Optically switched magnetism in photovoltaic perovskite CH3NH3(Mn:Pb)I3

Abstract

The demand for ever-increasing density of information storage and speed of manipulation boosts an intense search for new magnetic materials and novel ways of controlling the magnetic bit. Here, we report the synthesis of a ferromagnetic photovoltaic CH₃ NH₃ (Mn:Pb)I₃ material in which the photo-excited electrons rapidly melt the local magnetic order through the Ruderman-Kittel-Kasuya-Yosida interactions without heating up the spin system. Our finding offers an alternative, very simple and efficient way of optical spin control, and opens an avenue for applications in low-power, light controlling magnetic devices.

ARTICLE

Received 20 Nov 2015 | Accepted 30 Sep 2016 | Published 24 Nov 2016

DOI: 10.1038/ncomms13406 OPEN

Optically switched magnetism in photovoltaic perovskite CH3NH3(Mn:Pb)I3

B. Nfrdi1, P. Szirmai1, M. Spina1, H. Lee2, O.V. Yazyev2, A. Arakcheeva1, D. Chernyshov3, M. Gibert4, L. Forr1

& E. Horvth1

The demand for ever-increasing density of information storage and speed of manipulation boosts an intense search for new magnetic materials and novel ways of controlling the magnetic bit. Here, we report the synthesis of a ferromagnetic photovoltaic CH3NH3(Mn:Pb)I3 material in which the photo-excited electrons rapidly melt the local magnetic order through the RudermanKittelKasuyaYosida interactions without heating up the spin system. Our nding offers an alternative, very simple and efcient way of optical spin control, and opens an avenue for applications in low-power, light controlling magnetic devices.

1 Laboratory of Physics of Complex Matter, Ecole Polytechnique Fdrale de Lausanne (EPFL), CH-1015 Lausanne, Switzerland. 2 Institute of Physics, Ecole Polytechnique Fdrale de Lausanne (EPFL), CH-1015 Lausanne, Switzerland. 3 Swiss-Norwegian Beam Lines, European Synchrotron Radiation Facility,71 Avenue des Martyrs, F-38043 Grenoble Cedex 9, France. 4 DQMPUniversity of Geneva, 24 Quai Ernest Ansermet, CH-1211 Geneva 4, Switzerland. Correspondence and requests for materials should be addressed to B.N. (email: mailto:[email protected]

Web End [email protected] ) or to L.F. (email: mailto:[email protected]

Web End [email protected] ).

NATURE COMMUNICATIONS | 7:13406 | DOI: 10.1038/ncomms13406 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 1

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13406

The mechanism of magnetic interactions and eventually the magnetic order in insulating and conducting materials are fundamentally different. Diluted localized magnetic (M)

ions in insulating materials commonly interact over extended distances by the super-exchange (SE) interaction via atomic orbital bridges through nonmagnetic atoms, for example, oxygen,O. Common schemes for interactions in perovskite structures are the MOM or MOOM-like bridges. The strength and sign (antiferromagnetic/ferromagnetic (AFM/FM)) of these interactions are determined by the geometry of the bonds. Thus, the in situ ne-tuning of the interactions is usually difcult because it would call for structural alterations. A limited continuous change is possible by application of pressure1. Discrete changes in the lattice are achieved by chemical modications like replacing the bridging element with halides creating MClM, MBrM or MIM bonds2.

In conducting hosts the long-range double-exchange or the RudermanKittelKasuyaYosida (RKKY) interactions also come into play between the magnetic M ions. For the RKKY interaction, the key control parameters are the density of the localized moments and the density of itinerant electrons. The RKKY coupling strength oscillates between AFM or FM as a function of the MM distance and of the size of the Fermi surface. These parameters, however, similarly to the case of the SE, are intrinsic to the system and in situ modications are not feasible.

New magnetic materials and efcient, faster ways of controlling the magnetic bit are continuously searched for in order to sustain the needs for ever-increasing density and speed of information storage and manipulation37. Technologically relevant materials emerge when magnetic interactions of localized and itinerant spins are simultaneously present and compete in determining the ground state. This competition is usually controlled by the carrier concentration and a small external perturbation may result in an extremely large change, for instance, in resistivity. A well-known example is (La:Sr)MnO3 perovskite where at ne-tuned chemical substitutions ferromagnetic double-exchange interactions mediated by chemically doped electrons compete with the antiferromagnetic SE interaction of the parent insulating compound. Consequently, a metal-insulator transition and a ferromagnetic order develops8. Mechanical and electronic control of the carrier concentration and of the magnetic transition was also shown1,9,10.

Here we demonstrate an alternative way of controlling the competition of magnetic interactions between itinerant and localized electrons by using visible light illumination. By virtue of photodoping we modulate the carrier concentration and thus the magnetic order in the magnetic photovoltaic perovskite CH3NH3(Mn:Pb)I3. This method presents considerable advantages over chemical doping since it is continuously tuneable by light intensity, spatially addressable by moving the illuminating spot and, last but not least, provides a fast switching time (in the ns range required for relaxation of photo-excitations11,12). The observed optical melting of magnetism could be of practical importance, for example, in a magnetic thin lm of a hard drive, where a small magnetic guide eld will trigger a switching of the ferromagnetic moment into the opposite state via the light-induced magnetization melting. This kind of ferromagnetic moment reversal is rapid and represents several indisputable advantages over other optical means of manipulation of the magnetic state reported earlier3,1319. The central ingredient is a high-efciency photovoltaic material which orders magnetically. Taking advantage of the outstanding light-harvesting characteristics20 and chemical exibility21 of the organometallic perovskite CH3NH3PbI3 (hereafter MAPbI3), which has recently triggered a breakthrough in the eld of photovoltaics, we have developed a magnetic photovoltaic perovskite CH3NH3(Mn:Pb)I3

(hereafter MAMn:PbI3), (Fig. 1) by substituting 10% of Pb2 ions with Mn2 ions. This material provides a unique combination of ferromagnetism (TC 25 K) and high efciency

of photoelectron generation. It turns out that in our material these two properties are intimately coupled, thus optical control of magnetism is achieved. Furthermore, we expect this mechanism to be universally present in other magnetic photovoltaics, as well.

