Semiconductor Yield Modeling
Monday 2 June, 2:40--4:30
- Organizer: Mark HANSEN
- Bell Labs
- Moderator: Mark HANSEN
- Bell Labs
- Charles M. DRUM
- Lucent Technologies, Allentown, PA, USA
As a supplement to statistical yield models, the mechanistic
yield model for IC chips is based on random yield-loss mechanisms.
These include lateral shorts between runners in the various
conductor levels, interlevel shorts between the conductors on
different levels, open circuits in the various conductor runners,
and open circuits at the various window configurations. Each
mechanism is characterized two other quantities. These are the
critical area per mechanism, which is design specific; and the
defect density per mechanism, which is design independent. The full
model combines the contributions of random yield with those of
non-random yield. This model is largely empirical since the critical
areas per mechanism are deduced from the particular IC
design, and the defect densities per mechanism are measured with
test structures.
The results are used to rank order the effects of particular random
yield-loss mechanisms, and to determine the contributions of
design and of processing. Some examples will be discussed,
including various ICs (memory, ASICs, processors, mixed-signal) and
various processes including deep sub-micron MOS.
[Charles M. Drum, Lucent Technologies,
555 Union Boulevard,
Allentown, PA 18103 , cdrum@lucent.com]
- C. Neil BERGLUND
- Oregon Graduate Institute of Science and Technology, Portland, OR, USA
In a recent publication it has been shown that large-area
yield loss contributors such as those caused by poor parametric control
can be modeled through a straightforward extension of the Murphy yield
integral1. Of particular importance in this work is the conclusion that
such large-area defects result in a die-size-dependent reduction in yield
that is similar to that caused by small-defect clustering. In this paper
the model is extended both to help distinguish between such large-defect
problems and small-defect clustering, and to relate yield loss to
parametric control parameters such as the capability indexes Cp and Cpk
as well as to the often-observed yield drop near the edges of wafers.
Experimental data is compared to the predictions of the model.
1. C. Neil Berglund, "A Unified Yield Model Incorporating Both Defect
and Parametric Effects," IEEE Trans. Semiconduct. Manufact., vol. 9, no.
3, pp. 447-454, 1996.
[C. Neil Berglund, Oregon Graduate Institute of Science and Technology,
PO Box 91000,
Portland, OR, 97291-1000, USA, berglund@eeap.ogi.edu]
Back to JRC program
Back to JRC Home Page