Charge-controlled magnetism in colloidal dopedsemiconductor nanocrystalsStefan T. Ochsenbein, Yong Feng, Kelly M. Whitaker, Ekaterina Badaeva, William K. Liu, Xiaosong Li

and Daniel R. Gamelin*

Electrical control over the magnetic states of doped semiconductor nanostructures could enable new spin-basedinformation processing technologies. To this end, extensive research has recently been devoted to examination of carrier-mediated magnetic ordering effects in substrate-supported quantum dots at cryogenic temperatures, with carriersintroduced transiently by photon absorption. The relatively weak interactions found between dopants and charge carriershave suggested that gated magnetism in quantum dots will be limited to cryogenic temperatures. Here, we report theobservation of a large, reversible, room-temperature magnetic response to charge state in free-standing colloidal ZnOnanocrystals doped with Mn21 ions. Injected electrons activate new ferromagnetic Mn21–Mn21 interactions that are strongenough to overcome antiferromagnetic coupling between nearest-neighbour dopants, making the full magnetic moments ofall dopants observable. Analysis shows that this large effect occurs in spite of small pairwise electron–Mn21 exchangeenergies, because of competing electron-mediated ferromagnetic interactions involving distant Mn21 ions in thesame nanocrystal.

The manipulation of spins in semiconductor nanostructures is acentral theme in the development of spin-based informationtechnologies1,2. Prototype semiconductor ‘spintronics’

devices have used carrier–dopant magnetic exchange interactionsin magnetically doped semiconductors (so-called diluted magneticsemiconductors, DMSs) to manipulate either carrier spin polariz-ations or the spins of the dopants themselves3–6. Recently, theoreti-cal studies have proposed the use of charge carrier injection tocontrol pairwise Mn2þ–Mn2þ magnetic exchange interactions inDMS quantum dots7–11. Pairwise carrier–dopant magnetic exchangeinteractions in charged self-assembled II–VI quantum dots contain-ing single Mn2þ ions have been investigated12,13, and transientphotoinduced Mn2þ–Mn2þ exchange coupling mediated by exci-tons in self-assembled II–VI DMS quantum dots has been exam-ined14,15, but charge-controlled Mn2þ–Mn2þ magnetic exchangeof the type described in refs 7–11 has not yet been observed exper-imentally. More generally, charge effects in colloidal doped semi-conductor nanostructures remain wholly unexplored. Here, wereport the discovery of a surprisingly large charge-state-dependentmagnetic response in colloidal Mn2þ-doped ZnO (Mn2þ:ZnO)nanocrystals that is fully reversible and stable at room temperature.This phenomenon is closely related to the bound magnetic polaron(BMP) of bulk DMSs, in which dopant–carrier exchange couplinginduces magnetic ordering in the vicinity of a donor or acceptordefect, but has the important distinction that carrier localizationstems from the spatial confinement potential of the nanocrystalrather than from defect binding energies16.

Experimental observations of charge-controlled magnetismThe samples investigated in this study were colloidal wurtziteMn2þ:ZnO nanocrystals with average diameters d between 4.7and 5.6 nm and Mn2þ dopant concentrations between 0.0 and0.5 cation percent (that is, x¼ 0.00–0.005 in Zn1–xMnxO, deter-mined analytically; see Methods), which were synthesized andhandled as described previously17,18. Transmission electronmicroscopy (TEM) images of representative Mn2þ:ZnO

nanocrystals are shown in Fig. 1a, including a high-resolutionimage of a single quantum dot, highlighting the crystallinity.Anaerobic ultraviolet (UV) photoexcitation in the presence of ahole trap (EtOH) was used to generate kinetically stable conductionband electrons (e�CB) in the colloidal nanocrystals that are stableessentially indefinitely until exposed to air19,20. Electron paramag-netic resonance (EPR) spectroscopy was used to probe the effectof the added conduction band electrons at room temperature.Figure 1b shows 298 K derivative-mode X-band EPR spectra of col-loidal d¼ 5.6 nm 0.25% Mn2þ:ZnO nanocrystals before and afterthe introduction of e�CB. Upon charging, the EPR signal intensityincreases and the hyperfine structure broadens considerably.From comparison of these EPR spectra with that of conductionband electrons in undoped ZnO nanocrystals19,20 (Fig. 1b, inset)it is evident that this broadening does not simply arise from super-position of the Mn2þ and e�CB EPR signals, but instead reflectse�CB–Mn2þ magnetic exchange interactions. The EPR spectra inFig. 1b also do not show superimposed sharp and broad Mn2þ fea-tures, indicating that all Mn2þ ions are affected by the added elec-trons. Exposure of the charged Mn2þ:ZnO nanocrystals to airresults in rapid reoxidation and regeneration of the original EPRspectra (Fig. 1b, dotted). Figure 1c shows the same EPR signalsplotted in absorption mode, from which the resonance of thecharged Mn2þ:ZnO nanocrystals is seen to be �25% moreintense than that of the uncharged sample. This charging/reoxida-tion cycle can be repeated multiple times on the same sample withno detectable degradation. Figure 2a shows the EPR intensityduring two cycles of charging and reoxidation. Also plotted inFig. 2a is the infrared (IR) absorption intensity. As shown pre-viously18,19,21–23, a new intra-conduction band transition appearsin the IR region of the electronic absorption spectrum of ZnOnanocrystals upon introduction of conduction band electrons.The fact that the Mn2þ EPR intensity changes track the changesin IR absorption intensity provides a strong indication that theformer do indeed derive from the addition of conductionband electrons.

