With the aggressive scaling of semiconductor devices, the increasing complexity of device structure coupled with tighter metrology error budget has driven up Optical Critical Dimension (OCD) time to solution to a critical point. Machine Learning (ML), thanks to its extremely fast turnaround, has been successfully applied in OCD metrology as an alternative solution to the conventional physical modeling. However, expensive and limited reference data or labeled data set necessary for ML to learn from often leads to under- or overlearning, limiting its wide adoption. In this paper, we explore techniques that utilize process information to supplement reference data and synergizing physical modeling with ML to prevent under- or overlearning. These techniques have been demonstrated to help overcome the constraint of limited reference data with use cases in challenging OCD metrology for advanced semiconductor nodes.
We have developed a novel in-line solution for the characterization and metrology of high-aspect ratio (HAR) semiconductor structures using transmission small-angle X-ray scattering (SAXS). The solution consists of the Sirius-XCD® tool, NanoDiffract for XCD (NDX) analysis software and high-performance computing infrastructure. The solution provides quantitative information on the orientation and shape of HAR structures, such as 3D NAND channel holes and DRAM capacitors, and can be used for development and control of the critical etch processes used in the formation of such structures. The tool has been designed to minimize expensive cleanroom space without sacrificing performance with typical measurements taking only a few minutes per site. The analysis is done using real-time regression in parallel to the measurements to maximize the throughput of the solution. We will illustrate the key features of the solution using data from a HAR reference wafer and provide results for hole shape and tilt across the wafer together with complimentary data from other techniques. We will also discuss future opportunities for both stand-alone XCD applications and possibilities of XCD-OCD synergies including hybrid metrology in solving complex high-aspect ratio (HAR) and other applications.
Scatterometry performance enhancement is demonstrated through a holistic approach by utilizing comprehensive information from various sources, including data from different process steps, different toolsets, multiple structures, and multiple optical channels using samples from magnetic hard disk drive manufacturing. Parameter and spectrum feed-forward are performed across multiple targets at the photo step and the photo results are fed forward to the post-reactive ion etch (RIE) step. For an isolated structure with critical dimensions (CD) much smaller than the incident light wavelengths, feed-forward methods improve CD correlation with a general improvement of 20 to 60% in precision and fleet measurement precision (FMP). A second technique examined is hybrid metrology, where inputs from source tools, such as CD-SEM and CD-AFM, are used to determine critical parameters. Hybridization of line edge roughness results in CD and sidewall angle (SWA) FMP improvement of ∼60%. We also demonstrate improved CD accuracy using azimuthal scatterometry at 0, 45, and 90 deg azimuth angles measuring resist lines with CD larger than the incident light wavelengths. FMP reductions of ∼60 and 30% are obtained for CD and SWA. SWA hybridization after RIE results in CD and SWA FMP improvements by >50 and 30%, respectively.
Reducing parameter correlations to enhance scatterometry measurement accuracy, precision and tool matching is a crucial component of every modeling effort. Parameter sensitivity can largely depend on the orientation of the plane of incidence relative to the grating orientation. Conventional scatterometry is done with the plane if incidence normal to the grating orientation, whereas azimuthal scatterometry allows measurements at an arbitrary angle or set of angles. A second technique examined in this paper is hybrid metrology where inputs from source tools such as CD-SEM and CD-AFM are used to determine values of critical parameters. The first examples shows LER sensitivity gains by measuring narrow resist lines in an orientation parallel with the long axis of the grating. Hybridization of LER results in a CD and SWA FMP improvement of about 60%. We also showcase the benefits of azimuthal scatterometry measuring resist lines with CD larger than the wavelengths of the incident light. A CD and SWA FMP reduction of about 60% and 30% is obtained using azimuthal scatterometry at 0, 45 and 90 degrees azimuth angles. Hybridization of the ARC SWA after RIE results in CD and resist SWA FMP improvements by over 60% and 30%, respectively.