ResultsMagnetic properties in dark. The substitution of Mn2 ions into the MAPbI3 perovskite network is revealed by synchrotron powder X-ray diffraction and energy-dispersive X-ray measurements (Supplementary Fig. 1, Supplementary Table 1, and Supplementary Fig. 2). Mn2 ions in the host lattice are isoelectronic with Pb2 . Hence, they do not dope the system as also conrmed by our rst-principles electronic structure calculations discussed below. The substituted sample is semi-conducting in dark with few MOcm resistivity similarly to the parent compound. Moreover, the high level of Mn substitution does not diminish the photocurrent (Iph) generation. A strong Iph response is observed below 830 nm wavelength (Supplementary Fig. 3). It is worth mentioning that the optical gap decreased relative to the pristine material, which facilitates photovoltaic applications. The photocurrent and thus the carrier density can be ne-tuned by the incident light intensity in broad frequency and intensity ranges. The Mn substitution, however, dramatically modies the magnetic properties of the system as seen by electron spin resonance (ESR) measurements performed in an exceptionally broad 9315 GHz frequency range (Figs 2 and 3, Supplementary Figs 46)22,23. The pristine material is nonmagnetic, only ppm level of paramagnetic impurities could be detected. As expected, Mn substitution gives an easily observable signal. At low concentration, ESR shows well-resolved

MW detector

a

c

MW source

Sample

Switch

b

LED

MW mirror

Magnet

y

z

x CH3NH3

Pb Mn I

Figure 1 | Sample and measurement conguration. (a) Photo of a typical CH3NH3(Mn:Pb)I3 crystal, 1015 were assembled for the ESR measurement.

(b) Sketch of the crystal structure of CH3NH3(Mn:Pb)I3. (c) The experimental conguration for the high-eld ESR measurements showing the assembly of small crystals (sample). The absorption of the microwave eld provided by the microwave source (MW source, up to 315 GHz) is monitored (MW detector) in resonant conditions in dark and under illumination in reection geometry (MW mirror). The light source is a red (l 655 nm,

4 mWcm 2) LED activated by an external switch (switch).

2 NATURE COMMUNICATIONS | 7:13406 | DOI: 10.1038/ncomms13406 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13406 ARTICLE

a b c d

3.18

3.21

250

Total

I p

3 Pb p

Mn d

3.20

3.19

200 20

4

B 0(100 K)-B 0 (mT)

3.22

150

3.20

EE VBM(eV)

148.68

165.92154.33

164.19

2 1 0 1 2 3 4

3.42

2.88

Majority spin Minority spin

B(mT)

3.14

100

10

0

2.99

2.97

2.78

3.60

50

0 0 40 80 120 160 200

T (K)

y

240 280

2.94

DOS/PDOS (a.u.)

z

x

Pb Mn I

Figure 2 | Magnetic properties of CH3NH3(Mn:Pb)I3 in dark. (a) ESR linewidth (red dots) and resonant eld (blue dots, offset by a reference value B0) as a function of temperature recorded at 9.4 GHz. Their temperature-independent behaviour is characteristic for the paramagnetic phase (PM). The upturn below 25 K corresponds to the on-set of the FM phase. Inset: SQUID magnetometry of MAMn:PbI3. The temperature dependence of the spontaneous magnetization measured in 1.2 mT magnetic eld shows a clear increase below TC. The red line represents the M0(1 (T/TC)3/2) temperature dependence

given by Blochs Law. (b) First-principles calculations of the atomic congurations and magnetic order show total density of states (DOS) and projected density of states (PDOS) calculated for the in-plane model of CH3NH3(Mn:Pb)I3 in its neutral FM conguration. (c) Calculated PbI and

MnI distances for a single Mn dopant. (d) Calculated bond angles and bond distances for the I mediated super-exchange paths in the FM ground state of the in-plane model of CH3NH3(Mn:Pb)I3.

hyperne lines indicating the uniform dispersion of Mn2 ions24 (Supplementary Fig. 5). The MAMn:PbI3 sample shows a strong ESR signal (Fig. 2 and Supplementary Fig. 5) and most importantly, a ferromagnetic order developing below TC 25 K

on cooling in dark. The ferromagnetic order causes a rapid shift of the resonant eld, B0, and the broadening of the linewidth, DB, below TC (Fig. 2a and Supplementary Fig. 6) which are sensitive measures of the magnetic interactions and the internal magnetic elds25. The absence of additional ESR lines in the entire 9315 GHz frequency range indicates that the magnetic order is homogeneous, the MAMn:PbI3 material is free of secondary phases corroborating with the structural renement.

Static magnetization measurements by SQUID (inset to Fig. 2a and Supplementary Fig. 7) conrm the observations of ESR. The ground state is ferromagnetic as spontaneous magnetization Mspontaneous appears below TC in the same temperature range

where the ESR line shifts and broadens. Moreover, a small magnetic hysteresis appears at low temperatures indicated by the nite remanent magnetization (Supplementary Fig. 7). The deviation of Mspontaneous from the mean-eld description (Blochs

Law) in the TC to 2TC temperature range is characteristic to diluted magnetic systems indicating the phase homogeneity of the system. Temperature and eld dependence of M also revealed dominant ferromagnetic correlations at high temperatures, and the presence of magnetocrystalline anisotropy, K1 380 104 J m 3,

below TC (Supplementary Fig. 7). The appearance of ferromagnetic order stabilized by short-range SE interactions in the insulating sample at 10% doping levels indicates that both MnIMn and MnIIMn interactions are active to exceed percolation limits26. It should be emphasized that the homogeneous magnetic ordering itself in this insulating photovoltaic perovskite is already a remarkable observation. Such ordering was extensively searched for in homogeneously diluted magnetic semiconductors, and unambiguously observed only in few cases24,2729.

This surprising FM order is supported by a rigorous density functional theory (DFT) calculations (see the Methods section for calculation details). The model of MAMn:PbI3 was constructed starting from the experimentally determined low-temperature orthorhombic (Pnma) crystal structure of undoped material30, which was then extended to the 2 1 2

supercell. Two Pb atoms in the supercell were replaced by Mn

atoms to allow investigation of the exchange interactions between Mn dopants. Overall, one Pb atom of eight was substituted, which corresponds closely to the 10% doping concentration of experimentally investigated samples. Three different arrangements of Mn dopants were studied and are shown in Supplementary Fig. 8.

