High-power semiconductor lasers are the most efficient man-made light sources and convert up to three-quarters of electric energy into light. Such operation, however, is connected with very high internal power densities of more than 10¹⁰ W/cm³, strong local heating, and sometimes even device degradation. Catastrophic optical damage (COD) of the mirrors of diode lasers is a relevant degradation mechanism and major limit for reaching ultra-high optical powers. So far, the detailed evolution of COD in space and time is not understood, making the effect unpredictable. We give an overview on research activities targeting the mechanisms being involved in the COD process, which turns out to be essentially a thermal effect. In particular the sequence of events and the fast kinetics taking place on a nanosecond to microsecond time scale are addressed. Paving the way for knowledge-based solutions towards more robust diode lasers represents the ultimate goal of this work. Figure 1 shows a typical laser structure and the way, how COD affects such diode lasers.

The basic idea of our current work is to monitor COD processes online during single current pulses (500 ns-2 µs) and thus to resolve them temporally. For this purpose we constructed a special setup that allows for simultaneous monitoring of the emission power (with 1 ns temporal resolution) and inspection of the front facet with a thermocamera, see Fig. 2. The latter allows detecting COD events by the ‘thermal flash’, i.e. an impulsive release of Planck’s radiation during the thermal runaway.

Limits of semiconductor laser operation in 808 nm emitting broad-area devices

Figure 3 shows current and spatially averaged power transients (left panel) obtained in single-pulse-experiments while the right panel gives temporally averaged thermal images. The presence or absence of the thermal runaway is unambiguously detected by the thermocamera. Furthermore, this data demonstrates the compression of the 'time to COD' to a few tens of nanoseconds for sufficiently high operation currents. Such data are compiled in Figure 4 (a). Threshold power for reaching the COD (PCOD) are plotted as full circles in case a thermal flash was observed. Open circles represent experiments where no COD and no thermal flash was detected. For this case, the ordinate gives the pulse length. Obviously, there is a borderline dividing the diagram (and hence the operation regimes) into two regions: one with and one without COD. The blurred region in-between is explained by the scattering of individual device properties within the sample set. The dashed line shows the 'square-root law', known from earlier literature. Thus, compliance with the data obtained in pulsed operation with multiple pulses is demonstrated. Such a diagram defines regimes in which COD is avoided by the proper choice of operation parameters for a particular batch of devices. It is important to note that the 'time to COD' scale is determined by the accumulation time of excess energy which is absorbed from the optical output and results eventually in a local melting of the device. This has been proven by making-up the energy balance within the tiny heated front facet volume; see Fig. 4 (c). The experimental value Wexp as derived from data sets such as given in Fig. 3 is compared with the energy required to reach the melting temperature of the active region material.

The time resolved experiments presented here demonstrate nanosecond time intervals to COD. This time scale is determined by the accumulation time of excess energy which is absorbed from the optical output and results eventually in a local melting of the device. Our work identifies a regime in which COD of high-power semiconductor lasers is avoided by the proper choice of operation parameters, allowing the ultimate limit of semiconductor laser operation to be reached. Such results potentially serve as a benchmark for future modeling of fast COD dynamics.

Time-resolved analysis of catastrophic optical damage in 975 nm emitting diode lasers

Monitoring a COD diagram, as described in the preceding section, requires a homogeneous set of devices which eventually gets destroyed. We developed an alternative approach based on a less complex step test with single, subsequently increasing current pulses. Before increasing the current by a further step, an additional test measurement with a regular operation current is made. This method was demonstrated with 975 nm emitting broad area lasers. Figure 5 shows selected current and emission power transients and the corresponding thermal images, while Fig. 6 summarizes the results from 4 devices. Fig. 6 (a) shows the power evolution in test pulses and Fig. 6 (b) quantifies the thermal flashes. There is a clear correlation between device degradation (power drop in the test pulse) and the first appearance of a ‘thermal flash’. Most striking, however, is the finding of multiple thermal flashes giving direct evidence of the presence of multiple thermal runaway processes in subsequent pulses. Figure 6 (c) shows even a lateral motion of the flash across about half of the emitter stripe width. A further important finding is illustrated by Fig. 5: the onset of the thermal runaway process at 32 A is seen as (weak) thermal flash but not as distinct power drop, as observed for shorter wave length devices; see Fig. 3. The time constant of power loss for the 975 nm emitting devices was determined to 400-2000 ns, while for the 808 nm emitting devices 30-400 ns have been observed. This is in agreement with the empirically known higher robustness of longer wavelength devices against COD.

The proposed test method shows how to artificially slow down the explosion-like COD process. The COD single-pulse step test bears the potential to become a standard COD test, because it reduces at the same time the expenses and effects related to thermal load and gradual aging.

Summary and Outlook

COD represents an important sudden degradation mechanism and major limit preventing diode lasers to reach ultra-high powers. COD is dominated by thermal mechanisms. This time flow can be separated into three phases:

• In the 1^st phase the facet temperature is approaching a critical temperature Tcrit. Depending on the surface status of the facet and on operation conditions this phase takes from years (continuous wave operation) down to a few tens of nanoseconds (short high current pulses).
• The 2^nd phase involves the thermal runaway. Strongly localized melting takes place and the 'hot spot' spread spatially. Model calculations predict typical rise times on the order of one ns. So far, this phase has never been experimentally resolved, but our energy balances (see Fig. 3) point to this timescale as well.
• The 3^rd phase of COD begins when the thermal runaway stops because of a lack of energy provided by the laser emission. Further degradation takes place, in particular if the operation current is not abruptly terminated. This represents the standard case. This phase involves the creation of collateral damage that is not necessarily directly related to the thermal runaway. The duration of this phase depends much on operation conditions and might last up to minutes. Typical values are rather in the µs-ms range (if the operation current remains on).

• The ratio between bulk and facet heating can be adjusted.
• Single-pulse operation helps for separating the role of the actual chip architecture from that of the thermal properties of the package.
• It allows for the intentional preparation of pre-stages and very early phases of the COD process. Devices intentionally degraded in such a way are interesting for further (destructive) analysis.
• COD can be analyzed even for diode lasers that fail under regular operation conditions by other mechanisms than COD.
• The single-pulse approach could help for establishing standard COD-test procedures.

Acknowledgements

The authors would like to thank the following scientist for helpful discussion: Ute Zeimer, Bernd Sumpf, and Götz Erbert of the Ferdinand Braun Institute Berlin, Stella Elliott and Peter Smowton, of the University of Cardiff, Wlodzimierz Nakwaski of the Technical University of Lódz, Martina Baeumler, of the Fraunhofer Institute for Applied Solid-State Physics Freiburg, Henning E. Larsen, Peter E. Andersen, and Paul M. Petersen of the Danish Technical University, Jose Manuel G. Tijero and Ignacio Esquivias of the Universidad Politecnica de Madrid, Michel Krakowski and Nicolas Michel of Alcatel-Thales III-V lab Palaiseau, Julien Nagle of Thales Research and Technology Palaiseau, Peter Brick, Martin Reufer, Martin Müller, Harald König, and Uwe Strauss of Osram Opto Semiconductors Regensburg, and Marwan Bou Sanayeh of Notre Dame University Louaize, Lebanon.

Funding by the European Commission within the project WWW.BRIGHTER.EU under Contract No. IST-2005-035266 is acknowledged.