US20060105676A1

US20060105676A1 - Robust Signal Processing Algorithm For End-Pointing Chemical-Mechanical Polishing Processes

Info

Publication number: US20060105676A1
Application number: US10/904,586
Authority: US
Inventors: Eugene O'Sullivan; Shom Ponoth; Wei-Tsu Tseng
Original assignee: International Business Machines Corp
Current assignee: International Business Machines Corp
Priority date: 2004-11-17
Filing date: 2004-11-17
Publication date: 2006-05-18
Also published as: CN1792552A

Abstract

A signal processing system has the detected mechanical, chemical, optical, electrical, or thermal signals generated during chemical-mechanical polishing (CMP) process collected, analyzed and differentiated with respect to time in-situ, in order to reveal the different stages during CMP for process control and end-pointing purposes. This control and/or end-pointing scheme may be used to detect the interface between two material layers sharing similar properties such as those of low-k dielectric stacks for semiconductor applications.

Description

FIELD OF THE INVENTION

The field of the invention is chemical-mechanical polishing of material, in particular of integrated circuit wafers.

BACKGROUND OF THE INVENTION

Since its advent, Chemical-mechanical polishing or planarization (CMP) has become an enabling process technology for IC manufacturing. Implementation of CMP into manufacturing environment calls for the capability to monitor the process and end-point (EP) it, before it erodes into the metal or dielectric underneath the layer that is mean to be removed.
(CMP) processes are used for the manufacturing of high-speed microprocessors, application specific integrated circuits (ASICs), Microelectromechanical Systems (MEMS), and other IC's or MEMS devices. The invention presents a set of simple and precise algorithm to process the raw physical signals emitting from the events of polishing into distinguishable and repeatable high-resolution symbols that provide end-point control for CMP process.
To achieve this, most of the modern CMP tools are equipped with a sensor to detect the thermal (e.g., U.S. Pat. No. 5,196,353), frictional (e.g., U.S. Pat. No. 5,069,002), optical (e.g., U.S. Pat. No. 5,433,651), vibrational (e.g., U.S. Pat. No. 5,222,329), electrochemical (e.g., U.S. Pat. No. 5,637,185), or electrical (e.g., U.S. Pat. No. 6,072,313) signals emanating from the pad or wafer during polishing. Changes in the magnitude of these signals (referred to as discriminants because they are used to discriminate the correct stopping time) suggest the crossover from one material layer to the other, and hence can be interpreted as reaching the interface for the desired end-point. Quite often, however, the signals emitting from the interface is weak, ambiguous, and indistinguishable so that the end-point is overlooked. Detecting such end-point signals is even more challenging when the material layers of interests share similar thermal, optical, and mechanical properties.
In accordance with the invention, a sensor is attached to component(s) of a CMP tool, in order to detect changes in physical or chemical signals during polishing process. Such signals are collected, amplified, and transported to a controller or computer for further processing. Processed signals with a “cut-off” criterion for identifying the end-point are then fed back to the polisher, in order to terminate the polish process. The signals to be detected and collected can be those from the temperature change during polishing, as described in U.S. Pat. No. 5,196,353 entitled Method for Controlling a Semiconductor (CMP) Process by Measuring a Surface Temperature and Developing a Thermal Image of the Wafer of Gurtej. S. Sandhu et al., which is assigned to Micron Technology, Inc; or as in a series of U.S. Pat. Nos. 5,597,442, 5,643,050, and 5,647,952 of Chen et al. on detecting pad temperature change, which is assigned to Industrial Technology Research Institute of Hsinchu, Taiwan. These signals can also be electrical, as described in U.S. Pat. No. 5,337,015 entitled In-situ Endpoint Detection Method and Apparatus for Chemical-mechanical Polishing Using Low Amplitude Input Voltage of Naftali E. Lustig et al., which is assigned to International Business Machines corporation; or optical (e.g., reflectivity from wafer surface layer), as described in U.S. Pat. No. 6,159,073 entitled Method and Apparatus for Measuring Substrate Layer Thickness during Chemical Mechanical Polishing of Andreas N. Wiswesser of Applied Materials, Inc; or platen/carrier torque change (mechanical) as described in U.S. Pat. No. 5,036,015 entitled Method of Endpoint Detection during Chemical/mechanical Planarization of Semiconductor Wafersof Gurtej. S. Sandhu et al. of Micron Technology, Inc; or vibrational/acoustical as described in U.S. Pat. No. 5,222,329 entitled Acoustical Method and System for Detecting and Controlling Chemical-mechanical Polishing (CMP) depths into Layers of Conductors, Semiconductors, and Dielectric Material of Chris C. Yu of Micron Technologies, Inc.
Although mathematicians are aware that the first derivative of a curve identifies changes in slope, all commercially available CMP tools of which the inventors are aware use direct signals, not derivatives.

