In my research, I investigate how outlier patents can serve as early carriers of disruptive technologies in the solid electrolyte cell industry. The solid electrolyte cell is widely regarded as a next-generation energy storage device because it promises higher energy density, improved safety, and longer cycle life than conventional liquid-electrolyte lithium-ion batteries. However, the transition from incremental improvement to radical replacement is not linear. Disruptive technologies often emerge at the periphery of established technological trajectories, where they appear semantically distant, weakly connected, and economically uncertain. I therefore construct a three-stage progressive early identification model that combines outlier patent recognition, weak-signal detection, and multi-dimensional disruptive potential evaluation. My empirical focus is the solid electrolyte cell industry, and my goal is to show that outlier patents can be systematically converted into actionable intelligence for technology forecasting.
The phrase solid electrolyte cell appears throughout my analysis because the industry is still in a pre-paradigmatic phase. No single material system, manufacturing route, or cell architecture has achieved full consensus. Sulfide, oxide, polymer, thin-film, and composite electrolytes coexist. Lithium-metal anodes remain attractive but difficult to stabilize. Manufacturing costs, interfacial resistance, dendrite growth, and scale-up yield continue to challenge commercialization. In such an environment, mainstream patent clusters may reflect incremental improvements, whereas outlier patents may reveal alternative pathways that are not yet recognized as mainstream. My first-person research position is therefore simple: I treat outlier patents not as noise but as weak early signals of potential disruption in the solid electrolyte cell industry.
1. Introduction
Technological development can be divided into continuous and discontinuous trajectories. Disruptive technologies belong primarily to the discontinuous type because they change the rules of competition and reshape future industrial landscapes. A solid electrolyte cell is a clear example: if it succeeds at scale, it can replace flammable liquid electrolytes, enable lithium-metal anodes, and redefine the value chain of battery manufacturing. Yet the early stage of such a technology is difficult to detect. Patent citations, publication counts, and market reports often lag behind the emergence of radical ideas. I therefore ask how I can identify disruptive technologies in the solid electrolyte cell industry before they become visible in conventional indicators.
My answer is to focus on outlier patents. In statistics, an outlier is an observation that deviates markedly from the overall pattern. In patent analysis, an outlier patent is a patent that occupies a distinctive or abnormal position in technological space. It may be rare, semantically distant from mainstream clusters, and weakly connected to established citation networks. Previous work has often treated outliers as noise, but I treat them as early signals of technological reconfiguration. The solid electrolyte cell industry is particularly suitable for this approach because its knowledge base is fragmented and its dominant design is still unsettled.
I define a disruptive technology as a technology that, in its early formation, deviates from the mainstream trajectory, exhibits high novelty and uncertainty, and can later fundamentally reshape an existing technological system, industrial structure, or market. This definition implies four core characteristics: outlierness, novelty, forward-looking orientation, and potential influence. I use these four characteristics to guide my model. The model consists of three stages: outlier patent identification, disruptive early-signal recognition, and disruptive potential evaluation. Each stage reduces uncertainty while preserving the possibility of detecting weak signals. The empirical setting is the solid electrolyte cell industry, from which I extract high-potential disruptive technologies.
2. Theoretical Background and Research Gap
2.1 Disruptive Technologies
The concept of disruptive technology was popularized in the 1990s and has since been refined by many scholars. Although a unified definition remains elusive, most studies agree that disruptive technologies are transformative and forward-looking. They can produce a zeroing effect, resetting the value of incumbent technologies. They can restructure technological and economic landscapes. They can also become the mainstream direction of a future technology system. In the context of the solid electrolyte cell, disruption would mean not merely a better electrolyte material but a new battery architecture that changes manufacturing, safety standards, and application boundaries.
