1: \begin{abstract}

2: Terahertz ultra-massive multiple-input multiple-output (THz UM-MIMO)

3: is envisioned as one of the key enablers of 6G wireless systems. Due

4: to the joint effect of its large array aperture and small wavelength,

5: the near-field region of THz UM-MIMO systems is greatly enlarged.

6: The high-dimensional channel of such systems thus consists of a stochastic

7: mixture of far and near fields, which renders channel estimation extremely

8: challenging. Previous works based on uni-field assumptions cannot

9: capture the hybrid far- and near-field features, and will suffer significant

10: performance loss. This motivates us to consider hybrid-field channel

11: estimation. We draw inspirations from fixed point theory to develop

12: an efficient deep learning based channel estimator with adaptive complexity

13: and linear convergence guarantee. Built upon classic orthogonal approximate

14: message passing, we transform each iteration into a contractive mapping,

15: comprising a closed-form linear estimator and a neural network based

16: non-linear estimator. A major algorithmic innovation involves applying

17: fixed point iteration to compute the channel estimate while modeling

18: neural networks with arbitrary depth and adapting to the hybrid-field

19: channel conditions. Simulation results will verify our theoretical

20: analysis and show significant performance gains over state-of-the-art

21: approaches in the estimation accuracy and convergence rate. 

22: \end{abstract}

23: