# Reviewer's notes: A Periodic Chart for Magnetic Resonance

Companion to `mrs/periodic_table.html`. Preprint, version 0.3, 13 September 2026.

This file records every value in the chart that the people who compiled and audited it were
not fully confident of, in their own words, along with what they checked and what they could
not. It exists so that anyone fact-checking the chart can start where the weaknesses actually
are rather than sampling at random.

## How to read it

Each entry is a specific reservation about a specific value. They fall into a few kinds:

- a number taken from a secondary or remembered source that could not be re-verified;
- a number where two respectable sources disagree, and the choice between them is stated;
- a field deliberately left null because no defensible value was found;
- a figure derived by arithmetic rather than quoted, which is sound but inherits the
  assumptions behind it;
- a drawing hint in a spectrum, which is not a measurement at all.

Anything **not** listed here was either checked against a primary table or is computed by the
page from primitives that were. The whole isotope set is additionally checked by program
against Stone's IAEA compilation as distributed with EasySpin, by
`mrs/periodic_table_src/crosscheck.py`; that check is described on the page itself and is not
repeated here.

An entry being listed does not mean the value is wrong. It means nobody has yet shown it to be
right. Outright errors are not recorded here: they are fixed, and the fix appears in the git
history.

## Coverage

Every element slice, every dossier and every shift table in the chart now has notes here.
Version 0.1 covered Z = 43 upward, two of the ten dossiers and four of the seven shift tables,
and said plainly that Z = 1 to 42 deserved the same treatment and had not had it. Version 0.2
closed that gap: Z = 1 to 42, the remaining eight dossiers and the remaining three shift
tables were audited, and the questions the chart itself had flagged as open were worked
through. Coverage is not depth. The newer sections come from one audit pass each, run against
an egress proxy that refused most primary tables, and they say where that bit.

## Reviews applied so far

**Version 0.1.** Four reviewers went over the page's physics, the dossiers, the shift tables
and the JavaScript. Thirty-one findings held up and were fixed. Two were accepted on the
reviewer's own reasoning rather than a fetched source and remain open: the 17O
indirect-detection linewidth at natural abundance, corrected to agree with the calibration
quoted beside it but with the original figure's provenance never identified; and the fields at
which Nb3Sn enters NMR magnet design and gives way to HTS inserts. The second of those is now
settled: the page says 14 T.

**Version 0.2.** Eight auditors re-derived the numbers in their assigned slice, dossier or
shift table from the chart's own primitives and from whatever primary sources the proxy would
pass, and were asked to report errors separately from reservations. Thirty-three corrections
came back. Thirty-two were applied and are described in the git history, three of those narrowed
to what could be checked rather than adopted as proposed; the thirty-third, a 0.13 percent
revision to the 73Ge natural abundance, was declined for want of a reachable source and is
recorded below as a reservation instead. What follows is what the audit could not settle.

**Version 0.3.** The CRC susceptibility table that note 117 named as unreachable was supplied directly,
and the whole molar susceptibility column was rebuilt on it: forty of the 118 cells moved, two of them by
a factor of six. The column had never been the CRC table it claimed to be. Eight auditors adjudicated the
differences element by element, and every proposal was then put to a separate agent whose only job was to
refute it; nine were refuted, and three of those changed the value finally written. It is now a standing
check rather than a one-off, so the column cannot drift back: `crosscheck.py` compares every cell against
that table and prints the two deliberate departures, helium and fluorine, with their reasons.

---


## Elements Z = 1 to 9 (H through F)

1. chiMolar for all nine elements in this slice is an element-table value in the CGS convention (10^-6 cm3/mol, per mole of the standard-state substance, so H2, N2, O2 and F2 are per mole of the diatomic). The convention is applied consistently here: H -3.98, He -1.88, Li +14.2, Be -9.0, B -6.7, C -6.0, N -12.0, O +3449, F -9.63. Only fluorine was confirmed against a primary measurement this session: the isotropic susceptibility of F2 is reported as (-9.627 +/- 0.062) x 10^-6 cgs (J. Phys. Chem. A 1999), which is where the CRC -9.63 comes from, and even that paper drew a published Comment (J. Phys. Chem. A 1999, 'Comment on: On the Magnetic Susceptibility of Fluorine'), so the value is contested at the level of its own error bar. The other eight are recalled element-table numbers and could not be re-verified: every susceptibility-table host tried was refused by the egress proxy (fizika.si, periodictable.com, hyperphysics, wikidoc). Separately, O chiMolar +3449 is a Curie-law quantity quoted at one temperature (about 293 K) which the chart never states, and the O mrNote reproduces it as a bare fact ('Molecular O2 is paramagnetic at +3449e-6 cm3/mol'); at body temperature it is about 4 percent lower, and it is a gas-phase number being used to reason about dissolved O2. No primary measurement of it was reachable; searches returned only solid-O2 and high-pressure work.

2. **Narrowed.** The anisotropy is now stated on the carbon cell itself, which says the -6.0e-6 is graphite's in-plane value, that the two single-crystal components are about -5.2e-6 and -253e-6, that a randomly oriented powder averages near -88e-6, and that carbon-fibre and pyrolytic-graphite hardware therefore perturbs B0 far more than the cell suggests. What stands is the underlying complaint: a single scalar is the wrong shape for this quantity, and carbon is the entry where that costs a reader of this particular chart the most.

3. Chemical shift ranges in this slice, sorted by how well anchored they are. Well anchored: 13C '0 to 220' (the routine organic window, and the mrNote's derived 7 kHz at 3 T checks out at 7067 Hz), 1H '-1 to 14', 2H '-1 to 14' (2H shifts track 1H to within the isotope shift, so copying the window is defensible). Constructed windows, defensible but not tabulated endpoints: 11B/10B '-120 to +100', 9Be '-2 to +30' (the isotope note's 'about 30 ppm' just restates the window; published 9Be spans are usually quoted nearer 40 ppm), 14N/15N '-400 to +600', 17O '-40 to +1100' (whose upper end is too low: oxometalate delta(17O) alone spans about 1200 ppm and permanganate sits near +1230), 19F '-300 to +100'. Arguably should be null or explicitly labelled as solution-chemistry windows: 3He and 6Li/7Li, for the reasons in the two entries below. Nothing in this slice was checked against a Bruker or IUPAC shift compilation, because every host carrying one was blocked by the egress proxy (kodu.ut.ee, triton.iqfr.csic.es, nmr.oxinst.com, chem.ch.huji.ac.il, old.iupac.org, www2.chem.wisc.edu).

4. 1H shiftRange 'in vivo 0 to 5' is a constructed window that excludes an established measurement band. Downfield 1H MRS is published in human brain at both 3 T and 7 T and resolves NAA amide near 7.8-7.9, GSH/NAD+/homocarnosine near 8.0-8.2 and mobile-protein amides at 8.1-8.4 ppm; homocarnosine's imidazole doublets sit at about 7.1 and 8.1. The upfield end is also soft: macromolecule and lipid resonances run below 0.9 ppm. Since this is the anchor nucleus of the whole chart, the window should probably read '0 to 5 for the routine window, 6.5 to 8.5 for downfield MRS'.

5. 3He shiftRange '-70 to +10 for endohedral and dissolved 3He' overstates the low end. The endohedral-fullerene 3He literature is quoted as spanning about -50 to +10 ppm: 3He@C60 at -6.3 to -6.4, 3He@C70 at -28.8, and the extremes reached by the hexa-anions (He@C60(6-) shielded a further ~43 ppm, He@C70(6-) deshielded ~37 ppm) land near -49 and +8. I found nothing at -70. Dissolved 3He in ordinary solvents moves only a few ppm. Search-confirmed against the PCCP 2023 symmetry paper and the original Nature 1994 report; the primary tables themselves were blocked.

6. 7Li and 6Li shiftRange '-10 to +5' is a solution-chemistry window and is not flagged as one. Lithium metal and intercalated lithium carry a Knight shift of roughly +250 ppm, far outside the stated range, exactly the situation these notes already flag for 133Cs. Nothing in this slice warns the reader. The in vivo Li+ line sits at 0 and the window is fine for the MR use case, but the endpoints themselves are not quoted from a compilation.

7. 14N and 15N both carry reference 'CH3NO2 (neat)' and nothing else. That is the IUPAC primary reference, but IUPAC additionally recommends liquid NH3 as a secondary reference for 15N in aqueous solution, and the biomolecular and hyperpolarized-15N literature, which is the use case this chart is about, reports on the ammonia scale. The two scales differ by 380.2 ppm (search-confirmed). The shiftRange '-400 to +600' is only meaningful on the nitromethane scale and would be '-20 to +980' on the ammonia scale. A reader who takes a 15N shift from a hyperpolarization paper and reads it against this row will be 380 ppm out.

8. H mrNote 1 mixes two conventions in one sentence: '63.87 MHz at 1.5 T, 127.73 MHz at 3 T and 297.2 MHz at 7 T'. The first two are exact-field arithmetic from the page's own gamma (1.50008 T and 2.99994 T), but 297.2 MHz corresponds to 6.9802 T, the nominal field of a 7 T human scanner, not 7.000 T, which gives 298.04 MHz. Both 297.2 and 298 are quoted in the literature for '7 T'. The sentence should either be arithmetic throughout (63.87 / 127.73 / 298.0) or scanner-nominal throughout (63.6 / 123.2 / 297.2 for Siemens, 63.87 / 127.73 / 297.2 for the mix now written).

9. F mrNote 2 gives 19F as '120.15 MHz at 3 T'. The page's own gamma of 40.0775 MHz/T gives 120.233 MHz at 3.000 T, and scaling the quoted 1H 127.73 MHz by the IUPAC Xi of 94.094011 percent gives 120.186 MHz. 120.15 is 0.07 percent below both, and the chart's own fluorine shift table implicitly uses 120.2 (its '126 ppm, roughly 15 kHz at 3T' only works at 120.2). Small, but this is a value a coil builder would read off the page, and crosscheck.py's prose pass does not look at slice files at all, only at dossier_*.json and shifts_*.json, so nothing caught it.

10. F mrNote 4, 'The 19F shift range spans about 400 ppm (roughly 48 kHz at 3 T), so multi-resonance agents such as PFOB need broadband excitation', is internally inconsistent with the chart's own 19F shift table, which says organofluorine covers more than 300 ppm and over 800 ppm if inorganic fluorides are counted, that the in vivo agents together span about 126 ppm (15 kHz at 3 T), and that PFOB itself spans more than 60 ppm. PFOB's own lines run -63.7 to -126.6 ppm, 62.9 ppm, 7.6 kHz at 3 T. The 400 ppm figure is just the shiftRange field restated and does not describe any real 19F acquisition. The conclusion (no single pulse covers the table) is right; the number attached to it is the wrong one.

11. H mrNote 2's arithmetic does not support its own conclusion: 80 M against 1 to 10 mM is a factor of 8,000 to 80,000, four to five orders of magnitude, not the 'about 3 orders of magnitude of water suppression' the sentence concludes. Three orders is the right figure for what CHESS/VAPOR actually deliver in practice, so both halves are individually defensible and the 'so' joining them is not. There is also a unit mismatch: 80 M is a proton molarity (the blurb says so, this note does not) while the 1 to 10 mM metabolite figures are molecule molarities.

12. H mrNote 3, '2H ... its short T1 near 300 ms', is one value standing for a family that spans a factor of six. Measured 2H T1 values: HDO 346 +/- 5 ms at 4 T and 362 +/- 6 ms at 9.4 T, grey/white matter 320/290 ms at 7 T, CSF 510 ms, and the [6,6'-2H2]glucose downstream metabolites run 60 to 350 ms with glucose itself near 60-64 ms. 'Near 300 ms' is the tissue-water number and the TR argument built on it holds, but as written it reads as a property of the nucleus rather than of one metabolite at one field.

13. Li mrNotes carry four in vivo numbers that are all single-study or across-study means: brain lithium '0.2 to 0.8 mM', brain/serum ratio 'around 0.5', 'brain 7Li T1 runs to several seconds', and 'needs 3 T or higher'. Published brain/serum ratios by 7Li MRS are 0.59 +/- 0.12, 0.52 +/- 0.16 (range 0.30 to 0.80), 0.39 +/- 0.08 for average brain against 0.92 +/- 0.16 at the local maximum, and 0.5 to 1.3 over a 48 h dosing cycle; the ratio also depends on age and on time since dose. 0.5 is a fair central value but it is a mean of studies that disagree, which is what these notes already say about lithium under the single-line shift table. The '3 T or higher' floor is also a judgement: human 7Li MRS was done at 1.5 T.

14. B mrNote 1, '11B MRS and MRI of boronophenylalanine and BSH have been used to estimate tumour boron before neutron irradiation', reads as clinical practice. A 2026 review of boron quantification for BNCT (Med. Res. Rev., Selg et al.) states that none of the 11B, 10B or 19F MRI/MRS methods has been tested in humans or implemented in clinical BNCT; the published work is rodent, at 9.4 T. The 'tens of ppm boron in a large voxel' detection limit is corroborated (about 20 ppm boron at 6 x 6 x 6 mm in rat MRI) but is a preclinical high-field figure, not a clinical one. The blurb's hedge ('makes non-invasive boron dosimetry by MRS plausible') is right; the mrNote is not hedged.

15. C mrNote 1, 'an in vivo T1 of 30 to 60 s' for hyperpolarized [1-13C]pyruvate, is field- and compartment-dependent and the range as written spans the solution and in vivo cases at once. Reported values: 68 +/- 0.5 s for the C1 carboxylate at 1 T, 31 +/- 1 s in blood in vitro, about 50 s in brain tissue, and an apparent in vivo T1 of 7.5 +/- 1.3 s where the decay includes washout and dilution as well as relaxation. 30 to 60 s is the figure usually quoted for clinical 3 T work and I did not find it wrong, but it is not a single measured quantity.

16. N mrNote 3 says liquid nitrogen 'is the thermal shield coolant and the working fluid for cryogenic RF probes; N2 gas is the standard purge in magnet rooms'. The first clause is true of high-resolution NMR magnets and of older MRI cryostats but not of modern whole-body magnets, which cool the radiation shield with a two-stage cryocooler and carry no LN2 reservoir, and the same page's He mrNote already says modern MRI cryostats are sealed or zero boil-off. The cryoprobe clause is also partial: Bruker-style cryoprobes cool the coil with helium gas near 20 K and use LN2 only for the preamplifier stage. And magnet rooms are not N2-purged; dry N2 purges the bore, shim stack and probe of high-resolution systems. Stated as general MRI practice, it is not.

