while maintaining very narrow distributions to optimize the MHT performance of NCs, and thus the SAR values. This protocol allows the size of the NCs to be tuned by four independent parameters: 1 The volume capacity of the autoclave vessel 2 The filling percentage of the autoclave (filling percentage (FP) = the stock solution volume/reactor volume × 100) 3 The time of the reaction 4 The solvothermal crystallization temperature Table 3 summarizes the key experimental parameters varied to tune the size of the NCs from 8 to 42 nm.
The first two parameters are interconnected. We used standard Teflon or polyphenylene polymerlined (PPL) autoclave reactors with capacities of 25, 50 or 100 mL (for the detailed experimental PROTOCOL NATURE PROTOCOLS methodology, see Table 3 and the electronic supplementary information, ESI) and with FP varied from 20% to 70%. • The size obtained in the 25 mL autoclave is~18-20 nm (standard deviation, 0.1) for a FP of 46% • For 50-100 mL reactors maintaining the same 46% FP, the IONC size is much lower at~10 nm NCs (standard deviation,~0.07) To further tune the size on the chosen autoclave, the volume filling percentage was systematically changed. For instance, in the 50 mL reactor, the stock solution containing 0.6 mL of OA, 0.2 g of hexadecylamine (HAD), 1 mL of benzaldehyde and 2 mL of Fe(CO) 5 was diluted with 8, 16, 24 and 30 mL of 1-octanol respectively, to change the FP from 23% to 67%. The size of IONCs changed from 8 ± 1 to 12 ± 1 nm, while the cubic-like shape was maintained in all cases ( Supplementary Fig. 5 and see samples 4-7 from Table 3 for exact experimental parameters). At a low filling percentage, the dispersability of the NCs was greater than at a higher filling percentage, which could be related to the absolute amount of benzaldehyde present in the reaction pot.
As a third parameter, it was found that reaction time plays a role in varying the size control: for instance, in a 25 mL autoclave, when the reaction time was varied from 3 to 12 h, while keeping all the other parameters the same (filling percentage, stock solution, etc), the IONCs' edges varied in a wide range from 16 ± 1 to 42 ± 8 nm (Fig. 3b).
Crystal size determined through XRD is in agreement with the size determined through TEM (Supplementary Table 2), thus we can confirm the high crystallinity of the NCs synthetized with our solvothermal protocol. Cubic shape is maintained in all cases; however, at short times (~3 h), the NCs edges appear sharper and the cubic faces slightly concave, with XRD data confirming the inverse spinel structure of this sample ( Supplementary Fig. 6). These types of NPs, 'nano-octopods' or 'nanostars', have previously been obtained using thermal decomposition route, 62 but never obtained with solvothermal route.
Finally, as a fourth parameter, the temperature of the oven at which the solvothermal crystallization occurs also enables tuning of the NCs' size. For instance, for 25 mL autoclave, when Time of reaction (h) 10 12 Oven temperature (°C)  Table 3 contains the exact experimental conditions used in each case. Panels a and c adapted from ref. 80 , CC BY 4.0 (https://crea tivecommons.org/licenses/by/4.0/). modifying the temperature from 160°C to 240°C, a notable change in size from 11 ± 1 to 24 ± 3 nm was observed (Fig. 3c); however, below 200°C, the NPs displayed a truncated octahedral shape ( Supplementary Fig. 7).

Ligand exchange
The as-synthesized IONCs, which are surfactant coated, are easily transferred from CHCl 3 to water by applying a simple ligand exchange protocol using TMAOH, according to a reported protocol 35 , with certain modifications found in Supplementary Box 1. The short TMAOH ligand replaces the organic surfactants at the NP's surface and provides negative charge at physiological pH, thus offering stability to NPs via charge repulsion 35 .
It is important to note that ligand exchange is not effective if the NPs are prepared in the absence of aldehyde. Initially, this was noticed when we compared ligand exchange reactions on NCs prepared in the presence and absence of benzaldehyde. In the presence of benzaldehyde, the ligand exchange protocol is straightforward and provides NCs with a hydrodynamic size centered at 69 ± 16 nm (intensity-weighted peak) for TEM magnetic cores of 21 ± 2 nm ( Supplementary Fig. 3). The TMAOH-coated NCs have a Z-potential value of −53 ± 8 mV, with high stability in water for several weeks. On the other hand, the NPs synthesized in the absence of benzaldehyde had a hydrodynamic size that is much larger and broader (800 ± 160 nm) than the NC sample, indicating the poor and limited stability of this sample ( Supplementary Fig. 3).
It is possible to use different ligands (Supplementary Methods), such as PEG-GA, ND-PEG-COOH and an amphiphilic polymer shell of PMAO, which have been found to be suitable coating agents for in vitro/in vivo tests of MNPs for biomedical purposes 22,23,29,64,65 .
Among the different water-soluble protocols, the TMAOH ligand here was selected for the whole study because it is straightforward and much quicker (overall time is 2 h) than other water protocols, since they require an overall time of 8-70 h, as described in Supplementary Methods. Therefore, unless stated, we always selected TMAOH ligand for the water transfer of the NPs.
