### Abstract:

Two-phase gas-liquid flows in microscale fractal-like branching channel flow networks were experimentally studied to assess the validity of existing void fraction correlations and flow regimes based on superficial gas and liquid velocities. Void fractions were assessed using two different methods. First, void fraction data were acquired using a High-Speed-High-Resolution (HSHR) camera and computed by area-based two-dimensional image analysis. Void fraction data were also computed using a slip ratio, defined as gas velocity over liquid velocity. Liquid velocity represents the bulk-averaged liquid velocity as determined by microscale particle image velocimetry (micro-PIV). Gas velocity was determined by averaging gas-liquid interface velocities made at the channel centerline.
The fractal-like branching channel flow network has five bifurcation levels of different channel widths varying from 400 µm to 100 µm with a fixed channel depth of 250 µm. Each downstream width decreases by 30% whereas the downstream lengths increase by 40%. The total flow length through a single path is approximately 18 mm. Filtered air and deionized water were used as the gas and liquid working fluids, respectively. Mass flow rates of air and water into each k=0 branch were varied from 0.3 g/min to 2.5 g/min and from 5.2x10⁻⁵ g/min to 1.3x10⁻² g/min, respectively. These flow rates yielded superficial air and water velocities through the same branch level between 0.007 m/s and 1.8 m/s and between 0.05 m/s and 0.42 m/s, respectively.
For each branching level, due to an increase in flow area, the superficial liquid and gas flow rates change. A two-phase flow regime map was generated for each level of the fractal-like branching flow network and compared to maps developed using the Taitel and Dukler (1976) model and to maps presented in Chung and Kawaji (2004). Flow regime transitions are well predicted with the Taitel and Dukler (1976) model for each branching level.
Void fraction assessed using the slip ratio shows very good agreement with the homogeneous void fraction model for all branching levels. On the other hand, void fraction determined by area-based two-dimensional image analysis shows better agreement with the void fraction correlation of Zivi (1964).