17. N mrNote 4, 'N2 is the quenching gas in Rb spin exchange optical pumping cells for 3He and 129Xe, typically a few hundred torr', is the 129Xe figure applied to both. Xe SEOP cells run roughly 10 percent N2 in a several-atmosphere mix, so a few hundred torr; 3He cells run about 0.05 to 0.1 amagat of N2 against 8 to 12 atmospheres of 3He, i.e. tens of torr. Not re-verified against a primary source this session, but the two cell designs differ by an order of magnitude in N2 partial pressure and the note gives one number for both.

18. O mrNote 1, 'the sensitivity penalty makes it a 7 T and above method', is contradicted by published human work. Regional CMRO2 in human brain has been quantified with dynamic 17O MRI on a clinical 3 T system, and review text states 17O MRSI/MRI has been applied to human brain CMRO2 'at varied magnetic field strength from 3 T to 9.4 T'. 7 T and above is where it is practical, not where it is possible; as written it is a rule rather than a preference.

19. Role tokens are assigned inconsistently at two places in this slice. He carries ['hyperpolarized','magnet'] but not 'optical-pumping', although 3He is itself optically pumped (MEOP directly, SEOP with Rb) and helium is the buffer gas in 129Xe SEOP cells, while N carries 'optical-pumping' purely for being the quench gas in those same cells; Xe carries both tokens. O carries ['invivo-mrs','tracer'] but not 'contrast', although its own mrNote makes molecular O2 the mechanism behind oxygen-enhanced MRI, and 'contrast' is otherwise given to every element that shortens a relaxation time (Mn, Fe, and the lanthanides). The page renders 'optical-pumping' with the short label 'SEOP', which is an argument for leaving He out on MEOP grounds but not for leaving it out as a SEOP buffer gas.

## Elements Z = 10 to 19 (Ne through K)

20. chiMolar throughout Z = 10 to 19 is the CRC element-table value carried over from recall, and the CRC table itself could not be opened (both mirrors of 'Magnetic susceptibility of the elements and inorganic compounds', plus periodictable.com and every other tabulation tried, were refused by the egress proxy). Only two of the ten were confirmed against quotable text, and both only at second hand: Cl2 at -40.5e-6 cm3/mol and P at -20.8e-6 cm3/mol, from element-infobox text that itself cites CRC. Na +16.0, Mg +13.1, Al +16.5, Si -3.9, S -15.5 and K +20.8 were confirmed against nothing this session. The per-mole convention also breaks down at P and S and the file does not say how it was resolved: the halogens here follow the diatomic rule (Cl2 -40.5, consistent with F2 -9.63 and Br2 -56.4 next door), but phosphorus's standard state is molecular P4 and sulfur's is S8, and -20.8 and -15.5 are per mole of ATOMS. Neither allotrope is named either: -20.8 is white phosphorus, where red and black differ, and -15.5 is alpha-rhombic sulfur.

21. **Still open, and now sharper.** Ne and Ar have been moved onto the CRC values, -6.96e-6 and -19.32e-6, for consistency with the rest of the column, but the objection in this note survives the move and applies to CRC too: the exact relativistic calculations (Phys. Rev. A 109, 012820, 2024, converted here from a search snippet rather than from the paper) give -7.57e-6 for neon and -20.5e-6 for argon, so both tabulations are low, and argon has moved slightly further from theory rather than closer. Helium is the one noble gas where the departure was acted on, because its diamagnetism is calculable essentially exactly: the cell carries -1.89e-6 against CRC's -2.02, and that exception is recorded in crosscheck.py. Ne and Ar deserve the same treatment once the 2024 paper can be read first-hand; Kr and Xe probably do too, and no modern value for either was found.

22. The magnetic moments in this slice are not Stone's 2025 recommended values. They are printed to seven significant figures and reproduce the stored gamma through mu*mu_N/(I h) to better than 25 ppm everywhere (worst case 41K at 24 ppm), which is the signature of a standard NMR property table rather than of Stone; against the EasySpin/Stone file they run up to 0.09 percent off - 31P 1.1316 vs 1.130925, 33S 0.6438212 vs 0.64325, 25Mg -0.85545 vs -0.85533, 27Al 3.641507 vs 3.64070. None of that trips crosscheck.py's 0.2 percent threshold, so the mixed provenance is invisible on the page, and I could not open the table the seven-figure values come from in order to name it.

23. No IUPAC Xi in this slice was verified against IUPAC; every route to the recommendations was blocked (publications.iupac.org, the tandfonline reprint, the mriquestions.com PDF copy, pascal-man.com). The check I could do is internal: each Xi read against the moment-derived gamma implies a shielding for the reference species, and those come out 567, 554, 714, 788, 774, 911, 1055, 1149 and 1245 ppm for 21Ne, 23Na, 25Mg, 27Al, 29Si, 31P, 33S, 35Cl and 39K - a smooth rise with Z of the size a free-ion Lamb term should have, with 35Cl/37Cl agreeing to 1 ppm and 39K/41K to 2 ppm. 40K is the exception: its Xi = 5.802018 implies a shielding 108 ppm larger than 39K and 41K give, and all three are referenced to the same 0.1 M KCl in D2O, where the shielding must be identical. Either that Xi or mu(40K) = -1.2981 (five figures, so +/-40 ppm on its own) is the weak entry. 40K also carries a reference compound and a frequency ratio although no routine 40K NMR exists, and I could not establish whether IUPAC lists 40K at all or whether the number was computed from the moment.

24. The 21Ne isotope note says 'Stone's compilation lists +10.3(8) fm2'. The Stone table as actually distributed with EasySpin and cached in this tree prints +0.102 b, that is 10.2 fm2. A search snippet on the 2013 introduction does support 'Q(21Ne) = 103(8) selected by Stone', so the two may be different editions or the intro may differ from the data table, but I could not open either (www-nds.iaea.org blocked) and as written the note contradicts the file the repository checks itself against.

25. Two quadrupole-moment provenance items. Q(25Mg) = +19.94 fm2 is Pyykko's atomic value, confirmed by snippet as 199.4(20) mb, but a 2025 configuration-interaction-plus-coupled-cluster determination gives 203(2) mb, 1.8 percent higher; the 27Al record names its own 2019 revision (+14.66 against the proposed +14.82 fm2, both confirmed exactly as 146.6(10) and 148.2(5) mb), so by the file's own convention 25Mg should say the same and does not. Separately, Q(33S) = -6.94 fm2 is a deliberate Pyykko 2017 substitution for Stone's -6.78 and the isotope note says so, but unlike the other three Pyykko substitutions in the data set it is missing from crosscheck.py's EXPLAINED dictionary, so it reappears as an unexplained 2.4 percent 'minor difference' on every run of the checker.

26. Three shift ranges in this slice are constructed windows rather than quoted ones. 25Mg '-15 to +25 for most Mg(II); the full diamagnetic range approaches 200 ppm' has no endpoint I could anchor to anything and should arguably be null, or keep only the Mg(II) part. 27Al is half anchored: AlO6 near 0 and AlO4 +55 to +80 are confirmed (search gives 6 +/- 5 and 64 +/- 5 ppm), but '+35' for AlO5 is a midpoint I chose between reported values running from about 27 to 40 ppm, and the -50 and +300 outer endpoints come from no source I reached. 33S 'roughly -100 to +400' understates the nucleus: SO2 at +375 is confirmed to +375 +/- 2 ppm from sulfate and the reference at 0 is right, but the 33S range is reported as exceeding 800 ppm and often quoted near 1000, and the parenthetical 'metal sulfides span over 600 ppm' does not say that they lie BELOW the window, so a reader takes -100 as the low end of the nucleus.

27. 23Na shiftRange '-60 to +15 for Na compounds (sodide near -60)' is a Na(I)-compound window, not the 23Na range. The sodide endpoint is solid, search-confirmed at -61 to -63 ppm for Na- in cryptand and methylamine solutions, but sodium metal's Knight shift sits at +1131 ppm, seventy-five times outside the stated window, and nothing in the 23Na record flags it. Item 14 of these notes flags exactly this situation for caesium metal; 23Na deserves the same flag, or the window should be relabelled.

28. 29Si shiftRange '-400 to +50' mixes two populations. The common-compound window is about -200 to +50 (search-confirmed as 'the majority of 29Si shifts lie between +50 and -200'), while the benchmark set that reaches -400 at the bottom reaches +828 at the top, and silylium cations alone are confirmed at +225.5. So the low endpoint comes from the extreme population and the high endpoint from the ordinary one. Either -200 to +50 or -400 to +830 would be defensible; the printed pair is not.

29. 31P shiftRange is the entry a reader will check first and it carries three soft things. Its in vivo anchors are not fixed positions: Pi at +4.9 is a pH readout and beta-ATP at -16.3 is the free-Mg readout, the two quantities the same entry's mrNotes say are measured FROM these shifts, so both move with the physiology being reported (a 7T study quotes beta-ATP at -16.15, others -16.26). The window -25 to +25 is also wider than the real in vivo spread, which runs about -20 to +10. And 'about -250 to +250 in chemistry' is a working organophosphorus window, not the 31P range: white phosphorus P4 sits near -490 ppm, phosphinidene complexes run past +1000, and published 31P shift scales span roughly -500 to +2500.

30. 31P reference and shiftRange sit on two different zeros and the page prints them side by side. The reference field says 85 percent H3PO4, which is the chemistry scale; the shiftRange says 'PCr 0', which is the in vivo scale, and PCr lies about -2.5 ppm on the H3PO4 scale. I could confirm only that PCr is conventionally re-referenced to 0 in vivo and that its H3PO4-scale position is near -2.5 ppm; the exact offset, usually printed as -2.52, is not pinned to a retrievable primary source here.

31. Larmor frequencies quoted in the mrNotes (23Na 33.8 MHz at 3T and 78.9 at 7T; 31P 51.8 and 120.8; 35Cl 29.2 at 7T; 39K 13.9 at 7T) are gamma times a nominal field and inherit two assumptions, neither stated. They use exactly 3.000 and 7.000 T, while systems sold as 3T and 7T run near 2.89 and 6.98 T - the 31P frequency on a 2.8936 T scanner is about 49.9 MHz, two megahertz from the printed 51.8. And the stored gamma is the bare-nucleus value, whereas what a console tunes is Xi times the measured 1H frequency, 0.09 percent lower for 31P (51.71 against 51.75 MHz at 3.000 T exactly). Both figures are arithmetic and both are defensible; the card by the console will not match either exactly.

32. 31P mrNote 'T1 is long, roughly 3 to 6 s for PCr at 3T' is brain-weighted and reads as general. The best-measured 3T value is calf muscle, where PCr T1 is 6.4 +/- 0.2 s and 6.7 +/- 0.4 s in two studies, at or above the top of the stated window, falling to about 4.0 s at 7T. Muscle is where the fast exercise-recovery work described two bullets later is done, and it sits outside the window given.

33. 23Na mrNotes give relaxation windows that are true of parenchyma and read as general. T2* short 0.5 to 5 ms and long 15 to 30 ms is the literature brain-parenchyma range and is confirmed, but at 7T white matter the long component is measured at 35.2 +/- 12.1 ms, above the window, and CSF relaxes monoexponentially near 43 ms. Likewise 'T1 is short, about 15 to 40 ms' holds for tissue (white matter 35.1 +/- 8.4 ms at 7T) but not for CSF at 65.5 +/- 15.9 ms, so the advice that follows it, 'TR can be 100 ms or less', saturates the very compartment the 140 mM figure in the same entry refers to.

34. Mg blurb explains 25Mg's invisibility by the wrong pool. 'Free intracellular Mg2+ is only a few tenths of a millimolar' is a brain-specific 31P-MRS number (0.2 to 0.4 mM on the Iotti-type calibrations, and calibration-dependent at that; muscle runs higher), but a 25Mg experiment does not see only free Mg2+ - total tissue magnesium is of order 10 to 20 mM, comparable to the tissue sodium the same chart calls detectable. The real reasons are the 2.7e-4 receptivity and the quadrupolar broadening of bound Mg, both of which the entry also gives; the concentration clause should not carry the 'so' that follows it.

35. 29Si invivo is set true, which puts it in the same column as 23Na and 31P, on the strength of hyperpolarized silicon particles imaged in MICE by direct injection into gut and peritoneum. There is no human work. The T1 'near 40 min' is confirmed as 39 +/- 3 min for roughly 2 micrometre particles (Cassidy 2013), but silicon particle T1 varies from minutes to hours with size and surface chemistry, so it is a value for one preparation rather than for the nucleus.

36. Ar mrNote gives the 39Ar half-life as 269 y. That is the long-accepted Stoenner 1965 value, but a direct measurement from 3.4 years of DEAP-3600 data reports 302 +/- 8 +/- 6 y, 12 percent longer and outside the old error bar. I kept 269 as the tabulated value; the disagreement is not stated in the entry.

37. Al mrNote 'RF skin depth in aluminium is about 7 micrometres at 128 MHz' is my own arithmetic, not a quoted figure: sqrt(2 rho / omega mu0) with rho = 2.65e-8 ohm m gives 7.2 micrometres. That is right for pure annealed aluminium at room temperature, but the 5000- and 6000-series alloys actually used for screens and formers have resistivities 30 to 60 percent higher, which moves it to 8 or 9 micrometres. The conclusion drawn from it, that foil is thick enough, survives either way.

## Elements Z = 20 to 30 (Ca through Zn)

38. chiMolar throughout Z = 20 to 30 is the CRC element-table value carried from memory, and I could re-verify none of it against a primary table: every host that carries that table (fizika.si, gemstonemagnetism.com, periodictable.com, tandfonline, en.wikipedia.org) is refused by the egress proxy. Search snippets corroborated Ca +40.0, Ti +153 (293 K), Mn +511, Cu -5.46 and Zn -11.4 x 10^-6 cm3/mol; V +255 and Cr +180 were not corroborated at all. The convention is the right one here (CGS, per mole of the element, and every element in this slice is monatomic in its standard state, so the diatomic caveat that bites H, N, O and the halogens does not apply).

39. **Resolved.** CRC gives scandium as +295.2e-6 cm3/mol, so the file's +295 was right and the two search snippets giving +315e-6 'at 292 K', one of them attributing it to CRC, were both wrong. Leaving the value standing rather than swapping it for a second-hand number was the correct call.

40. Fe, Co and Ni have chiMolar null because they are ferromagnetic at room temperature, which is right for the column as the page draws it (the legend puts ferromagnets in solid ink rather than on the diverging ramp). Filling them would need a stated convention that the rest of the column does not use: an initial susceptibility, a technical saturation value, or a Curie-Weiss extrapolation from above Tc. Nothing short of that would be comparable with the +180 printed next to Cr.

41. The magnetic moments for 47Ti, 49Ti, 50V, 51V, 53Cr, 55Mn, 57Fe, 59Co and 61Ni are all 0.25 to 0.64 percent larger in magnitude than the Stone/EasySpin values in the repository's own cached reference table, consistently in one direction; crosscheck.py lists every one of them under 'minor differences' because they fall under its 1 percent hard threshold, and no source is named anywhere for the larger values. The pattern is what a diamagnetic-shielding correction looks like, but it reverses sign for 63Cu, 65Cu and 67Zn, where the file sits 0.01 to 0.03 percent BELOW Stone. The consequence is visible in the shielding you get by dividing the IUPAC frequency ratio by the file's own moment-derived gamma: 1461 ppm (43Ca), 1570 (45Sc), 1572/1576 (47/49Ti), 1262 (50/51V), 1986 (53Cr), 2045 (55Mn), 2168 (57Fe), -2513 (59Co, negative as expected because K3[Co(CN)6] is strongly deshielded), 1746 (61Ni), 750/791 (63/65Cu), 1702 (67Zn). Copper at about 780 ppm, against nickel at 1746 and zinc at 1702 on either side of it, is the one place in this slice where the shielded-versus-bare split does not behave; it wants either a different Cu moment or a sentence in the isotope note.

42. q for 51V is -5.2 fm2, which matches neither of the two tables the file could be drawing on: the repository's cached Stone/EasySpin reference says -0.043 b (-4.3 fm2), and the current recommended value that search returns for 51V is -4.8 fm2. Pyykko's 2017 table itself is on chem.helsinki.fi and tandfonline, both blocked, so I could not read the recommended value directly and did not substitute a number I have only from a snippet. See also the correction below on the isotope note's attribution.

43. q for 43Ca is -4.44 fm2, from combining theory with the measured electric-quadrupole hyperfine constant of the 3d 2D5/2 state of 43Ca+. That determination is real and is what crosscheck.py's allowlist records, but it sits inside a live disagreement: reported values for 43Ca span -4.08(8) to -6.5(20) fm2, and a 2025 hybrid configuration-interaction / coupled-cluster recalculation gives -4.79(6) fm2. The file's -4.44 is a defensible middle, not a settled number.

44. q is null for 41Ca and for 60Co, and both isotope notes justify it as 'not well established' / 'not confidently tabulated'. The repository's own reference table does carry values for both: crosscheck.py reports 'Q is null; the reference has -6.650 fm2' for 41Ca and '+46.000 fm2' for 60Co. So these are judgments against the reference rather than absences in it, and the notes should say which, or the values should go in. Neither nucleus is an NMR nucleus, so nothing on the page depends on it.

45. The 43Ca shift reference is written 'CaCl2 in D2O' with no concentration. The 43Ca shift is strongly concentration dependent, which is why IUPAC specifies 0.1 M CaCl2 in D2O and why the solid-state community works instead on a 1.0 M CaCl2 in H2O scale (solid CaO sits at +136 ppm on that scale). The same omission runs through Sc(NO3)3 in D2O, K2CrO4 in D2O, KMnO4 in D2O and Zn(NO3)2 in D2O, but 43Ca is the one where it changes the zero.

46. Chemical shift ranges, which are well anchored: 53Cr -1800 to +100 has real endpoints (chromate at 0, Cr(CO)6 near -1800 and the note names both species); 59Co -3000 to +15000 reproduces the standard 18000 ppm figure for the widest shift range in NMR; 63Cu/65Cu -100 to +1000 sits against a published overall 63Cu span of 980 ppm, though the -100 end is mine and not sourced.

47. Chemical shift ranges that are constructed and narrower than the literature: 43Ca -30 to +160 against a reported 43Ca isotropic range of more than 250 ppm, and 45Sc -100 to +300 against ligand-resolved ranges that reach +570 ppm for nitrogen donors and -370 ppm for carbon donors, with a 2020 paper whose title is about extending the 45Sc range further. Both windows describe aqueous and oxide chemistry and quietly exclude organometallics. They are usable as drawn but they are not the range of the nucleus.

48. Chemical shift ranges that are constructed windows with no endpoint I could source: 47Ti/49Ti -1600 to +400, 51V -2000 to +500, 55Mn -3000 to +100, 57Fe -1000 to +12000, 67Zn -50 to +400. The 57Fe window is the most defensible of these, since an overall 57Fe range of about 12000 ppm is quoted in the literature and ferrocene (+1531) and the ruffled iron porphyrins (+7258 to +8036) fall inside it. The 50V entry is the one that should arguably be null: it inherits 51V's window, and essentially no 50V chemical shifts are reported at all. 61Ni is already null on exactly that reasoning and its note says so, which is the right treatment.

49. The 59Co isotope note says the IUPAC frequency ratio 'sits about 2400 ppm above the moment-derived frequency'. Recomputing from the file's own numbers (gamma 10.07707 MHz/T, Xi 23.727074 percent, 1H gamma 42.5774806) gives 2513 ppm. The physics of the sentence is right and the rounding direction is wrong; it is arithmetic from stored primitives, not a quoted figure.

50. Fe mrNotes: 'fully deoxygenated blood is shifted about 0.27 ppm per unit haematocrit from fully oxygenated'. The number is the correct modern one (Spees 2001 at 37 C, confirmed at 0.273 +/- 0.006 by Jain 2012) but it is a CGS value, and 4 pi times larger, about 3.4 ppm, in SI. Everywhere else the chart quotes susceptibility in ppm SI (titanium +182, copper -10, tissue baseline -9), so the unit convention silently changes inside one bullet. Two smaller points in the same sentence: the older Weisskoff and Kiihne figure is 0.18 ppm CGS and the file chose the newer one without saying so, and a susceptibility difference is being described as a 'shift'.

51. Cr mrNotes gives CoCrMo a 'volume susceptibility of order +1000 ppm SI'. The implants dossier deliberately declines to put a number on CoCrMo, saying only that it 'sits between titanium and 316L', and the implants dossier entry in these notes records that as a decision because no defensible ppm value was found. +1000 ppm SI does sit between the dossier's +182 and +3520, so it is not inconsistent in magnitude, but the two places should agree on whether the number exists.

52. Mn mrNotes: 'Mn(II) and Mn(III) chelates are now in trials as gadolinium-free alternatives with r1 of roughly 2 to 6 s^-1 mM^-1 at 1.5 T'. No agent, medium or temperature is named. Mangafodipir itself is quoted near 1.5 mM^-1 s^-1, and the current trial agents sit nearer 2 to 4 at 1.5 T in plasma; I could not tie the upper bound of 6 to a specific compound. As written it reads as a property of the class.

53. Zn mrNotes: 'Zn-responsive Gd(III) smart agents, which switch r1 by roughly a factor of two on binding'. That is one family of agents (the GdDOTA-diBPEN line), the factor depends on serum albumin being present to bind the ternary complex, and the reported enhancement runs from about 1.7 to over 3 depending on the agent and the medium. True of the leading example, written as true of the approach.

54. Tier: Ca is 'exotic' while both its isotopes carry invivo false and its own mrNote says no in-vivo 43Ca spectrum has been reported. The legend prints 'not observed in vivo' for tier 'none', which would make Ca 'none'. Against that, the chart's practice everywhere is looser than the legend (Gd and Nd are 'workhorse', Pb 'exotic', and Mn is 'established' although the file states 55Mn is never detected in vivo), so Ca is consistent with the chart and inconsistent with the chart's own key. This is a chart-wide wording problem that this slice happens to expose.

55. Role tokens: 'tracer' is glossed as nuclear medicine and is applied to 29 elements, including Sr, Ga, Mo, Zr and Au on exactly this basis. It is missing from Co, whose mrNotes name the 60Co Gamma Knife and teletherapy source and the 57Co gamma-camera flood source and which carries isotope records for both, and from Ti, whose mrNotes name 45Ti as an emerging PET nuclide. Co is the stronger case of the two; Ti's 45Ti has no isotope record, so leaving it untagged is defensible. (Cu and Cr are the clear-cut cases and are filed as corrections.)

56. Co blurb calls 59Co 'the largest chemical shift range of any nucleus, around 18000 ppm'. The figure is standard and search-confirmed, but the chart's own 99Ru and 101Ru entries in slice_4 are drawn at 0 to +18000, an exact tie, and the shift-range entry for Z = 43 to 54 records that Ru range as search-confirmed at '>18000 ppm'. Either the superlative should be softened or the two entries should be reconciled.

57. Q(51V) is -5.2 fm2 here, the reference table gives -4.3 fm2, and a third value of -4.8 fm2 circulates. Which of Stone's successive compilations and which of Pyykko's tables carries which value could not be established this session: the v0.2 audit and this repository's own cross-check script disagreed with each other about the attribution, and neither could open the tables to settle it. The isotope note now states only the spread, which is a 20 percent disagreement on a quadrupole moment and therefore on every coupling constant derived from it.

## Elements Z = 31 to 42 (Ga through Mo)

58. chiMolar for all twelve elements is the CRC-derived element table (10^-6 cm3/mol, CGS, per mole of element in its standard state), not a primary measurement. Confirmed from search snippets of element data sheets that trace to CRC: As -5.5 (gray As), Se -25.0 (298 K), Br -56.4 (liquid, per mole of Br2, matching the convention the earlier slices used for I2), Kr -28.8 (298 K), Rb +17.0 (303 K), Sr +92.0 (paramagnetic, positive - I specifically checked for a sign error here and there is none), Ga -21.6 (290 K), Ge -76.84, Mo +89.0 (298 K). NOT confirmed against any source this session: Y +191.0, Zr +122.0, Nb +208.0. Those three are consistent with the mass susceptibilities I remember for the metals (Y about +2.15e-6, Zr +1.34e-6, Nb +2.28e-6 cm3/g) to within a few percent, which is a plausibility check and not a verification. Every attempt to reach the CRC table itself failed: fizika.si, gemstonemagnetism.com, googleschnoogleresearchinstitute.org, periodictable.com and schoolmykids.com were all refused by the egress proxy.

59. **Resolved, in this note's favour.** It predicted -11.5 to -12.0e-6 cm3/mol for germanium from a 2023 ab initio study and declined to act only because the column's stated convention was the element table. CRC gives -11.6, inside that interval, so the physics and the stated source turned out to agree and the element table was the thing that was wrong. The cell now reads -11.6. This is the clearest single case of the failure the retracted spot-check describes, and it was caught before the source was in hand.

60. Ga chiMolar -21.6 is a single scalar standing in for a strongly anisotropic and strongly state-dependent quantity. The single-crystal measurement (J. Appl. Phys. 31, 1802, 25 C) gives -0.119, -0.416 and -0.229 x 10^-6 cm3/g along a, b and c, i.e. -8.3, -29.0 and -16.0 x 10^-6 cm3/mol, a 3.5:1 spread; the polycrystalline average in the same work is -0.257 x 10^-6 cm3/g = -17.9 x 10^-6 cm3/mol at 25 C, while -21.6 corresponds to -0.310 x 10^-6 cm3/g, closer to that work's -196 C value. Worse for a chart a reader may use at room temperature: gallium melts at 29.8 C and liquid gallium was measured at +0.0031 x 10^-6 cm3/g, i.e. +0.2 x 10^-6 cm3/mol, essentially non-magnetic and of the opposite sign. The element label 'diamagnetic' is true of the solid only.

61. Chemical shift ranges, sorted by how much they are worth. Well anchored: 95Mo and 97Mo '-2000 to +4000', which is the standard literature statement of the 95Mo range (search-confirmed as 'from +4000 to -2000 ppm'), with Mo(CO)6 at -1857 and thiomolybdates above +2000 both consistent with it. Partly anchored: 73Ge '-1100 to +50', an envelope drawn around the three compounds named in its own note (GeI4 near -1080, Me4Ge 0, GeCl4 +31) rather than a quoted range, and it takes no account of Ge(II) species; 89Y '-100 to +1400', whose 1500 ppm width is supported indirectly by the 89Y DFT-prediction literature quoting +/-70 ppm as '5 percent of the chemical shift range', though neither endpoint is quoted anywhere I reached. Constructed windows: 69Ga and 71Ga '-700 to +750' and 77Se '-1000 to +2500'. No compound is named at any of those four endpoints, and the Se blurb's 'shift range near 3500 ppm' is nothing more than the width of the window it is describing. 93Nb '-2200 to +300' is the odd case: the niobate assignments inside its isotope note ARE sourced (search-confirmed against the 93Nb niobia shift scale, which gives -600 to -2000 overall and -650 to -950 for four-coordinate Nb), but the two endpoints of the chart's own range are not, and nothing is named that resonates downfield of the NbCl5 reference at all.

62. shiftRange is null for 75As, 79Br, 81Br, 83Kr, 85Rb, 87Rb, 87Sr and 91Zr. For As, Sr, Zr and Kr that is defensible and the isotope notes say why. For 87Rb it is arguably inconsistent: the isotope note itself states 'solid-state 87Rb isotropic shifts span roughly 150 ppm across common salts', so the chart asserts a range in prose while leaving the field empty. Either the field should carry roughly -50 to +100 with a solid-state qualifier, or the prose claim should be dropped. Filling it would need a solid-state 87Rb shift compilation, which I could not reach.

63. Q(93Nb) = -32.0 fm2 is -0.32(2) barn in the IAEA 2021 table, adopted from Pyykko 2018 and traceable to a 1973 muonic X-ray hyperfine measurement (1973Po15, Nucl. Phys. A217 573). That is a stated 6 percent uncertainty on a fifty-year-old determination with no modern redetermination that I could find, and it is the number the 93Nb line of the hardware dossier leans on. The value itself is right for the chart's stated scale; the precision it implies by being printed as '-32.0' is not.

64. The Nb entry underpins the implant line of the hardware dossier by arithmetic, not by an independent source. chiMolar +208 x 10^-6 cm3/mol with the density of niobium, 8.57 g/cm3, gives 4*pi*208e-6*8.57/92.906 = 241.1 ppm SI, which is exactly the 'niobium +241 ppm' printed in the dossier's implant-metals fact. The figure is sound but inherits both the unverified +208 (see the chiMolar entry above) and a density taken as exactly 8.57; it is not a quoted volume susceptibility for niobium.

