betaFeedback

WorksPhysics and Astronomy

On the role of grazing atmospheric particles in hypersonic boundary-layer receptivity

Abstract

The role of small atmospheric particulates in the initiation of the laminar-turbulent transition is investigated for the HIFLIER vehicle during Mach 6 ascent condition. High-fidelity simulations are used to examine the disturbance wavepackets generated by particles following either grazing trajectories through the boundary layer or trajectories that lead to impact and rebound at the cone surface. Grazing particles that penetrate less than halfway through the boundary layer are found to generate wavepackets with amplitudes comparable to those produced by wall-impinging particles. To identify the receptivity mechanisms responsible for this behaviour, a multiple-scales asymptotic formulation is developed that accounts for the extended streamwise region over which a particle interacts with the boundary layer. The analysis shows that a grazing particle can generate a strong disturbance response when crossing the critical layer, where the streamwise-momentum component of the adjoint mode is largest and the boundary layer is most receptive. Thus, direct surface impact is not required for atmospheric particulates to excite dynamically significant instability waves. Root-mean-square disturbance estimates for submicron grazing particles predict transition-onset locations close to those measured in flight, suggesting they may have contributed significantly to the onset of transition on HIFLIER at the trajectory condition considered.

Introduction

Laminar-turbulent transition (LTT) for high-speed vehicles has several persistent challenges to its full understanding and, therefore, prediction [5]. It appears desirable to move towards the amplitude method [11], but doing so requires a receptivity solution for the initial disturbance amplitudes specific to a disturbance source. While acoustics provide the dominant disturbance source initiating transition in wind tunnels [14], the source(s) in free flight have only been postulated [4]. The most likely candidates for causing transition in hypersonic free flight include kinetic fluctuations and atmospheric aerosols, which have measurements [9].

Fedorov [8] provided the first theoretical analysis of boundary-layer receptivity to disturbances generated by particulates in supersonic flow. The model was formulated within a temporal stability framework and assumed that the particle–boundary-layer interaction occurred over a sufficiently short streamwise distance for the locally parallel approximation to remain valid. Subsequent studies have employed full Navier–Stokes simulations to investigate particulate receptivity over a range of geometries and flow conditions. Good agreement was achieved for all the available numerical solutions [3, 10], but the theoretical predictions of particle-excited first mode amplitudes differed from the numerical solutions by up to 40%. Efforts have begun very recently to capture the particle-induced transition scenario in a statistical model like Al Hashine et al. (2023). Chuvakhov & Fedorov (2025) used the receptivity solutions from Fedorov [8] to model LTT onset for particles distributed in radius. They found micron-sized particles contributed most to the predicted LTT location, but in some instances submicron particles were also relevant to the calculation. While these tiny particles are able to reach the surface of the 14° sharp wedge considered there, they tend to only graze the boundary layer as the vehicle geometry becomes more slender and are unable to reach the surface. Habeck et al. [10] showed through 2D numerical simulation how some of these ‘grazing’ particle trajectories excite second-mode wavepackets of comparable strength as those impinging the vehicle’s surface. Few details regarding why this occurs (i.e., the associated receptivity process) were provided, however.

Figure 1 highlights that a grazing particle can generate a wavepacket of comparable amplitude to that produced by an impinging particle, despite generating weaker acoustic radiation and fewer small-scale flow structures. The importance of grazing trajectories was also demonstrated in the statistical model developed by the present authors in Al Hasnine et al. [1], where their inclusion increased the mean flux of particles capable of exciting wavepackets by approximately 185%. Since the particle flux is a key input to statistical models, this study seeks to clarify the role grazing particles play in initiating LTT in hypersonic free flight. We first develop the method of multiples scales (MMS) receptivity formulation within a spatial stability framework to model the wavepackets excited by grazing particulates over extended streamwise distance. After verifying the formulation against the numerical method of Browne et al. [3], we use the MMS analysis to elucidate the receptivity mechanisms associated with both impinging and grazing particle trajectories. Finally, the statistical model of Chuvakhov & Fedorov [6] is used to show that the transition location predicted by grazing particles agree more closely with recent flight data Borg et al. [2].