Department of Chemistry, University of Washington, Seattle, Washington 98195-1700, USA. *e-mail: gamelin@chem.washington.edu

ARTICLESPUBLISHED ONLINE: 16 AUGUST 2009 | DOI: 10.1038/NNANO.2009.221

NATURE NANOTECHNOLOGY | VOL 4 | OCTOBER 2009 | www.nature.com/naturenanotechnology 681

© 2009 Macmillan Publishers Limited. All rights reserved.

Because EPR intensities can be complicated by relaxation effects,quantitative magnetization measurements of dilute suspensions ofthe same colloidal nanocrystals were performed between 2 and300 K and up to H¼ 5 T using a superconducting quantum inter-ference device (SQUID). 2 K magnetization data for the colloidal0.25% Mn2þ:ZnO nanocrystals of Fig. 1b,c are shown in Fig. 1dand reveal a similar increase upon charging as observed by EPR.The solid lines in Fig. 1d plot the Brillouin function calculated forparamagnetic Mn2þ ions with spin S¼ 5/2 at 2 K for two differentvalues of the effective concentration xeff. The magnetic moment (inBohr magnetons per Mn2þ) before charging is Msat¼3.74 mB/Mn2þ (xeff/x� 0.75), which is significantly smaller thanwhat would be observed if all Mn2þ ions contributed fully(xeff/x¼ 1.0, Msat¼ 5.0 mB/Mn2þ). This reduction arises from thewell-known antiferromagnetic superexchange interaction activewithin nearest-neighbour Mn2þ–Mn2þ dimer pairs24,25. Dimerpair formation is expected to exceed statistical probability in thesenanocrystals because of correlated substitution26, deliberateremoval of Mn2þ from the nanocrystal surfaces17 and undoped criti-cal nuclei27.

The nearest-neighbour Mn2þ–Mn2þ superexchange interactionin Mn2þ:ZnO is known from inelastic neutron scattering measure-ments to be antiferromagnetic, with an exchange coupling constantof JAFM,iso� –1.61 meV (ref. 25) in the Heisenberg-type dimer spinHamiltonian of equation (1), where Si is the spin operator of the ithMn2þ ion:

H ¼ �2JS1 � S2 ð1Þ

The total dimer exchange splitting thus amounts to DEFM-AFM¼230J� 48 meV. Under normal circumstances, therefore, Mn2þ–Mn2þ nearest-neighbour pairs contribute only negligibly to thetotal magnetic moment at 2 K, 5 T or to X-band EPR intensitiesat room temperature. The antiferromagnetic superexchange inter-action strength decays rapidly with increasing Mn2þ–Mn2þ separ-ation such that already by the next-nearest-neighbour separation,Mn2þ ions are so weakly coupled that they can be treated as isolatedmonomers under our experimental conditions. Remarkably, char-ging increases the saturation moment to the full 5 mB/Mn2þ

8

Diff

eren

tial E

PR in

tens

ity e−CB: ZnO

4

0

−4

0.25

0.25

0.30 0.35Field (T)

0.40

5

4

3

2

1

0

3

dM/dH2

1

0

15

EPR

inte

nsity

10

5

00.30 0.35

Field (T)

Field (T)

0.40

5432

Magnetization (µ

B /Mn

2+)

10

Field (T)543210

2 K298 K

50 nm

2 nm

0.8

0.6dM/d

H

0.80.40Field (T)

Figure 1 | Structural characterization and charge-induced magnetic effects at 298 K and 2 K. a, TEM image of representative Mn2þ:ZnO nanocrystals. The

inset shows a high-resolution TEM image of a single nanocrystal. b,c, Derivative-mode (b) and absorption-mode (c) EPR spectra of as-prepared (solid red

line), photochemically charged (solid blue line) and re-oxidized (dotted grey line) colloidal Mn2þ:ZnO nanocrystals (with diameter d¼ 5.6 nm and Mn2þ

concentration x¼0.25%, 298 K). d, Field dependence of the magnetization of as-prepared (open red triangles) and charged (filled blue triangles) colloidal

Mn2þ:ZnO nanocrystals (d¼ 5.6 nm, x¼0.25%, 2 K). The activated fraction (filled blue diamonds) is the difference between the magnetization of the

charged and the as-prepared nanocrystals. e, The numerical derivative dM/dH of the as-prepared (open red triangles) and the activated fraction (filled blue

diamonds) of the Mn2þ:ZnO nanocrystals. The expanded low-field region of the activated fraction is shown in the inset.

ARTICLES NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2009.221

NATURE NANOTECHNOLOGY | VOL 4 | OCTOBER 2009 | www.nature.com/naturenanotechnology682

© 2009 Macmillan Publishers Limited. All rights reserved.

expected for the complete population of Mn2þ ions (Fig. 1d),determined analytically.

Figure 2b summarizes the charge-induced changes in 2 K, 5 Tmagnetization and 298 K EPR intensity as a function of Mn2þ con-centration for a series of colloidal Mn2þ:ZnO nanocrystals. The 2 Ksaturation moment decreases smoothly with increasing Mn2þ con-centration, consistent with the formation of more antiferromagneti-cally coupled nearest-neighbour Mn2þ–Mn2þ pairs. It is importantto note that the change in magnetization and EPR intensity uponcharging practically goes away as the Mn2þ concentration isdecreased, which is further evidence that the observed changes arenot due to the paramagnetic signal of the added conduction bandelectrons. Instead, the change in magnetization with chargingincreases with higher Mn2þ doping levels, always reaching5 mB/Mn2þ upon charging. The 298 K EPR results are essentiallyidentical (Fig. 2b), confirming that these two independent tech-niques indeed measure the same effect. From these data, we con-clude that the antiferromagnetically coupled Mn2þ–Mn2þ dimersmust become magnetically visible upon addition of conductionband electrons.