Scaling up the synthesis of IONCs with optimal magnetic hyperthermia performance to the gram Given the ease and robustness of the solvothermal method proposed here, it is possible to perform multiple and parallel synthesis using high-volume vessels, thus increasing the number of reactors per oven and scaling up the production of NCs by performing parallel synthesis, while preserving the high quality of the cubic shape, monodispersity and crystallinity of the obtained NPs. The mass of IONCs obtained per batch depends on various factors, including the autoclave volume, filling percentage and oven temperature (Supplementary Table 1), and it can be varied between 0.04 and 0.7 g of NCs per autoclave. This amount is maximized in the case of 100 mL autoclaves, setting the oven at 200°C and for an autoclave filling percentage of 46%; however, the size of the NCs is generally small (d crystal < 12 nm). With small autoclaves, to increase the mass of NCs obtained per autoclave with optimal size for MHT (d crystal > 12 nm), temperature was exploited as the parameter that had a positive impact on the mass of NCs per autoclave. In this case, we could successfully increase the mass production of NCs fourfold (i.e., for standard 20 ± 2 nm NCs, the obtained mass increased from 0.1 to 0.4 g when passing from 200°C to 240°C; however the resulting NCs obtained at high temperature had a size of 24 ± 3 nm). In terms of iron conversion per vessel, considering the ratio between the amount in iron of formed NCs and the initial amount of iron introduced as iron pentacarbonyl precursor, the iron conversion corresponded to~9-13% when using 25 mL vessels. The iron conversion yield could be further improved (~25-35%), by either decreasing the autoclave filling percentage or increasing the oven temperature from 200 to 240°C, at which solvothermal crystallization occurs (Supplementary Table 1).
Alternatively, a Parr's stirred scaled reactor can be used in place of the autoclaves to increase the iron conversion (Supplementary Methods). The Parr reactor and its different parts is shown in Fig. 4a. When reproducing the typical synthesis of IONCs up-scaled four times with minor modifications (the capacity of the reactor is 100 mL) (for the exact experimental details used in this case, see Supplementary Information), over 1 g of IONCs per reactor were obtained, with well-defined shape and a size of 18 ± 2 nm (Fig. 4b). The use of this kind of reactor, by increasing the iron yield conversion to over 70%, allows one to scale the production of ferrite NCs with the optimal size to the gram scale in one single reactor. Furthermore, the IONCs produced in Parr's reactor displayed SAR values of up to 630 W/g Fe at 180 kHz ( Supplementary Fig. 8), in agreement with the NCs of similar cube edge obtained in the Teflon autoclave (Fig. 3e,f).
It is worth noting that, when comparing the SAR values of our NCs sample of 18 ± 2 nm and those of NPs used in clinic (NanoTherm), our NCs display a SAR value of 240 W/g Fe (Fig. 3e) versus the 30 W/g Fe of the NanoTherm product at the same clinical field conditions (16 kA/m and 100 kHz) (ref. 66 ). Possibly, the smaller and spherical shape of the NanoTherm IONPs may account for such lower heating performance with respect to our NCs product.

Tuning the composition of the NCs with Mn and Zn
Having established a high level of control over the size and shape of the NCs, we adapted our solvothermal protocol to obtain NCs that have different material compositions, expanding the materials range from iron oxide to other mixed ferrite compositions. This adaptation is described in Procedure 2, with exact chemical amounts and critical parameters summarized in Table 4. This is achieved quite simply by introducing a mixture of metal precursors, i.e., iron pentacarbonyl (14.3 mmol) and a divalent metal acetylacetonate (0.5 mmol), such as either Mn acetylacetonate or Zn acetylacetonate, instead of iron pentacarbonyl alone. While the procedure is very similar, a few parameters needed to be changed and optimized.
After dissolving the amine and the surfactant in the alcoholic solvent at 60°C, the solution was maintained at 60°C under magnetic stirring, and first the Zn(II) acetylacetonate, Zn(acac) 2 for Znbased ferrite NCs (or Mn (II) precursor when aiming for Mn-based ferrite NCs) were dissolved for at least 30 min. Next, the addition of iron pentacarbonyl and, finally, benzaldehyde was introduced in the reaction vessel.
To obtain the well-defined cubic-shape NPs, the precursors feed Fe/metal ratio was a key parameter. Indeed, the Fe/Zn ratio must be kept above 1:28 to achieve the synthesis of 16 nm Zn 0.6 Fe 2.4 O 4 NCs with narrow size distribution (at σ =~0.1), high crystallinity and pure chemical composition of ferrite (Fig. 5a,d and Supplementary Fig. 9). This is far from the Fe/Zn stoichiometric feed ratio of 2:1, which have been previously reported in literature to successfully synthesize mixed ferrite compositions 26,27,67 . When the Fe/Zn feed ratio was kept at 1:2, the nucleation of both cubic Zn ferrite and irregular 'bullet'-like (elongated triangular) ZnO NPs occurs (TEM and XRD data available in Supplementary Fig. 10). When higher Fe/Zn feed ratios were used (e.g., 1:4 and 1:6), the population of 'bullet'-like NPs was sequentially smaller, and the relative intensities of the XRD peaks attributed to the ZnO phase were also less prominent (Supplementary Fig. 10).
When introducing Mn instead of Zn, also at an Fe/Mn feed ratio of 1:28, ferrite 12 ± 1 nm Mn 0.6 Fe 2.4 O 4 NCs could be easily obtained with very good shape and size control (Fig. 5c,f).