Pressure disturbance field comparison between impinging (left) and grazing (right) trajectory scenarios

Figure 1. Pressure disturbance field comparison between impinging (left) and grazing (right) trajectory scenarios with ρp∗=0.50\rho_p^* = 0.50 g/cm3^3 and ρp∗=0.25\rho_p^* = 0.25 g/cm3^3, respectively.

Methodology

Flow conditions, geometry, and base flow

The significance of grazing particles is examined using the geometry and flight conditions of the recent HIFLIER hypersonic flight experiment Borg et al. [2]. We consider the freestream conditions at 26.95 seconds into ascent, when the vehicle is at an altitude of approximately h∞=21.5h_{\infty} = 21.5 km. Based on the data reported by Borg et al. [2] and 1993 ICAO standard atmosphere, the prescribed freestream conditions are P∞=4.359P_{\infty} = 4.359 kPa, T∞=218.1T_{\infty} = 218.1 K, M∞=6.25M_{\infty} = 6.25, and Re∞=9.05(106)Re_{\infty} = 9.05(10^{6}) 1/m. The geometry consisted of a slightly blunted straight cone with 7∘7^{\circ} half angle, 2.5 mm nose radius, and total length of 1.6 m.

An isothermal wall condition of Tw=347T_{w} = 347 K is imposed to compute a laminar base flow using the CHAMPS CFD framework utilising a body-fitted grid, 5th-order WENO for convective terms, and 2nd-order central discretisation for the viscous terms (McQuaid & Brehm [12]). Grid convergence was verified by changing the initial wall spacing from Δηw=4\Delta\eta_{w} = 4 μ\mum to 1 μ\mum. The resulting change in surface heat flux near the outlet was less than 0.025%, indicating that the laminar base flow solution is effectively grid independent.

Solving the Disturbance Flow Equations

The numerical framework developed by Browne et al. [3] for simulating particle-induced LTT is employed within the in-house CHAMPS CFD solver. The method solves the Navier-Stokes equations in nonlinear disturbance equation (NLDE) form and uses an efficient adaptive-mesh-refinement, wavepacket tracking (AMR-WPT) algorithm. The present simulations use the particle-source-in-cell formulation, in which the particulate-induced momentum and energy source terms (see Browne et al. [3]) are added to the right-hand-side of the NLDE through a spatial kernel, δ∗(x∗−xp∗(t∗))\delta^{*}(\mathbf{x}^{*} - \mathbf{x}_{p}^{*}(t^{*})), defined as

δ3D∗=1(σ∗2π)3exp⁡(−∣x∗−xp∗(t∗)∣22(σ∗)2),δ2D∗=12π(σ∗)2exp⁡(−∣x∗−xp∗(t∗)∣22(σ∗)2),(1)\delta^{*}_{3D} = \frac{1}{(\sigma^{*}\sqrt{2\pi})^{3}} \exp\left(\frac{-\lvert\mathbf{x}^{*} - \mathbf{x}_{p}^{*}(t^{*})\rvert^{2}}{2(\sigma^{*})^{2}}\right), \qquad\delta^{*}_{2D} = \frac{1}{2\pi(\sigma^{*})^{2}} \exp\left(\frac{-\lvert\mathbf{x}^{*} - \mathbf{x}_{p}^{*}(t^{*})\rvert^{2}}{2(\sigma^{*})^{2}}\right), \tag*{(1)}

where (⋅)∗(\cdot)^{*} denotes dimensional quantities. The 2D kernel chosen this way directly yields the spanwise spectral density flow field for spanwise wavenumber β∗=0\beta^{*} = 0. This equivalence is demonstrated for the planar, second Mack mode in § 3.1.

The particle position, xp∗=xp∗(t∗)\mathbf{x}_{p}^{*} = \mathbf{x}_{p}^{*}(t^{*}), in eq. (1) is obtained by integrating the Lagrangian equations of motion from prescribed initial conditions located immediately downstream of the conical shock. The drag coefficient for a spherical particle of radius Rp∗R_{p}^{*} is evaluated using the correlation of Singh et al. [15], which is applicable from the continuum to the freemolecular regime. The adaptive Cartesian mesh used to resolve both the particle source kernel and the resulting wavepacket had minimum cell spacings of (Δxmin⁡∗,Δymin⁡∗)=(11.7,6.29)(\Delta x_{\min}^{*}, \Delta y_{\min}^{*}) = (11.7, 6.29) μ\mum for the two-dimensional simulations. The kernel width is specified as σ∗/Rp∗=2.5\sigma^{*}/R_{p}^{*} = 2.5. Previous work has shown this choice accurately captures the particle-disturbance interaction while rendering the results insensitive to σ∗\sigma^{*}. We further verified this behaviour using simulations with σ∗/Rp∗=5.0\sigma^{*}/R_{p}^{*} = 5.0, which produced a maximum difference 4.5% in downstream wavepacket amplitude. Only aerodynamic drag is included in the particle equations of motion. Local thermal equilibrium between the particle and the surrounding fluid is assumed, such that Tp=TT_{p} = T and the particle heat-source term vanishes, i.e., Q˙p=2πNu(Tp−T)=0\dot{Q}_{p} = 2\pi Nu(T_{p} - T) = 0.

