A multi-process node testing and monitoring method for thin-film lithium niobate modulators

By setting up process control monitoring areas and device-level verification areas in the thin-film lithium niobate modulator process chain, and implementing parameter testing in the manufacturing sequence, the problem of difficult anomaly source location in the thin-film lithium niobate modulator process chain is solved, achieving efficient anomaly detection and process control, and reducing costs and iteration cycles.

CN123111373APending Publication Date: 2026-11-17CENT SOUTH UNIV +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611480411.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-22
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In the existing technology, the process chain of thin-film lithium niobate modulators is difficult to monitor independently at each node, which makes it difficult to locate the source of the anomaly and makes it impossible to stop the subsequent high-cost process in time when the preceding node is abnormal, thus prolonging the iteration cycle and increasing the cost.

Method used

Within the same process batch, process control monitoring areas and device-level verification areas are set up. By mapping structural identifiers with process information, parameter tests are carried out in the manufacturing sequence to locate anomalies in layers, including testing and correction of nodes such as thin film deposition, pattern transfer, lithium niobate etching, cladding, and planarization.

Benefits of technology

It achieves the separation and extraction of propagation loss and bending loss, and can immediately stop subsequent processes when an anomaly occurs, thereby improving failure analysis efficiency, reducing invalid processing and duplicate fabrication, and lowering costs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123111373A_ABST
    Figure CN123111373A_ABST
Patent Text Reader

Abstract

This invention provides a multi-process node testing and monitoring method for thin-film lithium niobate modulators, comprising the following steps: setting up a process control monitoring area and a device-level verification area within the same process batch, and establishing a mapping relationship between structural identifiers and process information for each test structure; sequentially performing parameter tests on the test structures corresponding to each process node in the set process control monitoring area according to the manufacturing sequence, and comparing the test results with preset control conditions to determine whether to proceed to the next process node; performing baseline tests on independent electrode structures and waveguide-electrode superposition structures in the set device-level verification area, and performing device tests on the complete Mach-Zehnder modulator; binding the obtained test results of each structure with batch and coordinate information to form process monitoring records, trend warnings, and batch feedback. This invention prevents propagation loss and bending loss from mixing, eliminating the systematic error caused by comparing only the total insertion loss.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of optoelectronic device testing technology, specifically a multi-process node testing and monitoring method for thin-film lithium niobate modulators. Background Technology

[0002] Thin-film lithium niobate materials, with their high electro-optic coefficient, wide transparency window, low intrinsic optical loss, and excellent high-speed modulation potential, have become an important material platform for high-speed electro-optic modulators. Thin-film lithium niobate modulators typically use lithium niobate-on-insulator wafers as a substrate, and the process involves the following steps: hard mask deposition and pattern transfer, deep and shallow etching of the lithium niobate waveguide, cladding deposition and chemical mechanical planarization, thermally modulated electrode fabrication, high-frequency traveling wave electrode fabrication, and device-level optical, electrical, and electro-optic testing.

[0003] The aforementioned process chain contains multiple interconnected nodes. Deviations in any parameter, such as hard mask parameters, critical dimensions, etch depth and sidewall angles, cladding steps, electrode resistivity, and high-frequency transmission characteristics, can be transmitted through propagation loss, bending loss, mode constraints, electrode-waveguide overlap, and impedance matching, ultimately affecting the modulator's insertion loss, extinction ratio, half-wave voltage, and electro-optic bandwidth. Therefore, it is crucial to implement process control and monitoring and device-level verification for this complete manufacturing chain.

[0004] In existing technologies, the complete Mach-Zehnder modulator is typically tested directly after all processes are completed to determine whether the final device meets specifications. However, the aforementioned existing technologies still have the following shortcomings:

[0005] 1. An increase in the number of bends is often accompanied by an increase in the total waveguide length. If only the total insertion loss of different bend structures is compared, the propagation loss per unit length and the additional loss of a single bend will be mixed together, making it difficult to obtain independent parameters that can be used for process evaluation. Loss testing based on the resonator quality factor is also affected by coupling state, resonant line shape, local dimensional deviations, and fitting models, and not all process stages are suitable for arranging or testing resonators.

[0006] 2. Existing monitoring solutions mostly involve detecting a single test structure or a single process step, which cannot promptly stop subsequent high-cost processes when anomalies occur at preceding nodes. Furthermore, when devices exhibit increased insertion loss, decreased extinction ratio, increased half-wave voltage, or insufficient electro-optic bandwidth, it is difficult to directly distinguish whether the anomaly originates from the hard mask, etching, bent structure, cladding, electrodes, or test interface. If slicing, cross-sectional observation, or repeated fabrication are required to check each item, it will significantly extend the iteration cycle and increase costs. Summary of the Invention

[0007] The main objective of this invention is to provide a multi-process node testing and monitoring method for thin-film lithium niobate modulators that can separate and extract propagation loss and bending loss and can perform layered localization of anomalies in a complete Mach-Zehnder modulator.

[0008] The multi-process node testing and monitoring method for thin-film lithium niobate modulators provided by this invention includes the following steps:

[0009] S1. Set up a process control monitoring area and a device-level verification area within the same process batch, and establish a mapping relationship between structural identifiers and process information for each test structure.

[0010] S2. Perform parameter tests on the test structures corresponding to each process node in the process control monitoring area set in S1 in sequence according to the manufacturing order, and compare the test results with the preset control conditions to determine whether to enter the next process node.

[0011] S3. Perform baseline testing on the independent electrode structure and waveguide and electrode superposition structure in the device-level verification area set in S1, and perform device testing on the complete Mach-Zehnder modulator.

[0012] S4. Bind the structural test results obtained in S2 and S3 with batch and coordinate information to form process monitoring records, trend warnings and batch feedback.

[0013] The process control and monitoring areas include areas A to F. Area A is used for testing the thickness, refractive index, stress, surface roughness, and on-wafer uniformity of hard masks or dielectric films. Area B includes isolated lines, dense lines, long-distance line edge roughness test strips, and micro-loading arrays with different spacings, used to evaluate exposure, development, and hard mask pattern transfer. Area C includes deep etching windows, shallow etching windows, step measurement grooves, X and Y direction overlay scales, double-groove sacrificial waveguides, and interlayer relational units, used to evaluate lithium niobate etching depth, sidewalls, undercuts, redeposition, and deep / shallow etching alignment. Area D includes non-resonant long-path waveguide groups with fixed bending contributions and different centerline lengths, used to extract propagation loss. Area E includes waveguide matrices with different bending numbers under a fixed bending radius, used to extract single-bend additional loss. Area F includes dielectric film thickness blocks, arrays with different pattern densities, and trans-waveguide overlay structures, used to evaluate upper cladding film thickness, surface roughness, planarization density effect, and trans-waveguide steps.

[0014] The device-level verification area includes structures M1 to M6; M1 is an independent high-frequency metal electrode or transmission line structure used to measure at least one of the following: metal thickness, linewidth, sheet resistance, open / short circuit, scattering parameters, characteristic impedance, and group delay; M2 is an independent thermally tunable electrode structure used to measure thermally tunable electrode resistance, on-chip uniformity, temperature coefficient, and power withstand capability; M3 is a waveguide and high-frequency metal electrode superimposed structure used to check the continuity of the high-frequency metal across the cladding step, the relative position of the electrode and the waveguide, residual adhesive, bridging, and local metal absorption risk; M4 is a waveguide and thermally tunable electrode superimposed structure used to measure heater resistance, heating power, thermal stability, and additional optical losses introduced by heating; M5 is a complete Mach-Zehnder modulator; and M6 is a coupler, straight waveguide, bent waveguide, beam splitter / combiner, or bypass reference structure that uses the same or equivalent optical interface as the complete modulator.

[0015] In S1, a unique structure identifier is assigned to each test structure, and a mapping relationship is established between the structure identifier and the wafer number, chip number, wafer coordinates, layout layer number, process node, test equipment, test conditions, process recipe version and judgment rules. The structure identifier is preferably composed of region code, structure serial number, wafer coordinates and version number.

[0016] In S2, the manufacturing process is carried out in the following order: thin film deposition, pattern transfer, lithium niobate deep and shallow etching, passive optical testing, cladding and planarization. After the parameter test is completed at each process node, the results are compared with the preset control conditions. If the control conditions are not met at the previous node, the subsequent process is stopped and the adjustment direction is given.

[0017] In the thin film deposition node test, the main measurement window and multiple on-chip sampling points are selected in area A to measure the hard mask film thickness, refractive index at the target working wavelength, residual stress, root mean square surface roughness and on-chip uniformity.

[0018] For the i-th sampling point, the film thickness deviation Represented as:

[0019]

[0020] in, For the first The measured film thickness at each sampling point Target film thickness;

[0021] Intra-film relative non-uniformity Calculate using the following formula:

[0022]

[0023] in, This represents the maximum film thickness at the sampling point. This represents the minimum film thickness at the sampling point. This represents the average film thickness at the sampling points;

[0024] When at least one of the key parameters, such as film thickness, refractive index, stress, roughness, or intra-film uniformity, fails to meet the control conditions, the abnormal location, batch, and equipment status are recorded. Subsequent high-precision patterning is stopped, and the deposition temperature, gas flow rate, RF power, deposition time, or pretreatment conditions are adjusted.

[0025] In the graphic transfer node test, the key dimensions of isolated graphics, dense graphics and graphics with different spacings were measured in area B, and the edge roughness and line width roughness of long-distance graphics were statistically analyzed.