Existing identification methods can be grouped into expert-based approaches, text-mining approaches, technology-evolution approaches, and mathematical-model approaches. These methods have been applied in many fields and have demonstrated value. However, they also have limitations. Concepts are not always clearly defined. Data sources are often incomplete. Many methods are better at retrospective verification than at prospective early warning. The solid electrolyte cell industry faces these limitations acutely because the technology is still emerging and the noise level is high.
| Characteristic | Meaning for Disruptive Technology | Relevance to Solid Electrolyte Cell |
|---|---|---|
| Outlierness | Deviates from mainstream technological clusters | Alternative electrolyte compositions and cell architectures |
| Novelty | Introduces unprecedented technical solutions or knowledge combinations | New composite electrolytes, in-situ formation, interfacial engineering |
| Forward-looking orientation | Points to future directions rather than incremental improvement | Lithium-metal compatibility, bipolar stacking, flexible solid electrolyte cell design |
| Potential influence | Can reshape industrial structure or open new markets | Vehicle electrification, grid storage, consumer electronics, aerospace |
2.2 Outlier Patents
Outlier patents are derived from the statistical concept of outliers. Compared with ordinary patents, they show significant differences or anomalous features in patent technology space. Traditional statistics often removes outliers as noise, but innovation studies have increasingly recognized that outliers can provide new information. They represent deviations from established rules, trends, or patterns, and they may indicate a potential shift in technology trajectories or the emergence of a new knowledge paradigm.
In disruptive technology identification, outlier patents have positive foresight value. They tend to be unique, rare, potentially disruptive, and influential. They are more likely to trigger non-continuous leaps in technology trajectories and can become early carriers of disruptive technologies. In my study of the solid electrolyte cell, I therefore interpret an outlier patent as a patent whose semantic content is distant from the dense center of the patent population, but which may still contain a coherent and potentially valuable technical direction.
2.3 Weak Signals
Disruptive technologies often begin as weak signals. They are low-visibility, ambiguous, and not yet widely diffused. Traditional monitoring systems may overlook them because they do not appear in high-frequency keywords or citation hot spots. Weak-signal theory provides a way to detect such early traces. I use visibility and diffusion as two dimensions. A weak-signal keyword should have low visibility and low diffusion, but a relatively high growth rate over time. When combined with outlier patents, weak signals help me distinguish truly promising outliers from random noise.
In the solid electrolyte cell context, weak signals may appear as technical terms that are not yet dominant. Examples from my empirical analysis include fiber, electrolyte membrane, in-situ, adsorption, boric acid, and semi-solid battery. These terms do not yet define the mainstream solid electrolyte cell industry, but they point to possible future pathways. By linking them to outlier patents, I can form a candidate patent group that is both semantically unusual and temporally early.
3. Methodology
3.1 Overall Framework
I construct a three-stage progressive early identification model. Stage one identifies outlier patents using semantic embeddings and a local outlier factor algorithm. Stage two recognizes disruptive early signals by combining keyword visibility and diffusion in a combination graph. Stage three evaluates disruptive potential through a multi-dimensional indicator system and objective weighting methods. The overall logic is to move from a large patent population to a small set of candidate patents, and then to a ranked list of high-potential disruptive technologies. The solid electrolyte cell industry is used as the empirical case.
| Stage | Objective | Main Method | Output |
|---|---|---|---|
| Stage 1 | Identify semantically distant patents | BERT embeddings and LOF | Outlier patent set |
| Stage 2 | Detect early weak signals | LDA, NMF, KMeans, combination graph | Weak-signal keywords and candidate patent group |
| Stage 3 | Rank disruptive potential | Entropy weight and CRITIC | High-potential disruptive technologies |
3.2 Stage One: Outlier Patent Identification with BERT-LOF
In the first stage, I collect patent data and extract the abstract and independent claims. I use a pre-trained BERT model to encode each patent into a high-dimensional semantic vector. This step captures the meaning and context of technical descriptions, overcoming the shallow semantic representation of keyword-based or classification-based methods. I then standardize the embedding vectors so that each feature dimension has a mean of zero and a variance of one. This prevents differences in scale from distorting distance calculations.