Figure 1d also plots the difference between charged anduncharged magnetization curves measured at 2 K: Mactivated¼Mcharged–Muncharged. Mactivated shows S� 5/2 saturation, whichwould be inconsistent with the extra magnetization arising solelyfrom the added electrons themselves. Control experiments onundoped ZnO nanocrystals revealed no comparable increasein magnetization upon charging, confirming that the changesobserved here are associated with Mn2þ. Importantly, Mactivatedand Muncharged have a different curvature at low fields,which can be seen clearly in the first derivatives of theexperimental data sets (dM/dH, Fig. 1e). An inflection is observed

at dM/dH� 0.34+ 0.01 T for the activated fraction (see inset ofFig. 1e) that is not present for the uncharged Mn2þ:ZnO nano-crystals. This inflection is reminiscent of field-induced level cross-ings in molecular spin clusters28, or metamagnetic transitions inextended lattices29. Field-induced level crossings have been observedpreviously in bulk DMSs, but only at magnetic fields much greaterthan that of Fig. 2 (for example, at 15 T in Mn2þ:ZnTe, see ref. 30).From JAFM,iso determined by inelastic neutron scattering, the firstlevel crossing should also occur at �28 T in Mn2þ:ZnO (ref. 25).The inflection at 0.34 T in Fig. 1e thus indicates a surprisinglylarge reduction of dimer spin-state splittings upon charging ofMn2þ:ZnO nanocrystals. This conclusion implicates a newcharge-induced ferromagnetic contribution to the Mn2þ–Mn2þ

dimer exchange interaction that is strong enough to compete withthe antiferromagnetic superexchange interaction.

Electron-mediated exchange interactionsOne hypothesis to explain these observations is that the added con-duction band electron directly introduces new exchange termswithin the Mn2þ–Mn2þ pairs that are of comparable magnitudeto the normal dimer superexchange interaction but of oppositesign. To evaluate this hypothesis for the Mn2þ:ZnO nanocrystals,we can replace J in equation (1) by J*, which is the sum of competingantiferromagnetic superexchange and ferromagnetic e�CB-mediatedcontributions: J*¼ JAFMþ JFM. From the inflection point indM/dH of the activated fraction (Fig. 1e), we can estimate J* tobe ,0.025 meV. JFM must therefore be on the order ofþ1.59 meV. Unfortunately, this value of JFM is much larger thancan be reasonably anticipated from existing knowledge aboute�CB–Mn2þ exchange interactions in Mn2þ:ZnO. For example, ans–d exchange energy of N0a¼þ0.19 eV has been estimated fromexperimental magneto-transport data analysed within the mean-field approximation31. For a nanocrystal of 5.6 nm diameter, thisamounts to an average pairwise exchange interaction parameterJsd¼ 0.05 meV, and from the relationship DEFM–AFM¼ 6Jsd¼30JMn–Mn for electron-mediated Mn2þ–Mn2þ coupling, this Jsd

yields JFM� 0.01 meV per unpaired conduction band electron.This interaction is thus approximately two orders of magnitudetoo small to account for the large change in 2 K magnetizationobserved in Figs 1 and 2. Even in the possible scenario of multipleunpaired conduction band electrons (possibly three)23, this JFM isstill well over an order of magnitude too small to competewith JAFM. In short, the impact of charging on Mn2þ:ZnOnanocrystal magnetism is unexpectedly large and not readilyexplained using normal e�CB–Mn2þ exchange energies acting uponMn2þ–Mn2þ dimers.

It is conceivable that this discrepancy could imply a significantlygreater value of N0a in Mn2þ:ZnO nanocrystals than found in theMn2þ:ZnO thin films studied previously. Modification ofe�CB–Mn2þ exchange interactions due to quantum confinement inother DMSs has been suggested32–35. Although the s–d interactionin bulk DMSs involves only weak potential exchange coupling,recent work has described confinement-induced kinetic e�CB–Mn2þ

exchange contributions in Mn2þ:CdTe and GaMnAs quantumwells and in Mn2þ:ZnSe nanocrystals32–34, and theoreticalmodelling has suggested that N0a can even approach the magnitudeof the p–d exchange energy (N0b) in quantum dots32. Such a scenariocould thus possibly be invoked to explain the unexpectedly largechange in Mn2þ:ZnO nanocrystal magnetism observed upon charging.

We propose instead that the large effect of nanocrystal chargingobserved here can indeed be explained using typical bulk e�CB–Mn2þ

s–d exchange energies once additional contributions from otherwell-separated Mn2þ ions are also explicitly taken into account.For illustration, consider the simplest extension of the dimermodel described by equation (1), namely a trimer formed byaddition of one more Mn2þ ion into the nanocrystal at a large

1.0

0.9

0.8

0.7

Rela

tive

EPR

inte

nsity

1.0

0.5

0

Relative IR intensity

1.0

0.8

0.6Rela

tive

EPR

inte

nsity

0.50.40.30.20.10Mn2+ concentration (%)

5

4

3

5 T, 2 K M

agnetization (µ

B /Mn

2+)

OFF OFF OFFON ON

+eCB−

+eCB−

−eCB−

+eCB−

−eCB−

−eCB−

Figure 2 | Reversibility and Mn21 concentration dependence of charge-

controlled magnetism. a, Integrated 298 K EPR (Mn2þ:ZnO, open red

circles; e�CB:Mn2þ:ZnO, filled blue circles) and IR absorption (Mn2þ:ZnO,

open brown squares; e�CB:Mn2þ:ZnO, filled blue squares) intensities

normalized at the highest observed intensity with repeated stepwise

anaerobic ultraviolet (UV) irradiation (charging) and re-oxidation in air for

Mn2þ:ZnO nanocrystals (d¼ 5.6 nm, x¼0.25%). b, Relative integrated EPR

intensity of as-prepared (open red circles) and charged (filled blue circles)

Mn2þ:ZnO nanocrystals, and 2 K, 5 T magnetization of as-prepared (open

brown triangles) and charged (filled blue triangles) Mn2þ:ZnO nanocrystals

plotted versus Mn2þ concentration.

NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2009.221 ARTICLES

NATURE NANOTECHNOLOGY | VOL 4 | OCTOBER 2009 | www.nature.com/naturenanotechnology 683

© 2009 Macmillan Publishers Limited. All rights reserved.

distance from the original dimer pair (Fig. 3a, inset). Equation (2)presents the spin Hamiltonian for this trimer arrangement ofMn2þ ions within the ZnO nanocrystal:

H ¼ �2JdimerS1 � S2 � 2Jfar S1 � S3 þ S2 � S3ð Þ ð2Þ

For simplicity, it is assumed that the electron-mediated inter-action between distant Mn2þ ions described by Jfar is the same foreach of the Mn2þ ions of the dimer pair. From this spinHamiltonian, analytical solutions to the trimer energy levels canbe found and are plotted in Fig. 3a as a function of Jfar/Jdimer,where Jdimer represents the exchange interaction betweennearest-neighbour Mn2þ ions. In the absence of the added conduc-tion band electron, Jfar¼ 0 and the nanocrystal magnetization is

essentially the same as that of the isolated S¼ 5/2 paramagnetbut with a 2 K, 5 T moment of �1/3 . (5 mB/Mn2þ) because thedimer is practically silent. Charging this nanocrystal leads to anindirect e�CB-mediated exchange interaction between the distantMn2þ and the Mn2þ–Mn2þ dimer pair; that is, Jfar = 0. AsFig. 3a illustrates, turning on Jfar always reduces the gap betweenthe lowest and first excited energy levels relative to the dimer, andin many circumstances leads to a trimer ground spin state otherthan S¼ 5/2. In this way, even weak e�CB–Mn2þ s–d exchange coup-ling can strongly diminish the apparent antiferromagnetic energylevel splittings of a Mn2þ–Mn2þ dimer when other Mn2þ mono-mers are also present.

To test this interpretation of the data more quantitatively, wehave simulated the magnetization of various Mn2þ trimers with

5

4

3

2

1

0

Mag

netiz

atio

n (µ

B/M

n2+)

5

4

3

2

1

0

Mag

netiz

atio

n (µ

B/M

n2+)

4 53210Field (T)

543210Field (T)

+e−CB

+e−CB

1

23

1

23 4

a

1

2

3 JdimerJfar

Jfar

−40

−20

0

20

40

Ener

gy/J

dim

er

−2 −1 0 1 2

Jfar /Jdimer

|Sdimer,Stotal|5,15/2

|4,13/2

|3,11/2

|2,9/2

|1,7/2

|0,5/2 J av (m

eV)

M (µ

B/M

n2+)

0.4

0.2

0

2.252.001.751.501.25QD diameter (nm)

10

8

6

4

2

0

nM

n2

1

0

Figure 3 | Influence of Jfar on dimer magnetization. a, Energy levels of a Mn2þ(S¼ 5/2) trimer as a function of the ratio Jfar/Jdimer within the coupling

scheme indicated in the inset (Jdimer , 0). b,c, Calculated 2 K magnetization curves for a nearest-neighbour dimer coupled to one (b) or two (c) additional

distant Mn2þ ions via the added e�CB, following the coupling schemes shown in the insets. Solid red line: Jdimer¼21.61 meV, Jfar¼0. Dotted green line:

Jdimer¼20.60 meV, Jfar¼0. Solid blue line: Jdimer¼20.60 meV, Jfar¼þ0.68 meV. The saturation magnetization for a full contribution of all Mn2þ dopants

is 5 mB/Mn2þ. With one additional distant Mn2þ dopant (b), charging induces a significant increase in nanocrystal magnetization, but full Mn2þ

magnetization is not achieved. With two additional Mn2þ dopants (c), the nearest-neighbour antiferromagnetic exchange can be overcome completely upon

charging, and the full 5 mB/Mn2þ is observed. d, Calculated decrease (lower plot) in the average value of Jfar (Jav, solid line) and increase in the number of

Mn2þ ions (nMn, dashed line) with increasing nanocrystal diameter at a constant Mn2þ concentration x. The open sphere represents the calculated average

Jfar from DFT. Upper plot, calculated dependence of the change in magnetization on nanocrystal diameter for the nanocrystals with nMn¼ 3,4, . . . ,7 described

in the lower plot. The large size of the Hilbert space ((2Sþ 1)n� 1.7� 106) has precluded numerical calculation of one dimer plus eight monomers, the

average experimental situation for the Mn2þ:ZnO nanocrystals of Fig. 1. The nearly constant value of the change in magnetization with charging indicates that

the decreasing Jfar is compensated by the increasing nMn as the nanocrystal diameter increases at fixed x.

