Chip packaging TLCP molecular design method suitable for low-temperature BGA solder balls

Abstract

The invention relates to the technical field of semiconductor packaging materials, and particularly discloses a TLCP molecular design method suitable for chip packaging of a low-temperature BGA solder ball, and the method comprises the following specific steps: 1, constructing a target-driven TLCP molecular structure space; 2, establishing a molecular digital characterization and performance prediction model; 3, performing multi-objective reverse design and optimization screening; 4, structure-activity relationship analysis and experimental verification are carried out; a target driving molecular structure space with the melting temperature and the dielectric constant as core performance indexes is constructed, a quantitative prediction model between a TLCP molecular structure and Tm and Dk is established through machine learning, and global search is conducted in a virtual TLCP molecular library through a multi-target optimization algorithm. The Pareto optimal molecular structure capable of realizing synergistic balance between high Tm and low Dk is obtained. The invention further constructs a TLCP material research and development normal form of application requirements, performance objectives, AI reverse design and experimental verification, and the method can be popularized and applied to design of other electronic packaging high polymer materials.

Description

Technical Field

[0001] This invention relates to the field of semiconductor packaging materials technology, specifically to a TLCP molecular design method for chip packaging of low-temperature BGA solder balls. Background Technology

[0002] With the continuous increase in chip integration and operating frequency, advanced packaging technologies such as chiplet heterogeneous integration and 2.5D / 3D packaging are increasingly used in high-performance computing chips (such as CPUs, GPUs, and accelerator chips). The overall performance and long-term reliability of the packaging substrate and interconnect solder joint system have become one of the key factors limiting the overall performance and lifespan of the system. Ball grid array (BGA) solder balls are currently the mainstream packaging interconnect form.

[0003] To reduce thermal damage to the chip body and sensitive materials during packaging processes, lower thermal stress, and improve yield, low-temperature BGA solder balls with low reflow temperatures are gaining increasing attention. Their typical alloy systems (such as Sn-Ag-Cu-Bi and Sn-Ag-Cu-In-Bi) typically have liquidus temperatures below 220°C. Meanwhile, substrate materials, as packaging carriers, undergo multiple thermal cycles during manufacturing and service. To ensure dimensional stability and creep resistance, substrate materials generally require high melting temperatures and good thermal rigidity. Thermotropic liquid crystal polymers (TLCPs), with their high modulus, low coefficient of linear expansion, and excellent heat resistance, are considered important candidate materials for high-end packaging substrates.

[0004] In existing technologies, the modification of TLCP mainly focuses on the following aspects: increasing melting temperature and thermal deformation temperature by adding rigid aromatic rings and coplanar structures; reducing dielectric constant and dielectric loss by introducing fluorine-containing structures, reducing the content of polar groups, or blending with low-dielectric fillers; and improving mechanical properties and processing flowability by adjusting copolymerization ratio and molecular weight. However, these studies mostly focus on optimizing a single property, such as only improving heat resistance or only reducing Dk, making it difficult to achieve synergistic optimization of high melting temperature and low dielectric constant within a unified molecular design framework.

[0005] More importantly, existing TLCP designs are disconnected from specific packaging applications. They often fail to prioritize thermo-mechanical matching with specific low-temperature BGA solder ball systems as the ultimate design goal, making it difficult to address reliability issues such as interface stress concentration and solder joint fatigue crack propagation caused by differences in thermal expansion coefficients between the substrate and solder balls. Furthermore, traditional molecular design methods, which rely on experience and trial-and-error, have limited search efficiency in large-scale molecular structure spaces, resulting in long development cycles, high costs, and an insufficiently systematic understanding of complex structure-property relationships.

