The present Application relates to semiconductor laser devices and methods of manufacturing laser devices.
The production yields of single mode, single section, edge emitting laser diodes are much lower than those of ridge waveguide Fabry Perot lasers. There are two principal reasons for this, the general reliability of such devices and the fabrication tolerances required to make them.
A reason for the poor reliability of these single mode devices is the fact that the fabrication of these devices usually requires several epitaxial growth steps. After the first epitaxial growth a grating is formed on the surface of the wafer using well photolithographic techniques. This means that the second wafer growth step is initiated upon a damaged uneven crystal surface. As a result lasers fabricated in this way have relatively high defect densities and are thus intrinsically less reliable than devices fabricated in a single epitaxial growth step.
A second reason for the reduced production yields of these devices compared with Fabry Perot lasers is due to the inability to cleave the wafer into laser bars with sufficient accuracy. This means that in a significant proportion of the devices the cleaved facets are incorrectly positioned with respect to the wavelength selective filters (gratings in the case of DFB's) within these devices. This incorrect positioning in turn results in large numbers of devices which have unsatisfactory side mode suppression ratios, or do not operate in a single longitudinal mode over a useful temperature range.
It would be advantageous to alleviate the effects in the fabrication process of laser devices.
The present application uses the fact that in single stripe laser diodes only a very small percentage of the device's active region is used for the purpose of generating light and that accordingly there is space for including more than one laser device, with a subsequent step selecting the best laser device. It should be noted that some single mode, single section devices do not require two re-growth steps, the application as discussed below is also applicable to such devices.
Accordingly, the present application provides a laser device fabricated on a piece of semiconductor material, the laser device comprising at least a first lasing device and a second lasing device, whereby at least one of the lasing devices is intended to be redundant. This will be explained further with reference to and as shown in the accompanying drawings in which:
As described above and as illustrated in
Before attempting to quantify improvements in production yields which this technique brings, a number of factors which affect how accurately the cleaved facets may be positioned with respect to the wavelength selective filters will be discussed. The two main factors which affect this are, the accuracy of the cleaver itself (typically ±2 microns), and the rotational accuracy with which lithographic patterns may be aligned to the wafer (this is typically better than ±0.005 degrees for contact lithography and better than ±0.04 degrees for optical steppers). Consider an exemplary bar of single stripe laser diode devices 15 mm in length, the bar consisting of 50 devices, 300 microns wide, each of which is supposedly identical. If the mask patterns and the crystal axis were perfectly aligned, then corresponding features on different devices would be equidistance from either facet. However because this is not the case, the perpendicular distances of corresponding features on different devices from either facet, varies linearly (or nonlinearly if the cleave jumps to a different crystal plane) from one device to the next (
If for example in a single stripe device the probability that a randomly positioned wavelength selective filter is aligned correctly with respect to the facets is Pf. Then the probability that the filter will not be aligned correctly with the facets is 1−Pf. Then on an a device with N stripes, the probability that at least one of the wavelength filters will be correctly aligned to the cleaved facets is
1−(1−Pf)N.
However we may do better than this, consider
(typically the slot separations are a large multiples of this length) additionally the slot patterns are staggered with respect to each other by
It is noted here that the material wavelength λm is equal to
(where neff is the effective index of the mode). We can see that this situation provides a high chance of at least one lasing stripe meeting the criteria for a given application. In the case where
this approach should provide a yield of 100%. This is also the case if the distance between the slots on each ridge is a large multiple of,
as is typically the case. However for this method to work well the two ridge waveguides guides should be positioned close enough to each other, so that the error in positioning lithographic features relative to the facet, on one ridge with respect to the other should be approximately
In the discussion above we described one way of staggering the slot patterns on adjacent ridges with respect to each other. This was done directly by designing mask plates (optical lithography) or exposure patterns (e-beam lithography) in which these patterns were staggered. When the wavelength selective filters (slot patterns) are defined using optical or e-beam lithography this is by far the most straight forward way of staggering these filters with respect to each other. However it is also possible to change the phase between the wavelength filters and the facets by changing the effective index of part of the waveguide, or by intentionally misaligning the lithographic patterns with respect to the crystal axis by about 0.4 degrees. It is noted the last of these two methods may be applied to single mode lasers which have holographically defined gratings or wavelength selective filters defined by optical or e-beam lithography. The first of these methods however could only applied to structures in which holographically defined gratings extend along only portion of the cavity length.
First we discuss the effect of altering the effective index of part of one of the stripes of a dual stripe laser.
It is also possible to alter the phase difference between the facet and wavelength selective filters on adjacent ridges by deliberately misaligning the mask with respect to the crystal axis. This angular this alignment is chosen so as to have the same effect as wavelength filters directly staggered with respect to each other, and misaligned to the crystal axis by less than 0.05 degrees. So in the new situation the angular misalignment would be something in the region of 0.5±0.05. The smaller the error in the angular alignment, the smaller the actual angular misalignment needs to be. As the optical properties of the devices may be affected by large angular misalignments, they should be kept as smaller possible. This may be achieved by increasing the distance between adjacent stripes. This may be accomplished (without increasing the overall device size) by placing one of the bondpads of the two devices between the two devices (see
Depending on the precise positioning of the wavelength selective filters with respect to the facets of the device a discrete mode laser will lase in one several modes. This effect is most readily seen by testing a number of adjacent dual devices (approx 55) on a cleaved bar. It will be appreciated that the number of modes which have the capability to lase can be specified by using an appropriate pattern of slot features.
with respect to the pattern on the left stripe for each device. This is why the variation of the emission wavelength with respect to device position is similar on the left and right stripes, but shifted with respect to each other. It is noted the more central bands of the wavelength distribution perform best, because of this the performance benefit is obvious. In devices in which the left hand stripe lases in one of the extremum wavelengths (e.g. device number 11), then the right hand stripe will lase in one of the more central wavelengths. The situation is similar if one of the right hand stripes lases in one of the extremum wavelengths (e.g. device number 30).
Finally we consider an exemplary device containing two different types of device, for example a Fabry Perot chip and a single frequency laser diode. The motivation for this, with reference to the example is as follows, the yield of Fabry Perot device is almost 100%, if the single mode laser doesn't work correctly there is almost a 100% chance that this chip will still be useful.
Number | Date | Country | Kind |
---|---|---|---|
S2005/0587 | Sep 2005 | IE | national |
Filing Document | Filing Date | Country | Kind | 371c Date |
---|---|---|---|---|
PCT/EP2006/066061 | 9/6/2006 | WO | 00 | 8/5/2008 |