Shape control with other aromatic/conjugated aldehydes and comparison with aliphatic aldehydes The structure of the aldehyde affects the shape of the NPs. In our initial experiments we replaced benzaldehyde with other aromatic aldehydes at the same equivalent molarity (9.8 mmol). The first choices were focused on either 4-methylbenzaldehyde (M-BZD) or 4-phenylbenzaldehyde (for experimental conditions, see Fig. 6a,b and Table 1), which could be considered as derivatives of benzaldehyde having a para substitution of a hydrogen atom with a methyl or a phenyl group, respectively. In both cases, NCs of iron oxide with well-defined size and shape were obtained. Next, benzaldehyde was replaced with another aromatic derivative namely, 3-methoxy-benzaldehyde, and NCs were still obtained. In this case, the edge of the NCs appeared more roundish, along with a marginal amount of smaller spherical NPs (Fig. 6c). We may speculate that the moieties (methyl, phenyl and methoxy) and the position of the substituents on the phenyl ring of the benzaldehyde (para or meta in this case) may affect the preferential binding of the aromatic aldehydes toward specific nanocrystal facets. This may possibly arise from their positive or negative inductive effects (+I or −I), which may stabilize radical species of aromatic aldehyde that might be forming during the solvothermal synthesis at 200°C. To test this hypothesis, we conducted the synthesis in the presence of aromatic aldehyde derivatives that cannot be considered a direct derivate of benzaldehyde, but that also have an aldehyde group close to the aromatic phenyl ring to form a highly resonant structure with a higher possibility of electron/radical delocalization. For instance, when choosing α-(2-methylpropylidene)-benzeneacetaldehyde (α-BZAD) as the shape-directing agent, with the formyl and the phenyl groups both bound directly to an α carbon-carbon double bond (sp 2 carbon atoms), both the size and the cubic shape of resulting NPs were well controlled (Fig. 6d).
Interestingly, when employing aldehyde molecules containing an aldehyde group and a phenyl ring having one sp 3 carbon atom separating the formyl and the phenyl group, and no possibility of electron delocalization or conjugation effect (e.g., 2-phenylacetaldehyde) the resulting NPs had a faceted structure, corresponding to truncated nano-octahedra with a size of 20 ± 3 nm.
With 3-phenylpropionaldehyde, having the phenyl and the formyl group separated by two sp 3 carbon atoms, quasi-spherical NPs were obtained. Such morphologies are clearly different from the cubic-shaped NPs (Fig. 7b,c).   When aliphatic aldehydes are used instead of benzaldehyde, the morphologies of the resulting NPs are clearly spherical and far from the cubic-shaped NPs. For instance, by selecting aliphatic aldehyde with different lengths of the alkyl group such as heptaldehyde, pentanal and decanal, spherical shapes with a very narrow size distribution could be obtained (Fig. 7d-f).
In general, the diameter of the NPs obtained in the presence of aliphatic aldehydes was small (~12 nm), when applying standard experimental conditions (25 mL autoclave, FP 46%, 200°C and 6 h); these same conditions would lead to NCs of 18(20) ± 2 nm in the presence of an aromatic aldehyde, i.e., benzaldehyde. The size of these NPs can also be tuned by simply tuning the FP, temperature and time of the reaction (for the exact conditions used, see Table 2, and Fig. 7 and Supplementary Fig. 11 for the corresponding TEM images), obtaining nanospheres of 15 ± 2 nm in the case of pentanal, 17 ± 2 nm with heptanal and 18 ± 2 nm with decanal. Besides saturated aliphatic aldehyde, the unsaturated aliphatic aldehydes, such as (cis-4-hepten-1-al) could also provide quasi-spherical NPs ( Supplementary Fig. 12).

Star-like NCs: growth control of the MNPs through benzaldehyde and amine ligands
Another parameter that can be changed and optimized is the choice of amine ligand.
NPs with star-like shapes can be made by keeping the same aldehyde directing agent (benzaldehyde) and playing with the selection of the alkyl-amine ligands. Thus, the shape of NCs could be further tuned from highly regular NCs to star-like NPs presenting concave faces and elongated edges when the hexadecylamine was replaced with tertiary amines, such as trioctylamine or tridodecylamine (for experimental conditions, see Fig. 8a and Table 5).
Interestingly, there is a similar effect with secondary amines, such as dioctylamine or didodecylamine, but the star shape is less pronounced (Supplementary Fig. 13). Star-shaped NPs can also be obtained when hexadecylamine is replaced with oleylamine 39,62 (Fig. 8a). Here, however, with the use of benzaldehyde and a tertiary amine, we obtained the synthesis of highly crystalline ferrite starlike NPs, as confirmed by XRD diffraction characterization (Supplementary Fig. 14).
The replacement of the primary amine with a secondary/tertiary amine or with oleylamine could have an impact on the decomposition rate of the iron precursor, and thus on the shape, promoting the star-like NPs, since this shape is also obtained for primary amines at short periods of solvothermal crystallization times (i.e., 3 h). Moreover, zero-field-cooled (ZFC)/field-cooled (FC) curves show that the blocking temperature of the NCs and the star-like NPs was very different, despite the very similar size, in that the star-like NPs blocked at higher temperatures ( Supplementary Fig. 15). However, the magnetization versus field, M(H), hysteresis cycles at 5 K show almost no difference in terms of M s or H c (even H c is slightly less for the star-like NPs).  c CRITICAL STEP The basic reaction uses hexadecylamine. This could be replaced with different primary, secondary or tertiary amines to tune the degree to which the shape looks like star. Table 5 summarizes all the used amines and the main parameters of the synthesis. 2 Add the metal precursor (14.8 mmol) to Solution 1. 3 Add the aldehyde (9.8 mmol) to Solution 1 and stir for 30 min at room temperature (RT (25-28°C); final stock solution). c CRITICAL STEP The choice of aldehyde affects the shape of the nanoparticle (Tables 1 and 2). For NCs, use benzaldehyde; chemical amounts and conditions specified in Tables 1-3. See also Table 6 for a complete list (magnetic material mass for protocols employing different aldehydes). 4 Fill the autoclave(s) with the final stock solution, up to a volume filling percentage between 20% and 70% (to achieve these values the volume of the solvent is adjusted). c CRITICAL STEP The autoclave size affects the size of the NCs. For samples 2 and 3, use 50 and 100 mL autoclave vessels,

NATURE PROTOCOLS
respectively. More stock solution will be needed for larger vessels.
c CRITICAL STEP The filling percentage affects the NC size. The filling percentage is increased by adding more of the alcohol solvent. For instance, for the 50 mL autoclave reactors, 0.2 g of HAD, 0.6 mL of OA, 2 mL of Fe(CO)5, 1 mL of BZD and the 1-octanol volume was 8, 16, 24 and 30 mL to achieve filling percentages of 23%, 40%, 55% and 67%, respectively. 5 Place the autoclaves in an oven preheated at 160-240°C for between 3 and 12 h (pressure achieved 20-30 bar). c CRITICAL STEP The temperature and reaction time both affect the NC size; these could be explored using the conditions for samples 9-15 of Table 3, in these examples, we used the 25 mL autoclave, the time of the reaction was set at 3, 6, 9 or 12 h to increase the size of the NPs and the temperature of the oven was preheated at 160, 180 or 240°C. j PAUSE POINT The autoclave can be left to cool down to RT naturally and can be left for 1-3 d before washing the NPs. 6 Cool down the reaction mixture and collect the NPs in 45 mL Falcon tubes (2, 4 and 6 Falcon tubes for the autoclaves of 25, 50 and 100 mL, respectively) with CHCl 3 (15 mL for each Falcon tube). 7 Sonicate the NPs in CHCl 3 for 5 min at RT. 8 Add acetone (30 mL for each Falcon tube), then centrifuge (3,893g, for 20 min at RT) to precipitate the NPs fraction. 9 Redisperse the final product in CHCl 3 (20, 40 or 80 mL), to obtain NP solution of~6-8 mg/mL. j PAUSE POINT For optimal storage conditions, keep the batches in capped glass vials of appropriate volume at 4°C. Samples that will be used in the short term (up to few months) can be stored at RT. 10 Characterize the product. The important things to determine are: • The composition • The size; including hydrodynamic size (DLS) • The shape • How well it performs the desired function, e.g., magnetic properties • Magnetic hyperthermia performance measured in terms of specific adsorption rate Supplementary Information provides advice on how to perform the following assays, some of which require a water transfer before they can be carried out ( Supplementary Boxes 1-4). • Elemental analysis: inductively coupled plasma optical emission spectroscopy • TEM • DLS to determine hydrodynamic size and charge surface (water transfer) • XRD • Magnetic characterization: magnetic measurements were performed on a SQUID (water transfer) • AC magnetic measurements: the inductive magnetic characterization was carried out with a homemade AC magnetometer (water transfer) • Calorimetric measurements of the specific absorption rate: a commercially available magnetic nano-heating device (water transfer) (Optional) Water transfer of IONCs using TMAOH c CRITICAL For any applications or assays where the NCs need to be suspended in water, perform a water transfer procedure. There are many ways in which this can be done ( Supplementary Boxes 1-4). We have included three alternative procedures using gallic polyethylene glycol, α-nitrodopamine-ω-carboxy-poly(ethylene glycol) or poly(maleic anhydride-alt-1-octadecene) in Supplementary Methods. This one using TMAOH was chosen, because it is the shortest. 11 Centrifuge (3,893g, 20 min at RT) 1.4 mL of a solution of NPs in chloroform ([Fe] = 7 mg/mL) after addition of acetone (5 mL) to precipitate and collect the NPs. 12 Discard solvents and dry the pellet gently using a nitrogen flux. 13 Add 1 mL of a TMAOH solution in anhydrous ethanol (0.27 M) to the nanoparticle powder. 14 Sonicate the solution for 30 min at RT. 15 Add Milli-Q water (3 mL) and centrifuge the solution using centrifuge membrane tubes (432g, 10 min at RT). 16 For the NPs solution in the centrifuge tube, repeat Step 5 at least two more times to wash out the excess of TMAOH ligands. 17 Recover the NP solution and redisperse in 2 mL Milli-Q water.
j PAUSE POINT For optimal storage conditions, keep the samples in capped containers of appropriate volume at 4°C. Samples that will be used in the short term (up to few days) can be stored at RT.

Procedure 2 c
CRITICAL Ferrite NCs are prepared by introducing a non-stoichiometric mixture of iron pentacarbonyl, Fe(CO) 5 and Zn acetylacetonates or Mn acetylacetonates. The experimental process is very similar to that for the preparation iron oxide NCs (Procedure 1). Detailed experimental conditions for adjusting the size and chemical composition of the NCs are summarized in Table 4. 1 Prepare a homogeneous solution (Solution 1) containing: • The OA surfactant (1.9 mmol) • The aliphatic amine (0.8 mmol) • Iron pentacarbonyl (14.3 mmol) • One divalent metal precursor (0.5 mmol) that is either zinc(II) acetylacetonate (Zn(Acac) 2 ) or manganese (II) acetylacetonate, (Mn(Acac) 2 ) • The alcoholic solvent (8 mL) If this mixture of precursors is used to produce Zn ferrite or Mn ferrite NCs, make the mole ratio between Zn/Mn and Fe 1:28. j PAUSE POINT The autoclave can be left to cool down to RT naturally and can be left for 1-3 d before washing the NPs. 6 Cool down the reaction mixture and collect the NPs in 45 mL Falcon tubes (2, 4 and 6 Falcon tubes for the autoclaves of 25, 50 and 100 mL respectively) with CHCl 3 (15 mL for each Falcon tube). 7 Sonicate the NPs in CHCl 3 for 5 min at RT. 8 Add acetone (30 mL for each Falcon tube), then centrifuge (3,893g, 20 min at RT) to precipitate the NP fraction. 9 Redisperse the final pellet, after discarding the solution supernatant, in CHCl 3 (20, 40 or 80 mL), to obtain a NP solution of~6-8 mg/mL. j PAUSE POINT For optimal storage conditions, keep the batches in capped glass vials of appropriate volume at 4°C. Samples that will be used in the short term (within few days) can be stored at RT. 10 Characterize as described in Step 10 of Procedure 1. 11 Ligand exchange with TMAOH as described for water transfer of Fe 3 O 4 nanocubes prepared using the benzaldehyde-based solvothermal protocol summarized in Fig. 1.

Procedure 3
DC magnetic characterization • Timing 56 h 1 Prepare aqueous samples at concentrations above 2 mg/mL. 2 Dropcast 50 µL of the sample in a polycarbonate capsule with 60 mg of gypsum and let it dry overnight. 3 Enter the settings for the SQUID. In a typical experiment, hysteresis loops are recorded within the magnetic field of ± 60,318 Oe at 298 and 5 K. The ZFC and FC temperature-dependent magnetization measurements were performed on samples prepared in the same way as point 2. In the FC measurements samples were cooled in presence of a cooling field of 50 Oe. In the ZFC, samples were heated in presence of a 50 Oe magnetic field. 4 Correct the magnetization data with respect to the diamagnetic and paramagnetic contributions of water and gypsum using the automatic background substraction routine. 5 Normalize the curves to the metal concentration obtained from the elemental analysis. 6 Determine the average and standard deviation for the magnetic area values.
AC magnetic measurements • Timing 1 h 7 Prepare samples at concentrations above 2 mg/mL in metal content of Fe or Fe+Zn or Fe+Mn. 8 Enter the settings for the AC magnetometer (Hyster 2). In a typical experiment, AC hysteresis loops are recorded at different field intensities (8, 16 or 24 kA m −1 ) and frequencies (100-300 kHz). The measurements are conducted at a fixed temperature of 20°C.
9 Record three AC magnetization curves for each sample (using 40 μL). Determine the average and standard deviation for the magnetic area values. 10 Calculate the SAR value using the equation: SAR = A·f·m −1 where A is the area of the acquired hysteresis loops, f is the frequency of the applied AMF and m is the total metal mass (the AC magnetization signal was normalized to the mass of magnetic elements Fe, Zn and Mn present in the studied magnetic colloid).
Calorimetric measurements of the specific absorption rate • Timing 3 h 11 Prepare 300 μL samples at concentrations (Fe, Fe + Mn or Fe + Zn) above 2 mg/mL in water. 12 Enter the settings for the AC applicator. In a typical experiment, NCs are exposed to an AMF (from 12 up to 32 kA m −1 ) at two different frequencies (105 and 300 kHz) and three heating curves are recorded. 13 Determine the initial slope (dT/dt) of the temperature enhancement versus time considering only the first 60 s through the linear fitting of these points. Each datapoint is the average dT/dt value of three independent measurements. 14 SAR values were calculated using the following formula: where C d is the specific heat capacity of the dispersion medium (for water C d = 4,185 J g −1 K −1 ), m d is the medium mass and m is defined as the total metal mass per gram of the dispersion. 15 Normalize SAR values using the metal mass, m (Fe, Fe + Mn or Fe + Zn) determined by ICP. 16 For all the samples, SAR values were measured at the frequencies 100-120 kHz, because this frequency range is used in-clinic, while for samples that had very low SAR at the clinical range, SAR were also measured between 200 and 300 kHz. The two instruments (the magnetic calorimeter and the AC magnetometer), measure at frequency values that are close to each other but not always identical.

Troubleshooting
Troubleshooting advice can be found in Table 7.
Timing Procedure 1 Step

Anticipated results
Procedure 1 can be used to prepare a series of highly monodispersed IONCs in the size range of 8-24 nm ( Fig. 3d and Supplementary Fig. 16), and the methods of Procedure 3 can be used to study the corresponding heating capabilities. Each NC size sample is achieved by tuning the abovementioned experimental parameters (autoclave volume, autoclave filling percentage, temperature and solvothermal crystallization time) with the experimental conditions summarized in Table 3 (samples marked with an asterisk correspond to those selected for the water transfer). SAR measurements for samples at systematically larger cube edges were determined at different magnetic field conditions (H = 8, 12, 16, 20, 24 kA/m and f = 100 or 200 kHz) with an AC magnetometer (Fig. 3e,f). SAR values were calculated according to Procedure 3. We found that the optimal size (d opt ) of our IONCs to maximize SAR values is 18 nm NCs. In this sample, SAR values reaches values from 450 W/g Fe (H = 24 kA/m, f = 100 kHz, which corresponds to an Hf of 2.4 × 10 9 A/ms) to 790 W/g Fe (H = 24 kA/m, f = 200 kHz, which corresponds to an Hf of 4.8 × 10 9 A/ms), both complying safe magnetic hyperthermia field conditions (Hf ≤ 5 × 10 9 A/ms) 68 . NCs larger than 20 nm in cube size have a decrease in SAR values with respect to 18 nm NCs. We attribute this observation to aggregation effects of MNPs in solution, especially for the 24 nm cores. At RT, in the range between 20 and 24 nm, the transition from superparamagnetic to ferromagnetic single domain occurs for IONPs and, the larger the magnetic core is, the stronger the magnetic interactions among particles are, affecting in turn, the hydrodynamic volume of the NCs in solution and their heating performance 23,69 .
Regardless of the size, the NCs here obtained have very high magnetic moment, reaching saturation magnetization values of 98-117 emu/g Fe at 298 K and 123-132 emu/g Fe at 5 K ( Supplementary Fig. 17), which are equivalent to maximum values of 74 (at 298 K) and 84 (at 5K) emu/g Fe3O4 (Table 8). These values are very close to those of bulk magnetite (92 emu/g Fe3O4 at RT) 17,18 and sometimes even higher then M s values of NCs of similar sizes (below 30 nm) obtained through either nonhydrolytic or hydrolytic approaches 8,33 . IONCs with sizes below 21 nm display a superparamagnetic behavior at RT and possess coercive fields of 209-355 Oe at 5 K (Table 8).
For comparison, the SAR values of some of the samples in the size range of 11-18 nm were also determined through calorimetric measurements using a commercial calorimetric apparatus (nanoScale Biomagnetics). In this case, we exposed the colloidal IONCs to an AC coil at specific magnetic field conditions (220 kHz and 20 kA/m, thus the Hf factor is equal to 4 × 10 9 A/ms) and recorded the temperature rise over time ( Supplementary Fig. 18). As observed in the curves, the IONCs caused a fast temperature increase. The SAR values (SAR Calor. ) are outstanding and are in the range of 50-600 W/g Fe . When SAR Calor. values are compared with those determined through AC magnetometry (SAR Magnet ), we observe that SAR Calor. values are slightly lower than SAR Magnet. values (Supplementary Table 3). This trend is in agreement with other reported comparative studies 70 , as SAR values determined through the calorimetric methods are expected to be lower than those derived from the magnetometer due to the non-adiabatic heat losses that have more of an affect on the calorimetric measurements than the AC measurements. When comparing the heating properties of NC samples of IONCs of 20 ± 2 nm with the irregular magnetite NPs obtained following exactly the same solvothermal protocol but in the absence of benzaldehyde, these irregular NPs obtained have very poor SAR values, well below 200 W/g Fe at the highest field amplitude frequency ( Supplementary  Fig. 19, see the comparative temperature versus time curves for both samples under the same measurement conditions). This can reasonably be attributed to the higher polydispersity in size and shape of this sample, as well as its worse colloidal stability. On the other hand, IONCs synthesized in the presence of benzaldehyde have SAR values well above 200 W/g Fe , reaching SAR values of 800 at the highest frequency and field amplitude (300 kHz and 24 kA/m).

Water transfer of IONCs using ligands/polymers to suitable for in vitro and in vivo tests
As-synthesized IONCs coated with OA molecules are soluble in chloroform or other organic solvents. However, they can be easily transferred into aqueous media by different protocols. To the same core batch of 14 ± 1 nm iron NCs, well-established procedures were applied by employing different water transfer ligands, including gallic-polyethylene glycol (PEG-GA), ND-PEG-COOH or with a polymer coating protocol employing the amphiphilic polymer PMAO. Supplementary Boxes 1-4 summarize the step-by-step procedures used for each ligand/polymer coating of the NCs, following previously published protocols, with minor modifications 23,27,32,71 . For comparison, we also exchanged the same initial batch of NCs with TMAOH 35 as reported in Supplementary Box 1. Generally, all the IONC samples obtained by ligand exchange protocols had a nanoparticle yield in water of nearly 100%. The ligand exchange/polymer-coated NCs were fully soluble in water with no sign of aggregation: the hydrodynamic sizes of these samples were compatible with individually coated NCs in the hydrated forms (Fig. 9a). For instance, the mean hydrodynamic diameters in water for TMAOH, PEG-GA and ND-PEG-COOH peaked at 27 ± 8 nm, 38 ± 10 nm and 39 ± 10 nm, respectively (Fig. 9a,b), while for the PMAO polymer-coated ligand (sample P1), the hydrodynamic size peaked at 120 ± 30 nm.
The long-term stability of the colloids is generally very good (nearly more than 1 year) given the steric hindrance exerted by polyethylene glycol molecules and/or the negative surface charge of the coatings. Indeed, the Z-potential values recorded were in the range of −52 ± 4 mV for TMAOH, −36 ± 3 mV for PEG-GA, −28 ± 2 mV for ND-PEG-COOH and −49 ± 4 mV for P1 (Fig. 9c,d).
Independently of the type of coatings, very similar SAR values were measured for the NCs transferred to water via different ligands (Fig. 9e,g). Only for the polymer coated sample were the SAR values slightly lower. This is probably due to the larger hydrodynamic size of this sample. Moreover, these SAR values recorded on our NCs are very similar to those reported for NCs prepared by thermal decomposition methods and coated with PEG-GA ligands 23 , again confirming the high quality of the NCs produced by our solvothermal protocols.
Among the different water-soluble protocols, the TMAOH ligand was selected for the whole study because it is straightforward and much quicker (overall time is 2 h) compared with the other water protocols, since they require an overall time of 8-70 h (as described in Supplementary Boxes 1-4). Therefore, unless stated, we always selected TMAOH ligand for the water transfer of the NPs. The magnetic parameters were recorded at room temperature (298 K) and at 5 K. The saturation magnetization (M s ) and the coercive field (H c ) are reported for each of the samples.
Water transfer of mixed ferrite NCs using TMAOH ligand Table 4 summarizes the precise experimental conditions, size and chemical composition of the final Zn and Mn ferrite NCs obtained following Procedure 2. When carrying out the same synthesis with only the iron precursor, the NCs obtained had a size of 18 ± 2 nm (Fig. 5b,e). Zn-and Mn-ferrite NCs were solubilized in water using the above-mentioned protocol with TMAOH ligand. In the case of Zn-and Mn-ferrite NCs, the mass of NPs with an exhanged ligand was 10 mg (in this case all the magnetic elements were considered to calculate the concentration: Fe + Zn or Fe + Mn). Of note, even the Zn-and Mn-ferrite NCs stabilized in aqueous media can efficiently increase the temperature of the solution under the application of AMFs and provide efficient SAR values at AMF conditions relevant for the clinic, i.e., 100 kHz and 12-32 kA/m (Fig. 5g,h Fig. 22). These values are very similar and slightly lower than that of IONCs of similar cube edge (15 ± 1 nm). On the other hand, Mn-ferrite NCs of 12 ± 1 nm, have SAR values up to 140 W/g Fe+Mn for a frequency of 100 kHz (8-24 kA/m) and 340 W/g Fe+Mn for a frequency of 200 kHz (8-24 kA/m), being generally higher than SAR values of IONCs of a very similar size (11 ± 1 nm) (   Supplementary Fig. 22). In this case (IONCs), SAR values of 90 and 280 W/g Fe were measured for frequencies of 100 and 200 kHz, respectively. Interestingly, at low-field amplitudes (<20 kA/m), the SAR enhancement was well above 20%. Surprisingly, M s of Mn-ferrite NCs is not higher than that of IONCs of similar size ( Supplementary Fig. 23), as previously reported 24 . Nevertheless, our M s value (100 emu/g Fe+Mn at 300 K) is in agreement with that reported for Mn-ferrite quasi-spherical NPs of 12 nm produced by Leal et al. through the thermal decomposition method 72 . Overall, we could show that the MHT heat performance of the NCs produced here can be tuned by the incorporation of other divalent metals and Zn and Mn ions. Both samples can efficiently increase the temperature of the solution under the application of AMFs, and Mn stands as improving the heat performance when compared with IONCs produced by the same solvothermal approach.

Results from experiments using different aldehydes
The SAR values of MNPs with quasi-spherical/faceted shapes of 15 ± 2, 17 ± 2 and 18 ± 2 nm and the truncated nano-octahedra of 20 ± 3 nm were also measured, displaying increasing SAR values in accordance with their size (Fig. 7h) at field conditions of f = 100-300 kHz and H = 12, 16 and 24 kA/m. Surprisingly, at 100 and 300 kHz and 24 kA/m field amplitude, nano-octahedra of 20 ± 3 nm present SAR Magnet. values up to 480 and 1,500 W/g Fe , respectively, which are even slightly higher than SAR values of the NCs of similar size (20 nm) obtained with benzaldehyde ( Fig. 3) or the other benzaldehyde derivatives such as M-BZD (Fig. 6f). Probably, such high values are related to a particular behavior of this sample in solution, i.e., chains of magnetically interacting crystals, which have shown to play a crucial role in MHT 8,14,40 . Indeed, as reported, similar nano-octahedra of 22 ± 2 nm synthesized by thermal decomposition methods displayed SAR values of up to 300 W/g Fe (Hf = 3.1 × 10 9 A/ms) 73 . On the other hand, quasi-spherical/faceted shaped particles of 15 ± 2 (pentanal), 17 ± 2 (heptanal) and 18 ± 2 nm (decanal) displayed SAR values of up to 500 W/g Fe (Fig. 7h), and if we compared them with NCs of similar size and composition, the heat performances were lower (for example, 18 nm IONCs for a Hf of 4.0 × 10 9 A/ms have a SAR Cal of 500 W/g Fe and 18 nm faceted MNPs for a Hf of 4.8 ×10 9 A/ms have a SAR of 200 W/g Fe ). Our quasi-spherical/faceted shape of 15-18 nm have SAR values that are at least two times higher than the corresponding ones of similar size obtained by Salas et al 74 . Finally, spherical MNPs with a diameter less than 12 nm were likewise characterized, but they had very limited SAR values at any field conditions. It is worth noting that, even for the syntheses of quasi-faceted and spherical NPs, the mass amount of magnetic materials produced is between 0.1 g and 0.7 g per 25 mL autoclave (Table 6) and correspond to 1-7 g per thermalyzation cycle when using ten autoclaves in parallel in the same oven.

Results from the experiments using different amines
Our protocol can be adapted to obtain Fe 3 O 4 star-shaped NPs. We have compared the SAR values of star-like IONCs of 16 nm with regular cubic IONCs of similar size (14 nm) using AC magnetometer (f = 100 kHz and H = 12, 16 or 24 kA/m). Star-like NPs with SAR values of 120-240 W/gFe at clinically relevant AMF conditions (100 kHz and 12-24 kA/m) have higher SAR values than IONCs of similar size (Fig. 8b). The same trend was observed at higher frequency (200 kHz, Supplementary  Fig. 15,~20% increase at H = 20 kA/m and~40% increase at H = 24 kA/m), achieving SAR values as high as 950 W/g Fe at f = 200 kHz and H = 24 kA/m).

Conclusions
In this work, we report a gram-scale solvothermal protocol to prepare highly monodisperse NPs with peculiar shapes including NCs, nanostars and nano-octahedra that display benchmark specific absorption rates for MHT applications.
We have discovered the crucial role of aldehyde molecules as shape-directing agents for the control of high-quality ferrite NPs while maintaining size, size distribution and crystallinity control.
To promote the growth of NCs, the use of benzaldehyde, its derivates or a subclass of aromatic aldehydes at peculiar molecular structure (that is, having the formyl and the phenyl ring in resonant via the carbon-carbon double bond) must be used. The transition from NCs towards faceted NPs occurs when choosing aldehyde molecules having the phenyl ring and the formyl group separated by one or two sp 3 carbons. Instead, the use of bare aliphatic aldehydes enables the synthesis of highly monodispersed spherical NPs. Moreover, nanostars, having modified NC shapes with concave facets and sharper edges were obtained by simply replacing hexadecylamine with tertiary amines (such as trioctylamine or tridodecylamine) or with secondary amines or with oleylamine in the NC synthesis protocol. SAR values for the NCs produced by our solvothermal method are very similar or even superior to those obtained by thermal decomposition methods. The SAR values could be tailored by tuning the NP size (in the range between 8 and 42 nm), the shape (NCs, nanostars, nano-octahedra or nanospheres) and the composition (Fe 3 O 4 , Mn 0.6 Fe 2.4 O 4 and Zn 0.6 Fe 2.4 O 4 ). In particular, the nanostars and the NCs heat more than the truncated octahedra, which in turn, heat more than the spherical NPs. For Hf factor of 3.6-3.8 ×10 9 A/ms, the following order in terms of heat performance can be established: 20 ± 1 nm nano-cubes (480 W/g Fe ) >20 ± 3 nm nano-octahedra (300 W/g Fe ) >nano-spheres (below 300 W/g Fe for sizes of 15-18 nm and below 200 W/g Fe for 8-12 nm). Peculiarly at Hf >4.0 × 10 9 A/ms, the nanostar shapes performed even better than the NC shapes of fairly similar size.
Finally, as demonstrated, our protocol is easily scaled by performing multiple and parallel reactions for each oven cycle to obtain tens of grams of high-quality MNPs with unmatched MHT performance or by employing a Parr reactor, which provides a higher iron conversion than the autoclave solvothermal method. To note that, recently, in an in vitro study, NCs produced with the protocol here reported and capped with PEG-GA, were tested on U87 glioblastoma cells, and clearly indicate that they possess optimal colloidal stability and a low-toxicity profile in the absence of magnetic hyperthermia actuation 75 . Indeed, for IONPs, in which cytotoxicity is dictated mostly by the surface-capping agents, the choice of water-soluble ligands that are non-toxic, such as the PEG-GA ligand, enable the NPs toxicity to be tuned in a well-controlled way.
The gram-scale production of such high-quality NPs could be attractive not only for the medical field, but we also envision their exploitation in other applications such as MHT-mediated catalysis 76,77 , batteries 78 , photovoltaics and electronic devices, where high-heat performing NPs can make the difference 79 .
In future, the replacement of iron pentacarbonyl with other iron precursors may be considered to make the NC synthesis more eco-friendly.

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.