Method of Multiple Scales Particle Receptivity Model

A method of multiple scales (MMS) formulation is developed to solve the boundary-layer receptivity problem subject to distributed forcing. A related formulation was derived by Edwards & Tumin [7] for receptivity of kinetic fluctuations. The asymptotic analysis assumes that the base flow and the eigenfunctions of the locally parallel, homogeneous linear stability problem vary slowly in the streamwise direction. Accordingly, a slow streamwise coordinate is introduced as x1=ϵxx_{1} = \epsilon x with ϵ=1/R\epsilon= 1/R where the local Reynolds number is defined using boundary-layer-edge quantities as R=Ue∗Δ∗νe∗=RexR = \frac{U_{e}^{*}\Delta^{*}}{\nu_{e}^{*}} = \sqrt{Re_{x}}. Here, Ue∗U_{e}^{*}, Δ∗\Delta^{*} and νe∗\nu_{e}^{*} denote the boundary-layer-edge velocity, characteristic length scale and kinematic viscosity.

Substituting a WKBJ ansatz for the disturbance vector, together with the slowly varying (in xx) base flow into the forced linearised Navier-Stokes equations, and applying the solvability condition yields an ODE for the amplitude correction to the leading-order stability solution

dcdx1+cW+ϵ−1qK−1=0,(2)\frac{dc}{dx_1}+cW+\epsilon^{-1}qK^{-1}=0, \tag*{(2)}

where W=Γ−1⟨H2∂A0∂x1+H3A0,B0⟩W=\Gamma^{-1}\left\langle\mathbf{H}_2\frac{\partial\mathbf{A}_0}{\partial x_1}+\mathbf{H}_3\mathbf{A}_0,\mathbf{B}_0\right\rangle represents the non-parallel terms, q=Γ−1⟨G^,B0⟩q=\Gamma^{-1}\left\langle\widehat{\mathbf{G}},\mathbf{B}_0\right\rangle contains the Fourier transformed (in time and span) particle source terms G^\widehat{\mathbf{G}}, and K(x1)=exp⁡[iϵ−1∫x1,ux1α(η)dη]K(x_1)=\exp\left[i\epsilon^{-1}\int_{x_{1,u}}^{x_1}\alpha(\eta)d\eta\right] is the WKBJ ‘phase.’ The normalisation factor is defined as Γ=⟨H2A0,B0⟩\Gamma=\left\langle\mathbf{H}_2\mathbf{A}_0,\mathbf{B}_0\right\rangle where the solutions of the direct and adjoint O(1)O(1) stability problems for a particular mode with eigenvalue α∈C\alpha\in\mathbb{C} are A0\mathbf{A}_0 and B0\mathbf{B}_0, respectively. The definitions of matrices H2\mathbf{H}_2 and H3\mathbf{H}_3 are provided in Nayfeh [13]. Because the present study focuses on the excitation and subsequent exponential amplification of Mack’s second mode, the non-parallel correction is neglected. (2) is integrated from an upstream location x1,ux_{1,u} at which the particle forcing is negligible, i.e., q≈0q\approx0. With the direct eigenfunctions A0\mathbf{A}_0 normalised such that the flow quantity of interest has unity at the wall-normal location yiy_i, its dimensionless physical amplitude spectral density at the downstream location x1x_1 is given by