SUMMARY OF THE INVENTION

A feature of the invention is the provision of an algorithm that yields robust, easily distinguishable signals to end-point the CMP process for improved process control and stability.
A feature of the invention is that the signals are collected and transported to a computer or signal processor where their intensity (magnitude) is differentiated with respect to time in-situ, according to the following equation: dl/dt=ε
where l represents the intensity of the detected signals such as temperature, motor current (proportional to torque), acoustics, reflectivity, interference, impedance, electrical current (e.g., eddy current), or capacitance; t is the time in seconds; and ε is the incremental change of the intensity with time.
A feature of the invention is that when a desired end-point signal, ε_epis reached, the computer sends a command to the polisher to stop the process.
Another feature of the invention is that, when an interface of interest, ε_inis reached, the computer sends a command to the polisher to continue the polishing process for certain duration of time before stopping it (overpolishing), in order to meet the designed thickness specification.
Yet another feature of the invention is that the collected signals can be further processed to reveal more of the physical events during polishing, in order to assist the detection of desired interface. For example, second derivative of the signals with time, d²l/dt²can be generated and monitored in conjunction with dl/dt for end-pointing purpose.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a schematic drawing illustrating the cross-sectional view of a structure being chemical-mechanically polished.
FIG. 2 provides a schematic view of the hardware set-up for the signal collection and processing scheme described in this invention.
FIG. 3 is a temperature-time trace collected during CMP of a 200 mm Si-based semiconductor wafer. Referring to the features in FIG. 1, in this case, Layer 1 is F-TEOS ILD. Layer 2 and Layer 3 are not present.
FIG. 4 is the trace of first derivative of temperature with respect to time, reproduced from FIG. 3.
FIG. 5 is the temperature-time trace collected during CMP of another 200 mm semiconductor wafer. Referring to the features in FIG. 1, in this case, Layer 3 is 350 Å of PECVD SiN_x, Layer 2 is 700 Å of a PECVD-based Si_wC_xO_yH_zCMP stop layer, and Layer 1 is a low-k dielectric material. Traces are collected on different CMP tools of the same type and, despite some observed curvature change on these traces, they appear ambiguous and can not be clearly identified as the interface between SiN_xand Si_wC_xO_yH_zstop layer. In addition, there is much variation between different tools.
FIG. 6 is the trace of first derivative of temperature with respect to time, reproduced from FIG. 5.

