ABSTRACT (100 WORDS) 1. INTRODUCTION

Transcription

1 Overlay target selection for 20-nm process on A500 LCM Vidya Ramanathan b, Lokesh Subramany a, Tal Itzkovich c, Karsten Gutjhar a, Patrick Snow a, Chanseob Cho a Lipkong ap b a GLOBALFOUNDRIES 400 Stone Break Extension, Malta, N 12020; b KLA-Tencor Corp, 1 Technology Drive, Milpitas, CA 95035; c KLA-Tencor Israel, 1 Halavyan St. Migdal Ha emek ABSTRACT (100 WORDS) Persistently shrinking design rules and increasing process complexity require tight overlay control thereby making it imperative to choose the most suitable overlay measurement technique and complementary target design. In this paper we describe an assessment of various target designs from FEOL to BEOL on 20-nm process. Both scatterometry and imaging based methodology were reviewed for several key layers on A500LCM tool, which enables the use of both technologies. Different sets of targets were carefully designed and printed, taking into consideration the process and optical properties of each layer. The optimal overlay target for a given layer was chosen based on its measurement performance. Keywords: Overlay, Target design, Accuracy, Litho Metrology, TMU, CDSEM overlay, SCOL, 1. INTRODUCTION Persistently shrinking design rules and increasing process complexity require tight overlay (OVL) control thereby making it imperative to choose the most suitable overlay measurement technique and complementary target design. ITRS specifies the overlay metrology budget for the 14nm node as 3.6 nm [1]. Advancements in overlay metrology systems such as the KLA-Tencor s, A500LCM tool enable the user to choose between image based and scatterometry based targets. The A500LCM is an all-inclusive metrology solution combining both imaging and scatterometry to enable the flexibility needed for 1 node overlay metrology. The technology behind the tool allows the user to choose judiciously between image based (IBO) or scatterometry based overlay technology (SCOL), based on the resist profile of the litho-layer and its underlying stack properties, and overlay budget requirements for that layer. These 2 independent optical overlay methods may exhibit different performance for each layer. This arises primarily due to the fundamental technical differences between the two methodologies and the design of the target. In this paper we explore these two methodologies for several layers with different stacks and thereby optical properties. In this paper we describe an assessment of various target designs from FEOL to BEOL on 20-nm process. Both scatterometry and imaging measurement techniques were reviewed for several key layers. Multiple sets of SCOL and IBO targets were carefully designed and printed while taking into consideration the process and optical properties of each layer. IBO targets primarily included µm (Advanced Imaging Metrology). Scatterometry targets include 4-cell SCOL targets, with a unit cell size of 10 µm. Once targets were printed, performance comparison for each layer was done between the different target types: SCOL 4 cell targets, measured by scatterometry, and and BLOSSOM targets measured by the imaging technique. More details about the SCOL technique can be found in reference 2. The optimal overlay target for a given layer was determined based on its performance on the tool and correlation to reference Blossom or large imaging (24 24 µm) targets, which were used as overlay targets in production. Where applicable, a comparison to CDSEM measured overlay was made with both (16 16 µm) and SCOL targets. Target sensitivity to overlay was explored by exposing certain layers with a DOE rotation or expansion offset. For some layers, scaling of targets to etch process was studied as well. Correlations to IBO or scatterometry based overlay were made with CDSEM reference after development (ADI) and after etch (AEI). In the next section we outline the details of the experimental results of evaluating (16 16 µm) and SCOL target performances on 8 different layers. Section # 3 details the experimental results of CDSEM correlation after the etch process on a BEOL, double-patterning layer. Metrology, Inspection, and Process Control for Microlithography I, edited by Jason P. Cain, Martha I. Sanchez, Proc. of SPIE Vol. 9424, SPIE CCC code: /15/$18 doi: / Proc. of SPIE Vol

