Skip to main content

Rare Earth Elements
Patent Technology Landscape

Global innovation in rare earth extraction, compounds, alloys, permanent magnets, phosphors & recycling

PATSTAT Global, Spring 2026 32,931 families, 2014–2024 15 August 2026 mtc.berlin
32,931
Patent Families
First filing 2014 – 2024
82%
Filed at CNIPA
26,990 families
+75%
Volume 2014 → 2022
1,940 → 3,395 families
14.1%
Recycling & Recovery
4,654 families
62.5%
Grant Rate
2014–2019 cohort

Executive Summary

Rare earth elements — the fifteen lanthanides plus scandium and yttrium — are the enabling materials behind permanent magnets, phosphors, catalysts, optical glasses and high-strength alloys. This report maps where that innovation is created and protected, using the rare-earth-specific places of the IPC and CPC classification schemes as the definition of scope.

The landscape comprises 32,931 DOCDB patent families whose earliest filing falls between 2014 and 2024. Filing volume rose steeply in the first years of the period, from 1,940 families in 2014 to 3,373 in 2016, and has since run on a plateau, every year between 2,860 and 3,395 families — the 2022 figure of 3,395 is the highest in the series and only marginally above 2016.

Geographically the field is concentrated: 26,990 families — 82% — include a filing at CNIPA. Comparing complete filing years only, so that the unfinished filing programmes of 2023 and 2024 cannot distort the picture, China’s share of period volume rises from 77.1% in 2014–2018 to 83.9% in 2019–2022. The ground it gains comes from the national routes — Japan loses 3.3 percentage points, the United States 2.8, Korea 0.8, Germany 1.0 — while the EPO share is unchanged at 13.4% and the PCT route edges up from 18.5% to 19.3%.

Beneath the flat headline volume, the composition of the field shifts. Recycling and secondary recovery grow from 250 families in 2014 to 690 in 2022, lifting their share of annual filings from 12.9% to 20.3%. Extraction and refining metallurgy (C22B) grows across the period, compounds of rare earth metals (C01F) grow modestly, while luminescent materials (C09K) decline from a 2018 peak. Academic filing grows while corporate filing declines: universities average 1,051 families a year in 2019–2024 against 855 in 2014–2018, and the university-affiliated public research bodies grow faster still in relative terms, while corporate filings fall from 1,661 to 1,436 a year.

The headline finding is compositional, not volumetric. Total filing activity has been flat since 2016, but within it China gains 6.8 percentage points of share at the expense of the national routes of Japan, the United States and Korea, recycling doubles its share of annual filings from its 2016 trough, and academic filing grows while corporate filing declines. The field is not expanding; it is being redistributed.

Overall Filing Activity

DOCDB patent families by earliest filing year. Each family is counted once, in the year of its earliest filing anywhere in the world.

* 2023 and 2024 are provisional: applications are published about 18 months after filing, so the two most recent filing years are still filling up in the PATSTAT Global Spring 2026 edition.

The series has two regimes. Between 2014 and 2016 the landscape grew by 73.9%, from 1,940 to 3,373 families. From 2016 onwards it has been flat: every year since sits between 2,860 and 3,395 families, and the 2022 peak of 3,395 exceeds the 2016 level by less than one per cent. Measured across the last complete stretch, 2014 to 2022, the field grew 75.0%, a compound annual rate of 7.3% — but essentially all of that growth was booked in the first two years.

No growth rate is quoted through 2023 or 2024, because both years are still incomplete.

Data table: Filing activity 2014–2024
Filing yearPatent familiesYear-on-year
20141,940
20152,240+15.5%
20163,373+50.6%
20173,073−8.9%
20183,376+9.9%
20192,860−15.3%
20203,189+11.5%
20213,035−4.8%
20223,395+11.9%
2023*3,277−3.5%
2024*3,173−3.2%
Total32,931

Geographic Distribution

Where the families in scope seek protection. A family filed in several offices is counted at each of them, so the office figures sum above the 32,931 families in the landscape.

The share comparison is drawn on complete filing years only — 2014–2018 against 2019–2022 — because a family first filed in 2023 or 2024 has not yet completed its international filing programme, and including those years would credit the home office with a shift that is really a publication lag.

