The present disclosure pertains to a metrology device for determining metrological characteristics of a sample, with conductive features embedded in a layer of a material or a combination of materials having a conductivity different from that of the conductive features.
The present disclosure pertains to a method for determining metrological characteristics of a sample with conductive features embedded in a layer of a material or a combination of materials having a conductivity different from that of the conductive features.
For manufacturing integrated circuits a proper metrology of electrically conductive features is of ever increasing importance. Non-limiting examples of target metrology are measurement of buried electrically conductive features and subsurface material characterization. The first example may comprise a measurement of a (buried) metal layer thickness or a measurement of a height of an electrically conductive feature that is buried under a conductive layer, such as a fin height under a W or Co layer. This type of analysis also serves for defect detection in metallic layers and interconnects. The feature to be detected in that case is a defect or a void in an electrically conductive feature. In another example metrology is required for metal filled vias, BEOL interconnects, metal gates, metal lines, and the like. A still further example is hybrid CD-metrology for GAA and 3D NAND devices. Also it is desirable to identify material characteristics, so as the facilitate metrology analysis.
In particular, metrology for the next generation of the semiconductor devices like finFETs, GAAs, and 3D NANDs that incorporate complex device structures and new materials is currently one of the biggest challenges for the semiconductor industry.
The current e-beam based metrology solutions like CD-SEM are regarded damaging and have limitations for metrology in vertical dimensions (height/depth).
In some cases subsurface scanning probe microscopy (SSPM) is suitable as an alternative/complementary non-destructive technique to determine both lateral and vertical dimensions. However SSPM is not applicable when the mechanical properties (visco-elasticity or density) of the features to be measured are comparable with that of the substrate of the sample and/or neighboring features.
Accordingly, there is a need for a non-destructive device and method to measure and characterize for samples as specified above, for applications wherein SSPM is not applicable.
Reference is made to the following prior art.
CN 109 001 493 B relates to a high-precision magnetic detection microscopic device with integration of diamond nitrogen vacancy scanning and an AFM. This prior art is however silent about measurement of metrological characteristics of samples as specified above. Paragraph 60 merely mentions: “the fluorescence collecting mirror 4-8, and the photodetector 4-12 and photon counter 4-11 detect and calculate the magnetic information of the sample.”
US 2008/173812 A1 relates to a scanning magnetic microscope, for characterizing a sample of dilute paramagnetic particles, which has a photoluminescent nanoprobe implanted in the tip apex of an atomic force microscope (AFM), a scanning tunneling microscope (STM) or a near-field scanning optical microscope (NSOM) and exhibits optically detected magnetic resonance (ODMR) in the vicinity of impaired electron spins or nuclear magnetic moments in the sample material. The described spin microscope is said to have demonstrated nanoscale lateral resolution and single spin sensitivity for the AFM and STM embodiments.
WO 2014/051886 A1 discloses a sensing probe for magnetic media formed of a diamond material comprising one or more spin defects that are configured to emit fluorescent light and are located no more than 50 nm from a sensing surface of the sensing probe. The sensing probe includes an optical outcoupling structure formed by the diamond material and configured to optically guide the fluorescent light toward an output end of the optical outcoupling structure. An optical detector detects the fluorescent light that is emitted from the spin defects and that exits through the output end of the optical outcoupling structure after being optically guided therethrough. A mounting system holds the sensing probe and controls a distance between the sensing surface of the sensing probe and a surface of a sample while permitting relative motion between the sensing surface and the sample surface.
US 2016/282427 A1 discloses a diamond probe that is suitable to be attached to an Atomic Force Microscope and that is created with a tip that incorporates one or more Nitrogen Vacancy (NV) centers located near the end of the tip and with which an electrically charged sample, such as a field effect transistor or flash memory, is to be scanned. The probe arm acts as an optical waveguide to propagate the emission from the NV center with high efficiency and a beveled end directs excitation light to the NV center and directs photoluminescence light emanating from the NV center into the probe arm.