The energy differences between the FM and AFM congurations are of the order of 1020 meV, while the interaction sign varies across the studied models. We found that for the in-plane model (model 2 in Supplementary Fig. 8), the FM conguration is the ground state, which is 10.9 meV lower in energy compared with the AFM conguration. The density of states plot calculated for the charge-neutral conguration of in-plane model shows that Mn2 impurities substituting Pb2 ions do not give rise to charge-carrier doping and do not induce any mid-gap states (Fig. 2b). The FM interaction is the consequence of the strongly distorted orthorhombic perovskite structure with MnIMn bond angle signicantly reduced to B150 (Fig. 2c,d).

Melting ferromagnetic order by photoelectrons. Our major nding is a striking change of the magnetism when the sample is exposed to light illumination at wavelengths lower than the bandgap, ledge 830 nm (Supplementary Fig. 3). Typical ESR

absorption spectra taken by light-off and light-on (0.8 mW cm 2) at T 5 K are shown in Fig. 3c. The light-on spectrum is con

siderably narrower and of weaker intensity than the spectrum in dark. The difference between light-on and light-off signals is shown in orange. For the given light intensity, 25% of the initial spin-susceptibility disappears (wESR) on light exposure.

In a broad range of illumination intensities, after a threshold value, one can observe a monotonous decrease of the FM part of the signal (Fig. 3a). Presumably, below the threshold the photoelectrons ll up some trap sites. At larger intensities, they start to melt the FM state. (The same tendency is observed for DB versus illumination intensity, see Supplementary Fig. 9). The change is completely reversible. As wESR is directly proportional to the ferromagnetic volume, the results demonstrate that in a large part of the sample the ferromagnetic order is melted by light illumination. This effect could be closely followed in temperature, as well. The difference between the light-on and light-off signals

NATURE COMMUNICATIONS | 7:13406 | DOI: 10.1038/ncomms13406 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 3

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13406

a

b

1.1

1.0

0.9

0.8

0.7

0.6

0.5

T=5 K

1.00

[afii9851](light-on)/[afii9851](light-off)

B(light-on)/B(light-off)

0.95

0.90

0.4 0.01 0.1 1

CH3NH3Pbl3

CH3NH3(Mn:Pb)I3 Light-off

(Light-off)(Light-on)

Light-on

10

0.85 0 20 40 60 T (K)

80 100

(W cm2)

c d

T=5 K

A(a.u.)

A(light-on)-A(light-off)

TC

6.2

5.85

5.5

5.0 5.5 6.0 B (T)

4.5

6.5

T (K)

100 80 60 40 20 0 4.8

5.15

B (T)

Figure 3 | Illumination effect on the magnetic properties of CH3NH3(Mn:Pb)I3 measured by ESR. (a) The intensity change as the function of the illuminating red light intensity F at T 5 K. Above a threshold value, the FM part of the signal decreases monotonously. (b) Light-on ESR linewidth

normalized to the linewidth in dark. The narrowing of the linewidth on illumination starts below TC. (c) ESR spectra at 157 GHz and 5 K of pristine CH3NH3PbI3 (green lineno signal), of CH3NH3(Mn:Pb)I3 in dark (blue line) coming from the FM phase and its reduction (red line) on visible light illumination. The difference between light-off and light-on ESR signals is shown in orange. The effect is accompanied by narrowing of the ESR linewidth on illumination. (d) Difference of the ESR intensities between the light-off and light-on measurements as a function of temperature. (The third axis shows the resonant eld of the signal.) The intensity reduction on illumination is present only below TC 25 K, in the FM phase. Error bars represent the condence

interval of least square ts to the spectra.

both in DB and in wESR vary up to TC (Fig. 3b). The narrowing of DB in the remaining magnetic signal, only observed below TC (Fig. 3 and Supplementary Fig. 10), is a consequence of the surface melting of the magnetic order, as it is not accompanied by change of B0. The ferromagnetic DB is a strong function of sample shape and size. The light is absorbed in the rst few microns of the crystals11 where the FM is molten so the created magnetic core-shell structure effectively changes the morphology of the sample, thus DB. The observations shown in

Fig. 3b allow us to exclude heating effect by the light-emitting diode (LED) which means that this is an athermal magnetic change induced by photo-excited conduction electrons in the insulating magnetic phase.

Model of the melting of the ferromagnetic order. On the qualitative basis, one can interpret the light-induced melting of the magnetic structure as the competition between the SE-and the light-induced RKKY-interactions31. SE orders the entire sample magnetically in dark. It is known that halide bridges can mediate the interaction between localized Mn2

moments by SE in insulating perovskite crystals32. Illumination creates conduction electrons that alter the spin order established by SE as described by the RKKY Hamiltonian33. This mechanism is generic to all insulating magnets, where a high-efciency photoelectron generation is present.

Electrical transport measurements in MAMn:PbI3 support this qualitative interpretation. The highly crystalline insulating sample with MOcm range and thermally activated resistivity (not shown)

transforms even into a metal-like state by the low-intensity red light illumination in a broad temperature range promoting the RKKY interaction. The quadratic magnetoresistance together with the resistivity indicates that even at low temperatures the photoinduced free carrier concentration exceeds nB2 1017

cm 3. Furthermore, the weak, negative magnetoresistance in 02.5 T range (above this eld a positive, orbital contribution is observed) shows the coupling of conduction electrons to the magnetic moments.

The idea to model the melting of the FM order by photoelectrons was to consider the competition between the SE- and the RKKY-interactions. This has been performed by DFT

4 NATURE COMMUNICATIONS | 7:13406 | DOI: 10.1038/ncomms13406 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13406 ARTICLE

calculations (see the Methods section for calculation details). Technically, the effect of photo-excited charge carriers was addressed by considering separately electron- and hole-doped models since excitons cannot be described by DFT. Upon doping the in-plane conguration (Supplementary Fig. 8) with2.6 1020 cm 3 concentration charge carriers, the ground state

changes from FM to AFM with relative energies of 20.4 and10.9 meV for one hole and for one electron per supercell, respectively. The corresponding total and projected density of states plots for the doped models in their AFM state are shown in Supplementary Fig. 11. We have to mention that the carrier concentration used in the modelling is much higher than the measured photoelectron concentration, but the purpose of our calculations is to demonstrate the suppression of the FM order. In fact, calculations with an order of magnitude lower carrier concentrations gave qualitatively identical results.