On that basis the shift is real but narrower than the raw counts suggest. CNIPA appears in 77.1% of families first filed in 2014–2018 and 83.9% of those first filed in 2019–2022, a gain of 6.8 percentage points. What China gains is taken almost entirely from the national routes: Japan falls 3.3 points to 18.7%, the United States 2.8 points to 19.2%, Taiwan 0.9 to 3.7%, Korea 0.8 to 11.8% and Germany 1.0 to 2.0%.

European and international protection hold their ground: the EPO share is unchanged at 13.4% in both periods, and the PCT route edges up from 18.5% to 19.3%. The redistribution is between national filing routes, not away from multilateral ones.

Period shares are computed against the families first filed in each period (14,002 in 2014–2018 and 12,479 in 2019–2022) and therefore sum above 100%, because a single family reaches several offices. The periods cover five and four filing years respectively, so absolute counts are not comparable between them; shares are. The office totals in the first chart cover the full window 2014–2024.

Data table: Patent families by filing office
OfficePatent familiesShare of 32,931
CN — China (CNIPA)26,99082.0%
JP — Japan (JPO)5,75117.5%
US — United States (USPTO)5,74817.5%
WO — PCT5,61917.1%
EP — European Patent Office3,74711.4%
KR — Republic of Korea (KIPO)3,53510.7%
TW — Taiwan1,1943.6%
MX — Mexico9442.9%
DE — Germany (DPMA)7372.2%
CA — Canada6211.9%
RU — Russia5311.6%
AU — Australia4661.4%
BR — Brazil4401.3%
Data table: Office share by period (complete filing years)
Office 2014–2018 families Share of period 2019–2022 families Share of period Change
CN10,79977.1%10,47583.9%+6.8 pts
US3,08022.0%2,39319.2%−2.8 pts
JP3,06921.9%2,32818.7%−3.3 pts
WO2,58818.5%2,40819.3%+0.8 pts
EP1,87313.4%1,67513.4%±0.0 pts
KR1,77212.7%1,47711.8%−0.8 pts
TW6424.6%4603.7%−0.9 pts
DE4243.0%2502.0%−1.0 pts
All families in period14,002100%12,479100%

Technology Sub-Areas

The landscape broken down by the classification subclass of the rare-earth-specific code a family carries. Assignments are non-exclusive: a family classified both as an alloy and as a magnetic material appears under both, so the subclass counts sum above the 32,931 families in the landscape.

Data table: Patent families per classification subclass
SubclassFieldPatent families
C22CAlloys — rare-earth-containing steels, magnesium and nickel alloys9,585
C09KLuminescent materials containing rare earth metals9,524
H01FMagnets and magnetic materials — Nd-Fe-B and related6,149
C01FCompounds of rare earth metals4,294
C22BProduction and refining of rare earth metals3,869
C01GComplex oxides of other metals containing rare earths582
C30BSingle crystals containing rare earth metals547
B22FRare-earth intermetallic powders and powder metallurgy349
Sum of subclasses (non-exclusive)34,899
Distinct families32,931

How the sub-areas move over time

* 2023 and 2024 are provisional (≈18-month publication delay, PATSTAT Global Spring 2026 edition).

Two sub-areas move in opposite directions across the period. Extraction and refining metallurgy (C22B) grows from 264 families in 2014 to 473 in 2022, a rise of 79.2%, slightly ahead of the 75.0% recorded by the landscape as a whole over the same span. Luminescent materials (C09K) peak at 1,105 families in 2018 and fall to 822 by 2022, a decline of 25.6% from the peak, consistent with a maturing phosphor field.

Magnets (H01F) are the most stable large block, moving within a band of 468 to 663 families a year across the whole period with no clear direction.

Data table: Five largest sub-areas by filing year
Filing yearC22C AlloysC09K LuminescentH01F MagnetsC01F CompoundsC22B Production
2014380592468286264
2015550647548284285
20161,1981,011566362310
2017961978570341318
20181,0531,105529442308
2019853944476364318
2020923960594458336
2021912870526390367
20221,006822638483473
2023*883821663424450
2024*866774571460440

