LETTERS PUBLISHED ONLINE: 4 MAY 2015 | DOI: 10.1038/NNANO.2015.79
Voltage-controlled quantum light from an atomically thin semiconductor Chitraleema Chakraborty1, Laura Kinnischtzke2,3, Kenneth M. Goodfellow3,4, Ryan Beams5 and A. Nick Vamivakas1,3,4* Although semiconductor defects can often be detrimental to device performance, they are also responsible for the breadth of functionality exhibited by modern optoelectronic devices1. Artificially engineered defects (so-called quantum dots) or naturally occurring defects in solids are currently being investigated for applications ranging from quantum information science2,3 and optoelectronics4 to high-resolution metrology5. In parallel, the quantum confinement exhibited by atomically thin materials (semi-metals, semiconductors and insulators) has ushered in an era of flatland optoelectronics whose full potential is still being articulated6–18. In this Letter we demonstrate the possibility of leveraging the atomically thin semiconductor tungsten diselenide (WSe2) as a host for quantum dot-like defects. We report that this previously unexplored solid-state quantum emitter in WSe2 generates single photons with emission properties that can be controlled via the application of external d.c. electric and magnetic fields. These new optically active quantum dots exhibit excited-state lifetimes on the order of 1 ns and remarkably large excitonic g-factors of 10. It is anticipated that WSe2 quantum dots will provide a novel platform for integrated solid-state quantum photonics2,3 and quantum information processing19, as well as a rich condensed-matter physics playground with which to explore the coupling of quantum dots and atomically thin semiconductors. Groundbreaking work on the electronic and optical properties of monolayer graphene5–8 flakes has recently stimulated the search for other atomically thin material systems with appealing physical properties. Of particular interest are monolayer materials that exhibit semiconducting and insulating behaviour. Remarkably, it has been discovered that a large class of transition-metal dichalcogenides, when reduced to a single layer, exhibit emergent photoluminescence as a result of quantum well-like quantum confinement19–21. The two-dimensional nature of the materials, the availability of semiconductors, metals and insulators, and the possibility to assemble van der Waals heterostructures has resulted in numerous fundamental physics investigations as well as the demonstration of advanced optoelectronic and nanophotonic devices9–15. In addition to the conventional physical characteristics of these materials, the electronic excitations also exhibit a robust valley degree of freedom that coherently interfaces with photon polarization and can potentially provide the physical basis of next-generation information processing applications16–18. A natural question that arises when studying any material is the role that defects play in mediating the observed properties1. Although the label ‘defect’ sounds pejorative, it is no stretch of
the truth to say that the controlled introduction of defects is a driving force behind the myriad functionalities available with modern optoelectronic and nanophotonic devices. From the perspective of quantum technology, the potential for defects to localize excitons can enable novel sources of quantum light as well as provide a physical implementation of a stationary qubit. Along these lines, the last decade has witnessed intense quantum optical investigations of optically active defects in both GaAs and nitrogen vacancies (NVs) in diamond, as well as artificial defects such as InAs quantum dots2,3. In this Letter the quantum optical properties of light emission from quantum dots in electrically gated WSe2 are investigated22,23. The inset of Fig. 1a shows an optical microscope image of our gated device (Supplementary Fig. 1), in which the dashed box identifies the single-layer flake. The sample was prepared by mechanical exfoliation of single layers from a WSe2 bulk crystal onto a polydimethylsiloxane (PDMS) viscoelastic stamp (see Methods). Photoluminescence was used to identify the number of WSe2 layers (see Supplementary Fig. 2 for photoluminescence data). A dry transfer technique was applied to position the desired flake onto a doped silicon substrate, with 270 nm of SiO2 as a dielectric layer, into which electrode arrays were patterned for electrical gating (see Methods). The device was loaded into a magnetooptical cryostat with x–y–z positioners for sample positioning and the temperature was maintained at ∼4 K. The microscope objective had a numerical aperture of 0.8 and enabled diffractionlimited excitation and collection. Light emitted from the material following optical excitation at 532 nm was directed to a spectrometer that sent the dispersed light to a charge-coupled device (CCD) array for spectral measurements or through an exit slit to filter the emission before measuring the second-order intensity autocorrelation. An exemplary spectrum of a WSe2 monolayer, taken near the centre of the flake under 0 V gate bias, is presented in Fig. 1a. A characteristic emission spectrum composed of excitons and trions is observed22,23. This was confirmed by gate voltage-dependent photoluminescence measurements (Supplementary Fig. 3). More interesting results were obtained when the sample was positioned at an interface between the WSe2 monolayer and the WSe2 multilayer, where the spectrum in Fig. 1b was recorded. In stark contrast to Fig. 1a, the spectrum now consists of a number of narrow emission lines that are bright when compared to the extended twodimensional WSe2 excitons, even at an excitation power ∼20 times less than the excitation power used for the spectrum in Fig. 1a. For clarity, in the remainder of this Letter we will refer to these bright lines as quantum dots and adopt the labelling QD1
1
Materials Science, University of Rochester, Rochester, New York 14627, USA. 2 Department of Physics, University of Rochester, Rochester, New York 14627, USA. 3 Center for Coherence and Quantum Optics, University of Rochester, Rochester, New York 14627, USA. 4 Institute of Optics, University of Rochester, Rochester, New York 14627, USA. 5 Material Measurement Lab, National Institute of Standards and Technology Gaithersburg, Maryland 20899, USA. * e-mail: [email protected] NATURE NANOTECHNOLOGY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturenanotechnology
© 2015 Macmillan Publishers Limited. All rights reserved
1
LETTERS
NATURE NANOTECHNOLOGY b
QD2
QD1
Au pad
900 Intensity (c.p.s.)
Intensity (×103 counts)
a 30
DOI: 10.1038/NNANO.2015.79
20
10
X0
600
1.700 1.701 1.702 Energy (eV)
300
0
QD2
0 1.60
1.65
1.70
1.75
1.60
1.80
1.65
1.70
1.75
1.80
Energy (eV)
Energy (eV)
Figure 1 | Device and emission spectra. a, Fluorescence spectra from the single-layer region of the WSe2 flake at 4.2 K, integrated for 10 s. The peak at 1.75 eV is from neutral excitons in the WSe2 flake. Inset: Optical microscope image of the WSe2 device. The white dashed line outlines the single-layer region of the flake. b, Narrow emission lines identified as QD1 (1.65 eV) and QD2 (1.70 eV) in the photoluminescence spectrum around the interface of the single- and multilayer regions in the flake.
a
2.5
b 1.60 Counts (a.u.)
Energy (eV)
2.0 1.65 1.70
1.5 0 μm 4 μm
1.0 0.5
1.75
1.668 eV
1.716 eV
0.0 0
1
2
3
0
4
1
Position (μm)
30 T (K)
In(1 − (IPL/I4.2))
−0.2 −0.4
2
1
20
0.1
1.0
10 1.64 1.67 1.70 Energy (eV)
−0.6 −0.8
3
EA = 0.96 meV EB = 0.30 meV
−1.0
d Normalized coincidence counts
c
2
3
4
Position (μm)
1.00 0.75 Data Fit
0.50 0.25 0.00
0.05
0.10
0.15
−20
−10
1/T (K−1)
0
10
20
Time delay (ns)
Figure 2 | Quantum dot light emission. a, One-dimensional hyperspectral image of WSe2 quantum dot emission exhibiting spatial localization. Inset: Zoom-in of the black dashed box in the main image. b, One-dimensional line cuts of two different quantum dots. Inset: Optical microscope image of the flake. The start and finish locations of the line cut are identified by two arrows. Total line cut, 4 μm. c, Arrhenius plot for QD2, showing two different slopes. Inset: Temperature (T) dependence of multiple quantum dots. The bottom of the panel corresponds to 4.2 K and the top to 34 K. The colour bar is normalized such that the maximum number of photocounts is set to 1. d, Intensity autocorrelation measured on QD3. The horizontal dashed line corresponds to the zero delay value of 0.36. The zero delay value is corrected according to d.
