Laser action in nanowires: Observation of the transition from amplified spontaneous emission to laser oscillation

Direct evidence of the transition from ampliﬁed spontaneous emission to laser action in optically pumped zinc oxide (cid:1) ZnO (cid:2) nanowires, at room temperature, is presented. The optical power evolves from a superlinear to a linear regime as the pump power exceeds threshold, concomitant with a transition to directional emission along the nanowire and the emergence of well deﬁned cavity Fabry–Pérot modes around a wavelength of (cid:3) 385 nm, the intensity of which exceeds the spontaneous emission background by orders of magnitude. The laser oscillation threshold is found to be strongly dependent on nanowire diameter, with no laser oscillation observed for diameters smaller than (cid:4) 150 nm. Finally, we use an alternative “head on” detection geometry to measure the output power of a single nanowire laser. © 2008 American Institute of Physics . (cid:5) DOI: 10.1063/1.2965797 (cid:6)

Semiconductor nanowires have recently been proposed as the next frontier in the miniaturization of semiconductor lasers.While previous reports provided a clear evidence of amplified spontaneous emission ͑ASE͒ in zinc oxide ͑ZnO͒, [1][2][3][4][5] gallium nitride ͑GaN͒, 6 and cadmium sulfide ͑CdS͒ 7,8 semiconductor nanowires and some evidence of lasing in gallium antimonide ͑GaSb͒, 9 the conclusive demonstration of laser oscillation has remained elusive.In particular, no laser threshold marking the transition between a superlinear region, characteristic of ASE, and a linear region, characteristic of laser oscillation, 10 with increasing pump power or drive current, has yet been reported.In addition, the highly directional emission characteristic of lasers could not be observed since only light scattered from the nanowire ends was measured.
Laser oscillation occurs when the round-trip gain matches the round-trip losses; 10 in nanowires the latter are typically dominated by the mirror losses ͑i.e., transmission at the end facets͒.The reason for this is that in nanowires with diameters comparable to or smaller than the wavelength of light in the semiconductor diffraction causes the optical field to "spill out" of the nanowire.This contributes to the losses since only the fraction of the optical field inside the nanowire experiences a refractive index contrast at the end facets, thus reducing the amount of reflection.Here we present unambiguous evidence of optically pumped laser oscillation at ultraviolet frequencies in individual ZnO nanowires at room temperature.By means of systematic measurements on nanowires of different dimensions, we demonstrate the existence of a critical diameter below which no lasing occurs.Also, using an alternative "head on" detection geometry, we measure the output power of a single nanowire laser.
Single-crystal ZnO nanowires were synthesized by a simple vapor transport technique 11 and subsequently dispersed onto a Si/ SiO 2 substrate ͑500 nm thermal oxide͒.Individual nanowires were optically excited with the frequency-tripled output ͑355 nm͒ of a Nd:YAG ͑yttrium aluminum garnet͒ laser ͑500 Hz, 6 ns pulse length͒ focused to a diameter of ϳ100 m.The emission from individual nanowires was collected by a reflective microscope objective and coupled to a 55 cm spectrometer and a charge coupled device ͑CCD͒ camera.All measurements were performed at room temperature.
Figures 1͑a͒ and 1͑b͒ show the optical spectra and the corresponding CCD images, respectively, for a ZnO nanowire as a function of pump intensity I ex .For I ex Շ 200 kW cm −2 , the spectra are broad and featureless, centered around ϳ382 nm and with a full width at half maximum ͑FWHM͒ of 19.3 nm.In this regime, light is emitted essentially isotropically along the nanowire ͓Fig.1͑b͔͒, and the output power ͑Fig.2͒ depends linearly on the excitation intensity, consistent with spontaneous emission.For a͒ Electronic mail: capasso@seas.harvard.edu.200 kW cm −2 Շ I ex Շ 300 kW cm −2 , the spectra consist of a broad emission with the addition of sharp ͑FWHM Ͻ 0.4 nm͒ emission lines.In this regime, population inversion starts building up, leading to ASE along the nanowire at wavelengths corresponding to the longitudinal modes and thus to the enhanced emission from the nanowire ends ͓Fig.1͑b͔͒.Furthermore, the output power exhibits a superlinear increase with pump intensity ͓Fig.2͔, which is the expected behavior as the laser threshold is approached. 10,12For I ex տ 300 kW cm −2 , the spectra are dominated by sharp emission lines; their intensity is orders of magnitude greater than the spontaneous emission background.The output power depends linearly on excitation intensity ͓Fig.2͔ and is concentrated in a narrow emission range ͑385 nmϽϽ390 nm͒.The inset in Fig. 2 shows the output power on a log-log scale as well as a fit ͑solid line͒ with a multimode laser model, 12 which gives a threshold of Ӎ270 kW cm −2 . 12Note that the threshold in semiconductor lasers is "softer" than that in other lasers due to the small cavity volume and to the relatively high levels of spontaneous emission. 10,12In summary, the data in Figs. 1 and 2 show the expected behavior of the output power of the nanowire device as it passes through the threshold for laser oscillation.
The sharp features observed in Fig. 1͑a͒ correspond to longitudinal Fabry-Pérot modes.The mode spacing in a Fabry-Pérot cavity is given by 10 ⌬ = ͑1 / L͓͒͑ 2 / 2͒͑n − dn / d͒ −1 ͔, where L is the cavity length and n is the index of refraction at wavelength .Thus, for a fixed , the mode spacing ⌬ should scale in proportion to the inverse length 1 / L. This behavior is illustrated in Fig. 3, which shows above-threshold spectra for nanowires of different lengths and a plot of ⌬ for nine nanowires.The slope of the fit is 10765͑Ϯ1.7%͒nm 2.With = 385 nm and n = 2.4, this gives dn / d Ϸ −0.012 nm −1 .This value is in reasonable agreement with published data for ZnO, 13 which gives dn / d Ϸ −0.015 nm −1 .
The threshold condition for laser oscillation is that the round-trip gain inside the nanowire must equal the round-trip losses, 10 where ⌫ is the confinement factor, i.e., the fraction of the mode intensity contained within the nanowire, g th is the material gain, ␣ w is the waveguide loss, ␣ m accounts for the losses at the end facets ͑i.e., the mirror losses͒, L is the length of the nanowire, and R is the reflection coefficient, assumed the same for both laser ends.In contrast to conventional edge emitting semiconductor lasers where typically ␣ w Ͼ ␣ m , in nanowire lasers, ␣ m ӷ ␣ w due to the much smaller cavity length and the smaller reflection coefficient. 14s a result, the threshold gain is a strong function of L and the nanowire diameter D. Figure 4 shows the devices that lased and the ones that did not ͑the former ones indicated by filled circles͒ in the plane defined by two key dimensions: diameter and wire length.Nanowires with diameters smaller than ϳ150 nm did not reach threshold, independent of the nanowire length.Qualitatively, this can be explained by the diameter dependence of the reflection coefficient R, which is directly related to the confinement factor ⌫, 14 as was explained in the introduction.For example, for the most confined mode in our geometry, the fraction of the mode inten- sity inside the nanowire decreases from ϳ85% for D = 150 nm to Ͻ1% for D = 100 nm.The exact calculations need to be performed numerically to account for substrate effects and will be presented in a future publication.The detection geometry used above only permits collection of a small fraction of the total output power since abovethreshold light is emitted primarily in the direction along the nanowire axis. 15In order to measure the output power of a nanowire laser, it is therefore necessary to collect the emission head on.Such an alternative geometry is depicted schematically in Fig. 5: A nanowire is shown partially suspended in air and partially resting on a substrate ͑500 nm of thermal oxide on silicon͒.The nanowire is then excited uniformly along its entire length, and the emission is collected from one end, at an angle of 90°from the excitation beam, using a silicon detector.The data in Fig. 5 show the detected peak power versus pump peak power.
In conclusion, we have presented unambiguous evidence of laser action in ZnO nanowires at room temperature through a systematic study of the evolution from ASE to laser oscillation above threshold.We have demonstrated the key dependence of laser threshold on nanowire diameter and measured the output power of a single nanowire laser using a head on geometry, which provides a useful benchmark for the future development of these nanoscale devices.

FIG. 1 .
FIG. 1. ͑Color online͒ Laser oscillation in ZnO nanowires.͑a͒ Output spectra vs pump intensity of a 12.2 m long 250 nm diameter ZnO nanowire.͑b͒ Scanning electron microscopy and CCD images under different pump intensities for the same nanowire as in ͑a͒.The labels indicate the pump intensity in units of MW cm −2 .The color scale indicates the number of counts.

FIG. 2 .FIG. 3 .FIG. 4 .
FIG.2.Laser oscillation in ZnO nanowires.Pump intensity dependence of the total output power ͑circles͒ for the same nanowire as in Fig.1.The optical power was collected from the scattered light at one of the nanowire ends.The inset shows the same data on a log-log scale as well as a fit to the data using the model in Ref. 12.