2 2. TARGET EVALUATION A total of 8 layers from a 20-nm process, which spanned from the FEOL to the BEOL were studied for this project. For each target the optimal measurement settings was selected based on Total Measurement Uncertainty (TMU) performance and OVL performance (raw OVL values and modeled linear OVL residuals). TMU includes the TIS 3Sigma and precision calculated from 10 dynamic iterations. Matching was excluded from the calculations as all measurements were performed on a single A500 LCM tool. The TMU is calculated as follows - 3 Production sampling of 13 sites per field and a total of 13 fields was used to calculate TMU for each layer. Figure (1) describes the TIS Mean, TIS 3S, Precision and TMU for SCOL 1 st order targets. The bold horizontal bars in the plot indicate the pre-determined performance spec on the tool. (a) 115 Mean (nm) (b) (nn) 7,.J11,_, i s s 7 litkiti (c) Precision (nm) (d) 'MU (nm),1b, Fig.1. Performance specification for and axes for SCOL 1 st order targets on 7 different layers - (a) TIS Mean, (b) TIS 3S, (c) Precision and (d) TMU. The bars in the plots indicate the specified performance criteria. Figure (1) indicates that out of the 8 layers evaluated, SCOL 1 st order targets perform well within the specified performance criteria. In figure (2) we plot TIS mean, 3Sigma, precision and TMU values for µm targets on these layers. The bars in the plot indicate the performance criteria for the targets on the A500 tool. From the plots it is evident that for most layers, the targets performance is well under the defined success criteria. For layers 2 and 8 the respective and SCOL target performance exceeded the defined limit. As we only had a single wafer per layer we were not able to determine if the reason for low performance was a target design issue or a wafer issue. Wherever applicable comparison has also been made to a reference target that is currently used in production (POR). Raw overlay values for the available targets on these layers were modeled in KT Analyzer using a linear model that contained 6 wafer terms and 4 field terms. The scope of this comparison is to analyze the behavior of the residuals and the correctable terms for each target type and to determine the optimum target for each layer. Figure 3 (a) indicates the raw overlay values (mean +3S) for each of the available target on these layers and figure 3 (b) is the corresponding linear residuals (mean +3S). Figure 3 indicates that the raw overlay and linear residual values agree very well with both and SCOL targets for most layers and also compare well with the existing POR targets which are currently the large marks (24 24 µm in size) and the BLO marks. For layer # 2, while SCOL marks measure lower overlay values, the residuals for both target types are very similar with residual values slightly higher for SCOL marks along on the -axis. For layer # 8, 16 Proc. of SPIE Vol

3 measures higher overlay values and also slightly higher residuals. Given the option, SCOL marks would be the most (a) TIS Mean (nm) 3 4 SI 7 a r. Precision (m). (c) (d) TI5 35 (nm) (b) rb. ItILr S 6 7 B TMU (nm) suitable target type for this layer s 6 e Fig.2. Performance specification for and axes for µm targets on 7 different layers - (a) TIS Mean, (b) TIS 3S, (c) Precision and (d) TMU. The bars in the plots indicate the specified performance criteria. Raw Overlay (nm) - (b) uuuuuuriru.. ono m7r BLO - SCOL SCOL I SCOL SCOL SCOL SCOL SCOL 1 BLO - BLO - SCOL POR POR 1616 POR POR & Residuals (nm) r rrrmr mr- p u BIO- SCOL POR SCOL 1616 SCOL 1616 SCOL SCOL ISCOL POR SCOL 1 BLO - BLO - SCOL POR 1616 POR & Fig.3. (a) Raw overlay values (mean +3S) for each target and (b) corresponding linear residuals (mean +3S) calculated using KT-Analyzer, on layers 1 to 8. Proc. of SPIE Vol