(1.65 eV) and QD2 (1.70 eV). The inset of Fig. 1b shows the emission spectrum for QD2. Data from a third quantum dot, QD3 (emission energy 1.68 eV), are also presented in Fig. 3. We note that spectral wandering was observed on many of the emitters (see Supplementary Fig. 4 for a discussion). To explore the origin of these narrow emission lines, a one-dimensional hyperspectral image was acquired (Fig. 2a). One-dimensional line cuts for two distinct spectral lines (1.668 eV and 1.716 eV) are shown in Fig. 2b. The inset in Fig. 2b identifies the path for the hyperpectral line cuts. The start and stop positions are indicated by arrows. Interestingly, we observe these features along the interface of the single and multilayer flakes (data shown in Fig. 2a) and along the edges of flakes. The diffraction-limited resolution observed in the spatial line cut is typical of the spatial 2
localization observed from many of the narrow emission features investigated and is the first piece of evidence that these lines are defect-related. Another clue to the nature of these features is their temperature stability. In the inset of Fig. 2c, temperature-dependent photoluminescence is recorded for a number of quantum dots. As expected for defects24, above a certain temperature, in this case ∼35 K, light emission is nearly suppressed. Figure 2c presents the line cut for QD2 fit to the Arrhenius equation IPL = I4.2 (1 − e−Ea /kT ), where I4.2 is the low-temperature photoluminescence strength and Ea is the defect ionization energy. As is apparent in the data, there is a change in slope at ∼18 K. From the fit below 18 K we estimate a thermal ionization energy of 0.3 meV and from the fit above 18 K, 0.96 meV. We further find that the photoluminescence is restored when the temperature is lowered back to 4 K, and is robust
NATURE NANOTECHNOLOGY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturenanotechnology
© 2015 Macmillan Publishers Limited. All rights reserved
NATURE NANOTECHNOLOGY
Intensity (×103 cps) 4 8 12
a
LETTERS
DOI: 10.1038/NNANO.2015.79
c
b
60
Gate voltage (V)
40
X
Y
20 0 −20
X
Y
−40 −60 1.67
1.70
1.73
Normalized counts
1.76
Normalized counts
Energy (eV)
e 50 V
f 1.0
Counts (a.u.)
Counts (a.u.)
1.0
0.5
0V
1.0
1.0 0.5 Y 0.0 −20
0.5
0 Time (ns)
20
0.0
0.0 1.67
1.70
1.73
Counts (a.u.)
d
−50 V
1.0 0.5 X 0.0 −20
0.5
0 Time (ns)
20
0.0 1.67
Energy (eV)
1.70
1.73
Energy (eV)
1.67
1.70
1.73
Energy (eV)
Figure 3 | Voltage-controlled quantum light generation. a, WSe2 quantum dot emission spectra as a function of applied gate voltage (X = QD3 and Y = QD2). b,c, Voltage-dependent photoluminescence for QD3 and QD2. A bandwidth of 5 meV is used to generate each data set. d–f, Recorded spectra for QD3 and QD2 at the gate voltage values indicated in the panel. Insets: Intensity autocorrelation function for indicated gate voltages.
against temperature cycling (see Supplementary Fig. 5 for additional temperature-dependent data). After investigating the spatial and thermal character of the localized emitters, the photophysical properties were further examined. As is well known, generic two-level systems exhibit saturation as their excitation rate increases. Such behaviour has been observed in atomic systems, semiconductor quantum dots and defects in solids. The defect-related light emission in WSe2 is no different, and we observe a saturation of photoluminescence emission at 50 nW from power-dependent spectra (Supplementary Fig. 6). A natural consequence of these previous observations is that the localized emitter is, in fact, emitting a stream of non-classical light, in this case single photons. Figure 2d presents the second-order intensity autocorrelation function measured on the defect line QD3. From the pronounced anti-bunching dip, the excited-state lifetime is determined to be 1.8 ns. We observed single photon statistics on approximately half of the 20 localized emission sites measured, with lifetimes in the range 0.5–1.8 ns. Compared to extended WSe2 excitons, which have lifetimes on the order of ∼30 ps (ref. 25), the marked increase in excited-state lifetime is characteristic of three-dimensional quantum confinement26. We next investigated whether external d.c. electric and magnetic fields can control the spectral and/or emission properties of the WSe2 quantum dots. WSe2 is well known to support gate-voltagecontrolled exciton emission22 (Supplementary Fig. 3). Figure 3a presents the gate-voltage-dependent photoluminescence from two spectrally distinct quantum dots (QD2 and QD3). It is evident that the local electrical environment determines the activity of the observed emission. Figure 3b,c present the spectrally integrated voltage-dependent emission profiles for the lines labelled X and Y in Fig. 3a (this mitigates the impact of spectral wandering on the photoluminescence). Feature X exhibits a reproducible voltagedependent maximum in its photoluminescence. It is likely that the