65. 83Kr invivo is set true, the element carries the 'invivo-mrs' role and tier 'emerging', and nothing in the published record supports living-subject work. PubMed indexes seven hyperpolarized 83Kr MRI papers and every one is model surfaces or ex vivo lung: canine lung tissue (PNAS 2005), excised rat lungs in a ventilation chamber (J Magn Reson 2008), excised rat lungs (Phys Med Biol 2011), ex vivo rodent lung (Magn Reson Imaging 2013), and excised rat lungs in the elastase emphysema model (J R Soc Interface 2015). The 2013 J Magn Reson methods paper describes in vivo as a prospect, not an achievement. The tier is the specific problem: 'emerging' elsewhere in this chart means 7Li, 11B, 15N and 17O, all of which have human in vivo data. The isotope note's 'quadrupolar T1, of order 1 to 2 s in vivo' has the same defect twice over - the label should be ex vivo, and the published spread is wider than the window quoted: 0.7 to 3.7 s across excised rat lungs at 9.4 T (2008), narrowing to a reproducible 1.0 and 1.3 s for the respiratory zone and distal airways under a controlled inhalation scheme (2011).

66. 83Kr blurb: 'the third hyperpolarizable noble gas after 3He and 129Xe, and the only one whose quadrupole moment is the point'. Both halves overstate. 131Xe (I = 3/2) is polarized by the same SEOP cell that makes 129Xe, and its quadrupolar relaxation is likewise used as a surface probe, so it is neither later than 83Kr nor a counterexample-free claim; 21Ne has also been optically pumped. The sentence is true of the clinical hyperpolarized gases in routine use and reads as a statement about noble gases in general.

67. 83Kr mrNote: 'gas-phase T1 is seconds rather than minutes, so the polarizer must sit next to the subject'. I could not pin this to a measurement, and the one direct comparison I did reach cuts against the urgency it implies - the 2013 J Magn Reson methods paper reports that 83Kr bulk gas-phase T1 in mixtures containing more than about 1 percent O2 is SLOWER, not faster, than 129Xe in the corresponding mixtures. Gas-phase 83Kr T1 is dominated by wall collisions and therefore by container surface and pressure, so a single 'seconds' figure is a property of a particular cell, not of the isotope. The SEOP polarization claim in the same note IS confirmed: 4.4 +/- 0.5 percent equivalent in pure krypton (PLoS ONE 2012).

68. 77Se mu = 0.5338 departs deliberately from Stone/EasySpin (+0.53506), so it will flag in the programmatic cross-check described on the page. The departure is justified and I verified it: mu(77Se) = 0.533 80(3) muN, Hurajt and Antusek, J. Chem. Phys. 163, 144306 (2025), which re-derives the moment from ab initio shielding of Se(CH3)2. What I could NOT check is the residual: the chart's gamma (8.13788, derived from the new moment) and its IUPAC Xi (19.071513) together imply an absolute shielding for Me2Se of 2210 ppm, and the paper is paywalled, so I could not confirm that 2210 ppm is what its computed sigma(Me2Se) actually is. The isotope note's claim that the OLD pair is off 'by 0.45 percent, far more than any plausible shielding correction' does check out arithmetically.

69. The boilerplate 'magnetic moment not reliably tabulated, left null' was used for eight radionuclides in this slice (67Ga, 68Ga, 82Rb, 89Sr, 86Y, 90Y, 89Zr, 99Mo) and is not equally true of all of them. It was simply false for 67Ga, whose moment is tabulated as +1.8507(3) nuclear magnetons; that entry has since been filled in. It is honest but incomplete for 89Zr, where a value does exist - mu = -1.072(23) muN from NMR on oriented nuclei (Nucl. Phys. A 1997) - and what would be needed to fill the field is a determination better than the 2 percent that measurement carries, not the absence of any measurement. For 68Ga, 82Rb, 89Sr, 86Y, 90Y and 99Mo I could not check either way: the NNDC Stone moments PDF and the IAEA nuclear moments database were both blocked by the egress proxy. Given that the one case I could check was wrong, the other six should be checked individually rather than trusted as a class.

70. Two same-element frequency-ratio inconsistencies that exceed the precision the moments are printed to. Rb: Xi(87)/Xi(85) = 3.3889847 but mu(87Rb)/1.5 divided by mu(85Rb)/2.5 = 3.3896982, a 211 ppm disagreement - and since both isotopes are referenced to the same RbCl in D2O, chemical shielding cancels exactly in that ratio, so the discrepancy sits in the tabulated moments. Mo: the same comparison gives 122 ppm, and the Mo moments (-0.9142, -0.9335) are only four significant figures, so the derived gammas carry roughly 1e-4 relative uncertainty anyway. For contrast the same test on Ga gives 24 ppm and on Br 1.5 ppm, which is why I trust those. Neither Rb nor Mo is wrong enough to matter in practice; both are printed to more digits than they earn.

71. 87Sr mrNote: 'lower symmetry broadens the central transition past usefulness below about 18 T'. The physics is right - second-order quadrupolar broadening scales as 1/B0 - but 18 T is a threshold with no citation behind it and no definition of 'usefulness'. It reads as a specification and is a judgement. The same applies, less sharply, to the Zr note's assertion that almost all reported 91Zr work is solid state.

72. Mo mrNote: 'Molybdenum permalloy (about 4 percent Mo, 79 percent Ni, balance Fe)'. That is the nominal 4-79 Mo-permalloy composition and it varies by grade and vendor, exactly as the mu-metal composition already flagged in the hardware dossier notes does. Not verified against a standard this session.

73. Br blurb: 'Bromide is present in plasma at roughly 50 micromolar'. This is a round physiological figure, not a citation-exact value, and reported serum bromide reference ranges run roughly 20 to 100 micromolar depending on diet and assay. Same status as the 35Cl and 39K tissue concentrations already flagged in the single-line-nuclei notes.

74. 73Ge natural abundance is 7.75 % here. The v0.2 audit proposed 7.76 %, attributing it to the current CIAAW table; the reference table this repository cross-checks against (EasySpin, carrying Stone's compilation) gives 7.73 %. Three values within 0.4 percent of each other, no primary table reachable this session, and nothing in the chart depends on the difference, so 7.75 was left alone rather than moved on an unverifiable attribution.

## Elements Z = 43 to 54 (Tc through Xe)

75. 99mTc: magnetic moment not in any recommended compilation I could reach, so mu, gamma, xi, q are null; only I = 1/2, 6.0072 h and the 140.5 keV gamma are asserted.

76. mu(99Tc): the chart uses Stone's +5.6503 (via the EasySpin file, sourced to INDC(NDS)-0915, December 2025) and derives gamma = 9.5711 MHz/T from it. The value tabulated in the 99Tc NMR literature and in general isotope tables is +5.6847 nuclear magnetons, 0.61 percent higher, giving gamma = 9.6294 MHz/T. I could not reach Stone's 2025 table to see which measurement it adopted or whether a recent laser-spectroscopy redetermination lies behind the change (the 2020 LIST hyperfine study of 97,98,99Tc used the 99Tc moment as its own reference rather than remeasuring it). Until that is checked, treat gamma(99Tc) as uncertain at the half-percent level, not at the 0.13 percent level the isotope note implies.

77. 105Pd: EasySpin/Stone mu = -0.638 gives gamma = -1.94529; older NMR tables give -1.957 from mu = -0.642. 0.6% disagreement, noted in the file.

78. Quadrupole moments for 111In, 123I, 125I, 131I and 133Xe are set to null: values exist only on the pre-2018 scale (e.g. 111In 0.804 b, 125I -0.889 b, 131I -0.40 b, 133Xe 0.145 b) and no Pyykko-2018-scale revision is tabulated. 129I (-48.8 fm2) and 127I (-69.6 fm2) ARE on the 2018 scale and are given.

79. 124I: spin 2 is solid but the moment is not reliably tabulated, so mu/gamma are null (same convention slice_4 used for 68Ga and 90Y).

80. Chemical shift ranges are the softest numbers in the file. Confident: 125Te/123Te (-4100 to +1100, total span ~5000 ppm, search-confirmed), 99Ru/101Ru (>18000 ppm, search-confirmed), 129Xe/131Xe biomedical scale. Estimated: 99Tc (-2500 to +500), 103Rh (-1000 to +10000), 107/109Ag (-200 to +1200), 111/113Cd (-600 to +900), 113/115In (-200 to +500), Sn (-2200 to +2500). Left null where I had no defensible endpoints: 105Pd, 121Sb, 123Sb, 127I.

81. chiMolar values are the CRC element-table numbers. Iodine is quoted as -88.7, which is per mole of I2 in CRC, matching the convention slice_4 used for Br (-56.4); if the table wants per-mole-of-atoms, iodine should be -44.4 and bromine -28.2.

82. 115In shiftRange endpoints and the 107/109Ag range are the weakest single entries; both would benefit from a check against a Bruker/IUPAC shift compilation, which the egress proxy blocked (publications.iupac.org, Wikipedia, PMC, chemlin all refused).

## Elements Z = 55 to 70 (Cs through Yb)

83. Xi for 138La (13.194300) and 173Yb (4.821) are recalled IUPAC values I could not fetch directly (all IUPAC/PDF domains are blocked by the egress proxy). Both are corroborated indirectly: each reproduces its sister isotope's Stone moment ratio to 1e-5 and 2e-4 respectively, and the sister values 139La=14.125641, 171Yb=17.499306, 133Cs=13.116142, 135Ba=9.934457, 137Ba=11.112928 were all confirmed by search snippets.

84. 169Tm: no IUPAC frequency ratio was found for thulium in any reachable source, and the 3.531 MHz/T that circulates in the Bruker and chemlin tables could not be traced to a primary measurement. It corresponds to mu = 0.23161, an older tabulation, and it is almost certainly a moment-derived number rather than an observed frequency, since Tm3+ is 4f12 paramagnetic and no diamagnetic thulium reference sample exists. Keeping xi null is right for the same reason it is right for iridium.

85. chiMolar values are the CRC element table (10^-6 cm3/mol, ~293 K) from recall. Only Gd (+185000, explicitly at 350 K) and Er (+48000) were web-confirmed; every other value was cross-checked against Curie-Weiss chi = 0.12505*mu_eff^2/(T-theta) using literature mu_eff and paramagnetic Curie temperatures and agreed to 10-25%.

86. Pm chiMolar set null: no measurement exists for the metal. Search confirmed Pm2O3 at +2660e-6 cm3/mol vs a 5I4 free-ion prediction of +2960e-6; a widely circulated value of +980e-6 for Pm is inconsistent with both and was rejected.

87. 153Sm quadrupole moment set null (not in the EasySpin file and I could not verify it). Its mu = -0.00216 nuclear magnetons and I = 3/2 are web-confirmed, giving gamma = -0.011 MHz/T.

88. 133Cs shiftRange '-10 to +230' is a constructed window: solid CsCl at +223 ppm is confirmed, aqueous Cs+ salts sit within ~10 ppm of the CsNO3 reference. Cs metal's Knight shift lies far outside and is flagged in the note.

89. 137Ba/135Ba shiftRange left null: literature states the barium shift range is narrower than its own typical linewidth, so no useful window exists.

90. Quadrupole moments are Stone IAEA INDC(NDS)-650 (2013) via the EasySpin file, converted barn -> fm^2. Pyykko 2018 revises some lanthanide values (141Pr in particular); I did not substitute, and 141Pr's note names the source.

91. Receptivities for zero-abundance radionuclides (137Cs, 133Ba, 147Pm, 153Sm) are exactly 0.0 by the prescribed formula, not a missing value.

92. Group set null and block set 'f' for La through Yb, following the instruction that lanthanides take null group; note that La's configuration is [Xe]5d1 6s2, so a d-block assignment is defensible under the alternative convention.

## Elements Z = 71 to 83 (Lu through Bi)

93. Ir: IUPAC Xi 1.718/1.871 % imply gamma about 4 % below Stone's moments and are mutually inconsistent by 0.4 %; no Ir resonance has ever been observed in solution, so these frequency ratios are not experimentally anchored. Flagged in mrNotes.

94. The two lutetium frequency ratios cannot both be right, and no published moment pair reproduces their ratio. Xi(175)/Xi(176) = 11.404/8.131 = 1.4025335, while (mu175/3.5)/(mu176/7) is 1.4086709 for the chart's moments (+2.2257, +3.160) and 1.409088 for the older ABMR pair (+2.2327, +3.1692(45), Brenner, Buettgenbach, Rupprecht and Traeber, Nucl. Phys. A440 (1985) 407). Reproducing the Xi ratio would need mu(176Lu) near +3.174 to +3.184, three to four standard deviations above the measured +3.1692(45). Xi(175Lu) = 11.404 is internally sound (IUPAC's own lutetium entry carries gamma = 3.0552e7 rad s-1 T-1, i.e. 4.86258 MHz/T, which is exactly the +2.2327 moment, and that moment reproduces Xi with a +1447 ppm shielding, the right sign for a closed-shell 4f14 Lu3+ ion); Xi(176Lu) = 8.131 is the one that cannot be anchored. I read the IUPAC entry only through a search snippet: publications.iupac.org, old.iupac.org and every mirror I tried are blocked here, so the digits behind 11.404 and 8.131 were never seen directly.

95. 187Os and 189Os cannot be reconciled from the numbers now available, and I could not settle which of the four is wrong. Both isotopes are referenced to the same molecule (OsO4), so the ratio Xi(189)/Xi(187) must equal (mu189/I189)/(mu187/I187) exactly, shielding cancelling. Tabulated Xi give 3.4023987; the chart's moments give 3.4303599, 0.82 percent apart. Three published 187Os moments are in play: +0.0639 (chart, Stone via EasySpin), +0.06432(3) (direct NMR of the 187Os line in molten OsO4, 1967, with a free-atom diamagnetic correction of about +0.9 percent already applied), and +0.06465 (older NMR tables). Two 189Os moments are in play: +0.6576 (chart) and +0.659933(4). Only the pair (+0.064647, +0.659933) reproduces the Xi ratio, and that 187Os value is not one I could trace to a publication. The absolute shieldings implied for OsO4 by the chart's pairs are +2478 ppm from 187Os and +10717 ppm from 189Os; the second is the more plausible magnitude for a 5d nucleus, which argues the 187Os moment is the outlier, but I have no calculated sigma(OsO4) to decide with.

