Etching a substrate: processes – Nongaseous phase etching of substrate – With measuring – testing – or inspecting
Reexamination Certificate
1999-03-01
2001-10-30
Gulakowski, Randy (Department: 1746)
Etching a substrate: processes
Nongaseous phase etching of substrate
With measuring, testing, or inspecting
C216S084000, C438S014000, C438S016000, C451S006000
Reexamination Certificate
active
06309555
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention generally related to a method for determining a real-time thickness of a single material layer, and more particularly to a method for determining an endpoint for a chemical-mechanical polishing (CMP) process.
2. Description of the Related Art
Chemical-mechanical polishing (CMP) is currently regarded as the only technology to provide global planarization in fabricating very-large scale integrated (VLSI) circuits, or even ultra-large scale integrated (ULSI) circuits. Essentially, the CMP is based on a principle similar to a “polishing wheel” in mechanical polishing. With the aid of a reagent, the rough surface of a wafer can then be smoothly ground by using this technology.
As the technologies of fabricating semiconductor devices enter the deep sub-micron regime, the CMP has become a conventional technology in fabricating ICs, in which an endpoint detection (EDP) method is the most critical factor in extending process window during the CMP process and overcoming stability problem in mass production.
To perform the endpoint detection for a dielectric layer, optical methods are the most popular ones, which measure the intensity of a reflected light as a function of time. Working together with the “Window Logic”, turning-around points of slope on the curve can be identified, so that the endpoint of the CMP process can be determined. This method has advantages of real-time detection, no direct contact, and less noise induced.
The following paragraph describes the principles on which the endpoint detection method is based. Problems associated with this method are also described hereinafter.
Referring to
FIG. 1
, a light I
inc
is incident onto a dielectric layer
10
. Being reflected from the top surface of the dielectric layer
10
, and an interface between the dielectric layer
10
and an underlying substrate
12
, the intensity of a reflected light I can be obtained from principles of optical interference:
I=I
A
+I
B
+2{square root over (
I
A
I
B
+L )}cos&phgr; (1)
φ
=
4
π
nd
λ
0
cos
(
α
ref
)
and I is the intensity of a reflected light, t is the time, I
A
is the intensity of the first reflected light, I
B
is the intensity of the secondary reflected light, n is the refraction index of the material layer, d is the thickness of the material layer, &lgr;
0
is the wavelength of the incident light, &agr;
ref
is the refractive angle.
Eq. (1) shows that the intensity of the reflected light versus polishing time (I-t curve) comprises a cosine function as shown in FIG.
2
. However, neither the thickness nor the endpoint of the process can be obtained from Eq. (1), unless a correlation between the intensity curve and the dielectric layer thickness was established. Consequently, it is very different to detect the endpoint of a CMP process on which applied to a single dielectric layer, for example, inter-layer dielectric (ILD) or shallow trench isolation (STI), since there is no strong signal difference coming from the interfaces of two different materials. Thus, the endpoint of the CMP process needs to be determined depending on the periodical variation of I-t curve.
For &phgr;=m&pgr;,m=0, 1, 2, . . . , the corresponding thickness d
m
can be obtained by
d
m
=
m
λ
0
cos
(
α
ref
)
4
n
(2)
the peaks and valleys on the I-t curve correspond to a certain thickness provides the initial thickness of the dielectric layer is know. Referring to and I-t curve obtained from polishing the whole thickness of a layer, these peaks and valleys form a set of characteristic points or indices, which can be used to determine the endpoint of the CMP process.
As shown in
FIG. 2
, a local minimum point
202
on the curve which is closest to the desired endpoint of a CMP process can be obtained, which is called the “analytical endpoint”
204
because it can be explicitly defined on the curve. This local minimum point
202
can be identified by “window logic”, as it is the changing point of slope. After reaching the analytical endpoint
204
, an over polishing step is further performed until reaching a required final thickness
206
.
Thus, for a single dielectric layer (ILD) or a shallow trench isolation (STI) CMP process, the conventional endpoint detection technology identifies several local maximum/minimum points on the curve which have a thickness close to the desired thickness based on wavelength of the incident light.
However, there are still some problems for the above-mentioned principle to be applied practically. Although the peaks and valleys refers to a certain thickness are those identified with a slope of zero on the curve, it is not easy, however, to have an accurate calculation for these points because of the constraint on the speed of the polishing table. For example, sampling rate in a CMP process is generally only 1 sample/second due to a reasonable process parameter, which is not sufficient to accurately identify points with a zero slope. A more practical way to identify the peaks and valleys with zero slope is through the turning-around points on the curve by identifying changes between positive and negative slopes.
FIG. 2
shows a conventional “window logic” method to determine the analytical endpoint, in which observing windows
208
are used to identify the turning-around points on the curve. This method has an advantage of simplicity, but associated accuracy and reoccurrence are not so good. To minimize errors that might occur due to noises and interference on the curve, consecutive three to four windows are normally required to define a turn-around point. The inaccuracy of this method therefore guarantees no successful results in detecting peaks and valleys on the curve.
Due to different material used, environmental and process variations, problems encountered so far toward mass production includes that the thickness determination may not reflect the true thickness accurately since signals varies upon polishing material. In addition, the CMP processes demanding a very short polishing time, especially those having a polishing time of less than a signal period, the conventional method is not applicable. Furthermore, the corresponding I-t curves for each wafer can be dirrerent in both amplitude and period, because of difference in wafer and environment. Usually, different endpoint thickness is required for different processes, so that the corresponding endpoint on the I-t curve is different. Therefore, feasible rules need to be established to solve these problems.
The reason behind the above-mentioned problems is that the conventional signal analysis technology fails to accurately identify the peaks and valleys on I-t curve. The conventional method fails to quickly and correctly identify the slope of a point on the curve, whether it is descending or ascending.
SUMMARY OF THE INVENTION
It is therefore an objective of the present invention to provide a method for accurately identifying the peaks and valleys of intensity of the reflected light on the I-t curve, so that the thickness of the dielectric layer can be detected.
It is another objective of the present invention to provide a method for quickly and correctly determining whether the slope on the curve of the intensity of the reflected light is descending or ascending, so that the accuracy in detecting the dielectric layer can be improved.
It is another objective of the present invention to provide a method to overcome the problems mentioned earlier so that the endpoint detection method can be used in mass production processes.
In accordance with the foregoing and other objectives of the present invention, a method for real-time determining the thickness of a material layer is provided. The thickness of the material layer on a substrate decreases as time elapses. A reflected light is measured by using an incident light emitted toward the material layer. By integrating the
Ahmed Shamim
Gulakowski Randy
Thomas Kayden Horstemeyer & Risley LLP
United Microelectronics Corp.
LandOfFree
Method for determining thickness of material layer and... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method for determining thickness of material layer and..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method for determining thickness of material layer and... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2602339