Low pressure bottom-up synthesis of metal@oxide and oxide nanoparticles: control of structure and functional properties

Experimental activity on core@shell, metal@oxide, and oxide nanoparticles (NPs) grown with physical synthesis, and more specifically by low pressure gas aggregation sources (LPGAS) is reviewed, through a selection of examples encompassing some potential applications in nanotechnology. After an introduction to the applications of NPs, a brief description of the main characteristics of the growth process of clusters and NPs in LPGAS is given. Thereafter, some relevant case studies are reported: • Formation of native oxide shells around the metal cores in core@shell NPs. • Experimental efforts to obtain magnetic stabilization in magnetic core@shell NPs by controlling their structure and morphology. • Recent advancements in NP source design and new techniques of co-deposition, with relevant results in the realization of NPs with a greater variety of functionalities. • Recent results on reducible oxide NPs, with potentialities in nanocatalysis, energy storage, and other applications. Although this list is far from being exhaustive, the aim of the authors is to provide the reader a descriptive glimpse into the physics behind the growth and studies of low pressure gas-phase synthesized NPs, with their ever-growing potentialities for the rational design of new functional materials.


Introduction
Nanotechnology is now part of our everyday life. From electronic goods, to medicine and diagnostics, to food processing and packaging, nanomaterials are widely employed. Out of the many examples, it is possible to single out three types of commercial products employing nanoparticles (NPs): quantum dot (QD) televisions, the first types of food packages employing Ag NPs for antimicrobial purposes, and the contrast agents employed in magnetic resonance imaging, composed of superparamagnetic iron oxide (SPIO) NPs for cancer diagnostic.
In some commercial television screens, NPs of CdSe and other types of QDs are employed as filters in the backlighting units of liquid crystal displays (LCDs). It is well known that in colloidal NPs like CdS and CdSe the optical properties of QDs are size dependent: emission and absorption depends on the size d because of quantum size effects [1]. This means that the color of the emitted and absorbed light in a QD can be finely tuned with the proper choice of d, in a range approximately between 2-7 nm (see figure 1). In commercial QD televisions, radiation from a blue light emitting diode (LED) is transformed by QDs into pure red and green, and the combination of blue, green, and red light inhibits absorption of the extra tonalities by the filters behind the LCD screen [2]. In this way there is a substantial increase in light throughput and a significant improvement in color gamut. It must be noted, however, that QD TVs are still essentially based on LCD technology. The use of QDs as LEDs, or Q-LEDs, in displays is still in the experimental stage, although they may potentially be an alternative to the established organic LED displays [3]. These types of NPs are employed in medical imaging (with proper capping and functionalization to make them biocompatible [4,5]) and, more recently in TV screens. Still, the use of metals like Cd in these types of devices has some limitations because of health issues. For example, the European Union Restriction of Hazardous Substances Directive imposes strict limits on the use of these types of metals in consumer electronics.
The use of silver as antibacterial agent dates back in history. A common method to prolong the freshness of milk was to place a silver spoon or coin at the bottom of the milk container [6].
Recently the food-packaging industry in some countries has introduced products employing nanocomposite material, with nano-silver in containers and coated films. However, these developments raise some questions regarding the possibility of migration of the engineered nanoparticle particles (ENPs) into food and its potential toxicity (see figure 2) [7]. The investigation of the changes occurring to the physical and chemical properties of materials when passing from the micro-to nanoscale are therefore of paramount importance in addressing the many technological and safety challenges posed by the introduction of nanotechnology in the food processing and packaging industry [8].
A third example that is worth mentioning is taken from the field of medicine and diagnostics, in particular the employment of ENPs as a contrasting agent in magnetic resonance imaging (MRI) for the diagnosis of cancer in liver and other internal organs [8]. superparamagnetic iron oxide nanoparticles (SPIONs) have been used during the last 30 years in MRI for the detection of focal liver and spleen lesions. Their diameter is generally approximately 30 nm, but more recently ultrasmall SPIONs (U-SPIONs) have been in use, with a significantly smaller size. SPIONs are typically composed of magnetite (Fe 3 O 4 ) or maghemite (γ−Fe 2 O 4 ) single crystals. Other NPs that can be potentially used in MRI are Co, core@shell Au@Fe, and Pt@Fe. The possibility of producing core@shell NPs has extended the potentiality of applications in this field [9], with significant enhancement in diagnostic and therapeutic capabilities.
The huge advancements in NP synthesis and functionalization have allowed their use in many industrial products, but the requirements for the control of their properties, especially regarding their effects on human health and the environment are becoming more and more stringent. The crucial problem, as mentioned before, is the understanding of the modification of the physical and chemical properties of matter when passing from the micro-to the nanoscale. The . The particle diameters are respectively 2.1, 2.5, 2.9, 4.7, and 7.5 nm, going from blue to red. Reproduced from [4] with permission of the Royal Society of Chemistry. effects on the structure of metal NPs can be given as an example. It is well known that in small, stable metal clusters 4 with d of a few nm, atoms are arranged in closed geometric shells [10]. In fact, clusters do not possess translational symmetric, like 'bulk' crystals. However, in a conceptual bottom-up process, we can imagine building the particle by forming polyhedral concentric shells of atoms, in a sort of 'shell periodicity'; the concentric layers of atoms are added in a fashion that the overall symmetry of the clusters does not change. With this approach, when the shell of atoms is complete, the cluster has a stable configuration because the average atomic coordination number is maximized. Experimentally, this gives rise to the occurrence of 'magic numbers', i.e. sets of numbers of atoms for which clusters are energetically stable. For instance, in alkali metal clusters, the magic numbers are those obtainable by completing K polyhedral shells, based on 12-vertex icosahedra. This results in particles in the shape of icosahedrons (see figure 3), Ino decahedron, and face centered cubic cuboctahedron [11]. With increasing size, these types of structures can become metastable (i.e. they do not correspond to the absolute minimum of the free energy), but they can still be observed if the NPs are generated using gas-phase physical synthesis processes [12][13][14][15]. In a recent wide angle x-ray scattering experiment with pulsed free electron laser radiation, scattering patterns from in-flight single Ag NPs could be detected [16]. The most relevant result was the observation of patterns typical of metastable NPs, even for very large sizes (≈300 nm, N≈10 9 atoms). It can be predicted that these states (in many cases with icosahedral and decahedral shapes) show a phase transition to fcc-derived structures. Their occurrence for larger size NPs (with N≈10 6 atoms) was ascribed to a non-equilibrium growth process in the NP source from small seeds. Large NPs with a highly oblate shape could also be singled out, suggesting that the seed structure 'locks' the growth mode, in analogy to the formation of snowflakes.
The morphology and the surface structure of NPs can influence the properties of functional molecules absorbed on the exposed facets, or can determine some interfacial effects, like exchange bias (EB) occurring in ferromagnetic-antiferromagnetic (FM-AF) material interfaces, like in the case of core@shell configurations [17], so new architectures and functionalities can be designed. In most of applications, NPs are produced by chemical or green synthesis. These methods are generally scalable for industrial production. Nevertheless, because of their commercial applications and potential extended use in many fields, characterization and in-depth investigation of the 'case study' of NPs are necessary for the fine-tuning of the desired properties. Physical synthesis, bottom-up techniques have the advantage of 'reduced' complications due to absence of chemical by-products and ligands, and the possibility of onestep processing. The bottom-up approach allows for the precise control of the formation of NPs and of their characteristics. The aim of this review is to focus on some specific examples of growth and the investigation of low pressure gas-phase synthesized NPs, in order to demonstrate the potential of this research in realizing significant advances in nanotechnology. The paper is organized as follows: a brief summary of the principal techniques used for gas-phase synthesis of NPs will be given. Successively, some selected experiments, with the realization and study of metal@oxide and oxide NPs will be described, with an emphasis on the interplay between their structure and functional properties: tuning of EB in FM/AF systems, reversible reducibility, and oxidation in oxide NPs. Finally, some conclusions and future perspectives will be given.

Physical synthesis of NPs using a bottom-up approach
Physical synthesis of NPs using a bottom-up approach is a long tradition in the physics and chemistry of materials. The basic principle is to form small aggregates of atoms that can grow by attachments of single atoms, coagulation, or coalescence. The resulting clusters can be free particles in a 'cloud', in a collimated beam, or they can be adsorbed on a support. A variety of clusters and NP sources have been successfully developed. It is worth noting the seeded supersonic cluster source, the laser ablation source [18,19], the pulsed-arc cluster ion source, and pulsed microplasma cluster source [18][19][20]. These sources generally operate with a low pressure gas carrier. Recently, great advancements have been obtained in gas-phase synthesis with other methods, like flame-based synthesis [21] and non-thermal plasma based synthesis [22][23][24][25], which have been capable of producing core-shell NPs [26] and spark discharge-based synthesis [27], extending the possibilities in realizing scalable physical  4 We will adopt the term 'cluster' for agglomerate of atoms with size N<500, while the term 'nanoparticle' will be used for bigger size values. This choice is of course completely arbitrary, but it is convenient when the first stage of the growth of NPs is discussed. Moreover, many sources are also called cluster or nanocluster sources. synthesis methods for producing nanomaterials. In this review we focus our attention on the experiments performed with low pressure gas aggregation sources (LPGAS), like the one shown in figure 4 [28,29].
In LPGAS, atoms are generally evaporated from a furnace, or by sputtering from a target. The vapor is cooled by the interaction of atoms of a noble gas (typically Ar) at low temperature, and clusters start to grow in the so-called gas aggregation region (condensation region in figure 4). The noble gas acts also as a gas carrier, and the beam of clusters/NPs exit the gas aggregation region through an aperture into a second chamber with a higher vacuum. The NP/gas carrier beam can therefore be formed because of the differential pressure between the gas aggregation region and the second chamber. In a variant method, a pulsed laser is used to ablate material from a target and to inject it in a pulsed gas stream. This arrangement produces a pulsed cluster beam [30]. Another possibility of growing NPs can occur when a beam of atoms from a thermal source impinges on a surface. Nano-islands can grow spontaneously or after thermal treatment. In this case the NPs will be formed directly onto a substrate [31][32][33].
We now provide a more detailed description of sputter gas aggregation sources, based on the original design of Haberland [34][35][36], which is commonly used by many groups. As mentioned above, metal atoms are evaporated by the sputtering of a target using a D.C. magnetron (see figure 5(a)). A magnetron consists essentially of: • A cylindrical electrode (cathode) on which a disk-shaped target of the material is attached. The cathode is set to a negative voltage of few hundred volts (100-300 V). • An anode is set to ground, in the shape of a ring surrounding the cathode. The distance between the two electrodes is about 0.3-0.5 mm. Both electrodes are in vacuum. • A permanent magnet positioned behind the magnetron.
The system is also cooled, either with liquid nitrogen or water. Under operating conditions, Ar gas flows between the electrodes, giving rise to a glow discharge. Ar atoms get positively ionized and bombard the target, ejecting atoms and electrons out of the material. The electrons giving origin to the discharge are confined by the magnetic field to a region where magnetic field lines are parallel to the target surface. In this way the sputtering rate is enhanced, as more Ar atoms are positively ionized. The glow discharge, because of the magnetic field line distribution, has an annular shape, and the target gets eroded in a circular region. The sputtered atoms can be charged or neutral. While the neutral atoms escape the glow discharge region and form the clusters, the positively ionized sputtered atoms are deposited again on the target surface. After leaving the discharge regions, the small clusters  [28,29]. Reproduced from [29], Copyright 2001, with permission from Elsevier. start forming and increase in size, as they travel along the aggregation region with the gas carrier, which can be composed of Ar or a mixture of different species: He, Ne, or N 2 but also O 2 , depending on the type of NP to be realized (see figure 5(b)). It is generally accepted that condensation starts with the formation of dimers and it goes on by adding single neutral metal atoms [37][38][39][40]: However, in a recent paper by Haberland [36], it was argued that condensation starts first with the formation of Ar , 2 + which is given by

Ar
Ar Ar Ar Ar 2 + + -> + + + and then it continues with subsequent attachment of M atoms, although a detailed description of the initial stages of aggregation remains unknown. Also, some dimers can be formed by a single sputtering process, and can act as condensation germs. Once the homogeneous nucleation has lead to the formation of 'embryos', condensation occurs and (meta)stable NPs are formed, which can grow by single atoms attachment, coalescence, or coagulation. As mentioned above, the growth kinetics determines the structure and morphology of the resulting NPs. Up to now, we have summarized the process leading to the formation of metal NPs by using DC magnetron sputtering. However, RF magnetron sputtering can be used in the case of NPs composed of non-conductive materials. The magnetron sputtering gas aggregation source can produce beams of NPs with a very extended size range, from a few (d to less than 1 nm) to several hundreds of thousands of atoms, corresponding to a d greater than 30 nm. Also, the size distribution can be quite narrow, with a full width half maximum in the size distribution mass of about 20%-30%. The shape of the size distribution is generally well represented by the log-normal distribution, which was theoretically predicted by simulation of the aggregation processes using the Monte Carlo technique [41,42]. The experimental conditions affecting the characteristics of the NPs generated by the magnetron sputtering gas aggregation source are determined by the following parameters: • Ar gas flow in the discharge region and in the aggregation regions, which can be adjusted with EB flow controllers or with adjustable leak valves. • Discharge power.
