Electrically Driven Single-Cell Photonic Crystal Laser

See allHide authors and affiliations

Science  03 Sep 2004:
Vol. 305, Issue 5689, pp. 1444-1447
DOI: 10.1126/science.1100968


We report the experimental demonstration of an electrically driven, single-mode, low threshold current (∼260 μA) photonic band gap laser operating at room temperature. The electrical current pulse is injected through a sub-micrometer-sized semiconductor wire at the center of the mode with minimal degradation of the quality factor. The actual mode of interest operates in a nondegenerate monopole mode, as evidenced through the comparison of the measurement with the computation based on the actual fabricated structural parameters. As a small step toward a thresholdless laser or a single photon source, this wavelength-size photonic crystal laser may be of interest to photonic crystals, cavity quantum electrodynamics, and quantum information communities.

The laser physics and quantum optics communities have been interested for some time in extremely small, low-loss, low-power lasers (13). The potential to localize photons into photonic band gap semiconductor microcavities having wavelength-scale volumes and high quality factors enables us to study the cavity quantum electrodynamics in solids and to construct quantum optical devices such as ondemand single photon sources. Several optically pumped, ultra-small, photonic crystal lasers (48) or electrically driven light-emitting structures using the concept of photonic crystals (911) have been recently reported. Two kinds of electrically driven photonic band edge lasers, large-volume lasers with high output power (12) and quantum cascade lasers (13), especially draw our attention. However, the electronic activation of the wavelength-scale single-cell laser operating in a single mode, a crucial step toward a practical form of the thresholdless laser, has yet to be demonstrated.

One of the most daunting problems in trying to implement a single-cell, free-standing slab photonic crystal laser (47) is how to make electrical contact with the sub-micrometer-sized photonic crystal resonator structure (14). Locating the proper region inside the laser cavity to position an electrical contact requires an understanding of the resonant modes that are available in a single-cell triangular lattice photonic crystal cavity (15). Three potential candidates, each with a central node, were considered, because the introduction of a small central post as an electrical contact did not notably degrade the quality (Q) factor of the mode (15). The small central post functions as an electrical wire, a mode selector, and a heat sinker at the same time (16).

A sub-micrometer-sized semiconductor post is placed at the center of the single-cell photonic crystal resonator (Fig. 1A). The thickness of the semiconductor slab is 282.5 nm. Electrons are supplied laterally from the top electrode, whereas holes are injected directly through the bottom post. The carriers recombine in the six strain-compensated InGaAsP quantum wells that are designed to have an electroluminescence (EL) peak near communications wavelength of 1.5 μm. A doping structure that is inverted from that of a typical semiconductor laser is used to exploit the low mobility of the holes that are funneled through the sub-micrometer-sized post. The introduction of this heterojunction n-i-p structure also limits the occurrence of bimolecular recombination to the proximity of the central post and promotes an efficient overlap of the optical gain region with the mode profile. In addition, the modified single-cell photonic crystal cavity is surrounded by five heterogeneous photonic crystal lattices with the same lattice constant (a) but different air-hole sizes to improve the position and size of the central InP post (Fig. 2A). The peripheral dielectric material underneath the slab was added for mechanical support.

Fig. 1.

(A) Schematic diagram of current injection. The height of the central InP post is 1.0 μm. The post is diamond-shaped with 0.64a by 0.51a in diagonal directions. The diameter of etched mesa is 50 μm, and the inner radius of the AuGe n-electrode is 13 μm. Doping densities of top n layer and bottom p layer are ∼2.7 × 1019 cm–3 and ∼2.5 × 1018 cm–3, respectively. (B) Cross-sectional SEM image. From an intentionally broken sample, the region around the central post is clearly shown. Dusts around the post are remnants of the dielectric material, a photoresist, which are produced in the breaking process.

Fig. 2.

(A) Top view of fabricated sample. The lattice constant, a, is ∼510 nm, and the radii of the air holes in regions I, II, III, IV, and V are 0.28a, 0.35a, 0.385a, 0.4a, and 0.41a, respectively. (B) Monopole-mode image captured by an IR camera. The white bar represents 2-μm ruler. The white hexagon corresponds to region II-III interface in (A). (C) The vertical component of the Poynting vector obtained with the use of the structural data of (A) by 3D FDTD calculation. The calculation is performed at a vertical position of 3.0 μm above the slab, with consideration of a blurring effect by the objective lens. (D) The electric field intensity profile of the monopole mode calculated at the center of the slab (log scale).

The fabrication procedure contains two main steps: the mesa formation and the definition of photonic crystal patterns (17). The scanning electron microscope (SEM), cross-sectional view in Fig. 1B shows a fabricated sub-micrometer-sized InP post. It was discovered that the speed of wet etching depends on the radii of the air holes and that the position and size of the post can be improved by systematically modifying the size distribution of the air holes. This chirped photonic crystal resonator structure improves the Q factor slightly; however, the resonant frequency and the modal volume of the relevant modes remain almost unchanged. This was confirmed by a three-dimensional (3D) finite-difference time-domain (FDTD) calculation.

The fabricated single-cell photonic crystal cavities are electrically pulse-pumped at room temperature. The width and the period of the injected electrical current pulse are ∼6 ns and 2.5 μs, respectively (18). The emitted photons are collected by a 50× microscope objective lens with a numerical aperture of 0.42 and fed to a spectrometer. Single-mode lasing action was observed at a wavelength of 1519.7 nm (Fig. 3A). Above the threshold, a spectrometer-limited linewidth of 0.5 nm was measured from this nondegenerate lasing mode. The mode profile as captured by an infrared (IR) camera (Fig. 2B) exhibits a central intensity minimum and the characteristic features of a monopole mode (5, 15). Monopole mode operation was confirmed by comparing the measured resonant frequencies with those obtained from the 3D FDTD calculation. Numerical structural input data directly from the SEM image were used in the FDTD computation to truthfully compensate for any fabrication imperfections. In addition, no preferred direction of polarization is observed from the top, as expected from the monopole mode (5).