4 3. CDSEM CORRELATION In order to ascertain the accuracy of these newly designed target types, we were able to compare one of the BEOL layers to a reference CDSEM target upon etching the layer [3]. This particular layer did not have any SCOL marks on them but only the targets. CDSEM targets were scanned using 2 different scan directions. Very good correlation is observed for and CDSEM marks after separating the targets based on location and scan direction. Figure 4 (a) indicates an R 2 correlation of value of 0.99 for the -axis and between 0.94 for the -axis. y= x R'=0.99 OVL - (nm) 15 - y = 615x R'=0.94 OVL-(nm) U é _ a U - (b) Fig.4. Correlation of CDSEM vs. target raw overlay values for one BEOL layer (a) -axis and (b) -axis. As per figure 5 a, raw overlay values (mean +3S) and residuals values agree very well between the and CDSEM targets on this layer. On modelling the raw overlay values with a linear ploy nominal in KT-Analyzer, a comparison of the wafer term correctables yields excellent matching between the and the CDSEM marks (figure 5b), indicating that the mark is an accurate choice for this particular layer. (a) OVL and Residual (M +3S) nm OVL (M+3S) x OVL (M+3S) RES RES CDSEM (b) CDSEM Correctables Wafer Terms Tran (nm) N_Ortho (nm) Fig.5. Comparing correctables for and CDSEM marks for one BEOL layer after the etch process. Proc. of SPIE Vol

5 (a) OVL (M+3S) nm (b) Residuals (M+3S) nm r 1 I CDSEM 1616 SCOL 4. PROGRAMED OVERLA OFFSETS One of the ways to evaluate overlay targets for a certain layer is to observe their performance with a pre-programmed overlay offset after exposure at the litho step (ADI after develop inspection) and upon etch (AEI after etch inspection). We designed an experiment (DOE), wherein we chose one of the BEOL, double patterning layer that had a variety of different targets, 16 16, SCOL and also CDSEM overlay targets. The wafer was exposed with a programmed maximum rotation offset of about 30 nm. The experiment was conducted in 2 phases, in the first phase, overlay data was measured and analyzed at the ADI step. Measurements and analysis were repeated for the 2 nd phase at the AEI step. 4.1 Overlay target performance at ADI Full wafer map data was collected on all the 3 targets types ( and SCOL targets were measured on the A500 LCM). Each target had a total of 13 points per field. In figure 6a we plot the normalized raw overlay values (mean + 3sigma) as measured by these targets and figure 6b provides the corresponding normalized residuals when the raw data is modelled by a linear polynomial in KT Analyzer. We observe that while the CDSEM target measures just about 30% of the raw overlay values as compared to the other targets, and SCOL marks are in agreement over the raw data with values that correspond well to the 30-nm programmed rotation offset. Between the and SCOL marks, the SCOL targets measure about 1-2 nm lower residuals in the and axis respectively. Residuals for CDSEM targets are lower still (about 40% lower than marks) but this is a consequence of the much lower overlay values reported by these targets t CDSEM 1616 SCOL x v Fig.6. (a) Normalized raw overlay values and (b) normalized residuals for CDSEM, and SCOL marks on a DOE wafer exposed with a programmed rotation offset at the litho. We also observed that the rotation correctable for the wafer terms as reported by the CDSEM target is much lower than the rotation correctable for and SCOL targets. Based on the analysis of the phase 1 of the experiment, it is safe to conclude that CDSEM target is unreliable for this layer and does not serve as a reference metrology at litho. Also the most suitable target for this layer would be the SCOL mark. 4.2 Overlay target performance at AEI The same set of measurements on this wafer were repeated after the etch process. Upon etch, the CDSEM measured overlay values compare much better to the and the SCOL targets, as evident in figure 7 (a). In terms of residual values the marks outperform both CDESM and SCOL marks (figure 7b). Proc. of SPIE Vol

