ARTICLE

Room-temperature polarized spin-photon interfacebased on a semiconductor nanodisk-in-nanopillarstructure driven by few defectsShula Chen1, Yuqing Huang1, Dennis Visser2, Srinivasan Anand2, Irina A. Buyanova 1 & Weimin M. Chen 1

Owing to their superior optical properties, semiconductor nanopillars/nanowires in one-

dimensional (1D) geometry are building blocks for nano-photonics. They also hold potential

for efficient polarized spin-light conversion in future spin nano-photonics. Unfortunately, spin

generation in 1D systems so far remains inefficient at room temperature. Here we propose an

approach that can significantly enhance the radiative efficiency of the electrons with the

desired spin while suppressing that with the unwanted spin, which simultaneously ensures

strong spin and light polarization. We demonstrate high optical polarization of 20%, inferring

high electron spin polarization up to 60% at room temperature in a 1D system based on a

GaNAs nanodisk-in-GaAs nanopillar structure, facilitated by spin-dependent recombination

via merely 2–3 defects in each nanodisk. Our approach points to a promising direction for

realization of an interface for efficient spin-photon quantum information transfer at room

temperature—a key element for future spin-photonic applications.

DOI: 10.1038/s41467-018-06035-1 OPEN

1 Department of Physics, Chemistry and Biology, Linköping University, SE58183 Linköping, Sweden. 2 Department of Applied Physics, KTH Royal Institute ofTechnology, SE16440 Kista, Stockholm, Sweden. Correspondence and requests for materials should be addressed to S.C. (email: shuch@ifm.liu.se)or to W.M.C. (email: wmc@ifm.liu.se)

NATURE COMMUNICATIONS | (2018) 9:3575 | DOI: 10.1038/s41467-018-06035-1 | www.nature.com/naturecommunications 1

1234

5678

90():,;

Semiconductor nanopillars (NPs)/nanowires (NWs) in the1D geometry and their artificially designed lattices, likeperiodic arrays, represent a new class of metamaterials.

They exhibit interesting optical properties superior to their flat-surface counterparts of semiconductor materials in the 3D and2D geometry, such as enhanced light absorption and emission,waveguiding of light and light confinement thanks to a highrefractive index contrast with respect to the surroundings (e.g.,air), a large surface-to-volume ratio, etc. In the past decade, theyhave been extensively investigated as building blocks in nano-photonics, nano-photovoltaics, and nano-sensing1,2, includinglight emitting diodes (LEDs)3–5, photovoltaic (PV) cells6–9,photo-detectors10–13, nanolasers14,15, nonlinear optics16–19, andbio-sensors20–22, all on the nanoscale. NPs/NWs also function asnatural axial Fabry-Perot (FP) and radial whispering-gallery(WG) mode cavities with tailorable optical modes23,24. Theconfinement of photonic modes and their propagation along theNP/NW axis, promising for guiding light on the nanoscalebetween different photonic components, offer great potential forapplications in photonic integrated circuits. These superioroptical properties have also placed NPs/NWs in a pole position tobe ideal interfaces and building blocks bridging between spin-and photon-coded quantum information in future spin-photonicdevices and integrated circuits.

A prerequisite for the success of such spin-photonic devices isefficient generation of spin-polarized carriers in NPs/NWs.Unfortunately, reports of spin generation in NPs/NWs remainsparse, in sharp contrast to numerous studies being carried out insemiconductors in the form of thin films25–27, quantum wells(QWs)27–29 and QW lasers30–32, and quantum dots (QDs)33–35.The earlier studies in NPs/NWs have focused on all electricalspin-FET and have largely been restricted to low temperatures inaddition to requiring an external magnetic field36–39. To date, thehighest electron spin polarization achieved in a 1D system atroom temperature (RT) is 28.2% from a blue nanolaser based onGaN NWs in an applied magnetic field40. At present, generationof highly spin-polarized electrons at RT represents a majorchallenge in exploring the spin degree of freedom in NPs/NWs.

Moreover, a polarized spin-photon interface must exhibit bothstrong spin polarization and high radiative efficiency. These tworequirements have been found to be difficult to meet simulta-neously in a semiconductor system, as the conditions favoringspin polarization are generally unfavorable for radiative efficiencyor vice versa (see below for details). Moreover, the maximum spin

polarization achievable in a conventional approach is restricted tothe initially injected spin polarization which is commonly ratherlow. This imposes a seemingly fundamental obstacle in realizationof an efficient polarized spin-photon interface.

Here, we employ an approach that can effectively overcome theobstacle by dissociating the radiative efficiencies of the majorityand minority spins, i.e., by significantly enhancing the radiativeefficiency of the emitter with the desired spin while suppressingthat with the unwanted spin, which simultaneously ensuresstrong electron spin polarization. We report on our achievementof electron spin polarization up to 60% at RT without a magneticfield, which is in fact much higher than the initially injected spinpolarization, in a periodic array of dilute nitride GaNAs nanodisk(ND)-in-GaAs NP structure (referred to as GaNAs/GaAs DiPhereafter). We show that such a high electron spin polarizationdegree is facilitated by spin-dependent recombination (SDR) via,strikingly, merely 2–3 intrinsic point defects (i.e., gallium inter-stitials Gai) in each GaNAs ND. Furthermore, we demonstratethat the NP array outperforms its planar counterpart in terms ofboth light absorption and detection efficiency thanks to thephotonic crystal type of arrangement and the waveguide characterof the NP array, which efficiently converts the electron spinpolarization to polarized light in the fundamental HE11 cavitymode suitable for fiber-optic coupling. This work thereforedemonstrates the potential of the proposed approach in therealization of RT-operational, polarized nanoscale spin and lightemitters, e.g., the GaNAs/GaAs DiP structure, promising forpractical applications in future spin-photonics and quantumcommunication network technology.