ARTICLES NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2009.221

NATURE NANOTECHNOLOGY | VOL 4 | OCTOBER 2009 | www.nature.com/naturenanotechnology684

© 2009 Macmillan Publishers Limited. All rights reserved.

and without e�CB-mediated Jfar , using J values derived from densityfunctional theory (DFT). Previous calculations on pseudo-hydrogenpassivated wurtzite Mn2þ:ZnO nanocrystals of various sizes (33–84Zn2þ cations, d¼ 1.0–1.5 nm) using the UPBE1PBE/LanL2DZfunctional have reproduced the ZnO bandgap and charge-transferenergies accurately36,37. For the present study, an additional electronwas introduced into the Mn2þ:ZnO nanocrystals. The calculated e�CBprobability density is distributed over the entire nanocrystal in ans-like orbital (Fig. 4a), in agreement with EPR results for e�CB:ZnOnanocrystals20. The e�CB–Mn2þ pairwise exchange splitting energybetween spin-up and spin-down e�CB (DE"2#

sd ) was then calculatedfor various Mn2þ positions within the nanocrystal and found todecrease from –50 to �0 meV as the Mn2þ is moved away fromthe centre of the nanocrystal (Fig. 4b), yielding a volumetricallyaveraged value of DE"2#

sd (avg.)¼2 10.4 meV. From the relationshipDEsd"�#ðavg:Þ ¼ 3N0a=ncation, where ncation is the number of cations

per nanocrystal, this DFT result translates to N0a¼þ0.29 eV,which agrees well with the experimental value of N0a¼þ0.19 eV(ref. 31). The magnitude and sign of this result, and the absenceof e�CB probability density at the Mn2þ, support attribution ofthese e�CB–Mn2þ interactions to potential exchange. For DFT analy-sis of Mn2þ–Mn2þ exchange energies, energy differences betweenferromagnetic and antiferromagnetic configurations of variousMn2þ–Mn2þ pairs were calculated and converted to J values (JMn–

Mn¼ DEFM–AFM/30). Because of the axial crystal structure of wurt-zite ZnO, there are two nearest-neighbour Mn2þ–Mn2þ dimer pairconfigurations, with calculated exchange coupling strengths ofJdimer(1)¼22.33 meV for the Mneq–Mneq dimer, andJdimer(2)¼21.84 meV for the Mnax–Mneq dimer. These valuescompare well with those determined experimentally by inelasticneutron scattering ((1) 21.90 and (2) 21.39 meV)25. Jfar¼ 0 forall other Mn2þ–Mn2þ pairs. These calculations were then repeated

after addition of a conduction band electron. The computed Jdimervalues decrease only slightly upon addition of e�CB, to (1) 21.18and (2) 20.60 meV, consistent with the expectations discussedabove. Upon charging, Jfar increases from 0 to between �þ 0.24and þ0.68 meV depending on the radial positions of the twoMn2þ ions involved, again consistent with expectation38,39.Overall, the DFT calculations thus reproduce the available exper-imental e�CB–Mn2þ and Mn2þ–Mn2þ magnetic exchange couplingstrengths quite well.

With the above exchange energies and equation (2), we can nowsimulate the magnetization of a charged nanocrystal containingthree Mn2þ ions, one dimer pair and one distant Mn2þ ion.Figure 3b shows the magnetization of such a nanocrystal calculatednumerically. The small but non-zero Jfar in this case more thandoubles the trimer magnetization at 2 K, 5 T. For comparison, thesame electron has a negligible effect on the nanocrystal magnetismunder these conditions when Jfar is set to zero (dotted line inFig. 3b). Thus, although the added conduction band electron is insuf-ficient to alter the magnetism of the Mn2þ–Mn2þ dimer directly,introduction of just one additional distant Mn2þ ion completelychanges the Mn2þ:ZnO nanocrystal magnetism by greatly reducingthe trimer energy level spacings. Note that the trimer magnetizationin Fig. 3b still does not reach the full 5 mB/Mn2þ at 2 K, 5 T, butinstead increases from 1.7 to only �3.6 mB/Mn2þ. Figure 3c showsnumerical results calculated for an analogous tetramer, in which asecond distant Mn2þ ion has been added to the nanocrystal to intro-duce another set of Jfar interactions. As shown in Fig. 3c, the antiferro-magnetic interaction of the Mn2þ–Mn2þdimer pair is now completelyovercome, and the maximum value of 5 mB/Mn2þ is achieved at 2 K,5 T; that is, the contributions from multiple Mn2þ monomers arecumulative. The central conclusion drawn from these calculations isthat the presence of additional Mn2þ monomers strongly influencescarrier-mediated exchange interactions of dimer pairs.

Finally, extension of this concept to explain the experimental datain Figs 1 and 2 requires recognition that the average microscopic pair-wise e�CB–Mn2þ s–d exchange coupling strength decreases rapidlywith increasing nanocrystal volume, and can be expected to be roughly50 times smaller in the experimental nanocrystals of Figs 1 and 2(d¼ 5.6 nm) than in the computed nanocrystals (d¼ 1.56 nm).Likewise, however, there is a concomitant increase in the number ofMn2þ ions per nanocrystal. For example, the nanocrystals fromFig. 1 contain on average eight monomers and one Mn2þ–Mn2þ

dimer per nanocrystal. Numerical calculations confirm that thesetwo effects largely cancel one another, such that the central conclusiondrawn above remains valid for larger nanocrystals. Figure 3d (bottompanel) plots the average Jfar versus nanocrystal diameter, in compari-son with the average number of Mn2þ ions per nanocrystal for a con-stant value of x (nMn¼ 3,4,. . .,7). Figure 3d (top panel) plots theresults of numerical calculations of the increase in nanocrystal magne-tization upon addition of a conduction band electron across this series,and shows that there is only a small variation with particle diameter.Charging thus has a comparable impact on the magnetization oflarger nanocrystals, despite the sizeable reduction in average pairwisee�CB–Mn2þ s–d exchange coupling strengths, because of the additionalMn2þ monomers and the competing exchange interactions theybring. To our knowledge, the important role such carrier-mediatedcompeting interactions play in influencing dimer magnetism inDMSs has not been directly addressed in any previous theoretical orexperimental study, even though they can completely dominate theexperimental observations as shown in Figs 1 and 2.