For a patent pair \(p\) and \(o\), I calculate the Euclidean distance in the standardized semantic space:
$$ d(p,o) = \|\tilde{\mathbf{h}}_p – \tilde{\mathbf{h}}_o\|_2 $$
For a positive integer \(k\), the k-distance of patent \(o\) is the distance to its k-th nearest neighbor. The reachability distance from \(p\) to \(o\) is defined as:
$$ \text{reach-dist}_k(p,o) = \max\{k\text{-distance}(o), d(p,o)\} $$
The local reachability density of patent \(p\) is the inverse of the average reachability distance from its k-nearest neighbors:
$$ \text{lrd}_k(p) = \frac{1}{\frac{1}{|N_k(p)|}\sum_{o \in N_k(p)} \text{reach-dist}_k(p,o)} $$
The local outlier factor of patent \(p\) is then the ratio of the average local reachability density of its neighbors to its own local reachability density:
$$ \text{LOF}_k(p) = \frac{\frac{1}{|N_k(p)|}\sum_{o \in N_k(p)} \text{lrd}_k(o)}{\text{lrd}_k(p)} $$
A LOF value significantly above one indicates that the patent is located in a region of lower density than its neighbors. I tune two core parameters: the neighborhood size \(n\_neighbors\) and the contamination ratio. I search \(n\_neighbors\) from 2 to 10 with step 1 and contamination from 5 percent to 20 percent with step 3 percent. I use the mean difference between outlier points and normal points as the evaluation criterion. The larger the difference, the more separable the outliers. In my solid electrolyte cell dataset, \(n\_neighbors=6\) and contamination=5 percent produce the best separation. I therefore use these parameters to obtain the outlier patent set.
| Parameter | Search Range | Selected Value | Reason |
|---|---|---|---|
| n_neighbors | 2 to 10, step 1 | 6 | Best separation between outlier and normal patents |
| contamination | 5 percent to 20 percent, step 3 percent | 5 percent | Highest mean difference and stable core set |
3.3 Stage Two: Weak-Signal Recognition via Combination Graph
In the second stage, I identify weak signals from the outlier patent set. I preprocess the abstract and main claim text using a domain dictionary and a stop-word list. I then extract keywords and topics using three parallel methods: LDA, NMF, and KMeans. The three methods cross-validate each other and reduce the bias of a single model. For each patent, I determine its core topic words by calculating its semantic distance to topic centers.
I calculate two time-series indicators for each keyword: visibility and diffusion. Visibility is measured by term frequency, while diffusion is measured by document frequency. Both indicators are adjusted by a time weight. The formulas are:
$$ \text{DoV}_{ij} = \frac{TF_{ij}}{NN_j} \times (1 – t_{\omega} \times (n – j)) $$
$$ \text{DoD}_{ij} = \frac{DF_{ij}}{NN_j} \times (1 – t_{\omega} \times (n – j)) $$
Here, \(TF_{ij}\) is the frequency of keyword \(i\) in period \(j\), \(DF_{ij}\) is the number of documents containing keyword \(i\) in period \(j\), \(NN_j\) is the number of documents in period \(j\), \(n\) is the number of periods, and \(t_{\omega}\) is a time weight. Following prior weak-signal research, I set \(t_{\omega}=0.05\). A weak-signal keyword should have a low average value and a high growth rate. I construct a combination graph with visibility and diffusion axes. Keywords in the low-visibility and low-diffusion region with high growth are selected as weak signals. I then link these keywords back to the outlier patents to form a candidate patent group for the solid electrolyte cell industry.
| Weak-Signal Keyword | Interpretation in Solid Electrolyte Cell | Related Patent Count |
|---|---|---|
| fiber | Fiber-reinforced composite electrolytes and separators | 7 |
| electrolyte membrane | Thin, free-standing solid electrolyte membranes | 9 |
| in-situ | In-situ polymerization or in-situ interfacial formation | 8 |
| adsorption | Adsorption-controlled interphase and dendrite suppression | 6 |
| boric acid | Boron-based additives and surface modification | 5 |
| semi-solid battery | Hybrid semi-solid electrolyte cell architectures | 9 |
3.4 Stage Three: Disruptive Potential Evaluation
In the third stage, I evaluate the disruptive potential of the candidate patents. I build a multi-dimensional indicator system from three dimensions: technological value, industrial value, and economic value. Technological value includes technical strength, innovation, breakthrough, and cross-disciplinarity. Industrial value includes strategic positioning and upgrading capacity. Economic value focuses on market value and long-term economic drivers. This structure allows me to balance technological novelty with industrial and economic relevance, which is essential for the solid electrolyte cell industry.