[0026] The critical dimensional deviations for isolated and dense structures are expressed as follows:

[0027]

[0028]

[0029] The dense-isolation effect is expressed as:

[0030]

[0031] in, Design the dimensions for the layout; The deviation of the measured critical dimensions of the isolated structure from the layout design dimensions; For the measured critical dimensions of the isolated structure; The deviation of the measured key dimensions of the dense structure from the layout design dimensions; For the measured key dimensions of dense structures; This represents the difference in critical dimension deviations between dense and isolated structures under the same design dimensions.

[0032] For edge position sequences extracted along the length direction The roughness of the line edge is characterized by the standard deviation of the edge position or three times its standard deviation:

[0033]

[0034] in, Line edge roughness is expressed as three times the standard deviation of the samples at the edge location; The total number of edge location sampling points included in the statistics, taken as an integer greater than 1; This is the arithmetic mean of the sampled values ​​at the aforementioned edge locations;

[0035] When critical dimension deviations, dense-isolated differences, line edge roughness, or local defects do not meet the preset control conditions, lithium niobate etching is stopped, and the exposure dose, focus conditions, development time, hard mask etching gas ratio, or bias power are adjusted.

[0036] In the deep and shallow etching node testing of lithium niobate

[0037] After completing the first deep etching of lithium niobate, contour scanning or atomic force microscopy was performed on the deep etching window and step measurement groove in region C, and cross-sectional or tilt scanning electron microscopy was performed on the double-groove sacrificial waveguide to obtain the deep etching depth, sidewall angle, undercut, redeposition, and waveguide edge morphology. After completing the second shallow etching of lithium niobate, the depth of the shallow etching window was measured, and the fine and coarse range scales in the X and Y directions were read.

[0038] Let the overlay deviations in the X and Y directions be respectively and Then the composite overlay deviation Represented as:

[0039]

[0040] For separated, overlapping, intersecting, and enclosing units, check for local residues, over-etching, mask offset, and edge collapse; use deep etching depth, shallow etching depth, overlay deviation, sidewall corner, undercut, and redeposition as joint release conditions; when any key parameter does not meet the control conditions, stop the upper cladding deposition, and adjust the etching time, offset power, ion source power, reactive gas ratio, cavity state, resist removal cleaning, or overlay compensation according to the anomaly type.

[0041] In passive optical testing,

[0042] Propagation loss extraction in region D includes setting up at least three non-resonant long-path test waveguides in region D; each test waveguide has the same waveguide width W and the same bending radius. The same number of U-turns With identical input / output optical interfaces and port orientations, differing only in the length of the straight segment, the total length of the centerline of each structure is [not specified]. different;

[0043] Under the same wavelength grid, the same polarization state, the same input power, and the same automatic alignment rule, each structure should be measured at least repeatedly.

[0044] For the A structure, at wavelength Insertion loss at the point Represented as:

[0045]

[0046] in, Let be the output optical power of the i-th test structure in region D at wavelength λ; For the same structure, at the same wavelength, and under the same test conditions, the input optical power is measured in the same power unit.

[0047] No. The insertion loss obtained from repeated measurements is denoted as . Its mean and standard deviation They are respectively:

[0048]

[0049]

[0050] in, The total number of valid repeated measurements for the same D-region test structure under the same wavelength and test conditions, used to calculate the mean and standard deviation, is an integer not less than 2.

[0051] The average insertion loss of each structure in region D is linearly fitted to the total length of the centerline:

[0052]

[0053] in, The intercept is formed by input / output coupling loss, fixed bending loss, and other fixed losses; Propagation loss per unit length; The total propagation path length is calculated along the centerline of the i-th test waveguide in region D, including the centerline lengths of straight and curved segments, and uses a length unit that matches the propagation loss per unit length. Residuals caused by measurement and structural discretization;

[0054] When using the least squares method, the estimated propagation loss value for:

[0055]

[0056] in, The arithmetic mean of the total length of the centerline of each test waveguide participating in the same linear fitting in region D; This is the arithmetic mean of the average insertion loss of each test waveguide at the same wavelength, where the average insertion loss of each waveguide is calculated from its repeated measurement results;

[0057] Intercept estimate for:

[0058]

[0059] use Weighted least squares fitting is performed as the weights; when individual structures have isolated defects but the remaining structures satisfy a linear relationship, robust regression is used for verification while retaining the original data;

[0060] The quality of the fit is represented by the coefficient of determination:

[0061]

[0062] in, The coefficient of determination for the linear fit of insertion loss-centerline length is equal to 1 minus the ratio of the sum of squared residuals to the sum of squared total deviations.

[0063] When the coefficient of determination, residual distribution, or repeated measures standard deviation does not meet the preset conditions, abnormal structures are retested; when the fit quality meets the conditions, [the following is done]. As a correction factor for the propagation length in region E;

[0064] Extraction of additional losses from a single bend in region E includes grouping region E according to the bending radius R, with each radius group including at least three U-turns. Different test waveguides; each test waveguide uses the same or equivalent waveguide width, optical interface, port orientation and test conditions as region D;

[0065] Measure the total insertion loss of each structure and the total length of the centerline And use the propagation loss obtained in region D to perform propagation length correction:

[0066]

[0067] in, The corrected insertion loss is the value of the j-th test waveguide in the E-region with a bending radius of R, after deducting the propagation loss introduced by the total length of its centerline; this value still includes the fixed interface and transition loss and bending-related loss of the group, and is expressed in dB.

[0068] With a fixed bending radius Within the group, the corrected loss was fitted to the number of U-turns:

[0069]

[0070] in, This is a fixed interface and fixed transition loss for this radius group; radius Additional losses for a single 180° U-turn; The number of 180° U-turns contained in the j-th test waveguide within the group with bending radius R in region E is used as the independent variable for loss fitting of this radius group. For residuals;

[0071] Its least squares estimate for:

[0072]

[0073] in, It is the arithmetic mean of the corrected insertion loss of each test waveguide participating in the fitting within the same bending radius group at the same wavelength. The structural number of the test waveguide within this radius group; This is the arithmetic mean of the number of 180° U-turns contained in each of the test waveguides participating in the fitting of this group;

[0074] Each radius group is fitted separately, without directly merging data from different radii; data is obtained through multiple bending radius groups. The relationship between bending radius and bending radius.

[0075] The additional loss of a single 90° bend is only considered when the U-turn is equivalent to two independent 90° bends, and the effect of the intermediate connecting segment is negligible or has been corrected. Represented as:

[0076]

[0077] When the corrected loss is approximately linear with the number of U-turns, the fitted slope is used as the additional loss of a single 180° bend. When the corrected loss does not satisfy the approximately linear relationship with the number of U-turns, the single bend loss is not forced to be output. Instead, the mode coupling between adjacent bends, the straight length between bends, the excitation of higher-order modes, local etching anomalies, and the consistency of the test interface are checked, and the radius group is marked as not satisfying the condition for the superposition of independent bend losses.

[0078] In the upper cladding and planarization node tests, the upper cladding thickness, refractive index, intra-sheet uniformity and surface roughness were measured in a large area film thickness block in the F region; local depressions and erosions were measured in arrays with different pattern density; and residual steps, surface continuity and local voids or cracks were measured at the top and sides of the waveguide in the cross-waveguide covering structure.

[0079] Let the height of the reference surface before planarization be... The surface height of the densely packed graphic region is The amount of local depression Represented as:

[0080]

[0081] Let the surface height of the low-density area or adjacent area be... Then the amount of erosion Represented as:

[0082]

[0083] The cladding thickness, surface roughness, residual steps across the waveguide, local depressions, and erosion are used as joint release conditions before metal processing; if the control conditions are not met, the deposition thickness, deposition temperature, planarization time, polishing pressure, slurry formulation, or cleaning process are adjusted.

[0084] Baseline testing in S3 for independent electrode structures and waveguide-electrode stacked structures includes:

[0085] After the thermally tunable electrode and high-frequency metal electrode were fabricated, the structures from M1 to M4 were tested. For the M2 independent thermally tunable electrode array, small-signal resistance measurement and graded power loading were performed using a source measurement unit. At each power level, a preset time was maintained, and the resistance drift and recovery after cooling were recorded. When it was necessary to measure the temperature coefficient, at least two stable temperature points were set on the temperature control platform, and four-wire resistance testing was performed after the sample temperature stabilized.

[0086] For thermally adjustable electrodes, at voltage and current The resistance below and heating power They are represented as follows:

[0087]

[0088]

[0089] When at temperature and The resistance was measured below. and At that time, temperature coefficient Represented as:

[0090]

[0091] For the M1 independent transmission line, a high-frequency probe, a calibration substrate, and a vector network analyzer were used to move the reference plane to the probe tip and measure scattering parameters, insertion loss, return loss, and group delay. M3 was used to check the continuity of the metal across the cladding step and to verify the relative position between the electrode and the waveguide. M4 was used to simultaneously record the output optical power or spectrum when different heating powers were applied to obtain the heating-introduced additional loss and thermal stability baseline.

[0092] The results of M1 to M4 are correlated with the topography data of region F: when there is an anomaly in the cross-waveguide step of region F and M3 has a broken line, unstable contact or abnormal radio frequency transmission, the anomaly is preferentially located at the upper cladding or planarization node; when region F meets the control conditions but M1 or M2 is abnormal, the anomaly is preferentially located at the corresponding metal deposition, patterning, stripping or annealing node.