EP 3 376 245 A1 pertains to a sensor device that comprises a carrier, a force feedback sensor, and a probe containing a spin defect. The probe is connected to the force feedback sensor either directly or indirectly via a handle structure. In order to couple the spin defect to a microwave field in an efficient and robust manner, the sensor device comprises an integrated microwave antenna arranged at a distance of less than 500 micrometers from the spin defect. The sensor device can be configured as a self-contained exchangeable cartridge that can easily be mounted in a sensor mount of a scanning probe microscope.
According to one aspect of the disclosure, a metrology device is provided herein for determining metrological characteristics of a sample with conductive features embedded in a layer of a material or a combination of materials having a conductivity different from that of the conductive features. The metrology device comprises:
In operation the scanning mechanism displaces the probe relative to the sample along a surface of the sample. The SPM controller therewith controls a position of the probe in a direction transverse to the sample. In one embodiment the SPM controller is configured to maintain a contact mode of operation, wherein the probe is held in continuous contact with the sample surface during scanning. In an example thereof, the probe is provided as a cantilever. Typically therewith a deflection of an optical beam by the cantilever is used to provide a feedback signal with which the probe is maintained in contact with the sample.
In another embodiment the SPM controller is configured to maintain the probe close to the surface without contacting the surface. In an example of the second embodiment the probe is provided as a tuning fork, and a shear-force based feedback mechanism is used to maintain the probe close to the sample. An advantage of this embodiment is its simplicity as additional hardware components are avoided. In case the probe is provided as a cantilever, typically additional components are required, to enable z-position feedback, for example a light source directing a beam at the cantilever and an extra quad detector to sense a deflection of the beam by the cantilever. An advantage of this latter embodiment is its suitability for combination with other SPM cantilever-based metrology tools (like a subsurface SPM tool).
During scanning, the radiation source emits radiation to excite the tip to emit fluorescent light and in response thereto the optical sensor provides a sense signal indicative of an intensity of the emitted fluorescent light. The signal processor processes the sense signal to compute one more characteristics of a feature present in the sample. As set out in more detail hereinafter, the fluorescent response of the tip is affected by magnetic fields at the position of the tip. Therewith the signal processor computes characteristics of a feature at the location of the sample, such as its lateral dimensions, its thickness, a depth, a conductivity.
The small dimensions of the diamond tip, i.e. typically in the order of magnitude of about 10 nm to a few hundred nm already enable a highly accurate determination of feature positions in the sample. A further improvement of the positional accuracy is achieved in an embodiment of the sample characterization device wherein the diamond tip has a single NV center, which acts as an atom-sized probe.
In one embodiment the metrology device comprises a microwave antenna arranged near the tip and a microwave signal generator to supply the microwave antenna with a microwave signal. In one mode of operation the microwave signal generator is a sweeping microwave signal generator that is configured to supply the microwave antenna with a microwave signal of varying frequency. In that mode of operation it is detected at which frequency or pair of frequencies a reduction of a fluorescent response of the NV center occurs, which is indicative for a local magnitude of a magnetic field.
In another mode of operation of this embodiment the metrology device is used to determine a spin relaxation time T1. A value of the measured spin relaxation time T1 is indicative for a thickness and/or depth of electrically conductive features in the sample. In addition, changes in the value of T1 during scanning in a lateral direction indicate a lateral dimension of the features.
In another embodiment the metrology device alternatively or additionally comprises a magnetic coil and a current generator to supply the magnetic coil with a supply-current to induce a magnetic field in the sample. In case the supply current supplied in operation is an AC-current or comprises an AC-current component then an AC-magnetic field is generated by the magnetic coil that induces Eddy currents in electrically conductive features in the sample. The induced Eddy current in an electrically conductive feature induces a secondary magnetic field, which is detectable by the NV-center, so that a position or the dimensions of the feature can be determined from the secondary magnetic field as detected by the NV-center. In some examples, the metrology device further comprises a permanent magnet to provide a background magnetic field in the tip. Therewith the sensitivity and robustness of the metrology device is significantly increased. In some examples the metrology device further comprises a frequency control unit to control an operational frequency of the current generator. In one example thereof the frequency control unit is configured to subsequently select mutually different operational frequencies or to perform a frequency sweep. Low frequency magnetic fields tend to penetrate further into the sample than high frequency magnetic fields. Hence, when performing the Eddy current based measurements at mutually different frequencies at or near a same location mutually independent measurement results can be obtained that enable detecting mutually different features at mutually different depths at a same lateral position. In an alternative embodiment, the current generator generates a supply current having at least a first component and a second component of mutually different frequency. The composite response signal can be split into respective response signal components for mutually different frequency ranges, for example using a low-pass filter and a high pass filter.
