Process monitoring and surface characterization with in-line XPS metrology
10/01/2007
Executive OVERVIEW
An X-ray photoelectron spectrometer (XPS) is used at 65nm and 45nm nodes to monitor nitrogen dose and the SiON thickness of ultra thin nitrided gate oxides as well as the thickness of amorphous silicon on top of an amorphous carbon stack. XPS is also used as a process characterization technique in the fabrication line and allows for nondestructive chemical composition and bonding state analyses along with thickness measurements. Two examples of surface characterization are the influence of a dry cleaning plasma step on a silicon surface for the source/drain implantation process, and fluorine contamination of a TiN hard mask for interconnect applications.
Advanced CMOS technologies of 65nm and below are increasingly sensitive to process variations. Additional challenges come from the introduction of new materials and thinner films. The success of advanced characterization is increasingly linked to metrology tools’ capabilities. To face these challenges, measurement techniques such as x-ray metrology have to move from offline characterization laboratories to fabrication lines. Among these techniques, XPS is taking a growing place for process monitoring and characterization.
There are different applications in which the XPS technique is used on monitor wafers in the fabrication line. Process monitoring of 65nm node ultra thin nitrided gate oxides is performed by XPS. Nitrogen dose and SiON thickness are measured using N-1s and Si-2p peaks (Fig. 1) [1, 2].
To achieve gate and STI patterning with 65nm and 45nm nodes critical dimension (CD) requirements, an amorphous carbon mask is introduced for etch processes. The complex stack of this mask includes a thin amorphous silicon layer; its thickness is measured by XPS. The advantage of the XPS technique is that its analytical depth (~100Å) is larger than the film thickness and is not affected by the amorphous state of the silicon film.
 Figure 1. a) Si-2p and b) N-1s regions of a nitrided oxide spectrum. |
XPS is also a helpful technique to characterize chemical bonds and quantify surface contamination brought by processes in the fab. For example, the efficiency of resist dry strip and wet clean processes after lithography steps and doping processes can be evaluated. Wafer surface contamination brought by front opening united pod (FOUP) conditioning is another issue, since surface fluorination can lead to crystal growth, which strongly affects yield and defectivity.
While our process development results were achieved on 65nm and 45nm monitor wafers, it is believed that on-product measurements will be required for 45nm production control. It is to this end that Crolles2 has recently installed a next-generation ReVera tool with a 35µm spot size and pattern recognition capability to enable XPS metrology on patterned wafers.
Experimental set-up
XPS is an analytical technique that is surface-sensitive to the top 100Å of a film or material. Low energy x-rays illuminate the surface of the wafer and cause the emission of photoelectrons at various energies that correspond to the different elements of the periodic table for Z≥3. The information rich spectra can be analyzed to calculate compositions, film thicknesses, and to investigate bonding states and interfaces.
XPS measurements are performed on a RVX1000 metrology tool from Revera Inc. The Mg K
 Figure 2. Hard mask stack over amorphous carbon. |
Process monitoring
SiON gate dielectric thickness and N-dose measurements. For 90nm node and below, the gate oxide nitrogen dose is a critical parameter to monitor. Nitrogen dose drifts have been electrically identified as reliability detractor parameters. Currently, XPS is one of the best adapted metrology methods to ensure nitrogen monitoring. In comparison, traditional metrology methods use indirect material properties that weakly correlate to atomic composition. The direct atomic composition measurement via XPS produces much higher sensitivity and precision for nitrogen dose, making it ideal for inline metrology. Thickness and nitrogen dose provided by XPS have been successfully correlated to electrical parameters such as EOT, CET, and Vt [3, 4]. To determine SiON thickness, XPS measures the Si-2p region (Fig. 1a), which contains photoelectrons from the silicon substrate, but also photoelectrons from the silicon oxide.
The spectral information contains the bonding state of the silicon oxide, or Si-O as the input to a complex curve-fitting algorithm that deconvolutes the two peaks. Due to the attenuation of the photoelectrons from the underlying silicon through the silicon oxide, the ratio of the Si-Si peak to the Si-O peak is used to calculate film thickness, as shown in Eqn. 1. For the nitrogen dose calculation, XPS measures the N-1s peak (Fig. 1b) and Si-O component of Si-2p (Fig. 1a) intensities. As shown in Eqn. 3, material constants and the XPS thickness measurement are used to determine the number of nitrogen atoms present in (atoms/cm2) units. In both Eqns. 1 and 2, λ is the Si-2p mean free path (Å), assumed as constant in SiON. ISi-Si, ISiON and IN are Si-Si, Si-O (components of Si-2p intensity) and N-1s intensities in counts-per-second (cps) and IiSi and IiSiON are constants and represent intensities of Si and SiON infinite thickness layers. ρ is SiON atomic density (at/cm3) and ASF the atomic sensitivity factors of N and Si. The same attenuation theory has been used for thickness and dose measurements.
 |
The model accuracy is limited to a homogeneous nitrogen profile over the SiON layer which may not be the case for some nitrided oxides. For gate oxide process monitoring, thickness and dose are not corrected for nitrogen profile.