defect stability is modified by changes to the local potential energy landscape induced by the applied voltage. Line cuts at specific gate voltage values from the data in Fig. 3a are presented in Fig. 3d–f. Each of these voltage-controlled emitters is a quantum dot, as demonstrated by the intensity autocorrelation measurements in the insets of Fig. 3e,f. The lifetimes of these two transitions differ by a factor of 4. Whether the decay time of an isolated WSe2 quantum dot is voltagecontrollable will be the subject of a future study. Finally, examination of many quantum dot emission spectra revealed a characteristic doublet structure, as shown in the bottom spectra at 0 T in Fig. 4a (QD2). The doublet is highlighted by grey shading. The doublets exhibit correlated spectral wandering, suggesting they originate from the same localized state. To test if this doublet is related to the spin structure of emission, magneto-optical spectroscopy was performed in the Faraday configuration (magnetic field applied perpendicular to the sample plane and parallel to the optical axis). Figure 4a shows the emission spectra recorded for increasing magnetic fields. The doublet separation increases with field, as expected for spin-correlated optical transitions. In Fig. 4b, the peak spectral location is plotted as a function of field. The spectral wandering in the emission appears as a correlated jitter in the absolute emission energy of the doublet. This further confirms that the emission does indeed originate from the same quantum emitter. The relative separation of the magnetic field split transitions is plotted in Fig. 4c. The data are fit with the equation Δ = (( gμB B)2 + Δ2o ) where g is the exciton g-factor, μB is the Bohr magneton, and Δo is the zero-field splitting27. This equation is expected for electron– hole exchange-dominated magnetic field-dependent optical transitions in quantum dots of symmetry
Voltage-controlled quantum light from an atomically thin semiconductor Chitraleema Chakraborty1, Laura Kinnischtzke2,3, Kenneth M. Goodfellow3,4, Ryan Beams5 and A. Nick Vamivakas1,3,4* Although semiconductor defects can often be detrimental to device performance, they are also responsible for the breadth of functionality exhibited by modern optoelectronic devices1. Artificially engineered defects (so-called quantum dots) or naturally occurring defects in solids are currently being investigated for applications ranging from quantum information science2,3 and optoelectronics4 to high-resolution metrology5. In parallel, the quantum confinement exhibited by atomically thin materials (semi-metals, semiconductors and insulators) has ushered in an era of flatland optoelectronics whose full potential is still being articulated6–18. In this Letter we demonstrate the possibility of leveraging the atomically thin semiconductor tungsten diselenide (WSe2) as a host for quantum dot-like defects. We report that this previously unexplored solid-state quantum emitter in WSe2 generates single photons with emission properties that can be controlled via the application of external d.c. electric and magnetic fields. These new optically active quantum dots exhibit excited-state lifetimes on the order of 1 ns and remarkably large excitonic g-factors of 10. It is anticipated that WSe2 quantum dots will provide a novel platform for integrated solid-state quantum photonics2,3 and quantum information processing19, as well as a rich condensed-matter physics playground with which to explore the coupling of quantum dots and atomically thin semiconductors. Groundbreaking work on the electronic and optical properties of monolayer graphene5–8 flakes has recently stimulated the search for other atomically thin material systems with appealing physical properties. Of particular interest are monolayer materials that exhibit semiconducting and insulating behaviour. Remarkably, it has been discovered that a large class of transition-metal dichalcogenides, when reduced to a single layer, exhibit emergent photoluminescence as a result of quantum well-like quantum confinement19–21. The two-dimensional nature of the materials, the availability of semiconductors, metals and insulators, and the possibility to assemble van der Waals heterostructures has resulted in numerous fundamental physics investigations as well as the demonstration of advanced optoelectronic and nanophotonic devices9–15. In addition to the conventional physical characteristics of these materials, the electronic excitations also exhibit a robust valley degree of freedom that coherently interfaces with photon polarization and can potentially provide the physical basis of next-generation information processing applications16–18. A natural question that arises when studying any material is the role that defects play in mediating the observed properties1. Although the label ‘defect’ sounds pejorative, it is no stretch of
