"TDD (Time Division) Massive MIMO represents the only effective implementation of Massive MIMO at the frequency bands under consideration."!? So says Professor Erik Larssen, a leading researcher at Linkoping University in Sweden. Larsson blogged the controversial paper on April 4, the day after it was published at https://arxiv.org/pdf/1704.00623.pdf. The blog and the abstract of the paper is below.
On the other hand, China Telecom, China Unicom, Huawei, and ZTE have announced successful trials of FDD (Frequency Division) Massive MIMO. A joint press release by China Telecom and ZTE asserts the opposite: "The FDD Massive MIMO solution is predictable to be deployed in China Telecom in 2017." Obviously, I'm not qualified to judge engineers of that quality when they disagree. I'll do my best to explain the issues and direct you back to the principals.
The primary issue, as I understand it, is whether FDD overhead is inevitably too high for FDD to be practical. With line of sight (LOS), the two techniques appear to have similar results. Without decent Line of Sight. the new paper reports a significant difference.