[0006] Therefore, it is necessary to provide a TLCP material suitable for low-temperature BGA solder ball chip packaging applications, which can simultaneously achieve high melting temperature and low dielectric constant and match the thermo-mechanical behavior of low-temperature BGA solder balls. At the same time, it is necessary to establish a systematic method for target-driven reverse design using artificial intelligence technology to improve the efficiency and success rate of material development. Summary of the Invention

[0007] The purpose of this invention is to address the shortcomings of existing technologies by providing a TLCP molecular design method for chip packaging suitable for low-temperature BGA solder balls. By introducing packaging application constraints at the molecular design stage, a TLCP resin and substrate material with both high melting temperature and low dielectric constant, and thermo-mechanical behavior that is compatible with low-temperature BGA solder balls, is obtained, thereby improving the overall reliability and high-speed signal transmission performance of high-density chip packaging.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a TLCP molecular design method for chip packaging of low-temperature BGA solder balls, comprising the following steps: Step 1: Constructing the target-driven TLCP molecular structure space: The target is to obtain TLCP molecules with a melting temperature Tm of not less than 315℃ and a dielectric constant Dk of not more than 2.6 at 10 GHz frequency. The application constraint is "to be used in conjunction with low-temperature BGA solder balls with a liquidus temperature of less than 220℃ for chip packaging". Rigid aromatic monomers and fluorinated aromatic dicarboxylic acid monomers are selected to generate a virtual TLCP molecular library with multiple copolymer sequences within a preset molar fraction range. Step 2: Molecular digital characterization and performance prediction model establishment: Molecular descriptors are extracted for each molecular structure in the virtual TLCP molecular library described in Step 1 to obtain a descriptor set including elemental composition, rigid segment ratio, and quantum chemical properties. Combined with a TLCP sample dataset with known Tm and Dk, a machine learning algorithm is used to train a performance prediction model for Tm and Dk. The number of samples in the TLCP sample dataset used to train the performance prediction model shall not be less than 200. Step 3: Multi-objective reverse design and optimization screening: Construct a multi-objective optimization problem with the prediction of Tm as the objective to be maximized and the prediction of Dk as the objective to be minimized. Use a multi-objective optimization algorithm to search the virtual TLCP molecular library in Step 1 to obtain a Pareto optimal candidate molecular structure set that achieves a trade-off between Tm and Dk. Step 4: Structure-activity relationship analysis and experimental verification: The interpretability analysis of the candidate molecular structures in Step 3 is performed to identify the key structural factors affecting Tm and Dk, and the corresponding TLCP resin is synthesized according to the target ratio of the candidate molecular structures. Tm and Dk are measured and compared with the predicted values ​​for verification. If necessary, the measured data are fed back to the model and Step 2 and Step 3 are iterated.

[0009] As a preferred technical solution of the present invention, the molecular backbone in the virtual TLCP molecular library in step 1 is composed of rigid mesocrystalline units provided by p-hydroxybenzoic acid (HBA) and / or 6-hydroxy-2-naphthoic acid (HNA) and fluorinated aromatic dicarboxylic acid units copolymerized, wherein the total molar fraction of HBA and HNA is 70-85 mol%, and the molar fraction of fluorinated aromatic dicarboxylic acid monomer is 15-30 mol%. The fluorinated aromatic dicarboxylic acid monomer is 2,3,5,6-tetrafluoroterephthalic acid (TFTA) or a combination of fluorinated aromatic dicarboxylic acids with TFTA as the main component.

[0010] As a preferred embodiment of the present invention, the molecular descriptor in step 2 includes at least: fluorine atom molar fraction, aromatic ring content, rigid segment ratio, predicted molecular dipole moment, predicted molecular polarizability, and one or more quantum chemical energy level related indicators. The machine learning algorithm used to build the performance prediction model in step 2 is selected from at least one of gradient boosting decision tree, random forest, XGBoost and neural network; the algorithm used for multi-objective optimization in step 3 is selected from at least one of NSGA-II and MOEA / D. The TLCP sample dataset used to train the performance prediction model has no fewer than 100 samples.

[0011] A thermotropic liquid crystal polymer molecular structure for chip packaging, obtained by the design method described above, wherein the molecular structure is an aromatic copolyester, and its main chain contains: 70–85 mol% of rigid mesocrystalline structural units provided by p-hydroxybenzoic acid (HBA) and / or 6-hydroxy-2-naphthoic acid (HNA), and 15–30 mol% of modified structural units provided by fluorinated aromatic dicarboxylic acids; The fluorinated aromatic dicarboxylic acid is 2,3,5,6-tetrafluoroterephthalic acid (TFTA). TLCP resins prepared from molecular structures meet the following performance parameters: The measured melting temperature Tm is 315–335℃; The measured dielectric constant Dk at 10 GHz is 2.3–2.6.