the truth to say that the controlled introduction of defects is a driving force behind the myriad functionalities available with modern optoelectronic and nanophotonic devices. From the perspective of quantum technology, the potential for defects to localize excitons can enable novel sources of quantum light as well as provide a physical implementation of a stationary qubit. Along these lines, the last decade has witnessed intense quantum optical investigations of optically active defects in both GaAs and nitrogen vacancies (NVs) in diamond, as well as artificial defects such as InAs quantum dots2,3. In this Letter the quantum optical properties of light emission from quantum dots in electrically gated WSe2 are investigated22,23. The inset of Fig. 1a shows an optical microscope image of our gated device (Supplementary Fig. 1), in which the dashed box identifies the single-layer flake. The sample was prepared by mechanical exfoliation of single layers from a WSe2 bulk crystal onto a polydimethylsiloxane (PDMS) viscoelastic stamp (see Methods). Photoluminescence was used to identify the number of WSe2 layers (see Supplementary Fig. 2 for photoluminescence data). A dry transfer technique was applied to position the desired flake onto a doped silicon substrate, with 270 nm of SiO2 as a dielectric layer, into which electrode arrays were patterned for electrical gating (see Methods). The device was loaded into a magnetooptical cryostat with x–y–z positioners for sample positioning and the temperature was maintained at ∼4 K. The microscope objective had a numerical aperture of 0.8 and enabled diffractionlimited excitation and collection. Light emitted from the material following optical excitation at 532 nm was directed to a spectrometer that sent the dispersed light to a charge-coupled device (CCD) array for spectral measurements or through an exit slit to filter the emission before measuring the second-order intensity autocorrelation. An exemplary spectrum of a WSe2 monolayer, taken near the centre of the flake under 0 V gate bias, is presented in Fig. 1a. A characteristic emission spectrum composed of excitons and trions is observed22,23. This was confirmed by gate voltage-dependent photoluminescence measurements (Supplementary Fig. 3). More interesting results were obtained when the sample was positioned at an interface between the WSe2 monolayer and the WSe2 multilayer, where the spectrum in Fig. 1b was recorded. In stark contrast to Fig. 1a, the spectrum now consists of a number of narrow emission lines that are bright when compared to the extended twodimensional WSe2 excitons, even at an excitation power ∼20 times less than the excitation power used for the spectrum in Fig. 1a. For clarity, in the remainder of this Letter we will refer to these bright lines as quantum dots and adopt the labelling QD1
1
Materials Science, University of Rochester, Rochester, New York 14627, USA. 2 Department of Physics, University of Rochester, Rochester, New York 14627, USA. 3 Center for Coherence and Quantum Optics, University of Rochester, Rochester, New York 14627, USA. 4 Institute of Optics, University of Rochester, Rochester, New York 14627, USA. 5 Material Measurement Lab, National Institute of Standards and Technology Gaithersburg, Maryland 20899, USA. * e-mail: [email protected] NATURE NANOTECHNOLOGY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturenanotechnology
© 2015 Macmillan Publishers Limited. All rights reserved
1
LETTERS
NATURE NANOTECHNOLOGY b
QD2
QD1
Au pad
900 Intensity (c.p.s.)
Intensity (×103 counts)
a 30
DOI: 10.1038/NNANO.2015.79
20
10
X0
600
1.700 1.701 1.702 Energy (eV)
300
0
QD2
0 1.60
1.65
1.70
1.75
1.60
1.80
1.65
1.70
1.75
1.80
Energy (eV)
Energy (eV)
Figure 1 | Device and emission spectra. a, Fluorescence spectra from the single-layer region of the WSe2 flake at 4.2 K, integrated for 10 s. The peak at 1.75 eV is from neutral excitons in the WSe2 flake. Inset: Optical microscope image of the WSe2 device. The white dashed line outlines the single-layer region of the flake. b, Narrow emission lines identified as QD1 (1.65 eV) and QD2 (1.70 eV) in the photoluminescence spectrum around the interface of the single- and multilayer regions in the flake.
a
2.5
b 1.60 Counts (a.u.)