ResultsGaNAs/GaAs DiP array design. The GaNAs/GaAs DiP arraywas fabricated from a multiple quantum well (MQW) sample by atop-down approach employing nanosphere lithography (NSL)and inductively coupled plasma reactive ion etching (ICP-RIE)processes, see Methods section for details. The MQW samplecontains seven GaNAs QWs, each with a thickness of 9 nm. TheGaAs barrier between the neighboring QWs has a thickness of 20nm. The N content varies over the range of 1.1–1.32% among theseven GaNAs QWs, see Supplementary Figure 1. Figure 1a showsa scanning electron microscopy (SEM) image of the as-etchedDiP array, which exhibits a hexagonal lattice with an inter-pillardistance, a0, of ~460 nm as defined by the hexagonal close-packednanosphere mask. Each NP features a tapered cylindrical shape

CB

a b c

z

yx

Fig. 1 GaNAs/GaAs nanodisk in nanopillar array design. a Scanning electron microscopy (SEM) image of the arrayed GaNAs nanodisk-in-GaAs nanopillars(DiP). b Side-view SEM image of an individual DiP, where the seven GaNAs nanodisks are shown in red color. The energy profile of the conduction band(CB) edge along the nanopillar axis is indicated on the left side. The valence band edge is nearly unperturbed by N alloying or strain, and therefore its profileis approximately flat and not shown here. The scale bars in a and b denote 400 and 100 nm, respectively. c Schematic illustration of the experimentalconfiguration

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06035-1

2 NATURE COMMUNICATIONS | (2018) 9:3575 | DOI: 10.1038/s41467-018-06035-1 | www.nature.com/naturecommunications

with smooth side walls and a rather uniform size distribution,manifesting high quality of the nanofabrication. From the mag-nified SEM image on an individual NP shown in Fig. 1b, theaverage diameter, d, of the NPs is found to gradually increasefrom the top, dtop= 240 ± 10 nm, to the bottom, dbottom= 360 ±10 nm, over the height, h, of 370 ± 10 nm. The seven etched dilutenitride GaNAs NDs within each NP are highlighted in red, withthe first ND at a distance of 70 nm from the top surface. Theperiodic modulation of the conduction band (CB) edge along theNP axis is depicted on the left side of Fig. 1b, where the CB energyis lower in the NDs due to incorporation of N41. The energy ofthe valence band (VB) edge of the NPs, on the other hand, israther unperturbed42 (not shown here), as the incorporated Nstates predominantly interact with the CB and the alloy-inducedstrain in the NDs is relaxed owing to the DiP geometry (seeSupplementary Note 1).

Enhanced light absorption and extraction. To investigate theeffect of the NP array on the light emission, we measured RTphotoluminescence (PL) from the GaNAs/GaAs DiP underunpolarized laser excitation at 800 nm and compare it with theunetched reference MQWs. The results are shown in Fig. 2a. ThePL spectra from both DiP and reference MQWs are dominated bytwo peaks, i.e., the shorter-wavelength one at 870 nm from theGaAs barrier and the broader one at the longer-wavelength of1000 nm from the GaNAs NDs. Interestingly, the integrated PLintensities from GaNAs and GaAs of the DiP are stronger thanthat in the planar MQW structure by 1.55 and 3.04 times,respectively, under the same laser power Pexc. Given that theetched DiP has a smaller emitting volume than the MQWs and itadditionally suffers from enhanced non-radiative surface recom-bination associated with the unpassivated NP surface, theobserved stronger PL from the DiP array clearly demonstrates abetter light absorption and/or extraction efficiency than the pla-nar MQWs. This finding is consistent with the PL results fromother semiconductors with similar structures43,44. The measuredPL intensity, IPL, is affected by multiple factors45,

IPL / PexcαinVηαoutβ; ð1Þ

where αin(αout) is the in (out)-coupling coefficient which deter-mines the absorption (extraction) efficiency of light. The effective

volume of the emitter is denoted by V, which is proportional tothe material filling factor, f, i.e., V∝f. Here, f is 1 for the referenceMQW sample and 0.38 for the DiP structure that is averaged overseven size-tapered GaNAs NDs. The radiative recombinationefficiency (or the internal quantum efficiency) is denoted by η,and β represents the PL collection coefficient, i.e., the fraction offar-field emission that enters the numerical aperture (NA)-defined solid angle of a microscope objective. Based on thereflectivity measurements (see Supplementary Figure 2), wederive αin= 0.94 for the DiP structure and αin= 0.65 for theMQW sample, at the laser wavelength of 800 nm. From thetemperature-dependent PL measurements (see SupplementaryFigure 3), the GaNAs NDs in the DiP show a quantum efficiencyof η= 0.01, which is lower than η= 0.03 of the MQWs due tosurface recombination. We note that these values were obtainedunder unpolarized laser excitation, i.e., under the condition whenthe SDR effect is inactive. To gain quantitative insight into NP-tailored light output, we performed finite difference time domain(FDTD) calculations using Lumerical’s FDTD software. The PLradiation from the GaNAs NDs is modeled as unpolarized point-dipole emitters at 1000 nm, considering the band-to-band (BB)recombination involving both heavy-holes (HH) and light-holes(LH). The results are displayed in Fig. 2b for a single NP. The DiParray not only provides a vast density of nanostructured surfacefor efficient subwavelength light diffraction, but also it gives riseto a reduced contrast of effective refractive index with air, neff,which leads to higher light transmission into free space46,47. Moreimportantly, the shape of the DiP forms a free-standing wave-guide which can couple the emission into the cavity mode.Simulations on the emission intensity across the NP cross-sectiongiven in Fig. 2c–f show that light is mostly confined at the centerof the NPs, indicating a fundamental HE11-mode dominated PLoutput. All these factors can contribute to a higher PL detectionefficiency γ, defined by γ= αoutβ, for the DiP than the planarMQWs. To confirm this, we calculated γ as a function of dipoleemitter depth (or ND position) from the top surface for the twosample structures. As shown in the inset of Fig. 2a, γ of the DiPincreases with increasing ND depth, reaching a maximum of 2.9%for the deepest ND. Its averaged γ of 2.3% is almost eight timeshigher than γ= 0.3 % for the MQWs. By applying all thesederived parameters into eq. 1, the detected PL intensity from theGaNAs NDs in the DiP is calculated to be 1.4 times stronger than

x y

z

x

y

z

|Ex|2

|Ez|2

|Ey|2

e

850 900 950 1000 1050

1.45 1.4 1.35 1.3 1.25 1.2

70 120 170 2200.2

0.41

2

3

Wavelength (nm)

Photon energy (eV)

PL

inte

nsity

(a.

u.)

MQWs

DiPsDepth (nm)

� (%

)

a

DiPs

MQWs

|E|2

|E|2

b c d

f

0 10 V2 m–2 10 V2 m–20

Fig. 2 Enhanced light absorption and extraction at room temperature. a Photoluminescence (PL) spectra from the GaNAs nanodisk-in-GaAs nanopillars(DiPs) and the multiple quantum wells (MQWs) structure under unpolarized laser excitation at 800 nm. The inset shows the detection efficiency, γ, ofboth structures as a function of the dipole-emitter depth, i.e., the nanodisk (ND)/QW position from the top surface. b The side view of the finite differencetime domain (FDTD) simulations of the squared electric-field amplitude (proportional to the light intensity) from the GaNAs NDs with distances of 75, 104,133, 162, 191, 220, and 250 nm from the top surface. The simulation was taken at the middle cross-section of the DiP as marked by the dashed line in f. c–eThe top views of the FDTD simulations of the ND light emission intensity with x, y, and z polarization components, respectively, together with that of thetotal PL intensity (f). The images are obtained in the plane indicated by the dashed line in b

NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06035-1 ARTICLE

NATURE COMMUNICATIONS | (2018) 9:3575 | DOI: 10.1038/s41467-018-06035-1 | www.nature.com/naturecommunications 3

that from the MQWs, which is quite close to the value of 1.55deduced from the PL results. The good agreement between theexperiment and simulations verifies the optimized light man-agement in the DiP structure with a promoted light-matterinteraction and mode-profiled PL output. The obtained superiornano-photonic properties identify the GaNAs/GaAs DiP as anideal emitter to couple with fiber-optics-based quantum networksor in photonic integrated circuits.