[0012] A thermotropic liquid crystal polymer resin is prepared by melt polycondensation of the above-described thermotropic liquid crystal polymer molecular structure. The melt polycondensation method includes: performing a pre-esterification reaction at 180-240°C, followed by polycondensation at 260-320°C under reduced pressure for 1-3 h to obtain a TLCP melt with liquid crystal properties and high molecular weight, and then extruding and cooling to granulate to obtain TLCP resin particles.

[0013] A substrate material for chip packaging, the substrate material comprising at least the thermotropic liquid crystal polymer resin described above, and prepared into a plate or film by melt extrusion-stretching, calendering or lamination hot pressing, wherein the substrate material has a dielectric constant Dk of 2.3 to 2.6 at a frequency of 10 GHz and an out-of-plane coefficient of thermal expansion of 10 to 22 ppm / ℃ in the range of 25 to 260℃; the substrate material is used in conjunction with low-temperature BGA solder balls with a liquidus temperature below 220℃.

[0014] As a preferred embodiment of the present invention, the substrate material is used in conjunction with Sn-Ag-Cu-Bi or Sn-Ag-Cu-In-Bi alloy solder balls with a liquidus temperature of 200-220°C.

[0015] A high-density chip package includes: a chip body; a substrate material for chip packaging as described above; and low-temperature BGA solder balls soldered to pads on the substrate material, wherein the liquidus temperature of the low-temperature BGA solder balls is below 220°C.

[0016] An electronic device comprising the high-density chip package described above.

[0017] Compared with the prior art, the beneficial effects of the present invention are: Taking "packaging reliability matching low-temperature BGA solder balls" as the starting point for material design, reflow soldering temperature and interfacial thermo-mechanical behavior constraints were introduced at the molecular structure design stage, enabling high melting temperature and low dielectric constant to be synergistically achieved under a unified design framework. A TLCP structure-performance mapping and reverse search system was constructed using machine learning and multi-objective optimization algorithms, significantly improving material development efficiency and reducing trial-and-error costs. By rationally controlling the ratio window of HBA / HNA and TFTA, a TLCP resin with Tm in the range of 315–335℃, Dk in the range of 2.3–2.6 at 10 GHz, and a low out-of-plane coefficient of thermal expansion was obtained, effectively improving the thermal fatigue reliability and high-speed signal integrity of the substrate-low-temperature BGA solder ball interface. The established R&D paradigm of "application requirements → performance targets → AI reverse design → experimental verification" can be extended to the design of other electronic packaging polymer materials and composite materials. Detailed Implementation

[0018] The preferred embodiments of the present invention will now be described in detail so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby providing a clearer and more explicit definition of the scope of protection of the present invention.

[0019] This invention provides a TLCP molecular design method for chip packaging suitable for low-temperature BGA solder balls, comprising the following steps: First, based on the application requirements of chip packaging, the key performance targets of TLCP are established, including at least: a melting temperature Tm of not less than 315℃, preferably in the range of 315 to 335℃, and a dielectric constant Dk of not more than 2.6 at a frequency of 10 GHz, preferably in the range of 2.3 to 2.6. At the same time, the constraint of "using it in conjunction with low-temperature BGA solder balls with a liquidus temperature below 220℃" is set as a constraint, requiring the TLCP substrate to maintain good dimensional stability and interface reliability under typical low-temperature reflow soldering profiles (peak temperature of about 230 to 245℃) and temperature cycling conditions of -55 to 125℃.

