Reference table: byte-level harmonic fraction for 49 languages

Data from the irreversibility depth measurement. Method: Helmholtz-Hodge decomposition on byte-level de Bruijn graphs. Source: Wikipedia (~300k bytes per language).

The harmonic fraction f(D) measures the proportion of byte-level transition energy carried by irreversible cycle currents at context depth D. D* is the extrapolated depth where f vanishes. Sorted by D* (characters).

LanguageCode OrderMorphScript bpcD*cD*b f(1)f(2)f(3)f(4)f(5) Cycles
Chinesezh SVOisolatingcjk 2.41.23.0 0.9690.7770.5020.4570.405 1886
Japaneseja SOVagglutinatingcjk 2.51.53.7 0.9740.7610.5800.4720.427 844
Hindihi SOVfusionaldevanagari 2.51.94.8 0.9660.7420.6380.5450.488 152
Koreanko SOVagglutinatinghangul 2.12.04.2 0.9740.7640.5690.5100.464 721
Ukrainianuk SVOfusionalcyrillic 1.72.23.8 0.9700.7220.5610.4820.435 360
Russianru SVOfusionalcyrillic 1.72.33.9 0.9740.7250.5700.4900.441 435
Mongolianmn SOVagglutinatingcyrillic 1.72.34.0 0.9740.7310.5850.5000.450 210
Greekel SVOfusionalother 1.72.34.0 0.9720.7200.5770.4960.455 382
Arabicar VSOfusionalarabic 1.72.34.1 0.9670.7440.6300.5030.442 140
Hebrewhe SVOfusionalother 1.72.44.0 0.9650.7420.6020.5020.433 204
Bulgarianbg SVOfusionalcyrillic 1.72.74.5 0.9710.7390.5930.5210.476 337
Serbiansr freefusionalcyrillic 1.53.04.4 0.9720.7300.5820.5160.478 530
Czechcs freefusionallatin 1.13.33.6 0.9570.7120.5400.4760.448 1194
Slovaksk freefusionallatin 1.13.53.8 0.9580.7070.5480.4840.457 1006
Hungarianhu SOVagglutinatinglatin 1.13.53.8 0.9530.7130.5490.4910.466 1122
Turkishtr SOVagglutinatinglatin 1.13.84.0 0.9560.7510.5660.5010.474 854
Polishpl freefusionallatin 1.13.94.0 0.9600.7390.5650.5020.469 867
Croatianhr freefusionallatin 1.03.94.0 0.9540.7490.5680.4990.463 873
Vietnamesevi SVOisolatinglatin 1.24.04.9 0.9710.7210.5690.5220.496 290
Germande SOVfusionallatin 1.04.24.3 0.9510.7420.5800.5100.472 1111
Latvianlv freefusionallatin 1.14.24.5 0.9610.7500.5860.5190.485 807
Spanishes SVOfusionallatin 1.04.24.3 0.9600.7250.5650.5090.477 674
Lithuanianlt freefusionallatin 1.14.34.6 0.9580.7290.5760.5210.485 713
Portuguesept SVOfusionallatin 1.04.34.4 0.9600.7360.5730.5150.481 765
Italianit SVOfusionallatin 1.04.54.5 0.9580.7440.5710.5150.487 858
Danishda SVOfusionallatin 1.04.54.6 0.9540.7520.5860.5220.486 881
Estonianet SVOagglutinatinglatin 1.04.64.7 0.9590.7260.5680.5190.492 918
Frenchfr SVOfusionallatin 1.04.64.8 0.9600.7580.5970.5300.491 802
Romanianro SVOfusionallatin 1.04.64.8 0.9620.7290.5750.5230.495 726
Englishen SVOfusionallatin 1.04.74.8 0.9570.7610.5880.5230.492 702
Dutchnl SVOfusionallatin 1.0>5>5 0.9560.7630.6010.5430.511 667
Swedishsv SVOfusionallatin 1.0>5>5 0.9630.7500.6140.5680.546 451
Norwegianno SVOfusionallatin 1.0>5>5 0.9600.7400.5930.5430.517 620
Indonesianid SVOagglutinatinglatin 1.0>5>5 0.9560.7720.6130.5520.528 497
Malayms SVOagglutinatinglatin 1.0>5>5 0.9640.7620.6100.5560.533 575
Thaith SVOisolatingother 2.5>5>5 0.9690.7660.6440.5790.548 314
Swahilisw SVOagglutinatinglatin 1.0>5>5 0.9610.7790.6210.5610.533 431
Bengalibn SOVfusionalother 2.5>5>5 0.9700.7490.6290.5550.508 213
Persianfa SOVfusionalarabic 1.8>5>5 0.9650.7620.6590.5700.504 136
Finnishfi SVOagglutinatinglatin 1.0>5>5 0.9610.7330.5900.5460.515 908
Georgianka SOVagglutinatingother 2.3>5>5 0.9690.7780.6720.6140.577 186
Tamilta SOVagglutinatingother 2.6>5>5 0.9650.7800.6970.6380.598 110
Telugute SOVagglutinatingother 2.6>5>5 0.9620.7800.6720.5790.518 92
Burmesemy SOVisolatingother 2.6>5>5 0.9700.7790.6690.6100.574 159
Uzbekuz SOVagglutinatinglatin 1.0>5>5 0.9700.7540.5990.5460.521 750
Irishga VSOfusionallatin 1.1>5>5 0.9610.7350.5870.5330.509 394
Welshcy VSOfusionallatin 1.0>5>5 0.9580.7420.5930.5370.511 782
Tagalogtl VSOagglutinatinglatin 1.0>5>5 0.9600.7670.6000.5400.509 855
Latinla freefusionallatin 1.0>5>5 0.9610.7490.6010.5590.537 664

Column definitions

bpcMean bytes per Unicode character
D*cIrreversibility depth in characters (D*bytes / bpc)
D*bIrreversibility depth in bytes
f(D)Harmonic fraction at context depth D = ||Aharm||² / ||A||²
CyclesNumber of depth-3 harmonic cycles (trigram rotations)

Cite this data

@misc{hoekstra2026atlas,
  author = {Hoekstra, Richard},
  title = {Byte-Level Harmonic Fraction for 49 Languages},
  year = {2026},
  url = {https://richardhoekstra.nl/atlas/reference-table.html}
}