[0093] Device testing of the complete Mach-Zehnder modulator in S3 includes:

[0094] The M5 complete Mach-Zehnder modulator was tested in the following order: passive optics, thermally modulated bias, low-frequency electro-optics, and high-frequency electro-optics. Coupling, straight waveguide, beam splitting / combining, and bypass loss correction were performed using the M6 ​​reference structure. First, the common coupling loss and bypass loss were estimated using the M6 ​​reference structure. Second, thermally modulated power was applied to move the operating point to a preset bias position. Then, the low-frequency voltage was scanned to obtain the transmission curve and the half-wave voltage was fitted. Finally, after completing RF calibration and optoelectronic link normalization, the electro-optic response was measured.

[0095] In low-frequency electro-optic testing, the output power of the Mach-Zehnder modulator... Represented as:

[0096]

[0097] in, To normalize the input power, The interference visibility caused by the imbalance and loss difference between the two arms, The initial phase difference, It is a half-wave voltage;

[0098] By fitting the transmission curve through scanning DC or low-frequency voltage, the following can be obtained: And the half-wave voltage-length product:

[0099]

[0100] in, Effective modulation length;

[0101] Extinction ratio Represented as:

[0102]

[0103] in, The maximum output optical power on the transmission curve of the complete Mach-Zehnder modulator; The minimum output optical power on the same transmission curve is obtained under the same input optical power, wavelength and polarization conditions;

[0104] In high-frequency electro-optic testing, a preset low-frequency reference point is used. Normalize the electro-optic response:

[0105]

[0106] in, The electro-optic amplitude response is normalized relative to a preset low-frequency reference point, and the unit is dB. The frequency of the radio frequency modulation signal applied to the modulator; The complex electro-optic forward transmission response is obtained after normalization of the RF calibration and optoelectronic test link; the absolute value sign indicates that its amplitude is taken.

[0107] The lowest frequency that satisfies the following formula is defined as electro-optic. bandwidth :

[0108]

[0109] Record the insertion loss, extinction ratio, thermally tunable bias range, thermal drift, half-wave voltage, half-wave voltage length product, RF reflection, electro-optic response flatness, and other parameters of the complete modulator. bandwidth;

[0110] When a fault occurs in the complete modulator, the corresponding process parameters and device baselines are invoked for hierarchical localization based on the fault type:

[0111] When the insertion loss or extinction ratio is abnormal, check the propagation loss in region D, the bending loss in region E, the M6 ​​reference link, and the M3 metal superposition morphology in sequence. If the propagation loss in region D and the bending loss in region E are both abnormal, check regions B and C first. If regions D and E are normal but the M6 ​​reference link is abnormal, check the coupler, beam splitter / combiner, and port connections first. If the passive reference structure is normal, further check the M3 metal absorption and local superposition morphology.

[0112] When the thermal bias range is insufficient or the stability is abnormal, check the thermally adjustable resistor and temperature coefficient of M2, the heating power-optical response of M4, and the cladding data of the F region in sequence. If the resistance or temperature coefficient of M2 is abnormal, check the thermally adjustable material deposition, pattern integrity, and annealing first. If M2 is normal but the heating power-optical response of M4 is abnormal, check the cladding thickness, thermal isolation conditions, and the relative position of the waveguide and heater first.

[0113] When the half-wave voltage or half-wave voltage length product is abnormal, first check the overlay deviation of the C region and the relative position of the M3 electrode and waveguide; when the above results are normal, then combine the M1 electrode transmission and terminal status to determine the electrode field distribution and effective modulation length.

[0114] When the electro-optic bandwidth or high-frequency response flatness is abnormal, check the scattering parameters, impedance, group delay, termination matching, and test link of M1 in sequence. If M1 also shows increased RF reflection or increased insertion loss, prioritize locating the high-frequency metal and transmission line process. If M1 is normal, further check the microwave-optical speed matching, termination matching, and test link.

[0115] When an open circuit, short circuit, or abnormal contact occurs, check the electrical continuity and cross-step metal condition of M1, M2, and M3 in sequence.

[0116] Based on the above relationships, output the priority abnormal process nodes, associated structure identifiers, suggested retest items, and the direction of process adjustment for the next batch; when multiple nodes are abnormal at the same time, process nodes that appear earliest and can explain subsequent abnormalities are processed first according to the manufacturing order.

[0117] S4 involves binding the test results of each structure obtained in S2 and S3 with their mapped wafer number, chip number, wafer coordinates, exposure batch, etching recipe, deposition batch, annealing batch, and test time to form a process parameter vector. :

[0118]

[0119] in, The film thickness corresponds to the test node; The refractive index of the thin film at the target operating wavelength; This refers to the residual stress of the thin film. The surface roughness characterization values ​​are recorded according to the uniform measurement aperture of the corresponding test nodes; For testing the key dimensions of the graphic; This refers to the roughness of the line edge of the graphic. The measured depth of the first deep etching with lithium niobate; The measured depth of the second shallow etching of lithium niobate; This is the combined value of the overlay deviation in the X and Y directions; The waveguide propagation loss coefficient per unit length extracted from region D; The additional loss for a single 180° U-turn under a bending radius R; The residual step height of the cladding across the waveguide on the dielectric; For thermally adjustable electrode resistors; The temperature coefficient of resistance of the thermally adjustable electrode; These are the reflection and scattering parameters at the input port of the high-frequency electrode; Forward transport scattering parameters of the high-frequency electrode; The half-wave voltage of the complete Mach-Zehnder modulator; For the normalized electro-optic response of the complete modulator bandwidth;

[0120] Trend comparisons are performed on process parameter vectors at different locations within the same wafer, different wafers within the same batch, and between different batches. Coordinated display of results for regions A to F and M1 to M6 within the same wafer is generated, producing film thickness maps, critical dimension maps, etching depth maps, overlay maps, propagation loss maps, bending loss maps, cladding step maps, and resistance maps. For multiple wafers within the same batch, the wafer mean, wafer standard deviation, and inter-wafer differences are calculated separately. For consecutive batches, process control limits and drift trends are established.

[0121] For historically stable batch parameters, let the mean be... Standard deviation is Establish the following statistical control limits:

[0122]

[0123]

[0124] in, These are control coefficients set according to the process risk level; The statistical upper control limit for the monitored process parameters is determined based on historical stable batch data; Establish statistical lower control limits for the corresponding parameters; establish corresponding control limits for each parameter in the process parameter vector;

[0125] When a parameter has not exceeded the control limit but continues to drift in the same direction, a trend warning is output, and the batch is marked as a warning, increasing the sampling quantity for the next node; when multiple consecutive batches of the same parameter drift in the same direction, maintenance is performed on the equipment cavity status, consumable life, formula calibration, and measurement system before the final device fails; when a parameter exceeds the control limit, the corresponding subsequent process is stopped or the formula for the next batch is corrected; when multiple related parameters drift simultaneously, the priority inspection node is determined according to the manufacturing sequence and parameter causal relationship.

[0126] The beneficial effects of this invention are as follows:

[0127] 1. Substitute the extracted unit length propagation loss back into the correction bending test matrix, subtract the propagation loss introduced by the different total length of the center line of each structure, and then fit the number of bends to directly obtain the additional loss of a single bend and its variation with the bending radius, so that the propagation loss and bending loss are no longer mixed with each other, and the systematic error caused by only comparing the total insertion loss is eliminated.

[0128] 2. The waveguide width, bending radius, number of bends, optical interface and port direction of each test waveguide are consistent, and only the length of the straight segment is different. This makes the coupling loss and fixed bending loss fall into the intercept of the insertion loss - total length of the center line fitting. The propagation loss per unit length is given directly by the slope. There is no need to arrange resonators, and it does not depend on the quality factor and coupling state. Propagation loss parameters with better repeatability can be obtained in the process stage where it is not suitable to arrange resonators.

[0129] 3. Based on the baseline relationship between independent high-frequency electrodes, independent thermally tunable electrodes, waveguide-electrode superposition structure, and complete Mach-Zehnder modulator, the insertion loss, extinction ratio, half-wave voltage, thermal bias range, and electro-optic bandwidth anomalies of the complete modulator are decomposed and located to the corresponding process nodes according to the manufacturing sequence. Priority inspection objects can be determined without slicing or repeated fabrication to check each item, thus improving the efficiency of failure analysis.

[0130] 4. Connect the testing of process nodes such as thin film deposition, pattern transfer, lithium niobate deep and shallow etching, cladding and planarization, electrodes and complete modulator according to the actual manufacturing sequence, and release them together by setting parameters at critical nodes. If the preceding step is abnormal, stop the subsequent high-cost process and give the adjustment direction to reduce ineffective processing and repeated fabrication. Attached Figure Description

[0131] Figure 1 This is a flowchart illustrating an embodiment of the present invention.

[0132] Figure 2 A schematic diagram of the test structure group contributing to the propagation loss of the fixed bending in region D.

[0133] Figure 3 A schematic diagram of a test matrix for changing the number of bends while maintaining a fixed bending radius in region E. Detailed Implementation

[0134] The relevant technical solutions will now be clearly and completely described with reference to the accompanying drawings of the embodiments of the present invention. The described embodiments are only a part of the embodiments, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0135] like Figure 1 As shown in this embodiment, the multi-process node testing and monitoring method for thin-film lithium niobate modulators disclosed includes the following steps:

[0136] S1. Within the same process batch, set up a process control monitoring area and a device-level verification area, and establish a mapping relationship between structure identifiers and process information for each test structure; the specific steps are as follows:

[0137] This embodiment uses a lithium niobate-on-insulator wafer, which includes an upper thin-film lithium niobate layer, a silicon dioxide buried oxide layer, and a supporting substrate. The wafer crystal orientation, lithium niobate film thickness, buried oxide layer thickness, and supporting substrate type are determined according to the target modulation mode and electrode orientation. Preferably, an X-cut thin-film lithium niobate wafer is used, and the modulation arm direction, radio frequency electric field direction, and the electro-optic coefficient used are matched.