Efficient RT spin generation enabled by few defects. Besides theadvantage of enhanced light absorption and detection as estab-lished above, we shall now demonstrate efficient RT spin gen-eration in the same GaNAs/GaAs DiP via the defect-mediatedSDR processes. The Gai interstitial defects have been shown to beresponsible for SDR in GaNAs from the earlier optically detected-magnetic-resonance (ODMR) study27. The principle of the spingeneration via SDR is illustrated in Supplementary Figure 4 anddescribed in more detail below, which relies on electron spinpolarization of the Ga2þi defects with a single bound electron. If apreferential spin orientation of CB electrons is created bycircular-polarized (σ+ or σ−) light, even only slightly, SDR willlead to a buildup of spin polarization of both CB and defect

electrons during the optical pumping. For example, under σ+

light excitation along the NP axis as shown in Fig. 1c, a spinimbalance is generated for CB electrons with more spin-downthan spin-up CB electrons as shown in Supplementary Figure 4,following the optical selection rule of the BB transition (Fig. 3d).This imbalance can be characterized by the ratio n−/n+ > 1between the numbers of the spin-down and spin-up CB electrons,denoted by n− and n+, respectively. Dictated by the Pauliexclusion principle, the capture of a CB electron by the Ga2þidefect is allowed only if the CB electron and the electron bound toGa2þi have opposite spin orientations. As the probability of thespin-up (spin-down) Ga2þi defect in capturing a CB electronscales with the number of spin-down (spin-up) CB electrons, thespin-up Ga2þi defect should thus have a higher probability by afactor of n−/n+ in capturing a spin-down CB electron than thespin-down Ga2þi defect in capturing a spin-up CB electron, asillustrated in Supplementary Figure 4a. After the capture, one ofthe two spin-paired electrons at the Ga1þi defect will recombinewith a spin-unpolarized VB hole, leaving behind with an equalprobability either a spin-up or a spin-down electron at the defect.(Here, the VB hole spins are regarded as unpolarized in the SDRprocesses due to a stronger spin–orbit interaction and HH-LH

850 900 950 1000 1050

Wavelength (nm)850 900 950 1000 1050

0

5

10

15

20

25

0

20

40

60

1.00

1.05

1.10

1.15

1.20

1.45 1.4 1.35 1.3 1.25 1.2b e

f

VB

CB

+1/2–1/2 +3/2–3/2

+1/2–1/2

×10

CB

VB

a

PL PL

1.45 1.4 1.35 1.3 1.25 1.2

Wavelength (nm)

I(σ+

)/I(

σx )

PL

inte

nsity

(a.

u.)

PL

inte

nsity

(a.

u.)

Photon energy (eV) Photon energy (eV)

σ+

σx

σ+σX

σ+

σ+ σ+

σ–

σ– σ–

×10

c

No SDR SDR

PP

L (%

)

–Pe

(%)

GaAs GaNAs NDs GaAs GaNAs NDs

d

Fig. 3 Efficient room temperature spin generation enabled by few defects. a Schematic illustration of the defect-driven spin generation via spin-dependentrecombination (SDR) under σ+ excitation (the right panel), which shows stronger and co-polarized photoluminescence (PL) than that without the SDReffect under σx excitation (the left panel). The gray horizontal bars represent the defect level. b PL spectra excited by σ+ and σx laser light. c Spectraldependence of the I(σ+)/I(σX) ratio. d Band diagram illustrating the polarized band-to-band transitions involving valence band (VB) heavy-hole (±3/2) andlight-hole (±1/2) states. e σ+ and σ− polarized PL spectra under the σ+ excitation. f Spectral dependence of the PL and conduction band (CB) electron spinpolarizaion. In converting PPL–Pe, the light depolarization factor arising from the finite solid angle of light detection by the microscope objective with thegiven numerical aperture (NA) has also been taken into account. Since, the PL emission from GaAs is at least partly contributed from the planar GaAssubstrate, Pe in GaAs could be overestimated by a factor up to 1.5

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06035-1

4 NATURE COMMUNICATIONS | (2018) 9:3575 | DOI: 10.1038/s41467-018-06035-1 | www.nature.com/naturecommunications

coupling that randomize the initially polarized hole spins withinpicoseconds).48,49. As a result, after one round of the SDR pro-cess, more spin-up Ga2þi defects are converted to the spin-downGa2þi defect than the other way around, with a factor of n−/n+between their conversion rates. This leads to an imbalancebetween spin-down and spin-up Ga2þi defects, as illustrated inSupplementary Figure 4b, which can be characterized by theirratio N−/N+ > 1. With each round of such capture and recom-bination, more and more spin-up Ga2þi defects will be convertedto the spin-down Ga2þi defects. At the same time, an increasinglylarger ratio N−/N+ will in turn preferably deplete the spin-up(minority spin) CB electrons, i.e., a spin filtering effect, therebyfurther increasing the ratio n−/n+ and thus the CB electron spinpolarization. Eventually this positive feedback process betweenn−/n+ and N−/N+ will drive both the Ga2þi defect and CB elec-tron spins towards the majority-spin (i.e., spin-down) orientationvia the so-called dynamic spin polarization (DSP) process50. Asshown in the right panel of Fig. 3a, the resulting spin blockadeeffect prevents further capture and depletion of CB electrons viathe defects, thereby giving rise to a stronger and highly circular-polarized BB emission. During the time when the defect electronspin is preserved, any minority-spin CB electrons generated byeither poor optical spin generation by light or due to spinrelaxation will immediately be depleted by the SDR processes viathe defects as illustrated in Supplementary Figure 4c. This ensuresthat spin polarization of CB electrons can be maintained at asubstantially higher degree than the initial value created byoptical pumping without DSP. This is in sharp contrast to thecase without SDR and DSP under linear-polarized (σX) laserexcitation, see the left panel of Fig. 3a, when both spin-up andspin-down electrons are equally generated in CB, followed bytheir capture with an equal probability by the spin-unpolarizedGai defects and recombination with VB holes. No spin polar-ization of either the defect electrons or CB electrons is expectedduring the capture and recombination processes, except that thisfast non-radiative process equally depletes free carriers of bothspin orientations resulting in weak and unpolarized PL emission.

The most crucial condition for the success of the SDR and DSPprocesses in efficient spin generation is τsc > τs > τnr. Here, a muchlonger spin lifetime of the defect electron, denoted by τsc, ensuresthat the Ga2þi defects remain spin polarized during the time whenCB electrons undergo spin relaxation. Whenever that happens,the defects will immediately filter out these spin-flipped CBelectrons as long as their capture time by the defects (τnr) isshorter than the CB electron spin relaxation time τs. As τnr istypically in the order of ps, the spin filtering and spinamplification can be accomplished within a very short time.The efficiency of the SDR and DSP processes can also beaccelerated by increasing circularly polarized light intensitybecause it is accompanied by an increasing probability of theGa2þi defects in capturing CB electrons and thus an increasingnumber of the SDR cycles within the same period of time.

The boost in the PL intensity due to the spin blockade effectcan be characterized by the ratio of I(σ+ or σ−)/I(σX), where I(σ+/−/X) represents the PL intensity under right-circular/left-circular/linear-polarized laser excitation, respectively. Figure 3bshows representative PL spectra excited by the 800 nm laser lineat Pexc of 11.7 mW, where the SDR process in the GaNAs NDs isclearly evident from an apparently stronger emission under σ+

than σX excitation. This can be further corroborated with anunchanged PL signal from GaAs, where no SDR occurs. Thecorresponding I(σ+)/I(σX) ratio in Fig. 3c clearly exhibits avariation from a value of ~1.06 to ~1.15 with decreasing emissionenergy. The spectral variation of the I(σ+)/I(σX) ratio suggests thecoexistence of SDR processes with different efficiencies. Indeed,

our secondary ion mass spectroscopy (SIMS) measurements onthe reference MQWs reveal a N fluctuation from [Nmin]= 1.1%to [Nmax]= 1.32% among the seven QWs (see SupplementaryFigure 1), corresponding to a bandgap energy shift from 1.252 to1.228 eV which is comparable with the spectral width of thebroad GaNAs PL. Since a larger N content not only reduces thebandgap energy but also facilitates the formation of Gai27,51, andthus SDR efficiency, a higher I(σ+)/I(σX) ratio is expected at thelower emission energy that is in excellent agreement with ourexperimental finding. The lowest and highest values of the I(σ+)/I(σX) ratio are thus derived from the NDs with [Nmin] and [Nmax],respectively.

To directly determine spin polarization of CB electrons in theGaNAs NDs, we monitored the circular-polarized PL compo-nents under circularly polarized optical pumping (e.g., σ+). TheND emission is strongly co-polarized with the excitation light asshown in Fig. 3e. The derived PL circular polarization degree PPL,defined by PPL= (I+−I−)/(I++I−) where I+/− denotes the σ+/−-polarized PL intensity, is shown in Fig. 3f. It exhibits aspectral variation similar to the I(σ+)/I(σX) ratio, with PPL= 14and 20% from the NDs with [Nmin] and [Nmax], respectively. Itshould be noted that due to the degeneracy of HH and LH VBstates in the strain-relaxed NDs, see Supplementary Note 1 andSupplementary Figure 5–6, the detected PPL comprises contribu-tions from both HH and LH subbands of opposite helicity. Basedon the well-known 3-to-1 ratio between the oscillator strengths ofthe BB optical transitions involving the HH and LH states48,52, asillustrated in Fig. 3d, the spin polarization of CB electrons is twiceof PPL, i.e., Pe=−2PPL. By further taking into account the lightdepolarization factor arising from the finite solid angle of lightdetection by the microscope objective with the given NA (seeSupplementary Figure 7 and Supplementary Note 2), Pe generatedin the GaNAs NDs with [Nmin]= 1.1% to [Nmax]= 1.32% can bedetermined to be 43 and 60%, respectively. These values areestimated based on the results from Supplementary Figure 7 that100% electron spin polarization should generate only 33% lightcircular polarization under our experimental conditions andassuming full LH-HH degeneracy.

To evaluate to what extent the SDR effect via the defects havecontributed to the spin generation in the DiP, we carried out adetailed study of excitation power-dependent Pe and I(σ+)/I(σX)ratio for the NDs with [Nmin] and [Nmax]. A summary of theresults is shown in Fig. 4. For both N compositions, Pe increaseswith increasing Pexc owing to accelerated dynamic processes ofSDR. This can also be corroborated by a constant Pe of ~8% forthe GaAs emission, where the SDR effect is absent. From the bestfits of a nonlinear coupled rate equation analysis to the power-dependent Pe profiles, see Supplementary Note 3 for details, theconcentration of the Gai defects can be estimated to be ~2.6 ×1015 cm−3 and 4.1 × 1015 cm−3 in the NDs with [Nmin] and[Nmax], respectively. This approximately corresponds to 2 defectsin each ND with [Nmin] and 3 defects in each ND with [Nmax]. Itis remarkable that such few defects have led to strong Pe up to60% at RT, demonstrating the extremely high efficiency androbustness of the Gai-mediated spin generation processes.

The rate equation analysis also shows that the spin polarizationat very low Pexc before the optically pumped dynamic spinpolarization process of the defect electrons takes place, denotedby P0

e , is merely 5%. This value reflects the CB electron

polarization without the SDR effect. The low P0e ð¼ Pi

e1þτ=τs

Þ inthe GaNAs NDs, much lower than the maximum allowed 50%upon optical pumping of the GaAs barrier, originates from twoparts. The first part stems from spin loss of the light emitter itself,by a factor of 1

1þτ=τs(about 0.9 in our case), due to CB electron

spin relaxation within its lifetime. The second contribution, often

NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06035-1 ARTICLE

NATURE COMMUNICATIONS | (2018) 9:3575 | DOI: 10.1038/s41467-018-06035-1 | www.nature.com/naturecommunications 5

the more important one, is related to a low value of Pie (about

5.6% in our case), which reflects significant spin loss duringenergy relaxation of hot electrons/excitons within the GaAsbarriers before spin injection to the GaNAs NDs, spin transportacross the GaAs/GaNAs interfaces, and energy relaxation with theNDs after spin injection. In fact, it is such severe spin loss that hasprevented GaAs, or any other non-magnetic semiconductorwithout SDR, from efficient spin generation at RT. When theSDR effect is activated as Pexc increases, Pe increases andeventually approaches a saturation value of 60% (43%) for theNDs with [Nmax] ([Nmin]) due to the saturated SDR process. Theobserved defect-enabled spin gain can be characterized by a spinamplification factor (SAF), i.e., SAF ¼ Pe=P

0e , yielding a max-

imum SAF of 11 (7.6) for the NDs with [Nmax] ([Nmin]). Thesefindings further consolidate the effectiveness of the Gai defects forspin amplification in the GaNAs NDs, which could effectivelyovercome the severe spin loss in the way no any other approachcan.

Spin polarization versus internal quantum efficiency. An effi-cient polarized spin-photon interface must fulfill two stringentrequirements: strong electron spin polarization and efficient lightemitting. In a most commonly occurring case of a semiconductorwithout the defect-enabled spin amplification effect, these tworequirements favor contradicting conditions. This is because theelectron spin polarization P0

e can be described in its simplest formby the equation52

P0e ¼

Pie

1þ τ=τs: ð2Þ

Here, τ denotes the total lifetime of the electrons regardless oftheir spin orientations, which is contributed from both radiativeand non-radiative lifetime (i.e., τr and τnr) following therelationship 1

τ ¼ 1τrþ 1

τnr. ( 1

τnr¼ γeNc, where Nc is the defect

concentration responsible for the non-radiative recombinationand γe is the trapping coefficient of CB electrons by the defect.)

From eq. 2, it is apparent that P0e is restrained over the range of

Pie

1þτ=τs� P0

e � Pie. Achieving the maximum radiative recombina-

tion efficiency η= τ/τr= 100% is accompanied by the minimum

electron spin polarization P0e ¼ Pi

e1þτr=τs

. As τr is usually found to becomparable or longer than τs in semiconductor materials andnanostructures, the maximum η is obtained at the expense ofsevere spin loss. To approach the maximum value of P0

e � Pie, on

the other hand, τ needs to be much shorter than τs. This requiresintroduction of non-radiative recombination such that τ ≈ τnr ≪τs53, which simultaneously minimizes η, as η= τ/τr≪1 under thesame condition. Such an anti-correlation between Pe and η can beclearly visualized from the simulations shown by the gray dashedlines in Fig. 5.

This seemingly fundamental obstacle in simultaneouslyachieving both strong Pe and high η can be effectively overcomeby the defect-mediated SDR effect presented in this work. The keyto the success here is a strong spin polarization degree of thedefect electrons such that the non-radiative recombination timesof the spin-up and spin-down CB electrons, namely, τ ±

nr ¼ 1γeN�

,

becomes markedly different. Here, N± represents the concentra-tion of the Gai defects in the paramagnetic charge state occupiedby a single electron with the subscript ‘±’ denoting the spin-upand spin-down orientation of electrons. In the extreme case whenthe defect electrons are completely spin polarized, say N+ ≈Nc

and N− ≈ 0, non-radiative recombination of the CB electrons withthe majority-spin (i.e., spin-up) diminishes since τþnr ¼ 1

γeN�! 1

yielding η+→100%. This is in sharp contrast to the CB electronswith the minority-spin (i.e., spin-down) that suffer strong non-radiative recombination in the same way as the case without theSDR effect described above, because τ�nr ¼ 1

γeNþ. This strong

deviation in non-radiative recombination via the defects betweenthe majority-spin and minority-spin CB electrons in turn drivesCB electrons toward amplified spin polarization (Pe>P

ie) well

exceeding the maximum limit of P0e � Pi

e imposed in the absenceof the SDR effect. For a quantitative analysis, we performeddetailed simulations of Pe and η±= τ±/τr as a function of τ/τr with

0

10

20

30

40

50

60

0

2

4

6

8

10

12

0 2 4 6 8 10 120

5

10

15

0

1

2–P

e (%

)–P

e (%

)

Power (mW)

SA

F P

/Pe

e0

SA

F P

/Pe

e0GaAs

a

b

GaNAs NDs

Fig. 4 Room temperature spin amplification enabled by few defects. a Power-dependent Pe and spin amplification factor SAF ¼ Pe=P0e from the GaNAs

nanodisks (NDs) with the [Nmin] (the open triangles) and [Nmax] (the filled circles). The solid symbols are experimental data, and the solid lines are thefitting curves obtained by the rate equation analysis. b For comparison, power-dependent Pe from GaAs is also shown by the gray symbols. The verticalbars indicate the experimental error bars which are derived from the statistics of five measurements. The dashed horizontal line marks P0e

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06035-1

6 NATURE COMMUNICATIONS | (2018) 9:3575 | DOI: 10.1038/s41467-018-06035-1 | www.nature.com/naturecommunications

the aid of the coupled rate equations given in SupplementaryNote 3. The results are shown in Fig. 5, which demonstrate that,over the regime when the SDR is highly active, both Pe and η (forthe desired spins) can be significantly enhanced by orders ofmagnitude from that without the SDR effect. We should notethat, as the minority-spin CB electrons are depleted by the defect-mediated non-radiative recombination processes, the totalquantum efficiency ηtotal, including PL yield from bothmajority-spin and minority-spin CB electrons, should lie betweenthe quantum efficiency values of the two. The ηtotal valuescalculated based on the rate equations are shown as a function ofτ/τr in Fig. 5, together with the values of the majority-spin andminority-spin quantum efficiency. The simulations confirm thesignificant enhancement of not only the majority-spin quantumefficiency but also total PL quantum yield by activating the SDReffect. The upper bound of ηtotal achievable by the defect-mediated SDR effect is determined by the number of minority-spin CB electrons needed to be depleted via the non-radiativeprocesses. Theoretically, it can approach 50% when Pi

e is close tozero and 100% in the other extreme case of Pi

e close to 100%.It is clear from the simulations shown in Fig. 5 that the exact

extent of the enhancement in both Pe and η by the SDR effectcritically depends on the defect concentration (thus τ), CBelectron spin relaxation time τs and the initial spin polarizationPie. This provides a guideline for further improvements by

increasing the SDR defect concentration and by suppressing CBelectron spin relaxation through, e.g., quantum confinement. Thesimulations also show the potential of our approach tosimultaneously fulfill both requirements for strong electron spinpolarization with Pe approaching 100% and high radiativerecombination efficiency in the studied DiP structure. Futureeffort is also required to improve polarized spin-light conversionby removing the HH-LH degeneracy–the source of partialcancellation (by 1/2) in emitting light polarization, through

strain engineering and/or quantum confinement such that the BBtransition at RT only involves the HH states.

We note possible alternative approaches to reduce the radiativelifetime by exploring the Purcell effect and spin lasers54–57.Indeed, these alternative approaches in principle have thepotential to increase both Pe and η. However, as the Purcelleffect is not spin selective and therefore is incapable of spinamplification, it can only suppress spin loss of the emitter byreducing τ/τs. As a result, Pe is still fundamentally limited by the

initial spin polarization of the injected electrons Pie, since P0

e ¼Pie

1þτ=τs� Pi

e no matter how small value τ/τs is. Furthermore, toachieve the Purcell effect, either metal-based nanoplasmonics ordielectric Fabry-Perot/whispering-gallery cavities are required toreduce the spontaneous radiative lifetime by increasing the light-emitter interaction. It is commonly known that metal-basednanoplasmonics introduces additional Ohmic loss and could inaddition cause severe depolarization of the circularly polarizedemitting light from a 1D structure, which are undesirable for aspin-photon interface. Spin-lasers are attractive as they arecapable of spin amplification under the injection condition whenthe lasing threshold for the majority spin is reached whereas it isnot for the minority spin. The main drawback is that they involvemore complex nanofabrication processes and require a highinjection level (thus high power consumption). More critical isperhaps the requirement for well-separated injection levels atwhich the majority-spin and minority-spin start lasing57, whichdemands a high P0

e before lasing, i.e., before the spin amplificationtakes place. For example, P0

e = 20% in the spin laser based onGaAs/AlGaAs MQWs reported in ref.57. Considering CB electronspin polarization is typically rather weak in a commonsemiconductor at RT without spin amplification, e.g., P0

e = 5%in our case of the GaNAs/GaAs DiPs, applications of spin lasersin material systems with a week P0

e , where spin amplification is

+

100

100

101

101

102

10–2

10–4 10–3 10–2 10–1 100

10–1

102

a

Pe

(%)

� ±/�

r (%

)

�/�r

10–4 10–3 10–2 10–1 100

�/�r

10–4 10–3 10–2 10–1 100

�/�r

10–4 10–3 10–2 10–1 100

�/�r

dcb

hgfe

+ ++

Fig. 5 Spin polarization versus internal quantum efficiency. a–d Conduction band electron spin polarization Pe. e–h Radiative recombination efficiency of themajority-spin (spin-up) and minority-spin (spin-down), η±= τ±/τr, as well as the total radiative recombination efficiency. They are plotted as a function ofτ/τr, calculated from the rate equation analysis using the following parameters: a, e τsc= 2τs, τr= 10τsc, P0e ¼ 5%; b, f τsc= 2τs, τr= τsc, P0e ¼ 5%; c, g τsc=2τs, τr= 0.1τsc, P0e ¼ 5%; d,h, τsc= 2τs, τr= 0.1τsc, P0e ¼ 25%. τr is fixed at 3000 ps in all cases, and 1

τ ¼ 1τrþ γeNc. The simulations corresponding to the

cases with (without) the defect-mediated SDR are displayed by the colored solid lines (the gray dashed lines). The green and red vertical arrows indicatethe enhancements of electron spin polarization and radiative recombination efficiency induced by the few-defect-mediated SDR effect, respectively

NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06035-1 ARTICLE

NATURE COMMUNICATIONS | (2018) 9:3575 | DOI: 10.1038/s41467-018-06035-1 | www.nature.com/naturecommunications 7

most needed, could be rather restricted. In our work, we showboth experimentally and theoretically that the SDR-mediated spinamplification process remains highly effective even with P0

e = 5%or even lower. The most challenging task is to identify a suitablespin-filtering defect and to controllably incorporate them in ahost lattice. In GaNAs, the spin-filtering Gai defects have beenidentified by the optically detected-magnetic resonance techni-que27 and their concentrations can be controlled by e.g., Ncomposition, growth temperature and post-growth thermalannealing58.

DiscussionIn conclusion, we have proposed and realized an efficient novelpolarized spin-light emitter at RT based on a high-quality peri-odic array of the GaNAs/GaAs DiP. We observed a high PLcircular polarization degree of 20%, which corresponds to record-efficient spin generation of CB electrons in a semiconductormaterials system with a 1D geometry, up to 60% at RT without anapplied magnetic field. Strikingly, such achievement is enabled bythe SDR effect via merely 2–3 defects in each ND. The func-tionality of the spin amplification with a spin gain up to 11 timesat RT provided by such few defects is unique and could possiblybe found indispensable to achieve nearly 100% electron spinpolarization, which no any other approach is capable of. Weshowed that our approach holds the potential to simultaneouslyfulfill the requirements for both strong electron spin polarizationand high internal quantum efficiency, thereby overcoming aseemingly fundamental limit imposed on semiconductors in thisrespect. Combining with an optimized light in- and out-couplingstrength and the waveguide-tailored spin output for fiber-opticinterconnection, the nano-photonic structure of the GaNAs/GaAsDiP is promising for a highly sensitive and efficient nano-sizedinterface for spin-to-photon quantum information transfer at RT,paving the way for practical applications in future nanoscale spin-photonics and quantum communication networks.

The demonstrated defect-enabled spin generation and polar-ized spin-photon conversion also have the following addedadvantages: (i) the defect-mediated SDR and the resulting spinamplification processes are thermally activated, which warrantsefficient and fast spin generation and manipulation at RT or evenhigher temperatures; (ii) the giant bandgap bowing effectaccompanying dilute-alloying of GaAs with nitrogen in theGaNAs/GaAs DiP significantly shifts the emission wavelengthtoward the near infrared spectral range suitable for fiber-opticcommunications, while keeping the oscillator strength nearlyunchanged as a bright emitter; (iii) the heterostructures embed-ded in the DiP structures provide an additional degree of freedomin bandgap engineering and strain engineering for optimal deviceperformance; (iv) no magnetic fields required, thus avertingdesign complexities that are often associated with incorporatinglocal magnetic fields into device architectures; (v) freedom andease in selecting and switching spin direction by properlyorienting defect electron spins; (vi) availability of a rather maturetechnology base of GaAs, i.e., the parent material of GaNAs,which has widely been used today for high-frequency andoptoelectronic devices.

MethodsFabrication of the GaNAs/GaAs DiP array. We employed a cost-effective andmass-scalable top-down method, involving nanosphere lithography and inductivelycoupled plasma reactive ion etching processes59, to fabricate the dilute nitrideGaNAs/GaAs DiP array. A monolayer of SiO2 nanospheres with a diameter of~460 nm was dispersed by spin-coating onto a GaNAs/GaAs MQW sample,forming a hexagonal close-packed arrangement. The SiO2 nanosphere size wasreduced (on-site) to ~300–350 nm by subsequent fluorine-based RIE for 8 min.After that ICP-RIE using Cl2/H2/CH4 chemistry was performed under optimized

conditions for 2.5 min to etch the MQWs into a NP array. Finally, the top residualSiO2 NS mask was removed by HF etching.

