Optimizing the Chromium Dry Etch Process
Guenther Ruhl, Ralf Dietrich and Ralf Ludwig Infineon Technologies AG, Munich, Germany Norbert Falk Applied Materials, Ismaning, Germany Troy Morrison Surface/Interface Inc., Sunnyvale, Calif. Brigitte Stoehr Applied Materials, Sunnyvale, Calif. -- Semiconductor International, 7/1/2001
| At a Glance |  | ||
|
| ||
| |||
The minimum feature size requirements are outpacing the feature size reductions of each technology generation because of the increasing use of optical proximity correction (OPC) that requires small non-printable mask features.1 The size of the smallest features that can be fabricated on the mask is determined by two factors: the resolution of the exposure system, and the CD loss caused by the mask patterning process. The latter consists of the photoresist exposure/development and the chromium (Cr) etch processes.
In recent years, a major improvement was achieved by replacing the isotropic wet Cr etch process by anisotropic dry plasma etch. However, the use of oxygen-containing etch chemistry erodes the photoresist slightly, reducing linewidths (and widening spaces).
Mask metrology
The difference between the as-drawn CD and the final CD on the mask is called the CD bias. A key etch process development goal is to minimize the CD etch bias. To this end, it is important to determine the amount of etch bias contributed by the dry etch process. A straightforward method is to measure the difference between the CD in the photoresist and the final CD on the mask Cr.
However, the established methods — such as confocal optical microscopy and scanning electron microscopy — have the drawback that the measured CD value strongly depends on the sidewall shape of the measured structures. Calibrated CD results can only be obtained from structures with a sidewall shape identical to the calibration standard. In the real mask-making world, especially during process development, this often is not the case. Thus, another CD measurement method is needed that also images the sidewalls of the measured structures. One recently developed method is surface nanoprofiling (see "Surface Nanoprofiling").
Experimental
For this work, we used 6 × 6 in., 0.250 in.-thick binary chrome-on-glass blanks with a 100 nm thick Cr/CrOx layer and coated with 570 nm TOK iP3600 resist. The exposures were done on an Etec ALTA 3500 laser beam system. The dry etch experiments were performed on an Applied Materials plasma etch system. The etch process used an inductively coupled plasma (ICP) and chlorine/oxygen/helium etch chemistry. An oxygen descum etch was done prior to the Cr etch. Resist thickness measurements were done with a Nanometrics NanoSpec 6100.
CD measurements on both resist and the Cr mask were performed with a Surface/Interface SNP9000 Stylus NanoProfilometer (SNP) using an ~800 nm-long probe tip with a rectangular cross section 250 nm wide. The sidewall angles of the tip were ~88°. The pixel width of the measurement was 5 nm. To minimize the influence of tip variation of different tips, the tip was not changed during the experiments. The width, corner rounding and sidewall angles of the tips were first determined by a characterization procedure (see "Surface Nanoprofiling," p. 244). The tip was characterized before the resist measurements, before the Cr measurements, and after the Cr measurements. Resist measurements were made with a lower contact force setting than for the Cr to avoid damaging the resist or staining the probe tip. The initial resist CD measurements were done after an oxygen plasma descum; final CD measurements were done on the Cr pattern after resist stripping. CDs were calculated from SNP measurements of 50 nm-high structures. Optical CD measurements were also made with a Leica LWM 250 i-line transmission CD microscope.
We used a special test pattern with varying local pattern density consisting of 18 × 18 fields covering a 110 × 110 mm area. Pattern load (areal density) was 50%, which yields a total Cr load of 26%. The single fields contained clear and dark lines with varying linewidth and pitch. We used data from CD measurements of 800 nm-wide clear and dark lines, with a 1.6 µm pitch for dense structures and 75 µm pitch for isolated structures.
Results
The tip shape information gained from the characterization procedure was then applied to the raw data to obtain the true CD values. Corner rounding does not allow investigating the last few nanometers at the bottom corner of a feature. The sidewalls of a feature can only be determined exactly with tips that have a steeper sidewall slope than the feature itself. To determine the real CD value for transmitted light on an optical photomask, one can extrapolate the CD value at different heights of the feature to "zero" height at the bottom of a feature. Previous investigations have shown that this is a useful method.2 This assumes that there is no variation of the sidewall slope or the "foot" at the bottom of the feature. Since the sidewall slope of the resist features are not constant on the features and only CD difference values are required in this case, this method was not applied.
To determine the CD bias, the CD value at 50 nm height of the resist feature was subtracted from the CD value at 50 nm height of the Cr. This method proved to give the best measurement reproducibility for this application. Because changing corner rounding of different tips would influence the bias values, the same tip was used for both measurements in resist and Cr. To verify the tip shape integrity, tip characterization was performed at the beginning and the end of the measurement series. The resulting data showed no significant tip wear after the measurements. However, a possible error was taken into account in the design of experiment (DoE) matrix with centerpoint replicates within the series of experiments. The CD measurement repeatability was 12 nm (3 s) in photoresist and 9 nm (3 s) in Cr structures.
CD bias optimization
To reduce the minimum feature size and improve resolution, the CD etch bias must be minimized. Cr dry etch processes use a chlorine/oxygen-based chemistry. To maintain CD uniformity, especially at low Cr loads, a minimum amount of oxygen is required. Too much oxygen leads to significant lateral etching of the photoresist, increasing the etch bias. In previous experiments, we identified the key parameters for CD etch bias as oxygen concentration in the reactive gas, the amount of an inert gas such as helium, and the total pressure. Therefore, we studied the effects of these three key parameters on CD etch bias.
| 1. A plot of SNP vs. LWM 250 CD measurements shows good linear correlation. The small variation of the data may partly be caused by sidewall shape variation and the differing number of measurement points on the mask. |
The credibility of the data was checked by plotting SNP CD measurements vs. the LWM CD measurements for Cr masks (Fig. 1). The data show a good linear correlation. The small variation of the data may partly be caused by sidewall shape variation and the differing number of measurement points on the mask (five for the SNP and 84 for the LWM 250).
CD bias results
The DoE analysis showed that CD bias decreases significantly with decreasing oxygen concentration and helium flow, and with increasing pressure. At the same time, these parameters have an inverse effect on the Cr-to-photoresist selectivity that shows a nearly linear correlation to CD etch bias caused by the resist loss that occurred during pattern formation in the dry etch process (Fig. 2). These results are similar to those previously reported for Zeon ZEP7000 e-beam resist.3 Other data showed that this behavior is caused by inverse effects of all three factors on the etch rates of both the photoresist and the Cr. As a consequence, the optimal settings for minimum CD bias are high pressure, low oxygen concentration and low helium flow.
Isolated/dense CD bias results
| 2. Cr-to-resist selectivity shows a nearly linear correlation to CD etch bias caused by the resist loss that occurred during pattern formation in the dry etch process. |
In our test mask these measurements show that, at clear structures (trenches), this iso/dense CD bias is in the range of 10-15 nm (i.e. the isolated trenches are wider than dense trenches). In contrast, dark structures (lines) show an iso/dense bias of about -80 nm. Obviously, isolated lines end up significantly smaller than dense lines.
For the characterization of the etch process it is necessary to determine whether this effect originates during the resist patterning process (exposure and development) or in the dry etch process. To analyze these contributions, repeated measurements of resist and Cr structures were made. The Table shows these data together with the CD difference due to the dry etch process. These data are the mean values measured on four identically processed masks (center point of DoE).
| CD Values for Various Local Pattern Densities in Photoresist and Cr | |||
| Structure | Photoresist CD (nm) | Chrome CD (nm) | CD etch bias (nm) |
| Clear dense | 814 | 980 | 166 |
| Clear isolated | 831 | 1002 | 171 |
| Clear difference | 17 | 22 | 5 |
| Dark dense | 781 | 619 | -162 |
| Dark isolated | 731 | 531 | -200 |
| Dark difference | -50 | -88 | -38 |
From these results it was obvious that the main contribution to the large iso/dense effects, especially on dark structures, was due to the resist patterning process. This is probably caused by proximity effects during resist exposure.
After having characterized the resist contribution, we measured the iso/dense bias on all of the DoE samples with the higher-throughput LWM 250 tool. From these values the resist contributions measured with SNP were subtracted.
The DoE data showed the following: Iso/dense bias values on clear structures were always lower than 5 nm. The variation of the data within the DoE only reflected the repeatability of the CD measurement. On dark structures, the iso/dense bias varied within a range of 18 nm. The analysis of the DoE yielded a reasonable model for the effects of all three input factors: oxygen concentration, helium flow and pressure. The behavior was similar to the effects on the overall CD etch bias. Additionally, there is a strong interaction between He flow and pressure. At low He flow, pressure has only a weak effect on increasing iso/dense bias. At high He flow, the effect is strongly decreasing iso/dense bias with increasing pressure. This obviously reflects that iso/dense bias is partly determined by the absolute CD etch bias combined with a contribution of an interaction between the flow and pressure, which is not yet fully understood.
Optimized process
| 3. Optimum Cr etch process window (yellow region) for CD etch bias, selectivity and isolated/dense bias. |
The optimal settings are high pressure, low oxygen concentration and low helium flow. These parameter settings yield a process with a CD etch bias of 107 nm at the given pattern coverage, a selectivity of 1.5, and an iso/dense bias on clear structures of 6 nm and on dark structures of 15 nm. As the steep sidewall shape (Fig. 4) indicates, there is a chance to further minimize CD etch bias by reduction of the over-etch time.
The CD uniformity in Cr is typically lower than the 12 nm (3 s) measured on a Leica LMS IPRO, but this value includes the contributions of all previous process steps. In future work we plan to analyze the pure etch contribution to CD uniformity.
Conclusions
 |  |
The CD process bias was optimized for a Cr etch process on a novel dry etch platform independently from previous process step variations, e.g. from resist exposure or development. In the same manner CD iso/dense bias could be minimized. An overlay of model functions
4. SEM cross section of the optimized process.
derived from a DoE was used to provide the necessary optimization data.
The Surface/Interface Stylus NanoProfilometer provides information about CDs in photoresist and Cr structures and is appropriate for use in process development of Cr etch processes. The throughput needs to be increased for the SNP to become the primary CD measurement tool. SNP measurements allow a realistic separation of CD contributions from the resist patterning and Cr etch processes independently from sidewall shapes, which is difficult with conventional optical or SEM measurements.
Surface Nanoprofiling
The SNP employs a new type of force sensor2 designed to deliver precise metrology and be immune to surface charging effects. It uses a balance beam to achieve a force-controlled contact with the surface being scanned. The extremely small force (a few hundred nanonewton) is sufficient to overcome the effects of charging and surface adsorbed layers. The sensor performs a DC force measurement once per scan pixel, touching the surface with a controlled frequency of ~100 Hz. This scanning technique minimizes tip wear. Tip and sensor technology are independent, permitting the sensor mechanism to remain the same while the tip technology can be modified and customized for specific applications.
Measurement of probe tip shape True CD metrology requires a complete understanding of the sample probe or tip and sensor interaction. Therefore, a detailed characterization of the tip shape is necessary. This is done by profiling two special silicon "characterizers." Information about the tip bottom is obtained by scanning over a pattern with a sawtooth-shaped "nanoedge" profile. Scanning a characterizer that has an undercut profile gives information about the sidewall angles of the probe. The characterization data are then "stitched" together to form a complete tip profile. The probe profile is then automatically extracted from the raw data to achieve the true CD profile (Fig. 2). The SNP can provide information about feature CD, z-height and sidewall angle with a single scan. REFERENCES
| |||||
Guenther Ruhl, Infineon Technologies.
Phone: +49-89-234-23171
e-mail: guenther.ruhl@infineon.com
Norbert Falk joined Applied Materials in 1989 as a field service engineer, and in 1999 became the system specialist and process engineer dedicated to photomask etch at the Infineon Mask House. He has an education in information technology.
e-mail: norbert_falk@amat.com
Brigitte Stoehr is senior technology manager for photomask etch. She joined Applied Materials in 1994 as a product marketing manager after two years as a visiting scientist at the IBM Almaden Research Center. She has an M.S. in chemistry and a Ph.D. in physical chemistry.
e-mail: brigitte_stoehr@amat.com
REFERENCES
- W. Maurer, "Mask Specifications and OPC," SPIE, Vol. 3546, 1998, p. 232.
- R. Ludwig, Diploma Work, University of Applied Science Munich, 2000.
- J. Hochmuth, G. Ruhl, T. Coleman, "Control Methodology of Off-Target for Varying Pattern Densities With Cr Dry Etch," SPIE, Vol. 3873, 1999, p. 297.
- F. Chen, et al, "Impact of the Loading Effect on CD Control in Plasma Etching of Cr Photomasks Using ZEP7000 Resist," SPIE, Vol. 3546, 1998, p. 429.
The authors thank Thomas Schaetz (Infineon Technologies Mask House) for his support in CD measurements on the LWM 250.