In an embodiment the signal processor comprises signal processor components to determine respective changes in a magnetic field for a lateral position near the surface of the sample resulting from a source magnetic field for respective mutually different magnetic field frequencies, and to determine material properties of a material at that lateral position in the sample from the determined changes. The signal processor components for example comprise a signal processor component to decompose the response signal into two or more components, for example a component comprising a low-pass filter and a high-pass filter. Optionally the component to decompose the response signal into a plurality of components also comprises one or more band filters to select components of the response signal in respective frequency components. Further signal processor components are provided therewith to determine a magnitude of respective ones of the components therewith providing an indication of the change of the magnetic field associated with properties of the material in the sample, such as its relative magnetic permeability and its conductivity. Still further a signal analysis component may be provided that compares the indicated change magnitudes to identify the nature of the material in the sample near the sensed position.
According to a second aspect of the present disclosure a method is provided for determining metrological characteristics of a sample with conductive features embedded in a layer of a material or a combination of materials having a conductivity different from that of the conductive features, is provided. The method comprises:
The sensed signals are indicative for an extent to which the magnetic field near the surface of the sample is changed due to material properties. For the currently investigated materials it appears that at relatively low frequencies, dependent on the conductivity, e.g. for copper in the range of 0 to 1 kHz, the change in magnetic field is mainly determined by the relative magnetic permeability of the material present at that position. For relatively high frequencies, e.g. in the case of copper about 100 MHz the effect of Eddy currents in the material becomes more relevant. The contribution thereof depends on the conductivity of the material. Using a single frequency, it is possible to determine both vertical and lateral dimensions of a feature in a sample. Therewith an alternative is provided for relaxometry measurements. Using two frequencies, e.g. a first frequency of at most 10 kHz and a second frequency of at least 100 kHz, would make it possible already to determine material characteristics provided that its dimensions are known. Conversely, the dimensions of a feature can be determined according to refence measurements and/or benchmarked models if the material from which it is constituted is known. This is typically the case for semiconductor manufacturing applications
As the change in magnetic field strength is also determined by other circumstances, such as the dimensions of a material feature it is preferred to perform the measurements for a larger plurality of different frequencies. The contribution of Eddy currents in the material to the sensed magnetic field change is characterized by a monotonously increasing function of the frequency, which may be approximated as a linear relationship with the frequency. In an embodiment the at least a first frequency and a second frequency are part of a frequency sweep from a lower frequency of at most 10 kHz to a higher frequency of at least 100 kHz. In that case the magnitude of the magnetic field change can be observed during the frequency sweep to analyze various material characteristics and dimensions of features in the sample. The observed behavior of the magnitude of the magnetic field change during a lateral scan is indicative for the lateral dimensional properties of a feature in the sample. As noted above, the probe when being displaced relative to the sample may be held in continuous contact with the sample surface during scanning or the probe may be maintained close to the surface without contacting the surface.
These and other aspects are described in more detail with reference to the drawing. Therein:
Like reference symbols in the various drawings indicate like elements unless otherwise indicated.
The SPM controller 32 is configured to position the probe 20 in a direction z transverse to the sample S, for example to maintain the probe in contact with the surface Ssf of the sample S at a predetermined static force or to maintain the probe close above the surface Ssf for example in a shear-force mode (or oscillating with a predetermined amplitude above the sample for example in a tapping mode). The SPM controller 32 is controlled by the signal processor 60 with a signal C32 to select a particular mode of operation.
