1. Introduction

General illumination applications using III-nitride based phosphor-converted light emitting diodes (pc-LEDs) are beginning to gain wide adoption in the lighting marketplace and are well poised to succeed traditional incandescent and compact florescent illuminants as the primary light sources for indoor, outdoor, and automotive lighting applications. However, the nonthermal efficiency droop of light emitting diode (LEDs) with increasing current density limits the output power emitted per wafer area and is an ongoing topic of investigation. Suggested origins of efficiency droop include Auger recombination, carrier leakage out of the active region, and overflow of carriers from In-rich potential minima [1–7]. While peak luminous efficacies exceeding 200 lm/W have been reported in LED-based lighting products, these values are generally at low current densities on the order of ~10 A/cm² [8,9]. Since low current densities are required for high efficiency operation, many device die are needed for typical high power white light illumination systems operating at peak efficiency.

In contrast to LEDs, the efficiency of LDs does not nonthermally droop above threshold and high efficiencies can be achieved at very high current densities, resulting in a very high output power per wafer area [9–12]. Figure 1 illustrates the LED epitaxial area and LD epitaxial area needed to produce enough power to replace a 60 W incandescent bulb while operating at peak efficiency. In addition, LDs have highly directional light emission, which can be useful for a number of applications. This suggest that LDs could be used in certain white lighting systems where a small, directional emitter can be useful to reduce cost or meet certain design specifications.

 

Fig. 1 Illustration of the LED epitaxial area and LDepitaxial area needed to produce enough power to replace a 60 W incandescent bulb.

Download Full Size | PDF

Full laser-based lighting systems have shown that color rendering index (CRI) is not significantly hindered by the sharp spectral nature of lasers. A study performed by Neumann et al. set out to show if there is a statistical preference for a particular illumination type when comparing white LEDs, incandescent bulbs, and laser illumination. In the study, identical scenes were illuminated by a white light produced by mixing and scattering red, green, yellow and blue laser emission with CRI ranging from 83 to 91 and a reference illuminant consisting of a cool-white LED, a warm-white LED, a neutral-white LED, or an incandescent bulb. Participants in the study then rated their preference for either illumination with the traditional white source or the laser-based source tuned for a matching correlated color temperature to that of the reference. While the incandescent bulb was preferred over all sources, the laser illumination was favored equally to the warm-white LED and preferred over both the cool-white and neutral-white LEDs [13]. However, CRI has been shown to be a poor metric for color rendering of highly engineered white spectra composed of narrow wavelength peaks. A study performed by A. David demonstrated poor performance of CRI as a metric for color rendering for laser-based systems similar to those used by Neuman et al. Improved color rendering fidelity was found for sources including broad wavelength distribution phosphor emission versus emission composed of only narrow wavelength band emitters such as discrete LEDs [14]. This suggests the use of phosphor-converted LDs (pc-LD) or inclusion of other broad spectrum emitters is needed to maximize color rendering of laser-based white lighting systems.

In practical applications, using red, green, yellow and blue lasers for white lighting has similar implementation complexities as using red, green, and blue LEDs for white lighting. These multiple emitter white lighting systems will need discrete drivers for each color coupled with feedback to allow automated color control. This complexity can be avoided by using phosphor conversion of a single color solid state emitter coupled with appropriate wavelength converting phosphor material. Upon the advent of pc-LED white lighting, it was found that many phosphors quench due to heating from high optical power densities. Use of phosphors which saturate with increasing excitation leads to significant color variation with increasing excitation. One conventional phosphor, cerium-doped yttrium aluminum garnet (YAG:Ce), has been shown to not saturate under the increasingly high incident power densities from LEDs and LDs [15–17]. Previous demonstrations of pc-LDs for white lighting have shown stable color temperatures with increasing drive current but were limited by luminous efficacies of less than 40 lm/W (76 lm/W) and luminous fluxes less than 100 lm (252 lm) under CW (pulsed) operation [15–17]. Wall plug efficiencies (WPEs) for GaN-based LDs in the blue spectrum have rapidly improved over the past few years [18,19]. Due to these rapid developments in GaN-based LDs, laser-based white lighting systems may soon compete with LED-based white lighting systems for multiple applications. In this paper, we demonstrate that a single pc-LD die is capable of emitting enough luminous flux for use in a high power white light illumination system. Typical application of phosphor powder in silicone matrix is not suitable for use under high level of static excitation. High thermal conductivity of solid ceramic phosphors permits the use of greatly increased excitation without deterioration with increased efficiency [20].