96. Q(209Bi) is settled and no longer needs a reservation, but the scale it now sits on does: every other quadrupole moment in the chart comes from Stone INDC(NDS)-650 (2013) through the EasySpin isotope file, and adopting -42.2 fm2 for 209Bi puts one entry on a 2023 molecular/atomic scale while its neighbours stay on the 2013 one. I could not reach Stone's 2021 quadrupole table (INDC-NDS-0833) or the IAEA site at all (www-nds.iaea.org, nds.iaea.org and nndc.bnl.gov are all refused by the egress proxy), so I cannot say whether that table already carries the revision or still prints -516 mb.

97. Quadrupole moments left null for 186Re, 188Re, 198Au and 203Hg: not verified from a primary table, so omitted rather than guessed.

98. Chemical shift ranges for 183W (-3500 to +8000), 195Pt (-6000 to +8000), 199Hg (-3000 to +1000) and 187Os (-5200 to -2000) are literature spans assembled from secondary sources, not authoritative tabulations; the 207Pb (-5000 to +11000) and 205Tl (-1750 to +5100) ranges are better anchored to specific published endpoints.

99. Tl reference compound: IUPAC prints Tl(NO3)3 in H2O, which is unstable in water; the scale is in practice anchored on aqueous TlNO3 at infinite dilution. Used the latter, with the discrepancy noted.

100. No IUPAC reference compound assigned (set null) for 175/176Lu, 177/179Hf, 191/193Ir and 197Au; a compound is tabulated for some of these but could not be confirmed.

101. chiMolar values are CRC molar susceptibilities recalled from memory; WebFetch was blocked by the egress proxy for every host tried, so they could not be re-verified against a primary table.

102. mu(186Re)=+1.728 and mu(188Re)=+1.777 come from a 1965 atomic-beam triple-resonance measurement (diamagnetically corrected); no modern re-determination was located.

103. Blurb/mrNote claims not independently verifiable from a primary source: LSO/LYSO intrinsic activity of roughly 300 Bq per cm3, cisplatin at about -2100 ppm, and the 500 uM 195Pt detection limit (the last from a single cited study).

## Elements Z = 84 to 118 (Po through Og)

104. **Resolved.** Th chiMolar is now +97e-6, the CRC value, which agrees to 1.5 percent with the high-purity chi_g = +0.412e-6 cm3/g measurement this note already cited. The +132e-6 was the element-table figure, and the justification recorded here for keeping it, 'consistent with the rest of this data set', is exactly the reasoning the version 0.3 pass overturned: the rest of the data set was not CRC either. The Th mrNote has been rewritten.

105. U chiMolar +399e-6 and Pu chiMolar +534e-6 are element-table values; primary measurements scatter (alpha-U at 1.74e-6 cm3/g gives +414e-6; alpha-Pu is quoted near +5.3e-4 cm3/mol).

106. **Narrowed.** Am, Ra, Ac, Po, At, Rn, Fr and everything from Z = 96 up still have chiMolar null, and the reason is now specific rather than general: they are not listed in the CRC element table. Pa and Np were null for the same stated reason and should not have been. Both are listed there, unqualified, at +277e-6 and +575e-6, and both are now filled, which also removes the oddity of carrying U and Pu while leaving their neighbours blank.

107. 209Po mu = +0.68 muN is tabulated to two significant figures, so its derived gamma carries about 1 percent uncertainty; half-life given as 124 y (older tables say 102 y).

108. 231Pa mu = 1.99 is the only entry in the EasySpin/Stone file printed without a sign, so the sign of the moment is not experimentally established; written as positive and flagged in notes.

109. 211At (I = 9/2) and 225Ac (I = 3/2) have established ground-state spins but no adopted magnetic moment, so mu, gamma and q are null and recH/recC are 0.0 (abundance 0), following the convention used for radionuclides in slices 5 and 6.

110. 223Ra mu = +0.2705 muN from ISOLDE collinear resonance ionization spectroscopy; its spectroscopic quadrupole moment was not confirmed, so q is null.

111. 229Th uses mu = +0.360(7) muN and Q = +311 fm2 from the 2013 reanalysis, superseding EasySpin's +0.46 muN and +4.3 barn; noted in the isotope record.

112. 227Th (spin not confirmed here) and 241Am are mentioned only in mrNotes, not listed as isotope records, to avoid quoting unverified spins or moments.

113. Magnetism for Z >= 100 is an inference from electron configuration or the period-6 homolog, never a measurement; each such element says so in its mrNotes. Category is 'unknown' for Z >= 109 and 'transition-metal' for Rf-Hs, whose group chemistry has been demonstrated.

114. Superheavy half-lives are described qualitatively (hours, seconds, tens of milliseconds) rather than with exact numbers, except 294Og at about 0.7 ms and 285Cn at about half a minute.

115. Po assigned category post-transition-metal and At metalloid; both labels vary between reference data sets.

116. No IUPAC Xi values, shift references or shift ranges exist for any nucleus in this slice, so all three fields are null throughout, and invivo is false for every isotope (none is detected by MR in vivo).

## The molar susceptibility column

117. **Resolved, and this is the entry the version 0.3 pass answers.** The CRC table named here was supplied directly rather than fetched, and the whole column was rebuilt on it. Forty of the 118 cells moved. The finding behind them is a provenance error rather than a set of slips: the column was never the CRC element table it claimed to be, but the element-data compilation behind Wikipedia's and periodictable.com's susceptibility fields, which agrees with CRC for about half the table and departs from it by up to a factor of six. Two identifications pin that down: Wikipedia's hafnium infobox carries +75.0e-6 cm3/mol against CRC's +71, and a search returns palladium at +567.4e-6 against CRC's +540, both of them the chart's old values to the digit. **Every per-slice chiMolar provenance note below, which says in one form or another that the values are recalled element-table figures that could not be checked, is superseded by this one.** What is not resolved: the egress obstacle described in the original note was never overcome, so anything else in the chart needing a primary compilation is still where it was.

118. **Retracted.** That spot-check reported 'I found no value in the column that is wrong', and about forty were. The mechanism is worth keeping, because it is the failure mode of the whole method: the check passed because the chart and the search snippets it was checked against came from the same non-CRC lineage, so they agreed with each other. Two of its ten confirmations were wrong in the same way. Th +132.0 was reported as 'confirmed against search text quoting the CRC element table'; CRC gives +97, and the note's own slice-level entry already recorded a high-purity measurement near +96. Pt +201.9 was reported as matching; CRC gives +193. Confirming a value against text that claims to quote a table is not the same as reading the table.

119. **Resolved.** The per-mole-of-X2 convention now appears in the page's own provenance section rather than only in these notes: the text states that the values are per mole of the element as the source names it, that hydrogen, nitrogen, oxygen and the halogens are therefore per mole of the diatomic molecule, and that sulfur, whose standard state is S8, is per mole of atoms. The elements where the source offers more than one form now say which one their cell carries, in the cell's own note.

120. **Partly resolved.** Oxygen's cell now carries its own temperature: its note states that +3449e-6 is a value near 293 K, that oxygen is the one Curie paramagnet in the column, that it falls to about +3260e-6 at body temperature, and that the liquid and solid figures are a different regime rather than the same number colder. Chromium's cell now names the 311 K Neel point and the 160 to 180e-6 spread across published room-temperature values. What still stands: terbium's +170000e-6 is quoted bare, and its susceptibility moves about 1.4 percent per kelvin, so the source's own 285 to 300 K latitude is worth about eleven percent on that cell.

## Dossiers: proton, phosphorus, carbon and sodium

121. dossier_h1.json, metabolite T2 at 3 T: the entry contradicts itself. It gives NAA CH3 as 301 +/- 18 ms in occipital white matter and then says the full metabolite set 'spans 116 to 295 ms' across gray and white matter, so a value it quotes lies above the upper bound of the span it quotes. The two clauses are plainly from two different papers (a regional NAA/tCr T2 study and a whole-profile T2 survey) stitched into one value string, with no field, region, sequence or refocusing-train attribution for the 116 to 295 ms figure. Neither source was reachable this session; publisher domains are blocked and the numbers are not in any abstract I could retrieve.

122. dossier_h1.json, water-reference practice bullet: 'Allow at least 9 s without RF excitation beforehand so T1 saturation loss stays under 10 percent even in cystic or CSF-rich voxels.' Against this dossier's own CSF water T1 of 4425 ms the residual saturation after 9 s is 1 - exp(-9/4.425) = 13.1 percent, and at 3 T (CSF T1 about 4.3 s) 12.3 percent. A 9 s delay holds the loss under 10 percent only for T1 at or below about 3.9 s, i.e. for tissue water, not for the CSF-rich case the sentence singles out. I read the Oz 2020 experts' consensus in full (PMID 31922301, doi 10.1002/nbm.4236) and it contains no 9 s figure, so the provenance of the number was not identified and I cannot tell whether the delay or the 10 percent claim is the one carried over wrongly.

123. dossier_h1.json, chemical shift displacement: 'Consensus caps CSD at 4 percent per ppm for SVS and 2 percent per ppm for MRSI.' The 4 percent per ppm SVS figure is confirmed verbatim in the Oz 2020 consensus, read in full. The 2 percent per ppm MRSI figure I could not source: it is in neither that paper nor any MRSI consensus statement reachable by search this session, and it may be a transposition of the roughly 2 percent per ppm CSD of the sLASER adiabatic refocusing pair, which is a sequence property and not a recommendation. The same consensus also records that at 7 T the 4 percent per ppm limit cannot presently be met in the excitation direction with commercial volume coils, where implementations sit at 7 percent per ppm; the entry states the cap without that caveat.

124. dossier_h1.json, MEGA-PRESS macromolecule fraction: the fact says macromolecules contribute 'more than 50 percent' of the 3 ppm edited signal at 3 T and below, and the matching pitfall says 'over half'. The standard wording in the editing literature is 'as much as 50 percent' or 'approximately half', i.e. a ceiling rather than a floor, and the fraction is not a constant: it depends on the editing pulse bandwidth, on TE, and on voxel tissue composition. Search text confirmed the co-edited MM arises from lysine groups coupled at about 1.7 ppm, which the entry gets right; I found no study putting the fraction above 50 percent, so the direction of the inequality is asserted rather than sourced.

125. dossier_p31.json, 'T1 values fall by up to 62 percent from 3T to 7T': traced to Bogner 2009 (PMID 19526487, doi 10.1002/mrm.22057), whose abstract I retrieved. The 62 percent is the PME T1 fall, 8.1 s to 3.1 s, and PME is the one resonance the entry's own table omits; of the peaks it does list the largest fall is beta-NTP at 54 percent. The same source also records that Pi was the single metabolite whose T1 change from 3 T to 7 T was not statistically significant (6.9 to 6.3 s, a 9 percent fall), which undercuts the blanket 'as chemical shift anisotropy takes over' framing. Every other number in the entry matches Bogner exactly, including the T2 set.

126. dossier_p31.json, CK and ATPase flux: the rate constants k(PCr to gamma-ATP) 0.38 +/- 0.02 per s and k(Pi to gamma-ATP) 0.19 +/- 0.02 per s are confirmed by search text as Ren 2017 band inversion transfer at 7 T, a single laboratory and a single cohort. The fluxes 'near 1.3 mM/s' and 'near 0.2 mM/s' are my-side-checkable arithmetic (0.38 x 3.5 and 0.19 x 1.0) but they multiply those rate constants by PCr and Pi concentrations taken from a different entry, which are themselves conditional on the assumed 3 mM ATP reference flagged in the brain-concentration note. Three assumptions therefore ride on each flux, and none of the three came from the same measurement.

127. dossier_p31.json, cardiac PCr/ATP below 1.6 'associated with poor survival in dilated cardiomyopathy': this is the Neubauer 1997 dilated-cardiomyopathy cohort, a single 1.5 T study in one disease, and the threshold was derived without the blood and saturation corrections the same entry says are mandatory at 7 T. Quoted here next to 7 T healthy-range values of 1.85 to 2.14, it reads as a field-independent and disease-independent cut-off, which it is not. The primary paper was not retrievable this session.

128. dossier_p31.json, free cytosolic Mg2+ 'about 0.18 mM in human occipital lobe': a single-laboratory number from a single calibration curve, and it sits at the low end of reported human brain values (roughly 0.18 to 0.3 mM is in print). The entry's own note correctly says the spread is calibration rather than physiology, which makes quoting one unqualified point value for brain harder to defend than the explicit 0.3 to 1.0 mM range it gives for muscle. Not re-verified.

129. dossier_c13.json, why hyperpolarized 13C has stayed at 3 T: the note attributes it to carbonyl T1 shortening and to the signal coming 'from the polarizer rather than from B0'. I read the Punwani 2025 multi-centre consensus in full (PMID 40342865) and the community's own stated reason is different: all published human studies have used 3 T because major vendors only offer non-proton capability at 3 T and above, and there is limited evidence at other fields. The physics claim is also only approximate: with magnetization fixed by the polarizer, detected SNR is field-independent in the sample-noise-dominated regime but still rises with B0 for small or lightly loaded coils. The sentence is the compiler's reasoning, not a sourced explanation.

130. dossier_c13.json, 'Use small, constant-through-time flip angles for hyperpolarized dynamics': this is a defensible engineering position but it is not what the 2025 consensus says. That document reached consensus only that multi-centre studies should harmonize the flip angle scheme, and separately recorded that the delivered flip angle influences both AUC ratios and kinetic rate constants; it took no position against variable flip angle schemes. The bullet is stated in the same register as the two bullets around it, which are consensus-backed (gadolinium after the 13C acquisition, 24/25; centre frequency from the 1H water frequency, 23/24 - both confirmed verbatim). Similarly, the advice to use dual-tuned 1H/13C coils runs past a consensus item that explicitly permits repositioning onto a better 1H coil.

131. dossier_c13.json, the mandatory agent parameters: the five named items are confirmed in Punwani 2025, but as mandatory dose *release criteria* (pH and residual EPA 26/26, pyruvate concentration 25/26, temperature 23/26, dose volume 21/26), not as a reporting list. The paper does say a minimum reporting set was agreed, but it lives in a table the PubMed full-text route does not render, so the identity of the two lists is an inference. The claim that polarization failed to reach consensus 'only because current measurement methods are not trusted' also drops the second reason the paper gives, that polarization is difficult to measure quickly inside the workflow. Polarization did not reach consensus at 14/25, 56 percent.

132. dossier_c13.json, kPL and kPB in normal human brain: 0.012 +/- 0.006 per s and 0.002 +/- 0.002 per s are confirmed by search text as the first healthy-volunteer cerebral study at 3 T, a single small cohort from one site. kPB's standard deviation equals its mean, so the entry's derived statement that 'kPB is about a sixth of kPL' is a ratio of two numbers one of which is not distinguishable from zero at one sigma. Both are also model-dependent: they assume a two-site exchange fit, and the consensus above records that the applied flip angle changes kinetic rate constants, so they are not portable between acquisition schemes.