The largest individual classification places

CodeClassification placePatent families
C22C38/005Ferrous alloys containing rare earths (Sc, Y, lanthanides)6,691
H01F1/057Permanent magnets of rare earth and group IIIa elements, e.g. Nd2Fe14B4,364
C22B59/00Obtaining rare earth metals3,869
C09K11/77Luminescent materials containing rare earth metals2,407
H01F1/0577Sintered rare-earth permanent magnets1,937
C09K11/80Rare-earth luminescent materials containing aluminium or gallium1,603
C01F17/00Compounds of rare earth metals1,593
C01F17/10Preparation, separation or purification of rare earth compounds1,553
C22C23/06Magnesium alloys with a rare earth metal as next major constituent1,329
C09K11/78Rare-earth luminescent materials containing oxygen1,311
C09K11/85Rare-earth luminescent materials containing halogen1,251
H01F41/0293Grain-boundary diffusion of rare earth elements (Tb, Dy, Ho) into magnets1,241
C09K2211/182Organic electroluminescent complexes of the rare earth metals1,208
C22C28/00Alloys based on a metal not covered elsewhere (rare-earth-based alloys)1,159
H01F1/055Permanent magnets of rare earth and transition metals936

Non-exclusive: a family carries several classification codes and appears in every row that applies to it.

Top Applicants

The twenty most active applicant groups in the landscape. Counts are consolidated across the harmonised name variants that PATSTAT records for the same organisation; the consolidation rule is stated in the Methodology section.

# Applicant group Type Patent families
1Nippon SteelCompany646
2JFE SteelCompany534
3Jiangxi University of Science and TechnologyUniversity346
4NichiaCompany329
5Baotou Iron and SteelCompany320
6TDKCompany317
7Hitachi Metals / ProterialCompany283
8Northeastern UniversityUniversity241
9Panasonic / MatsushitaCompany235
10Central South UniversityUniversity232
11Baotou Research Institute of Rare EarthsGov / non-profit227
12Grirem Advanced MaterialsCompany206
13ToyotaCompany204
14Ningbo Institute of Materials Technology and Engineering (CAS)Research201
15Sumitomo Metal MiningCompany179
16ToshibaCompany138
17OsramCompany137
18Shin-Etsu ChemicalCompany127
19Beijing Zhong Ke San Huan Hi-TechCompany106
20Zhenghai Magnetic MaterialCompany80

The applicant ranking does not mirror the geography. China accounts for 82% of the filings, yet the two largest applicant groups are Japanese steelmakers: Nippon Steel with 646 families and JFE Steel with 534. That is consistent with the shape of the classification breakdown, where the single largest place in the landscape is C22C38/005 — ferrous alloys containing rare earths — with 6,691 families.

The Chinese side of the ranking is led by academic and state research institutions: Jiangxi University of Science and Technology (346), Northeastern University (241), Central South University (232) and the Baotou Research Institute of Rare Earths (227), with Baotou Iron and Steel at 320 the largest corporate name among them. The twenty groups together account for 5,088 family mentions, at most 15.4% of the landscape — the count is non-exclusive, since a co-filed family counts for each partner, so the true share is lower still. A field in which the twenty largest filers hold under a sixth of the volume is a dispersed one.

Data quality note — name variants. PATSTAT’s harmonised applicant name (han_name) does not group an organisation across spellings, scripts and corporate history. In this landscape the same steelmaker appears as NIPPON STEEL CORP and NIPPON STEEL & SUMITOMO METAL CORP, and JFE Steel appears under three spellings including a full-width Japanese form. The counts above merge such variants under a documented rule; the underlying han_name counts are lower. Because no consolidation rule catches every variant, all applicant counts are a lower bound.

PATSTAT records no applicant country for the majority of families in this landscape, so organisations that share a name across countries — Northeastern University among them — cannot be separated on the data alone.

International Applicant Landscape

Applicants for which PATSTAT records a country of residence outside China. Because a country code is recorded for only a minority of applicants in this field, this ranking is a lower bound rather than a complete picture of non-Chinese activity.

Of the eleven applicants above the reporting threshold, ten are Japanese and one is German. No applicant recorded in the United States, the Republic of Korea, France or the United Kingdom reaches the threshold of 85 families, even though the USPTO receives 5,748 families and KIPO 3,535 — those filings are dispersed across many applicants rather than concentrated in a few portfolios.

The Japanese presence spans the whole value chain: steel and alloys (Nippon Steel, JFE Steel), magnets and electronic components (TDK, Toshiba), phosphors and lighting (Nichia, Denka), silicon and specialty chemicals (Shin-Etsu) and end-use manufacturing (Toyota, Panasonic). Osram Opto Semiconductors is the single European name in this group, in the phosphor and LED domain.