ConclusionsThe results and conclusions drawn above are summarized schema-tically in Fig. 5. Figure 5 (left) depicts a colloidal Mn2þ:ZnO nano-crystal before photochemical reduction. Within the nanocrystal,Mn2þ–Mn2þ pairs exist that are silent to both EPR and magnetic

+eCB−

Mn2+

Jsd (R)

ΔEsd

(meV

)

−60

−40

−20

00.60.40.20

R (nm)

eCB−

Figure 4 | Density functional theory results. a, Electron density probability

distribution (purple) for a conduction band electron (e�CB) added to a

Zn83MnO84 nanocrystal. Zn2þ (grey), O22 (red), Mn2þ (black sphere).

b, The e�CB–Mn2þ s–d exchange splittings (DE"2#sd ) calculated for an

e�CB:Zn83MnO84 nanocrystal, plotted versus Mn2þ radial position within the

nanocrystal (green circles). The black line is a fit to the square of a zero-

order spherical Bessel function, yielding a mean-field s–d exchange

parameter of N0a¼þ0.29 eV (DE"2#sd (R)¼2 6Jsd (R)¼23a j C (R) j2,

where C(R) represents the radial dependence of the wavefunction of a

conduction band electron). Inset, schematic representation of the e�CB–Mn2þ

exchange interaction within the nanocrystal.

NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2009.221 ARTICLES

NATURE NANOTECHNOLOGY | VOL 4 | OCTOBER 2009 | www.nature.com/naturenanotechnology 685

© 2009 Macmillan Publishers Limited. All rights reserved.

susceptibility due to their antiferromagnetic superexchange coup-ling (Jdimer), resulting in S¼ 5/2 magnetization at 2 K, 0–5 T, butwith a reduced average saturation moment. After injection of con-duction band electrons, 298 K EPR intensities and 2 K magneticsusceptibilities both increase markedly, and the full saturationmoment of 5 mB/Mn2þ is obtained (Fig. 5, right). Despite smallpairwise e�CB–Mn2þ s–d exchange energies, Jdimer is effectively over-come by the introduction of multiple new e�CB-mediated ferromag-netic interactions between the dimer pair and distant Mn2þ ions(Jfar), which lead to small energy spacings among the lowest mag-netic states of the nanocrystal as a whole, directly analogous toweak BMP formation but in colloids. These carrier-mediated ferro-magnetic interactions are fully eliminated again upon removal of theadded electron, achieved here simply by exposure of the chargednanocrystals to air. This discovery of charge-controlled magnetismin free-standing colloidal DMS nanocrystals that is large, reversible,and stable at room temperature presents new opportunities for fun-damental studies of carrier–dopant magnetic exchange interactionsin semiconductor nanostructures under well-defined and syntheti-cally controlled conditions. The new motif of colloidal BMPanalogs may also raise interesting possibilities for the developmentof spin-based information technologies from solution-processableDMS nanostructures.

MethodsSynthesis and characterization. Colloidal Mn2þ-doped ZnO nanocrystals wereprepared and routine characterization was performed as described in detailpreviously17. TEM experiments were performed on a JEOL 2010 microscope(200 kV) equipped with a LaB6 filament as the electron source, located at the PacificNorthwest National Laboratories (PNNL). Mn2þ concentrations were determinedanalytically by inductively coupled plasma-atomic emission spectroscopy (ICP-AES,Jarrel Ash model 995) using the standard addition technique and analytical Zn2þ

and Mn2þ standards of known concentration (High-Purity Standards). In this way,Mn2þ contents of the colloidal ZnO nanocrystal samples used for the magnetizationmeasurements were determined with an absolute accuracy of+5% and with anaccuracy of+0.01% in concentrations relative to Zn2þ. Typical absolute Mn2þ

concentrations used for magnetization measurements were on the order of0.1–1.0 ppm, or well over the ICP-AES detection limit.

Photochemical reduction. Conduction band electrons were introduced into thecolloidal Mn2þ:ZnO nanocrystals by irradiating suspensions of tri-octylphosphineoxide capped nanocrystals in toluene with UV light (mercury arc lamp) underrigorously anaerobic conditions in the presence of �1–5% ethanol (by volume).Once charged, these conduction band electrons are kinetically stable, with decay rateconstants of kdecay , 0.01 per week, allowing the various physical measurements tobe performed. Re-oxidation was achieved by exposing these suspensions to air.Experimental details can be found in ref. 18. The photochemical reduction was

monitored by electronic absorption spectroscopy18 with a Varian Cary 5Espectrophotometer.

Physical measurements. Colloidal suspensions of Mn2þ:ZnO nanocrystals weresealed in quartz tubes of 4 or 5 mm outer diameter for all physical measurements.Continuous-wave EPR spectra were recorded at room temperature on Bruker EMXand Bruker E580 spectrometers with X-band microwave sources (�9.5 GHz).Magnetic measurements were performed on a Quantum Design SQUID MPMS 5magnetometer between 300 and 2 K. The magnetic data were corrected fordiamagnetic contributions from the sample holder, the solvent, and the sample bysubtracting a straight line derived from fitting the magnetization measured at 300 K.

Calculations. DFT calculations were performed using the PBE1PBE hybrid DFTfunctional and the Los Alamos double-z pseudo-core potential (LANL2DZ) usingthe development version of the Gaussian program package40. Zn84–nMnnO84wurtzite ZnO nanocrystals (diameter �1.56 nm) possessing C3v symmetrywere constructed using lattice parameters from experimental data: a¼ 3.249 Å,c¼ 5.204 Å and u¼ 0.382 (ref. 41). Nanocrystal surface dangling bonds werepassivated with pseudo-hydrogen atoms having modified nuclear charges of 0.5 and1.5 to terminate surface O22 and Zn2þ ions, respectively. The O–H (1.057 Å) andZn–H (1.731 Å) bond lengths were taken from optimized H4O and ZnH4 tetrahedra.This pseudo-hydrogen capping scheme leads to a well-defined bandgap and stablenanocrystal geometry. Replacing Zn2þ with a Mn2þ dopant ion is a charge-neutralsubstitution, and retains the overall neutral charge of the nanocrystal. Charging ofthe nanocrystals was achieved by adding one additional electron into the lowestunoccupied molecular orbital of the ground-state configuration, followed by a fullelectronic wavefunction optimization. Further details of the calculations can befound in ref. 36. Numerical magnetization simulations were carried out using theFortran program MAGPACK42.

