Electrically Driven Single-Photon Source

See allHide authors and affiliations

Science  04 Jan 2002:
Vol. 295, Issue 5552, pp. 102-105
DOI: 10.1126/science.1066790


Electroluminescence from a single quantum dot within the intrinsic region of a p-i-n junction is shown to act as an electrically driven single-photon source. At low injection currents, the dot electroluminescence spectrum reveals a single sharp line due to exciton recombination, while another line due to the biexciton emerges at higher currents. The second-order correlation function of the diode displays anti-bunching under a continuous drive current. Single-photon emission is stimulated by subnanosecond voltage pulses. These results suggest that semiconductor technology can be used to mass-produce a single-photon source for applications in quantum information technology.

The emerging field of quantum information technology requires the development of a new type of light source, in which the photon number can be carefully controlled. Especially useful is an emitter of single photons at predetermined times. However, it is impossible to generate single photons with even very faint laser pulses, for which the photon number obeys Poissonian statistics. It has been shown that the unavoidable multiphoton pulses produced by a laser render quantum cryptography insecure from certain types of eavesdropping attack (1). Thus, unconditionally secure quantum cryptography, in addition to other applications in photonic quantum computing (2) and communications, requires the development of a true single-photon source. The generation of light with sub-Poissonian fluctuations in the photon number is also useful for performing optical measurements at lower noise levels than is possible with classical light. We report here the realization of an electrically driven source of single photons based on integrating quantum dots into a conventional semiconductor light-emitting diode structure.

Single photons can be generated through laser excitation and subsequent fluorescence of a single quantized system, such as an atom or an ion (3). Recently, this concept has been extended to several other quantum systems, such as single dye molecules (4–8), single quantum dots (9–13), and single nitrogen vacancy centers in diamond (14, 15). In each of these experiments or proposals, the emission was stimulated by optical excitation with an incident laser. However, from the viewpoint of practical application, one would prefer an electrically driven photon source. Early proposals for constructing an electrical single-photon source suggested using the Coulomb blockade effect to inject single electrons and holes into an etched double-barrier mesoscopic heterojunction (16). However, experimental studies on such a structure required operation at milli-kelvin temperatures, whereas the collected photon rate was too weak to allow the second-order correlation function to be verified (17). It was also theoretically proposed that a single quantum dot could be used, not just for emission of single photons but also for generation of entangled pairs of photons (18). Another proposal was based on injecting single electrons, confined within the moving potential wells defined by a surface acoustic wave and a split gate, into a p-type region (19).

We study the photo- and electroluminescence of a p-i-n diode containing a layer of InAs self-organized quantum dots in the intrinsic region. In order to avoid the nonradiative recombination associated with etched surfaces, relatively large mesas were prepared containing many quantum dots and their surface apertured, so as to collect the emission from just one dot. The electroluminescence spectra of the diode display sharp lines due to excitonic transitions in individual dots, which, as verified by photon pair correlation measurements, generate sequential emission of individual photons.

In our device structure (Fig. 1A), the semiconductor layers were grown by molecular beam epitaxy on a GaAs substrate and consist of a GaAs p-i-n diode with a layer of nominally nanometer-scale InAs self-organized quantum dots (20) inserted into the intrinsic region. Atomic force microscopy of dot layers grown under similar conditions (Fig. 1C) reveals that the dots have an areal density of ∼5 × 108cm−2. The structure was etched into mesas with lateral dimensions of 10 × 10 μm and ohmic contacts were formed to the n- and p-type regions (Fig. 1B). The emissive area was defined by means of an aperture in the opaque metal layers on the device surface. The p-i-n diodes were found to display nearly ideal current-voltage (I-V) characteristics, with the injected current increasingly rapidly around a forward bias of 1.5 V (Fig. 2C).

Figure 1

(A) Schematic of the single-photon–emitting diode in cross section. (B) Image of the device in an optical microscope. (C) Atomic force micrograph of a quantum dot layer grown under similar conditions to those in the device.

Figure 2

The electroluminescence spectra of the single-photon–emitting diode, showing sharp line emission characteristic of individual quantum dots. (A) Electroluminescence intensity from the device as a function of emission photon energy (vertical axis) and drive current (horizontal axis). Sharp line emission (marked X andX 2) is seen arising from a single quantum dot in the structure. (B) Measured (symbols) dependence of the intensities of the X andX 2 lines on the injection current. The solid lines show a linear fit to the low-current data. (C)I-V characteristic of the device shows ideal diode-like behavior.