Data quality note. Two entries in the chart belong to the same corporate group: NIPPON STEEL CORP and NIPPON STEEL & SUMITOMO METAL CORP are successive names of one filer, together 646 families after consolidation. The chart shows the harmonised names as PATSTAT records them, so that each bar corresponds to a value that can be reproduced from the query; the consolidated figures are in the Top Applicants section.

Data table: Applicants with a recorded non-Chinese country
Applicant (harmonised name)CountryPatent families
NIPPON STEEL CORPJP547
JFE STEEL CORPJP534
NICHIA CORPJP218
MATSUSHITA ELECT IND CO LTDJP206
TDK CORPJP191
NIPPON STEEL & SUMITOMO METAL CORPJP162
TOYOTA JIDOSHA CO LTDJP106
SHIN ETSU CHEM CO LTDJP101
CO LTD TOSHIBAJP92
DENKA CO LTDJP90
OSRAM OPTO SEMICONDUCTORS GMBHDE89

Recycling & Recovery of Rare Earth Elements

Families in the landscape that additionally carry a recycling or secondary-recovery classification: working-up of raw materials other than ores (C22B7), reclaiming of battery materials (H01M10/54), and the CPC climate tags for waste-management recycling (Y02W30) and metal-processing recycling (Y02P10/20).

* 2023 and 2024 are provisional (≈18-month publication delay, PATSTAT Global Spring 2026 edition).

Recycling is the clearest growth story in an otherwise flat landscape. Recycling and recovery filings rise from 250 families in 2014 to 690 in 2022, an increase of 176.0% and a compound annual rate of 13.5% — against 75.0% and 7.3% for the landscape as a whole over the same span. Their share of annual filings roughly doubles, from 12.9% in 2014 and a trough of 10.2% in 2016 to 20.3% in 2022, holding near 19% in the two provisional years.

Across the whole period, 4,654 of 32,931 families (14.1%) carry a recycling or recovery classification.

Scheme-coverage caveat. Two of the four recycling anchors used here — Y02W30 and Y02P10/20 — are CPC tagging schemes with no IPC equivalent. Of the 4,654 recycling families, 3,314 are reachable only through CPC and 1,340 through IPC, and 3,268 carry a Y tag without either of the scheme-neutral codes. Offices that assign IPC but little CPC are therefore systematically under-represented in this cut, which in a landscape that is 82% Chinese means the recycling figures are a floor rather than a measurement of the whole.

Data table: Recycling and recovery by filing year
Filing yearAll familiesRecycling / recoveryShare
20141,94025012.9%
20152,24029012.9%
20163,37334510.2%
20173,07332310.5%
20183,37635310.5%
20192,86033411.7%
20203,18939412.4%
20213,03545415.0%
20223,39569020.3%
2023*3,27762319.0%
2024*3,17359818.8%
Total32,9314,65414.1%

Collaboration Networks

Families filed by more than one applicant, and the relationships behind them.

3,875 of the 32,931 families — 11.8% — carry more than one applicant. Put the other way round, roughly nine families in ten are filed by a single organisation, in a field that spans mining, metallurgy, materials chemistry and device manufacturing.

The second thing it says is that the largest co-filing relationships are not partnerships between independent organisations but joint filings inside a corporate group or a research group. The four largest relationships that resolve to complete harmonised names on both sides are all of this kind.

Co-applicant pairCo-filed families
NIPPON STEEL & SUMITOMO METAL CORP + NIPPON STEEL CORP61
CO LTD TOSHIBA + TOSHIBA MATERIALS CO LTD25
HYUNDAI MOTOR CO + KIA CORP25
GRIREM ADVANCED MATERIALS CO LTD + GRIREM HI TECH CO LTD22

The single largest co-filing relationship in the landscape, at 78 families, is centred on Fujian Changting Golden Dragon Rare-Earth Co., Ltd.

Read together with the applicant ranking, the picture is of a field organised around vertically integrated groups rather than around consortia. Every one of the four largest fully resolvable pairs joins two entities of the same group — a parent and its materials subsidiary, or two successive corporate names of the same filer. Cross-organisational collaboration exists, but not at a scale that changes the shape of the landscape.

How the pairs were counted. Co-applicants are counted from applications with two or more distinct harmonised applicant names, each unordered pair counted once. The query therefore cannot place a single applicant in its own partner list. It does not merge affiliated entities, which is exactly why intra-group pairs remain visible above — that is a property of the data worth seeing, not an error to be filtered away.

Applicant Type Distribution

Applicant sectors as recorded in PATSTAT, expressed as families per filing year so that the five-year and six-year periods can be compared. This section uses the full window, unlike the office comparison in Geographic Distribution, which is restricted to complete filing years. Assignments are non-exclusive: a family filed jointly by a company and a university counts in both sectors.

Across the whole period companies hold 16,921 of the 32,931 families (51.4%) and universities 10,582 (32.1%): roughly one family in three carries a university applicant. The direction of travel is as notable as the level: on a per-year basis, university filings rise 22.9% between the two periods while corporate filings fall 13.6%. Public research organisations are flat at about 155 families a year, and the university-affiliated research bodies grow 35.2%, from 93 to 125 a year.

Individual inventors fall from 164 to 62 families a year, a decline of 61.9%.

Data quality note — unassigned sectors. Families with no sector recorded rise from 80 to 743 per year across the two periods. This is a property of PATSTAT’s sector assignment, which lags for recently added person records; it is not a new class of applicant. Because those 4,858 families are missing from every sector line, the per-sector figures for 2019–2024 are understated to an unknown degree, and the comparison between periods should be read as directional rather than exact.

Data table: Applicant sectors by period
Sector Families 2014–2018 Per year Families 2019–2024 Per year All families
Company8,3071,6618,6141,43616,921
University4,2768556,3061,05110,582
Government / non-profit7651539291551,694
Government-non-profit university463937511251,214
Individual818164374621,192
No sector assigned398804,4607434,858
Distinct families in the landscape32,931

Minor sectors below 100 families in total (hospital, mixed company-and-public assignments) are omitted from the table; they add 141 families. Sector sums exceed the distinct family total because a family can carry applicants from several sectors.

Citation Analysis

Forward citations at family level: how many distinct DOCDB families cite the families in this landscape.

20,402 of the 32,931 families (61.9%) have been cited at least once, and the landscape has attracted 89,510 citing-family links in total, an average of 2.72 per family. On the completed part of the period — families first filed in 2014–2019 — the cited share rises to 79.2% (13,356 of 16,862). 2,111 families (6.4%) have been cited by ten or more families.

The declining bars are a measurement artefact, not a finding: a family first filed in 2024 has had almost no time to be cited. Only the left-hand side of this chart carries information about impact.

Forward-citation counts for recent filing years are structurally near zero and must not be read as impact. The most heavily cited family in the landscape has been cited by 317 distinct DOCDB families; the second, an atomic-layer-deposition coating of fluorinated rare earth oxide for semiconductor chamber components, by 238.

Data table: Forward citations by filing year
Filing yearFamiliesCiting familiesPer family
20141,94013,0916.75
20152,24012,9325.77
20163,37314,6664.35
20173,07312,5434.08
20183,37610,8523.21
20192,8608,2982.90
20203,1897,0672.22
20213,0354,9561.63
20223,3953,1500.93
20233,2771,3130.40
20243,1736420.20
Total32,93189,5102.72

Methodology

Search strategy, code-set construction, data basis and known limitations.

Data basis. All figures in this report were computed on 15 August 2026 against EPO PATSTAT Global, Spring 2026 (patstat-mtc.patstat_2026a), with classification metadata from IPC edition 20260101 (in force 1 January 2026) and CPC edition 202608 (in force 1 August 2026). The classification code set was completed in three ways before any family was counted: every code was expanded to its full subtree, retired symbols were added by a prefix pass bounded to inactive codes, and predecessor symbols were unioned in from the reclassification chain. No code list was filtered on is_active, so deprecated symbols that still sit on documents remain in scope.

Search strategy, code-set construction, counting rules and data source

How the code set was found

The landscape is defined by the classification places that are themselves about rare earths, discovered through three independent entry points rather than by hand-picking codes:

  • Catchword index — WIPO’s natural-language index into the IPC (dpt_catchword), searched for rare earth and lanthanides.
  • Title breadcrumbSEARCH(title_full, …) on the IPC and CPC views, which carries the ancestor chain down to every subgroup, so that a subgroup whose own title reads only “containing rare earths” is still findable.
  • Definition textsSEARCH(definition_scope, …) on CPC, the scope prose that WIPO writes using terms deliberately alternative to those in the title. This is a CPC capability; IPC carries definition text for only a small minority of symbols.

Because SEARCH() does not stem, singular and plural were searched separately and unioned: rare earth and rare earths, lanthanide and lanthanides. The plural is not cosmetic here — on CPC titles the singular form returns 327 symbols and the plural 589. The three entry points together yield 962 joinable CPC symbols and 143 joinable IPC symbols.

How the code set was completed over time

  • Subtree expansion — a recursive walk of the parent relation, not a symbol prefix. Subject matter belonging to a subgroup is never classified into its parent, so a single node holds only the residue of its field. CPC grows from 962 to 1,336 symbols; IPC does not grow, because a breadcrumb search is already subtree-closed.
  • Retired symbols — deprecated codes carry no searchable breadcrumb and cannot be reached by a parent walk, because deprecated symbols have no parent. They were added by a prefix pass over the canonical symbol, bounded to NOT is_active and to descendants of a seed. This adds 18 CPC symbols.
  • Predecessor union — the reclassification chain (dpt_cpc_reclassification, dpt_ipc_reclassification) was joined so that documents still sitting on a superseded symbol are found. Without it, a young symbol produces a time series that measures reclassification progress and reads like invention.

Where the predecessor union was restricted, and what that costs

A predecessor union buys recall at the cost of precision, and in this field the cost is measurable. Three of the predecessor symbols are broad parent groups that merely happened to spawn a rare-earth subgroup: C01G53/00 (compounds of nickel), C01G51/00 (compounds of cobalt) and C01G45/12 (complex oxides containing manganese). Their contents are overwhelmingly lithium-ion cathode chemistry, not rare-earth technology.

A predecessor was therefore kept only where the predecessor place itself names rare earths or lanthanides. The unrestricted union yields 45,796 families, the restricted one 32,931. The 12,865-family difference is a deliberate precision decision, and the honest reading is that the true figure lies between the two.

Core places versus application-domain tags

The discovery step also returns a large body of codes in which a rare earth appears as a constituent of some other technology — polymerisation catalysts (C08F), ceramics (C04B), heterogeneous catalysis (B01J), optical glass (C03C), lasers (H01S), electrochemical cells (H01M) and others. Those tags cover 100,175 additional families that carry no core rare-earth code. A random sample of 24 of them read as a mixture of polypropylene compositions, refractory bricks, cordierite beads and polycrystalline diamond — far past the one-in-five off-topic threshold at which a landscape profile is too broad.

The scope was therefore restricted to the eight subclasses in which the rare earth is the classified subject matter: C22B, C01F, C01G, C22C, H01F, C09K, C30B and B22F.

IPC or CPC — measured, not assumed

Both schemes were searched and both were used. Running the same scope through each in turn: 16,162 families are found by both schemes, 11,307 only through CPC and 5,462 only through IPC. A CPC-only landscape would lose 5,462 families here and an IPC-only landscape 11,307, so neither scheme alone is sufficient and the union is what the report counts.

Family counting

  • The counting unit is the DOCDB patent family, counted with COUNT(DISTINCT docdb_family_id). An invention filed in CN, JP, US and at the EPO is one family, not four.
  • Each family is assigned to the earliest filing year of the family, so that the per-year counts sum exactly to the landscape total (they do: 32,931).
  • Only patents of invention are counted (ipr_type = 'PI'); utility models and design rights are excluded. Pseudo-families (docdb_family_id = 0) are excluded.
  • Applicants are taken from applt_seq_nr > 0; inventors are not counted as applicants.

Multi-assignment note

Patent families can carry several classification codes, reach several offices and involve applicants from several sectors. Sub-totals in the technology, geography and sector breakdowns therefore exceed the total family count, and that is correct behaviour rather than double counting. Every such table states its own sum next to the distinct total.

Data source and stack

EPO PATSTAT Global, Spring 2026 edition (patstat-mtc.patstat_2026a) on Google BigQuery, with classification metadata from the patstat-mtc.classification views. Queries were written and executed through mtc.berlin’s patstat-mcp server; analysis and visualisation with Claude AI. All SQL ships with the report in queries.sql.

Acknowledgement

The initial framing of this subject — the idea of combining classification codes for rare-earth extraction and recovery with a proximity text search — was kindly provided by Riccardo Priore, PhD, Centro Patlib — Ufficio Valorizzazione della Ricerca, Struttura Ricerca e Innovazione, AREA SCIENCE PARK, Padriciano 99, 34149 Trieste, Italy.

Scope Limitations

  • Title-only families are not in scope. 9,796 families whose English-language title contains “rare earth” carry no rare-earth classification code and are therefore absent from every figure in this report — an additional 29.7% on top of the 32,931 counted. A sample of 20 of them read as genuinely on topic (rare-earth-doped fibres, magnet motors, molten-salt electrolysis cells, flotation agents), so this is a real recall gap rather than noise. It is left open deliberately: an English-title layer is not available in every language and would make the landscape inconsistent across offices.
  • The recycling cut is CPC-weighted. Two of its four anchors are CPC-only Y tags; 3,314 of the 4,654 recycling families are reachable only through CPC. Offices that assign IPC but little CPC are under-represented.
  • Applicant country is largely unrecorded. PATSTAT holds no country code for the applicants of most families in this landscape, so applicant-origin geography is not reported. The geography in this report is filing-office geography — where protection is sought, not where the invention was made.
  • Applicant names are a lower bound. Harmonised names do not group an organisation across spellings, scripts and corporate history. The consolidation applied here catches the largest variant groups, not all of them.
  • The two most recent filing years are provisional. Applications are published about 18 months after filing, so 2023 and 2024 are still filling up. No growth rate in this report terminates in either year.
  • Grant rates carry a pendency caveat. The headline grant rate of 62.5% is computed on the 2014–2019 cohort, which is old enough to be mostly decided. Across all filing years the figure is 55.2%, and that lower number measures examination speed as much as success.
  • Design rights and utility models are excluded, which in a field with heavy Chinese activity removes a real body of protected technology from view: 2,942 families in the period are utility-model-only.
  • External magnitude check. No independent published family count for a comparable rare-earth scope could be obtained for this report; the qualitative findings of recent peer-reviewed patent studies — a China-centred filing landscape and rising recycling activity — are consistent with what is measured here, but no numeric cross-check is claimed.
Glossary — Patent Terms Explained
Rare earth elements (REE)
The fifteen lanthanides plus scandium and yttrium. Critical raw materials for permanent magnets, phosphors, catalysts, optical glasses and high-strength alloys.
DOCDB patent family
A group of patent applications protecting the same invention in different countries. Counted once, so that an internationally filed invention does not outweigh a nationally filed one.
Earliest filing year
The year of the earliest filing anywhere in a patent family. Used here so that each family is counted in exactly one year.
IPC / CPC
International and Cooperative Patent Classification. IPC is assigned by virtually every office; CPC is maintained by the EPO and USPTO and applied consistently across family members. Both are used in this report.
Y02 tags
A CPC-only tagging scheme for climate-change mitigation technologies, applied across the whole classification rather than within one technical field. It has no IPC equivalent.
NdFeB
Neodymium-iron-boron, the strongest class of permanent magnet, classified in H01F1/057 and its subgroups.
Grain-boundary diffusion
A process that drives heavy rare earths such as terbium or dysprosium into the grain boundaries of a sintered magnet to raise its coercivity while using less of the scarce element. Classified in H01F41/0293.
han_name
PATSTAT’s harmonised applicant name. It merges spelling variants of a name but does not merge subsidiaries, translations or successive corporate names.
PCT (WO)
The Patent Cooperation Treaty route. A WO filing buys the option of protection in many countries; the national decisions come later.
CAGR
Compound annual growth rate, computed over the number of intervals between the first and last year quoted.
Forward citation
A later patent family citing an earlier one. Counts for recent filing years are structurally near zero and cannot be read as impact.

Report Files

This report is published as a set of source files: meta.json (metadata), cover.html and kpi-strip.html (hero and key figures), sections/ (the twelve content sections) and queries.sql — the complete, commented SQL behind every number shown here, including the classification code-set construction, so that the analysis can be reproduced against the stated PATSTAT and classification editions.

Like what you see?

This report was built with a fully reproducible pipeline: EPO PATSTAT Global on BigQuery, a custom MCP server, and Claude AI for analysis and visualization.

Custom Patent ReportNeed a similar analysis for your technology area, company, or research question? We build tailored patent intelligence reports.
The ToolchainWant to run your own AI-powered patent analyses? We help you set up PATSTAT on BigQuery with MCP server integration.