[0138] S1-1. Within the same process batch, establish a process control monitoring area and a device-level verification area; the process control monitoring area is formed using the same hard mask deposition, exposure, lithium niobate etching, cladding, and metallization processes as the final device, including areas A to F; the device-level verification area includes structures M1 to M6; such as... Figure 1 and Figure 2 As shown, area A to area F constitutes the process control and monitoring link, and M1 to M6 constitute the device-level verification link.

[0139] In the process control and monitoring area, Zone A is used for testing the film thickness, refractive index, stress, surface roughness, and on-wafer uniformity of hard masks or dielectric films; Zone B includes isolated lines, dense lines, long-distance line edge roughness test strips, and micro-loading arrays with different spacings, used to evaluate exposure, development, and hard mask pattern transfer; Zone C includes deep etching windows, shallow etching windows, step measurement grooves, X and Y direction overlay scales, double-groove sacrificial waveguides, and interlayer relational units, used to evaluate lithium niobate etching depth, sidewalls, undercuts, redeposition, and deep / shallow etching alignment; Zone D includes non-resonant long-path waveguide groups with fixed bending contributions and different centerline lengths, used to extract propagation loss; Zone E includes waveguide matrices with different bending numbers under fixed bending radii, used to extract single-bend additional loss; Zone F includes dielectric film thickness blocks, arrays with different pattern densities, and trans-waveguide overlay structures, used to evaluate upper cladding film thickness, surface roughness, planarization density effect, and trans-waveguide steps;

[0140] In the device-level verification area, M1 is an independent high-frequency metal electrode or transmission line structure used to measure at least one of the following: metal thickness, linewidth, sheet resistance, open / short circuit, scattering parameters, characteristic impedance, and group delay; M2 is an independent thermally tunable electrode structure used to measure thermally tunable electrode resistance, on-chip uniformity, temperature coefficient, and power withstand capability; M3 is a waveguide and high-frequency metal electrode superposition structure used to check the continuity of the high-frequency metal across the cladding step, the relative position of the electrode and waveguide, residual adhesive, bridging, and local metal absorption risk; M4 is a waveguide and thermally tunable electrode superposition structure used to measure heater resistance, heating power, thermal stability, and additional optical losses introduced by heating; M5 is a complete Mach-Zehnder modulator; and M6 is a coupler, straight waveguide, bent waveguide, beam splitter / combiner, or bypass reference structure that uses the same or equivalent optical interface as the complete modulator.

[0141] In this embodiment, the first etching layer of lithium niobate is used to form the main waveguide and the synchronous deep etching structure, with a preferred target depth of 300 nm ± 20 nm; the second etching layer is used to form a local shallow etching structure, preferably about 100 nm ± 15 nm; the cladding layer on the dielectric is preferably about 1.0 μm of silicon dioxide; the high-frequency metal preferably includes an adhesion layer and a gold layer, with the gold layer thickness being about 500 nm; the thermally tunable electrode is preferably about 100 nm of titanium nitride. All of the above parameters can be adjusted according to the film thickness, waveguide cross-section, electrode simulation, and manufacturing capabilities.

[0142] S1-2. Assign a unique structure identifier to each test structure and establish a mapping relationship between the structure identifier and wafer number, chip number, wafer coordinates, layout layer number, process node, test equipment, test conditions, process recipe version, and judgment rules. The structure identifier preferably consists of a region code, structure serial number, wafer coordinates, and version number to ensure that test data can be traced back to a specific layout and process recipe. Set unified positioning marks for the input and output ports of areas D and E, and ensure that the coupler structure, orientation, process parameters, and surrounding environment of the port are consistent when adding coupling interfaces.

[0143] S2. Perform parameter tests on the test structures corresponding to each process node in the process control monitoring area set in S1 according to the manufacturing sequence, and compare the test results with the preset control conditions to determine whether to proceed to the next process node; the specific steps are as follows:

[0144] This step is carried out in the actual manufacturing sequence of thin film deposition, pattern transfer, lithium niobate deep and shallow etching, passive optical testing, cladding and planarization. After the parameter test of each process node is completed, it is compared with the preset control conditions. If the control conditions of the previous node are not met, the subsequent high-cost process is stopped and the adjustment direction is given.

[0145] S2-1, Thin film deposition node testing and first process release;

[0146] After cleaning, dehydration and surface pretreatment of the lithium niobate wafer on the insulator are completed, a hard mask layer for etching lithium niobate is deposited; the hard mask is preferably silicon nitride, but may also be other media or metal materials that match the etching selectivity of lithium niobate.

[0147] In area A, a main measurement window and multiple on-chip sampling points are selected to measure the hard mask film thickness, refractive index at the target working wavelength, residual stress, root mean square surface roughness, and on-chip uniformity. The film thickness and refractive index are preferably fitted using a multi-wavelength elliptic model, the residual stress is preferably calculated by the change in wafer curvature before and after deposition, and the surface roughness is preferably reported as the root mean square value under the same scanning range and the same filtering conditions. Wafer map mapping is performed on the results at different locations to identify center-edge differences and local anomalies.

[0148] For the i-th sampling point, the film thickness deviation Represented as:

[0149]

[0150] in, For the first The measured film thickness at each sampling point Target film thickness;

[0151] Intra-film relative non-uniformity Calculate using the following formula:

[0152]

[0153] in, This represents the maximum film thickness at the sampling point. This represents the minimum film thickness at the sampling point. This represents the average film thickness at the sampling points;

[0154] Compare the test results in area A with the preset process window or statistical control limits:

[0155] When at least one of the key parameters, such as film thickness, refractive index, stress, roughness, or intra-wafer uniformity, fails to meet the control conditions, record the abnormal location, batch, and equipment status, stop subsequent high-precision patterning, and adjust the deposition temperature, gas flow rate, RF power, deposition time, or pretreatment conditions.

[0156] When all key parameters meet the control conditions, proceed to the graphic transfer node test.

[0157] S2-2, Graphical transfer node test and second process release.

[0158] After coating, exposure, development and hard mask etching of the first deep etching layer of lithium niobate and the pattern layer processed simultaneously, the key dimensions of isolated patterns, dense patterns and patterns with different spacings were measured in area B, and the edge roughness and line width roughness of long-distance patterns were statistically analyzed.

[0159] In this embodiment, at least multiple measurement locations are selected for each design linewidth; isolated lines and dense lines use the same imaging magnification, acceleration voltage, threshold algorithm and edge extraction rules; long-distance line edge roughness test strips are sampled in segments along the propagation direction and the end transition area is removed; the micro-loading array is sorted according to the graphic spacing or duty cycle, and the local size and residual changes are compared.

[0160] The critical dimensional deviations for isolated and dense structures are expressed as follows:

[0161]

[0162]

[0163] The dense-isolation effect can be expressed as:

[0164]

[0165] in, Design the dimensions for the layout; The deviation of the measured critical dimensions of the isolated structure from the layout design dimensions; For the measured critical dimensions of the isolated structure; The deviation of the measured key dimensions of the dense structure from the layout design dimensions; For the measured key dimensions of dense structures; The difference in critical dimension deviation between dense and isolated structures with the same design dimensions is used to characterize the dense-isolation effect.

[0166] For edge position sequences extracted along the length direction The roughness of the line edge can be characterized by the standard deviation of the edge position or three times its standard deviation:

[0167]

[0168] in, Line edge roughness is expressed as three times the standard deviation of the samples at the edge location; The total number of edge location sampling points included in the statistics, taken as an integer greater than 1; This is the arithmetic mean of the sampled values ​​at the aforementioned edge locations;

[0169] Simultaneously inspect for bridging, fractures, residual adhesive, bottom residue, and localized micro-loading. If a dense structure exhibits systematic dimensional shrinkage or expansion relative to an isolated structure, adjust the exposure proximity compensation and hard mask etching conditions accordingly.

[0170] When critical dimension deviations, dense-isolated differences, line edge roughness, or local defects do not meet the preset control conditions, lithium niobate etching is stopped, and the exposure dose, focus conditions, development time, hard mask etching gas ratio, or bias power are adjusted back; when the control conditions are met, the lithium niobate deep and shallow etching node test is initiated.

[0171] S2-3, Lithium niobate deep and shallow etching node test and third process release.

[0172] After completing the first deep etching of lithium niobate, the deep etching window and step measurement groove in region C are subjected to contour scanning or atomic force microscopy, and the double-groove sacrificial waveguide is subjected to cross-sectional or tilt scanning electron microscopy to obtain the deep etching depth, sidewall angle, undercut, redeposition and waveguide edge morphology. The sidewall angle is preferably measured on both sides of the waveguide and the average value is taken. The undercut and redeposition are preferably recorded in sections at the top of the waveguide, the middle of the sidewall and the bottom of the trench.

[0173] After completing the second shallow etching of lithium niobate, the depth of the shallow etching window is tested, and the fine and coarse ranges in the X and Y directions are read and scaled: first, the coarse range is used to determine the direction and order of magnitude of the deviation, and then the fine range is used to obtain the precise deviation.

[0174] Let the overlay deviations in the X and Y directions be respectively and Then the composite overlay deviation Represented as:

[0175]

[0176] Examine separated, overlapping, intersecting, and enclosing cells for local residues, over-etching, mask offsets, and edge collapse to confirm that local exposure, etching, and cleaning do not cause abnormal residues.