A(x1)=rp2c(x1)A0(x1,yi)K(x1),(3)A(x_1)=r_p^2c(x_1)\mathbf{A}_0(x_1,y_i)K(x_1), \tag*{(3)}

where rp=Rp∗/Δ∗r_p=R_p^*/\Delta^*. Finally, the wavepacket envelope is obtained from the magnitude of the analytic signal of the disturbance quantity of interest. For example, for the wall-pressure fluctuation, the envelope is given by ∣pa(x,z,t)∣\lvert p_a(x,z,t)\rvert. Its maximum is estimated to leading order by applying the method of steepest descent about the saddle point (ωs,βs)(\omega_s,\beta_s), yielding the following expressions for two- and three-dimensional wavepackets:

∣pa(x,zs,ts)∣∼(21−jπ)1−j2eN(xd,x)∣A(xd)(SωωSββj−jSωβ2)1/2∣(ωs,βs)(4)\left\lvert p_a(x,z_s,t_s)\right\rvert\sim\left(\frac{2^{1-j}}{\pi}\right)^{\frac{1-j}{2}}e^{N(x_d,x)} \left\lvert\frac{A(x_d)}{\left(S_{\omega\omega}S_{\beta\beta}^{j}-jS_{\omega\beta}^{2}\right)^{1/2}}\right\rvert_{(\omega_s,\beta_s)} \tag*{(4)}
{j=0,2Dj=1,3D\begin{cases} j=0,\quad2\mathrm{D}\\ j=1,\quad3\mathrm{D} \end{cases}

where zsz_s and tst_s are the spanwise location and moment in time the envelope maximum associated with the saddle point wave of maximum NN-factor, N(xd,x)=−Im⁡[S(xd,x)]N(x_d,x)=-\operatorname{Im}[S(x_d,x)], S(xd,x)=∫xdxα(η;ω,β)dηS(x_d,x)=\int_{x_d}^{x}\alpha(\eta;\omega,\beta)d\eta, passes over the given downstream observation location, xx (see Chuvakhov & Fedorov [6] for more details). The lower bound of integration, xdx_d, is some downstream location where the receptivity process is complete and the amplitude develops only due to exponential growth. The A(xd)A(x_d) in eq. 2.4 is evaluated with eq. 2.3 at that location.

The MMS particle receptivity model is implemented using the LST code from Tumin [16] to solve the direct and adjoint stability problems. Therefore, we take the inner product without complex conjugation, i.e., ⟨a,b⟩≡∫0∞∑i=18aibi dy\langle\mathbf{a},\mathbf{b}\rangle\equiv\int_{0}^{\infty}\sum_{i=1}^{8}a_i b_i\,dy, making B0\mathbf{B}_0 the eigenfunction of the complex conjugate of the conventional adjoint stability problem. Since the time signal of the particle source terms is given implicitly through the spatial kernel and its trajectory, for each (x,y)(x,y) we generate a time signal and take the Fourier transforms for use in eq. 2.3. We normalise the eigenfunction to have unit pressure fluctuation at the wall for ease of comparison with NLDE simulation data. Finally, x1,ux_{1,u} was chosen to be the position where the particle first reaches a wall-normal distance of 20Δ∗20\Delta^* and qq is effectively null.

Grazing Study

With all other parameters held fixed, the particle density, ρp\rho_p, and initial (post-shock) position, xp,0\mathbf{x}_{p,0}, are varied to generate two classes of trajectories: impinging trajectories, for which the particle strikes and reflects elastically from the cone surface, and grazing trajectories, for which the particle penetrates the boundary-layer edge—defined by 99.5% of the freestream stagnation enthalpy — but does not contact the wall. To ensure a consistent comparison, the initial particle positions are selected such that all trajectories cross the boundary layer edge at the same streamwise location, x∗=0.6x^* = 0.6 m. This construction isolates the influence of the particle trajectory on modal excitation, since the particle source terms do not depend explicitly on ρp\rho_p. The particle parameters considered in this study are summarised in Table 1.

Trajectory TypeImpingingGrazingGrazing
ρp∗\rho_p^* (g/cm3^3)0.500.250.20
xp,0∗x_{p,0}^* (m)(0.2790, 0.0259)(0.2090, 0.0208)(0.1790, 0.0188)

Table 1. Grazing study particle densities, initial positions and trajectory type.