6 1.2 (a) OVL (M+3S) nm (b) Residuals (M +35) nm i II CDSEM SCOL CDSEM SCOL Fig.7. (a) Normalized raw overlay values and (b) normalized residuals for CDSEM, and SCOL marks on a DOE wafer exposed with a programmed rotation offset after the etch process. Overlay bias between ADI and AEI occurs due to the distribution of the stress to thin film layers, micro loading effect at the pattern etching and so on [3]. The KT Analyzer lets us compare this bias in the ADI and AEI overlay values for each target. The bias values are simply a process correction, where the after etch data (AEI) is subtracted from its corresponding ADI values. In figure 8, we represent the bias signature for each of these targets and in figure 9 we quantify the bias for each target type. There is no significant scaling of overlay values for the and SCOL marks, whereas a large rotation signature is observed for CDSEM marks. This large bias for CDSEM target is the result of the underestimation of the overlay values at the litho step. The mean + 3sigma of the bias values for and SCOL marks is about 20% and 40% of the bias value of the CDSEM target, respectively. This indicates that both and SCOL marks are suitable overlay marks for this layer. Due to the small overlay bias between ADI and AEI, the marks can be considered as dedicated overlay target for this layer at the AEI step. One of the conclusions that can be drawn from this experiment is that CDSEM measured overlay cannot be used as a reliable reference metrology at the ADI step. CDSEM cm SCOL Fig.8. Process corrected overlay bias signature (ADI AEI) for CDSEM, and SCOL targets. In figure 10 represents the correlation between overlay measured using targets with the values obtained using CDSEM. We observe a reasonable correlation with an R 2 of 1 for the -axis and 8 for the -axis, after excluding the edge dies. One of our hypotheses for the low R 2 values on this layer, compared to the layer described in section 3, could be the large rotation offset on the wafer that could induce some amount of asymmetry to the overlay targets. The extreme value of the rotation offset occurs on the edges of the wafer which are the outliers in the correlation plot (figure 10). Proc. of SPIE Vol

7 Bias Signature ADI - AEI (M +3S) nm 1.2 i CDSEM SCOL Fig.9. Quantifying the process corrected overlay bias signature (ADI AEI) for CDSEM, and SCOL targets, mean + 3sigma values for the wafer maps in figure 8. Axis Axis Z y=1.0038x R.= 165 y =1.0261x R'= î Fig.10. Correlation of CDSEM vs. target raw overlay values for BEOL layer; (a) -axis and (b) -axis for DOE wafer with ~ 30 nm rotation offset 5. CONCLUSION In this work we demonstrated the evaluation of overlay targets that use several different technologies for a 20-nm process with layers that extended from the FEOL to BEOL. The most suitable overlay target for each layer can be selected based on the measurement uncertainty values for each target type. We also demonstrated target accuracy based on comparison to CDSEM reference metrology at the AEI step. Our DOE experiment on a BEOL double patterning layer indicates that target selection can also be based on comparing target performance at the ADI and the AEI steps. 6. ACKNOWLEDGMENTS The authors would like to thank Christian Sparka, Bill Pierson, from KLA-Tencor for extending KT-Analyzer to support overlay result files from CDSEM tools and iaoxiao Zhang, from Globalfoundries, for her support in providing the CDSEM overlay measurements for this project. REFERENCES [1] International Technology Roadmap for semiconductors ( [2] Philippe Leray ; Shaunee Cheng ; Daniel Kandel ; Michael Adel ; Anat Marchelli; Irina Vakshtein ; Mauro Vasconi and Bartlomiej Salski "Diffraction based overlay metrology: accuracy and performance on front end stack", Proc. SPIE 6922, Metrology, Inspection, and Process Control for Microlithography II, 69220O (March 22, 2008); doi: / ; [3] Osamu Inoue, Takeshi Kato, utaka Okagawa, Hiroki Kawada, In-die Overlay Metrology by using CD-SEM, Proc. SPIE 8681, Metrology, Inspection, and Process Control for Microlithography VII, (SPIE 2013); doi: / Proc. of SPIE Vol