[0177] The deep etching depth, shallow etching depth, overlay deviation, sidewall angle, undercut, and redeposition are used as joint release conditions. When any key parameter fails to meet the control conditions, the upper cladding deposition is stopped, and the etching time, bias power, ion source power, reactive gas ratio, cavity state, resist removal cleaning, or overlay compensation are adjusted according to the anomaly type. When the joint control conditions are met, the D-zone propagation loss and E-zone bending loss are jointly extracted.

[0178] S2-4, D region propagation loss and E region bending loss are extracted jointly.

[0179] like Figure 2 As shown, region D includes at least three non-resonant long-path test waveguides; each test waveguide in region D has the same waveguide width W and the same bending radius. The same number of U-turns With identical input / output optical interfaces and port orientations, only the length of the straight segment is changed to ensure the total length of the centerline of each structure is the same. The difference is that the input / output coupling losses of each structure and the additional losses caused by a fixed number of bends remain constant and do not change with the total length of the centerline.

[0180] In this embodiment, the waveguide core width of the test waveguide in region D is preferably 1.2 μm, the bending radius is preferably 100 μm, and the length of the straight segment can be set to approximately 500 μm, 1000 μm, 1500 μm and 3300 μm respectively, so that the total length of the centerline forms at least four separate data points.

[0181] During testing, the tunable laser, polarization controller, input coupling interface, output coupling interface, and optical power detector form an optical test link. Before testing, the input and output interfaces are automatically optimized, and the optimal position, coupling angle, and polarization state are saved. Under the same wavelength grid, polarization state, input power, and automatic alignment rule, each structure in region D is measured at least twice, preferably more than three times. Forward and reverse coupling retests are performed on each structure to identify single-port contamination or end-face defects. The reference channel power is recorded synchronously during the sweep process to correct laser power drift. When a reference optical path is set, the synchronous reference power can be used to normalize the source drift and common coupling drift.

[0182] For the A structure, at wavelength Insertion loss at the point Represented as:

[0183]

[0184] in, Let be the output optical power of the i-th test structure in region D at wavelength λ; For the same structure, at the same wavelength, and under the same test conditions, the input optical power is measured in the same power unit.

[0185] No. The insertion loss obtained from repeated measurements is denoted as . Its mean and standard deviation They are respectively:

[0186]

[0187]

[0188] in, The total number of valid repeated measurements for the same D-region test structure under the same wavelength and test conditions, used to calculate the mean and standard deviation, is an integer not less than 2.

[0189] For each wavelength point, first calculate the mean and standard deviation of repeated measurements, and then perform D-region fitting within the preset working wavelength or working band.

[0190] The average insertion loss of each structure in region D is linearly fitted to the total length of the centerline:

[0191]

[0192] in, The intercept is formed by input / output coupling loss, fixed bending loss, and other fixed losses; Propagation loss per unit length; The total propagation path length is calculated along the centerline of the i-th test waveguide in region D, including the centerline lengths of straight and curved segments, and uses a length unit that matches the propagation loss per unit length. Residuals caused by measurement and structural discretization;

[0193] When using the least squares method, the estimated propagation loss value for:

[0194]

[0195] in, The arithmetic mean of the total length of the centerline of each test waveguide participating in the same linear fitting in region D; This is the arithmetic mean of the average insertion loss of each test waveguide at the same wavelength, where the average insertion loss of each waveguide is calculated from its repeated measurement results;

[0196] Intercept estimate for:

[0197]

[0198] Based on structural repeatability, the following can be adopted: Weighted least squares fitting is performed as the weights; when individual structures have isolated defects but the remaining structures satisfy linear relationships, robust regression can also be used for verification while retaining the original data.

[0199] The quality of the fit can be represented by the coefficient of determination:

[0200]

[0201] in, The coefficient of determination for the linear fit of insertion loss to centerline length is equal to 1 minus the ratio of the sum of squared residuals to the sum of squared total deviations. It characterizes the extent to which the fitted model explains the variation in insertion loss. This statistic is dimensionless and is used in conjunction with the value of the bending radius (described later). Different meanings;

[0202] When the coefficient of determination, residual distribution, or repeated measurement standard deviation does not meet the preset conditions, the abnormal structure is retested: first, automatic alignment is re-executed; second, the input / output coupling end, waveguide path, and local particles are inspected using a microscope, and waveguide breaks and port alignment are checked; if it is confirmed to be an isolated defect, the original results of the structure are retained in the report, and robust regression is used to give the verification propagation loss.

[0203] When the fit quality meets the condition, As a correction factor for the propagation length in region E.

[0204] The propagation loss results in region D are also correlated with the edge roughness in region B and the sidewall morphology in region C, and are used to determine whether the increase in propagation loss has a basis in the process morphology.

[0205] like Figure 3 As shown, region E is grouped according to the bending radius R, and each radius group includes at least three U-turns. Different test waveguides; each E-region structure adopts the same or equivalent waveguide width, optical interface, port orientation and test conditions as the D-region.

[0206] In this embodiment, there are three bending radius groups in region E: 50 μm, 100 μm, and 200 μm. The number of U-turns in the 50 μm and 100 μm groups can be set to 0, 4, 8, and 12, and the number of U-turns in the 200 μm group can be set to 0, 2, 4, and 6.

[0207] Measure the total insertion loss of each structure and the total length of the centerline Propagation length correction was performed using the same wafer, the same polarization state, and the same wavelength for the D-region propagation loss.

[0208]

[0209] in, The corrected insertion loss is the value of the j-th test waveguide in the E-region with a bending radius of R, after deducting the propagation loss introduced by the total length of its centerline; this value still includes the fixed interface and transition loss and bending-related loss of the group, and is expressed in dB.

[0210] With a fixed bending radius Within the group, the corrected loss was fitted to the number of U-turns:

[0211]

[0212] in, This is a fixed interface and fixed transition loss for this radius group; radius Additional losses for a single 180° U-turn; The number of 180° U-turns contained in the j-th test waveguide within the group with bending radius R in region E is used as the independent variable for loss fitting of this radius group. For residuals;

[0213] Its least squares estimate for:

[0214]

[0215] in, It is the arithmetic mean of the corrected insertion loss of each test waveguide participating in the fitting within the same bending radius group at the same wavelength. The structural number of the test waveguide within this radius group; This is the arithmetic mean of the number of 180° U-turns contained in each of the test waveguides participating in the fitting of this group;

[0216] Each radius group is fitted separately, without directly merging data from different radii; data is obtained through multiple bending radius groups. The relationship between bending radius and bending radius.

[0217] The additional loss of a single 90° bend is only considered when the U-turn can be equivalent to two independent 90° bends, and the effect of the intermediate connecting segment is negligible or has been corrected. It can be represented as:

[0218]

[0219] When the corrected loss is approximately linear with the number of U-turns, the fitted slope is used as the additional loss of a single 180° bend. When the corrected loss does not satisfy the approximately linear relationship with the number of U-turns, the single bend loss is not forced to be output. Instead, the mode coupling between adjacent bends, the straight length between bends, the excitation of higher-order modes, local etching anomalies, and the consistency of the test interface are checked, and the radius group is marked as not satisfying the condition for the superposition of independent bend losses.

[0220] S2-5, Upper cladding and flattening node testing and fourth process release.

[0221] After cleaning the lithium niobate waveguide, a cladding layer is deposited on the dielectric, preferably a silica cladding layer. Chemical mechanical planarization is then performed depending on the requirements for subsequent metal crossing of steps. The cladding thickness, refractive index, intra-sheet uniformity, and surface roughness are measured in a large-area film thickness block in the F region. Local depressions and erosion are measured in arrays with different pattern densities. Residual steps, surface continuity, and local voids or cracks are measured at the top and sides of the waveguide in a cross-waveguide covering structure.

[0222] Cross-waveguide steps are preferably scanned along the direction perpendicular to the waveguide and along the waveguide direction, respectively, to distinguish between local steps and long-distance undulations.

[0223] Let the height of the reference surface before planarization be... The surface height of the densely packed graphic region is The amount of local depression Represented as:

[0224]

[0225] Let the surface height of the low-density area or adjacent area be... Then the amount of erosion Represented as:

[0226]

[0227] The cladding thickness, surface roughness, residual steps across the waveguide, local depressions, and erosion are used as joint release conditions before metallization. If the control conditions are not met, the deposition thickness, deposition temperature, planarization time, polishing pressure, slurry formulation, or cleaning process are adjusted; if the control conditions are met, electrode and multilayer stacked baseline testing is performed in the device-level verification area.

[0228] The F-region test results were used in conjunction with the subsequent M3 cross-step metal continuity to determine whether planarization was sufficient to support high-frequency metal fabrication.

[0229] S3. Perform baseline testing on the independent electrode structure and waveguide-electrode superposition structure in the device-level verification area set in S1, and perform device testing on the complete Mach-Zehnder modulator; the specific steps are as follows:

[0230] S3-1, Electrode and Multilayer Baseline Testing.

[0231] After the thermally modulated electrode and high-frequency metal electrode were fabricated, the structures of M1 to M4 were tested.

[0232] After completing the titanium nitride thermally tunable electrode, the M2 independent thermally tunable electrode array was tested first: small-signal resistance measurement and graded power loading were performed using the source measurement unit; a preset time was maintained at each power level, and resistance drift and recovery after cooling were recorded; when it was necessary to measure the temperature coefficient, at least two stable temperature points were set on the temperature control platform, and four-wire resistance testing was performed after the sample temperature stabilized.

[0233] For thermally adjustable electrodes, at voltage and current The resistance below and heating power They are represented as follows:

[0234]

[0235]

[0236] When at temperature and The resistance was measured below. and At that time, temperature coefficient It can be represented as:

[0237]

[0238] After completing the high-frequency metal electrode, the M1 independent transmission line was tested: using a high-frequency probe, calibration substrate and vector network analyzer, the reference plane was moved to the probe tip to measure scattering parameters, insertion loss, return loss and group delay.

