XMCD Proof of Ferromagnetic Behavior in ZnO Nanoparticles

Article pubs.acs.org/JPCC

XMCD Proof of Ferromagnetic Behavior in ZnO Nanoparticles C. Guglieri,†,‡ M. A. Laguna-Marco,§ M. A. García,∥ N. Carmona,∥ E. Céspedes,§,⊥ M. García-Hernández,§ A. Espinosa,§ and J. Chaboy*,†,‡ †

Instituto de Ciencia de Materiales de Aragón, Consejo Superior de Investigaciones Científicas and ‡Departamento de Física de la Materia Condensada, Universidad de Zaragoza, 50009 Zaragoza, Spain § Instituto de Ciencia de Materiales de Madrid, CSIC, Cantoblanco 28049 Madrid, Spain ∥ Instituto de Cerámica y Vidrio, CSIC & IMDEA Nanociencia, Cantoblanco 28049 Madrid, Spain ∥ Dpto. Física de Materiales, Universidad Complutense de Madrid, 28040 Madrid, Spain ⊥ Institute for Science and Technology in Medicine (ISTM), Guy Hilton Research Centre, Keele University, Stoke-on-Trent ST4 7QB, United Kingdom S Supporting Information *

ABSTRACT: The combined element-specific X-ray absorption spectroscopy (XAS) and X-ray magnetic circular dichroism (XMCD) study of ZnO nanoparticles (NPs) capped with organic molecules confirms the occurrence of intrinsic ferromagnetic-like (FML) behavior up to room temperature (HTFM). Zn K-edge XMCD measurements reveal the coexistence of two different magnetic contributions: a paramagnetic response from the core of the NP, and a ferromagnetic-like contribution stemming from the interface formed between the ZnO core of the NP and the organic molecule. The extent and conformation of this interface depends on the capping molecule, and as demonstrated in ZnO/ZnS heterostructures, the FML behavior is reinforced when neat interfaces are formed.

O

reliability of the macroscopic results revealing HTFM behavior obtained by accurate full sets of control experiments13 has been further confirmed by X-ray magnetic circular dichroism (XMCD) experiments performed at the L2,3 and K absorption edges of Zn. Whereas the localized 3d states of Zn do not carry any measurable ferromagnetic moment, the magnetic polarization of the Zn 4p states suggests the existence of saturation effects overimposed to a background paramagnetic contribution.14 The confirmation of the existence of saturation effects associated with the conduction band of ZnO NPs renders crucial in the quest of the origin of the HTFM in these materials. First, it would unambiguously demonstrate that HTFM is an intrinsic phenomenon ending a longstanding issue in the field. In addition, it would contribute to determine which mechanism occurring at the surface/interface of the NPs4,15−21 is responsible for the observed HTFM. To this aim, we have investigated the magnetic field dependence of the XMCD spectra recorded at the Zn K-edge in the case of 20 nm ZnO NPs capped with three different organic molecules: trioctylphosphine (hereafter TOPO), dodecylamine (hereafter AMINE), and dodecanethiol (here-

ne of the most fascinating topics in magnetism that emerged in the several last years lies on the possibility of obtaining ferromagnetic behavior in traditionally nonmagnetic materials. In the particular case of semiconductor materials and initially connected to the research on the so-called dilute magnetic semiconductors (DMSs), the development of hightemperature ferromagnetism (HTFM) in nonmagnetic oxides has led to a new branch of research. The occurrence of this new class of magnetism seems to be triggered by the size reduction of the new nanostructures and linked to the near-surface regions (surface, grain boundaries, or interfaces),1−3 where the breaking of translational symmetry leads to surface effects such as the modification of the electronic structure and the occurrence of surface magnetic anisotropy.4 The subtle interplay between localization, necessary for the moment formation, and extended states leading to strong and long-range magnetic coupling can be reached at these near-surface regions.5 Indeed, recent works have addressed the importance of disordered interfaces and intergranular regions into driving this HTFM behavior.2,6,7 Within this framework, ZnO constitutes a paradigmatic case. Although initially magnetic elements (Fe, Mn, or Co) were employed as impurities, reports appeared soon of magnetism in the absence of magnetic elements, such as nonmagnetic transition metals (Cu, Ti, etc.)8,9 or even nonmetals like N or C.10,11 Moreover, HTFM has been reported in ZnO nanoparticles (NPs) capped with organic molecules.12 The © 2012 American Chemical Society

