The geometry of the test model airfoil in this study and that of the owl airfoil given by Liu et al. [8] are compared in Figure 1. The cross section of the owl

airfoil (denoted by the red dash line) in Figure 1 is measured at 40% of the semi-span of the actual owl wing (2z/b = 0.4). The maximum airfoil thickness is as thin as 5.5% (t/c = 0.055) at x/c = 0.11. The wing thickness near the feathery trailing edge (x/c > 0.9) is zero. Since it is exceedingly difficult to create an airfoil model of this thickness dimension to withstand the wind tunnel test, a test model made of balsa wood with a thickness of 1.25% near the trailing edge, a chord length of 80 mm, and a span length of 180 mm was created, as shown in Figure 2. There were also slight differences in shape on the lower surface of the model (x/c = 0.2 - 0.6). Regardless, the baseline of the owl wing was reproduced as closely as possible and any effects on measurement were negligible since deviation in design was less than 1% of the chord length. The geometry of the test model, which is denoted by the black dash line in Figure 1, is measured using a noncontact three-dimensional laser scanning system (KONICA MINOLTA RANGE7) with 4 μm measurement accuracy.

A series of experiments were performed using an open-circuit low-speed wind tunnel at Kyushu University. The rectangular cross section of the test section was 180 mm × 360 mm. The test section was covered on every side with an acrylic sidewall. Turbulence intensity was approximately 0.3% at 5.0 m/s. Lift and drag measurements were made by vertically mounting the airfoil model on the test section and attaching a three-component microforce balance system (LMC-3501-5N, NISSHOELECTRIC-WORKS) at the bottom of the test section. Each related load of the balance system was set at 5 N for the lift and drag forces, and 0.5 Nm for the momentum. The uncertain accuracy of the aerodynamic force measurements is estimated at approximately 0.2%, based on our micro-force calibration tests. Lift and drag forces were calibrated beforehand using standard weights. The test model was installed in the test section with a gap of 0.5 mm between the sidewall and the model tip in accordance with Rae and Pope [16] , who indicated that the gap should be lower than 0.5% of the span, and Burns and Mueller [17] , who showed that gap sizes between 0.1 and 1.4 mm were acceptable and should not affect the results.

The same setup for the test model was used to perform the smoke visualization

experiments using the smoke wire technique. Two nichrome wires with 0.1 mm diameter were horizontally set up at the inlet of the test section and coated with liquid paraffin as the working substance. The wires were located 75 mm forward of the leading edge of the airfoil and perpendicular to the leading edge. An alternating electric voltage from a Volt slider (S-130-10, YAMABISHI ELECTRICS) was applied across the wires to generate smoke. A green laser with a 2 W output (SDL-532-2000T, Scitec Instruments Ltd.) was used as a light source. The smoke-line was recorded using a high-speed video camera (Miro C110, Phantom) at a frame rate of 400 frames per second with an exposure time of 2400 μs, and image resolution of 1280 × 1024 pixels.

Aerodynamic force measurements were conducted in the Reynolds number range 23,000 - 60,000. The Reynolds number was based on the chord length of the test model. Since the cruise Reynolds number of the Mars airplane currently under consideration by the Japan Aerospace Exploration Agency is Re = 23,000, we recognized that value as a standard Reynolds number. The corresponding free-stream velocities for each Re are shown in Table 1. The angle of attack (α) was changed from α = −10 deg to 20 deg in 1 deg increments. The flow visualization was performed on each of the upper and lower surfaces of the wing at Re = 23,000. For Re = 30,000, only the flow field on the suction side was visualized.

Figure 3 illustrates the aerodynamic force characteristics at Re = 23,000. Error bars indicate standard deviations of three repeat measurements. As shown in Figure 3, the standard deviations are negligibly small, indicating highly reproducible measurement results. For purpose of comparison, CFD results of the two-dimensional laminar analyzes [12] were also plotted.

With respect to lift measurements, the experimental results in Figure 3(a) show the test model to be in reasonable agreement with CFD data, although a slight difference was observed at α = 5 deg. The lift coefficients increased almost

linearly at low angles of attack from α = −5 deg to 4 deg. A conventional symmetric airfoil, such as the NACA0012, at low angles of attack (around Re = 20,000) has a lift curve slope that is much lower than 2 πα due to trailing-edge separation on its suction side and attached flow on its pressure side [18] . In contrast, the lift curve in Figure 3(a) for the test model nearly equaled to 2 πα with a convex slope that rose steadily from α = 0 deg to 4 deg. Presumably, variations in the flow field around the airfoil affected the aerodynamics (to be described later). A nonlinear lift increase was observed in the range between α = 4 deg and α = 6 deg where the experimental and CFD results diverged. The lift curve slope began flattening out at α > 6 deg, and the lift coefficient reached almost the maximum value at α = 9 deg, corresponding to the stall angle of attack. At larger angles of attack, conventional stall phenomena accompanied by a rapid lift reduction found at high Reynolds numbers was not observed here. Lift coefficients were almost the same or showed a modest decrease.