Experimental techniques. The PL measurements were performed on a Horibamicro-PL (µ-PL) setup. A continuous-wave (CW) Ti: Sapphire laser was used as anexcitation source, with output at 800 nm to excite carriers across the GaAsbandgap. For optical orientation of CB electron spins, a set of linear polarizer andquarter-wave plate were used to create σ+ or σ− excitation, which could generate amaximum Pe of 50% in GaAs according to the selection rule of the BB opticaltransitions involving the VB HH and LH states. A microscope objective, ×50/NA0.5, was used to focus the excitation beam along the NP axis (z-axis in Fig. 1c)to a spot size of ~4 µm2, by which ~20 NPs were excited. At the highest excitationpower of 12 mW used in our experiments, the electron generation rate in each NDunder the above and below GaAs excitation is estimated to be ~2 × 1012 s−1 (~2 ps−1) and ~1 × 1012 s−1 (~1 ps−1), respectively. The resulting PL signal from theGaNAs DiP array was dispersed by a single-grating monochromator and detectedby a CCD camera. The circular-polarized emission components were resolved by asecond set of linear polarizer and quarter-wave plate.

Data availabilityAll relevant data are available from the authors upon request.

Received: 24 February 2018 Accepted: 13 August 2018

References1. Yan, R. X., Gargas, D. & Yang, P. D. Nanowire photonics. Nat. Photon. 3,

569–576 (2009).2. Zhang, A. Q., Zheng, G. F. & Lieber, C. M. Nanowires: Building blocks for

nanoscience and nanotechnology. (Springer, Switzerland, 2016).3. Guo, W., Zhang, M., Banerjee, A. & Bhattacharya, P. Catalyst-free InGaN/

GaN nanowire light emitting diodes grown on (001) silicon by molecularbeam epitaxy. Nano Lett. 10, 3355–3359 (2010).

4. Nguyen, H. P. T. et al. Breaking the carrier injection bottleneck of phosphor-free nanowire white light-emitting diodes. Nano Lett. 13, 5437–5442(2013).

5. Zhao, C. et al. Droop-free, reliable, and high-power InGaN/GaN nanowirelight-emitting diodes for monolithic metal-optoelectronics. Nano Lett. 16,4616–4623 (2016).

6. Krogstrup, P. et al. Single-nanowire solar cells beyond the Shockley–Queisserlimit. Nat. Photon. 7, 306–310 (2013).

7. Wallentin, J. et al. InP nanowire array solar cells achieving 13.8% efficiency byexceeding the ray optics limit. Science 339, 1057–1060 (2013).

8. Law, T., Greene, L. E., Johnson, J. C., Saykally, R. & Yang, P. D. Nanowire dye-sensitized solar cells. Nat. Mater. 4, 455–459 (2005).

9. Tang, J. Y., Huo, Z. Y., Brittman, S., Gao, H. W. & Yang, P. D. Solution-processed core-shell nanowires for efficient photovoltaic cells. Nat. Nanotech.6, 568–572 (2011).

10. Wang, J. F., Gudiksen, M. S., Duan, X. F., Cui, Y. & Lieber, C. M. Highlypolarized photoluminescence and photodetection from single indiumphosphide nanowires. Science 293, 1455–1457 (2001).

11. Kim, J. et al. Photon-triggered nanowire transistors. Nat. Nanotech. 12,963–968 (2017).

12. Bulgarini, G. et al. Avalanche amplification of a single exciton in asemiconductor nanowire. Nat. Photon. 6, 455–458 (2012).

13. Hayden, O., Agarwal, R. & Lieber, C. M. Nanoscale avalanche photodiodes forhighly sensitive and spatially resolved photon detection. Nat. Mater. 5,352–356 (2006).

14. Eaton, S. W., Fu, A., Wong, A. B., Ning, C. Z. & Yang, P. D. Semiconductornanowire lasers. Nat. Rev. Mater. 1, 16028 (2016).

15. Zimmler, M. A., Capasso, F., Müller, S. & Ronning, C. Optically pumpednanowire lasers: invited review. Semicond. Sci. Technol. 25, 024001 (2010).

16. Ren, M. L., Liu, W. J., Aspetti, C. O., Sun, L. X. & Agaral, R. Enhancedsecond-harmonic generation from metal-integrated semiconductor nanowiresvia highly confined whispering gallery modes. Nat. Commun. 5, 5432 (2014).

17. Camacho-Morales, R. et al. Nonlinear generation of vector beams fromAlGaAs nanoantennas. Nano Lett. 16, 7191–7197 (2016).

18. Sanatinia, R., Swillo, M. & Anand, S. Surface second harmonic generationfrom vertical GaP nanopillars. Nano Lett. 12, 820–826 (2012).

19. Timofeeva, M. et al. Polar second harmonic imaging to resolve pure andmixed crystal phases along GaAs nanowire. Nano Lett. 16, 6290–6297 (2016).

20. Cui, Y., Wei, Q. Q., Park, H. K. & Lieber, C. M. Nanowire nanosensors forhighly sensitive and selective detection of biological and chemical species.Science 293, 1289–1292 (2001).

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06035-1

8 NATURE COMMUNICATIONS | (2018) 9:3575 | DOI: 10.1038/s41467-018-06035-1 | www.nature.com/naturecommunications

21. Xie, P., Xiong, Q. H., Fang, Y., Qing, Q. & Lieber, C. M. Local electricalpotential detection of DNA by nanowire–nanopore sensors. Nat. Nanotech. 7,119–125 (2012).

22. Patolsky, F. et al. Detection, stimulation, and inhibition of neuronal signalswith high-density nanowire transistor arrays. Science 25, 1100–1104 (2006).

23. Gong, S. H., Ko, S. M., Jang, M. H. & Cho, Y. H. Giant rabi splitting ofwhispering gallery polaritons in GaN/InGaN core−shell wire. Nano Lett. 15,4517–4524 (2015).

24. Hua, B., Motohisa, J., Ding, Y., Hara, S. & Fukui, T. Characterization of Fabry-Pérot microcavity modes in GaAs nanowires fabricated by selective-area metalorganic vapor phase epitaxy. Appl. Phys. Lett. 91, 131112 (2007).

25. Malajovich, I., Berry, J. J., Samarth, N. & Awschalom, D. D. Persistentsourcing of coherent spins for multifunctional semiconductor spintronics.Nature 411, 770–772 (2001).

26. Kotissek, P. et al. Cross-sectional imaging of spin injection into asemiconductor. Nat. Phys. 3, 872–877 (2007).

27. Wang, X. J. et al. Room-temperature defect-engineered spin filter based on anon-magnetic semiconductor. Nat. Mater. 8, 198–202 (2009).

28. Kioseoglou, G. et al. Electrical spin injection from an n-type ferromagneticsemiconductor into a III–V device heterostructure. Nat. Mater. 3, 799–803(2004).

29. Zhu, H. J. et al. Room-temperature spin injection from Fe into GaAs. Phys.Rev. Lett. 87, 016601 (2001).

30. Rudolph, J., Hägele, D., Gibbs, H. M., Khitrova, G. & Oestreich, M. Laserthreshold reduction in a spintronic device. Appl. Phys. Lett. 82, 4516 (2003).

31. Rudolph, J., Döhrmann, S., Hägele, D., Oestreich, M. & Stolz, W. Room-temperature threshold reduction in vertical-cavity surface-emitting lasersby injection of spin-polarized electrons. Appl. Phys. Lett. 87, 241117(2005).

32. Hövel, S. et al. Optical spin manipulation of electrically pumped vertical-cavitysurface-emitting lasers. Appl. Phys. Lett. 92, 041118 (2008).

33. Warburton, R. J. Single spins in self-assembled quantum dots. Nat. Mater. 12,483–493 (2013).

34. Atature, M. et al. Quantum-dot spin-state preparation with near-unity fidelity.Science 312, 551–553 (2006).

35. Gerardot, B. D. et al. Optical pumping of a single hole spin in a quantum dot.Nature 451, 441–444 (2008).

36. Park, T. E. et al. Large spin accumulation and crystallographic dependence ofspin transport in single crystal gallium nitride nanowires. Nat. Commun. 8,15722 (2017).

37. Kum, H. et al. Room temperature single GaN nanowire spin valves with FeCo/MgO tunnel contacts. Appl. Phys. Lett. 100, 182407 (2012).

38. Heedt, S. et al. Electrical spin injection into InN semiconductor nanowires.Nano Lett. 12, 4437–4443 (2012).

39. Zhang, S. X. et al. Electrical spin injection and detection in Silicon nanowiresthrough oxide tunnel barriers. Nano Lett. 13, 430–435 (2013).

40. Chen, J. Y., Wong, T. M., Chang, C. W., Dong, C. Y. & Chen, Y. F. Self-polarized spin-nanolasers. Nat. Nanotech. 9, 845–850 (2014).

41. Shan, W. et al. Band anticrossing in GaInNAs alloys. Phys. Rev. Lett. 82,1221–1224 (1999).

42. Bellaiche, L., Wei, S. H. & Zunger, A. Localization and percolation insemiconductor alloys:GaAsN and GaAsP. Phys. Rev. B 54, 17568–17576 (1996).

43. Radulaski, M. et al. Scalable quantum photonics wit single color centers insilicon carbide. Nano Lett. 17, 1782–1786 (2017).

44. Naureen, S., Sanatinia, R., Shahid, N. & Anand, S. High optical quality InP-based nanopillars fabricated by a top-down approach. Nano Lett. 11,4805–4811 (2011).

45. Zhang, L. et al. How much better are InGaN/GaN nanodisks than quantumwells—Oscillator strength enhancement and changes in optical properties.Appl. Phys. Lett. 104, 051116 (2014).

46. Sun., Y. R. & Forrest, S. R. Enhanced light out-coupling of organic light-emittingdevices using embedded low-index grids. Nat. Photon. 2, 483–487 (2008).

47. Yang, X. Y. et al. Light extraction efficiency enhancement of colloidalquantum dot light-emitting diodes using large-scale nanopillar arrays. Adv.Funct. Mater. 24, 5977–5984 (2014).

48. Dyakonov, M. I. Spin physics in semiconductors. (Springer, Berlin,Heidelberg, 2008).

49. Uenoyama, T. & Sham, L. J. Hole relaxation and luminescence polarization indoped and undoped quantum wells. Phys. Rev. Lett. 64, 3070–3073 (1990).

50. Lagarde, D. et al. Electron spin dynamics in GaAsN and InGaAsN structures.Phys. Stat. Sol. (a) 204, 208–220 (2007).

51. Wang, X. J. et al. Dominant recombination centers in Ga(In)NAs alloys: Gainterstitials. Appl. Phys. Lett. 95, 241904 (2009).

52. Meier, F. & Zakharchenya, B. P. Optical orientation (Elsevier, North-Holland,1984).

53. Beyer, J., Buyanova, I. A., Suraprapapich, S., Tu, C. W. & Chen, W. M. Hanleeffect and electron spin polarization in InAs/GaAs quantum dots up to roomtemperature. Nanotechnology 23, 135705 (2012).

54. Altug, H., Englund, D. & Vučković, J. Ultrafast photonic crystal nanocavitylaser. Nat. Phys. 2, 484 (2006).

55. Englund, D. et al. Controlling the spontaneous emission rate of singlequantum dots in a two-dimensional photonic crystal. Phys. Rev. Lett. 95,013904 (2005).

56. Ren, Q. et al. Spin-resolved Purcell effect in a quantum dot microcavitysystem. Nano. Lett. 12, 3455 (2012).

57. Iba, S., Koh, S., Ikeda, K. & Kawaguchi, H. Room temperature circularlypolarized lasing in an optically spin injected vertical-cavity surface-emittinglaser with (110) GaAs quantum wells. Appl. Phys. Lett. 98, 081113 (2011).

58. Buyanova, I. A. & Chen, W. M. Physics and applications of dilute nitrides.(Taylor & Francis, New York, 2004).

59. Naureen, S., Shahid, N., Sanatinia, R. & Anand, S. Top-down fabrication ofhigh quality III–V nanostructures by monolayer controlled sculpting andsimultaneous passivation. Adv. Funct. Mater. 23, 1620–1627 (2013).

AcknowledgementsWe are grateful for C.W. Tu for providing the MQWs samples. This work was supportedby the Linköping University and KTH through the Professor Contracts, the SwedishResearch Council (Grant No. 2016–05091 and 349–2007–8664), the Swedish Govern-ment Strategic Research Area in Materials Science on Functional Materials at LinköpingUniversity (Faculty Grant SFO-Mat-LiU # 2009–00971), and the Swedish Energy Agency(Grant No. P40119–1).

Author contributionsS.C. and Y.H. carried out the experiments and analyzed the data under the supervision ofW.M.C. and I.A.B. D.V. and S.A. fabricated the DiP structures. S.C. and W.C. wrote themanuscript, with contributions from all co-authors.

Additional informationSupplementary Information accompanies this paper at https://doi.org/10.1038/s41467-018-06035-1.

Competing interests: The authors declare no competing interests.

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

Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims inpublished maps and institutional affiliations.

Open Access This article is licensed under a Creative CommonsAttribution 4.0 International License, which permits use, sharing,

adaptation, distribution and reproduction in any medium or format, as long as you giveappropriate credit to the original author(s) and the source, provide a link to the CreativeCommons license, and indicate if changes were made. The images or other third partymaterial in this article are included in the article’s Creative Commons license, unlessindicated otherwise in a credit line to the material. If material is not included in thearticle’s Creative Commons license and your intended use is not permitted by statutoryregulation or exceeds the permitted use, you will need to obtain permission directly fromthe copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

© The Author(s) 2018

NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06035-1 ARTICLE

NATURE COMMUNICATIONS | (2018) 9:3575 | DOI: 10.1038/s41467-018-06035-1 | www.nature.com/naturecommunications 9