Optical critical dimension (OCD) metrology using scatterometry has been widely adopted for fast and non-destructive in-line process control and yield improvement. Recently there has been increased interest in metrology performance enhancement through a holistic approach. We investigate the benefits of feed-forward of metrology information from prior process steps using samples from magnetic hard disk drive manufacturing. The scatterometry targets are composed of rather isolated gratings that are designed to have better correlation with device features. Two gratings, one with pitch ≈ 10CD, and the other with pitch ≈ 15CD, are measured at post develop and post reactive ion etch (RIE) steps. Two methods: parameter feed-forward (PFF) and spectrum feedforward (SFF) are studied in which the measurement results or spectrum collected on the blanket target at photo step are fed forward to the measurements on the grating structures at post develop or post RIE step. Compared with standard measurement without FF, for the more isolated grating at photo step, both PFF and SFF improve CD correlation from R2=0.96 to R2=0.975 using CD-SEM results measured on device as the reference. Dynamic precision and fleet measurement precision are improved by 20-60%. For post RIE step, PFF and SFF significantly improve CD correlation from R2=0.95, slope=1.09 to R2=0.975, slope=1.03 for the denser grating, and from R2=0.90, slope=0.79 to R2=0.96, slope=0.96 for the more isolated grating. Dynamic precision is generally improved by 20-40%. It is observed that both PFF and SFF are equally efficient in reducing parameter correlation for the application studied here.
CD and shape control of extreme ultraviolet lithography (EUVL) structures is critical to ensure patterning performance at the 10 nm technology node and beyond. The optimum focus/dose control by EUV scanner is critical for CD uniformity, and the scanner depends on reliable and rapid metrology feedback to maintain control. The latest advances in scatterometry such as ellipsometry (SE), reflectometry (NISR), and Mueller matrix (MM) offers complete pattern profile, critical dimensions (CD), side-wall angles, and dimensional characterization. In this study, we will present the evaluation results of CD uniformity and focus dose sensitivity of line and space EUV structures at the limit of current ASML NXE 3100 scanner printability and complex 3D EUV structures. The results will include static and dynamic precision and CD-SEM correlation data.
Optical critical dimension (OCD) metrology using scatterometry has been demonstrated to be a viable solution for fast and non-destructive in-line process control and monitoring. As extreme ultraviolet lithography (EUVL) is more widely adopted to fabricate smaller and smaller patterns for electronic devices, scatterometry faces new challenges due to several reasons. For 14nm node and beyond, the feature size is nearly an order of magnitude smaller than the shortest wavelength used in scatterometry. In addition, thinner resist layer is used in EUVL compared with conventional lithography, which leads to reduced measurement sensitivity. Despite these difficulties, tolerance has reduced for smaller feature size. In this work we evaluate 3D capability of scatterometry for EUV process using spectroscopic ellipsometry (SE). Three types of structures, contact holes, tip-to-tip, and tip-to-edge, are studied to test CD and end-gap metrology capabilities. The wafer is processed with focus and exposure matrix. Good correlations to CD-SEM results are achieved and good dynamic precision is obtained for all the key parameters. In addition, the fit to process provides an independent method to evaluate data quality from different metrology tools such as OCD and CDSEM. We demonstrate 3D capabilities of scatterometry OCD metrology for EUVL using spectroscopic ellipsometry, which provides valuable in-line metrology for CD and end-gap control in electronic circuit fabrications.
In this work, we report the first demonstration of scatterometry Optical Critical Dimension (OCD) characterization on advanced Ge Multi-Gate Field-Effect Transistor (MuGFET) or FinFET formed on a Germanium-on-Insulator (GeOI) substrate. Two critical process steps in the Ge MuGFET process flow were investigated, i.e. after Ge Fin formation, and after TaN gate stack etching process. All key process variations in the test structures were successfully monitored by the floating or fitting parameters in the OCD models. In addition, excellent static repeatability, with 3σ lower than 0.12 nm, was also achieved. The measurement results from OCD were also compared with both Scanning Electron Microscopy (SEM) and Transmission Electron Microscopy (TEM) measurements. Excellent correlation with both SEM and TEM was achieved by employing OCD characterization, confirming scatterometry OCD as a promising metrology technique for next generation multi-gate transistor with an advanced channel material.
We evaluate diffraction-based overlay (DBO) metrology using two test wafers. The test wafers have different
film stacks designed to test the quality of DBO data under a range of film conditions. We present DBO results using
traditional empirical approach (eDBO). eDBO relies on linear response of the reflectance with respect to the overlay
displacement within a small range. It requires specially designed targets that consist of multiple pads with programmed
shifts. It offers convenience of quick recipe setup since there is no need to establish a model. We measure five DBO
targets designed with different pitches and programmed shifts. The correlations of five eDBO targets and the correlation
of eDBO to image-based overlay are excellent. The targets of 800nm and 600nm pitches have better dynamic precision
than targets of 400nm pitch, which agrees with simulated results on signal/noise ratio. 3σ of less than 0.1nm is achieved
for both wafers using the best configured targets. We further investigate the linearity assumption of eDBO algorithm.