133. dossier_na23.json, 'at 78.8 MHz the deposited power is comparable to 1H at about 1.9T': this compares Larmor frequencies only. SAR scales as sigma x omega^2 x B1^2, and for the same flip angle and pulse duration a 23Na pulse needs B1 larger by gamma_H/gamma_Na = 3.78, so omega^2 B1^2 is matched to 1H at the same B0, not to 1H at 1.9 T; the residual difference is tissue conductivity at 79 versus 298 MHz. The practical conclusion the sentence supports - that at TR 100 to 150 ms SAR is not the binding constraint and gradient duty cycle is - is separately well attested, but the frequency-equivalence justification offered for it is not, and I found no source stating it.

134. dossier_na23.json, triple quantum filtering specificity: '32 +/- 6 percent of the TQF signal still came from extracellular sodium' is confirmed by search text, but the preparation is a Langendorff-perfused *mouse* heart with Tm(DOTP) shift reagent, ex vivo. Neither the species nor the preparation is stated. The pitfall then generalises it to 'a third of the TQF signal can be extracellular' as a property of in vivo human TQF, where no shift reagent can be used, the tissue is brain or muscle rather than myocardium, and the extracellular fraction has not been measured the same way.

135. dossier_na23.json, the compartment sum: '0.2 x 145 mM plus 0.8 x 12 mM is about 39 mM' uses an extracellular volume fraction of 0.2, a round textbook figure rather than a measurement (brain extracellular space is reported over roughly 0.15 to 0.25), and 12 mM, the midpoint of the stated 10 to 15 mM intracellular range. At the ends of those ranges the same arithmetic gives 32 mM and 45 mM. The near-exact agreement with the meta-analytic 40.5 mM in the next entry is therefore a coincidence of round inputs, not an independent confirmation of either. The 40.5 mM figure itself I did verify: Sci Rep 2023, PMID 36828873, pooled 40.51 mM, 95 percent CI 37.59 to 43.44, 141 measurements from 28 studies and 400 individuals, with the authors' own emphasis on high between-study heterogeneity.

136. dossier_na23.json, claims tied to current practice that will age: the PETALUTE cartilage timing (2:06 against 3:36 for density-adapted radial) is one very recent result on one sequence at one site and is the kind of number that is superseded rather than corrected; the multiple sclerosis cohort values are quoted to 0.1 mM with no field strength, calibration method or relaxation-correction status, which the same dossier's own practice bullet identifies as the dominant source of between-study scatter; and the summary's '3 to 4 mm and 10 to 15 minutes' is 7 T practice stated without the field, where the corresponding fact entry does name 7 T and the dossier elsewhere notes that matching 7 T at 3 T costs five to seven times the averaging.

## Dossiers: noble gases, deuterium and oxygen, fluorine, the difficult nuclei

137. dossier_gas.json: the summary's '0.2 to 0.9' polarization range mixes methods and the 0.9 end is attached to the wrong experiment. Search confirmed the MEOP figures as 'above 80 percent at 2, 3 and 4 T, with 89 percent reached at 3 T', but that is a low-pressure sealed cell in a polarized-target/neutron-optics context, measured in the pumping cell before the roughly three orders of magnitude of compression that the same dossier says is where polarization is lost. No SEOP system and no gas ever delivered to a patient reaches it. The SEOP entry then gives two different numbers for the same case in one sentence, 'roughly 25 percent for a 1 L clinical accumulation' and 'modern systems routinely at or above 30 percent for 1 L', and names neither the polarizer generation nor the point of measurement (in-cell, in-bag, at the scanner, at inhalation); 30 percent is also the value the Dose Equivalent arithmetic in 'Dose and breath-hold' silently assumes.

138. dossier_gas.json: the ventilation defect percent block splices two cohorts with different definitions. '1.6 +/- 1.2 percent with an upper limit of normal near 2.0 percent' is a single healthy cohort (n = 27, search-confirmed), while the age strata 1.3 / 2.5 / 3.8 percent come from a multi-centre age-dependent ULN database that explicitly spans sites, platforms, acquisition protocols and VDP quantification methods. As printed, every subject over 40 has an upper limit of normal above the flat 2.0 percent ULN quoted one clause earlier. VDP is strongly segmentation-dependent (linear binning versus thresholding versus adaptive clustering) and no method is named for either number.

139. dossier_gas.json: 'In IPF the ratio falls 3.3-fold to 0.16 +/- 0.03' belongs to the 1.5 T single-centre cohort whose healthy mean is the 0.55 +/- 0.13 in the same sentence, not to the 3 T multisite value quoted immediately before it: 0.55 x 0.48/1.58 = 0.167, so the stated 52 percent RBC drop and 58 percent barrier rise reproduce it exactly, whereas against the 3 T pooled healthy mean of 0.49 the same IPF value reads 3.1-fold. A separate three-peak IPF series (search-confirmed) gives healthy 0.58 +/- 0.12 and IPF 0.18 +/- 0.07. The fold-change is presented as a property of the disease when it is a property of which healthy reference you pair it with.

140. dossier_gas.json: several numbers in this dossier are arithmetic or practice hints rather than measurements and inherit their assumptions. The 'about 20 s' 129Xe gas T1 in the lung is the 0.482 s^-1 amagat^-1 oxygen relaxivity evaluated at an assumed alveolar pO2 of 100 mmHg (17.9 s at 37 C), a round physiological number, and the 3He figure is the same calculation from T1 x pO2 = 2.61 bar.s; real in-lung T1 moves with inflation, inspired oxygen fraction and breath-hold phase. The 208 ppm dissolved-phase centre with its 7.35 kHz and 3.67 kHz offsets, the 'roughly 70 times' gas-to-dissolved ratio (1/0.0145 from the multisite amplitudes) and the '390 mL of 86 percent enriched gas at 30 percent polarization' for 100 mL DE are all my own arithmetic: correct, but only as good as their inputs. The 'roughly 0.7 ms' sinc is a rule of thumb whose gas suppression depends on time-bandwidth product and field, not a specified pulse. The 2.55 +/- 0.22 h bulk-gas T1 is a sealed-cell value dominated by that vessel's wall relaxation. And the 3He and 129Xe ADCs (0.190 and 0.038 cm2/s) are quoted with no b value, diffusion time or lung inflation, none of which is a property of the gas.

141. dossier_gas.json: 'Xenon MAC is about 71 percent' is the 1969 determination. Search confirmed that the modern sevoflurane-interaction study gives 63.1 percent and that MAC in the elderly is sex-dependent at 69.3 percent (men) and 51.1 percent (women), so the literature spans roughly 51 to 71 percent and the dossier picks the top end without saying so. I found no source at all for 'noticeable sedation starts above roughly 30 percent inhaled fraction', which reads as a round clinical number. The 75 to 100 mL Dose Equivalent target and the claim that 129Xe 'is now the approved agent' are the current US label, which search shows has already been revised once (the paediatric indication now reaches age 6 while the 3 T chest coil is cleared for 12 and above), and no other jurisdiction is addressed. This is the entry most likely to age.

142. dossier_gas.json: the five human brain 129Xe resonances are given here as 187, 192.5, 195.4, 199 and 216.2 ppm, while the chart's own xenon shift table draws the same five species as 188, 192, 196, 200 and 217 ppm and its notes describe those as single-centre 1.5 T assignments rounded to whole ppm. The two sets differ by up to 1 ppm and neither file says which fit each came from, so the chart carries two versions of one measurement. In the same entry, '500 mL inhaled with a 20 s breath-hold and a dynamic spectrum every 2 s' is one group's protocol described as typical. Separately, the cardiogenic oscillation note's claim that sampling slower than about 50 ms 'aliases into the mean' is not a Nyquist statement (a 1 to 1.7 Hz cardiac fundamental would only alias beyond roughly 300 ms) but a practical requirement for resolving the waveform shape; 20 ms is simply the TR of the single study the 10.0 +/- 2.5 percent amplitude comes from.

143. dossier_h2o17.json: the background concentrations are a chain of assumptions rather than measurements, and the summary and the facts disagree. The summary says 17O gives 'a 20 mM natural abundance H2-17O background' while the fact entry gives 20.56 umol per gram of water and 'roughly 15 to 17 mM in tissue after the water fraction': the summary is quoting the per-gram-of-water figure as if it were the tissue background, a 25 percent difference that propagates into any percent-signal-change statement built on it. On the 2H side, the '1.1e-6 of 1H' natural-abundance receptivity is 9.65e-3 x 1.15e-4, which silently selects the 0.0115 percent end of the 0.0115 to 0.0156 percent abundance interval (the other end gives 1.5e-6), and the 8.9 to 13.7 mM 'theory' span is my own arithmetic over that interval crossed with a 70 to 80 percent brain water fraction. The measured 8.96 +/- 0.7 mM is a single 7T study. The '15 percent across the USA' and '5 percent within a single county' deuterium variation figures are isotope-hydrology map claims, US-specific and uncited here.

144. dossier_h2o17.json: the slice-selection note is my own derivation, not a quoted result, and its two halves do not share a premise. The first sentence is correct that at a bandwidth fixed in ppm the gradient for a slice of given thickness is gamma-independent. The second sentence switches to a pulse of fixed duration, under which the gradient needed for the same slice is 6.5 times LARGER at 2H, not 'unchanged'; the chemical-shift-displacement conclusion survives either reading but the sentence as written invites the wrong one. Separately, every Larmor frequency in this dossier is computed at nominal integer field. The human 3 T 17O study actually ran its coil at 16.7 MHz rather than the 17.3 MHz quoted here, because a clinical '3 T' magnet is 2.89 T, and the same 4 percent applies to the 19.6 MHz 2H figure and to the 'watch the RF environment at 19.6 and 45.7 MHz' bullet.

145. dossier_h2o17.json: the practice bullet tells you to 'check that the HDO linewidth is in the 11 to 26 Hz range before you start the clock', but that is a 7T figure used as a field-independent acceptance gate. The one accessible measurement of human brain 2H linewidth across field (PMC7141779, doi 10.1002/nbm.4235) gives water 11.8 +/- 1.1 Hz at 4 T and 11.7 +/- 1.7 Hz at 7 T, glucose 9.4 and 10.3 Hz, Glx 10.0 and 11.1 Hz: near-constant in hertz and sitting at the very bottom of the quoted range. The 20 to 26 Hz numbers in that same paper are post-mortem rat head across the whole head. I could not identify a human source for the 26 Hz upper bound, so as a shim gate the stated range may be far too loose.

146. dossier_h2o17.json: everything human in the 17O section comes from one centre and five subjects, and the dossier strips the qualifiers. Full text (PMC11910252, doi 10.1016/j.zemedi.2023.07.004) confirms the regional CMRO2 values, the 7.3 percent / 5 percent GM/WM signal rise, the 32 to 48 percent WM overestimate and the voxel-count rule, but also shows that the cohort was five healthy men aged 31 to 52; that the regional CMRO2 values carry standard deviations of 0.14 to 0.41 umol/g/min which are not printed here; that the 25/50-voxel ROI rule is a Rician-noise simulation at that study's own baseline SNR of 19 (GM) and 15 (WM) at 8 mm and does not transfer to another SNR; and that the '8 to 39 percent low' deep-grey figure merges two structure-specific ranges (caudate 13 to 39 percent, insula 8 to 31 percent) while the thalamus actually agreed with PET. The 7.3 / 5 percent rise is specific to that protocol's gas volume, enrichment and inhalation length. The note that 'any signal rise is metabolically generated H2-17O' is the standard justification but omits recirculating H2-17O generated elsewhere in the body, which that study's own kinetic model carries as a separate gain term. The RF-environment bullet (one group killing the technical-room lighting, one site's eightfold gradient-filter capacitors, 'up to 25-fold' SNR loss) is two site anecdotes stated as general practice.

147. dossier_f19.json: the summary calls 19F 'the second most favourable nucleus in the table' in the same sentence that defines favourability per nucleus. The chart itself lists 3H at 45.414839 MHz/T, above 1H's 42.577, so on a per-nucleus basis 19F is third. The claim is defensible only on natural-abundance receptivity, where 3H's zero abundance removes it, a different quantity from the one the sentence names.

148. dossier_f19.json: the detection limits, 11.8 +/- 3.0 mM for PFCE at 24 C rising to 379.9 +/- 51.8 mM for PFOB at 37 C, are one 3 T phantom study (Colotti 2017, PMID 27385530, abstract confirmed). Its two endpoints differ in both agent and temperature, so the '30-fold spread across agents and temperature' cannot be apportioned to either variable; and a detection limit is not a number without a voxel volume, a scan time and a coil, none of which the dossier states. The 'bounded at about 13 percent' efficiency loss in the practice bullet is the same paper's own simulation across its own set of cell types, fixation states and temperatures, not a general bound on optimising at the wrong condition. The summary compounds this by putting 'high micromolar to millimolar' on one scale, since the micromolar end is unlocalised whole-head spectroscopy over tens of minutes and the millimolar end is a per-voxel imaging limit.

149. dossier_f19.json: the clinical 19F figures each rest on one small study and the dossier does not say so. 'About 10^5 cells per voxel at 3T in under 10 min' carries no voxel volume and comes from the first-in-human dendritic-cell vaccine trial, as does the roughly 50 percent fall in apparent cell number at 24 h. The isoflurane head washout, 9.5 and 130 min, is Lockwood 1997 (PMID 9422894, doi 10.1093/bja/79.5.581; abstract confirmed, including the dossier's honest note that the signal was too weak to localise the compartments), a single surface-coil human study whose field strength and subject number are omitted here. The fluvoxamine and fluoxetine brain concentrations are one group's 1990s unlocalised work, with 'brain-to-plasma ratios near 10' as a round ratio. And the practice bullet's claim that labelling efficiency 'varies by cell type across more than an order of magnitude' is not supported by the dossier's own numbers, which span 4.7-fold (7.9e11 per macrophage to 3.7e12 per dendritic cell); the cell type that would justify the claim is never named.