[0020] Based on the aforementioned performance targets and application constraints, rigid aromatic monomers with mesocrystalline properties and high coplanarity (such as p-hydroxybenzoic acid HBA and / or 6-hydroxy-2-naphthoic acid HNA) were selected as the main chain backbone units. Simultaneously, fluorinated aromatic dicarboxylic acid monomers were introduced as low-dielectric modification units, with 2,3,5,6-tetrafluoroterephthalic acid (TFTA) being the preferred choice. By presetting a composition window with a total molar fraction of 70–85 mol% for HBA and HNA and a molar fraction of 15–30 mol% for the fluorinated aromatic dicarboxylic acid, a virtual TLCP molecular library containing various copolymer sequences and ratio combinations was automatically generated using an algorithm.

[0021] Based on this, each molecular structure in the virtual TLCP molecular library was digitally characterized, extracting multidimensional molecular descriptors including fluorine atom molar fraction, aromatic ring content, rigid segment ratio, predicted molecular dipole moment, predicted polarizability, and several quantum chemical energy level related indicators. Combining literature data, existing experimental data, and necessary molecular simulation calculations, a TLCP sample dataset containing at least 100, preferably at least 200, samples with known Tm and Dk was constructed. Machine learning algorithms such as gradient boosting decision trees, XGBoost, or neural networks were used to train performance prediction models for Tm and Dk, ensuring the determination coefficient R of the Tm prediction model was [value missing]. 2 Ideally, the coefficient of determination R of the Dk prediction model should be no less than 0.85. 2 The preferred value is not less than 0.80.

[0022] Subsequently, maximizing the predicted Tm and minimizing the predicted Dk were set as mutually constraining dual objectives, constructing a multi-objective optimization problem. A non-dominated sorting genetic algorithm (NSGA-II) or other multi-objective evolutionary algorithms were used to perform a global search in a virtual TLCP molecular library to obtain a set of Pareto front candidate molecular structures with the optimal trade-off between high Tm and low Dk. Further considering packaging process and thermo-mechanical matching requirements, the Pareto front candidate molecules were screened, with candidate molecular structures having a melting temperature greater than 320℃ and a dielectric constant less than 2.6 being preferred.

[0023] By employing interpretable machine learning tools (such as feature importance-based analysis or SHAP value analysis), key structural factors influencing Tm and Dk are identified, including the content of fluorinated units, the proportion of rigid segments, and the proportion of naphthalene rings introduced, thus forming a synergistic molecular design principle of "highly rigid aromatic units + appropriate amounts of fluorinated dicarboxylic acids." Based on the optimized candidate molecular structures, specific copolymerization formulations are designed, and monomers such as HBA, HNA, and TFTA are weighed according to the set ratios. The target TLCP resin is prepared using melt polycondensation, followed by Tm and Dk tests. The measured results are compared with the model predictions. When the deviation exceeds the preset range, the measured data is fed back into the training set, and the prediction model and multi-objective design process are iteratively optimized.

[0024] The TLCP molecular structure obtained by the above AI-aided design method is an aromatic copolyester. The main chain contains 70-85 mol% rigid mesocrystalline structural units provided by HBA and / or HNA, and 15-30 mol% modified structural units provided by fluorinated aromatic diacids, preferably TFTA. The TLCP resin prepared from this molecular structure has a measured melt temperature Tm of 315-335℃, a dielectric constant Dk of 2.3-2.6 at 10 GHz, and preferably a coefficient of out-of-plane linear expansion of 10-22 ppm / ℃ in the range of 25-260℃.

[0025] The TLCP resin can be prepared by melt polycondensation: under an inert atmosphere, HBA, HNA and TFTA esterification precursors are mixed with a catalyst and heated to 180-240°C for pre-esterification reaction. After the byproducts are discharged, the temperature is further increased to 260-320°C and the pressure is gradually reduced to vacuum conditions for polycondensation for 1-3 hours to obtain a TLCP melt with liquid crystal properties and high molecular weight. The melt is cooled, extruded and granulated to obtain TLCP resin particles, which are then processed into films or plates by extrusion-stretching, calendering or lamination hot pressing and other methods, and used as chip packaging substrate materials.

[0026] When the TLCP substrate material described in this invention is used in conjunction with low-temperature BGA solder balls with a liquidus temperature below 220°C, it maintains good flatness and dimensional stability under low-temperature reflow soldering curves and significantly slows down the propagation of interface cracks at solder joints under temperature cycling conditions of -55 to 125°C. At the same time, the substrate has a low dielectric constant and low insertion loss in the high-frequency band, making it suitable for high-speed signal transmission at 40 GHz or even higher frequencies.