Energy (eV)
2.0 1.65 1.70
1.5 0 μm 4 μm
1.0 0.5
1.75
1.668 eV
1.716 eV
0.0 0
1
2
3
0
4
1
Position (μm)
30 T (K)
In(1 − (IPL/I4.2))
−0.2 −0.4
2
1
20
0.1
1.0
10 1.64 1.67 1.70 Energy (eV)
−0.6 −0.8
3
EA = 0.96 meV EB = 0.30 meV
−1.0
d Normalized coincidence counts
c
2
3
4
Position (μm)
1.00 0.75 Data Fit
0.50 0.25 0.00
0.05
0.10
0.15
−20
−10
1/T (K−1)
0
10
20
Time delay (ns)
Figure 2 | Quantum dot light emission. a, One-dimensional hyperspectral image of WSe2 quantum dot emission exhibiting spatial localization. Inset: Zoom-in of the black dashed box in the main image. b, One-dimensional line cuts of two different quantum dots. Inset: Optical microscope image of the flake. The start and finish locations of the line cut are identified by two arrows. Total line cut, 4 μm. c, Arrhenius plot for QD2, showing two different slopes. Inset: Temperature (T) dependence of multiple quantum dots. The bottom of the panel corresponds to 4.2 K and the top to 34 K. The colour bar is normalized such that the maximum number of photocounts is set to 1. d, Intensity autocorrelation measured on QD3. The horizontal dashed line corresponds to the zero delay value of 0.36. The zero delay value is corrected according to d.
(1.65 eV) and QD2 (1.70 eV). The inset of Fig. 1b shows the emission spectrum for QD2. Data from a third quantum dot, QD3 (emission energy 1.68 eV), are also presented in Fig. 3. We note that spectral wandering was observed on many of the emitters (see Supplementary Fig. 4 for a discussion). To explore the origin of these narrow emission lines, a one-dimensional hyperspectral image was acquired (Fig. 2a). One-dimensional line cuts for two distinct spectral lines (1.668 eV and 1.716 eV) are shown in Fig. 2b. The inset in Fig. 2b identifies the path for the hyperpectral line cuts. The start and stop positions are indicated by arrows. Interestingly, we observe these features along the interface of the single and multilayer flakes (data shown in Fig. 2a) and along the edges of flakes. The diffraction-limited resolution observed in the spatial line cut is typical of the spatial 2
localization observed from many of the narrow emission features investigated and is the first piece of evidence that these lines are defect-related. Another clue to the nature of these features is their temperature stability. In the inset of Fig. 2c, temperature-dependent photoluminescence is recorded for a number of quantum dots. As expected for defects24, above a certain temperature, in this case ∼35 K, light emission is nearly suppressed. Figure 2c presents the line cut for QD2 fit to the Arrhenius equation IPL = I4.2 (1 − e−Ea /kT ), where I4.2 is the low-temperature photoluminescence strength and Ea is the defect ionization energy. As is apparent in the data, there is a change in slope at ∼18 K. From the fit below 18 K we estimate a thermal ionization energy of 0.3 meV and from the fit above 18 K, 0.96 meV. We further find that the photoluminescence is restored when the temperature is lowered back to 4 K, and is robust
NATURE NANOTECHNOLOGY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturenanotechnology
© 2015 Macmillan Publishers Limited. All rights reserved
NATURE NANOTECHNOLOGY
Intensity (×103 cps) 4 8 12
a
LETTERS
DOI: 10.1038/NNANO.2015.79
c
b
60
Gate voltage (V)
40
X
Y
20 0 −20
X
Y
−40 −60 1.67
1.70
1.73
Normalized counts
1.76
Normalized counts
Energy (eV)
e 50 V
f 1.0
Counts (a.u.)
Counts (a.u.)