Received 14 May 2009; accepted 9 July 2009;published online 16 August 2009

References1. Zutic, I., Fabian, J. & Das Sarma, S. Spintronics: fundamentals and applications.

Rev. Mod. Phys. 76, 323–410 (2004).2. Wolf, S. A. et al. Spintronics: a spin-based electronics vision for the future.

Science 294, 1488–1495 (2001).3. Jonker, B. T. et al. Robust electrical spin injection into a semiconductor

heterostructure. Phys. Rev. B 62, 8180–8183 (2000).4. Ohno, Y. et al. Electrical spin injection in a ferromagnetic semiconductor

heterostructure. Nature 402, 790–792 (1999).5. Ruster, C. et al. Very large magnetoresistance in lateral ferromagnetic

(Ga,Mn)As wires with nanoconstrictions. Phys. Rev. Lett. 91, 216602 (2003).6. Abolfath, R. M., Petukhov, A. G. & Zutic, I. Piezomagnetic quantum dots.

Phys. Rev. Lett. 101, 207202 (2008).7. Fernandez-Rossier, J. & Brey, L. Ferromagnetism mediated by few electrons in a

semimagnetic quantum dot. Phys. Rev. Lett. 93, 117201 (2004).8. Govorov, A. O. Voltage-tunable ferromagnetism in semimagnetic quantum dots

with few particles: magnetic polarons and electrical capacitance. Phys. Rev. B 72,075359 (2005).

9. Qu, F. & Hawrylak, P. Magnetic exchange interactions in quantum dotscontaining electrons and magnetic ions. Phys. Rev. Lett. 95, 217206 (2005).

10. Qu, F. & Hawrylak, P. Theory of electron mediated Mn–Mn interactions inquantum dots. Phys. Rev. Lett. 96, 157201 (2006).

11. Abolfath, R. M., Hawrylak, P. & Zutic, I. Electronic states of magnetic quantumdots. New J. Phys. 9, 353 (2007).

12. Leger, Y., Besombes, L., Fernandez-Rossier, J., Maingault, L. & Mariette, H.Electrical control of a single Mn atom in a quantum dot. Phys. Rev. Lett. 97,107401 (2006).

13. Besombes, L., Leger, Y., Maingault, L. & Mariette, H. Spin properties of chargedMn-doped quantum dot. J. Appl. Phys. 101, 081713 (2007).

14. Seufert, J. et al. Dynamical spin response in semimagnetic quantum dots. Phys.Rev. Lett. 88, 027402 (2002).

15. Bacher, G. Optical spectroscopy on epitaxially grown II–VI single quantum dots(review). Top. Appl. Phys. 90, 147–185 (2003).

16. Maksimov, A. A. et al. Magnetic polarons in a single diluted magneticsemiconductor quantum dot. Phys. Rev. B 62, R7767–R7770 (2000).

17. Norberg, N. S. et al. Synthesis of colloidal Mn2þ:ZnO quantum dots andhigh-TC ferromagnetic nanocrystalline thin films. J. Am. Chem. Soc. 126,9387–9398 (2004).

18. Liu, W. K., Whitaker, K. M., Kittilstved, K. R. & Gamelin, D. R. Stablephotogenerated carriers in magnetic semiconductor nanocrystals. J. Am. Chem.Soc. 128, 3910–3911 (2006).

19. Liu, W. K. et al. Room-temperature electron spin dynamics in free-standing ZnOquantum dots. Phys. Rev. Lett. 98, 186804 (2007).

20. Whitaker, K. M., Ochsenbein, S. T., Pollinger, V. Z. & Gamelin, D. R. Electronconfinement effects in the EPR spectra of colloidal n-type ZnO quantum dots.J. Phys. Chem. C 112, 14331–14335 (2008).

H H

M5T,2K < 5 µB/Mn2+ M5T,2K = 5 µB/Mn2+

+eCB−

−eCB−

Figure 5 | Colloidal analogues to bound magnetic polarons. Schematic

representation of an as-prepared free-standing Mn2þ:ZnO nanocrystal (left)

and an e�CB:Mn2þ:ZnO nanocrystal (right) at 5 T, 2 K. Images show colloidal

suspensions of as-prepared Mn2þ:ZnO nanocrystals (left) and of

e�CB:Mn2þ:ZnO nanocrystals (right). The red arrows represent the Mn2þ

spins, and the blue arrow and the shaded circle represent the electron spin

and its delocalization over the entire nanocrystal. Charging causes the

antiferromagnetic dimer exchange coupling to be effectively eliminated and

the magnetization of all Mn2þ ions to be observed. The charged and doped

colloidal nanocrystal shown on the right represents a colloidal analogue to

the bound magnetic polaron (BMP) of bulk and thin-film materials.

ARTICLES NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2009.221

NATURE NANOTECHNOLOGY | VOL 4 | OCTOBER 2009 | www.nature.com/naturenanotechnology686

© 2009 Macmillan Publishers Limited. All rights reserved.

21. Haase, M., Weller, H. & Henglein, A. Photochemistry and radiation chemistry ofcolloidal semiconductors. 23. Electron storage on ZnO particles and sizequantization. J. Phys. Chem. 92, 482–487 (1988).