DiscussionAs an outlook, the observed optical melting of magnetism could be of practical importance, for example, in a magnetic thin lm of a hard drive, where a small magnetic guide eld will trigger a switching of the ferromagnetic moment into the opposite state via the light-induced magnetization melting. Its principle is illustrated in Fig. 4. This kind of ferromagnetic moment reversal is rapid and represents several indisputable advantages over other optical means of manipulation of the magnetic state reported earlier3,1319. It does not require high-power or femtosecond laser instrumentation, which, besides the complexity of the techniques, raises the stability issue due to photochemistry and fatigue coming from the high intensity and the rapid local thermal cycling of the material16. Our method needs only a low-power visible light source, providing isothermal switching, and a small magnetic guide-eld to overcompensate the stray eld of neighbouring bits. Although this is a simple and elegant method for magnetic data storage, it has never been discussed in literature, because magnetic photovoltaic materials have not been developed.

We have shown the extension of photovoltaics into magnetism by preparing a ferromagnetic MAMn:PbI3. It has been demonstrated that the high-efciency photocurrent generation by low-power visible light illumination results in a melting of the ferromagnetic state and a small local eld can set the direction of the magnetic moment. It should be emphasized that this mechanism is radically different from switching the orientation of magnetic domainshere the photoelectrons tune the local interaction between magnetic moments and thus change the

magnetic ground state. This study provides the basis for the development of a new generation of magneto-optical data storage devices where the advantages of magnetic storage (long-term stability, high data density, non-volatile operation and re-writability) can be combined by the fast operation of optical addressing. Such a technology should be developed with thin lms with higher TC (which is by far a non-trivial challenge)

where the total melting of the magnetism in MAMn:PbI3 could be achieved on illumination. Last but not least, this study highlights that besides photovoltaics, lasing and LED operation there is one more extraordinary feature of the CH3NH3PbI3 perovskite material.

Methods

Sample preparation. CH3NH3(Mn:Pb)I3 single crystals were prepared by precipitation from a concentrated aqueous solution of hydriodic acid (57 w% in H2O, 99.99% Sigma-Aldrich) containing lead (II) acetate trihydrate (99.999%,

Acros Organics), manganese (II) acetate tetrahydrate (99.0%, Fluka) and a respective amount of CH3NH2 solution (40 w% in H2O, Sigma-Aldrich). The solubility of the Pb- and Mn-acetate provides indirect evidence of the homogeneous distribution of the Mn dopants. A constant 5542 C temperature gradient was applied to induce the saturation of the solute at the low temperature part of the solution21. Besides the formation of hundreds of submillimeter-sized crystallites (polycrystalline powder) large aggregates of long MAMn:PbI3 needle-like crystals with 520 mm length and 0.1 mm diameter were grown after 7 days (Fig. 1). Leaving the crystals in open air resulted in a silver-grey to green-yellow colour change. To prevent this unwanted reaction with moisture the as synthesized crystals were immediately transferred and kept in a desiccator prior to the measurements. Millimetre size undoped (CH3NH3PbI3) single crystals were also synthesized and kept as a reference material for qualitative analysis.

Synchrotron X-ray powder diffraction (XRD). Synchrotron XRD pattern of the CH3NH3(Mn:Pb)I3 sample was measured at room temperature at the

SwissNorwegian beam lines of the European Synchrotron Radiation Facility (ESRF). The wavelength of the used synchrotron radiation was 0.9538 . All data were collected in the DebyeScherrer geometry with a Dectris Pilatus2M detector. The sample-to-detector distance and the detector parameters were calibrated using a LaB6 NIST reference powder sample. CH3NH3(Mn:Pb)I3 powders were placed into 10 mm glass capillaries and mounted on a goniometric spinning head. For Rietveld renement Jana crystallographic programme was used. Crystal structure was rened in I4/mcm tetragonal space group. Rened atomic parameters of Pb, I, C, and N are very similar to those published for CH3NH3PbI3 (ref. 34). In addition, H atoms were also localized. The XRD prole together with the results of the Rietveld prole tting is shown in Supplementary Fig. 1 and in Supplementary Table 1. The synchrotron X-ray diffraction proles revealed a sample without observable secondary phases or phase inhomogeneity.

Scanning electron microscope. Scanning electron microscope images were taken with a MERLIN Zeiss electron microscope. Individual single needle-like crystallites were broken off from the rod-like bundles of MAMn:PbI3 for scanning electron microscope micrographs (Supplementary Fig. 2). Aluminium pucks were used for sample support. Conducting carbon tape served as electric contact between the sample and the support.

Energy-dispersive X-ray spectroscopy (EDS). The elemental composition of the MAMn:PbI3 crystallites were analysed by EDS (accelerating voltage of 8 kV, working distance of 8.5 mm). Samples were mounted on Al pucks with carbon tape with electrical contact to the surface also formed by carbon tape. The measurement was performed with an X-MAX EDS detector mounted at a 35 take-off angle with a SATW window. EDS spectra were obtained at a working distance of 8.5 mm with 8 keV accelerating voltage and a current held at 184 pA. 2,048 channels were used for the acquisitions, corresponding to energy of 5 eV per channel. Spectra were acquired over 1,573 s of live time with detector dead time averaging of 4% and a dwell time per pixel of 500 ms. Quantitative EDS analysis utilized Aztec software provided by Oxford Instrument Ltd.

To obtain information on the homogeneity of Mn substitution of the MAMn:PbI3 crystals, EDS were performed on several positions on the as-grown surface of the needle-like MAMn:PbI3 crystallites. For the purpose of gathering bulk information as well EDS spectrum were taken also on broken-off surfaces. These experiments systematically yield (Mn0.1Pb0.9)I3 stoichiometry indicating homogeneous Mn substitution.

Electron spin resonance spectroscopy. Polycrystalline assembly of 1015 rod-like MAMn:PbI3 samples with typical 1 0.1 0.1 mm are sealed in a quartz

capillary. ESR at 9.4 GHz microwave frequency was performed on a Bruker X-band

a b c

Figure 4 | Schematic illustration of writing a magnetic bit. In the dark (a) the spin alignment corresponds to a given orientation of the magnetic moment in the FM state, representing a bit. On illumination (b) the FM order melts and a small magnetic eld of the writing head will set the orientation of the magnetic moment once the light is switched off (c).