Simulation results indicate that as the pitch of DBO targets gets smaller, the nonlinearity error, i.e., the error in the
overlay measurement results caused by deviation from ideal linear response, becomes bigger. We propose a nonlinearity
correction (NLC) by including higher order terms in the optical response. The new algorithm with NLC improves
measurement consistency for DBO targets of same pitch but different programmed shift, due to improved accuracy. The
results from targets with different pitches, however, are improved marginally, indicating the presence of other error
Diffraction-based overlay (DBO) technologies have been developed to address the overlay metrology
challenges for 22nm technology node and beyond. Most DBO technologies require specially designed targets that
consist of multiple measurement pads, which consume too much space and increase measurement time. The traditional
empirical approach (eDBO) using normal incidence spectroscopic reflectometry (NISR) relies on linear response of the
reflectance with respect to overlay displacement within a small range. It offers convenience of quick recipe setup since
there is no need to establish a model. However it requires three or four pads per direction (x or y) which adds burden to
throughput and target size. Recent advances in modeling capability and computation power enabled mDBO, which
allows overlay measurement with reduced number of pads, thus reducing measurement time and DBO target space. In
this paper we evaluate the performance of single pad mDBO measurements using two 3D targets that have different
grating shapes: squares in boxes and L-shapes in boxes. Good overlay sensitivities are observed for both targets. The
correlation to programmed shifts and image-based overlay (IBO) is excellent. Despite the difference in shapes, the
mDBO results are comparable for square and L-shape targets. The impact of process variations on overlay measurements
is studied using a focus and exposure matrix (FEM) wafer. Although the FEM wafer has larger process variations, the
correlation of mDBO results with IBO measurements is as good as the normal process wafer. We demonstrate the
feasibility of single pad DBO measurements with faster throughput and smaller target size, which is particularly
important in high volume manufacturing environment.
Resolution enhancement techniques such as double patterning (DP) processes are implemented to achieve
lower critical dimension (CD) control tolerances. However the design complications, overlay resulting
from multiple exposures, and production cost limit the DP usage. EUVL offers the most promising
patterning technology to be adopted for 14nm and beyond due to simplicity and cost advantage estimates.
However, EUVL is also prone to number of patterning challenges that are unique to EUV, such as
orientation dependent pattern placement errors resulting from mask shadowing effect, flare(leads to CD
non-uniformity) and non-flatness (leads to overlay errors). Even though the shadowing effects can be
corrected by means of OPC and mask stack design, there is a need to monitor the systemic errors due to HV
bias in order to control the lithographic process. In this paper, we will report the measurement sensitivity
of EUVL targets (CD, height and sidewall angle), systemic CD errors (H-V bias) and feedback for OPC
correction by scatterometry. We will also report the measurement precision, accuracy and matching for
As the dimensions of integrated circuit continue to shrink, diffraction based overlay (DBO) technologies have
been developed to address the tighter overlay control challenges. Previously data of high accuracy and high precision
were reported for litho-etch-litho-etch double patterning (DP) process using normal incidence spectroscopic
reflectometry on specially designed targets composed of 1D gratings in x and y directions. Two measurement methods,
empirical algorithm (eDBO) using four pads per direction (2x4 target) and modeling based algorithm (mDBO) using two
pads per direction (2x2 target) were performed. In this work, we apply DBO techniques to measure overlay errors for a
different DP process, litho-freeze-litho-etch process. We explore the possibility of further reducing number of pads in a
DBO target using mDBO. For standard targets composed of 1D gratings, we reported results for eDBO 2x4 targets,
mDBO 2x2 targets, and mDBO 2x1 target. The results of all three types of targets are comparable in terms of accuracy,
dynamic precision, and TIS. TMU (not including tool matching) is less than 0.1nm. In addition, we investigated the
possibility of measuring overlay with one single pad that contains 2D gratings. We achieved good correlation to blossom
measurements. TMU (not including tool matching) is ~ 0.2nm. To our best knowledge, this is the first time that DBO
results are reported on a single pad. eDBO allows quick recipe setup but takes more space and measurement time.
Although mDBO needs details of optical properties and modeling, it offers smaller total target size and much faster
throughput, which is important in high volume manufacturing environment.