The probe 20 has a diamond tip 21 with one or more nitrogen-vacancy (NV) centers. The radiation source 40 irradiates the tip with radiation to excite the tip to emit fluorescent light. The optical sensor 50 provides a sense signal S50 indicative of an intensity of the emitted fluorescent light. A confocal imaging member 45 maps the radiation irradiated by the radiation source 40 onto the tip and maps the fluorescent light emitted by the tip onto the optical sensor 50. The signal processor 60 process the sense signal S50 to compute an extent of a feature present in the sample S.
An exemplary embodiment of the optical components of the metrology device of
In operation, the scanning mechanism 30 (See
During a relative displacement in a lateral direction, e.g. in the direction x, or at mutually different relative lateral positions of the sample S and the probe 20 the following is performed.
In an embodiment, at each relative lateral position, the spin relaxation time T1 is measured. Then the signal processor 60 processes the sense signal S50 to compute a dimension of a feature present in the sample. Although various NV-based measurement approaches are available, T1 relaxometry is considered the best option as it is specifically suitable to sense the “magnetic noise” from the random movement of electrons in a metal at a few GHz. The various T1 relaxometry methods have in common that the NV center is polarized with a laser pulse (having a duration in the order of few μs). After a waiting time, another laser pulse (normally shorter) is applied to read-out the spin state of the NV center. This laser pulse excites the NV center. The resulting fluorescence is recorded and compared to the reference obtained at t=0. This procedure is repeated for several waiting times (t) to extract the relaxation time. Variations to this method can be also done, e.g. by adding a microwave pulse before readout. Small magnetic noises due to electron movements in a material will cause the NV spin to relax. The amount of spin relaxation therewith depends on the magnetic noise spectral density. i.e. the more noise, the faster the pace at which relaxation occurs. This implies that features with high conductivity (e.g. metals) will cause the spin to relax faster than features with lower conductivity (e.g. non-metals), due to their free electrons moving around and making more magnetic noise than non-conductive features.
In those regions in the lateral direction between the metal features M1, the spin relaxation time T1 is relatively large. In the regions where the metal features M1 are present, the spin relaxation time T1 drops below a reference level Tr. Therewith the signal processor 60 determines that the regions for which the spin relaxation time T1 is less than the reference level Tr as having the metal feature M1 and the regions where the spin relaxation time T1 is greater than the reference level Tr as not having the metal feature. The critical dimension CD1, i.e. the width of the metal feature, is determined as the length of the range in the lateral direction where the spin relaxation time T1 is less than the reference level Tr. The critical dimension CD2, i.e. the distance between subsequent metal features, is determined as the length of the range in the lateral direction where the spin relaxation time T1 exceeds the reference level Tr. By way of example, close to the metal features a spin relaxation time T1 in the order of 1 ms is measured, whereas in a region between subsequent metal features, the spin relaxation time is in the order of 5 ms.
Therein dnv,t is the distance from the NV-center to the tip, and dt,s is the distance from the tip to the surface of the sample.
Also it is shown in this example how the thickness tf, herein denoted as CD4, of the metallic feature/film M12 over a dielectric background is determined. The relationship between the features d, tf on the hand and the relaxation rate Γmet is provided by the following expression:
Therein.
Provided that the depth d of a feature is known, this relationship can be used to determine the thickness tf of the feature with a component 61 (See
In an illustrative example, the sample has lateral dimensions of 5×5 mm, and the scanning mechanism 30 provides for a scanning range of 50 μm×50 μm in the x-y plane. However any other lateral dimensions are conceivable. The same principles would apply for example when applied to a full wafer with a cross-section of 450 mm. Furthermore, the scanning range can be selected from a larger range by a manual setting. Alternatively or additionally, the scanning mechanism 30 may provide for a larger scanning range when desired. The SPM controller 32 is capable to position the probe 20 within a scanning range in the direction z of 2 μm at a bandwidth of 50 kHz to maintain the tip 21 in a contact mode or a tapping mode or at a predetermined distance to the sample dependent on a selected setting. In other examples the z-scanning range may be larger, e.g. up to 15 μm. Alternatively or additionally, the z-scanning range can be selected from a larger range by a manual setting. In the embodiment shown, the radiation source 40 is a laser having a wavelength of 515 nm and an output power of 150 mW. Output noise in the wavelength range of 250 Hz-2 MHZ (rms) is less than 0.2%. The optical sensor 50 is a photon counter having a detection efficiency over the NV emission band of 67%. The fluorescence collection efficiency from the NV center up to and before the optical sensor is about 1.7%. In a DC magnetometry operational mode the sensitivity of the metrology device is in a range of 1-10 μT/√Hz at a resolution of 10-20 nm.