22. Shim, M. & Guyot-Sionnest, P. Organic-capped ZnO nanocrystals: synthesis andn-type character. J. Am. Chem. Soc. 123, 11651–11654 (2001).

23. Germeau, A. et al. Optical transitions in artificial few-electron atoms stronglyconfined inside ZnO nanocrystals. Phys. Rev. Lett. 90, 097401 (2003).

24. Furdyna, J. K. & Kossut, J. (eds) Diluted Magnetic Semiconductors (AcademicPress, 1988).

25. Kolesnik, S. et al. Determination of antiferromagnetic interactions in Zn(Mn)O,Zn(Co)O and Zn(Mn)Te by inelastic neutron scattering. J. Appl. Phys. 99,08M122 (2006).

26. Droubay, T. C. et al. Correlated substitution in paramagnetic Mn2þ-doped ZnOepitaxial films. Phys. Rev. B 79, 155203 (2009).

27. Schwartz, D. A., Norberg, N. S., Nguyen, Q. P., Parker, J. M. & Gamelin, D. R.Magnetic quantum dots: synthesis, spectroscopy and magnetism of Co2þ- andNi2þ-doped ZnO nanocrystals. J. Am. Chem. Soc. 125, 13205–13218 (2003).

28. Taft, K. L. et al. [Fe(OMe)2(O2CCH2Cl)]10, a molecular ferric wheel. J. Am.Chem. Soc. 116, 823–832 (1994).

29. Kahn, O. Molecular Magnetism (VCH, 1993).30. Shapira, Y. et al. Nearest-neighbor exchange constant and Mn distribution in

Zn1–xMnxTe from high-field magnetization step and low-field susceptibility.Phys. Rev. B 33, 356–365 (1986).

31. Andrearczyk, T. et al. Spin-related magnetoresistance of n-type ZnO:Al and Zn1–

xMnxO:Al thin films. Phys. Rev. B 72, 121309 (2005).32. Merkulov, I. A. et al. Kinetic exchange between the conduction band electrons

and magnetic ions in quantum-confined structures. Phys. Rev. Lett. 83,1431–1434 (1999).

33. Bussian, D. A. et al. Tunable magnetic exchange interactions in manganese-doped inverted core–shell ZnSe–CdSe nanocrystals. Nature Mater. 8,35–40 (2009).

34. Myers, R. C., Poggio, M., Stern, N. P., Gossard, A. C. & Awschalom, D. D.Antiferromagnetic s–d exchange coupling in GaMnAs. Phys. Rev. Lett. 95,017204 (2005).

35. Besombes, L. et al. Probing the spin state of a single magnetic ion in anindividual quantum dot. Phys. Rev. Lett. 93, 207403 (2004).

36. Badaeva, E., Feng, Y., Gamelin, D. R. & Li, X. Investigation of pure and Co2þ-doped ZnO quantum dot electronic structures using density functional theory:choosing the right functional. New J. Phys. 10, 055013 (2008).

37. Badaeva, E. et al. Theoretical characterization of electronic transitions in Co2þ-and Mn2þ-doped ZnO nanocrystals. J. Phys. Chem. C 113, 8710–8717 (2009).

38. Maingault, L., Besombes, L., Leger, Y., Bougerol, C. & Mariette, H. Inserting onesingle Mn ion into a quantum dot. Appl. Phys. Lett. 89, 193109 (2006).

39. Leger, Y., Besombes, L., Maingault, L., Ferrand, D. & Mariette, H. Geometricaleffects on the optical properties of quantum dots doped with a single magneticatom. Phys. Rev. Lett. 95, 047403 (2005).

40. Frisch, M. J. et al. Gaussian Development Version (Gaussian, Inc., 2007).41. Kisi, E. H. & Elcombe, M. M. u Parameters for the wurtzite structure of ZnS and

ZnO using powder neutron diffraction. Acta Cryst. C45, 1867–1870 (1989).42. Borras-Almenar, J. J., Clemente-Juan, J. M., Coronado, E. & Tsukerblat, B. S.

MAGPACK: A package to calculate the energy levels, bulk magnetic propertiesand inelastic neutron scattering spectra of high nuclearity spin clusters.J. Comput. Chem. 22, 985–991 (2001).

AcknowledgementsThis work was supported by the US National Science Foundation (CHE 0628252-CRC toD.G. and X.L., DGE-0504573 (IGERT) to K.W.). Additional support from Gaussian Inc,the Research Corporation (Cottrell Scholar, D.G.), the Dreyfus Foundation(Teacher/Scholar, D.G.) and the University of Washington is gratefully acknowledged. E.B.thanks the Center for Nanotechnology at the University of Washington for UIF fellowshipsupport. S.O. was supported by a fellowship for prospective researchers by the SwissNational Science Foundation (contract no. PBBE2-115064). EPR instrumentation supportfrom the Center for Ecogenetics and Environmental Health UW Center grant no. P30ES07033 from the National Institutes of Environmental Health Sciences, NIH, is gratefullyacknowledged. TEM data were collected at Environmental Molecular Science Laboratory(PNNL), a national scientific user facility sponsored by the Department of Energy, andC. Wang (PNNL) is thanked for valuable assistance with this instrument.

Author contributionsS.O., K.W. and W.L. performed the experiments. Y.F. and E.B. performed the DFTcalculations. All authors discussed the experimental and computational results andanalysis. S.O, Y.F., K.W., E.B., W.L., X.L. and D.G. co-wrote the paper.

Additional informationReprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/. Correspondence and requests for materials should beaddressed to D.R.G.

NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2009.221 ARTICLES

NATURE NANOTECHNOLOGY | VOL 4 | OCTOBER 2009 | www.nature.com/naturenanotechnology 687

© 2009 Macmillan Publishers Limited. All rights reserved.