Spacer defined double patterning processes consists of multiple deposition, post strips and etch steps and is
inherently susceptible to the cumulative effects of defects from each process step leading to higher rate of
defect detection. CD distortions and CD non-uniformity leads to DPT overlay errors. This demands
improved critical dimension uniformity (CDU) and overlay control. Scatterometry technique enables the
characterization and control the CD uniformity and provision to monitor stepper and scanner characteristics
such as focus and dose control. While CDSEM is capable of characterizing CD and sidewall angle, is not
adequate to resolve shape variations, such as footing and top rounding and spacers with leaning angles,
during the intermediate process steps. We will characterize direct low temperature oxide deposition on
resist spacer with fewer core films and reduced number of processing and metrology control steps.
Metrology characterization of SADP and resist core transferred spacers at various process steps will be
performed by scatterometry using spectroscopic ellipsometry and reflectometry. We will present CD
distribution (CDU) and profile characterization for core formation, spacer deposition and etch by advanced
optical scatterometry and also validate against CDSEM.
Scatterometry has been used extensively for the characterization of critical dimensions (CDs) and detailed sidewall profiles of periodic structures in microelectronics fabrication processes. In most cases devices are designed to be symmetric, although errors could occur during the fabrication process and result in undesired asymmetry. Conventional optical scatterometry techniques have difficulties distinguishing between left and right asymmetries. We investigate the possibility of measuring grating asymmetry with Mueller matrix spectroscopic ellipsometry (MM-SE) for a patterned hard disk sample prepared by a nanoimprint technique. The relief image on the disk sometimes has an asymmetrical sidewall profile, presumably due to the uneven separation of the template from the disk. Cross section SEM reveals that asymmetrical resist lines are typically tilted toward the outer diameter direction. Simulation and experimental data show that certain Mueller matrix elements are proportional to the direction and amplitude of profile asymmetry, providing a direct indication to the sidewall tilting. The tilting parameter can be extracted using rigorous optical critical dimension (OCD) modeling or calibration methods. We demonstrate that this technique has good sensitivity for measuring and distinguishing left and right asymmetry caused by sidewall tilting, and can therefore be used for monitoring processes for which symmetric structures are desired.
Diffraction based overlay (DBO) technologies have been developed to address the tighter overlay control
challenges as the dimensions of integrated circuit continue to shrink. Several studies published recently have
demonstrated that the performance of DBO technologies has the potential to meet the overlay metrology budget for
22nm technology node. However, several hurdles must be cleared before DBO can be used in production. One of the
major hurdles is that most DBO technologies require specially designed targets that consist of multiple measurement
pads, which consume too much space and increase measurement time. A more advanced spectroscopic ellipsometry (SE)
technology-Mueller Matrix SE (MM-SE) is developed to address the challenge. We use a double patterning sample to
demonstrate the potential of MM-SE as a DBO candidate. Sample matrix (the matrix that describes the effects of the
sample on the incident optical beam) obtained from MM-SE contains up to 16 elements. We show that the Mueller
elements from the off-diagonal 2x2 blocks respond to overlay linearly and are zero when overlay errors are absent. This
superior property enables empirical DBO (eDBO) using two pads per direction. Furthermore, the rich information in
Mueller matrix and its direct response to overlay make it feasible to extract overlay errors from only one pad per
direction using modeling approach (mDBO). We here present the Mueller overlay results using both eDBO and mDBO and compare the results with image-based overlay (IBO) and CD-SEM results. We also report the tool induced shifts (TIS) and dynamic repeatability.
The extension of optical lithography to 22nm and beyond by Double Patterning Technology is often challenged by CDU
and overlay control. With reduced overlay measurement error budgets in the sub-nm range, relying on traditional Total
Measurement Uncertainty (TMU) estimates alone is no longer sufficient. In this paper we will report scatterometry
overlay measurements data from a set of twelve test wafers, using four different target designs. The TMU of these
measurements is under 0.4nm, within the process control requirements for the 22nm node. Comparing the measurement differences between DBO targets (using empirical and model based analysis) and with image-based overlay data indicates the presence of systematic and random measurement errors that exceeds the TMU estimate.