• Length of the gas aggregation tube, i.e. the travel length of the cluster during the process of aggregation and growth.
Additionally, a mixture of gas in the aggregation region can affect the size distribution: for instance, He is used for producing smaller clusters and NPs. In some experiments, O 2 flow has been used to generate oxide NPs [43,44].
Once the NP beam exits the LPGAS, the size distribution can be analyzed with mass spectra. Quadrupole mass spectrometers/filters (QMF) or time of flight spectrometers (TOFS) are generally used [29]. The QMF can select and measure the ionized NPs, which can be present in the beam in a consistent fraction (for some instruments, up to 80%). After this stage, the NPs can be deposited on a support and investigated with a variety of techniques, in situ or ex situ. In addition, if the NPs are electrically charged, their impact energy can be varied with the help of an electrostatic field, for instance by setting the support at high voltage. Some important studies on the effect of the impact of the energy on the NP morphology and structure when deposited on a substrate have been performed experimentally [34,35] and with molecular dynamics simulations on Mo clusters deposited on Cu and Mo substrates [45], revealing three different regimes: • If the kinetic energy of the cluster just before the impact onto the surface is E k~0 .1 eV/atom (low energy regime), the cluster structure is only slightly deformed (soft landing). • If E K lies between 1-10 eV/atom (medium energy), the morphology of the cluster can change and there may be some induced defects on the support. • If E k >10 eV/atom the cluster is disrupted, the substrate is damaged in the impact area, and the damage can extend down to several atomic layers.
During normal operation, the magnetron based LPGAS generate beams of NPs at low energies, although some deformation of NPs after landing was observed. This change in shape (a sort of slight 'flattening') was ascribed to electrostatic interaction between the NPs and the substrate [46].
There are additional possibilities for more complex NP architectures, finalized to the realization of core@shell, dumb-bell geometries, nanoalloys etc [47][48][49][50]. Also, chemisorption of functional molecules on in-flight NPs or growth of NPs in solid matrix and/or multilayered NP films are possible, with some implementations of the equipment. Some examples will be discussed in the following section. In figure 6, a sketch of a typical experimental apparatus equipped with a NP source is shown. The system, designed and used by our group [15,46], consists of four interconnected ultrahigh vacuum chambers: • The NP source and the QMF, where the NP beam is generated and mass-filtered. • The deposition chamber where the substrate is inserted through a fast entry arm and positioned across the NP beam. In this chamber there are also evaporators for capping and/or co-deposition treatment, a gas-line for letting in O 2 and other gas species, a quartz microbalance for monitoring the amount of the deposited material, and an oven for thermal treatment. • Two analysis chambers where the NP film deposited on the substrate is transferred for x-ray photoelectron spectroscopy (XPS) and auger electron spectroscopy analysis.
XPS is mostly used, for an in situ chemical analysis of the produced NP film. Other techniques are employed ex situ; for instance scanning electron microscopy (SEM), atomic force microscopy, transmission electron microscopy (TEM), high-resolution TEM (HRTEM), and scanning TEM, to determine the structure and morphology, but a great number of experiments can be performed, depending on the properties and on the functions of the NPs that need to be investigated.

Native oxide shells in metal NPs
The investigation of metal clusters and NPs produced by low pressure gas-phase synthesis is a vast subject that has been carried out over the last five decades. Structure, electronic, magnetic, and optical properties have been studied in-depth. The reader is referred to some recent books and excellent reviews [29,[51][52][53]. This section is dedicated to the activities on core@shell NP, as this configuration presents interesting opportunities for nanomaterial functionalities. The simplest way to obtain a shell around metal NPs is to expose them to atmosphere. The surface of the NPs will be oxidized, and the result will be NPs with a metallic core and a shell of native oxide. In a strict sense, most of the metal NPs exposed to atmosphere (even for a short time) while they are transferred to a different apparatus, like for instance a TEM, will have a shell of native oxide; some notable exceptions are Au NP, or core@shell NP with an Au shell [40]. The thickness of the native oxide shell will depend on different factors, like the exposure time and the NP reactivity to air exposure. The shell can modify the original properties of the NPs. The structure and physical properties of the external layer will change, as a new material is now formed, as well as the NP morphology and size. Some specific examples of this effect have been studied on Fe [54], Co [55], FePt [56], FeCo [57] and Ni [15] NPs. In the case of Fe [54], the TEM image shows a regular shell, with a thickness t≈2 nm over a core diameter of size d≈5 nm (see figure 7).
In the Co NPs of diameter d≈8 nm, a value of t≈1 nm was obtained for the native CoO. For smaller sizes (d≈2 nm) the NPs were reported to be completely oxidized [55]. The effect of exposure to atmosphere and controlled oxidation of Ni NPs of size d between 4-8 nm has been systematically investigated with XPS and TEM [15]. The Ni NPs were deposited on inert substrates (namely, Si wafers with their native oxide Si/SiO x and HOPG), and were oxidized with different procedures: co-deposition in the presence of O 2 , exposure to O 2 after deposition, exposure to atmosphere for 12 h, and exposure for 20 d. The treated NP assemblies were then analyzed with XPS (see figure 8(a)), focusing on the photoelectron spectra from Ni 2p core levels. The spectra were deconvolved in two chemicalshifted components, corresponding to the emission of electrons from Ni in a metallic state and from Ni in an oxidized state (see figure 8(b)). The relative intensities of the two components Figure 6. Sketch of the experimental apparatus used in the work described in [15,46], with the NP source and QMF filter, deposition chamber and XPS and Auger electron spectroscopy chambers for NP film analysis. allowed estimating the thickness of the oxide shell of no more than a few atomic layers after 12 h of exposure. Moreover, a metallic Ni core was still present after more than 20 d of air exposure.
Similar to many other metal NPs realized with low pressure gas-phase synthesis, HRTEM images taken from a Ni NP deposited on TEM grids revealed a multitwinned icosahedral structure for the metallic core [12][13][14] (see figure 9). Local strain analysis showed also deformation of the peripheral Ni region which can be associated with the growth of epitaxial NiO crystalline islands on the NP surface. The overall behavior was consistent with an an image often given by surface oxidation of metals consisting of three phases: The velocity of the third process depends strongly on the temperature and is hampered by the limited diffusion of the nickel cations through the oxide film. For temperatures below 500-600 K [58,59], like in this case, the number of oxide layers that can be actually grown is limited. After their formation, they passivate the metallic nickel surface and stop the increase of the oxide film thickness.
A more recent investigation on the process of Fe NP oxidation was carried out with aberration-corrected STEM [60]. In this case, cuboid-shaped Fe NPs of width 5-20 nm were gas-phase synthesized and imaged after prolonged exposure times to air, even after 2 yr (see figure 10). A coreshell geometry was observed after 6 months of deposition, with a shell showing an inverse spinel structure composed by an Fe 3 O4 and γ-Fe 2 O 3 phase. Interestingly, no equilibrium oxide thickness (which was estimated to be 2-3 nm in the Cabrera-Mott theory of metal oxidation [61]) is reached. After approximately 2 yr, the particles were fully oxidized. The strain analysis of the STEM images explained the complete oxidation (and the presence of Kinkerdall voids at the center of the NP [62]), see figure 10(d), as due to positive strain gradients leading to increased ionic conductivity. As a result, there is diffusion of cations towards the NP surface, occurring by the exchange of vacancies, and a diffusion of anion through the interstitial sites. The oxidation process in these NPs does not stop, as in the case of bulk metals. The reason given by the authors is the impossibility of creating dislocations that allow relaxation of the elastic strain at the metal-oxide interface, which prevents the oxidation from progressing further. The different behavior between the two reported cases (Ni and Fe NP) can be probably ascribed to the different NP shape and structure.  [15]. © IOP Publishing Ltd. All rights reserved.

Magnetic properties of metal@oxide NP: EB
As already mentioned, the creation of an oxide shell can have major effects on the electronic and magnetic properties of the NP. A typical example is given by metal@oxide NP with a ferromagnetic (FM) core and an antiferromagnetic (AF) shell, like in Co@CoO and Ni@NiO NPs. It is well known that FM bulk materials show magnetic domains, as a consequence of energy minimization [63]. In fact, the competition between magnetostatic energy and the energy associated with the formation of domain walls favors the presence of domains. Nevertheless, when the finite size of the NP decreases below a certain limit, formation of multiple domains is not energetically convenient anymore, as the total energy must be minimized, including the magnetostatic energy, the exchange, and the anisotropic terms as well as the domain contribution. In this case the NP can exhibit a single magnetic domain, with all the atomic spins oriented along a certain preferred direction, depending on particle shape and anisotropy. The NP can be considered in this case as a single 'giant' spin. The transition from multi-domain to the single-domain regime in small particles has been studied theoretically by Kittel [64]. He estimated a value for the critical size of a small particle of the order of 10 nm. It was found that the critical size can vary between 20-800 nm. This value depends on the energies of anisotropy and of exchange. In Fe spherical NPs, for instance, the transition from multi-to single-domain occurs with particles of diameter of about 30 nm, while for Co the critical size is typically 35 nm [65]. As a consequence, an assembly of singledomain NP can be conceived as an ensemble of magnetic dipoles, and inside the NP the atomic spins will be subject to coherent rotation. The giant spin of a free NP will depend, of course on the NP size, but also on its shape and structure. A series of experiments on beams of clusters/NPs have been performed by a number of groups, aimed at the measurements of their magnetic moment [66]. Some of these works followed the conceptual idea of the Stern-Gerlach experiments, i.e. a beam of mass selected clusters or small NPs were deflected by the force exerted by the gradient of a magnetic field. The deflection was recorded generally with a time of flight (TOF) mass spectrometer. By knowing the field gradient dB/dz and other quantities like the length of the magnet L and the length of the flight tube between the magnet and the mass spectrometer D, the magnetic moment per atom, m, of the sizeselected clusters could be obtained [67][68][69][70][71][72][73][74][75]. Another important technique used for determining the magnetic moment per atom of free and supported NP is x-ray magnetic circular dichroism in absorption spectra (XMCD), where synchrotron radiation x-rays are employed [76][77][78]. The knowledge of NP magnetic properties is crucial for the design of advanced magnetic memory, in particular regarding the requirement of high density storage devices. To this purpose, the miniaturization of a single element of memory can be hampered by the loss of magnetic stability due to thermal fluctuations [79].
In a small magnetic particle, the existence of a certain uniaxial magnetic anisotropy will give rise to a preferred orientation of its magnetic moment along an easy axis of magnetization. In order to reverse the orientation of this moment, an energy barrier (called anisotropy energy) ΔE=K a V has to be overcome, where K a is the anisotropy constant and V is the particle volume. When V is reduced, the energy barrier is lowered, and the NP thermal energy K b T can give rise to fluctuations that are increasingly frequent that can overcome the anisotropy barrier. In this case the NP magnetic moment will no longer be blocked, and an assembly of NPs showing this behavior will be in the superparamagnetic (SPM) phase (see figure 11). On the contrary, if the NPs remain stable during a certain measurement time, t m , it is said that the NPs are blocked. The blocking temperature, conventionally defined by the formula T B =K a V/(25K B ), corresponds to the temperature of transition from blocking to superparamagnetic phase [65]. T B therefore depends on the product between the NP volume and the anisotropy constant. For Ni, given an anisotropy constant K=0.5·10 4 J m −3 , it can be obtained that T B is equal to room temperature (RT) for a NP volume V crit =2.1·10 4 nm 3 . If the NP is assumed to have a spherical shape, this volume corresponds to a diameter d≈34 nm. The 'superparamagnetic limit' can be therefore a serious obstacle to design and realization of ultrahigh density magnetic memories [79].
In recent years, the possibility of obtaining stable magnetic NPs has been investigated by exploiting the EB effect at the interface between a FM and an AF material. In the case of NPs, this effect would be obtained by realizing a core@shell configuration, where the core is FM and the shell is AF. Co@CoO, Ni@NiO and Ni@CoO NPs should present this effect, as NiO and CoO are both AF materials. The origin and the explanation of the EB effect still present some difficulties [80]. A first, simple explanation for the EB effect occurring at FM/AF planar interface, like in the case of films and multilayers, lies in the additional torque exerted by the exchange coupling between the FM and AF spins when there is a switching of the external magnetic field [80] (see figure 12). Let us suppose that the Néel temperature of the AF material, T N , is lower than the Curie temperature, T C , of the FM material, T N <T C . When the system is cooled below T C in the presence of a magnetic field H with a large enough intensity, all the spins of the FM material are aligned along H. If the temperature is further cooled down below T N , the spin interaction at the FM/AF interface forces the first layer of the spins of the AF material to be aligned along the direction parallel to the spins of the FM medium. This will give rise to an AF ordering with spins parallel or antiparallel to the magnetization M of the FM. The FM/AF spin interface coupling gives rise to an extra torque exerted on the FM spins, which has to be overcome by H. On the basis of these simple considerations, if the AF/FM coupling is strong enough, there is a rigid shift in the hysteresis loop, while in the case of small AF/FM interaction the result is an increase in the coercition field, i.e., there is some hardening of the magnetic material. Very often both phenomena occur because of inhomogeneity of the AF anisotropy due to grain size distribution and structural defects. EB can therefore lead to the stabilization of magnetic NPs, if the magnetic core is either surrounded by and AF shell or embedded in an AF medium. This behavior was directly verified on chemically and physically synthesized core-shell NPs. In particular, low pressure gas-phase synthesized NPs permitted great flexibility in realizing oxide shells, either by simple oxidation [15,81], or by co-deposition of the NP core and shell [82]. Moreover shells of different oxide species could be realized [48,82]. For instance, in a seminal work [83] assemblies of Co@CoO NPs embedded in an AF matrix (CoO) or paramagnetic (Al 2 O 3 ) matrix were successfully produced with LPGAS and a conventional sputtering apparatus; structure, composition, and magnetic properties were investigated with HRTEM and with Superconducting QUantum Interference Device (SQUID) magnetometry, obtaining field cooled/zero field cooled (FC/ZFC) curves and hysteresis cycles at different temperature values (see figure 13).
The results demonstrated that FM NPs embedded in an AF matrix can lead to magnetic stabilization and a substantial increase of T B , even for Co NPs as small as d=3 nm.
Similar studies were carried out by the same group on Co@CoO NPs in Al 2 O 3 , with increasing NP density, revealing a significant increase of T B , of the coercition and of the bias fields [84]. This effect was ascribed to the 'recovery' of the AF properties in the CoO shells when they came into contact, rather than to the interaction between the NP cores. Ni@NiO NPs present some intriguing, different properties with respect to Co@CoO. Although the FM anisotropy constant of Ni (K 1 =−5×10 3 J m −3 ) and the AF anisotropy constants of NiO (K=−500 J m −3 ) are significantly smaller Figure 11. Onset of superparamagnetism in a single-domain NP. If the diameter of the NP d is sufficiently small, the anisotropy energy E a =KV will be comparable to the thermal energy and the particle will start to oscillate at high frequency between the two energy minima, resulting in magnetic instability. Picture reproduced courtesy of A Ponti. than the ones of Co (K 1 =45×10 4 J m −3 ) and of CoO (K = 2.7×10 7 J m -3 ), for CoO T N =291 K, a value which is lower than the one of Ni (T N =525 K), with potential problems in magnetic devices operating at room temperature. Several studies have been carried out on Ni@NiO NPs [85][86][87]. The results on EB and stabilization are sometime controversial. This puzzling behavior can be ascribed to the differences in NP synthesis procedures, which can influence many factors (presence of interface defects and/or dislocations, crystal quality in core, and shell, etc). In a recent comparative work, the oxide shell was formed by using different methods, namely exposition to oxygen after NP growth and deposition on substrates, annealing in atmosphere, and the 'sequential layer deposition' procedure. [88,89]. The sequential deposition method consisted of three phases: (1) A NiO layer was grown on Si/SiO x , by means of thermal evaporation of atomic Ni in the presence of oxygen.
(2) A beam of metal NPs was generated by LPGAS and impinged on the ultrathin NiO layer. (3) An additional oxide film was deposited on the Ni NP assembly, with the same method used in step 1.
This procedure permits the growth of oxide shells with variable thicknesses, as the Ni core diameter remains fixed, allowing in this way considerable simplification of the analysis. The morphology of the samples prepared with sequential layer deposition was analyzed with STEM in high angle annular dark field (HAADF) mode and HRTEM from NP assemblies deposited on carbon-coated copper grids. From the image of the NP shown in figure 14(a), there is evidence of a core-shell morphology, where the shell can be assigned to NiO, while the core, which from the contrast appears heavier, is composed of Ni. It is deduced that NiO grows as a shell around the Ni NP synthesized by LPGAS. The STEM-HAADF image has a shape compatible with a McKay icosahedron. The icosahedral structure is due to multiple twinning of Ni small single crystal exposing (111) facets, a common phenomenon occurring in the fcc metal NP (see also figure 9).