Electroluminescence spectra recorded on the diode at 5 K with different biases applied between the n- and p-type contacts (Fig. 2A) show that at low injection currents, a sharp line is observed near 1.3942 eV, the intensity of which increases approximately linearly with I(Fig. 2B); a best fit to the data shows that the intensity strengthens as I 1.1. Based on this current dependence, we ascribe this line to recombination of the single exciton (X) of one electron and one hole within a quantum dot. At higher injection currents, the X line weakens and a second strong line (marked X 2) appears at 4.7 meV higher energy. This line, which strengthens with current asI 2.0, is ascribed to the biexciton transition of the dot. The strength of X drops for currents in excess of 5 μA because of competition from the biexciton state. On the other hand, the biexciton intensity is seen to saturate at the highest currents, which suggests that tri- and higher order excitons cannot be excited in these dots. This is presumably because, due to the small size of the dots studied here, they have only one electron- or hole-confined level. The maximum intensity of the biexciton line is stronger than that of the exciton, because of its shorter lifetime and the fact that at high injection currents, the average occupancy of the biexciton state can approach unity, whereas at the intermediate currents for which X dominates, there is always a statistical distribution in the occupancy of the empty, exciton, and biexciton states.

The same transitions, with similar photon energy and linewidth, can also be observed in the photoluminescence spectra recorded on the structure (21). The X line showed an approximately linear dependence on laser power, whereas a quadratic increase was seen for X 2, lending further credence to our assignment of the two peaks. Time-resolved photoluminescence measurements on the same dot determined the exciton and biexciton lifetimes to be 1.02 and 0.47 ns, respectively. We also observe the X decay to be delayed relative toX 2 at high laser power, as one would expect, because the biexciton photon is emitted before the exciton (13). Similar characteristics are seen in the time-resolved electroluminescence.

We measured the photon pair correlation statistics of the electroluminescence using a Hanbury-Brown and Twiss arrangement (Fig. 3D) of a 50/50 beamsplitter and two Silicon photon-counting avalanche photodiodes (22). A time-correlated single-photon counting card was used to repeatedly measure the time delays, τ, between photons registered by the two detectors, from which a histogram of the number of occurrences of different time delays was compiled. This histogram is proportional to the second-order correlation function g (2)(τ) in the low count rate limit, where the average time between detected photons is much longer than the measured delays τ. For an ideal single-photon emitter, we would expect to find no occurrences of zero time delay; that is, two photons should never be emitted simultaneously.

Figure 3

Experimental proof of photon anti-bunching and single-photon emission in the electroluminescence of the device. (A) Second-order correlation functiong (2)( τ) of the electroluminescence of the single exciton line measured for different injection currents of (i) 2.0, (ii) 2.5, and (iii) 4.0 μA, as well as the biexciton line for 6 μA (iv). For comparison, the correlation trace of the wetting layer (WL) electroluminescence (v) is also shown. The curves are offset vertically for clarity. The solid lines are identical to the broadened calculated curves in (B). (B) Calculated second-order correlation function (thin black lines) for electron-hole pair injection rates of (i) 0.45, (ii) 0.55, (iii) 1.31, and (iv) 2.00 pairs per nanosecond and after convolution with a Gaussian function to mimic the finite temporal resolution of the measurement system (thick red lines). (v) Second-order correlation function of an emitter with Poissonian statistics. (C) Correlation measured for pulsed electrical injection for (i) quantum dot exciton and (ii) wetting layer emission. (D) Schematic of the experimental arrangement for correlation measurements.

Figure 3A (i through iii) plots the correlation signal recorded for the single exciton electroluminescence with different injection currents. The dip in the correlation signal g(2)(τ) at zero time delay, τ = 0, which is the signature of a single-photon source, is clearly observed. This demonstrates the strong suppression of two photon emission events from the quantum dot, which occurs because, after photon emission, there is a finite delay before the dot is repopulated and can re-emit. In contrast, emission from the two-dimensional wetting layer on which the dots form [Fig. 3A (v)] displays a flat correlation trace, as expected for Poissonian statistics.