[0239] M3 checks the continuity of the metal across the cladding step and verifies the relative position between the electrode and the waveguide; M4 synchronously records the output optical power or spectrum when different heating powers are applied to obtain the additional loss introduced by heating and the baseline of thermal stability.

[0240] The results of M1 to M4 are correlated with the topography data of region F: when there is an anomaly in the waveguide step of region F and M3 has a broken wire, unstable contact or abnormal radio frequency transmission, the anomaly is preferentially located at the upper cladding or planarization node; when region F meets the control conditions but M1 or M2 is abnormal, the anomaly is preferentially located at the corresponding metal deposition, patterning, stripping or annealing node.

[0241] S3-2, Complete Mach-Zehnder Modulator Test.

[0242] The M5 complete Mach-Zehnder modulator was tested in the following order: passive optics, thermally modulated bias, low-frequency electro-optics, and high-frequency electro-optics. The M6 ​​reference structure was used for coupling, straight waveguide, beam splitting / combining, and bypass loss correction. Specifically, the common coupling loss and bypass loss were first estimated using the M6 ​​reference structure. Then, thermally modulated power was applied to move the operating point to the preset bias position. Next, the low-frequency voltage was scanned to obtain the transmission curve and the half-wave voltage was fitted. Finally, the electro-optic response was measured after RF calibration and optoelectronic link normalization.

[0243] In low-frequency electro-optic testing, the output power of the Mach-Zehnder modulator... Represented as:

[0244]

[0245] in, To normalize the input power, The interference visibility caused by the imbalance and loss difference between the two arms, The initial phase difference, It is a half-wave voltage;

[0246] By fitting the transmission curve through scanning DC or low-frequency voltage, the following can be obtained: And the half-wave voltage-length product:

[0247]

[0248] in, Effective modulation length;

[0249] Extinction ratio Represented as:

[0250]

[0251] in, The maximum output optical power on the transmission curve of the complete Mach-Zehnder modulator; The minimum output optical power on the same transmission curve is obtained under the same input optical power, wavelength and polarization conditions;

[0252] In high-frequency electro-optic testing, a preset low-frequency reference point is used. Normalize the electro-optic response:

[0253]

[0254] in, The electro-optic amplitude response is normalized relative to a preset low-frequency reference point, and the unit is dB. The frequency of the radio frequency modulation signal applied to the modulator; The complex electro-optic forward transmission response is obtained after normalization of the RF calibration and optoelectronic test link. It is used to characterize the transmission characteristics from RF electrical input to optical modulation output. The absolute value symbol in the formula indicates that its amplitude is taken.

[0255] The lowest frequency that satisfies the following formula is defined as electro-optic. bandwidth :

[0256]

[0257] Record the insertion loss, extinction ratio, thermally tunable bias range, thermal drift, half-wave voltage, half-wave voltage length product, RF reflection, electro-optic response flatness, and other parameters of the complete modulator. bandwidth.

[0258] S3-3, Layered anomaly localization of process control monitoring results, device baseline and complete device results.

[0259] When a fault occurs in the complete modulator, the corresponding process parameters and device baseline are called first according to the fault type, rather than directly attributing the fault to the complete device itself.

[0260] When the insertion loss or extinction ratio is abnormal, check the propagation loss in region D, the bending loss in region E, the M6 ​​reference link, and the M3 metal superposition morphology in sequence. If the propagation loss in region D and the bending loss in region E are both abnormal, check regions B and C first. If regions D and E are normal but the M6 ​​reference link is abnormal, check the coupler, beam splitter / combiner, and port connections first. If the passive reference structure is normal, further check the M3 metal absorption and local superposition morphology.

[0261] When the thermal bias range is insufficient or the stability is abnormal, check the thermally adjustable resistor and temperature coefficient of M2, the heating power-optical response of M4, and the cladding data of the F region in sequence. If the resistance or temperature coefficient of M2 is abnormal, check the thermally adjustable material deposition, pattern integrity, and annealing first. If M2 is normal but the heating power-optical response of M4 is abnormal, check the cladding thickness, thermal isolation conditions, and the relative position of the waveguide and heater first.

[0262] When the half-wave voltage or half-wave voltage length product is abnormal, first check the overlay deviation of the C region and the relative position of the M3 electrode and waveguide; when the above results are normal, then combine the M1 electrode transmission and terminal status to determine the electrode field distribution and effective modulation length.

[0263] When the electro-optic bandwidth or high-frequency response flatness is abnormal, check the scattering parameters, impedance, group delay, termination matching, and test link of M1 in sequence. If M1 also shows increased RF reflection or increased insertion loss, prioritize locating the high-frequency metal and transmission line process. If M1 is normal, further check the microwave-optical speed matching, termination matching, and test link.

[0264] When an open circuit, short circuit, or abnormal contact occurs, check the electrical continuity and cross-step metal condition of M1, M2, and M3 in sequence.

[0265] Based on the above relationships, the system outputs the priority of the abnormal process nodes, associated structure identifiers, suggested retesting items, and the direction of process adjustments for the next batch. When multiple nodes are abnormal simultaneously, the earliest occurring process node that can explain subsequent abnormalities is processed first, according to the manufacturing sequence.

[0266] S4. Bind the structural test results obtained in S2 and S3 with batch and coordinate information to form process monitoring records, trend warnings, and batch feedback; the specific steps are as follows:

[0267] S4-1, Create process monitoring records.

[0268] The test results of each structure are linked to its mapped wafer number, chip number, wafer coordinates, exposure batch, etching recipe, deposition batch, annealing batch, and test time to form a process parameter vector. :

[0269]

[0270] in, The film thickness corresponds to the test node; The refractive index of the thin film at the target operating wavelength; This refers to the residual stress of the thin film. The surface roughness characterization values ​​are recorded according to the uniform measurement aperture of the corresponding test nodes; For testing the key dimensions of the graphic; This refers to the roughness of the line edge of the graphic. The measured depth of the first deep etching with lithium niobate; The measured depth of the second shallow etching of lithium niobate; This is the combined value of the overlay deviation in the X and Y directions; The waveguide propagation loss coefficient per unit length extracted from region D; The additional loss for a single 180° U-turn under a bending radius R; The residual step height of the cladding across the waveguide on the dielectric; For thermally adjustable electrode resistors; The temperature coefficient of resistance of the thermally adjustable electrode; These are the reflection and scattering parameters at the input port of the high-frequency electrode; Forward transport scattering parameters of the high-frequency electrode; The half-wave voltage of the complete Mach-Zehnder modulator; For the normalized electro-optic response of the complete modulator bandwidth;

[0271] S4-2, Trend Comparison and Statistical Control Limits.

[0272] Trend comparisons are performed on process parameter vectors at different locations on the same wafer, different wafers in the same batch, and different batches.

[0273] The results of regions A to F and M1 to M6 at different locations within the same wafer are displayed in coordinate form to generate film thickness maps, critical dimension maps, etching depth maps, overlay maps, propagation loss maps, bending loss maps, upper cladding step maps, and resistance maps. For multiple wafers in the same batch, the mean value within the wafer, the standard deviation within the wafer, and the difference between wafers are calculated separately. For consecutive batches, process control limits and drift trends are established.

[0274] For historically stable batch parameters, let the mean be... Standard deviation is The following statistical control limits can be established:

[0275]

[0276]

[0277] in, These are control coefficients set according to the process risk level; The statistical upper control limit for the monitored process parameters is determined based on historical stable batch data; Establish the statistical lower control limit for the corresponding parameter; establish corresponding control limits for each parameter in the process parameter vector to determine whether it exceeds the statistical control range;

[0278] S4-3, Trend Warning and Batch Feedback.

[0279] When a parameter has not exceeded the control limit but continues to drift in the same direction, a trend warning is output, and the batch is marked as a warning, increasing the sampling quantity for the next node; when multiple consecutive batches of the same parameter drift in the same direction, maintenance is performed on the equipment cavity status, consumable life, formula calibration, and measurement system before the final device fails; when a parameter exceeds the control limit, the corresponding subsequent process is stopped or the formula for the next batch is corrected; when multiple related parameters drift simultaneously, the priority inspection node is determined according to the manufacturing sequence and parameter causal relationship.

[0280] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although detailed descriptions have been provided with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for testing and monitoring multiple process nodes of a thin-film lithium niobate modulator, characterized in that, Includes the following steps: S1. Set up a process control monitoring area and a device-level verification area within the same process batch, and establish a mapping relationship between structural identifiers and process information for each test structure. S2. Perform parameter tests on the test structures corresponding to each process node in the process control monitoring area set in S1 in sequence according to the manufacturing order, and compare the test results with the preset control conditions to determine whether to enter the next process node. S3. Perform baseline testing on the independent electrode structure and waveguide and electrode superposition structure in the device-level verification area set in S1, and perform device testing on the complete Mach-Zehnder modulator. S4. Bind the structural test results obtained in S2 and S3 with batch and coordinate information to form process monitoring records, trend warnings and batch feedback.

