Electric heating – Metal heating – By arc
Reexamination Certificate
2002-01-07
2004-09-14
Heinrich, Samuel M. (Department: 1725)
Electric heating
Metal heating
By arc
C219S121690
Reexamination Certificate
active
06791059
ABSTRACT:
BACKGROUND OF THE INVENTION
This invention relates to employing a laser beam to vaporize or otherwise alter a portion of a circuit element on a silicon substrate and is particularly applicable to vaporizing metal, polysilicide and polysilicon links for memory repair.
Semiconductor devices such as memories typically have conductive links adhered to a transparent insulator layer such as silicon oxide, which is supported by the main silicon substrate. During laser processing of such semiconductor devices, while the beam is incident on the link or circuit element, some of the energy also reaches the substrate. Depending upon the power of the beam, length of time of application of the beam, and other operating parameters, the silicon substrate can be overheated and damaged.
Laser processes of this kind have typically been conducted at wavelengths of 1.047 &mgr;m or 1.064 &mgr;m. Silicon has sufficiently low absorption at these wavelengths that the amount of beam energy employed to evaporate typical polysilicide and polysilicon links has not harmed the underlying silicon substrate.
It has been recognized, e.g., by the present inventor, by Lapham et al. in U.S. Pat. No. 4,399,345, and by others, that in laser processing of semiconductor devices, it can be advantageous to use wavelengths beyond the “absorption edge” of silicon (i.e., wavelengths greater than about 1.1 &mgr;m, where the absorption of silicon drops precipitously). This makes the silicon substrate more transparent to the laser beam, and reduces heating of the silicon caused by absorption of the beam. The preferred wavelength generally mentioned for this purpose has been 1.32 or 1.34 &mgr;m, though a broad theoretical range of low absorption has been identified. The 1.32 or 1.34 &mgr;m wavelength has been proposed as it is close to the minimum absorption of silicon.
Removal of memory links for memory repair is an application in which these considerations are relevant. As noted in “Computer simulation of target link explosion in laser programmable redundancy for silicon memory” by L. M. Scarfone and J. D. Chlipala, 1986, p. 371, “It is desirable that laser wavelengths and various material thicknesses be selected to enhance the absorption for the link removal process and reduce it elsewhere to prevent damage to the remainder of the structure.”
The usefulness, in general, of thicker insulative layers underneath links or circuit elements, and the usefulness of limiting the duration of heating pulses has also been recognized, as in the paper of which I was author, “Laser Adjustment of Linear Monolithic Circuits”, Litwin and Smart, 166/L.I.A., Vol. 38 ILAELO (1983).
Makers of semiconductor devices typically continue production of earlier developed products while developing and entering production of more advanced versions that typically employ different structures and processes. Many current memory products employ polysilicide or polysilicon links while smaller link structures of metal are used for more advanced products such as the 256 megabit memories. Links of 1 micron width, and ⅓ micron depth, lying upon a thin silicon oxide layer of 0.3 to 0.5 microns are being used in such large memories. Production facilities typically have lasers and related equipment capable of operating at the conventional wavelengths of 1.047 &mgr;m or 1.064 &mgr;m and also wish to have lasers and related equipment capable of operating in the wavelength region recognized for its lower absorption by silicon.
SUMMARY OF THE INVENTION
I have realized that, while taking advantage of the lower absorption of silicon at a wavelength beyond the absorption edge of about 1.1 &mgr;m, and by tightly constraining the duration of the laser pulse in which energy is delivered, it is possible in many instances to obtain an improved overall result in throughput of successful repair of memory links, e.g. in high density memories, at high repetition rate of the laser pulses. Preferably the pulse duration for imparting sufficient energy to remove the link is constrained to less than 10 nanoseconds, more preferably to less than about 5, and even more preferably to about 4 nanoseconds or less, while using a pulse repetition rate of at least 5 Khz and preferably 8 Khz, 10 Khz or higher.
By so limiting the duration in which the energy required to produce link removal is delivered, temperature rise in the silicon can be limited to prevent damage.
I have realized that, with the very small spot size used with small metal links, the heat may be considered to spread in essentially an exponential gradient by conduction from the portion of the beam striking the target. By employing a peak beam power so high that sufficient energy for evaporation of the link is delivered in a pulse of 8 nanoseconds, and preferably substantially less, the conductive component of heat transfer can be substantially confined to a metal link and the underlying oxide layer, despite its being very thin, such that the temperature rise in the silicon attributable to conduction and the temperature rise attributable to absorption of the beam in silicon, can cumulatively be kept below the temperature threshold at which unacceptable silicon damage occurs.
I have further realized that, to achieve the results with current practical technology, a laser having a wavelength that is not beyond the absorption edge of silicon, capable of high gain at high repetition rate conditions, should be employed to produce an original output beam and a shifting system should be interposed to shift the wavelength of the beam beyond the absorption edge of silicon. Specifically, I have realized that a laser with a high-gain wavelength, e.g. at 1.047 or 1.064 microns, shifted to a longer wavelength beyond the absorption edge, can produce the desired short pulse in which the requisite energy is delivered.
Using a neodymium vanadate laser with a short cavity length, a pulse width of less than 5 nanoseconds can be achieved at repetition rates up to 10 Khz, and of 4 nanoseconds up to about 8 Khz.
By employing a system in which the pulse rate is thus at least 5 Khz, and preferably 8 Khz or 10 Khz, at a wavelength beyond the absorption edge of silicon with a pulse width of less than 8 nanoseconds, and preferably less than 5 nanoseconds or 4 nanoseconds, one can obtain clean removal of a link without leaving conductive residue.
Furthermore, I have realized that by observing an upper limit on the wavelength of less than about 1.2 &mgr;m, the overall process advantage of better focusability gained at such wavelength can outweigh any advantage that could otherwise be obtained by using longer wavelengths at which the absorption of silicon is more minimized. The sharpness of spot size and good depth of focus lead to reliable processing of very large die such as are carried on an 8-inch wafer, which may have significant variation in flatness over the die surface. In particular, many 8-inch wafers are less than about 300 microns thick and suffer a condition sometimes referred to as the “potato chip” effect, in which the surface curves or otherwise is not flat. The smaller spot size and better depth of focus obtainable by maintaining the wavelength less than 1.2 &mgr;m can improve the defect rate with such wafers in excess of what might be obtained when longer wavelengths, e.g. 1.32 &mgr;m, are employed.
Thus, very clean removal of metal links of current design, e.g. of 1 micron width and ⅓ micron thickness, can be obtained, with high throughput of memory repair and with relatively few rejects.
According to another aspect of the invention, an optical system is employed to accurately focus a laser spot when operating at a wavelength that is not beyond the absorption edge and the same optical system is used, without modification, at a wavelength beyond the absorption edge, constrained less than 1.2 &mgr;m. This provides a versatile new laser system that can be inexpensively retrofitted into existing laser systems. For instance, the initial 1.064 &mgr;m laser beam output of the system is useful directly for processing conventional polys
Fish & Richardson P.C.
GSI Lumonics Corp.
Heinrich Samuel M.
LandOfFree
Laser processing does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Laser processing, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Laser processing will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3262008