Received: January 25, 2012 Revised: February 24, 2012 Published: February 27, 2012 6608

dx.doi.org/10.1021/jp300837f | J. Phys. Chem. C 2012, 116, 6608−6614

The Journal of Physical Chemistry C

Article

Figure 1. Comparison of the normalized Zn K-edge XMCD spectra recorded as a function of the applied magnetic field at T = 5 K in the case of capped ZnO NPs: AMINE, TOPO, and THIOL. In all cases, the normalized Zn K-edge XANES spectra are also shown for the sake of clarity. In the lower right panel, a detailed comparison of the XMCD recorded at different magnetic fields on TOPO and THIOL samples as well as their difference are shown.

Figure 2. Scheme of the proposed formation of an interface between the core and the surface of the ZnO nanoparticles capped with the organic molecules in the case of THIOL and TOPO samples.

signal following a Curie−Weiss law and a ferromagnetic-like (FML) contribution. The FML behavior of the magnetization is characterized by the existence of remanence, coercivity, and saturation up to room temperature.14 Moreover, the value of the coercive field, HC ≈ 200 Oe, does not change with the temperature. This behavior is in contrast with that obtained by applying the same procedure in the case of reference bulk and

after, THIOL). Details of the synthesis and characterization of these samples have been previously reported.14 In addition, both bulk ZnO and ZnS compounds and ZnO/ZnS multilayers have been also studied for reference purposes. The magnetic characterization of theses samples obtained by standard macroscopic methods yields that in addition to an overall diamagnetic behavior at 5 K there is a paramagnetic 6609



Article

single-crystal ZnO samples for which the magnetization is at least one order of magnitude smaller; it shows a clear linear trend with the applied magnetic field as well as non significant hysteresis. (See Figure 13 in the Supporting Information.) The dependence of the Zn K-edge XMCD signals as a function of the external magnetic field, XMCD(H), for the three investigated molecules is shown in Figure 1. The spectral shape of the XMCD signals of TOPO and AMINE shows a narrow positive peak in correspondence to the maximum of the XAS absorption. In contrast, this main peak broadens in the case of THIOL, appearing to be composed of two superimposed single peaks. Each of these contributions occurs close to the energy at which the main peaks (A, B) of the XAS spectra appear. The enhancement of peak A in THIOL with respect to TOPO or AMINE is due to the ZnS-like phase formed through the capping of ZnO with dodecanethiol molecules.22 These results suggest that the observed XMCD signals are due to the magnetic polarization of the conduction band of Zn in both ZnS and ZnO components. This is verified by directly subtracting the dichroic signals of both THIOL and TOPO samples recorded under the same experimental conditions. As shown in Figure 1, the XMCD spectrum of THIOL after subtracting the TOPO one is similar to that of both TOPO and AMINE but shifted towards lower energies, which confirms the presence of two different magnetic contributions in the THIOL samples. Because the low-energy XMCD component of THIOL is not present in TOPO and AMINE, we can conclude that this contribution stems from the Zn atoms forming the ZnS shell, whose existence has been previously evidenced22 exclusively. (See Figure 2.) Having no 3d localized moment in the materials,23 the observed XMCD signals could be due to Pauli-paramagnetism (PP), induced by the action of the external magnetic field, or to the existence of an intrinsic magnetic polarization of these Zn electronic states. It should be noted that the possibility of a Curie−Weiss-like paramagnetic contribution is excluded as the XMCD signals do not depend on temperature. (See Figure 14 in the Supporting Information.24) In the first case, a linear dependence of the XMCD signal with the applied magnetic field is expected for the studied range of magnetic fields because for T = 5 K the field required to observe deviation from linearity, even assuming magnetic moments on the order of 1 μB, is over 25 T. The expected linearity of the PP contribution has been verified in the case of reference ZnS and ZnO bulk samples. (See Figure 3 and also Figure 15 in the Supporting Information.) In contrast, the dependence on the applied field of the XMCD spectra recorded in the capped ZnO NPs (Figure 4) shows a different trend. While the XMCD versus H dependence is linear, within the signal-to-noise ratio, for AMINE, it clearly deviates from this trend for H ≥ 4 T in the case of TOPO. The behavior of THIOL is more complex: the low-energy XMCD peak, associated with the ZnS component, shows a linear trend, while the XMCD(H) of the high-energy peak, associated with ZnO, clearly departs from linearity. These results are in agreement with the behavior of the integral of the XMCD signals performed in the energy range from −5.5 to 20.5 eV, as shown in Figure 4d. The integrated XMCD(H)/H is approximately constant for AMINE, as expected for a linear dependence, whereas it decreases as H increases for both TOPO and THIOL, as expected for a saturated regime. All of these results are an unambiguous proof of the existence of an intrinsic FML behavior in these capped ZnO NPs. The fact that this saturated signal is overimposed to a paramagnetic

Figure 3. Top: Comparison of the normalized Zn K-edge XAS and XMCD spectra of bulk reference samples ZnS (red, ○), ZnO (●), and metallic Zn (blue, solid line) recorded at H = 10 T and T = 5 K. Bottom: Comparison of the normalized Zn K-edge XMCD spectra of ZnS recorded at different applied magnetic fields and those derived from the XMCD signal recorded at H = 10 T by assuming a linear dependence with the applied field.

one suggests that the magnetic response is not the same for all Zn atoms in the material. It should be noted in this respect that Zn K-edge X-ray absorption measurements, XAS and XMCD, probe all Zn atoms in the material, that is, both at the core and at the surface of the NPs. The fact that the paramagnetic contribution to the XMCD signal dominates over the ferromagnetic one suggests that the latter is confined near the surface or at the interface formed between the ZnO NP and the capping molecule. The formation of an interface between the NP and the capping molecule, schematized in Figure 2, has been evidenced in the case of ZnO NPs capped with dodecanethiol.22 The question now is to assess if the same occurs for different capping molecules. To this end, we have compared in Figure 5 the Zn K-edge XANES spectra of both TOPO and AMINE with that of bulk ZnO. The XANES of both NPs and bulk ZnO reference show the same spectral profile for energies higher than the white line (peak B). In particular, there is no difference in the shape nor in the relative intensity of the three-peak structure (D1, D2, D3) occurring at ∼20 eV above the edge. This spectral feature is characteristic of the wurtzite-ZnO (wZnO) crystal structure and very sensitive to small modifications of the local structure around Zn. (See Figure 17 in the Supporting Information.) The fact that the white line decreases for TOPO and AMINE while this structure remains unchanged is indicative of the presence of Zn atoms, forming a short-range phase overimposed to that of bulk-like ZnO. (See Figures 16 6610



Article

Figure 4. Comparison of the normalized Zn K-edge XMCD spectra of AMINE, TOPO, and THIOL recorded at different applied magnetic fields and those derived from the XMCD signal recorded at H = 10 T by assuming a linear dependence with the applied field. The dependence on the applied field of the integrated signals is shown in panel d.

and 18 in the Supporting Information.) Detailed ab initio computations demonstrate that the differences between the experimental XANES spectra recorded for bulk ZnO and for the capped ZnO NPs can be accounted for in terms of a twophase scheme: in addition to the expected bulk-like ZnO contribution, a short-range ZnO phase, extending only to a few coordination shells, also contributes to the XANES spectra. Making a parallel to the THIOL case, this new phase should be located at the near-surface region, that is, at the interface between the ZnO bulk-like core of the NP and the capping molecules. (See Figure 2.) As shown in Figure 5, the theoretical spectra obtained from the addition of these two contributions reproduce the observed decrease in the white-line intensity, whereas the rest of the spectrum remains unvaried. The existence of two different phases containing Zn has large implications in the normalization of the XMCD signals. This is clearly exemplified in THIOL. Previous works have shown that in the case of the ZnO NPs capped with dodecanethiol both wZnS and w-ZnO are present with a relative weight of ∼50%. As discussed above, each of the two main peaks of the XMCD signal of THIOL (Figure 1) corresponds to the ZnS and ZnO phases of the capped NP. Because both the XANES and the XMCD spectra are normalized to the absorption jump, that is, to the total amount of Zn atoms in the sample and not only to those present in each phase, the intensity of the XMCD peaks is artificially reduced by the normalization procedure. Indeed, if one takes into account that approximately only one-half of the total Zn atoms contributes to each peak, then the overall XMCD signal should be enhanced by a factor of two, which renders the amplitude of the ZnO contribution to the XMCD

in THIOL of similar magnitude as that for both TOPO and AMINE samples. Similar reasoning should be applied to the normalization of the contribution to the XMCD coming from the Zn atoms at both the core and the interface of the NPs. Accordingly, the total XMCD signal should correspond to the addition: XMCDTot = (1 − x) × XMCDbulk + x × XMCDinter, where x is the percent of the interface with respect to the total volume of the NP. (In the case of THIOL, XMCDbulk stands for the bulk contribution of both wurzite-like ZnS and ZnO regions). Estimates of the amount of Zn atoms at this interface have been derived from the ratio of the white line intensity with respect to that of bulk ZnO. In this way, we obtain that the amount of Zn atoms at the interface is about 15, 10, and 5% for AMINE, TOPO, and THIOL, respectively. Now, by assuming that the contribution of the Zn atoms in the ZnO core to the XMCD is the same as that for bulk ZnO, we can extract the XMCD corresponding to the interface by subtracting from the total XMCD signal the ZnO contribution according to its relative weight. The result of this procedure is shown in Figure 6. In all cases, the obtained signals show similar spectral shape. This is of special significance in the case of THIOL, in which two corelike signals, corresponding to the observed ZnS and ZnO contributions, have been subtracted from the total XMCD. As shown in Figure 6, the extracted XMCD signal of TOPO is clearly saturated for H ≥ 4 T. In the case of THIOL, the extracted XMCD is saturated in the region corresponding to the Zn−O bonds, whereas it shows a linear dependence with the applied magnetic field in the Zn−S bonds region. The case of AMINE is unclear as the lower signal-to-noise ratio prevents 6611



Article

Figure 5. Top panel: Comparison of the normalized Zn K-edge XAS spectra of capped ZnO NPs (AMINE (blue, dots), TOPO (red solid line), and THIOL (green, dashes)) and of bulk ZnO (●) as well as their difference. Bottom panel: Comparison between the experimental Zn K-edge XANES of wurtzite-ZnO (●) and TOPO (green, ○) and the weighted sum of computations performed for long- and shortrange ZnO clusters. (See the text for details.)

assurance that there is a real saturation effect for magnetic fields higher than H = 6 T. As discussed above, the FML XMCD signal stems from ∼5 to 15% of the total amount of Zn atoms in the material. Taking into account that the size of the NPs is ∼20 nm, we estimate that the FML behavior arises from a 3 to 8 Å thickness region. The extent of this region is in agreement with the result of the ab initio computations, suggesting that the FML saturated signal stems from the Zn atoms at the interface formed between the ZnO core of the NP and the organic molecule. To verify further these findings, we have grown two ZnO/ZnS multilayers in which the thickness of each ZnS and ZnO layer is 2 and 4 nm. Then, if our hypothesis is correct, the bulklike ZnO and ZnS contribution to both the XANES and XMCD should decrease. Accordingly, the paramagnetic component associated with the core should decrease, whereas the FML contribution of the interface should maximize. Therefore, the XMCD spectra will be directly comparable to the FML contribution extracted from the XMCD of the capped NPs after renormalization to the relative percentage of the interface. The comparison, shown in Figure 7, is in agreement with our starting hypothesis. A final comment is deserved to the slight differences observed in the XANES of both ZnO/ZnS multilayered samples. The intensity of the white line is smaller and the

Figure 6. Comparison of the extracted Zn K-edge XMCD(H) of AMINE (top panel), TOPO (middle), and THIOL (bottom)-capped ZnO NPs after subtracting the contribution of the Zn atoms at the core: H = 10 T (●), 8 T (red, ○), 6 T (blue, ■), and 4 T (green, □) and (magenta, ▲).

spectral features are smoother for (ZnO2nm/ZnS2nm)20 than for the (ZnO4nm/ZnS4nm)10 heterostructure, as it should correspond to the more disordered interface developed in the former sample. This is in agreement with X-ray reflectometry and cross-sectional TEM images results (Figures 10 and 11 in the Supporting Information), which confirm neat interfaces for (ZnO4nm/ZnS4nm)10 and large interdiffusion and disorder for (ZnO2nm/ZnS2nm)20. The fact that the HTFM behavior is reinforced in the case of the (ZnO4nm/ZnS4nm)10 heterostructure (Figure 12 in the Supporting Information) suggests that the FM interactions are weaker at an ill-defined interface and, consequently, the existence of a well-conformed interface is a prerequisite to observe the FML behavior. In conclusion, we have demonstrated that the modification of the surface of ZnO NPs through the capping with organic molecules enables the development of ferromagnetic behavior 6612



Article

at SPring-8 (long-term proposal no. 2009B0024). The assistance of the BL39XU staff during the SR experiments is acknowledged. C.G. acknowledges the Ministerio de Eduación y Ciencia of Spain for a Ph.D. grant. M.A.L.-M. acknowledges the Ministerio de Ciencia e Innovación of Spain for a Juan de la Cierva grant, and N.C. acknowledges the financial support of the FSE-MEC, Ramón y Cajal program (ref RYC-2007− 01715). The Spanish FECYT (postdoc program) is acknowledged by E.C.

■

(1) Straumal, B. B.; Mazilkin, A. A.; Protasova, S. G.; Myatiev, A. A.; Straumal, P. B.; Schütz, G.; van Aken, P. A.; Goering, E.; Baretzky, B. Phys. Rev. B 2009, 79, 205206. (2) Céspedes, E.; Laguna-Marco, M. A.; Jiménez-Villacorta, F.; Chaboy, J.; Boada, R.; Guglieri, C.; deAndrés, A.; Prieto, C. J. Phys. Chem. C 2011, 115, 24092. (3) Straumal, B. B.; Mazilkin, A. A.; Protasova, S. G.; Myatiev, A. A.; Straumal, P. B.; Goering, E.; Baretzky, B. Thin Solid Films 2011, 520, 1192. (4) Hernando, A.; Crespo, P.; García, M. A.; Ayuela, A.; Echenique, P. M. Phys. Status Solidi B 2011, 248, 2352. (5) Muñoz, M. C.; Gallego, S.; Sanchez, N. J. Phys.: Conf. Series 2011, 303, 012001. (6) Straumal, B. B.; Protasova, S. G.; Mazilkin, A. A.; Baretzky, B.; Myatiev, A. A.; Straumal, P. B.; Tiezte, T.; Schütz, G.; Goering, E. Mater. Lett. 2012, 71, 21. (7) Guglieri, C.; Céspedes, E.; Preito, C.; Chaboy, J. J. Phys.: Condens. Matter 2011, 23, 206006. (8) Coey, J. M. D. Curr. Opin. Solid State Mater. Sci. 2006, 10, 83. (9) Yan, W.; Sun, Z.; Liu, Q.; Li, Z.; Pan, Z.; Wang, J.; Wei, S. Appl. Phys. Lett. 2007, 91, 062113. (10) Herng, T. S.; Lau, S. P.; Wei, C. S.; Wang, L.; Zhao, B. C.; Tanemura, M.; Akaike, Y. Appl. Phys. Lett. 2009, 95, 133103. (11) Yu, C.-F.; Lin, T.-J.; Sun, S.-J.; Chou, H. J. Phys. D: Appl. Phys. 2007, 40, 6497. (12) García, M. A.; Merino, J. M.; Pinel, E. F.; Quesada, A.; de la Venta, J.; González, M. L. R.; Castro, G. R.; Crespo, P.; Llopis, J.; González-Calbet, J. M.; et al. Nano Lett. 2007, 7, 1489. (13) García, M. A.; Pinel, E. F.; de la Venta, J.; Quesada, A.; Bouzas, V.; Fernández, J. L.; Romero, J. L.; Martn-González, M. S.; CostaKrämer, J. L. J. Appl. Phys. 2009, 105, 013925. (14) Chaboy, J.; Boada, R.; Piquer, C.; Laguna-Marco, M. A.; GarcíaHernández, M.; Carmona, N.; Llopis, J.; Ruíz-González, M. L.; González-Calbet, J.; Fernández, J. F.; et al. Phys. Rev. B 2010, 82, 064411. (15) Hong, N. H.; Sakai, J.; Brizé, V. J. Phys.: Condens. Matter 2007, 19, 036219. (16) Wang, X.; Leung, T. C.; Harmon, B. N.; Carra, P. Phys. Rev. B 1993, 47, 9087. (17) Patterson, C. H. Phys. Rev. B 2006, 74, 144432. (18) Sanchez, N.; Gallego, S.; Cerdá, J.; Muñoz, M. C. Phys. Rev. B 2010, 81, 115301. (19) Gallego, S.; Beltrán, J. I.; Cerdá, J.; Muñoz, M. C. J. Phys.: Condens. Matter 2005, 17, L451. (20) Dietl, T. Nat. Mater. 2010, 9, 965. (21) Hernando, A.; García, M. A. J. Nanopart. Res. 2011, 1−8. (22) Guglieri, C.; Chaboy, J. J. Phys. Chem. C 2010, 114, 19629. (23) It should be noted that Zn L2,3-XMCD experiments performed on the same samples demonstrated that the 3d electronic shells of Zn do not carry any measurable ferromagnetic moment (≥0.005μB) so that the presence of defects and vacancies do not yield a partially unfilled magnetically polarized 3d shell. (24) It should be also noted that according to the data of bulk and single crystal ZnO shown in Figure 13 of the Supporting Information, the magnetization at T = 5 K and H = 40 kOe is ∼3 × 10−4 emu/g, which results in a paramagnetic susceptibility of 7.5 × 10−9 emu/g·Oe (i.e., 3 × 10−9 in the S.I. using 5.6 g/cm3 as the ZnO density). By

Figure 7. Comparison of both the extracted XMCD (T = 5 K, H = 10 T) Zn K-edge of THIOL (black, solid squares) after normalizing to the nonbulk phase content (see text for details) and the XMCD recorded on the (ZnO2nm/ZnS2nm)20 (○, red) multilayers at room temperature. For the sake of completion, the XAS spectra are also shown: THIOL (black, solid line), (ZnO2nm/ZnS2nm)20 (○, red), and (ZnO4nm/ZnS4nm)0 (●, blue).

up to room temperature. This work also demonstrates that both the XAS and XMCD signals of core−shell-like NPs are made of two different contributions, whose disentangling is mandatory to determine the magnetic behavior of each component. In addition, the existence of Pauli paramagnetism contribution, hindered by diamagnetism to macroscopic magnetometry tools, in the conduction band of ZnO has been demonstrated. Zn K-edge XMCD versus H measurements reveal, from the observed saturation at moderate applied fields, an intrinsic FML contribution stemming from the formed interface that is estimated to extent over 3 to 8 Å depending on the capping molecule. Our results also point out that the ferromagnetic behavior is favored when Zn−O bonds rather than Zn−S ones are involved at this interface. These results end the longstanding controversy about the existence of intrinsic HTFM in ZnO-based systems, providing new insights to establish finally the mechanism that sets on the ferromagnetic order in these systems and bringing support to this new route for roomtemperature semiconductor spintronics.

■

ASSOCIATED CONTENT

S Supporting Information *

Details on the XAS and XMCD experimental methods,25,26 including additional XMCD data, thin films samples preparation and characterization,27 and ab-initio XANES calculations. This material is available free of charge via the Internet at http://pubs.acs.org.

■ ■

REFERENCES

AUTHOR INFORMATION

Notes

The authors declare no competing financial interest.

ACKNOWLEDGMENTS This work was partially supported by Spanish MAT2008− 06542-C04−01, MAT-2010−16022, MAT2010−09346-E, MAT2011−27573-C04−04, MAT2011−27470-C02−02, FIS2008−06249, and CSD2009−00013 and by the Madrid Government project NANOBIOMAGNET S2009/MAT1726. The synchrotron radiation experiments were performed 6613



Article

assuming that it follows a Curie law and that there are 4 × 1028 Zn atoms/m3, the magnetic moment per Zn atom should be ∼4 × 10−7 μB/atom, which demonstrates that the observed XMCD are not due to the Curie−Weiss contribution. (25) Maruyama, H. J. Synchrotron Radiat. 2001, 8, 125. (26) Suzuki, M.; Kawamura, N.; Mizumaki, M.; Urata, A.; Maruyama, H.; Goto, S.; Ishikawa, T. Jpn. J. Appl. Phys. 1998, 37, L14C88. (27) Espinosa, A.; García-Hernández, M.; Menéndez, N.; Prieto, C.; de Andrés, A. Phys. Rev. B 2010, 81, 064419.

6614


XMCD Proof of Ferromagnetic Behavior in ZnO Nanoparticles

Recommend Documents