The aim of the present study was to investigate the multi-leaf collimator (MLC) gravity e ect on intensity modulated radiation therapy (IMRT) dose-dynamic delivery at di erent gantry angles. The non-uniform uence for IMRT was generated by means of dose-dynamic MLC (DMLC) performances, which were multiple static segments (MSS) or fully dynamic sliding window (SW) delivery methods at di erent gantry angles. SW delivery is quite similar to MSS with regard to MLC sliding pattern through the X1 to X2 jaw direction, except that it is irradiated during DMLC movement. To determine how the DMLC gravity a ects the IMRT dose uence, we mounted a two-dimensional ion chamber matrix (MatriXX, Scanditronix-Wellhofer, Schwarzenbruck, Germany) to a linear accelerator (CL 21EX, Varian, Palo Alto, CA, USA). Individual IMRT dose-dynamic segments were applied by using MSS and SW for 6-MV photon beams at ve di erent gantry angles: 0 (neutral gravity), 45 (semi-along gravity), 90 (vertical-along gravity), 315 (semi-against gravity) and 270 (against gravity). To test the correlation of leaf speed and gravity, we delivered half monitor units (MUs) for planned doses for absolute dose comparison. Strict -index (dose di erence: 2 %; distance to agreement: 2 mm) histograms were used for quantitative analyses of the discrepancy. The dose distribution by MSS under neutral gravity (gantry 0 ) with 137 MUs at a prescribed dose (45 cGy) point, which was in the low gradient area, was used as a reference for the other results. The -index histograms showed an increased tendency of the dose discrepancy toward the gravity-along direction rather than toward the gravity-against direction. The acceptable proportional ranges below 1 of the -index were 96.2 { 99.6 % (mean: 97.6 %), 92.6 { 93.5 % (mean: 93.0 %), 90.4 { 92.6 % (mean: 91.3 %), 94.6 { 98.0 % (mean: 96.4 %) and 92.9 { 96.8 % (mean: 95.0 %) for neutral, semi-along, vertical-along, semi-against and vertical-against gravity of the positions of MLC segments, respectively. we noted that MSS deliveries were more stable than were the SW deliveries (unacceptable -index range 1.0 { 2.0: mean 3.6 % for MSS and 5.4 % for SW). For the dose measurement on the prescribed dose point, all measurements showed a good agreement within an average of 2 %. Our experiment conclusively demonstrated that the DMLC gravity a ects the