[0027] Example 1: AI-assisted TLCP molecular design process In this embodiment, the objective is to design a TLCP resin for matching low-temperature BGA solder balls with a liquidus temperature of approximately 212°C, ensuring its melting temperature Tm is not lower than 315°C, preferably greater than 320°C, and its dielectric constant Dk at 10 GHz is not higher than 2.6. Using HBA, HNA, and TFTA as basic structural units, a composition window is set with a total molar fraction of 70–85 mol% for HBA and HNA and a molar fraction of 15–30 mol% for TFTA. Approximately 1500 copolymer sequences and ratio combinations are automatically generated by a computer program to construct a virtual TLCP molecular library.

[0028] Molecular modeling software was used to construct and geometrically optimize the structures of each molecule in a virtual molecular library, extracting 142-dimensional molecular descriptors including elemental composition, topological structure, and quantum chemical properties. Combining literature data, existing experimental data, and necessary molecular simulation results, a TLCP performance database containing 300 records was established, with each record including the Tm and Dk at 10 GHz for that TLCP. XGBoost algorithm was used to train Tm and Dk prediction models separately, and five-fold cross-validation was used to evaluate model performance. The coefficient of determination R0 of the Tm prediction model was calculated. 2 The coefficient of determination R of the Dk prediction model is approximately 0.91. 2 It is approximately 0.87.

[0029] With the dual objectives of maximizing the predicted Tm and minimizing the predicted Dk, the NSGA-II algorithm was used to perform a multi-generational evolutionary search on a virtual molecular library, obtaining a set of Pareto front candidate molecules. Considering both synthetic feasibility and application requirements, several candidate formulations were selected. One preferred formulation is an HBA / HNA / TFTA molar ratio of 70 / 15 / 15, corresponding to a predicted Tm of approximately 325℃ and a predicted Dk of approximately 2.51.

[0030] Interpretive analysis was used to analyze the structure-activity relationship of the candidate molecular structures. The results showed that: appropriately increasing the TFTA content is beneficial to reducing Dk, but excessively high content will lead to a slight decrease in Tm; the introduction of HNA is beneficial to further increase Tm and slightly reduce Dk; the proportion of mesocrystalline rigid segments and the content of fluorinated units are the key factors determining the balance between Tm and Dk. Based on this, the synergistic design criterion of "high rigidity aromatic unit + appropriate amount of fluorinated dicarboxylic acid" was established.

[0031] Example 2: Preparation and Performance Testing of AI-Designed TLCP Resin According to the preferred formulation obtained in Example 1, with a molar ratio of HBA / HNA / TFTA of 70 / 15 / 15, the hydroxy acids or esterification precursors of the corresponding monomers were weighed out, and antimony acetate catalyst (approximately 150 ppm of the total molar amount of monomers) was added. Under nitrogen protection, the mixture was heated to approximately 200°C for a pre-esterification reaction for about 2 hours. After removing water or lower alcohol byproducts, the temperature was further increased to approximately 280°C, and the reaction was maintained under a nitrogen flow for about 1 hour. Subsequently, the system pressure was gradually reduced to approximately 300 Pa, and melt polycondensation was carried out at 290–300°C for about 2 hours until the system viscosity significantly increased. After stopping the reaction, the melt was extruded, cooled, and granulated to obtain pale yellow TLCP resin particles.

[0032] After drying the obtained TLCP resin, a film with a thickness of approximately 100 μm was prepared under extrusion-calendering conditions. The melting behavior was tested using DSC under a nitrogen atmosphere and a heating rate of 10 °C / min, yielding a Tm of approximately 322 °C. The dielectric constant, Dk, was measured at 10 GHz using the parallel-plate capacitor method combined with a vector network analyzer, resulting in a Dk of approximately 2.55. The measured values ​​showed little deviation from the model predictions (Tm_pred = 325 °C, Dk_pred = 2.51), demonstrating the good accuracy of the AI ​​design method.

[0033] The out-of-plane linear expansion coefficient (CTE_z) in the range of 25–260℃ was further tested using a thermomechanical analyzer, and the CTE_z was found to be approximately 15 ppm / ℃. Under repeated reflow soldering conditions with a simulated peak temperature of approximately 235℃, the warpage of the thin plate was observed. The AI-designed TLCP thin plate remained largely flat with no significant warpage.

[0034] Example 3: Packaging Compatibility and Reliability Verification Using the AI-designed TLCP resin obtained in Example 2 (denoted as AI-TLCP) and a commercially available high-performance TLCP resin (Tm approximately 305°C, Dk approximately 2.8, denoted as Ref-TLCP), a packaging substrate with a thickness of approximately 0.3 mm was prepared under the same process conditions.

[0035] BGA soldering was performed on Al-TLCP and Ref-TLCP substrates using Sn-Ag-Cu-In-Bi low-temperature solder paste with a liquidus temperature of 212℃ and a reflow soldering profile with a peak temperature of approximately 235℃. Scanning electron microscopy results showed that the Al-TLCP substrate exhibited virtually no surface warping and good solder joint wetting after soldering; however, the Ref-TLCP substrate showed micron-level warping and localized stress concentration areas.

[0036] The soldered samples were subjected to 1500 temperature cycles within the range of -55 to 125°C. The crack length at the solder joint interface was observed by metallographic sectioning and scanning electron microscopy. The results showed that the average crack length of the BGA solder joint on the AI-TLCP substrate was approximately 15 μm, while the corresponding crack length on the Ref-TLCP substrate was approximately 28 μm. The AI-TLCP material of this invention significantly reduces the interface crack propagation rate and significantly improves the thermal fatigue resistance of the solder joint.

[0037] In the 40 GHz band, S-parameter tests were performed on microstrip transmission lines fabricated based on AI-TLCP substrates and Ref-TLCP substrates. The results showed that the insertion loss of the AI-TLCP substrate was reduced by about 15-20% compared with that of the Ref-TLCP substrate, indicating that the low dielectric properties of the TLCP of the present invention are beneficial to improving the high-speed signal transmission performance.

[0038] In summary, the TLCP material obtained by the AI-aided design method of this invention can achieve synergistic optimization between high melting temperature and low dielectric constant, and achieve good matching with low-temperature BGA solder balls in terms of thermo-mechanical and electrical properties, meeting the stringent requirements of high-density, high-speed chip packaging.

[0039] The above embodiments merely illustrate implementation methods of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention.

Claims

1. A TLCP molecular design method for chip packaging of low-temperature BGA solder balls, characterized in that, Includes the following steps: Step 1: Constructing the target-driven TLCP molecular structure space: The target is to obtain TLCP molecules with a melting temperature Tm of not less than 315℃ and a dielectric constant Dk of not more than 2.6 at a frequency of 10GHz. The application constraint is "to be used in conjunction with low-temperature BGA solder balls with a liquidus temperature of less than 220℃ for chip packaging". Rigid aromatic monomers and fluorinated aromatic dicarboxylic acid monomers are selected to generate a virtual TLCP molecular library with multiple copolymer sequences within a preset molar fraction range. Step 2: Molecular digital characterization and performance prediction model establishment: Molecular descriptors are extracted for each molecular structure in the virtual TLCP molecular library described in Step 1 to obtain a descriptor set including elemental composition, rigid segment ratio, and quantum chemical properties. Combined with a TLCP sample dataset with known Tm and Dk, a machine learning algorithm is used to train a performance prediction model for Tm and Dk. The number of samples in the TLCP sample dataset used to train the performance prediction model shall not be less than 200. Step 3: Multi-objective reverse design and optimization screening: Construct a multi-objective optimization problem with the prediction of Tm as the objective to be maximized and the prediction of Dk as the objective to be minimized. Use a multi-objective optimization algorithm to search the virtual TLCP molecular library in Step 1 to obtain a Pareto optimal candidate molecular structure set that achieves a trade-off between Tm and Dk. Step 4: Structure-activity relationship analysis and experimental verification: The interpretability analysis of the candidate molecular structures in Step 3 is performed to identify the key structural factors affecting Tm and Dk, and the corresponding TLCP resin is synthesized according to the target ratio of the candidate molecular structures. Tm and Dk are measured and compared with the predicted values ​​for verification. If necessary, the measured data are fed back to the model and Step 2 and Step 3 are iterated.

2. The design method according to claim 1, characterized in that, The molecular backbone in the virtual TLCP molecular library in step 1 is composed of rigid mesocrystalline units provided by p-hydroxybenzoic acid (HBA) and / or 6-hydroxy-2-naphthoic acid (HNA) and fluorinated aromatic dicarboxylic acid units copolymerized, wherein the total molar fraction of HBA and HNA is 70-85 mol%, and the molar fraction of fluorinated aromatic dicarboxylic acid monomer is 15-30 mol%. The fluorinated aromatic dicarboxylic acid monomer is 2,3,5,6-tetrafluoroterephthalic acid (TFTA) or a combination of fluorinated aromatic dicarboxylic acids with TFTA as the main component.

3. The design method according to any one of claims 1-2, characterized in that, The molecular descriptor in step 2 includes at least: fluorine atom mole fraction, aromatic ring content, rigid segment ratio, predicted molecular dipole moment, predicted molecular polarizability, and one or more quantum chemical energy level related indicators. The machine learning algorithm used to build the performance prediction model in step 2 is selected from at least one of gradient boosting decision tree, random forest, XGBoost and neural network; the algorithm used for multi-objective optimization in step 3 is selected from at least one of NSGA-II and MOEA / D. The TLCP sample dataset used to train the performance prediction model has no fewer than 100 samples.

4. A thermotropic liquid crystal polymer molecular structure for chip packaging obtained by the design method according to any one of claims 1-3, characterized in that, The molecular structure is an aromatic copolyester, and its main chain contains: 70–85 mol% of rigid mesocrystalline structural units provided by p-hydroxybenzoic acid (HBA) and / or 6-hydroxy-2-naphthoic acid (HNA), and 15–30 mol% of modified structural units provided by fluorinated aromatic dicarboxylic acids; The fluorinated aromatic dicarboxylic acid is 2,3,5,6-tetrafluoroterephthalic acid (TFTA). TLCP resins prepared from molecular structures meet the following performance parameters: The measured melting temperature Tm is 315–335℃; The measured dielectric constant Dk at 10 GHz is 2.3–2.

6.

5. A thermotropic liquid crystal polymer resin, characterized in that, The thermotropic liquid crystal polymer molecular structure described in claim 4 is prepared by melt polycondensation, wherein the melt polycondensation method includes: performing a pre-esterification reaction at 180-240°C, followed by polycondensation at 260-320°C under reduced pressure for 1-3 h to obtain a TLCP melt with liquid crystal properties and high molecular weight, and then extruding and cooling to granulate to obtain TLCP resin particles.

6. A substrate material for chip packaging, characterized in that, The substrate material comprises at least the thermotropic liquid crystal polymer resin of claim 5, and is prepared into a plate or film by melt extrusion-stretching, calendering or lamination hot pressing. The substrate material has a dielectric constant Dk of 2.3 to 2.6 at 10 GHz and an out-of-plane coefficient of thermal expansion of 10 to 22 ppm / ℃ in the range of 25 to 260℃. The substrate material is used in conjunction with low-temperature BGA solder balls with a liquidus temperature below 220℃.

7. The substrate material for chip packaging according to claim 6, characterized in that, The substrate material is used in conjunction with Sn-Ag-Cu-Bi or Sn-Ag-Cu-In-Bi alloy solder balls with a liquidus temperature of 200-220°C.

8. A high-density chip package, characterized in that, include: Chip body; substrate material for chip packaging according to claim 6 or 7; And low-temperature BGA solder balls soldered onto the pads of the substrate material, wherein the liquidus temperature of the low-temperature BGA solder balls is below 220°C.

9. An electronic device, characterized in that, The electronic device includes the high-density chip package as described in claim 8.