2. Experimental setup

For these experiments, a commercial GaN-based LD emitting at 442 nm was pressed into a copper heat sink and a single aspheric lens was used to collimate and focus the beam. The heat sink was mounted in the side access port of a 50 cm diameter integrating sphere (Instrument Systems ISP 500), with the beam entering through a port 8 mm in diameter. Table 1 lists properties of the LD including the threshold current (I_th), the threshold voltage (V_th), the slope efficiency, and the WPE and forward voltage (V_f) at 1.4. The LD had a slope efficiency of 1.66 W/A, corresponding to a differential efficiency of 59.2%, and a peak WPE of 31.6% at 1.4 A. A 2 mm thick YAG:Ce single crystal phosphor with 0.15% (0.03%) weight (atomic) percent of Ce and a diameter of 1 inch (purchased from Marketech International) was held in place by a Teflon fixture in the center of the integrating sphere and the laser beam was aligned so that it was incident on the disc. The YAG:Ce single crystal phosphor disc was oriented so that laser beam was incident at an angle 30° from the surface normal of the disc to avoid reflections back to the LD. A photograph of the experimental setup is shown in Fig. 2. Color adjustments were made by directing a portion of the defocused laser beam away from the single crystal phosphor to allow direct control of the blue to yellow ratio present in the integrated white emission. For this experiment 77.4% of the blue laser emission was incident on the phosphor. Full integrated measurements were taken using SpecWin Light and an Instrument Systems MAS40 spectrometer. All measurements were taken under CW conditions using a 2440 Keithley Sourcemeter.

Table 1. Laser diode properties.

View Table

 

Fig. 2 Photograph of the experimental setup showing a 50 cm integrating sphere, a YAG:Ce single crystal phosphor disc at the center of the integrating sphere, and a commercial GaN-based LD located at the side access port of the integrating sphere that was aligned so that it was incident on the phosphor disc. The YAG:Ce single crystal phosphor disc was oriented so that laser beam was incident at an angle 30° from the surface normal of the disc to avoid reflections back to the LD.

Download Full Size | PDF

3. Experimental results and discussion

As indicated in Fig. 3(a), the pc-LD had a peak luminous efficacy of 86.7 lm/W at 1.4 A with a luminous flux of 515.4 lm. The luminous efficacy decreased to 75.6 lm/W at a peak luminous flux of 1.10 klm at 3.0 A, which was limited by the maximum source current of the power supply. The decrease in the luminous efficacy at high currents was caused by an increase in V_f (see Table 1) and a slight decrease in the slope of the luminous flux with current. The luminous flux increased linearly between the onset of lasing at 0.28 and 1.4 A, but increased sublinearly from 1.4 to 3.0 A, due to heating of the LD. Even under CW operation, there were no adverse impacts on the YAG:Ce single crystal such as discoloration, bleaching, or other physical damage that might be expected under high incident optical power density. Among the advantages of a pc-LD white lighting system is a significant reduction of luminous efficacy droop. The luminous efficacy of a 1 mm² high power pc-LED reported by Narukawa et al. showed a droop of 17.6% over an increase in current of only 0.3 A [8], while the luminous efficacy of the pc-LD exhibited a droop of only 12.8% over a much larger increase in current of 1.6 A. This nearly droop free performance is due to the fact that carrier density clamps at threshold for LDs. While LED carrier density continues to increase with increasing drive current, LD carrier density remains unchanged above threshold, so all recombination rates other than stimulated emission also remains unchanged above threshold [10–12]. Although the emission from LDs is negligible below threshold, at very high current densities this is offset by the nearly linear and relatively high differential efficiencies (e.g. > 50%) that can be achieved with high power LDs.

 

Fig. 3 (a) Dependence of luminous flux and luminous efficacy of the pc-LD on current. (b) Spectra of the pc-LD at 1.4 and 3 A.

Download Full Size | PDF

Spectra of the pc-LD at 1.4 and 3 A normalized to the intensity of the peak emission wavelength of the LD are showen in Fig. 3(b). To achieve a light source that falls on the Plankian locus, the integrated power of the phosphor was adjusted so that it comprises 67% of the total power for a total system WPE of 26%. Although this spectrum produced a relatively high correlated color temperature (CCT) of 7252 K and a relatively low CRI of 60, the CCT can be decreased and the CRI can be increased by using LDs with different emission wavelengths and different phosphors or mixtures of phosphors. These spectra also do not show evidence of phosphor quenching in the single crystal YAG:Ce as LD current is increased from 1.4 to 3 A; the relative phosphor emission region of the two spectra match within instrument error. It is possible for increased thermal loss to occur at increased current without change in the relative spectra which cannot be identified without direct measurement of the system efficiency. At a drive current of 1.4 A the power emitted from the LD without the YAG:Ce single crystal phosphor was 1.90 W, whilethe power emitted from the LD with the YAG:Ce single crystal phosphor was 1.54 W. Compensating for the Stokes shift from the emission from the LD at 442 nm to the YAG:Ce emission with centroid wavelength of 568 nm, we calculate a quantum yield of 95%

Figure 4 shows a calculations of luminous efficacy and CCT for a pc-LD where the LD wavelength was varied from 400 nm to 470 nm in steps of 5 nm and the spectrum of the YAG:Ce phosphor is the same as the spectrum that was presented in Fig. 3(b). The calculation assumes that the WPE of the LD is 31.6% and that the quantum efficiency of the YAG:Ce phosphor is 95%. These simulations were performed for comparison with our experimental results and also to see the effect of changing LD emission wavelength on efficacy and CCT when YAG:Ce is used as the phosphor. Each simulated LD-phosphor combination was adjusted for white emission on the Plankian locus with Duv < 0.006. Since the system simulates mixing two emitter colors (LD emission and YAG:Ce phosphor emission) the resulting combined emission may only coincide with the Planckian locus at one point on the CIE 1931 x y chromaticity diagram. At a lasing wavelength of 442 nm, the calculated luminous efficacy of 86.1 lm/W and CCT of 7083 K agree well with our experimental data of 86.7 lm/W and of 7252 K, respectively. The simulation is calculated assuming phosphor quenching does not occur. Agreement between the simulation and experimental results suggests phosphor quenching is not significant while operating at peak LD efficiency. The simulation was performed again using powdered YAG:Ce and 442 nm LD spectral emission data. The calculated luminous efficacy for the powdered phosphor-LD emission of 88.7 lm/W, and CCT of 4608 K shows a small increase in efficiency for emission with CCT more suitable for lighting. For each change in LD emission wavelength, variations in the white color can be observed in the plot of CCT. Along with the change in CCT, the ratio of blue LD emission to phosphor emission must vary. While the loss due to the Stokes shift decreases with increasing LD emission wavelength, the simulation predicts that the peak luminous efficacy should occur at 455 nm before declining with an increase in wavelength. This decrease in luminous efficacy at wavelengths longer than 455 nm occurs because it is necessary to increase the blue LD to phosphor emission ratio to achieve white emission on the Plankian locus. Since the blue LD emission has a much lower luminous efficacy of radiation that then YAG:Ce emission the luminous efficacy decreases at longer LD emission wavlengths. Simulations of combining the YAG:Ce phosphor with LD emission wavelengths above 470 nm did not allow a white combined emission with Duv < 0.006 as no color combination falls on the Planckian locus. Nevertheless, we should emphasize that these results are exclusive to the combination of a LD with a YAG:Ce phosphor. Alternative phosphors or mixtures of phosphors can be used with LDs with different emission wavelengths to a achieve a more flexible range of luminous efficacies, CCTs, and CRIs.

 

Fig. 4 Calculations of luminous efficacy and CCT for a two component pc-LD where the LD wavelength was varied from 400 nm to 470 nm in steps of 5 nm and the spectra of the YAG:Ce phosphor is the same as the spectra that was presented in Fig. 3(b). The calculations assume that the WPE of the LD is 31.6% and that the quantum efficiency of the YAG:Ce phosphor is 95%.

Download Full Size | PDF

4. Conclusion

We have demonstrated a phosphor-converted single LD-based white light source with a luminous flux that meets general illumination requirements. A peak luminous flux of 1100 lm with a CCT of 7300 and CRI of 62 were demonstrated under CW operation with a peak luminous efficacy of 86.7 lm/W at 1.4 A. If the WPE of GaN-based LDs is able to approach the WPE of state-of-the-art GaAs-based LDs (~70%), it will be possible to achieve luminous efficacies in excess of 200 lm/W. These results show that a single pc-LD die is capable of emitting enough luminous flux for use in a high power white light illumination system.