The dip in g (2)(τ) does not fall completely to zero. For the lowest injection current of 2 μA, we findg (2)(0) = 0.34 [Fig. 3A (i)]. However, this is largely due to the finite time resolution of the measurement system, which was measured to be 0.85 ns (full width at half maximum) rather than two photon emission events from the dot. We calculatedg (2)(τ) using a simple rate equation model to describe the occupancies of the biexciton, exciton, and empty states of the dot at different injection current densities (Fig. 3B, thin lines). The simulation also takes account of the dark counts of the single-photon detectors, as well as background electroluminescence, which add a flat contribution to g (2)(τ) that is always less than 0.14. The correlation dip sharpens as the current increases, due to an increase in the electron/hole capture rate. Figure 3B also displays the effect of convoluting the calculated curves with a Gaussian function, so as to simulate the effect of the finite temporal resolution of the measurement system. The agreement of the measured and calculated curves, for which there are no fitting parameters, demonstrates that the finite value of g (2)(0) derives mostly from the limited time resolution of the measurement system, whereas two-photon emission from the device contributes tog (2)(0) < 0.07 at an injection current of 2 μA. This represents more than an order of magnitude reduction in multiphoton emission events as compared to a classical light source displaying Poissonian statistics.

We also measured the second-order correlation function of the biexciton electroluminescence line of the dot at a diode current of 6 μA [Fig. 3A (iv)]. We find that the electroluminescence of theX 2 transition also displays photon anti-bunching, as reported recently for optical injection experiments (13, 22). After emission of a biexciton photon, the dot is occupied by a single exciton and must therefore capture another electron-hole pair before a second biexciton photon can be emitted. In this case, the zero delay dip is shallower,g (2)(0) = 0.75, than for the single exciton. However, as demonstrated by the calculated curve in Fig. 3A (iv), this is largely because the recovery ing (2)(τ) is faster for the biexciton, which has a shorter lifetime, than for the exciton, and thus is more difficult to resolve experimentally for the biexciton. These results show that at higher injection currents, the diode can generate pairs of photons, one at the exciton energy of the dot and the other at the biexciton.

The emission time of the single photons from the diode can be regulated through pulsing of the injection current. Pulsed electrical injection leads to pulsed emission from the dot, which is single-photon emission provided that the pulse width is much less than the exciton lifetime. We biased the diode with rectangular voltage pulses at a repetition rate of 80 MHz, superimposed on a dc bias of 1.50 V. This dc voltage is just below the “turn-on” of the diode and was chosen so as to result in relatively little electroluminescence. For pulse durations considerably less than the exciton lifetime, we observe that the shape of the time-resolved electroluminescence is independent of pulse width and is similar to that excited by the 1-ps-wide laser pulses.

The second-order correlation function recorded for theX line with a pulse height of 0.15 V and a width of 400 ps [Fig. 3C (i)] shows only a weak peak at zero time delay, indicating a strong suppression of the multiphoton emission pulses from the dot. This contrasts with the pair correlation recorded for the wetting layer electroluminescence [Fig. 3C (ii)], for which a second photon is just as likely to be found in the same emission pulse as the first, as in any other. The area of the zero delay peak of the dot correlation is about 11% of those at finite delays, again indicating an approximate order of magnitude decrease in multiphoton emission pulses as compared to a Poissonian source of the same average intensity. This rate of multiphoton emission could be further diminished by using shorter electrical pulses and reducing the backgound luminescence from the sample.

As for the prospects of realizing a single-photon–emitting diode for practical applications, the emission wavelength of the quantum dots chosen for study here is dictated by the spectral response of the Si single-photon detectors used in the experiments. However, InAs self-organized quantum dots can be tailored to emit at the wavelength used for long-distance fiber optic communications, around 1.3 μm (24). Furthermore, these longer wavelength quantum dots have much deeper carrier confinement potentials and thus emit efficiently at higher temperatures. In addition, the collection efficiency from the dot can be enhanced by means of an integrated cavity structure (25). Quantum dot light-emitting diodes may therefore provide an attractive source of either single photons or photon pairs for applications in quantum information technology and experiments in quantum optics.

  • * To whom correspondence should be addressed: E-mail: andrew.shields{at}


View Abstract

Navigate This Article