NATURE COMMUNICATIONS | 7:13406 | DOI: 10.1038/ncomms13406 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 5

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13406

spectrometer. A conventional eld modulation technique was employed with lock-in detection which results the rst derivative of the ESR absorption spectra. Experiments in the mm-wave frequency range were performed on a home-built quasi-optical spectrometer operated at 75, 105, 157, 210 and 315 GHz frequencies in 016 T eld range (Fig. 1). The spectral resolution of ESR is linearly proportional to the frequency, thus we extended the precision of ESR by about a factor 30 compared with X-band ESR technique. More details about the set-up can be found inrefs 22,23. A red LED was placed underneath the sample as a light source. Magnetic eld strength at the sample position was calibrated against a KC60 standard sample.

In contrast to the low-eld ESR experiments, at millimetre-wave frequencies microwave power chopping was combined with lock-in detection. This detection scheme results directly the ESR absorption signal instead of its rst derivative. The working principles of the two methods are shown in Supplementary Fig. 4.

Supplementary Fig. 5 compares pristine MAPbI3 with 1% and 10% substituted MAMn:PbI3 at room temperature. Pristine MAPbI3 crystals show no intrinsic ESR signal. Only low, ppm levels of paramagnetic impurity centres were observed (Fig. 3 and Supplementary Fig. 5). In contrast, Mn substitution to MAMn:PbI3 results in a strong ESR signal.

The presented ESR experiments prove that the magnetic transition is not driven by temperature change. ESR unambiguously demonstrates that we are not dealing with a temperature effect. By ESR at each temperature one obtains the spin susceptibility, ESR linewidth and resonance eld simultaneously. All three parameters are strongly temperature-dependent as shown in Fig. 2 and Supplementary Fig. 6. Temperature change modies all the three parameters concurrently. Our careful ESR experiments performed in dark provide us an internal thermometer. Increasing the temperature by B1 K would change all three aforementioned ESR parameters simultaneously. Switching on the light does not show this effect as demonstrated in Fig. 3. It changes the spin susceptibility, ESR linewidth and resonance eld by an amount that corresponds to different temperature changes. This cannot be explained by a temperature effect. This shows unambiguously that we are not dealing with a temperature effect. Accordingly, the phenomenon we discovered is an athermal effect. Instead, we would like to point out that our ESR experiments are demonstrating the change of Curie temperature with photo-excitation. The spin susceptibility, measured by the ESR intensity, at ToTC decreases by B25% on light illumination (Fig. 3b,d). This demonstrates the disappearance of 25% of the FM volume upon illumination. It means that in that 25% volume the TC decreased from25 K to below 5 K, the lowest temperature in our experiments.

The spectra at 1% Mn2 concentration consist of two signals. One set of sextet lines and an about 50 mT broad line (Supplementary Fig. 5). The sextet signal is characteristic of a hyperne splitting of 55Mn with g 2.001(1) g-factor and

Aiso 9.1 mT hyperne coupling constant24. This spectrum corresponds to both

allowed (sextet) and forbidden (broad component) hyperne transitions between the Zeeman sublevels. It is characteristic to Mn2 ions in octahedral crystal elds.

Since strong forbidden transitions are observed, Mn2 ions do not occupy strictly cubic sites, as strictly cubic centres have zero probability of forbidden transitions, rather distorted octahedral sites. The well-resolved hyperne also testies the homogeneous distribution of Mn ions in MAMn:PbI3 (ref. 24).

These ESR characteristics are in good agreement with both powder X-ray diffraction and DFT calculations showing distorted octahedral Mn coordination. The ESR spectra of MAMn:PbI3 at high Mn2 concentration (10%) consist of one broad ESR line only. This is a common resonance of both allowed and forbidden transitions. We explain the uniformity of the g-factor by strong exchange narrowed spinorbit interaction dominated linewidth of the Mn2 ions. Following the calculations of ref. 35 and assuming a spinorbit width contribution of the order of (Dg/g)J, yields a value of the order of 100 K for exchange integral J.

The broad ESR and isotropic g-factor is strongly intrinsic for the system. We nd no evidence of frequency dependence at high temperatures in the 9315 GHz frequency range. The eld- and temperature-independent DB and B0 is

characteristic to exchange coupled paramagnetic insulators. Below 25 K, both DB and B0 acquires strong temperature dependence indicative of ferromagnetic ordering. The shift in B0 measures the temperature dependence of the internal ferromagnetic eld of MAMn:PbI3. DB scales to B0 at all measured elds and temperatures (Fig. 2 and Supplementary Fig. 6) indicating an inhomogeneous broadening induced by spatial distribution of the local internal ferromagnetic eld. The inhomogeneity of the local internal ferromagnetic eld is partially of geometrical origin. The demagnetizing eld of our irregularly shaped particles is inhomogeneous. Additionally, the statistical uctuations of the Mn concentration across the sample also increase the inhomogeneity by modulating the strength of the ferromagnetic order.

The magnetic phase purity can be further conrmed by comparing the two described ESR method. The microwave chopping method (Supplementary Fig. 4b), which yields the integrated ESR signal, would reveal possible broad ESRsignals. However, Fig. 3c,d show the absence of any broad magnetic impurity signals. The magnetic eld modulation method (Supplementary Fig. 4a) would help to identify narrow signals with a linewidth in the order of the modulation. Supplementary Fig. 5f proves the absence of the narrow impurity signals as well.

We note here that the signal of itinerant electrons generated by the illumination is not detected, either. Detection of a so-called conduction ESR (CESR) line would be a major challenge (see, for example, refs 24,36). The two main difculties in the order of the importance are: (i) low Pauli spin susceptibility of a CESR signal, and(ii) the spinorbit coupling provokes a broadening in the signal37,38.

In our system, the presence of conduction electrons can be excluded in dark, as MAMn:PbI3 is an insulator without light. We would only observe the CESR signal in the presence of photoexcited carriers on illumination. As seen in Fig. 3c and in Supplementary Fig. 5f, however, we do not observe the CESR on illumination.

In our case, both issues of CESR detection are signicant. The weak illumination results in the small spin-susceptibility of the generated conduction electrons. In fact, the expected spin susceptibility of the CESR (Pauli susceptibility) of the photoexcited state is 56 orders of magnitude smaller than the paramagnetic Mn2 ESR signal. Furthermore, the large spinorbit coupling broadens this small signal. These two effects prevent the observation of the CESR.