150. dossier_quad.json: the 'receptivity times tissue concentration' table is this dossier's central argument, and while every product is arithmetically correct, the concentrations are round physiological stand-ins chosen by the compiler rather than measurements tied to a stated tissue: 23Na 40 mM, 31P 30 mM, 35Cl 12 mM, 39K 83 mM, 7Li 0.5 mM, 25Mg 10 mM total, 43Ca 2 mM total, and tissue water at 'about 80 M' of protons (pure water is 111 M; 80 M implies a 72 percent water fraction). Several are tissue-specific and are then used as if they were general: the 83 mM potassium and the 12 mM chloride are both calf muscle at 7 T, the 0.5 mM lithium is a whole-brain average that the same dossier elsewhere says varies twofold by region. '14N at a 10 mM site' is not a measured concentration at all, it is a reference value invented to make that row comparable, and the 1/2e6 feasibility threshold in the note is drawn from these same numbers.

151. dossier_quad.json: 'which implies a 7Li T1 near 2.1 s in vivo' is my inference from the reported 4.6 degree Ernst angle at TR 6.7 ms. Full text (PMC5955212, doi 10.1038/s41380-018-0016-6) confirms both parameters but never states the T1 the authors assumed, and the chart's own single-line-nuclei shift table quotes 'about 4.6 s' for 7Li in the human head; the chart therefore carries two 7Li T1 values a factor of two apart, neither measured in the imaged cohort. The same paper also shows that the fact entry's 'reconstructed 15 x 15 x 25 mm' and the practice bullet's '25 mm isotropic in 8 minutes' are two ends of one scan: acquisition was 24.0 x 25.3 x 25.0 mm and the 15 mm figure is an interpolated reconstruction grid, so quoting it as resolution overstates what was resolved. The 'coefficient of variation near 28 percent within brain' in the pitfalls is that study's 27.9 +/- 3.6 percent within-slice CoV of b-SSFP signal intensity in eight euthymic bipolar subjects, i.e. of signal, not of concentration.

152. dossier_quad.json: 'the 39K voxel is 22 times the volume of the 23Na voxel, which is the 194-fold receptivity deficit converted into geometry' does not close arithmetically. 39K is 194-fold less receptive but sits at roughly twice the concentration, a net in-vivo deficit near 93-fold, so a 22-fold voxel absorbs less than a quarter of it and the 39K images from that session are also simply noisier than the 23Na images, not iso-SNR at matched time. In the same spirit, the 'relative to 13C' receptivities in the first fact are recomputed from the 1H column using 13C = 1.70e-4 rather than quoted: standard tables print 7Li = 1540 and 35Cl = 20.2 against the 1593 and 21.0 here, so 'which match the standard tables' is true to a few percent, not exactly.

153. dossier_quad.json: three superlatives and one scaling claim need qualifying. '133Cs ... Q -0.343 fm^2, the smallest of any alkali' and '7Li has the second smallest quadrupole moment of the alkali metals' both count one isotope per element and quietly exclude 6Li, whose Q of about -0.081 fm^2 is four times smaller than 133Cs's. 'About seven times gentler in quadrupolar relaxation rate' than 23Na is Q^2 arithmetic ((10.4/4.01)^2 = 6.7) that assumes Li+ and Na+ see identical electric field gradients and correlation times in tissue, which is the very thing the compartmentation entry treats as unknown. The hyperpolarized 15N lifetimes, 41 s for nicotinamide, 196 s for the GGT substrate and up to 477 s for the pyridine N-oxides, are all quoted at 1 T while the imaging described in the same sentence was done at 3 T; CSA relaxation scales as B0^2, so those lifetimes do not carry to the field they were used at. Finally, 'intracellular Rb+/K+ partitioning is about 20:1' reads as an intracellular Rb-to-K ratio, which nothing I could reach supports; it is presumably an intracellular-to-extracellular Rb+ ratio, and the perfused-organ papers that would settle it were behind blocked publisher domains.

## Dossier: the scanner as a materials problem

154. Quench pipe bore diameter omitted: the vendor datasheet (euro-emc.co.uk) was blocked by the egress proxy and I would not quote a number I could not check.

155. MRI's share of world helium consumption omitted: commonly repeated as 20 percent but I found no primary source I would defend; helium spot price also omitted for the same reason.

156. The 1.4 nano-ohm persistent-circuit resistance is my own derivation from R = L(dI/dt)/I for 0.1 ppm/h on a 50 H, 500 A magnet, not a quoted vendor figure. It is arithmetically sound but the drift spec and inductance vary by magnet.

157. 5 gauss line distances: sources disagreed (one gave 2 to 3 m for actively shielded 1.5 T, another 4 to 6 m axial for unshielded). I widened the stated ranges to cover both rather than pick one.

158. NdFeB (BH)max upper bound of 470 kJ/m3 is near the laboratory/theoretical ceiling; routine commercial N52 is closer to 400 to 420 kJ/m3. The range as written spans both.

159. Mu-metal composition (77 Ni, 16 Fe, 5 Cu, 2 Mo) is nominal and varies by grade and vendor; saturation 0.74 to 0.8 T is well attested.

160. NbTi Tc is quoted as 9.2 K in most references but one source gave 9.8 K for a particular Nb-Ti ratio. I used 9.2 K as specified in the brief and as the majority value.

161. Gradient amplifier peak currents above 900 A are premium/research systems; the 1500 A figure came from a single high-power driver spec and is flagged as such in the text.

162. Aluminium RF screen thickness (0.5 to 1.5 mm) and copper foil (0.1 to 0.2 mm) are typical installed practice rather than a standardised spec; the 90 to 100+ dB attenuation requirement is well sourced.

## Dossier: contrast, susceptibility and implants

163. Cobalt-chromium (CoCrMo) volume susceptibility: no defensible ppm value found, so it is stated only comparatively (between titanium and 316L) rather than numerically.

164. Cortical bone susceptibility relative to water: literature spans about -0.3 ppm (1990s NMR methods) to -2 ppm and lower (modern UTE-QSM). Quoted as 1.4 to 2 ppm diamagnetic relative to water, which brackets the modern values only.

165. Rohrer 2005 per-agent 1.5 T values: secondary sources disagree on whether gadoterate or gadoteridol is higher (one Bayer brochure gives gadoterate 4.1 / gadoteridol 3.6, the opposite of the usual table). I therefore used the independently confirmed Szomolanyi 2019 human-plasma set and quoted only gadobenate 6.3 and gadodiamide 4.3 from the linear group.

166. Thermodynamic stability constants log K 25.6 / 22.1 / 16.9 (gadoterate / gadopentetate / gadodiamide) are standard-table values I could not re-confirm in an accessible source this session.

167. Gadoteridol acid dissociation half-life stated loosely as 'a few hours'; the Port 2008 figure is around 3.9 h at pH 1.2 but I could not verify it directly.

168. Eu3+ room-temperature effective moment of about 3.4 Bohr magnetons: the Van Vleck origin and 7F0 J=0 ground state are confirmed, the exact 3.4 figure is not.

169. Nitinol +230 to +250 ppm SI is derived by me from a measured mass susceptibility (2.87e-6 cgs cm3/g, electropolished austenite) times density 6.45 g/cm3 times 4 pi; surface oxide and martensite/R-phase shift it by more than 6 percent.

170. 316L range +3520 to +6700 ppm SI is attributed to Schenck 1996 via secondary sources only; the primary table was not reachable (publisher domains blocked).

171. Ferumoxytol particle size: hydrodynamic 17 to 31 nm is well supported, but reported core sizes vary widely by method (TEM about 7 nm, cryo-TEM about 2 nm), so core size is deliberately left vague.

172. The practice bullet saying r1 at 3 T is 'a few percent to 15 percent below' the 1.5 T value is a generalisation; gadobutrol actually rises slightly from 4.78 to 4.97 in human plasma.

## Shift tables: proton, phosphorus and carbon

173. shifts_h1.json, reference field ('Tetramethylsilane at 0 ppm'): TMS is not the convention in general use for an aqueous in vivo scale, and it is not what the source of these positions used. The compilation these numbers match (Govindaraju, Young and Maudsley 2000) reports its shifts against the DSS trimethyl singlet at 0.0000 ppm, and DSS or TSP, not TMS, is what IUPAC recommends for aqueous samples; TMS is insoluble in water and is never present in an in vivo experiment. The numerical consequence is a few thousandths of a ppm, which is why no stem should move, but the sibling shifts_h2.json writes 'DSS or TMS at 0 ppm' for this same scale while this file writes only TMS. The DSS referencing was confirmed only from search-result text summarising the paper; the PDF host (pfeifer.phas.ubc.ca) and Wiley are both blocked by the egress proxy, so the primary table was not re-read. The same field also puts tissue water 'near 4.7 ppm' while shifts_h2.json fixes HDO at exactly 4.8 ppm, which is the same resonance on the same scale: water moves about -0.01 ppm per degree C, so 4.7 is the body-temperature value and 4.8 the value most DMI papers write, and neither table says which it means.

174. shifts_h1.json, the ppm column as a whole: these are high-resolution solution shifts of the kind used to build a fitting basis set, measured in vitro at one temperature and pH, not positions measured in vivo. The internal evidence that they come from the standard compilation is strong (creatine 3.913 against phosphocreatine 3.930, GPC 3.212 against PC 3.208 against choline 3.185, mI H1/H3 3.52 and H4/H6 3.61, Ala J about 7.2 Hz, Lac J about 6.9 Hz are all the familiar values), but I could not re-read that table this session, so the two-decimal figures are from a secondary or remembered rendering of it. In tissue these positions move with temperature and intracellular pH and are never determined to the 0.01 ppm printed here. One entry is a further step removed: glucose is drawn at 3.43 and described as 'centre of the ring-proton cluster', but the H2 to H6 centroid of the two anomers lies nearer 3.6 ppm; 3.43 is the conventional glucose fitting position (the alpha H4 and beta H3/H5 region), not a centre of mass.

175. shifts_h1.json, what is actually resolved at clinical field: the table note calls the heights 'peak heights in a short-TE 3T spectrum', yet several stems are not peaks in such a spectrum. GABA 3.01 and GSH 2.95 are recovered only by editing, as their own peak notes say, so their rel values of 0.1 are heights in an edited difference spectrum, a different experiment from the one the note describes. Four stem pairs are drawn closer than an in vivo linewidth: Lac 1.31 against Lip13 1.30 (0.01 ppm is 1.3 Hz at 3T, against a metabolite linewidth of 5 to 8 Hz), Glc 3.43 against Tau 3.42, tCr CH3 3.03 against GABA 3.01 (2.6 Hz), and NAA 2.01, NAAG 2.04 and MM20 2.05 within 5 Hz end to end. Each position is defensible; the visible separation between them is a property of the drawing, and only the NAA/NAAG pair carries a note saying the two are reported together.

176. shifts_h1.json, the rel column taken as a set: it is not one spectrum. Lip13 at 0.25 and Lip09 at 0.30 of NAA are a fat-contaminated voxel or frank necrosis (the peak notes say so, but the heights are drawn unconditionally), Lac at 0.06 is above what normal brain shows at short TE, and Ala at 0.04 is a meningioma and abscess marker that is effectively absent from healthy tissue. The table note nevertheless describes a single short-TE 3T spectrum and makes no statement of mixed provenance, unlike the 13C table ('two different experiments on one axis') and the 129Xe table ('these lines come from two different experiments'). The macromolecule heights carry a quieter assumption of the same kind: MM20 at 0.25 and MM17 at 0.12, with the age remarks in their notes ('larger with age', 'increases from about age 30 to 40'), are single-cohort findings written as general properties, and I found no source for either age claim this session.

177. shifts_p31.json, PC at 6.20 and GPC at 2.76: the PC note says the value given is the centre of a reported in vivo range of 5.9 to 6.3 ppm, but 6.2 is not that centre. The 3T phantom study this table otherwise tracks almost line for line (Molecules 2021, PMC8703310, doi 10.3390/molecules26247571, read in full through the PubMed MCP) measures PC at 6.2 to 6.4 ppm in vitro at 37 C and pH 7.0 to 7.5 and explicitly contrasts that with an in vivo PC of 5.9 ppm, so for this peak an in vivo table is carrying the phantom number. 7T brain work does report PC near 6.2, which is why I would not simply move the stem, but the choice is a choice and the note misdescribes it. GPC is the mirror case: 2.76 is that paper's in vivo value and the bottom of the range the note quotes (2.76 to 2.95), while dossier_p31.json in the same publication lists GPC at +2.9. PE at 6.77 is corroborated (the same paper gives 6.78 in vivo); GPE at 3.5 is a literature value it adopted for its own basis set rather than measured.

178. shifts_p31.json, the table as a whole mixes three tissues without saying so. PCr at rel 1.0 with PE and PC is brain or muscle; the two 2,3-DPG lines exist only in erythrocytes; UDPG is essentially a liver signal (about 2 mM there against 0.2 to 0.3 mM in brain, per the same 3T paper). No voxel produces this spectrum, and unlike the 13C and 129Xe tables the table note does not say so - it explains only the pH and Mg2+ dependence of two gaps. The individual peak notes do say 'blood only' and 'prominent in liver', so the facts are present peak by peak; what is missing is the statement at table level. There is a referencing consequence too: the declared zero is PCr, which liver and kidney do not contain, so for the liver lines the scale is in practice anchored on alpha-ATP or Pi and then quoted as though PCr-referenced.

179. shifts_p31.json, Pi at 4.85 labelled 'At pH 7.0': the titration data quoted in the same note do not put Pi at 4.85 at pH 7.0. With the endpoints stated there (acid limit 3.29, base limit 5.68) the Henderson-Hasselbalch form returns 4.79 ppm at pH 7.0 for pKa 6.77 and 4.82 for the pKa 6.75 the note names, so 4.85 corresponds to pH 7.04 or 7.03 respectively. The three worked points the note does quote (4.13 at pH 6.5, 5.02 at 7.2, 5.23 at 7.4) are reproduced by pKa 6.77 with those endpoints to better than 0.01 ppm and by 6.75 only to 0.02, so the sentence pairs one calibration's endpoints and worked values with another calibration's pKa. Offered below as a correction to the pH label rather than to the shift. Note also that dossier_p31.json states the same titration as pKa 6.75 with endpoints 3.27 and 5.63 and puts Pi at +4.9, and itself warns that the variants are worth up to 0.05 pH units of absolute offset; the two files should end on one constant set. I could not reach Petroff 1985 itself - search-result text quotes the formula with a third endpoint pair, 3.23 and 5.70, at the same pKa 6.77.

