Background Ultrahigh dose-rate radiation (UHDR) produces less hydrogenperoxide (H2 O2 ) in pure water, as suggested by some experimental studies, andis used as an argument for the validity of the theory that FLASH spares the normal tissue due to less reactive oxygen species (ROS) production. In contrast,most Monte Carlo simulation studies suggest the opposite. Purpose We aim to unveil the effect of UHDR on H 2 O2 production in purewater and its underlying mechanism, to serve as a benchmark for Monte Carlosimulation.We hypothesized that the reaction of solvated electrons (e−aq ) remov-ing hydroxyl radicals (•OH), the precursor of H 2 O2 , is the reason why UHDRleads to a lower G-value (molecules/100 eV) for H2 O2 (G[H2 O2 ]), because: 1,the third-order reaction between e−aq and •OH is more sensitive to increasedinstantaneous ROS concentration by UHDR than a two-order reaction of •OHself -reaction producing H 2 O2 ; 2, e−aq has two times higher diffusion coefficientand higher reaction rate constant than that of •OH, which means e−aq woulddominate the competition for •OH and benefit more from the inter-track effectof UHDR. Meanwhile, we also experimentally verify the theory of long-lived rad-icals causing lower G(H2 O2 ) in conventional irradiation, which is mentioned insome simulation studies. Methods and materials H2 O2 was measured by Amplex UltraRed assay.430.1 MeV/u carbon ions (50 and 0.1 Gy/s), 9 MeV electrons (600 and0.62 Gy/s), and 200 kV x-ray tube (10 and 0.1 Gy/s) were employed. For threekinds of water (real hypoxic: 1% O 2 ; hypoxic: 1% O2 and 5% CO2 ; and nor-moxic: 21% O 2 ), unbubbled and bubbled samples with N2 O, the scavenger ofe−aq , were irradiated by carbon ions and electrons with conventional and UHDRat different absolute dose levels. Normoxic water dissolved with sodium nitrate(NaNO3 ), another scavenger of e−aq , and bubbled with N2 O was irradiated byx-ray to verify the results of low-LET electron beam. Results UHDR leads to a lower G(H 2 O2 ) than conventional irradiation. O2 andCO2 can both increase G(H2 O2 ). N2 O increases G(H2 O2 ) of both UHDR andconventional irradiation and eliminates the difference between them for carbonions. However, N2 O decreases G(H2 O2) in electron conventional irradiation butincreases G(H2 O2 ) in the case of UHDR, ending up with no dose-rate depen-dency of G(H2 O2 ). Three-spilled carbon UHDR does not have a lower G(H2 O2 )than one-spilled UHDR. However, the electron beam shows a lower G(H 2 O2 )for three-spilled UHDR than for one-spilled UHDR. Normoxic water with N2 O orNaNO3 can both eliminate the dose rate dependency of H 2 O2 production forx-ray. Conclusions UHDR has a lower G(H 2 O2 ) than the conventional irradiation forboth high LET carbon and low LET electron and x-ray beams. Both scavengersfor e−aq ,N2 O and NaNO3 ,eliminate the dose-rate dependency of G(H2 O2 ),whichsuggests e−aq is the reason for decreased G(H2 O2 ) for UHDR. Three-spilledUHDR versus one-spilled UHDR indicates that the assumption of residual radi-cals reducing G(H2 O2 ) of conventional irradiation may only be valid for low LETelectron beam.