PRT for esophageal cancer has potential for improved conformality and reduced toxicity, but the location of the esophagus poses special challenges for scattered beam PRT. Passage through lung tissue and upper abdominal bowel gas can compromise the geometry of the spread out Bragg peak (SOBP). We sought to evaluate the effect of PRT beam arrangement on dosimetric parameters in lower esophageal tumors. Using a model for scattered beam PRT (Varian Eclipse), we compared the dose distribution for PA, AP, AP-PA, LPO, and LPO/RPO beam arrangements in 2 patients with lower esophageal tumors. Coverage was promoted by customizing the SOBP width from each beam to overlap with the thickest part of the target with 12 mm distal and 6 mm proximal margins. Symmetric 15 mm transverse margins were set between the CTV and block-edge. Custom compensators were calculated for each beam, ensuring shape conformity to the distal CTV surface. All plans were normalized to yield 100% dose at isocenter. For Patient #1, 4-field photons resulted in excellent CTV coverage (98% min, 110% max), but modest doses to lung (V5 = 34%), heart (V35 = 43%), and spinal cord (max = 45 Gy). PA PRT had worse coverage (93% - 108%), lower lung (V5 = 6%) and heart (V35 = 6%), but higher cord (51 Gy) doses. AP PRT had poor CTV coverage (89% - 110%), higher heart (V35 = 40%) and lower cord (18.5 Gy) doses. AP-PA PRT had acceptable CTV coverage (95% - 106%) with low lung (V5 = 6%), heart (V35 = 7%), and cord (29 Gy) doses. LPO PRT at 2 different gantry angles (150/165°) had worse coverage (93% - 113%/109%), higher left lung dose (V5 = 19% and 12%, respectively), but lower cord dose (47/49 Gy) compared to PA PRT. The LPO/RPO plan (140, 220°) had poor CTV coverage (88% - 106%) with higher lung dose (V5 = 22%). In Patient #2, photons had good CTV coverage (99% - 105%), but high lung (V5 = 48%) and heart (V35 = 70%) doses. PA PRT had worse coverage (92% - 110%), but very low lung (V5 = 13%) and heart (V35 = 11%) doses. AP-PA PRT improved homogeneity (92% - 106%) with a small rise in heart (V35 = 15%) and decrease in cord (23 Gy) doses. Variable weighting of AP-PA fields did not improve homogeneity. The LPO and LPO/RPO plans had very poor CTV coverage (<86% min) mainly due to inhomogeneity from lung tissue in the beam's entrance path. The heart and lung doses can be significantly reduced with PRT, but the presence of air degrades coverage. Air in the upper abdomen may limit anterior field reliability due to variability in bowel gas, especially in low thoracic tumors. PA only fields may be the ideal arrangement, but this must be tailored individually with attention to cord dose. Scanning PRT may help the problem of tissue inhomogeneity, but issues with organ motion, especially near the diaphragm, would still need to be addressed.