180. shifts_p31.json, NADH -8.15, NAD+ -8.31, UDPG a -8.2 and UDPG b -9.8: all four come from the single 3T phantom paper above, which reports NADH at 8.16 and NAD+ at 8.32 ppm from PCr and the two UDPG doublets at -8.2 (ribose-side phosphorus) and -9.8 (glucose-side), including which doublet is which. This table rounds the first two 0.01 ppm the other way, which changes nothing visible but shows the values were transcribed rather than independently sourced. None of the three is resolved in vivo - they lie on the upfield foot of alpha-ATP and are recovered by fitting - so these are prior-knowledge positions in a basis set, not observed in vivo positions, and the drawn heights (0.04, 0.08, 0.04, 0.05) are what a fit apportions rather than what anyone sees. The NADH note's threshold '1H decoupling and 4T or above' is stricter than that paper, which puts in vivo NAD+/NADH determination at 3 to 7 T and cites positions measured from 4 to 11.7 T. The redox ratio 'about 5.7 in brain' and total NAD 0.3 to 0.5 mM both match that paper, which is itself citing others for the 5.7.

181. shifts_p31.json, 2,3-DPG note, 'the pair spans roughly 5.4 to 6.9 ppm over the physiological pH and oxygenation range': the two drawn positions are confirmed - the same 3T paper gives exactly 5.5 and 6.3 ppm as the in vivo blood singlets and explains their offset from the phantom values (4.1 and 5.3 at pH 7.0, 4.6 and 5.8 at pH 7.5) by binding to haemoglobin - but the 5.4 to 6.9 window itself is unattributed and I found no source for it in any reachable paper. The pH and oxygenation sensitivity is real and is why a window belongs in the note; its endpoints are the part I cannot defend. Labelling the two lines 'downfield' and 'upfield' rather than 2-phosphate and 3-phosphate is a deliberate and, I think, correct evasion of an assignment the in vivo literature does not settle.

182. shifts_c13.json, reference field and Lip CH2 at 30.2: the whole natural-abundance half of the axis hangs on the lipid methylene, which the Lip CH2 note itself says is 'quoted between 29.9 and 30.5 ppm'. Anchoring at 30.2 therefore imports an offset of up to plus or minus 0.3 ppm into every natural-abundance position drawn here, which is 10 Hz at 3T and comparable to separations the table asks the reader to take seriously (Glu C3 to Gln C3 is 0.8 ppm, Glu C2 to Gln C2 0.6 ppm). The choice is at least consistent inside the publication - dossier_c13.json also uses 30.2 ppm - but I found no standard fixing the in vivo anchor at 30.2 rather than 30.0, groups do both, and work referenced instead to NAA C6 at 22.9 or to the glycerol carbons will not agree with either. The hyperpolarized half is anchored separately on pyruvate C1 at 171.0, so the two halves of this axis are not tied to one another by anything measured.

183. shifts_c13.json, Glu C5 note, the carbonyl-region list: it is quoted to 0.1 ppm and the in vivo literature disagrees at exactly that level. Against the 11.7T values in J Neurosci Methods 2011 (PMC3090476, doi 10.1016/j.jneumeth.2011.02.021, read in full) - Glu C5 182.0, Gln C5 178.5, Asp C4 178.3 to 178.4, Glu C1 175.3, Asp C1 175.1, Gln C1 174.8, bicarbonate 161.0 - this table gives Gln C5 178.6, Glu C1 175.4, Asp C1 175.0 and Gln C1 174.9, each 0.1 ppm away, and the companion paper (J Magn Reson 2012, PMC3471529, doi 10.1016/j.jmr.2011.11.012) puts Glu C4 at 34.1 where this table draws 34.2 and Asp C3 at 37.1 where it draws 37.4. The weakest single item is 'NAA C1 179.8'. NAA's two aspartyl carboxyls both lie in this region and the numbering is not standardised: Bluml 2000 (PMID 10729255, doi 10.1006/jmre.1999.2001) lists them as 'N-acetyl-aspartate C(1-4, C=O)' while this table's NAA C6 note numbers the acetyl carbonyl C5 and the methyl C6; and by analogy with free aspartate in the same list (C1 175.0, C4 178.4) the downfield 179.8 line ought to be the beta carboxyl, which most schemes would call C4. No 13C table that settles it was reachable, so I flag the label rather than the position.

184. shifts_c13.json, the hyperpolarized heights and the Pyr C1 decay figure: the five hyperpolarized positions (171.0, 183.2, 179.4, 176.5, 161.0) are the standard set and I have no quarrel with them, but the heights are one tissue at one moment. Lac C1 at 0.45 of pyruvate is a high lactate-to-pyruvate ratio for normal brain - search-result text for the quantitative human brain study gives bicarbonate-to-pyruvate ratios of 0.07 to 0.37 in normal tissue, and lactate ratios depend on whether peak height, a single frame or area under the curve is compared - and in prostate or liver the same drawing would be wrong in the other direction. The note's 'in vivo T1 near 20 to 45 s at 3T' is an apparent decay constant that folds in RF depletion, perfusion and exchange rather than a relaxation time: the solution T1 of [1-13C]pyruvate at 3T is usually quoted near 60 to 70 s, and search-result text this session gave about 70 s. The 0.43 mL/kg bolus of 250 mM is the standard human dose and needs no reservation.

185. shifts_c13.json, HCO3- note, 'kPB runs about a sixth of kPL': verified for healthy human brain and only there. Search-result text from the quantitative human brain study gives kPL 0.012 +/- 0.006 per s and kPB 0.002 +/- 0.002 per s, a ratio near one sixth but with 100 percent uncertainty on the numerator of it; the sentence as written reads as a property of the method. In tumour, where lactate rises and pyruvate dehydrogenase flux falls, and in prostate, where bicarbonate is routinely undetectable, the ratio is much smaller; in heart it is larger. The same note's claim that bicarbonate is 'the only in vivo readout of pyruvate dehydrogenase flux' is true of the [1-13C]pyruvate experiment as drawn but not of 13C MRS generally - Glu C4 labelling from [1-13C]glucose, three stems to the right on this same axis, also reports flux through PDH.

186. shifts_c13.json, organ provenance inside the natural-abundance half: the table note declares the hyperpolarized/thermal split, which is the harder one, but not the organ split underneath it. NAA C6 at 0.25 is brain; glycogen C1 at 0.40 is liver or muscle (its own note says brain glycogen at 3 to 5 mM is not detectable at natural abundance); Lip CH2 at 0.85 is the subcutaneous fat a surface coil sees rather than a metabolite; and Glu C4, Gln C4, Glu C3, Gln C3, Asp C3 and the two glucose anomers reach these heights only during an infusion of labelled substrate. So the axis carries at least four provenances and the note names two. This is the same objection as for shifts_p31.json and the remedy is a sentence, not a number; the 129Xe table shows the wording that works.

## Shift table: deuterium

187. Glx is quoted as 2.35 ppm by de Feyter/de Graaf and as 2.4 ppm by several other groups; glutamate and glutamine are never resolved in vivo so the exact centre depends on the fitting basis set. Used 2.35.

188. Lactate is quoted as 1.3 ppm in the original DMI papers and as 1.4 ppm in some reviews, partly because the overlapping natural-abundance lipid CH2 pulls the fitted centre. Used 1.3.

189. Glucose appears as 3.8 ppm in most DMI work but as 3.9 ppm in at least one recent review. Used 3.8.

190. The water/HDO reference is set to 4.8 ppm in nearly all DMI papers, but a few write 4.7 ppm (same convention as some 1H MRS groups). Used 4.8.

191. rel heights are drawing hints only: they sketch a post-dose steady-state brain spectrum (HDO dominant, Glc about a fifth of it, Glx smaller, Lac small). Actual ratios depend on tissue, time after dose, and tumour vs normal brain.

## Shift table: hyperpolarized xenon

192. The brain RBC peak (217.0) and the lung RBC peak (218.2) are the same chemical species drawn as two stems 1.2 ppm apart, because the pulmonary value comes from a 3T multisite standardized cohort and the cerebral value from single-center 1.5T head CSI. If the chart looks cluttered there, the brain stem can be dropped and folded into the lung RBC note.

193. The alveolar gas sub-peak at -2.3 ppm is real but rests on one careful time-domain-fitting study (Duke, 1.5T) that assigned the airway gas component to 0 ppm. Most pipelines report a single gas peak at 0 ppm, and if the reference is taken as the whole gas peak rather than the airway component this offset shrinks.

194. Brain compartment values (188, 192, 196, 200) are single-center 1.5T numbers rounded to whole ppm; the literature spread is about 1 ppm (gray matter 196 to 196.5, white matter 192 to 193, plasma 197 to 200). They are well established as assignments but not standardized the way the lung values now are.

195. The membrane peak is genuinely two components about 5 ppm apart under three-peak time-domain fitting, but their absolute positions are not consistently reported, so only the single 197.6 ppm membrane peak is drawn and the splitting is described in its note.

196. The oxygenation dependence of the RBC shift is nonlinear and the in-vitro calibration is referenced to plasma at 20 C, giving an RBC-to-plasma separation of 20.4 ppm (deoxygenated) to 25.5 ppm (fully oxygenated). That is larger than the 20.6 ppm membrane-to-RBC gap of the standardized in-vivo values, so the stated 'about 1 ppm per 10 percent sO2' is a working slope near normal saturation, not a calibration curve.

197. Relative heights are drawing hints only. The true gas peak is roughly 100x any dissolved peak; the RBC:membrane ratio of about 0.49 is the one height relationship in the file that is quantitatively faithful.

## Shift table: fluorine

198. Isoflurane CF3 at -81.5 ppm and OCHF2 at -88.0 ppm: sources disagree. An in vivo mouse table referenced to CFCl3 gives -80.55 and -86.85 (JMRI, PMC5484368); an 11.7T phantom/in vivo study gives -82.9 and -89.9 with PFCE at -91.5 (PMC8760223); high-resolution solution chemistry would put the CF3 several ppm further downfield near -75. The 6 to 7 ppm CF3-to-OCHF2 separation is consistent across the in vivo work, so the pair was centred inside the reported in vivo range as the brief asks.

199. Desflurane at -85.0 ppm: no verified 19F assignment was found for desflurane in any accessible source. The value is inferred from the CF3-CHF-O-CHF2 structure by analogy (beta-fluorine shielding of the isoflurane CF3, plus the OCHF2 of the same ether), and the CF3 and OCHF2 lines are deliberately drawn as one stem. The lone CHF is stated in the note as beyond -140 ppm without a numeric stem because that value could not be confirmed.

200. Fluoxetine at -61.5 ppm: the in vivo MRS papers (Komoroski 1994 and later) were not retrievable in full text, so the position comes from the para-alkoxy aryl-CF3 family (benzotrifluoride -63.7, p-methoxy analogue near -61.7) rather than from a quoted in vivo number. Confident to about 1 ppm, not better.

201. 5-fluorouracil at -169.3 ppm is confirmed to within about 0.15 ppm and the reservation can be retired, but two conditions attach to it. First, the value depends on how the secondary reference is placed: the confirming measurement sets internal trifluoroacetic acid at -75.43 ppm, while the chart's own reference note places TFA at -76.55 ppm, the neat/CDCl3 figure. Aqueous TFA is reported between about -75.4 and -76.5 depending on concentration and pH, so anything in this table anchored on TFA carries an unresolved 1 ppm offset. Second, the ring N-H of 5-FU titrates with pKa about 8.0 to 8.1, so the in vivo position moves with tissue pH between roughly -169 and -166 as the anion grows in; the quoted number is a pH 7.4 value, not a constant.

202. PFOB internal CF2 components at -117.5, -122.0 and -126.6 ppm: taken from a published summary table whose other entries (perfluorodecalin, F-44E) are clearly mis-referenced, though the PFOB and PFCE rows agree with independent sources. The number of resolvable CF2 components and their relative heights are a drawing approximation.

203. Solid-state 19F of bone fluorapatite is mentioned in the fluoride note without a ppm value on purpose; its position (commonly quoted near -103 ppm) was not verified and it is invisible to liquid-state sequences anyway.

204. Network egress was blocked for every WebFetch target in this session, so verification relied on WebSearch result text plus PubMed Central full text via MCP. Confirmed from source: PFCE -91.8 / -91.5, PFOB CF2Br -63.7 and CF3 -81.82, aqueous NaF -121.5 and KF -125.3, TFA -76.55 (PMC5484368, doi 10.1002/jmri.25564); PFCE -91.5 with isoflurane offsets (PMC8760223, doi 10.1007/s11307-021-01653-6); sevoflurane -75.13 and -155.65 (PMC4518101, PMID 26330861); 5-FU metabolite offsets (PMC2394413, doi 10.1038/sj.bjc.6601345); hexafluorobenzene -164.9 from 19F reference-standard tables.

## Shift table: the single-line nuclei

205. All five peaks are placed at exactly 0.0 ppm, which is the honest answer: each nucleus is referenced to its own aqueous standard (NaCl, water, LiCl, NaCl chloride, KCl) and the in vivo line sits on that zero, so the stems will coincide at the origin. Nothing here is identified by position, so there is no sign to get backwards.

206. Shift-reagent direction: I state only the Dy(PPP)2 (7-) case (extracellular Na 10 to 20 ppm UPFIELD of intracellular Na at 0 ppm), which sources word explicitly as upfield. Sources for TmDOTP5- were ambiguous about direction (one says 'downshifted by 2.2 ppm' while also quoting the peak at about +2 ppm), so I deliberately omitted TmDOTP rather than risk a sign error.

207. 7Li brain concentration is given qualitatively ('a few tenths of a millimolar, well below serum') because reported brain/serum ratios vary by study; the 4.6 s human head T1 comes from a single older in vivo report.

208. 35Cl tissue level is quoted as physiology (about 100 to 110 mM plasma, a few mM intracellular) rather than as a measured 35Cl MRI tissue concentration, which I could not pin to a specific value.

209. 39K muscle concentration of roughly 100 mM is a round physiological figure, not a citation-exact 39K MRI value. The 200 Hz quadrupolar satellite figure (calf parallel to B0, 7T, Rosler 2016) is solid; the derived 'roughly 14 ppm at 7T' is my own arithmetic from gamma = 1.989 MHz/T.

210. The 'rel' heights (1.0, 0.6, 0.35, 0.25, 0.15) are a compressed readability ranking, not a ratio. True relative in vivo signal spans about two orders of magnitude (23Na 1.0, 17O ~0.16, 7Li ~0.05, 35Cl ~0.02, 39K ~0.01 on a concentration-times-receptivity estimate); I noted the compression inside the 23Na peak note.

---

*Corrections and additions welcome; this file should shrink as the chart is checked. If you
settle one of these, the entry should be deleted rather than annotated, and the value it
concerned should carry its new source in the data.*