Fig. 3.

(A) Typical L-I curve of the monopole-mode laser. Threshold current is ∼260 μA, and output power indicates the peak value measured at the spectrometer. (Inset) The spectrum is taken at 700 μA. (B) Comparison of the measured L-I curves (red dots) with those obtained from the rate equations (black lines) for the monopole mode. Main parameters are as follows: internal efficiency ηi = 0.25, confinement factor Γ = 0.175, surface recombination velocity vs = 1.2 × 104 cm s–1, bimolecular radiative coefficient B = 1.6 × 10–10 cm3 s–1, Auger coefficient C = 5.0 × 10–29 cm6 s–1, transparent carrier density Ntr = 1.5 × 1018 cm–3, active volume Va = 1.72 × 10–13 cm3 and active surface area Aa = 1.47 × 10–8 cm2. (Inset) Typical electrical characteristics are shown. a.u., arbitrary units.

The calculated field profiles associated with the monopole mode are shown in Fig. 2, C and D. The measured near-field profile (Fig. 2B) represents the intensity of the propagating field in the proximity of the slab within the depth of focus of the 50× objective and compares well with the vertical component of the Poynting vector computed at a plane 3.0 μm above the laser cavity that has a small post (Fig. 2C). Even the asymmetry originating from the imperfect fabrication is reasonably reproduced in the FDTD computation with the use of numerical structural input data. As a reference, the energy distribution confined in the slab is shown in Fig. 2D.

Among several theoretically identified resonant modes in the lattice parameters of Fig. 2A, only the monopole mode was experimentally observed with a gain spectrum and resonant frequency in agreement with theoretic predictions. This is attributed to the fact that the other potential central node modes, such as quadrupole or hexapole modes (15), are located outside the spectral gain region. No adjustable parameter was used in our computation.

The measured Q factor of a cold cavity for the monopole mode, as estimated from the spectral line width associated with a transparent current of ∼225 μA (8), is ∼>2500 and compares well with the computed Q factor of ∼3480 obtained with a diamond-like post with 0.64a by 0.51a estimated from the SEM picture, where a is the lattice constant. The Q factor degrades rapidly when the post size becomes larger than the above value and improves slightly with smaller post size; however, the smaller post size leads to electrical resistance and thermal problems. Thus, it is important to optimize the post size considering both the optical and electrical characteristics.

The modal volume, V, of the monopole mode is found to be 5.87 × 10–2 μm3. This value corresponds to 0.684 (λ/n)3, where n is the refractive index of the slab material (3.4) and approaches the smallest theoretical value (4). The large estimated Purcell factor (389) of the current resonator implies the possibility of observing cavity quantum electrodynamic effects in an electrically driven, small, high Q cavity, photonic crystal laser (1).

A low threshold current of ∼260 μA was observed from the peak output intensity (Fig. 3A) and compares favorably with those estimated in the optical pumping experiment (58). Considering that there are nonnegligible current leakage paths in the structure, the actual threshold current may be even smaller.

The soft turn-on shoulder near the threshold (Fig. 3A) implies a large spontaneous emission factor (β). The β value can be estimated from simple laser rate equations, given the experimental values and typical parameters of InGaAsP quantum wells (15, 19). The critical parameter, such as the active surface area, is measured directly from the shape and size of the electroluminescent image obtained near the transparency. We found the β value is determined mostly by the shape of the light-current (L-I) curve below and near threshold, where slight variations of the parameters other than the surface recombination are tolerable. A β value of ∼0.25 can be calculated by comparing measured spectrally integrated output intensities with L-I curves obtained from the rate equation (Fig. 3B). This β value is considerably higher than previously reported from the semiconductor nanolasers (15, 19, 20) and attributed to the effective carrier localization by electrical pumping together with the nondegeneracy and the small modal volume (21, 22).

Typical electrical characteristics for a single-cell photonic crystal laser are shown (Fig. 3B, inset), where peak voltage and current values are used. The turn-on voltage is less than 1.0 V, and the electrical resistance is ∼2.2 kΩ. Noticeable current leakage is identified from the (I dV/dI) curve. The relatively high resistance is mainly attributed to the sub-micrometer size of the p-InP post and partly attributed to the lateral distance between the n electrode and the center. The current leakage is attributed to the nonradiative recombination at the air-semiconductor air hole interfaces and at the edge of the mesa (23).

There are various issues still remain to be addressed before this electrically driven, ultra-small cavity with a large Purcell factor becomes the practical on-demand single photon source. For example, one needs to find ways to place well-defined quantum dots or impurity atoms (24) at the antinode of the cavity and to inject single electron-hole pairs efficiently (25). The vertical coupling efficiency out of the cavity should be also improved by proper modifications of the cavity structure, e.g., the size of the air holes (26) and/or the reflectivity of the substrate. Alternatively, the photons localized in the cavity could be funneled horizontally into the neighboring low-loss photonic crystal waveguide (27) prepared by the quantum well intermixing (28). Together with all the challenging issues, the demonstration of electrically driven single-cell photonic crystal laser is believed to represent a small but meaningful step toward the ultimate photon source.

Supporting Online Material

Materials and Methods

Figs. S1 to S4

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article