| Dimension | First-Level Indicator | Attribute | Second-Level Indicator |
|---|---|---|---|
| Technological value | Technical strength | Quantitative | Citation count |
| Technological value | Technical strength | Qualitative | Patent type |
| Technological value | Technical strength | Quantitative | Claim count |
| Technological value | Technical strength | Quantitative | Patent page count |
| Technological value | Technical strength | Quantitative | Inventor count |
| Technological value | Innovation | Quantitative | Patent innovation index |
| Technological value | Innovation | Quantitative | Patent absorption rate |
| Technological value | Breakthrough | Quantitative | Technical breakthrough |
| Technological value | Breakthrough | Quantitative | Outlierness |
| Technological value | Cross-disciplinarity | Quantitative | Trajectory transfer degree |
| Technological value | Cross-disciplinarity | Quantitative | Ratio of non-patent literature citations |
| Industrial value | Strategic | Quantitative | Priority count |
| Industrial value | Strategic | Quantitative | Technology scope |
| Industrial value | Strategic | Quantitative | Patent family size |
| Industrial value | Upgrading | Quantitative | Industrial coverage |
| Economic value | Market value | Quantitative | Market competitiveness |
I use the entropy weight method and the CRITIC method to determine weights. The entropy weight method measures the variability of each indicator. The CRITIC method considers both contrast intensity and conflict among indicators. I combine the two sets of weights to obtain a comprehensive weight. This reduces the risk of relying on a single objective weighting scheme.
For the entropy weight method, I first standardize the original data and calculate the proportion \(P_{ij}\). The information entropy of indicator \(j\) is:
$$ e_j = -\frac{1}{\ln m}\sum_{i=1}^{m} P_{ij}\ln P_{ij} $$
The redundancy degree is:
$$ d_j = 1 – e_j $$
The entropy weight is:
$$ W_j = \frac{d_j}{\sum_{j=1}^{n} d_j} $$
For the CRITIC method, I calculate the standard deviation \(\sigma_j\) and the Pearson correlation coefficient \(r_{jk}\) between indicators. The information amount of indicator \(j\) is:
$$ C_j = \sigma_j \sum_{k=1}^{m} (1 – r_{jk}) $$
The CRITIC weight is:
$$ w_j = \frac{C_j}{\sum_{j=1}^{m} C_j} $$
I then compute the comprehensive weight as a convex combination:
$$ W = \alpha W_{\text{entropy}} + (1 – \alpha) w_{\text{CRITIC}} $$
In my implementation, I set \(\alpha = 0.5\) to give equal importance to both objective weighting schemes. This is a transparent and reproducible choice for the solid electrolyte cell case.
3.5 Additional Technical Indicators
To capture innovation and breakthrough more precisely, I also compute several patent-level indicators. The patent innovation index \(N_i\) measures the dispersion of IPC classes among forward citations:
$$ N_i = 1 – \sum_{j} S_{ij}^{2} $$
where \(S_{ij}\) is the ratio of IPC class \(j\) to all forward-citation IPC classes of patent \(i\). The patent absorption rate \(A_{ij}\) measures the diversity of knowledge absorbed from backward citations:
$$ A_{ij} = \frac{1}{n}\sum_{j=1}^{n}\frac{C_j}{C_i} $$
where \(C_j\) is the number of IPC classes of backward-cited patent \(j\), and \(C_i\) is the number of IPC classes of the focal patent. The technical breakthrough indicator is:
$$ TB_i = \frac{\sum_{j \in i}\log\left(\frac{N}{N_g+1}\right)}{n_i} $$
where \(N\) is the total number of patents, \(N_g\) is the number of patents containing IPC class \(g\), and \(n_i\) is the number of IPC classes in technology topic \(i\). Finally, the trajectory transfer degree is:
$$ T = \frac{\sum_{j} CT_i}{n_j} $$
where \(CT_i\) is the number of IPC main groups in cited patent \(i\) that are not present in the original patent \(j\), and \(n_j\) is the number of IPC main groups in all backward citations of patent \(j\). These indicators enrich the technological value dimension and make the evaluation more sensitive to outlier characteristics.