Furthermore, the precursor Mn-acetate used for the Mn substitution has markedly different ESR spectra from the substituted material, thus inclusions of Mn-acetate islands can be excluded, as well.

SQUID magnetometry. SQUID magnetometry experiments reveal that the temperature dependence of spontaneous magnetization, the dening macroscopic property of ferromagnetism, appears below 50 K and dramatically enhances below25 K (Fig. 2). The theoretical behaviour of the paramagnetic magnetization in the same conditions is shown by the green line. Clearly, the spontaneous magnetization is orders of magnitude greater compared with a paramagnetic magnetization, testifying the ferromagnetic order.

The mean-eld theory of spontaneous magnetization is described by Blochs Law, which states that MspontaneousB1 (T/TC)a with a 3/2 (red line in Fig. 2).

However, deviations from the mean-eld exponent are recurrent, for example, Iron and Nickel show critical exponents a of 0.34 and 0.51, respectively. Similarly, the spontaneous magnetization in MAMn:PbI3 deviate from the mean-eld value (Fig. 2). The deviation from Blochs Law is indicative of the presence of strong magnetocrystalline anisotropy. The primary source of magnetocrystalline anisotropy is the spinorbit interaction, which is strong due to the involvement of Pb and I atoms.

In Supplementary Fig. 7a, we show the temperature dependence of the magnetization cooled in 1 T external eld. In agreement with the appearance of the remanent magnetization in the zero eld-cooled experiments, we nd a CurieWeiss temperature of TCW 14 K characteristic to predominant ferromagnetic

correlations. At low temperatures, however, the magnetization is suppressed relative to the isotropic CurieWeiss behaviour. This is characteristic to the presence of magnetocrystalline anisotropies in perfect agreement with the observed deviation of the spontaneous magnetization from the mean-eld description.

The magnetic eld dependence of the magnetization measured at T 2 K up to

7 T magnetic eld (see Supplementary Fig. 7b) shows a steady increase of magnetization with about HS 2K1/MS 9 T saturating magnetic eld. This again

underlines the presence of magnetic anisotropy of K1 380 104 J m 3 at T 2 K.

Note that this value is in the same range as those found at room temperature in haematite (K1 120 104 J m 3) and for YCo5 (K1 550 104 J m 3).

Finally, the temperature dependence of the remanent magnetization measured by decreasing the magnetic eld from 7 T shows small value in agreement with the magnetization isotherms. The temperature behaviour is similar to the behaviour of the spontaneous magnetization, and it increases below 50 K (inset of Fig. 2).

These SQUID experiments undoubtedly reveal the existence of magnetic order of our MAMn:PbI3 compound. It also shows high magnetic phase purity. No sign of additional magnetic or nonmagnetic phase was detected in perfect agreement with the multi-frequency ESR investigations.

Photocurrent spectroscopy. For photocurrent spectra a low-intensity monochromatic light was selected by a MicroHR grid monochromator from a halogen lamp. The wavelength resolution (full width at half maximum; FWFM) of the 600 g mm 1 grating was 10 nm. The photo-excited current was measured by a two-terminal method at xed bias voltage of 1 V while the wavelength was stepwise changed (Supplementary Fig. 3). Measurements were performed on pristine MAPbI3 and on MAMn:PbI3. The bandgap energy was determined by tting a

FermiDirac distribution to the data. The resulting gap energies at room temperature are 7831 nm and 8291.4 nm for the MAPbI3 and MAMn:PbI3, respectively. The intrinsic width of the FermiDirac distribution for both systems is thermally broadened. This indicates that the Mn substitution is homogeneous. Mn clustering would cause broadening of the band edge. It is also worth mentioning the strong, B46 nm upshift of the band edge on Mn substitution since the gap of

MAMn:PbI3 is reduced relative to MAPbI3, Mn substitution presents an alternative route to extend the light absorption range, hence increase photocell efciencies. The temperature dependence of the photocarrier generation in 50300 K temperature range was also studied in a closed-cycle cryostat equipped with an optical window (Supplementary Fig. 3). The gap energy increases by decreasing temperature due to thermal expansion, however, the photocarrier generation of MAMn:PbI3 remains effective down to the lowest studied temperatures.

First-principles electronic structure calculations. To corroborate the experimental ndings, we carried out rst-principles electronic structure calculations in the framework of DFT39,40 as implemented in the Quantum ESPRESSO package41. The exchange-correlation energy is given by the PerdewBurkeErnzerhof generalized gradient approximation42 while the electronion interactions are

6 NATURE COMMUNICATIONS | 7:13406 | DOI: 10.1038/ncomms13406 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13406 ARTICLE

treated by using the ultrasoft pseudopotentials43 which have been previously published44. Wave functions and charge densities are expanded using the plane-wave basis sets with kinetic energy cutoffs of 40 and 320 Ry, respectively. The Brillouin zone is sampled using 3 4 3 Monkhorst-Pack meshes of special

k-points45. The plane-wave cutoffs and k-point meshes are chosen to ensure the convergence of total energies within 10 meV. When performing calculations on charged models, a compensating jellium background was introduced in order to avoid the spurious divergence of electrostatic energy46.

The models of Mn-doped CH3NH3PbI3 were constructed starting from the experimentally determined crystal structure of undoped material (orthorhombic phase, space group Pnma)30, which was then extended to the 2 1 2 supercell by

doubling the lattice constants along the a and c directions. Two Pb atoms in the supercell were replaced by Mn atoms to allow investigating the exchange interactions between Mn dopants. Overall, one Pb atom of eight was substituted, which corresponds closely to the doping concentration of experimentally investigated samples (10%). Three different arrangements of Mn dopants, referred to as top, in-plane, and diagonal, are shown in Supplementary Fig. 8. Atomic coordinates of all these three congurations were optimized to the residual ionic forces smaller than 0.02 eV 1, whereas the lattice parameters were kept xed.

For each conguration both the FM and AFM arrangements of local magnetic moments of Mn atoms were investigated. Our calculations show that optimization of the internal atomic coordinates is crucial for reproducing the relative energies of FM and AFM congurations. Indeed, substitution of Mn atoms for Pb atoms leads to a pronounced lattice distortion around the Mn dopants due to different ionic sizes of Mn2 and Pb2. Specically, the MnI distances are about 2.9 , whereas the PbI distances are about 3.2 (Fig. 2c,d).