2. The multi-process node testing and monitoring method for thin-film lithium niobate modulators as described in claim 1, characterized in that: The process control and monitoring areas include areas A to F. Area A is used for testing the thickness, refractive index, stress, surface roughness, and on-wafer uniformity of hard masks or dielectric films. Area B includes isolated lines, dense lines, long-distance line edge roughness test strips, and micro-loading arrays with different spacings, used to evaluate exposure, development, and hard mask pattern transfer. Area C includes deep etching windows, shallow etching windows, step measurement grooves, X and Y direction overlay scales, double-groove sacrificial waveguides, and interlayer relational units, used to evaluate lithium niobate etching depth, sidewalls, undercuts, redeposition, and deep / shallow etching alignment. Area D includes non-resonant long-path waveguide groups with fixed bending contributions and different centerline lengths, used to extract propagation loss. Area E includes waveguide matrices with different bending numbers under a fixed bending radius, used to extract single-bend additional loss. Area F includes dielectric film thickness blocks, arrays with different pattern densities, and trans-waveguide overlay structures, used to evaluate upper cladding film thickness, surface roughness, planarization density effect, and trans-waveguide steps. The device-level verification area includes structures M1 to M6; M1 is an independent high-frequency metal electrode or transmission line structure used to measure at least one of the following: metal thickness, linewidth, sheet resistance, open / short circuit, scattering parameters, characteristic impedance, and group delay; M2 is an independent thermally tunable electrode structure used to measure thermally tunable electrode resistance, on-chip uniformity, temperature coefficient, and power withstand capability; M3 is a waveguide and high-frequency metal electrode superimposed structure used to check the continuity of the high-frequency metal across the cladding step, the relative position of the electrode and the waveguide, residual adhesive, bridging, and local metal absorption risk; M4 is a waveguide and thermally tunable electrode superimposed structure used to measure heater resistance, heating power, thermal stability, and additional optical losses introduced by heating; M5 is a complete Mach-Zehnder modulator; and M6 is a coupler, straight waveguide, bent waveguide, beam splitter / combiner, or bypass reference structure that uses the same or equivalent optical interface as the complete modulator. In S1, a unique structure identifier is assigned to each test structure, and a mapping relationship is established between the structure identifier and the wafer number, chip number, wafer coordinates, layout layer number, process node, test equipment, test conditions, process recipe version and judgment rules. The structural identifier preferably consists of a region code, structural serial number, wafer coordinates, and version number.

3. The multi-process node testing and monitoring method for thin-film lithium niobate modulators as described in claim 1, characterized in that, In S2, the manufacturing process is carried out in the following order: thin film deposition, pattern transfer, lithium niobate deep and shallow etching, passive optical testing, cladding and planarization. After the parameter test is completed at each process node, the results are compared with the preset control conditions. If the control conditions are not met at the previous node, the subsequent process is stopped and the adjustment direction is given.

4. The multi-process node testing and monitoring method for thin-film lithium niobate modulators as described in claim 3, characterized in that, In the thin film deposition node test, the main measurement window and multiple on-chip sampling points are selected in area A to measure the hard mask film thickness, refractive index at the target working wavelength, residual stress, root mean square surface roughness and on-chip uniformity. For the i-th sampling point, the film thickness deviation Represented as: in, For the first The measured film thickness at each sampling point Target film thickness; Intra-film relative non-uniformity Calculate using the following formula: in, This represents the maximum film thickness at the sampling point. This represents the minimum film thickness at the sampling point. This represents the average film thickness at the sampling points; When at least one of the key parameters, such as film thickness, refractive index, stress, roughness, or intra-film uniformity, fails to meet the control conditions, the abnormal location, batch, and equipment status are recorded. Subsequent high-precision patterning is stopped, and the deposition temperature, gas flow rate, RF power, deposition time, or pretreatment conditions are adjusted.

5. The multi-process node testing and monitoring method for thin-film lithium niobate modulators as described in claim 3, characterized in that, In the graphic transfer node test, the key dimensions of isolated graphics, dense graphics and graphics with different spacings were measured in area B, and the edge roughness and line width roughness of long-distance graphics were statistically analyzed. The critical dimensional deviations for isolated and dense structures are expressed as follows: The dense-isolation effect is expressed as: in, Design the dimensions for the layout; The deviation of the measured critical dimensions of the isolated structure from the layout design dimensions; For the measured critical dimensions of the isolated structure; The deviation of the measured critical dimensions of the dense structure from the layout design dimensions; For the measured key dimensions of the dense structure; This represents the difference in critical dimension deviations between dense and isolated structures under the same design dimensions. For edge position sequences extracted along the length direction The roughness of the line edge is characterized by three standard deviations at the edge location: in, Line edge roughness is expressed as three times the standard deviation of the samples at the edge location; The total number of edge location sampling points included in the statistics, taken as an integer greater than 1; This is the arithmetic mean of the sampled values ​​at the aforementioned edge locations; When critical dimension deviations, dense-isolated differences, line edge roughness, or local defects do not meet the preset control conditions, lithium niobate etching is stopped, and the exposure dose, focus conditions, development time, hard mask etching gas ratio, or bias power are adjusted.

6. The multi-process node testing and monitoring method for thin-film lithium niobate modulators as described in claim 3, characterized in that, In the deep and shallow etching node testing of lithium niobate After completing the first deep etching of lithium niobate, contour scanning or atomic force microscopy was performed on the deep etching window and step measurement groove in region C, and cross-sectional or tilt scanning electron microscopy was performed on the double-groove sacrificial waveguide to obtain the deep etching depth, sidewall angle, undercut, redeposition, and waveguide edge morphology. After completing the second shallow etching of lithium niobate, the depth of the shallow etching window was measured, and the fine and coarse range scales in the X and Y directions were read. Let the overlay deviations in the X and Y directions be respectively and Then the composite overlay deviation Represented as: For separated, overlapping, intersecting, and enclosing units, check for local residues, over-etching, mask offset, and edge collapse; use deep etching depth, shallow etching depth, overlay deviation, sidewall corner, undercut, and redeposition as joint release conditions; when any key parameter does not meet the control conditions, stop the upper cladding deposition, and adjust the etching time, offset power, ion source power, reactive gas ratio, cavity state, resist removal cleaning, or overlay compensation according to the anomaly type.

7. The multi-process node testing and monitoring method for thin-film lithium niobate modulators as described in claim 3, characterized in that, In passive optical testing, Propagation loss extraction in region D includes setting up at least three non-resonant long-path test waveguides in region D; each test waveguide has the same waveguide width W and the same bending radius. The same number of U-turns With identical input / output optical interfaces and port orientations, differing only in the length of the straight segment, the total length of the centerline of each structure is [not specified]. different; Under the same wavelength grid, the same polarization state, the same input power, and the same automatic alignment rule, each structure should be measured at least repeatedly. For the A structure, at wavelength Insertion loss at the point Represented as: in, Let be the output optical power of the i-th test structure in region D at wavelength λ; For the same structure, at the same wavelength, and under the same test conditions, the input optical power is measured in the same power unit. No. The insertion loss obtained from repeated measurements is denoted as . Its mean and standard deviation They are respectively: in, The total number of valid repeated measurements for the same D-region test structure under the same wavelength and test conditions, used to calculate the mean and standard deviation, is an integer not less than 2. The average insertion loss of each structure in region D is linearly fitted to the total length of the centerline: in, The intercept is formed by input / output coupling loss, fixed bending loss, and other fixed losses; Propagation loss per unit length; The total propagation path length is calculated along the centerline of the i-th test waveguide in region D, including the centerline lengths of straight and curved segments, and uses a length unit that matches the propagation loss per unit length. Residuals caused by measurement and structural discretization; When using the least squares method, the estimated propagation loss value for: in, The arithmetic mean of the total length of the centerline of each test waveguide participating in the same linear fitting in region D; This is the arithmetic mean of the average insertion loss of each test waveguide at the same wavelength, where the average insertion loss of each waveguide is calculated from its repeated measurement results; Intercept estimate for: use Weighted least squares fitting is performed using the weights; when individual structures have isolated defects but the remaining structures satisfy a linear relationship, robust regression is used for verification while retaining the original data; The quality of the fit is represented by the coefficient of determination: in, The coefficient of determination for the linear fit of insertion loss-centerline length is equal to 1 minus the ratio of the sum of squared residuals to the sum of squared total deviations. When the coefficient of determination, residual distribution, or repeated measures standard deviation does not meet the preset conditions, abnormal structures are retested; when the fit quality meets the conditions, [the following is done]. As a correction factor for the propagation length in region E; Extraction of additional losses from a single bend in region E includes grouping region E according to the bending radius R, with each radius group including at least three U-turns. Different test waveguides; each test waveguide uses the same or equivalent waveguide width, optical interface, port orientation and test conditions as region D; Measure the total insertion loss of each structure and the total length of the centerline And use the propagation loss obtained in region D to perform propagation length correction: in, The corrected insertion loss is the value of the j-th test waveguide in the E-region with a bending radius of R, after deducting the propagation loss introduced by the total length of its centerline; this value still includes the fixed interface and transition loss and bending-related loss of the group, and is expressed in dB. With a fixed bending radius Within the group, the corrected loss was fitted to the number of U-turns: in, This is a fixed interface and fixed transition loss for this radius group; radius Additional losses for a single 180° U-turn; The number of 180° U-turns contained in the j-th test waveguide within the group with bending radius R in region E is used as the independent variable for loss fitting of this radius group. For residuals; Its least squares estimate for: in, It is the arithmetic mean of the corrected insertion loss of each test waveguide participating in the fitting within the same bending radius group at the same wavelength. The structural number of the test waveguide within this radius group; This is the arithmetic mean of the number of 180° U-turns contained in each test waveguide participating in the fitting of this group; Each radius group is fitted separately, without directly merging data from different radii; data is obtained through multiple bending radius groups. The relationship between bending radius and bending radius. The additional loss of a single 90° bend is only considered when the U-turn is equivalent to two independent 90° bends, and the effect of the intermediate connecting segment is negligible or has been corrected. Represented as: When the corrected loss is approximately linear with the number of U-turns, the fitted slope is used as the additional loss of a single 180° bend. When the corrected loss does not satisfy the approximately linear relationship with the number of U-turns, the single bend loss is not forced to be output. Instead, the mode coupling between adjacent bends, the straight length between bends, the excitation of higher-order modes, local etching anomalies, and the consistency of the test interface are checked, and the radius group is marked as not satisfying the condition for the superposition of independent bend losses.