It is noted that the scanning mechanism 30 may perform a relative displacement in another lateral direction, e.g. in the direction y to determine critical dimensions in that direction. Alternatively, measurement data obtained for measurements obtained from scanning in the x-direction for mutually different values can be used for that purpose.
It is further noted that various signal processing operations may be applied to the sense signal S50. For example a deconvolution operation may be applied to compensate for a pointspread function involved in the imaging method.
An alternative mode of operation of the metrology device of
Reference is first made to
Upon irradiation of the NV-center with a suitable wavelength, e.g. less than 550 nm, transitions occur from the ground states to their corresponding exited states. These transitions are followed by a reverse transition to the corresponding ground state. Therewith a photon is emitted at a wavelength of about 637 nm. Instead of this reverse transition a transition may occur from an excited state to the singlet state succeeded by a transition from the singlet state to the ground state with spin 0. This sequence of transitions is more likely to occur as an alternative for the direct transition from a non-zero spin excited state than as an alternative for the direct transition from the zero spin excited state. I.e. the transition from the excited zero-spin state will almost always take place in a direct manner, i.e. as a direct transition to the zero-spin ground state. Contrary thereto, the transition from a non-zero spin excited state will in about 30% of the cases take place indirectly, via the singlet state. The indirect transition is non-radiative and therefore not detectable in the visible range.
Upon subjecting the NV-center to micro-wave (MW) radiation having a frequency corresponding to the energy gap between the zero-spin ground state and a non-zero spin ground state, a transition is induced from the zero-spin ground state to the non-zero spin ground state. This also implies that the irradiation causes a transition from the non-zero spin ground state to the non-zero spin exited state. As mentioned above, starting from the non-zero spin excited state, there is a substantial probability that the reverse transition will not take place directly, but will take place via the singlet state without an emission in the visible range. This effect is illustrated in
As illustrated in
As shown in
Δf=2γB
Therewith the scanning mechanism 30 displaces the probe 20 relative to the sample, along a surface of the sample, in this case by translating the sample carrier 10 in the x-y plane while the sweeping microwave signal generator 70 supplies the microwave antenna 72 with a microwave signal of varying frequency. The signal processor 60 determines at the scanned locations of the sample for which microwave frequencies a reduction in the luminance of the NV-center occurs, and therewith computes the magnetic field as a function of the position. In one embodiment the sweeping microwave signal generator 70 performs the frequency sweep in a frequency range which is exclusively at one side of the center frequency, e.g. from 2700 MHz to 2870 MHz, or from 2870 MHz to 3100 MHz. This is advantageous in that the range can be relatively modest, which may be more cost-effective. Alternatively, it may be configured to perform the frequency sweep bilaterally, e.g. in a range from 2700 to 3100 MHz, which is advantageous in that the magnitude of the magnetic field can be estimated from two measurements, i.e. the lower frequency and the higher where the photoluminescence is minimized, which contributes to an accuracy of the determination. According to another approach, the frequency of the microwave signal is maintained at a constant value, and the photoluminescence amplitude changes resulting from the local magnetic field is monitored. In this case the strength of the local magnetic field is computed from the monitored photoluminescence amplitude observed for the fixed microwave signal frequency. According to a still further approach the controller tracks the frequency at which a minimum photoluminescence response is achieved. An example thereof is shown in
In the simulation results for copper (a, . . . , f), it can be observed that for low frequencies the computed difference dB is in the order of magnitude of 0.2 μT. This difference is explained due to the fact that copper is diamagnetic, having a relative permeability of 0.999994. For frequencies of 100 kHz and higher the magnitude of the difference dB increases. For 100 kHz the difference dB is increased to about 0.3 μT. For a frequency of 1 MHz, the difference dB is increased to about 3 μT. This increase is explained by the Eddy current effect which monotonously increases with the frequency.