DESCRIPTION OF THE PREFERRED EMBODIMENTS

Referring to FIG. 1, a semiconductor structure has a first dielectric layer 1, and a hardmask/CMP polish stop layers 2 and 3. The interface between Layer 3 and Layer 2 is the interface to be detected or end-pointed. The fill material can be a metal or dielectric, depending on the application.
The invention is directed to the detection of CMP end-point during the polishing of chemically similar material layers comprising, for example, a single damascene or dual damascene thickness of a first dielectric (or multilayer dielectric stack including an embedded etch stop), and multiple spin-on or plasma-enhanced chemically vapor deposited (PECVD) CMP stop layer(s) (or “cap layers”) upon the first dielectric layer. The first dielectric layer may be comprised of SiLK™, GX-3™, porous SiLK™, GX-3p™, Black Diamond™, NCS™, or other non-porous or porous low k dielectric materials. The CMP stop layers may be comprised of one or combination of the following: TEOS-oxide, silane-oxide, SiN_x, BLok™, N-BLok™, PECVD-based Si_wC_xO_yH_zdielectric materials, AP 6000™, HOSP™, HOSP BESt™, Ensemble™ Etch Stop, Ensemble™ Hard Mask, hydrido silsesquioxanes, hydrido-organo silsesquioxanes copolymers, siloxanes, silsesquioxanes, or other spin-on or CVD material.
The substrate may contain electronic devices such as, for example, transistors and an array of conductor elements. A shallow-trench isolation structure composed of fill oxide, liner oxide, and SiN_xCMP stop layer, a tungsten plug structure employing TiN/Ti liner layers, or an interconnect structure composed of low-k dielectrics and Cu wires, in accordance with the invention can be formed on the substrate. Conventional CMP end-point algorithms would require the collection of the signals emitting from the interface as the structure is polished down from layer 3 to layer 2.
FIG. 2 shows a schematic view of a CMP system according to the invention. Block 100 represents a CMP tool. Four possible sensor detecting positions on the CMP tool are illustrated here, 1) the sensor is embedded in the carrier, 2) the sensor monitors a fixed spot on the pad, 3) the sensor is embedded in the platen or pad, or the sensor uses parameters of the motor, such as the current drawn by the motor. The signals are collected and transported to a computer or signal processor 150 for further analysis.
The first (or higher) derivative is calculated. A plot of ε vs. t is put forth instantaneously for monitoring purposes. Distinct peaks on this ε vs. t plot flag the times when a change in the intensity of the above signals takes place and hence corresponds to the change in the materials properties across the interface between two layers.
Referring to FIG. 3 and FIG. 4 or FIG. 5 and FIG. 6, in accordance with the invention, much improved clarity, fidelity, and reliability of the signals can be obtained after the raw temperature trace is processed and differentiated with respect to time.
FIG. 3 shows two classic traces of temperature vs. time. They were generated under two different sets of down force (DF) and back pressure (BP) in psi, during polishing. The start of liner (4) polishing shows up clearly, but the end point is not at all clear. The lower curve has a well-defined knee, but the upper curve is not well defined.
In this case, the structure being polished contains only one dielectric layer, Layer 1, which is PECVD fluorine-doped TEOS (F-TEOS) oxide. Layers 3 and 2 are not present in this example. The interface to be detected is thus liner/F-TEOS. The temperature monitored on the pad shows no obvious feature that can be identified as reaching the end of liner polishing for end-pointing purpose.
FIG. 4 showing the first derivative of temperature show the line end-point with much greater clarity. In this case, the quantity of dT/dt represents the instantaneous temperature change rate during polishing. Once the liner polish begins, the dT/dt trace quickly reaches a peak (point 1), after which it descent into a valley (point 2), which corresponds to the point where liner has been partially removed and the underlying layer 1 starts to be exposed. The trace resurges again and reaches the second peak (point 3) before it finally descends into point 4, where the temperature change rate, dT/dt maintains at zero afterwards, corresponding to the stage of single layer (layer 1 ) polishing. In the particular example illustrated—that of the interface between liner (4) and layer 1, that interface will be located somewhere between point 2, the end of the liner polish and point 4, the start of ILD 1 polish. To ensure complete removal of the liner, point 4 is the end-point of interests for practical application. Since the dT/dt trace remains flat after point 4, the end-pointing criteria to catch point 4 can be defined as one that the temperature change rate remains at zero for certain duration of time and that the second derivative of temperature with time stays below a certain finite (near-zero) value:
|dT/dt|≦m and |d ² T/dt ² |≦n within Δt=10 sec Eq. [1]
where m is a cut-off value (e.g., 0.5 in this case) for |dT/dt|; n is the cut-off value (e.g., 0.05 in this case) for |d²T/dt²|and Δt is the detection time window.
According to the invention, an end-point value is specified based on empirical data and the polishing is stopped when that value is reached.
The dT/dt vs. t traces on FIGS. 4 and 6 reveal characteristic peaks that can be identified as the desired interface for end-point unambiguously, as described in more detail below. The thickness of Layer 2 can be controlled and tailored by the detection of the interface from the characteristic peaks, in order to meet the desired device performance requirements.
FIGS. 5 and 6 present a more challenging situation. In the same structure as in FIG. 1, a set of 6 curves were recorded. In this case, layers 3 and 2 are present and the interface that it is desired to locate is the layer 3-layer 2 interface.
The temperature curve is shown in FIG. 5, showing that four of the curves are clustered together, with two “outliers’ that deviate significantly—a common occurrence.
It will readily be seen that this situation is not well suited to the use of the temperature as a discriminant, since there is no knee in the curves that is readily visible.
FIG. 6 show the first derivative of temperature for some of the curve of FIG. 5, showing much more structure. Similar to FIG. 4, during liner polish, the dT/dt trace reaches the 1^stpeak before it descends into the valley, which corresponds to partial removal of liner and exposure of the dielectric underneath (layer 3, SiN_xin this case). After resurging to the second peak (point 3), it quickly drops to the plateau where SiN_xpolishing begins (point 4). The trace climbs up again at last, as the Si_wC_xO_yH_zCMP stop layer (layer 2) is finally exposed. Depending upon the device performance requirements, the end-point to be detected can be set as the interface between liner and SiN_x, i.e., point 4 on the dT/dt trace, whereby the detection criteria can be defined as:
|dT/dt|≦u and |d ² T/dt ² |≦v within Δt=10 sec Eq. [2]
where u is the cut-off for |dT/dt| (e.g., can be 1.5 in this case); v is the cut-off value for |d²T/dt²| (e.g., can be 0.3 in this case) and Δt is the detection time window.
In another case, the interface between SiN_xand Si_wC_xO_yH_zstop layer (e.g., layer 3 and layer 2, respectively, based on FIG. 1) can be the desirable interface (end-point) to be detected. In this case, point 5 would be the end-point. This point can be reached by turning on the criteria to catch SiN_xas in Eq. [2] above first, followed by another criterion which identifies that the |d²T/dt²| rises above certain value (0.3 in this case) after a given amount of time (30 sec) since point 4 is first detected.
While the invention has been described in terms of a single preferred embodiment, those skilled in the art will recognize that the invention can be practiced in various versions within the spirit and scope of the following claims.

Claims

1. A system for identifying a change in a parameter of a workpiece comprising:

A sensor for sensing a discriminant of said parameter;

A signal processing unit connected to said sensor and capable of performing real-time signal processing including calculating at least a first derivative with time of said discriminant; and

A comparison module within said signal processing unit for comparing said at least first order derivative with a criterion.

2. A system according to claim 1, in which said signal processing unit calculates a second derivative with time of said discriminant.

3. A system according to claim 1, in which said parameter is the thickness of a film deposited on a surface of said workpiece.

4. A system according to claim 1, in which said sensor is connected to a chemical-mechanical polishing system.

5. A system according to claim 4, in which said chemical-mechanical polishing system is adapted to remove a first layer on said workpiece and to stop material removal when said first layer is removed, whereby said sensor senses an interface between said first layer and a second layer below said first layer.

6. A system according to claim 4, in which said criterion is a fixed value of the first derivative of said discriminant.

7. A system according to claim 4, in which the first dielectric layer is comprised of a material selected from the group consisting of SiLK™, GX-3™, porous SiLK™, GX-3p™, JSR LKD 5109™, JSR LKD 5130™, Black Diamond™, NCS™, porous spin-on or PECVD Si_wC_xO_yH_zor other low k or porous low k dielectric materials.

8. A system according to claim 7, in which said PECVD or spin-on CMP protective layer(s) is comprised of a material selected from the group consisting of TEOS-oxide, silane-oxide, SiN_x, BLok™, N-BLok™, PECVD-based Si_wC_xO_yH_zdielectric materials, AP 6000™, HOSP™, HOSP BESt™, Ensemble™ Etch Stop, Ensemble™ Hard Mask, hydrido silsesquioxanes, hydrido-organo silsesquioxanes copolymers, siloxanes, silsesquioxanes, or other spin-on or CVD material, or combination of the above.

9. A system according to claim 4, in which said criterion is a magnitude of said first derivative less than a first reference value for a detection period of time and a magnitude of a second order derivative with time below a second reference value during said detection time.

10. A method for identifying a change in a parameter of a workpiece comprising:

providing a sensor for sensing a discriminant of said parameter;

providing a signal processing unit connected to said sensor and capable of performing real-time signal processing including calculating at least a first derivative with time of said discriminant; and

A comparison module within said signal processing unit for comparing said at least first order derivative with a criterion, comprising the steps of:

Calculating said first order derivative with time of said discriminant;

Comparing a current value of said first order derivative with said criterion; and

Generating an output signal when said criterion is met.

11. A method according to claim 10, in which said signal processing unit calculates a second derivative with time of said discriminant.

12. A method according to claim 10, in which said parameter is the thickness of a film deposited on a surface of said workpiece.

13. A method according to claim 10, in which said sensor is connected to a chemical-mechanical polishing system.

14. A method according to claim 13, in which said chemical-mechanical polishing system is adapted to remove a first layer on said workpiece and to stop material removal when said first layer is removed, whereby said sensor senses an interface between said first layer and a second layer below said first layer.

15. A method according to claim 13, in which said criterion is a fixed value of the first derivative of said discriminant.

16. A method according to claim 14, in which said criterion is a fixed value of the first derivative of said discriminant.

17. A method according to claim 13, in which said criterion is a magnitude of said first derivative less than a first reference value for a detection period of time and a magnitude of a order derivative with time below a second reference value during said detection time.

18. A method according to claim 14, in which said criterion is a magnitude of said first derivative less than a first reference value for a detection period of time and a magnitude of a order derivative with time below a second reference value during said detection time.

19. A method according to claim 13, in which the first dielectric layer is comprised of a material selected from the group consisting of SiLK™, GX-3™, porous SiLK™, GX-3p™, JSR LKD 5109™, JSR LKD 5130™, Black Diamond™, NCS™, porous spin-on or PECVD Si_wC_xO_yH_zor other low k or porous low k dielectric materials.

20. A method according to claim 19, in which said PECVD or spin-on CMP protective layer(s) is comprised of a material selected from the group consisting of TEOS-oxide, silane-oxide, SiN_x, BLok™, N-BLok™, PECVD-based Si_wC_xO_yH_zdielectric materials, AP 6000™, HOSP™, HOSP BESt™, Ensemble™ Etch Stop, Ensemble™ Hard Mask, hydrido silsesquioxanes, hydrido-organo silsesquioxanes copolymers, siloxanes, silsesquioxanes, or other spin-on or CVD material, or combination of the above.