8. The multi-process node testing and monitoring method for thin-film lithium niobate modulators as described in claim 3, characterized in that, In the upper cladding and planarization node tests, the upper cladding thickness, refractive index, intra-sheet uniformity and surface roughness were measured in a large area film thickness block in the F region; local depressions and erosions were measured in arrays with different pattern density; and residual steps, surface continuity and local voids or cracks were measured at the top and sides of the waveguide in the cross-waveguide covering structure. Let the height of the reference surface before planarization be... The surface height of the densely packed graphic region is The amount of local depression Represented as: Let the surface height of the low-density area or adjacent area be... Then the amount of erosion Represented as: The cladding thickness, surface roughness, residual steps across the waveguide, local depressions, and erosion are used as joint release conditions before metal processing; if the control conditions are not met, the deposition thickness, deposition temperature, planarization time, polishing pressure, slurry formulation, or cleaning process are adjusted.

9. The multi-process node testing and monitoring method for thin-film lithium niobate modulators as described in claim 1, characterized in that, Baseline testing in S3 for independent electrode structures and waveguide-electrode stacked structures includes: After the thermally tunable electrode and high-frequency metal electrode were fabricated, the structures from M1 to M4 were tested. For the M2 independent thermally tunable electrode array, small-signal resistance measurement and graded power loading were performed using a source measurement unit. At each power level, a preset time was maintained, and the resistance drift and recovery after cooling were recorded. When it was necessary to measure the temperature coefficient, at least two stable temperature points were set on the temperature control platform, and four-wire resistance testing was performed after the sample temperature stabilized. For thermally adjustable electrodes, at voltage and current The resistance below and heating power They are represented as follows: When at temperature and The resistance was measured below. and At that time, temperature coefficient Represented as: For the M1 independent transmission line, a high-frequency probe, a calibration substrate, and a vector network analyzer were used to move the reference plane to the probe tip and measure scattering parameters, insertion loss, return loss, and group delay. M3 was used to check the continuity of the metal across the cladding step and to verify the relative position between the electrode and the waveguide. M4 was used to simultaneously record the output optical power or spectrum when different heating powers were applied to obtain the heating-introduced additional loss and thermal stability baseline. The results of M1 to M4 are correlated with the topography data of region F: when there is an anomaly in the cross-waveguide step of region F and M3 has a broken line, unstable contact or abnormal radio frequency transmission, the anomaly is preferentially located at the upper cladding or planarization node; when region F meets the control conditions but M1 or M2 is abnormal, the anomaly is preferentially located at the corresponding metal deposition, patterning, stripping or annealing node. Device testing of the complete Mach-Zehnder modulator in S3 includes: The M5 complete Mach-Zehnder modulator was tested in the following order: passive optics, thermally modulated bias, low-frequency electro-optics, and high-frequency electro-optics. Coupling, straight waveguide, beam splitting / combining, and bypass loss correction were performed using the M6 ​​reference structure. First, the common coupling loss and bypass loss were estimated using the M6 ​​reference structure. Second, thermally modulated power was applied to move the operating point to a preset bias position. Then, the low-frequency voltage was scanned to obtain the transmission curve and the half-wave voltage was fitted. Finally, after completing RF calibration and optoelectronic link normalization, the electro-optic response was measured. In low-frequency electro-optic testing, the output power of the Mach-Zehnder modulator... Represented as: in, To normalize the input power, The interference visibility caused by the imbalance and loss difference between the two arms. The initial phase difference, It is a half-wave voltage; By fitting the transmission curve through scanning DC or low-frequency voltage, the following can be obtained: And the product of half-wave voltage length: in, For effective modulation length; Extinction ratio Represented as: in, The maximum output optical power on the transmission curve of the complete Mach-Zehnder modulator; The minimum output optical power on the same transmission curve is obtained under the same input optical power, wavelength and polarization conditions; In high-frequency electro-optic testing, a preset low-frequency reference point is used. Normalize the electro-optic response: in, The electro-optic amplitude response is normalized relative to a preset low-frequency reference point, and the unit is dB. The frequency of the radio frequency modulation signal applied to the modulator; The complex electro-optic forward transmission response is obtained after normalization of the RF calibration and optoelectronic test link; the absolute value sign indicates that its amplitude is taken. The lowest frequency that satisfies the following formula is defined as electro-optic. bandwidth : Record the insertion loss, extinction ratio, thermally tunable bias range, thermal drift, half-wave voltage, half-wave voltage length product, RF reflection, electro-optic response flatness, and other parameters of the complete modulator. bandwidth; When a fault occurs in the complete modulator, the corresponding process parameters and device baselines are invoked for hierarchical localization based on the fault type: When the insertion loss or extinction ratio is abnormal, check the propagation loss in region D, the bending loss in region E, the M6 ​​reference link, and the M3 metal superposition morphology in sequence. If the propagation loss in region D and the bending loss in region E are both abnormal, check regions B and C first. If regions D and E are normal but the M6 ​​reference link is abnormal, check the coupler, beam splitter / combiner, and port connections first. If the passive reference structure is normal, further check the M3 metal absorption and local superposition morphology. When the thermal bias range is insufficient or the stability is abnormal, check the thermally adjustable resistor and temperature coefficient of M2, the heating power-optical response of M4, and the cladding data of the F region in sequence. If the resistance or temperature coefficient of M2 is abnormal, check the thermally adjustable material deposition, pattern integrity, and annealing first. If M2 is normal but the heating power-optical response of M4 is abnormal, check the cladding thickness, thermal isolation conditions, and the relative position of the waveguide and heater first. When the half-wave voltage or half-wave voltage length product is abnormal, first check the overlay deviation of the C region and the relative position of the M3 electrode and waveguide; when the above results are normal, then combine the M1 electrode transmission and terminal status to determine the electrode field distribution and effective modulation length. When the electro-optic bandwidth or high-frequency response flatness is abnormal, check the scattering parameters, impedance, group delay, termination matching, and test link of M1 in sequence. If M1 also shows increased RF reflection or increased insertion loss, prioritize locating the high-frequency metal and transmission line process. If M1 is normal, further check the microwave-optical speed matching, termination matching, and test link. When an open circuit, short circuit, or abnormal contact occurs, check the electrical continuity and cross-step metal condition of M1, M2, and M3 in sequence. Based on the above relationships, output the priority abnormal process nodes, associated structure identifiers, suggested retest items, and the direction of process adjustment for the next batch; when multiple nodes are abnormal at the same time, process nodes that appear earliest and can explain subsequent abnormalities are processed first according to the manufacturing order.

10. The multi-process node testing and monitoring method for thin-film lithium niobate modulators as described in claim 1, characterized in that, S4 involves binding the test results of each structure obtained in S2 and S3 with their mapped wafer number, chip number, wafer coordinates, exposure batch, etching recipe, deposition batch, annealing batch, and test time to form a process parameter vector. : in, The film thickness corresponds to the test node; The refractive index of the thin film at the target operating wavelength; This refers to the residual stress of the thin film. The surface roughness characterization values ​​are recorded according to the uniform measurement aperture of the corresponding test nodes; For testing the key dimensions of the graphic; This refers to the roughness of the line edge of the graphic. The measured depth of the first deep etching with lithium niobate; The measured depth of the second shallow etching of lithium niobate; This is the combined value of the overlay deviation in the X and Y directions; The waveguide propagation loss coefficient per unit length extracted from region D; The additional loss for a single 180° U-turn under a bending radius R; The residual step height of the cladding across the waveguide on the dielectric; For thermally adjustable electrode resistors; The temperature coefficient of resistance of the thermally adjustable electrode; These are the reflection and scattering parameters at the input port of the high-frequency electrode; Forward transport scattering parameters of the high-frequency electrode; The half-wave voltage of the complete Mach-Zehnder modulator; For the normalized electro-optic response of the complete modulator bandwidth; Trend comparisons are performed on process parameter vectors at different locations within the same wafer, different wafers within the same batch, and between different batches. Coordinated display of results for regions A to F and M1 to M6 within the same wafer is generated, producing film thickness maps, critical dimension maps, etching depth maps, overlay maps, propagation loss maps, bending loss maps, cladding step maps, and resistance maps. For multiple wafers within the same batch, the wafer mean, wafer standard deviation, and inter-wafer differences are calculated separately. For consecutive batches, process control limits and drift trends are established. For historically stable batch parameters, let the mean be... Standard deviation is Establish the following statistical control limits: in, These are control coefficients set according to the process risk level; The statistical upper control limit for the monitored process parameters is determined based on historical stable batch data; Establish statistical lower control limits for the corresponding parameters; establish corresponding control limits for each parameter in the process parameter vector; When a parameter has not exceeded the control limit but continues to drift in the same direction, a trend warning is output, and the batch is marked as a warning, increasing the sampling quantity for the next node; when multiple consecutive batches of the same parameter drift in the same direction, maintenance is performed on the equipment cavity status, consumable life, formula calibration, and measurement system before the final device fails; when a parameter exceeds the control limit, the corresponding subsequent process is stopped or the formula for the next batch is corrected; when multiple related parameters drift simultaneously, the priority inspection node is determined according to the manufacturing sequence and parameter causal relationship.