Ultra-high field (UHF) (>7T) transmit (Tx) (1) and transceiver (2) human head surface loop phased arrays in combination with RF shimming or parallel transmission allow improving Tx-efficiency (B₁⁺/√P) and homogeneity in comparison to single-channel volume coils. Further enhancement in performance with regard to Tx- and receive (Rx) parallel imaging and SNR can be achieved by increasing the number of elements. However, construction of such densely populated arrays is very difficult due to the requirement of a very large number of decoupling circuits. An analytical model providing a quick evaluation of the coupling within a multi-channel Tx-array is thus the enabling step towards array optimization. Previously full-wave analytical models using dyadic Green’s functions have been utilized to investigate SNR at various frequencies (3,4) and the impedance, Z, between a pair of planar loops at 63MHz (5). In this work, we developed an analytical model allowing for impedance analysis of two loops arranged in a cylindrical geometry and cross-validated the approach against experiments and numerical simulations. It is furthermore demonstrated that the geometry and relative positioning of loops can be optimized such that simultaneous resistive and inductive decoupling can be achieved at UHFs (9.4T) without need for additional decoupling circuits.The analytical model was developed using the same concepts as in (5), i.e. the Z-matrix was calculated using the reaction theorem and the framework of dyadic Green’s functions in the spectral domain. The model included two rectangular wire loops placed on a cylindrical surface (Fig.1) thus mimicking a pair of loops within a human head Tx-array. Then, the magnetic, k_m, and electric, k_e, coupling between the loops were evaluated. To validate the modeling we experimentally measured the S-parameter matrix at 64,124,300, and 400 MHz using a network analyzer and then extracted the Z-matrix. Loops were constructed of a 1.5 mm copper wire and placed on a cylindrical FR4 holder (OD-215 mm). Each loop measured 100 mm in length (along the holder's axis) and 80 mm in width (42.5º coverage), which is similar to what has previously been used in UHF head transceiver arrays (2). A cylindrical phantom (ID - 170 mm), constructed to mimic tissue properties at 400MHz (1), was placed inside. To further cross validate analytical model and experimental measurements we simulated the same setup using the CST microwave studio 2015 frequency domain solver (Darmstadt, Germany). Based on optimization results obtained from the validated analytical modeling approach 9.4T (400 MHz) overlapped 2-loop array with perfectly decoupled (-40 dB) loops was constructed without need of additional decoupling circuits and compared to a gapped array (1,2) of similar coverage based on conventional transformer decoupling (TD). Data were acquired on a Siemens Magnetom 9.4T human imaging system. In transmission arrays were driven with 45º phase shift.Fig.2 displays calculated and measured k_m and k_e dependences on the angle, α, between the centers of the loops. Comparison of the analytical, simulated and experimentally measured data (Fig.2) shows that our analytical model provides a good match with measured and simulated k_m and k_e at lower frequencies (64,124 MHz, Fig.2A). At UHFs (300,400 MHz, Fig.2B) analytical results for k_e match the CST and experimental data only when k_m< 0.02 (α>70º). However, all three k_e plots matched each other well (Fig.2C), when the magnetic coupling was cancelled. This fact, however, does not impair the validity of our results since compensation of the mutual inductance is a critical requirement in most Tx-arrays. Furthermore, analytical analysis (Fig.2D) shows that while k_m does not significantly vary with frequency, k_e changes substantially. The higher the frequency the closer are the first zero-crossing points (α_m, α_e) that correspond to perfect inductive (k_m~0) and resistive (k_e~0) decoupling respectively. Thus, a loop overlap can be found that minimizes inductive and resistive coupling simultaneously at UHFs. In contrast at 64 and 124 MHz (for α=α_m) k_e measured 0.5÷0.6, which corresponds to S₁₂ of -12÷-10dB. Decoupling at 400 MHz was further optimized analytically by varying the loop width (Fig.2E). For 10.5 cm loop width both k_m and k_e are perfectly cancelled at the same time. Compared to the common gapped 2-loop array overlapping improves both B₁⁺ and SNR (Sum-of-Square) maps by eliminating the void between the loops and increasing the penetration depth (Fig.3).The developed analytical model was comprehensively validated and allows for the optimization of the geometry and relative positioning of transmit array elements. The latter enabled simultaneous cancellation of resistive and inductive coupling without additional decoupling circuits at UHFs. The resulting overlapped array element arrangement improves both Tx- and Rx-performance in comparison to conventional gapped arrays.