Figure 3. Aerodynamic characteristics at Re = 23,000.

Figure 4 shows Reynolds number dependency on the aerodynamic performance from Re = 23,000 - 60,000. Figure 4(a) shows plots of the lift coefficient as a function of the angle of attack for five Reynolds values. Up to around α = 2 deg, the plots show a linear slope with little variation between the Reynolds numbers. Above that, the plots begin to diverge and assume a non-linear slope

to eventually flatten out or dip at α > 6 deg. At high angles of attack above α = 9 deg, the plots diverge again and the maximum lift coefficient increases with Reynolds number. The relation between Reynolds number and the maximum lift to drag ratio (L/D_max) is illustrated in Figure 5. The L/D_max gradually increased in proportion to the Reynolds number. In particular, an increment of the L/D_max from Re = 23,000 to 30,000 is the largest, and the L/D_max almost linearly increases above Re = 30,000. By contrast, the drag curves shown in Figure 4(b) have Reynolds number dependency, and the drag characteristics tended to decrease gradually with a decrease in the Reynolds number since the viscous drag relatively increased with a decreasing Reynolds number compared with the pressure drag.

Visualization of the flow fields on the suction side of the airfoil by a smoke wire method at Re = 23,000 from α = 1 deg to 13 deg is shown in Figure 6. Here a separation, transition, and reattachment are denoted by S, T, and R, respectively. At α = 1 deg [Figure 6(a)], the flow separated near the trailing edge, and the separation point gradually moved toward the leading edge as the angle of attack increased until α = 4 deg [Figure 6(b)]. The separated shear layer transitioned from laminar to turbulent and reattached to the surface at α = 5 deg [Figure 6(c)]. Then, a separation bubble formed between the separation and reattachment points. Since a large negative pressure region on the suction side was formed inside the separation bubble [19] , we attributed the nonlinear lift increase in Figure 3(a) to the formation of the separation bubble. The drag coefficients also increased at around α = 5 deg due to the increase in pressure drag associated with the formation of the separation bubble on the suction side. This separation bubble advanced to the leading edge as the angle of attack increased. However, after the separation point passed near the maximum airfoil thickness at α = 8 deg [Figure 6(e)], a short separation bubble formed near the leading

edge. Although the separation point remained fixed at the leading edge, the reattachment point gradually moved to the trailing edge, and the separation bubble stretched [Figure 6(f)]. The flow clearly separated at the leading edge without reattachment above α = 12 deg [Figure 6(g)], and large-scale vortices shed downstream. However, the lift coefficient at around α = 12 deg did not decrease much as shown in Figure 3(a) despite dramatic changes in the flow field such as the burst of the separation bubble. It is inferred that a high negative pressure region is secured on the suction side after the burst by exchanging momentum due to the mixing of the shedding vortices. The characteristics of the flow fields on the suction side at Re = 23,000 are summarized in Table 2.

Figure 7 shows the flow fields at Re = 30,000. Compared with Fig. 6, the flow fields on the suction side at α = 1, 6, 8 degrees are almost the same even if the Reynolds number increases. Since there is almost no Reynolds number dependency on the aerodynamic performance at the corresponding angles of attack as shown in Figure 4, we assume that the flow field is similar in larger Reynolds number region.

The flow fields on the pressure side of the airfoil at Re = 23,000 are shown in Figure 8. It is significant that the separation bubble formed on the pressure side at α = 0 deg (Figure 8(a)). The flow separated at around x/c = 0.1 and reattached to the surface near the center of the chord length with turbulent transition, thus thickening the separated shear layer. The location and length of the separation bubble changed little, but the state of the turbulent transition gently changes with the angle of attack [Figures 8(b)-(c)]. Furthermore, the boundary layer gradually thinned out. The flow field on the pressure side changed to almost attached flow at α = 5 deg [Figure 8(d)].

At α = 5 deg, the position of the separation bubble switched from the pressure side to the suction side. This tendency is very similar to the property observed in a circular arc airfoil [20] where the separation bubble is continuously present on the airfoil surface within a wide angle of attack range. The difference in lift at low angles of attack between the aforementioned NACA0012 airfoil and the owl airfoil corresponded to the presence and absence of the separation bubble on the lower surface. The positive pressure region on the pressure side inside the separation bubble worked to enhance the lift. In addition, the drag increase was not distinctly observed from α = 0 deg to 4 deg, as shown in Figure 3(b), despite the increase in pressure drag associated with the laminar separation on the pressure side near the leading edge. We deduced that the impact of the separation bubble thickness on the projected area viewed from the front was limited to low angles of attack because of the characteristically large undercamber shape line of the owl airfoil. Furthermore, the positive pressure region on the pressure side where the undercamber was strong worked in the thrust direction and had the effect of decreasing the drag. Therefore, the separation bubble formed on the pressure side during the low angles of attack contributed greatly to the lift increase such that drag was curbed.