In the simulation results for tungsten (g, . . . , l), it can be observed that for low frequencies the computed difference dB is in the order of magnitude of 2.2 μT. This difference is explained due to the fact that tungsten is paramagnetic, having a relative permeability of 1.000068. Also for high frequencies an increase of the difference dB is observed, but this increase is less prominent than for copper. For a frequency of 1 MHz, in the case of the tungsten disk of radius 2500 nm and thickness of 100 nm the difference dB increases from about 2.3 to about 2.4 μT. This is due to the fact that tungsten, due to its larger relative permeability as compared to copper already causes a relatively high change dB in the magnetic field and to the fact that the conductivity of tungsten is only about 0.3 times that of copper.
From the simulations it is apparent that the sample characterization device with the signal processing device of
As another example wherein the material used in a feature is known, a detected deviation between the response as measured and the expected response can be used as an indication for a possible defect of the feature. For example reference measurement data may be available for conductive interconnects, and a detected deviation of measurement data of a specific interconnect in the sample is indicative for a defect or a void within that specific interconnect. The magnitude of the deviation may further be indicative for the size of the defect or the void.
As another example, which is demonstrated with reference to
As a first example it is shown that the critical dimension CD1, i.e. the lateral size of a conductive feature M1 is determined as the distance in the lateral direction for which the measured value dB exceeds a threshold value Btr.
As a second example it is shown that the critical dimension CD2, i.e. the lateral distance between successive conductive feature M1 is determined as the distance in the lateral direction for which the measured value dB is less than a threshold value Btr.
Analogously, in a fourth example shown in
It is noted that metrology can be performed alternatively on the basis of a measurement of the relaxation time. As demonstrated above with reference to
Similarly, as further demonstrated above with reference to
Further, as an alternative of the third example the depth CD3 of a conductive feature M11 having a known thickness CD4 is determined to be the depth value for CD3 for a conductive feature having the known dimensions and for which the expected value of the relaxation time T1 is equal to the measured value of the relaxation time.
Analogously, as an alternative for the fourth example, the thickness CD4 of a conductive feature (M12) at a known depth CD3 is determined to be the thickness value for CD4 for a conductive feature at the known depth and for which the expected value of the relaxation time T1 for the corresponding lateral dimension is equal to the measured value of the relaxation time T1.
In a still further example, the inventive metrology device is configured to detect a void or defect in a conductive feature. In one embodiment it is configured for this purpose by determining that the detected relaxation time deviates from a reference value for said conductive feature. In another embodiment it is configured for this purpose by determining that the detected magnetic field deviates from a reference value for said conductive feature. In a still further embodiment it is configured to detect a void or defect by using both input data indicative for a deviation of a detected relaxation time and input data indicative for a deviation of a detected magnetic field.
In again another example the inventive metrology device is configured to detect a deviation of a material property of a conductive feature. In one embodiment it is configured for this purpose by determining that the detected relaxation time deviates from a reference value for said conductive feature. In another embodiment it is configured for this purpose by determining that the detected magnetic field deviates from a reference value for said conductive feature. In a still further embodiment it is configured to detect a deviation of a material property of a conductive feature by using both input data indicative for a deviation of a detected relaxation time and input data indicative for a deviation of a detected magnetic field.
While the present invention has been described with respect to a limited number of embodiments, those skilled in the art will appreciate numerous modifications and variations therefrom within the scope of this present invention as determined by the appended claims
Number | Date | Country | Kind |
---|---|---|---|
21201626.5 | Oct 2021 | EP | regional |
Filing Document | Filing Date | Country | Kind |
---|---|---|---|
PCT/NL2022/050564 | 10/7/2022 | WO |