Scatterometry has been used extensively for the characterization of critical dimensions (CD) and detailed
sidewall profiles of periodic structures in microelectronics fabrication processes. So far the majority of applications are
for symmetric gratings. In most cases devices are designed to be symmetric although errors could occur during
fabrication process and result in undesired asymmetry. The problem with conventional optical scatterometry techniques
lies in the lack of capability to distinguish between left and right asymmetries. In this work we investigate the possibility
of measuring grating asymmetry using Mueller matrix spectroscopic ellipsometry (MM-SE). A patterned hard disk
prepared by nano-imprint technique is used for the study. The relief image on the disk sometimes has asymmetrical
sidewall profile, presumably due to the uneven separation of the template from the disk. The undesired tilting resist
profile causes difficulties to the downstream processes or even makes them fail. Cross-section SEM reveals that the
asymmetrical resist lines are typically tilted towards the outer diameter direction. The simulation and experimental data
show that certain Mueller matrix elements are proportional to the direction and amplitude of profile asymmetry,
providing a direct indication to the sidewall tilting. The tilting parameter can be extracted using rigorous optical critical
dimension (OCD) modeling or calibration method. We demonstrate that this technique has good sensitivity for
measuring and distinguishing left and right asymmetry caused by sidewall tilting, and can therefore be used for
monitoring processes, such as lithography and etch processing, for which symmetric structures are desired.
The extension of optical lithography to 32nm and beyond is made possible by Double Patterning Techniques
(DPT) at critical levels of the process flow. The ease of DPT implementation is hindered by increased significance of
critical dimension uniformity and overlay errors. Diffraction-based overlay (DBO) has shown to be an effective
metrology solution for accurate determination of the overlay errors associated with double patterning [1, 2] processes. In
this paper we will report its use in litho-freeze-litho-etch (LFLE) and spacer double patterning technology (SDPT),
which are pitch splitting solutions that reduce the significance of overlay errors. Since the control of overlay between
various mask/level combinations is critical for fabrication, precise and accurate assessment of errors by advanced
metrology techniques such as spectroscopic diffraction based overlay (DBO) and traditional image-based overlay (IBO)
using advanced target designs will be reported. A comparison between DBO, IBO and CD-SEM measurements will be
reported. . A discussion of TMU requirements for 32nm technology and TMU performance data of LFLE and SDPT
targets by different overlay approaches will be presented.
As optical lithography advances to 32 nm technology node and beyond, double patterning technology (DPT)
has emerged as an attractive solution to circumvent the fundamental optical limitations. DPT poses unique demands on
critical dimension (CD) uniformity and overlay control, making the tolerance decrease much faster than the rate at which
critical dimension shrinks. This, in turn, makes metrology even more challenging. In the past, multi-pad diffractionbased
overlay (DBO) using empirical approach has been shown to be an effective approach to measure overlay error
associated with double patterning . In this method, registration errors for double patterning were extracted from
specially designed diffraction targets (three or four pads for each direction); CD variation is assumed negligible within
each group of adjacent pads and not addressed in the measurement. In another paper, encouraging results were reported
with a first attempt at simultaneously extracting overlay and CD parameters using scatterometry .
In this work, we apply scatterometry with a rigorous coupled wave analysis (RCWA) approach to characterize
two double-patterning processes: litho-etch-litho-etch (LELE) and litho-freeze-litho-etch (LFLE). The advantage of
performing rigorous modeling is to reduce the number of pads within each measurement target, thus reducing space
requirement and improving throughput, and simultaneously extract CD and overlay information. This method measures
overlay errors and CDs by fitting the optical signals with spectra calculated from a model of the targets. Good
correlation is obtained between the results from this method and that of several reference techniques, including empirical
multi-pad DBO, CD-SEM, and IBO. We also perform total measurement uncertainty (TMU) analysis to evaluate the
overall performance. We demonstrate that scatterometry provides a promising solution to meet the challenging overlay
metrology requirement in DPT.
The performance of a 0.25NA full-field EUV exposure tool is characterized in terms of CD uniformity, focus
and overlay control, as well as dose uniformity. In addition to the characterization of the scanner, we explore the use of
scatterometry techniques for the measurements of extremely fine resolution features, with critical dimensions below 40
nm. The stability of the scanner performance over an extended period of time is assessed.
The ability to manage critical dimensions (CDs) of structures on IC devices is vital to improving product yield
and performance. It is challenging to achieve accurate metrology data as the geometries shrink beyond 40 nm features.
At this technology node CDSEM noise and resist LER are of significant concerns1.
This paper examines the extendibility of scatterometry techniques to characterize structures that are close to limits of
lithographic printing and to extract full profile information for 2D and 3D features for OPC model calibration2. The resist
LER concerns are diminished because of the automatic averaging that scatterometry provides over the measurement pad;
this represents a significant added value for proper OPC model calibration and verification. This work develops a
comparison matrix to determine the impact of scatterometry data on OPC model calibration with conventional CDSEM
measurements. The paper will report test results for the OPC model through process data for accuracy and predictability.