For all considered arrangements of Mn dopants, the energy differences between the FM and AFM congurations are of the order of 1020 meV. We found that for model 2 (in-plane, Supplementary Fig. 8), the FM conguration is the ground state, which is 10.9 meV lower in energy compared with the AFM conguration. Due to intrinsic limitations of DFT calculations, the effect of photoexcited charge carriers was addressed by considering separately electron- and hole-doped models. One has to emphasize that the DFT calculations correspond to a 0 K case and xed number of photoelectrons. At nite temperatures and variable carrier density between the FM and AFM congurations it is reasonable to expect a paramagnetic state as seen in the experiment.

DC resistivity and magnetotransport under illumination. DC resistivity and magnetotransport under illumination were performed with the same light conditions as the ESR experiments. Resistivity and magnetoresistance were measured in a standard 4-terminal conguration in the 5300 K temperature and 016 T magnetic eld range. In dark, the resistivity of the samples is in the MOcm range and show thermally activated character (not shown). Under red light illumination, the resistivity monotonically drops by lowering temperature.

At the structural transition temperature around 150 K, however, the resistivity discontinuously jumps. Magnetoresistance at low temperatures increases quadratically by increasing magnetic eld. In the carrier/exciton ratio study of DInnocenzo et al.47, it was suggested that free charge carriers are predominant in perovskite solar cells at room temperature. Hence, the perovskites possess large built-in elds which can effectively drift photogenerated carriers to avoid charge recombination. These are in perfect agreement with our magnetoresistance and photocurrent spectroscopy measurements. The resistivity decreases by a factor more than 400 by cooling from 300 to 30 K. This clearly indicates the presence of free carriers at temperatures relevant to FM melting. The photocurrent generation is an active process. Also, the photocurrent spectra at 50 K clearly shows that free carriers are readily excited in our experiments.

Our experiments were performed under continuous illumination, which implies a constant number (time-independent after a few fs transient) of out-of-equilibrium photoexcited carriers next to the thermalized free carriers and excitons.

Data availability. The data that support the ndings of this study are available from the corresponding authors upon request.

References

1. Hwang, H., Palstra, T., Cheong, S. & Batlogg, B. Pressure effects on the magnetoresistance in doped manganese perovskites. Phys. Rev. B 52, 1504615049 (1995).

2. Snively, L. O., Tuthill, G. F. & Drumheller, J. E. Measurement and calculation of the superexchange interaction through the two-halide bridge in the eclipsed layered compounds [NH3(CH2)nNH3]CuX for n 25 and X Cl4 and Cl2Br2.

Phys. Rev. B 24, 53495355 (1981).3. Kimel, A. V., Kirilyuk, A., Tsvetkov, A., Pisarev, R. V. & Rasing, T. Laser-induced ultrafast spin reorientation in the antiferromagnet TmFeO3.

Nature 429, 850853 (2004).4. Ohno, H. et al. Electric-eld control of ferromagnetism. Nature 408, 944946 (2000).

5. Lottermoser, T. et al. Magnetic phase control by an electric eld. Nature 430, 541544 (2004).

6. Kovalenko, O., Pezeril, T. & Temnov, V. New concept for magnetization switching by ultrafast acoustic pulses. Phys. Rev. Lett. 110, 266602 (2013).

7. Li, T. et al. Femtosecond switching of magnetism via strongly correlated spin-charge quantum excitations. Nature 496, 6973 (2013).8. Moritomo, Y., Asamitsu, A., Kuwahara, H. & Tokura, Y. Giant magnetoresistance of manganese oxides with a layered perovskite structure. Nature 380, 141144 (1996).

9. Dhoot, A. S., Israel, C., Moya, X., Mathur, N. D. & Friend, R. H. Large electric eld effect in electrolyte-gated manganites. Phys.Rev. Lett. 102, 136402 (2009).

10. Cui, B. et al. Reversible ferromagnetic phase transition in electrode-gated manganites. Adv. Funct. Mater. 24, 72337240 (2014).

11. Stranks, S. D. et al. Electron-hole diffusion lengths exceeding 1 micrometer in an organometal trihalide perovskite absorber. Science 342, 341344 (2013).

12. Xing, G. et al. Long-range balanced electron- and hole-transport lengths in organic-inorganic CH3NH3PbI3. Science 342, 344347 (2013).

13. Stanciu, C. et al. All-optical magnetic recording with circularly polarized light. Phys. Rev. Lett. 99, 047601 (2007).

14. Astakhov, G. et al. Nonthermal photocoercivity effect in a low-doped (Ga,Mn)As ferromagnetic semiconductor. Phys. Rev. Lett. 102, 187401 (2009).

15. Vahaplar, K. et al. Ultrafast path for optical magnetization reversal via a strongly nonequilibrium state. Phys. Rev. Lett. 103, 117201 (2009).

16. Hui, L., Bo, L., Huanyi, Y. & Tow-Chong, C. Thermally induced stability issues of head-disk interface in heat-assisted magnetic recording systems. Jpn J. Appl. Phys. 44, 7950 (2005).

17. Khorsand, A. et al. Role of magnetic circular dichroism in all-optical magnetic recording. Phys. Rev. Lett. 108, 127205 (2012).

18. Zhang, G. P., Hubner, W., Lefkidis, G., Bai, Y. & George, T. F. Paradigm of the time-resolved magneto-optical Kerr effect for femtosecond magnetism. Nat. Phys. 5, 499502 (2009).

19. Vahaplar, K. et al. All-optical magnetization reversal by circularly polarized laser pulses: experiment and multiscale modeling. Phys. Rev. B 85, 104402 (2012).

20. Lee, M. M., Teuscher, J., Miyasaka, T., Murakami, T. N. & Snaith, H. J. Efcient hybrid solar cells based on meso-superstructured organometal halide perovskites. Science 338, 643647 (2012).

21. Horvth, E. et al. Nanowires of methylammonium lead iodide (CH3NH3PbI3) prepared by low temperature solution-mediated crystallization. Nano Lett. 14, 67616766 (2014).

22. Nfrdi, B., Gal, R., Sienkiewicz, A., Fehr, T. & Forr, L. Continuous-wave far-infrared ESR spectrometer for high-pressure measurements. J. Magn. Reson. 195, 206210 (2008).