The evolution of the NP film morphology with the increasing thickness of the NiO overlayer was investigated with SEM. As the thickness of the NiO overlayer increases, the NP gets bigger, with the presence of small structures that can be ascribed to the nucleation of the NiO islands. When the thickness of the NiO overlayer reaches the value of t NiO =6.5 nm, grains of size up to 60 nm are observed ( figure 14(b)), as a fractured film is formed. Therefore, it can be concluded that at increasing thickness of the overlayer, the NiO shells around the Ni NPs extend progressively, and eventually the Ni NP cores are embedded in a solid matrix. A closer inspection at the core-shell interface structure has been made possible by STEM and HRTEM, comparing two possible ways of obtaining the NiO: annealing in air of the metallic NP at T=250°C for t=30 min and sequential layer deposition. In synthesis, in the first case, the shell has a structure of crystalline NiO, growing in the (111) direction. A more detailed study of the core-shell interface with the geometrical phase analysis of the STEM images evidences the formation of dislocation that probably relaxes the stress in the interface region caused by the lattice mismatch between the crystalline fcc Ni and rock-salt NiO (111) planes, favoring crystal ordering in the NiO shells. This relaxation and the formation of dislocations can be ascribed to thermal annealing. In the case of the samples prepared with the sequential layer deposition, the interface sharpness is reduced and the shell atomic structure is less ordered. The exit wave reconstruction analysis of HRTEM images in some parts of the shell shows atomic fringes corresponding to a [243] direction for the NiO growth, different from the original (200) NiO planes observed for the annealed sample (see figure 14). In this way, the 3:4 NiO/Ni interface coincidence observed by TEM, which requires an expansion of 11% in the NiO lattice, can be explained. The reduced sharpness compared to the annealed sample is ascribed to the two different procedures for the realization of NiO shells. Nevertheless, with sequential layer deposition it is possible to change the shell thickness without varying the core diameter, with a thorough analysis of the effect of the NiO shell thickness on the NP magnetic properties. In practice, the EB effect, which is indeed observed with SQUID magnetometry, can be finely tuned by adjusting the shell thickness as a single parameter.
The FC/ZFC magnetization curves are reported in figure 15. The difference between M FC and M ZFC decreases upon increasing T, because the NP magnetic moment gains enough thermal energy to overcome the barriers due to magnetic anisotropy, and tend to align parallel to H. When M FC −M ZFC =0, the NPs pass to the SPM (reversible) phase. At the corresponding temperature T max , M ZFC is maximum, so that T max can be considered to be an estimate of T B . However, the behavior of the curves for low values of the shell thickness at high temperatures (200-300 K) suggest that the superparamagnetic NPs are coupled by exchange and magnetostatic interactions. Finally, when the NiO form a film in which the Ni NP cores are embedded, a larger T max is displayed. The reason of this behavior, in analogy with what observed on Co@CoO NP, is a stronger FM/AF interface coupling, together with a higher magnetic anisotropy, and a blocking temperature T B >RT, causing a reduction in the effect of the magnetostatic inter-NP interactions. A small EB field, H bias , is measured up to shell thickness t s =1.6 nm, but when t s increases, H bias is greater, with a steep increase and no sign of saturation (see figure 16). The maximum value of the EB field is H bias =0.57 kOe. The samples grown with sequential layer deposition have a lower H bias when their shell has a thickness similar to the one obtained for the annealed sample. The differences in the core/shell interface quality could be the cause of this result. On the other hand, with sequential layer deposition is possible to tune EB by simply adjusting the NiO overlayer thickness, i.e. the shell thickness keeping the NP core diameter fixed.
As mentioned before, physical synthesis of NPs with LPGAS and some auxiliary methods (like evaporation from other sources) allows creating different species of oxide shells and matrices [48,82]. Relevant results were recently obtained in the case of Co@CoO NP (Co core diameter d≈5 nm) embedded in an NiO matrix [90]. The CoO shells were obtained during the NP synthesis in LPGAS, while the NiO matrix (and Nb for a non-magnetic matrix in the reference samples) were realized with RF sputtering. SQUID magnetometry measurements revealed an important increase of the blocking temperature T B well beyond room temperature, in one case above T=400 K, in contrast with the Co@CoO embedded in Nb, where T B could reach a maximum value of 70 K. The Co@CoO/NiO samples also showed a relevant value of μ 0 H E =14 mT and some coercivity at room temperature. The outstanding enhancement of T B is explained as due to a combined magnetic effect involving both the CoO shell and NiO matrix, in particular the combination of the CoO high anisotropy and of the high Néel temperature of NiO (through a proximity effect), providing in this way a sufficiently high AF anisotropy at the core-shell anisotropy to enhance the effective anisotropy of the Co cores. It is worth mentioning some preliminary studies obtained by the 'reversed' system, in particular Ni@CoO NPs obtained with the sequential layer deposition method, previously described [91]. In situ XPS analysis revealed the existence of an interfacial NiO layer obtained during the growth procedure, at variance with observations on Ni@MgO [82] systems, synthesized with a similar method. A possible reason for this different behavior may lie in the electronegativity of Mg (1.31 Pauling), which is lower compared to the value for Ni (1.91 Pauling), so oxidation of Mg is favored. In a previous work [92] Ni@CoO NPs were synthesized with a physical method: evaporation of Ni and laser ablation of CoO with condensation by means of Ar [92], and also in that case a thin interfacial NiO layer was obtained. The shell around the Ni NPs was composed of Co 3 O 4 . EB was observed up to T=45 K, a temperature close to the Néel temperature of Co 3 O 4 . Since the electronegativity values of Ni and Co (1.88 Pauling) are very close, it is reasonable to deduce that formation of NiO can take place, as well as oxidized Co. XMCD data taken at remanence from Ni@CoO NP films synthesized with sequential layer deposition Figure 15. Left: stack of FC/ZFC curves from Ni@NiO NP films grown with sequential layer deposition, at increasing values of the NiO overlayer from t NiO =0 (sample a) to t NiO =6.5 nm (sample g). Right: corresponding hysteresis loops taken at T=5 K. Reproduced from [89]. © IOP Publishing Ltd. All rights reserved.
revealed the existence of a dichroic signal at room temperature, a possible indication of stabilization of the Ni NP cores in a FM phase [91].

'Hard' magnetic NP: the case of L1 0 FePt
An obvious solution to the problem of 'beating the superparamagnetic limit' could be the use of materials with high anisotropy constant. The choice of the material is crucial for the realization of the new techniques of recording, like heat assisted magnetic recording (HAMR) or energy assisted magnetic recording [93]. An example is the FePt alloy in the chemically ordered L1 0 phase. This alloy has a face centered tetragonal structure, with an anisotropy constant K= 6·10 6 J m −3 , so stable grains of 2.5-3 nm diameters are conceivable. In practice, several effects occurring at the surface of the NP (lattice distortion due to NP faceting, surface modification due to exposure to atmosphere or interaction with the embedding media etc) can decrease this value by an order of magnitude. Very often FePt NPs are synthesized in the fcc A1 phase, where chemical ordering is disrupted and the material shows soft ferromagnetic behavior [93,94]. Similar properties have been found in CoPt NPs [95]. The L1 0 phase can be obtained by thermal annealing a T=450°C, but some strategies are necessary to prevent the NPs from sintering and agglomeration, when they are already deposited on a substrate. In some experiments, after having grown the NPs with LPGAS, thermal annealing was performed in-flight, by allowing the NP beam to flow through an oven [96]. In other works, a careful optimization of the design of the magnetron, the sputtering target, and of the parameters governing the sputtering and the NP condensation (sputtering voltage, current density, and inert gas pressure) permitted the direct synthesis of monodispersed FePt NP in the L1 0 phase [93,97,98] (see figure 17).
Another possibility can be the formation of a shell around the NP or an inert matrix during the film growth, in order to avoid sintering [99,100]. This method has been applied, with partial success by using silica and MgO [101,102]. The necessity of controlling the surface effects, which can modify and degrade the required magnetic properties of FePt NP, has been addressed, and coating or capping with different metals has been seen as a possible solution [103]. This type of fabrication, when applied to physically synthesized NPs, can be obtained with the use of multiple sources of evaporation.

Core@shell NPs realized with multimagnetron approach
A recent instrumental development of LPGAS is the planar multimagnetron source, with three targets mounted on the same assembly [50]. Nanoalloy particles and dumb-bell NPs can then be realized in a one-step procedure inside the source, with great flexibility in varying the relative alloy composition and size. Another important newly conceived apparatus is the multiple ion cluster source (MICS) [47,48,104], which permits direct synthesis of nanoalloy particles, core@shell, and core@shell@shell NPs (see figure 18). The MICS hosts three independent magnetrons mounted on separate linear drives, so that the magnetrons can be positioned in different regions of the gas aggregation region. In this way, sputtering from different materials and condensation will occur at different distances from the exit aperture of the source (see figure 18). There is also an additional linear drive for the global positioning of the three magnetrons inside the source, with a considerable increase in the number of parameters controlling the NP growth, and a consequent increase in the complexity of the nanostructures to be fabricated. As an example, results on Au@TiO x NPs obtained with MICS have been reported [48]. This system presents an interesting application and may play a fundamental role in photocatalysis because of the presence of the Au core surface plasmon resonance which allows electron transfer to the TiO 2 shell, when the NPs are illuminated by visible light. The NPs were synthesized by first obtaining Au@Ti NPs in the source and oxidizing the Ti shell in-flight before the deposition on the substrate, in a similar way used to produce CeO x NPs [43,44].
The structure was studied with Cs-corrected STEM and the electronic properties with XPS, revealing the high degree of crystal order in the Au core and the amorphous phase of the TiO x shell, mainly composed of insulating TiO 2 .
4. Low pressure gas-phase synthesis of oxide NPs: interplay between structure and functional properties in CeO 2−x .
Oxide NPs play a major role in many fields of advanced technology: photocatalysis, energy conversion and storage, biomedicine, and memory are just a few examples. Physical synthesis of oxide nanoclusters with LPGAS from oxide targets may present some difficulties or it can be impossible, if DC sputtering is used, because of the intrinsic insulating nature of many of these materials. An alternative can be represented by the use of RF magnetron plasma sputtering [105]. Another possibility is the use of the low energy cluster beam deposition (LECBD) apparatus described in [106,107]. This system makes use of a ND:Yag laser for vaporization from a target and of a He condensation system, producing supersonic beams of clusters. With this technique, quasiperfect ZnO NPs (with diameters d≈6 nm) were produced, with a very good degree of stoichiometry. Additional incorporation of atomic oxygen during the cluster growth improved the degree of crystallinity in the wurtzite phase and removed native defects. ZnO NPs also presented interesting properties of organization when deposited. In particular, the oriented attachment process was observed, with the formation of large crystalline domains, envisaging a fascinating perspective in the growth of mesostructures and superarchitectures from smaller nanocrystals [107]. Another example of techniques for the synthesis of oxide NPs is pulsed laser vaporization (PLV), which has been extensively used for growing TiO x NPs [108]. In particular, with the careful choice of the PLV working parameters (laser fluence and background pressure) it was possible to obtain 'black' TiO 2 NPs, with a crystalline rutile core and disordered Ti 2 O 3 shells, which can in principle exhibit higher photocatalytic activity with respect to ordinary white TiO 2 [108].
An alternative way to produce oxide NPs by using DC magnetron sputtering from metallic target is to oxidize inflight the metal NPs, either inside the gas aggregation region, or immediately after their exit from the source, in a separate chamber, before deposition on the substrate. This method can be very effective for reactive metals, like for instance rare earths [43,44]. Among the rare earth oxides, certainly CeO x is one of the most studied because of its capability of storing and releasing oxygen atoms as a reducible oxide, by switching reversibly between two stable oxidation states, corresponding to the formation of CeO 2 and Ce 2 O 3 compounds. This property can be utilized in many areas of application: catalysis, energy conversion and storage, biomedicine, and memory [109]. Since the redox process is reversible, there are great implications in phenomena related to oxygen transport within the material. For this reason, a huge number of studies have been carried out  on ceria in different forms [109]. In recent years, great attention has been devoted to the influence of reduced dimensionality in its redox properties, focusing on CeO x ultrathin epitaxial films and nanostructures [110]. It was found that in cerium oxide ultrathin films and NPs a higher surface-to-bulk ratio can promote an increased degree of reducibility [111][112][113]. Other factors playing a major role in this increase are the modifications of the electronic structure and a more pronounced mobility of the oxygen ions. Also, when nanocystalline cerium oxide supports metal NPs, the energy formation of oxygen vacancies is highly reduced with respect to the case of large surfaces, so oxygen can be transferred more easily to the metal NPs [114]. Nanostructured ceria has previously been obtained by magnetron sputtering and inert gas condensation, with a resulting high degree of non-stoichiometry [115][116][117] because of the presence of a massive amount of Ce 3+ ions. In more recent works, LPGAS with magnetron sputtering from a pure Ce metal target was used, and a systematic study of the effects of NP size and of the different procedures of oxidation (inflight and post-deposition) on the structure and on reducibility have been carried out [43,44]. In particular three different samples were prepared with the same NP size distribution (〈d〉=8.5 nm with a size dispersion Δd=3.9 nm) [44]: (1) In the first sample, the ceria NP assemblies were obtained by the deposition of metal NPs followed by exposure to O 2 ('post-oxidized sample').
(2) A second film was realized by obtaining oxidized Ce NPs in the aggregation region of the source, using a mixture of Ar and O 2 as a gas carrier ('directly oxidized sample'). (3) In the last case metal Ce NPs were deposited on the support, and they were exposed to residual O 2 gas. The NPs were oxidized after a short time because of the high reactivity of Ce ('sample oxidized in high vaccum, HV').
The supports used were Si/SiO x for the samples measured with XPS, and lacey grids for electron microscopy analysis. Figure 19 shows STEM images of the NPs obtained with the three different procedures. The 'post-oxidized' NPs present a CaF 2 crystalline structure; the faces of the nanocrystal exhibit mainly a (111) and a (100) orientation. The 'directly oxidized' NPs (see figure 19(b)) have an irregular shape, and a detailed HRTEM analysis demonstrated that they are mainly polycrystalline. The STEM images taken from the NPs 'oxidized in HV' (figure 19(c)) were taken with a higher magnification compared to the previous cases. Although some NP agglomerates are visible, the structure of the single NP is similar to the 'post-oxidized NP'.  By means of electron energy loss spectroscopy (EELS)-STEM maps of the NP it was possible to obtain the distribution of Ce 3+ and Ce 4+ inside the single NP. This analysis was performed by performing a point-by-point analysis of the Ce M 4,5 edge EELS spectra. For the 'post-oxidized sample' (figure 20(a)), the Ce 3+ ions are located mainly in the surface region of the NP, as they form a shell of thickness t≈0.7 nm, probably because of exposure to atmosphere while moving the samples from the source chamber to TEM because of the adsorption of OH groups. In the directly oxidized NP ( figure 20(b)), the presence of Ce 3+ ions is significant also in the core. The explanation could lie in the different procedure used to obtain the NP: during the formation of the NP in the gas aggregation region, CeO 2 clusters with Ce 3+ are formed at the surface, interacting with each other and coagulating into a bigger NP; the process of coagulation gives as a result the presence of Ce 3+ ions also in the NP core, at the borders between the original smaller clusters, which now form grains. The EELS-STEM map reported from the sample oxidized in HV (figure 20(c)) shows instead an agglomerate of NPs with Ce 3+ and Ce 4+ ion distribution within the single NP similar to the case of figure 20(a).
The three types of samples were all transferred quickly to a different UHV chamber and treated with a reduction process, which was carried out by thermal annealing in UHV at increasing temperature (from RT to T=1020 K with steps of 250 K), and with an oxidation process by annealing at T=1020 K in presence of oxygen. The behavior of NPs under reducing and oxidizing conditions was monitored with XPS from Ce 3d core levels during the reduction-oxidation cycles and an example is shown in figure 21(a) [43,44].
The analysis was performed by fitting the Ce 3d corelevel spectra with a linear combination of the Ce 3+ and Ce 4+ reference spectra [43]. A comparison of the results from the three samples ( figure 21(b)) demonstrated that the sample oxidized in HV is generally more greatly reduced than the other two. Moreover, it was not possible to re-oxidize the NP thoroughly. Instead, the other two samples could be completely oxidized. The trend of the Ce 3+ concentration vs the annealing temperature in the directly oxidized sample is similar to the one of the sample oxidized in HV, but its reduction is not complete. By combining these results with the ones obtained by the TEM analysis, we can relate the reduction-oxidation properties of these three types of NPs with their structure. The post-oxidized NPs with their regular crystalline structure probably have a high mobility of oxygen ions. Also, the (100) exposed planes visible in the TEM images have energy of formation of O vacancies lower than the one of the (111) planes, and all these factors could be responsible for the high reducibility of such NP films. At variance, the directly oxidized NPs show a polycrystalline structure, resulting in irregularities and defects especially at the borders of their grains and this presence can hamper the mobility of the O ions. Regarding the NPs oxidized in HV, their tendency to form agglomerates can be explained by their high degree of reducibility. These agglomerates can hamper a complete re-oxidation, resulting in a lower catalytic efficiency for ceria NPs realized in this way.
The behavior of ceria NPs of different sizes (〈d〉=6 , 9, and 14 nm) was also analyzed. In order to investigate the (100) surface contribution to the reducibility we compared the results obtained from the NP with 〈d〉=9 nm with the ones obtained on epitaxial ceria films grown by reactive evaporation of Ce in UHV on the Pt(111) surface [43]. The exposed film surface corresponds to the (111) plane. The film growth conditions have been chosen in order to have the same surface-to-volume ratio as the 9 nm NP sample. By XPS analysis during the oxidation and reduction cycles, we observed a stronger reducibility for the NP, possibly because of the fact that the exposed Figure 21. (a) Ce 3d XPS spectra from a ceria NP sample. Spectra (A)-(D) were acquired at RT, and after different annealing steps under UHV at 520, 770, and 1020 K; spectrum E was acquired after annealing in O 2 at 1020 K. The main Ce 3+ -related features are indicated with a red arrow. (b) Evolution of Ce3 + concentration obtained by XPS Ce 3d core-level analysis. Left side: Ce 3+ concentration from samples obtained from XPS analysis of the asdeposited samples in the NP source chamber. Right side: Ce 3+ concentration from XPS data taken in the second UHV chamber after different steps during reduction and oxidation. Reproduced from [44]. © IOP Publishing Ltd. All rights reserved. surface is (100) which is more favorable to oxygen release, instead of the (111) surface as in the case of the film.
The high reducibility of NPs may have important implications for their catalytic efficiency. In summary, these studies are relevant for the comprehension of the mechanism of nanocatalysis, and may have some important implications in the production of catalytic NPs.

Conclusions
Some selected examples of experimental activities on low pressure gas-phase synthesis metal@oxide, core@shell, and oxide NPs have been reviewed. In particular, significant effort has been put into trying to explain the physical principles of the working of LPGAS, relating them with the properties of the produced NPs, and showing some points of the strength of the bottom-up physical synthesis methods: one-step synthesis procedure, absence of unwanted chemical by-products and ligand, and higher degree of control of the NP characteristics with respect to the widely used and more economic chemical routes that can alter the intrinsic properties of the 'model system' NPs. Among the many issues which can be of relevance in nanotechnology products,the stabilization of magnetic NPs and basic redox properties of ceria nanostructures have been reviewed, taking into consideration the possibilities of more complex NP architectures offered by new instrumental developments, like the codeposition and multimagnetron approaches, which can be of significant help in working to achieve new functionalities and rational designs in nanoscience and nanotechnology.