4. Empirical Study on the Solid Electrolyte Cell Industry
4.1 Data Collection and Preprocessing
I collect patent data from the National Intellectual Property Administration. The search covers invention and utility model patents related to the solid electrolyte cell, published between 2015 and 2025. I extract title, abstract, claims, publication or application year, IPC classification, and patent holder fields. I remove invalid, withdrawn, and duplicate records. The final valid dataset contains 5,891 patents. This dataset provides a rich basis for identifying outlier patents and weak signals in the solid electrolyte cell industry.
| Field | Description | Use in My Model |
|---|---|---|
| Title | Patent title | Topic extraction and interpretation |
| Abstract | Technical summary | BERT semantic encoding |
| Claims | Legal and technical scope | BERT semantic encoding and technical strength |
| Year | Publication or application year | Weak-signal time series |
| IPC | International Patent Classification | Breakthrough, cross-disciplinarity, industrial coverage |
| Patent holder | Applicant or assignee | Industrial and economic interpretation |
4.2 Outlier Patent Identification Results
After BERT encoding and standardized vector construction, I run the LOF algorithm with the selected parameters. The algorithm identifies 291 outlier patents from the 5,891 valid solid electrolyte cell patents. These 291 patents are not necessarily the most cited or the most commercially visible. Instead, they are semantically distant from the dense center of the patent population. Many of them relate to composite electrolytes, interfacial layers, in-situ formation, membrane engineering, and hybrid semi-solid designs. This result supports my expectation that outlier patents in the solid electrolyte cell industry may carry early signals of disruptive change.
To test robustness, I perform a parameter sensitivity analysis. I fix contamination at 5 percent and vary \(n\_neighbors\) as 5, 6, 7, and 8. I find that 228 patents are identified as outliers under all four parameter settings, accounting for 78.4 percent of the core set. A total of 282 patents are identified under at least three parameter settings, accounting for 96.9 percent. This indicates that the core outlier set is highly stable across parameter choices. The identification method therefore has good robustness for the solid electrolyte cell industry.
| Parameter Setting | Number of Outlier Patents | Overlap with Core Set | Share |
|---|---|---|---|
| n_neighbors=5, contamination=5 percent | 291 | 228 | 78.4 percent |
| n_neighbors=6, contamination=5 percent | 291 | 228 | 78.4 percent |
| n_neighbors=7, contamination=5 percent | 291 | 228 | 78.4 percent |
| n_neighbors=8, contamination=5 percent | 291 | 228 | 78.4 percent |
| At least three settings | 282 | 282 | 96.9 percent |
4.3 Weak-Signal and Candidate Patent Group
From the outlier patent set, I extract topics and keywords using LDA, NMF, and KMeans. I then compute visibility and diffusion for each keyword over time. The combination graph identifies six weak-signal keywords: fiber, electrolyte membrane, in-situ, adsorption, boric acid, and semi-solid battery. These six keywords map to 44 candidate patents. This candidate group is both outlier-rich and early-stage. It includes patents on fiber-reinforced solid electrolyte cell components, free-standing electrolyte membranes, in-situ polymerization, adsorption-controlled interfaces, boron-based additives, and semi-solid battery designs.