23. Nfrdi, B., Gal, R., Fehr, T. & Forr, L. Microwave frequency modulation in continuous-wave far-infrared ESR utilizing a quasi-optical reection bridge.J. Magn. Reson. 192, 265268 (2008).24. Szirmai, P. et al. Synthesis of homogeneous manganese-doped titanium oxide nanotubes from titanate precursors. J. Phys. Chem. C 117, 697702 (2012).

25. Vonsovskii, S. V. E. & Massey, H. S. H. Ferromagnetic resonance: the phenomenon of resonant absorption of a high-frequency magnetic eld in ferromagnetic substances 1 edn (Pergamon P., 1968).

26. Kirkpatrick, S. Percolation and Conduction. Rev. Modern Phys. 45, 574588 (1973).

27. Coey, J. M. D., Venkatesan, M. & Fitzgerald, C. B. Donor impurity band exchange in dilute ferromagnetic oxides. Nat. Mater. 4, 173179 (2005).

28. MacDonald, A., Schiffer, P. & Samarth, N. Ferromagnetic semiconductors: moving beyond (Ga, Mn) As. Nat. Mater. 4, 195202 (2005).

29. Dietl, T. A ten-year perspective on dilute magnetic semiconductors and oxides. Nat. Mater. 9, 965974 (2010).

30. Baikie, T. et al. Synthesis and crystal chemistry of the hybrid perovskite (CH3NH3)PbI3 for solid-state sensitised solar cell applications. J. Mater. Chem.

A 1, 56285641 (2013).31. Kolley, E., Kolley, W. & Tietz, R. Ruderman-Kittel-Kasuya-Yosida interaction versus superexchange in a plane in the limit. J. Phys. Condens. Matter 10, 657662 (1998).

32. Keffer, F. & Oguchi, T. Theory of superexchange. Phys. Rev. 115, 14281434 (1959).

33. Van Vleck, J. H. Note on the interactions between the spins of magnetic ions or nuclei in metals. Rev. Modern Phys. 34, 681686 (1962).

34. Kawamura, Y., Mashiyama, H. & Hasebe, K. Structural study on cubictetragonal transition of CH3NH3PbI3. J. Phys. Soc. Jpn 71, 16941697 (2002).

35. Monod, P. et al. Paramagnetic and antiferromagnetic resonance of CuO.J. Magn. Magn. Mater. 177181, 739740 (1998).36. Jnossy, A., Chauvet, O., Pekker, S., Cooper, J. R. & Forr, L. Conduction electron spin resonance in Rb3C60. Phys. Rev. Lett. 71, 10911094 (1993).

37. Yafet, Y. Calculation of the g factor of metallic sodium. Phys. Rev. 85, 478478 (1952).

38. Elliott, R. J. Theory of the effect of spin-orbit coupling on magnetic resonance in some semiconductors. Phys. Rev. 96, 266279 (1954).

NATURE COMMUNICATIONS | 7:13406 | DOI: 10.1038/ncomms13406 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 7

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13406

39. Hohenberg, P. & Kohn, W. Inhomogeneous electron gas. Phys. Rev. 136, B864B871 (1964).

40. Kohn, W. & Sham, L. Self-consistent equations including exchange and correlation effects. Phys. Rev. 140, A1133A1138 (1965).

41. Giannozzi, P. et al. QUANTUM ESPRESSO: a modular and open-source software project for quantum simulations of materials. J. Phys. Condens. Matter 21, 395502 (2009).

42. Perdew, J., Burke, K. & Ernzerhof, M. Generalized gradient approximation made simple. Phys. Rev. Lett. 77, 38653868 (1996).

43. Vanderbilt, D. Soft self-consistent pseudopotentials in a generalized eigenvalue formalism. Phys. Rev. B 41, 78927895 (1990).

44. Garrity, K. F., Bennett, J. W., Rabe, K. M. & Vanderbilt, D. Pseudopotentials for high-throughput DFT calculations. Comput. Mater. Sci. 81, 446452 (2014).

45. Monkhorst, H. & Pack, J. Special points for Brillouin-zone integrations. Phys. Rev. B 13, 51885192 (1976).

46. Leslie, M. & Gillan, N. J. The energy and elastic dipole tensor of defects in ionic crystals calculated by the supercell method. J. Phys. C Solid State Phys. 18, 973982 (1985).

47. DInnocenzo, V. et al. Excitons versus free charges in organo-lead tri-halide perovskites. Nat. Commun. 5, 3586 (2014).

Acknowledgements

We are grateful to J.M. Triscone for facilitating SQUID experiments. This workwas partially supported by the Swiss National Science Foundation (Grant No. 200021_144419) and ERC Advanced Grant PICOPROP (Grant No. 670918).

H.L. and O.V.Y. thank the Swiss NSF (Grant No. PP00P2_133552), the ERC starting grant TopoMat (Grant No. 306504) and NCCR-MARVEL. First-principles computations have been performed at the Swiss National Supercomputing Centre under Project No. s515.

Author contributions

L.F. initiated the research. B.N. suggested the idea, coordinated the efforts and designed the experiment. E.H. synthesized the substance. P.Sz., M.S., D.C., M.G. and B.N. derived the experimental results. A.A. analysed the XRD pattern. H.L. and O.V.Y. performed rst-principles calculations. All authors discussed the results and contributed to the writing of the manuscript.

Additional information

Supplementary Information accompanies this paper at http://www.nature.com/naturecommunications

Web End =http://www.nature.com/ http://www.nature.com/naturecommunications

Web End =naturecommunications

Competing nancial interests: The authors declare no competing nancial interests.

Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/

Web End =http://npg.nature.com/ http://npg.nature.com/reprintsandpermissions/

Web End =reprintsandpermissions/

How to cite this article: Nfrdi, B. et al. Optically switched magnetism in photovoltaic perovskite CH3NH3(Mn:Pb)I3. Nat. Commun. 7, 13406 doi: 10.1038/ncomms13406 (2016).

Publishers note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional afliations.

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 articles 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 to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

Web End =http://creativecommons.org/licenses/by/4.0/

r The Author(s) 2016

8 NATURE COMMUNICATIONS | 7:13406 | DOI: 10.1038/ncomms13406 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications