Magnetic properties in α-MnO2 doped with alkaline elements

α-MnO2 nanotubes were fabricated using a hydrothermal technique. Li, Na and K ions were introduced into MnO2 nanotubes to tailor their magnetic properties. It was found that with a doping concentration lower than 12 at%, the nanotubes showed ferromagnetic-like ordering at low temperature (<50 K), while antiferromagnetic coupling dominated their physical behavior with doping concentrations beyond 12 at%. Such experimental phenomenon was in very good agreement with associated first principle calculations. The ferromagnetic-like ordering originates from the breaking of equivalence between two different Mn-O octahedrals in α-MnO2 due to the filling of alkaline ions in the tunnels. Both small charge transfer and lattice distortion play important roles in the ferromagnetic ordering.

detailed investigation is needed to identify unambiguously the mechanism generating such ferromagnetic-like behaviour in this material.
In this work, we studied the ferromagnetic behaviour in Kdoped MnO 2 nanotubes. X-ray absorption near edge spectroscopy (XANES) indicates that Mn is very close to 41 valency rather than 31, suggesting very weak charge transfer and this weak charge transfer alone may not be the origin of ferromagnetic-like behaviour. Detailed investigations reveal that neighbouring tunnels which may be filled and unfilled lead to differences in the magnetic moment of Mn atoms in the apex site and plane site, thus breaking the balance of Mn-O-Mn bonds. This results in the ferromagnetic-like ordering. Furthermore, we can also fill Li or Na ions into these tunnels by exchanging Li/Na ions with K 1 and this leads to filling behaviour and induced ferromagnetic properties that are similar to that observed by K 1 doping. a-MnO 2 is one of the most promising anode materials for lithium batteries. However, the existence of K 1 during fabrication has impeded charging/discharging of these batteries. This work has shown that Li or Na ions can totally replace K 1 through a solution exchanging method. The understanding of Li or Na filling mechanism in MnO 2 may be of importance for the development of high performance batteries.

Results and Discussion
Characterization of MnO 2 nanorods. We first examine the X-ray diffraction (XRD) patterns of K-doped MnO 2 nanotubes. This demonstrates that the phase of all the samples is a-MnO 2 and no secondary phase is observed (Fig. S1a). In addition, the intensities of the peaks for 16 at% K-doped MnO 2 is much lower than that of the others, suggesting disordering in this sample due to the high concentration of K doping. If we enlarge the (211) peak, which has the strongest intensity (Fig. S1b), we can see that with increasing doping concentration, the peaks shift to a lower 2h value, suggesting d spacing expansion caused by the incorporation of K in the tunnel. The XRD spectra of the Li and Na-doped MnO 2 are similar to that of K-doped MnO 2 . No secondary phases were observed in either sample.
To further verify the effects of doping, Transmission electron microscopy (TEM) analysis was used to investigate the microstruc-ture of the doped oxides as shown in Fig. 1. The shape of the nanotubes is similar to that in the literature (Fig. 1a) 19 The inset of Fig. 1a shows a typical square shaped MnO 2 nanotube under low magnification TEM imaging. All the tubes with doping concentrations lower than 12 at% show strong crystalline structures as seen from Fig. 1b-1e. Whereas, the high resolution TEM image of 16 at% K-doped MnO 2 exhibits some localised disordering (Fig. S2), which is consistent with the XRD measurements. D-spacing analysis using digital micrograph software indicates that 2 at%, 6 at% and 12 at% K-doped MnO 2 samples have a d-spacing values of 0.506, 0.511 and 0.514 nm in the (200) plane of a-MnO 2 respectively (Fig. 1b to 1d). This indicates that K doping leads to lattice expansion and the higher the doping concentration the greater the lattice expansion, confirming that introducing K ions into MnO 2 results in geometry change. Li and Na-doped MnO 2 have a lattice spacing of 0.510 and 0.512 nm in the (200) plane, respectively, which are smaller than that of K-MnO 2 with 6 at% K doping. This may be due to the relatively lower dopant concentration and relatively smaller radii of Li 1 (0.9 nm) and Na 1 (0.116 nm) ( Fig. 1e and 1f).
As discussed, the mixture of Mn 31 and Mn 41 valence state in a-MnO 2 has been considered as one of the reasons for the ferromagnetic ordering/spin glass behaviour at low temperature 18 . To determine the valence state of Mn in doped MnO 2 , X-ray absorption fine structure (XAFS) measurements were performed for all the samples, as shown in Fig. 2. From the XAFS spectra in Fig. 2a, it can be seen that all the spectra have similar chemical shifts, which corresponds to the presence of Mn 41 . No trace of Mn 31 species can be observed in the spectra. Fig. 2b shows the XANES spectra of the 6 at% K-MnO 2 compared to other forms of manganese oxide, such as MnO, MnO 2 , Mn 2 O 3 and Mn 3 O 4 . It is found that the XANES of the 6 at% K-MnO 2 in the near edge overlaps with that of MnO 2 , indicating a Mn 41 valence state for this sample. These results indicate that charge transfer between K and Mn is very weak, which is not readily detected by XAFS analysis.
Fourier transform of the XAFS data and the fitting to the first Mn-O shell indicate that 2% K-MnO 2 has very small distortion in its structure (Fig. 2b). 6 at% and 12 at% K-MnO 2 has an eminent distortion in the first Mn-O shell (Table SI in the supporting informa-  tion), suggesting that the inclusion of K in the tunnels induces geometric frustration of the triangular lattices, which may destroy the balance of Mn-O-Mn antiferromagnetic ordering, leading to ferromagnetic-like ordering. Though Jahn-teller distortion due to the existence of Mn 31 has been proposed in MnO 2 23,24 , in this work, XAFS cannot detect Mn 31 in these samples. From the structure of a-MnO 2 , the tunnel has a diameter larger than 0.48 nm. A small amount of K doping should not induce lattice distortion. It has been reported that the tunnels in a-MnO 2 are usually supported by doped ions to avoid the collapse of these structures 23 . In our experiments, the K concentration has been strongly diluted by HCl exchange. Whilst a collapse of structure was not observed, suggesting that the OH 2 , H 2 O or H 3 O 1 may always act to support the scaffold 9,25 . Therefore, a small amount of K doping may still induce some degree of lattice distortion. For samples with a doping concentration higher than 16 at%, there is a small distortion in MnO 6 octahedrons of the first shell, but a large distortion in the high order shells, similar to that of a short range ordered material (Debye Waller factor 0.0114 from Table S1 in the supporting information). It is known that the a-MnO 2 has 12.5 at% tunnels, which may be fully filled if the concentration of K ions is higher than 12.5 at%. This uniform distribution may reduce the extent of distortion in MnO 6 . Certainly, some tunnels may be filled with more than one K ion, leading to small distortion. While, some other K ions may reside in the interstitial sites of MnO 6 rather than tunnels, producing a more disordered structure. Similarly, 6 at% Li or Na doping does not induce a large distortion in the first MnO 6 shell. This is due to the relatively lower dopant concentration of Li and Na. The small atomic radius of Li 1 (0.09 nm) and Na 1 (0.116 nm) may also explain the smaller distortion in MnO 6 .
MnO 2 is polymorphic with a, b, e and c phases. Fig. 2d shows the Fourier transform spectra of both 6 at% K-MnO 2 and standard traces for the a, b, e and c phases. From the comparison of Mn-O and Mn-Mn shells, it confirms that all the nanotubes fabricated in this work are a-MnO 2 and this is consistent with the XRD analysis.
Magnetic properties measurement. Ferromagnetic-like behavior of K-MnO 2 has been observed in previous studies 18 . In this work, we found that when K doping concentration is lower than 12 at%, the hysteresis loop with eminent coercivity appears at 5 K, indicating ferromagnetic ordering at low temperatures (Fig. 3a). 6 at% K-MnO 2 presents the highest saturation magnetization (5.2 emu/g at 30 kOe) and coercivity (7500 Oe). The mechanism for the high coercivity is not very clear. It may be associated with the exchange coupling between antiferromagnetic phase and ferromagnetic-like phase as that in NiO nanostructures 26,27 . The decreasing coercivity with increasing doping concentration may be due to the increasing antiferromagnetic phase with small ratio of ferromagnetic-like phase 26,28,29 . It is interesting to note when the doping concentration is higher than 12 at% a linear M-H curve is observed and the magnetization is much smaller than that of MnO 2 with the lower doping concentrations. This suggests that the sample may become antiferromagnetic again. In order to study the M-H curves in detail, we enlarged the M-H curves for 12 at% and 16 at% K-MnO 2 over a narrow scale as shown in Fig. 3b. For the 12 at% K-doped MnO 2 , the curve is almost linear, suggesting the antiferromagnetic signal in dominant. However, a small coercivity (25 Oe) can still be detected in the sample, suggesting very weak ferromagnetic-like ordering. For the 16 at% K-doped MnO 2 , a coercivity of nearly 650 Oe was observed, indicating relatively strong ferromagneticlike ordering in this sample in addition to the more dominant antiferromagnetic signal. This results suggest that ferromagneticlike behavior may not only originate from the electron charge transfer by K doping and also the Mn 31 related to Jahn-teller distortion as the increase of K doping concentration should lead to more charge transfer, thus enhances the ferromagnetic-like ordering if the ferromagnetic-like behavior is arisen from charge transfer alone. However, it has been discovered that 12 at% K doped MnO 2 has a very weak ferromagnetic-like ordering compared to that of 6 at% K doped MnO 2 . It should be noted that we did not measure pure a-MnO 2 nanotubes without K, Li or Na doping since it is impossible to achieve totally alkaline element free MnO 2 tubes with a chemical synthesis technique using KMnO 4 as a precursor. Fig. 3c shows the hysteresis loops of 6 at% K-MnO 2 , 6 at% Li-MnO 2 and 6 at% Na-MnO 2 taken at 40 K, respectively. K-MnO 2 has the highest Since Mn impurities, such as Mn 3 O 4 , is ferrimagnetic, which may contribute to the magnetic ordering in MnO 2 nanorods. Our TEM, XRD and EXAFS have not detected any impurities phases other than a-MnO 2 nanotubes. The resolution is better than 1%. From Ref. 30, the saturation magnetization of Mn 3 O 4 is around 20 emu/g. Hence, the contribution from Mn 3 O 4 is only 0.2 emu/g, which is negligible for the sample with a saturation magnetization of 5.2 emu/g. In addition, the Curie temperature of Mn 3 O 4 nanoparticles is around 40 K, while in our work, the Curie temperature is around 50 K. Furthermore, by increasing or decreasing alkaline doping concentration, the magnetization and coercivity will vary accordingly, supporting that the magnetization in our samples are not from Mn 3 O 4 impurities.
From zero field cooling/field cooling (ZFC/FC) measurements, the critical temperature of ferromagnetic-like ordering was measured as shown in Fig. 4. 2 at% K-doped MnO 2 has a Curie temperature of 50.4 K. The reverse susceptibility indicates that the nanotubes also have a negative susceptibility, which indicates the samples have a mixture of ferromagnetic-like and antiferromagnetic phase. It is known that pure MnO 2 is antiferromagnetic. Hence, the antiferromagnetic signal should come from antiferromagnetic MnO 2 itself. The ferromagnetic-like signal is from alkaline element doping. When the doping concentration is 6 at%, the Curie temperature increases to 55.8 K and the antiferromagnetic phase is undetectable. Continual increase of the doping concentration leads to a decrease of critical temperature. 12 at% K-doped MnO 2 has the lowest critical temperature (43.7 K). From the M-H measurements, a very weak ferromagnetic-like signal is detected (Fig. 3), thus having the lowest ordering temperature. There are two sharp peaks present in the ZFC curves. This suggests that three transitions may occur in the samples. This phenomenon has been reported elsewhere 18 . The irreversible peak at 24 K is related to spin glass behavior 18 . The peak at 34 K may be related to some disordered structures due to the spin frustration induced by the large amount of K doping, since there is no peak in the FC curve with an applied field as small as 500 Oe 18 . The reverse susceptibility in the inset shows that there is a large amount of anti-   ferromagnetic phase in 12 at% K-MnO 2 . Hence, the ferromagneticlike ordering is very weak. However, the 16 at% K-doped MnO 2 has similar ZFC and FC curves to that of 12 at% K-MnO 2 , as shown in the Fig. S6.
First principles calculations. In order to understand the mechanism of ferromagnetic-like behavior in MnO 2 nanotubes, we employ first principle calculations to investigate the origin of ferromagnetic-like behavior. For pure MnO 2 , the most stable spin configuration is the antiferromagnetic state as shown by the density of states (DOS) (Fig. 5b), consistent with previous calculations 23 . It shows semiconductor behavior and the Fermi level is inside the energy gap. The bandgap is approximately 1.44 eV, in good agreement with other theoretical and experiments results 23,25 . However, after the incorporation of 6.25% K, the spin degeneracy around the Fermi level is broken as shown in Fig. 5c and the entire system shows magnetic properties. The Fermi level is increased to the conduction band, indicating half metallic property.
In this case, one unit cell has a magnetic moment of 1 m B , corresponding to 0.0625 m B /Mn. It is noted that K and O do not show any magnetic moment from our calculations. A saturation magnetization of 5.2 emu/g (at 30 kOe) was experimentally observed in K doped MnO 2 , corresponding to 0.085 m B /Mn, agrees well with theoretical calculations. Further increasing the doping concentration, K ions will be incorporated in neighboring tunnels. The K doping then affects the two different Mn-O octahedrals (purple and blue in Fig. 5a) equally. The magnetic MnO 2 becomes antiferromagnetic again (Fig. 5d). The difference of the DOS between MnO 2 without K doping and with 12 at% K doping is the position of Fermi level. The latter one is inside the conduction band, indicating conductive antiferromagnetism instead of semiconductive/insulator antiferromagnetism for MnO 2 without K doping. Theoretically, if the doping concentration is increased to more than 12.5%, the balance of the homoge-neously distributed K atoms may be broken, resulting in the ferromagnetic-like phase again. The stronger ferromagnetic-like ordering in 16 at% K-doped MnO 2 (Fig. 3b) than that in 12 at% Kdoped MnO 2 was verified by the results of calculations. From our calculations, 18.5 at% K-MnO 2 shows a half-metallic behavior (Fig.  S7 in the supporting information), which means that the antiferromagnetic coupling in 12 at% K-doped MnO 2 changes to ferromagnetic-like coupling with increasing doping concentration. It is noted that in this case one tunnel of MnO 2 may contain more than one K ions. The magnetic measurement from these experiments shows that the M-H curve of 12 at% K-MnO 2 is near linear, indicating that antiferromagnetic coupling dominates. However, a very small coercivity has been observed if the X axis scale is enlarged (Fig. 3b). The minor discrepancy between the experimental and theoretical results may be due to the inhomogeneous distribution of K ions in these samples. Different from the ideal condition of theoretical calculation, in the 12 at% K doping sample, the slightly non-uniform K doping may induce some degree of ferromagnetic-like ordering. However, the main signal from the magnetic measurement by SQUID is antiferromagnetic. From the experimental results analyzed by XAFS, if the doping concentration is higher than 16 at%, the disordered structure was detected, suggesting that K may enter interstitial sites in MnO 2 . Such a disordered structure may result in a paramagnetic behavior. This is different from the periodical model of first principle calculations. Therefore, the sample with higher K doping concentration will not be further discussed in this work.
Similarly, first principle calculations were also employed to calculate Li and Na doped MnO 2 . It shows that both 6.25 at% Li-MnO 2 and 6.25 at% Na-MnO 2 show half-metallic behavior, which is similar to that for K doping (Fig. 5e and 5f). Experimentally, it shows that 6 at% Li or Na doped MnO 2 has very strong ferromagnetic-like ordering at low temperature (i.e. 5 K), which is similar to that 6 at% Kdoped MnO 2 , as shown in Fig. 3 and Fig. S5. However, 12.5 at% Li and Na doping in these samples result in the antiferromagnetic coupling again, similar to that of the K-doped MnO 2 . The results have shown that the ferromagnetic-like ordering can be achieved in all the three alkaline elements. The ferromagnetic-like ordering can be tailored by tuning the doping concentration.
This doping effect on magnetic properties can be understood by the special crystal structure of a-MnO 2 . a-MnO 2 is one of the hollandite-romanechite families with 2x2 tunnel structure, similar to b or rutile-MnO 2 . A Mn-O octahedral is expected in every side of tunnel (as shown in Fig. 5a). The two Mn-O octahedrals (blue and purple shown Fig. 5a) are equivalent in the pure a-MnO 2 and share the O atoms each other. For O atoms in a-MnO 2 , each O atom is shared by two Mn-O octahedrals and occupies the different sites of these two octahedrals: the apex of one octahedral and plane corner of another. Thus, the change of such an O atom has a different effect on these two octahedrals and also the neighboring Mn atoms. If one tunnel is doped with K ions (I in Fig. 6a) while the neighboring tunnel is undoped, i.e. empty (II in Fig. 6a), the interaction between K and such O atom breaks the symmetry between blue and purple Mn-O octahedrals, which ensures the antiferromagnetic coupling between neighboring Mn atoms in pure MnO 2 . Although the geometrical distortion and charge transfer are only minor due to such doping, symmetry breaking of two kinds of Mn-O octahedral transforms the antiferromagnetic state of pure MnO 2 into ferromagnetic-like state. Fig. 6b and Fig. 6c show the partial DOS (PDOS) projected on to the blue and purple Mn atoms (belong to two different kinds of octahedrals) close to the K atoms after doping. The remarkable difference of the two PDOSs can be observed. The blue Mn atom has a much larger magnetic moment (3.03 m B ) than that of the purple Mn (2.9 m B ) around the Fermi level. Hence, the symmetry between the purple and blue Mn-O octahedrals is broken, resulting in ferromagnetic-like properties observed in the experiment 31 . Our further calculation (Fig.  S7) indicates that the ferromagnetic-like ordering is a combined effect of K doping induced lattice distortion and small charge transfer between K and Mn. The distortion of the lattice leads to the asymmetry of energy splitting and the small charge transfer leads to the overlap of the Fermi level in the conduction band. Because the doping concentration is low, the overall valence state change is very small. For example, for the 6 at% K-doped MnO 2 sample there will be charge transfer of 6 electrons to MnO 2 given that the cell size has 100 Mn atoms. Then, the transferred electrons will change the Mn valance from 41 to 13.94, which is a very small change that may not be detected by XAFS examination.
Conclusion. We have synthesized K-MnO 2 with different doping concentrations using a hydrothermal method with KMnO 4 as the precursor. K-doped MnO 2 nanotubes with doping concentration lower than 12 at% show ferromagnetic-like ordering. 6 at% alkaline ions doping in MnO 2 leads to the maximum saturation magnetization. Doping concentration higher than 12 at% leads to disordered structure since some of K ions may enter the interstitial sites due to the higher doping concentration. Li and Na doping also lead to ferromagnetic-like behavior. The results of first principle calculations are consistent with the experimental data. The ferromagnetic-like ordering is due to incomplete filling of K ions in the tunnels, which affects the symmetry of Mn plane, forming ferromagnetic-like ordering.

Methods
Synthesis and characterization. a-MnO 2 nanotubes were prepared using a hydrothermal method, similar to that previously reported 18,19 . KMnO 4 was used as the precursor and was dissolved in HCl solution for 12 hours at 413 K. All chemicals in this work were purchased from Sigma-Aldrich with a purity of 99.99%. The prepared a-MnO 2 nanotubes contain approximately 8 at% K that was measured using inductively coupled plasma (ICP, ICPMS, PerkinElmer quadrapole Nexion ICPMS) analysis. In order to control the K doping concentration in the MnO 2 nanotubes, the prepared MnO 2 nanotubes were soaked into 1 M HCl or 1 M KOH solution for different durations from 1 hr to 6 hrs at 413 K. The nanotubes soaked in the KOH solution showed an increase in the K doping concentration. On the other hand, the nanotubes soaked in the diluted HCl solution resulted in a decrease in doping concentration. It was found that K doping concentration is proportional to soaking time. Through the control of the soaking time combined with the ICP analysis, the K doping concentration was estimated to be approximately 2 at%, 6 at%, 8 at%, 12 at% and 16 at% in atomic ratio respectively. Similarly, 1 M NaOH and 1 M LiOH solutions were also used for the doping process for Li or Na, respectively. MnO 2 nanotubes were first heated in a diluted HCl solution for more than 24 hrs at 413 K. The amount of residual K in the MnO 2 tubes was reduced and eventually could not be detected by energy disperse X-ray spectroscopy (EDX) attached to a scanning electron microscope. After soaking the nanotubes in either 1 M NaOH or 1 M LiOH, K ions could not be detected by ICP. The concentration of Li was subsequently measured with ICP. In this work, we prepared Li/Na doped samples with a concentration of approximately 6 at% by controlling the soaking time. X-ray diffraction (XRD, PANalytical Xpert Multipurpose X-ray Diffraction System, Cu Ka radiation), scanning electron microscopy (SEM, FEI Nova NanoSEM 230) and transmission electron microscopy (TEM, JEM-2010, JEOL) were used for the characterization of phases and microstructures. A superconducting quantum interference device (SQUID, Quantum Design XL-5) was used for magnetic property measurements. X-ray absorption fine structure (XAFS) spectra were measured in transmission mode at the XDD beamline at Singapore Synchrotron Light Source (Singapore).
First principles calculations. First-principles calculations were performed using density functional theory from Vienna ab initio simulation package with a plane wave basis 32 . The generalized gradient approximation (GGA) with spin-polarized Perdew-Burke-Ernzerhof (PBE) 33,34 scheme was employed for calculating the exchange and correlation functional. The core electrons were represented by the projectoraugmented-wave (PAW) potential. Kinetic energy cutoff was set at above 400 eV and k-point sampling on the unit cell was 232310. The structure optimization was performed with the criteria of force convergence at 0.01 eV/Å . The optimized lattice constant for a-MnO 2 is a5b50.96 nm and c50.28 nm, which is in agreement with previous experimental measurements 35-37 . This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder in order to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/ www.nature.com/scientificreports SCIENTIFIC REPORTS | 5 : 9094 | DOI: 10.1038/srep09094