The weak-signal keywords are important because they indicate where the solid electrolyte cell industry may be moving before the mainstream recognizes it. For example, in-situ formation can reduce interfacial resistance and improve contact between the solid electrolyte and electrodes. Adsorption may control lithium deposition and suppress dendrites. Boric acid and boron-based compounds may stabilize interfaces. Semi-solid battery designs may offer a practical bridge between liquid and all-solid-state systems. These are not yet dominant themes in the solid electrolyte cell industry, but they are technically plausible and patent-active.
| Weak-Signal Keyword | Visibility Level | Diffusion Level | Growth Trend | Candidate Patent Count |
|---|---|---|---|---|
| fiber | Low | Low | Increasing | 7 |
| electrolyte membrane | Low | Low | Increasing | 9 |
| in-situ | Low | Low | Increasing | 8 |
| adsorption | Low | Low | Increasing | 6 |
| boric acid | Low | Low | Increasing | 5 |
| semi-solid battery | Low | Low | Increasing | 9 |
4.4 Disruptive Potential Evaluation Results
I apply the multi-dimensional evaluation system to the 44 candidate patents. The weights reflect the relative importance of technological, industrial, and economic value. The largest weights appear in technical strength, strategic positioning, and market value. This is consistent with the nature of the solid electrolyte cell industry, where technical barriers and strategic patent layout are critical. I also observe that breakthrough and cross-disciplinarity receive meaningful weights because disruptive technologies often arise from knowledge recombination.
| Dimension | First-Level Indicator | Weight | Second-Level Indicator | Average Weight |
|---|---|---|---|---|
| Technological value | Technical strength | 30.16 percent | Citation count | 7.50 percent |
| Technological value | Technical strength | 30.16 percent | Patent type | 9.21 percent |
| Technological value | Technical strength | 30.16 percent | Claim count | 4.07 percent |
| Technological value | Technical strength | 30.16 percent | Patent page count | 3.70 percent |
| Technological value | Technical strength | 30.16 percent | Inventor count | 5.68 percent |
| Technological value | Innovation | 13.36 percent | Patent innovation index | 9.07 percent |
| Technological value | Innovation | 13.36 percent | Patent absorption rate | 4.30 percent |
| Technological value | Breakthrough | 8.06 percent | Technical breakthrough | 4.53 percent |
| Technological value | Breakthrough | 8.06 percent | Outlierness | 3.53 percent |
| Technological value | Cross-disciplinarity | 8.56 percent | Trajectory transfer degree | 4.99 percent |
| Technological value | Cross-disciplinarity | 8.56 percent | Non-patent literature ratio | 3.57 percent |
| Industrial value | Strategic | 22.82 percent | Priority count | 7.89 percent |
| Industrial value | Strategic | 22.82 percent | Technology scope | 5.62 percent |
| Industrial value | Strategic | 22.82 percent | Patent family size | 9.32 percent |
| Industrial value | Upgrading | 5.56 percent | Industrial coverage | 5.56 percent |
| Economic value | Market value | 11.47 percent | Market competitiveness | 11.47 percent |
Descriptive statistics show substantial variation across the candidate patents. Citation counts range from 0 to 49, claim counts from 2 to 36, patent pages from 5 to 62, and inventor counts from 1 to 9. Market competitiveness has an extremely wide range, reflecting the different scales and financing capacities of the applicants. This heterogeneity supports the need for a multi-indicator evaluation rather than a single-metric ranking.
| Indicator | Minimum | Maximum | Mean | Standard Error | Standard Deviation |
|---|---|---|---|---|---|
| Citation count | 0 | 49 | 3.83 | 1.40 | 8.97 |
| Patent type | 7 | 10 | 9.41 | 0.19 | 1.20 |
| Claim count | 2 | 36 | 10.59 | 1.00 | 6.41 |
| Patent page count | 5 | 62 | 14.83 | 1.59 | 10.17 |
| Inventor count | 1 | 9 | 4.59 | 0.33 | 2.14 |
| Patent innovation index | 0 | 1 | 0.60 | 0.08 | 0.49 |
| Patent absorption rate | 0.10 | 1.18 | 0.48 | 0.03 | 0.21 |
| Technical breakthrough | 0.03 | 0.25 | 0.10 | 0.01 | 0.05 |
| Trajectory transfer degree | 0 | 3.08 | 0.43 | 0.08 | 0.51 |
| Outlierness | 0.34 | 0.95 | 0.51 | 0.03 | 0.16 |
| Non-patent literature ratio | 0 | 0.86 | 0.22 | 0.02 | 0.15 |
| Priority count | 0 | 5 | 0.44 | 0.14 | 0.92 |
| Technology scope | 1 | 9 | 4.32 | 0.34 | 2.15 |
| Patent family size | 0 | 12 | 1.49 | 0.49 | 3.16 |
| Industrial coverage | 1 | 6 | 1.90 | 0.17 | 1.09 |
| Market competitiveness | 30 | 578603900 | 29146886.77 | 16043697.09 | 102729785.7 |
The top ten patents by disruptive potential score reveal several important directions for the solid electrolyte cell industry. The highest-ranked patents relate to solid ion conductors, solid electrolyte cell manufacturing, hybrid cation perovskite solid-state solar cells, solid-state batteries, solid electrolyte sheets, conductive polymer solid electrolyte secondary batteries, ceramic soft composites, composite binder compositions, solid thin-film hybrid electrochemical cells, and solid oxide fuel cell high-chromium waste pretreatment. Although some of these are adjacent to the core solid electrolyte cell field, they all contain outlier characteristics and early signals.
| Rank | Score | Patent Theme | Technology Direction |
|---|---|---|---|
| 1 | 63.83 percent | Solid ion conductor and preparation method | Solid electrolyte cell ion transport |
| 2 | 63.56 percent | Solid electrolyte cell and preparation method | Cell-level integration and manufacturing |
| 3 | 63.02 percent | Hybrid cation perovskite solid-state solar cell | Hybrid solid-state device architecture |
| 4 | 55.51 percent | Solid-state battery | Solid electrolyte cell system design |
| 5 | 51.83 percent | Sheet containing solid electrolyte | Electrode sheet and solid electrolyte integration |
| 6 | 49.95 percent | Conductive polymer solid electrolyte secondary battery | Polymer-based solid electrolyte cell |
| 7 | 41.56 percent | Ceramic soft composite for solid-state battery | Composite solid electrolyte cell |
| 8 | 41.55 percent | Composite binder composition for all-solid-state battery | Electrode slurry and binder engineering |
| 9 | 40.77 percent | Solid thin-film hybrid electrochemical cell | Thin-film solid electrolyte cell |
| 10 | 39.71 percent | Solid oxide fuel cell high-chromium waste pretreatment | Resource and recycling pathway |
To validate the model, I use the 2025 selected projects of the national key research and development program on disruptive technology innovation as a target set. The model achieves an accuracy of 0.52, a recall of 0.49, and a precision of 0.72. These values indicate that the model has promising effectiveness for identifying high-potential disruptive technologies. Precision is relatively higher than recall, meaning that the model is conservative but reliable when it flags a candidate. This is suitable for early-stage solid electrolyte cell intelligence, where false positives can be costly but false negatives are also important to reduce.
| Metric | Value | Interpretation |
|---|---|---|
| Accuracy | 0.52 | Moderate overall agreement with the target set |
| Recall | 0.49 | Captures about half of known disruptive projects |
| Precision | 0.72 | High reliability when a patent is flagged |
4.5 Interpretation of Top Technologies
The top-ranked technologies show that solid electrolyte cell innovation is not limited to a single electrolyte material. Instead, it covers core material innovation, composite membrane structure optimization, property control, application-specific design, and resource recycling. This broad coverage is consistent with the view that the solid electrolyte cell industry is moving from single-material breakthroughs toward system integration and process optimization. In my interpretation, this diversity is itself a signal of an unsettled technological paradigm.
Solid electrolyte material selection remains contested. Sulfide electrolytes offer high ionic conductivity, but they suffer from poor stability, difficult process control, and high cost. Oxide electrolytes offer better chemical stability and mechanical strength, but their rigid interfaces create large grain-boundary resistance. Polymer electrolytes offer good processability and flexible interfaces, but their room-temperature ionic conductivity is low. In this context, composite strategies are emerging. For example, ceramic soft composites combine high-conductivity sulfide electrolytes with organic ionic plastic crystals to improve solid-solid contact while maintaining ion transport. This is a representative outlier pathway in the solid electrolyte cell industry.
Lithium-metal anode preparation and protection also form a core competitive highland. The top patents include solid ion conductors, ceramic soft composites, and lithium-metal manufacturing methods. These inventions address dendrite growth, interfacial side reactions, and low-cost scalable manufacturing. For the solid electrolyte cell, the anode interface is often the bottleneck. Therefore, technologies that stabilize lithium metal are likely to have high disruptive potential. The white paper on the global solid electrolyte cell patent landscape also identifies solid lithium batteries as a core direction for next-generation power batteries, which supports my findings.
From a geographical perspective, Japan and South Korea appear to lead in the top-ranked solid electrolyte cell technologies. Among the top ten potential disruptive technologies, Japanese and Korean patents account for 40 percent, and many rank high. Chinese patents account for 20 percent in the top ten, but their share rises to 45 percent in the top twenty. United States patents account for 20 percent in the top twenty. The high-potential patents from Japan and South Korea also show broader claim protection, clearer industrialization orientation, and stronger industry-academia collaboration. This suggests that China still has gaps in core manufacturing and industrialization of solid electrolyte cell technologies.
| Rank Group | Japan and South Korea | China | United States | Other |
|---|---|---|---|---|
| Top 10 | 40 percent | 20 percent | 20 percent | 20 percent |
| Top 20 | 30 percent | 45 percent | 20 percent | 5 percent |
5. Discussion
My model offers three main contributions. First, it changes the research perspective. Traditional technology forecasting often focuses on the linear development of mainstream technologies and overlooks the disruptive potential of non-mainstream technologies. I start from patent outlierness and treat peripheral, non-consensus patents as early signals of disruptive technologies. This provides a new theoretical entry point for solid electrolyte cell intelligence.
Second, my model values weak-signal detection. The weak-signal keywords identified by the combination graph are captured before large-scale diffusion. This is important because once a technology becomes fully visible, the best window for strategic intervention may have passed. In the solid electrolyte cell industry, terms such as in-situ, adsorption, and semi-solid battery are still relatively weak, but they may grow into important technical routes.
Third, my multi-dimensional evaluation reduces the bias of single indicators. If I relied only on citation frequency, some high-potential patents could be underestimated because they are early-stage. By combining technological value, industrial strategy, and economic potential, I obtain a more balanced view. This is especially relevant for solid electrolyte cell patents, where early patents may have low citations but high strategic value.
My study also has limitations. The data source is primarily patent data, which may not capture market information, failed experiments, or tacit knowledge. The economic value indicator uses financing scale as a proxy, which may not reflect all dimensions of economic impact. The model is validated in one industry, the solid electrolyte cell industry, so its generalizability to other fields requires further testing. In future work, I plan to apply the model to multiple technology domains and compare it with traditional disruptive technology identification methods. I also intend to incorporate additional data sources, such as standards, investment records, and expert judgments, to improve early warning accuracy for the solid electrolyte cell industry.
6. Conclusion
In this research, I construct a three-stage progressive early identification model for disruptive technologies based on outlier patents. The model combines BERT semantic encoding, local outlier factor detection, weak-signal theory, combination graph analysis, and multi-dimensional evaluation with entropy weight and CRITIC methods. I apply the model to the solid electrolyte cell industry and identify 41 high-potential disruptive technologies. The top ten technologies cover solid ion conductors, solid electrolyte cell manufacturing, composite electrolytes, polymer solid electrolytes, thin-film cells, binder engineering, and resource recycling. These results validate the feasibility of using outlier patents to identify disruptive technologies and provide a quantitative method for early technological signal detection in the solid electrolyte cell industry.
My findings suggest that the solid electrolyte cell industry is in a phase of technological pluralism. No single material or design has achieved dominance, and composite strategies are likely to be important. Lithium-metal anode protection and scalable manufacturing are core competition areas. Japan and South Korea currently lead in high-potential patents, while China is catching up, especially in the broader top twenty group. For policymakers and firms, the implication is that outlier patents and weak signals should be monitored systematically. The solid electrolyte cell is not only a battery technology issue; it is a strategic industrial issue that requires early, evidence-based, and multi-dimensional intelligence. I believe that the model I present can support such intelligence and can be extended to other emerging technology fields in future research.