For the 65nm technology node, modern XPS-based metrology systems have been proven to offer sufficient productivity and system capabilities for active inline production. Process control via XPS is being executed on monitor wafers using standard metrology test sites and protocols. Process control on product wafers is being implemented in the Crolles2 fabrication line.
Top layer silicon thickness measurements for hard mask stack. For 65nm and 45nm nodes, gate and STI patterning requires the use of a new stack for the mask of plasma etching processes. Because CDs are constantly shrinking, aspect ratios must be reduced to avoid resist pattern collapse issues and CD deviations that appear during lithography and etch processes. Figure 2 shows the hard mask stack composed of a thin amorphous silicon layer on top of a SiO2 layer, which constitutes a nonreflective stack for lithography, on top of a thicker amorphous carbon layer. Photoresist is used for Si/SiO2 hard mask opening and amorphous carbon is used for gate and STI etching. It is then possible to reduce resist thickness and thus avoid pattern collapse issues as well as CD deviations induced by high aspect ratio stack plasma etching.
XPS is an ideal technique to measure the Si top layer thickness. As Eqn. 3 shows, the calculation method used is similar to that used for SiON thickness measurements. The SiO2 layer is thicker than the analyzed depth, so it is considered the substrate.
 |
In Eqn. 3, λ is Si inelastic mean free path (assumed constant in oxide and bulk); ISi and ISiO2 are respectively Si-Si and Si-O peak intensities of Si-2p bonds. IiSi and Iisi02 are constants that represent intensities of Si and the SiO2 infinite thickness layer.
Unlike optical techniques, XPS uses intensity ratios that are neither altered by ambient molecular contamination (AMC) nor by the fact that the amorphous Si and SiO2 optical indices are similar. This XPS characteristic enabled the improvement in the optical model shown in Fig. 3. The original model (Fig. 3a) was not correlated to XPS results and two different slopes were observed. Considering XPS results as references, a new model (Fig. 3b) has been created, and it is well correlated with XPS results. Currently in the Crolles2 fab, XPS is used to correct process excursions on the entire hard mask stack.
Characterization and surface contamination monitoring
Source/drain descum characterization. The surface state of implanted layers has to be carefully controlled for contamination to prevent any implantation of contaminants in the silicon active region. Contamination can be brought by the successive steps that occur in the implant module: photolithography, implant, resist strip, and surface cleaning processes. However, cleaning processes after the lithography step can change the photoresist profiles and CD. Therefore, it is necessary to characterize the influence of each process for a complete S/D implant process flow on different materials and surfaces impacted.
 Figure 3. XPS/optical measurement comparisons a) before and b) after optical model improvement. |
An O2-based microwave downstream plasma process that is used to remove resist residues after photolithography is termed “descum.” The influence of the S/D descum process on photoresist CD and silicon active surface were analyzed. Figure 4 shows CD measurements of the space between resist lines before and after a 60-sec standard N S/D descum process on two wafers. The CD increased by about 20nm after the descum process on both wafers. Complementary analyses show that identical results have been obtained after P S/D lithography followed by a descum step. This increase of CD between resist lines is due to resist consumption by the O2 plasma. However, a 20nm CD increase is not acceptable and descum parameters have to be optimized in order to reduce resist consumption and an increase in CD. Figure 5 shows top view SEM pictures of silicon active areas after P S/D lithography; the contrast of the silicon active region is different before and after descum, indicating a modification of the silicon surface.
To understand the nature of the modification induced by the descum process on the active silicon surface, XPS is used for chemical analysis of the silicon surface. Measurements from four wafers after silicon surface cleaning, spacer liner deposition (80Å TEOS), and spacer liner wet clean (TEOS removal) were taken to simulate real process conditions on the silicon surface. XPS analyses were performed after photolithography N S/D and after descum for different descum process times: 15, 30, 45, and 60 sec. XPS results show that the species present on the silicon surface are silicon, oxygen, carbon, fluorine, and nitrogen. Atomic compositions of carbon are reported in the table.
 Figure 4. CD measurements before and after descum. |
High amounts of carbon were detected before photo N S/D for wafers C and D. This is attributed to contamination coming from the spacer wet clean process, which lacks efficiency. The table also shows that carbon atomic composition decreased by a factor of 10 after a standard descum and a 45-sec descum, and reduced to 0.9-1.3% even for highly contaminated wafers. Those results are well correlated to contrast differences observed on SEM pictures before and after descum that are due to carbon removal on the silicon surface (Fig. 5). XPS analyses have demonstrated that it is possible to decrease the descum process time from 60 sec to 15 sec while removing all carbon containing layers on the silicon active area and reducing resist consumption and thus CD deviation.
Fluorine contamination monitoring on TiN layers. For 65nm and 45nm nodes, ultra low-k (ULK) dielectric constant materials are successfully integrated by using an enhanced trench first hard mask (TFHM) backend architecture [5, 6]. In this TFHM approach, a thin metallic hard mask layer is deposited on top of a dense dielectric layer that serves to encapsulate the underlying ULK material. Post-etch cleaning issues directly relate to the metal hard mask (TiN in this study) contamination.
Time dependent metal-fluoride residues can be observed on the TiN top surface. The impact of these crystalline defects on final yield has been clearly demonstrated, especially when their growth is located on edges of patterned structures preventing a good conformity of the Cu metal deposition. Fluorine can come directly from fluorocarbonated etch plasmas, but also from a FOUP, and its combination with moisture and queue time enhances the growth of crystals.
 |
There is a real necessity to characterize this contamination as well as its origin. XPS is an attractive method to characterize and monitor fluorine contamination. The measurement technique enables light element quantification and extreme edge measurements. Furthermore, the fluoride crystalline defect signature is directly correlated to the origin of fluorine contamination. Thus, XPS can be used as an identifier of contamination, with a center signature for CF4 plasma effects, or an edge signature when fluorine is outgassed by the FOUP sides. In the same way, XPS can be used to monitor wet etch clean treatment efficiency [6].
Fluorine dose quantification, shown by Eqn. 4, is similar to that used to quantify nitrogen dose in SiON. TiN is chosen as the substrate with thickness >100Å.
 |
In these equations, ρ is the atomic density of TiN (at/cm3); IF and ITi are the F-1s and Ti-2p intensities (cps); ASFF and ASFTi are fluorine and titanium atomic sensitivity factors; and t (cm) is chosen as 4× the electron attenuation length (EAL) because TiN is thicker than the analysis depth. Fluorine is assumed to be uniformly distributed over the analysis depth. An advantage of this method is that fluorine quantification is not affected by surface organic contamination such as carbon contamination.
 Figure 6. Evolution of Ti bonding states before and after wet clean as observed by photoemission of Ti-2p. |
XPS capability to obtain bonding state information is a strong advantage for the technique. Solid lines in Fig. 6 show that Ti-Fx bonds are present on the TiN surface before the wet clean process. After wet clean, peaks corresponding to Ti-F bonds are not detected anymore and the peak at 455eV corresponding to Ti-N bond peak intensity has increased. This result indicates that a TiFx layer induced by CF4 plasma is present on top of the TiN surface. The remaining fluorine concentration corresponds to a surface contamination where fluorine atoms are not bonded to titanium.
Conclusion
We used an XPS tool in a 65nm and 45nm node fabrication line. Both process monitoring of nitrided gate oxides for production and surface characterization for process development are performed on monitor wafers. However, measurements performed on monitor wafers may not fully represent actual properties at the device level. The recent installation in Crolles2 of a next generation XPS tool with a 35µm spot size and pattern recognition capability will enable material metrology on product wafers. In return, process excursions will be detected inline, thus reducing the number of product wafers at risk. Additional productivity improvements will be seen by reducing consumption of monitor wafers and their processing as well as data acquisition for correlations with device reliability parameters.
Acknowledgments
The authors would like to thank metrology contributors, especially A. Tarnowka and D. Neira, for information on optical methods; process contributors, especially J. Bienacel, for help on SiON thickness and dose calculations; and M. Kwan, senior applications technologist at ReVera, and Emir Gurer, director of applications at ReVera, for help with XPS application development.
References
- C. Wyon, et al., “FEOL and BEOL Applications of X-Ray Metrology,” SEMICON FABTECH, 28th edition.
- J. Moulder, et al., Handbook of X-ray Photoelectron Spectroscopy, Physical Electronics Inc., 1992.
- C. Thomas Larson, et al., “Using an XPS-based Metrology System to Determine Film Thickness and Composition,” MICRO Magazine, April 2005.
- J. Borland, et al., “Meeting Challenges for Engineering the Gate Stack,” Solid State Technology, July 2005.
- R. Fox, et al., “High Performance k = 2.5 ULK Backend Solution Using an Improved TFHM Architecture, Extendible to the 45nm Technology Node,” IEDM 2005 Technical Digest.
- L. Broussous, et al., “Post-Etch Cleaning Chemistries Evaluation for Low-k Cu Integration,” Solid State Phenomena, Vol. 92, pp. 263-266, 2003.
For more information, contact Nicolas Cabuil at STMicroelectronics, Crolles, France; ph 33/4389-22789, e-mail [email protected].