1.0
0.5
0V
1.0
1.0 0.5 Y 0.0 −20
0.5
0 Time (ns)
20
0.0
0.0 1.67
1.70
1.73
Counts (a.u.)
d
−50 V
1.0 0.5 X 0.0 −20
0.5
0 Time (ns)
20
0.0 1.67
Energy (eV)
1.70
1.73
Energy (eV)
1.67
1.70
1.73
Energy (eV)
Figure 3 | Voltage-controlled quantum light generation. a, WSe2 quantum dot emission spectra as a function of applied gate voltage (X = QD3 and Y = QD2). b,c, Voltage-dependent photoluminescence for QD3 and QD2. A bandwidth of 5 meV is used to generate each data set. d–f, Recorded spectra for QD3 and QD2 at the gate voltage values indicated in the panel. Insets: Intensity autocorrelation function for indicated gate voltages.
against temperature cycling (see Supplementary Fig. 5 for additional temperature-dependent data). After investigating the spatial and thermal character of the localized emitters, the photophysical properties were further examined. As is well known, generic two-level systems exhibit saturation as their excitation rate increases. Such behaviour has been observed in atomic systems, semiconductor quantum dots and defects in solids. The defect-related light emission in WSe2 is no different, and we observe a saturation of photoluminescence emission at 50 nW from power-dependent spectra (Supplementary Fig. 6). A natural consequence of these previous observations is that the localized emitter is, in fact, emitting a stream of non-classical light, in this case single photons. Figure 2d presents the second-order intensity autocorrelation function measured on the defect line QD3. From the pronounced anti-bunching dip, the excited-state lifetime is determined to be 1.8 ns. We observed single photon statistics on approximately half of the 20 localized emission sites measured, with lifetimes in the range 0.5–1.8 ns. Compared to extended WSe2 excitons, which have lifetimes on the order of ∼30 ps (ref. 25), the marked increase in excited-state lifetime is characteristic of three-dimensional quantum confinement26. We next investigated whether external d.c. electric and magnetic fields can control the spectral and/or emission properties of the WSe2 quantum dots. WSe2 is well known to support gate-voltagecontrolled exciton emission22 (Supplementary Fig. 3). Figure 3a presents the gate-voltage-dependent photoluminescence from two spectrally distinct quantum dots (QD2 and QD3). It is evident that the local electrical environment determines the activity of the observed emission. Figure 3b,c present the spectrally integrated voltage-dependent emission profiles for the lines labelled X and Y in Fig. 3a (this mitigates the impact of spectral wandering on the photoluminescence). Feature X exhibits a reproducible voltagedependent maximum in its photoluminescence. It is likely that the
defect stability is modified by changes to the local potential energy landscape induced by the applied voltage. Line cuts at specific gate voltage values from the data in Fig. 3a are presented in Fig. 3d–f. Each of these voltage-controlled emitters is a quantum dot, as demonstrated by the intensity autocorrelation measurements in the insets of Fig. 3e,f. The lifetimes of these two transitions differ by a factor of 4. Whether the decay time of an isolated WSe2 quantum dot is voltagecontrollable will be the subject of a future study. Finally, examination of many quantum dot emission spectra revealed a characteristic doublet structure, as shown in the bottom spectra at 0 T in Fig. 4a (QD2). The doublet is highlighted by grey shading. The doublets exhibit correlated spectral wandering, suggesting they originate from the same localized state. To test if this doublet is related to the spin structure of emission, magneto-optical spectroscopy was performed in the Faraday configuration (magnetic field applied perpendicular to the sample plane and parallel to the optical axis). Figure 4a shows the emission spectra recorded for increasing magnetic fields. The doublet separation increases with field, as expected for spin-correlated optical transitions. In Fig. 4b, the peak spectral location is plotted as a function of field. The spectral wandering in the emission appears as a correlated jitter in the absolute emission energy of the doublet. This further confirms that the emission does indeed originate from the same quantum emitter. The relative separation of the magnetic field split transitions is plotted in Fig. 4c. The data are fit with the equation Δ = (( gμB B)2 + Δ2o ) where g is the exciton g-factor, μB is the Bohr magneton, and Δo is the zero-field splitting27. This equation is expected for electron– hole exchange-dominated magnetic field-dependent optical transitions in quantum dots of symmetry