These particle densities are set below those of representative atmospheric aerosols so that some particles with the selected radius, Rp∗=5R_p^* = 5 μm, follow grazing trajectories. These cases are intended to emulate the trajectories of submicron particles while retaining the same particle cross-sectional area, ApA_p, thereby enabling direct comparison of the disturbance amplitudes generated by grazing and impinging trajectories. This is appropriate because the particle-induced momentum force scales as fm∝qrelApCDf_m \propto q_{\mathrm{rel}}A_pC_D, where qrelq_{\mathrm{rel}} is the dynamic pressure based on the particle–fluid relative velocity, and CDC_D is the drag coefficient.

Results

Trajectory and NLDE Simulations

The trajectories corresponding to the particle densities listed in Table 1 are shown in Figure 2(a). The intended design of the study is evident: the particle with ρp∗=0.50\rho_p^* = 0.50 g/cm3^3 impinges on the cone surface and reflects, the particle with ρp∗=0.25\rho_p^* = 0.25 g/cm3^3 penetrates deeply into the boundary layer, and the particle with ρp∗=0.20\rho_p^* = 0.20 g/cm3^3 reaches the middle of the boundary layer. Hereafter, particle densities are normalised by the maximum considered, such that ρp=ρp∗/ρp,max⁡∗\rho_p = \rho_p^*/\rho_{p,\max}^*. The inset of Figure 2(b) compares wall-pressure amplitude spectral densities extracted at x∗=0.635x^* = 0.635 m from two- and three-dimensional NLDE simulations for the ρp=1.0\rho_p = 1.0 case, with the three-dimensional result evaluated at β∗=0\beta^* = 0. The close agreement confirms the two- and three-dimensional equivalence discussed in § 2.2. The remainder of the analysis therefore employs two-dimensional NLDE simulations, which provide physically meaningful spanwise spectral densities at β∗=0\beta^* = 0.

Figure 2. Grazing study. (a) Laminar baseflow and particle trajectories, and (b) Hilbert maximum of particle-induced wavepackets.

For each case, the envelope of the excited second-mode wavepacket is obtained at each streamwise location from the maximum magnitude of the Hilbert transform of the wall-pressure disturbance, p~w\tilde{p}_w. The resulting distributions are shown in Figure 2(b). Despite their markedly different trajectories, the ρp=1.0\rho_p = 1.0 impinging particle and the ρp=0.5\rho_p = 0.5 grazing particle (deep inside the boundary layer) generate wavepackets of nearly identical amplitude downstream. For the shallower grazing trajectory, ρp=0.4\rho_p = 0.4, the wavepacket amplitude increases more gradually as the particle enters the boundary layer, but ultimately reaches a comparable second-mode amplitude. An additional calculation for the ρp=1.0\rho_p = 1.0 case was performed with the nonlinear terms suppressed. The linear and nonlinear solutions are indistinguishable until the onset of saturation near x∗=0.9x^* = 0.9 m, confirming that the disturbance evolution remains linear over the region of interest.

Comparison of NLDE and MMS for particle receptivity

We now compare the full NLDE simulation with predictions from the MMS particle-receptivity model combined with method of steepest descent denoted as ‘MMS’, for second-mode excitation by impinging and grazing particles. Figure 3 (top) compares the dimensionless ASDASD spectra at three streamwise locations. The MMS receptivity model tends to underpredict the spectral amplitudes at frequencies above the dominant frequency, fmax⁡∗f_{\max}^{*}. This figure includes a comparison with the particle-receptivity model of Fedorov [8], denoted ‘FM’. This model is evaluated only at f∗=433.7f^{*} = 433.7 kHz for which the neutral point coincides with the particle impingement location and where the temporal formulation is strictly valid, i.e. Gaster transformation. The FM model is not applied to the grazing trajectories because their particle-boundary-layer interaction extends over a finite streamwise distance that cannot be represented accurately within a localised formulation. The bottom row of Figure 3 compares the wall-pressure amplitude, p~w\tilde{p}_{w}, normalised by twice dynamic pressure. The MMS prediction is evaluated by substituting the initial disturbance amplitude spectral densities (ASDASD) computed from equation (3) into the 2-D steepest-descent expression in equation (4). For all three particle trajectories, MMS accurately predicts the downstream wavepacket amplitude beyond the receptivity region.

Figure 3. Comparison of NLDE and MMS particle model.

The receptivity mechanics of impinging and grazing particles

We next examine the principle mechanisms governing particle receptivity for impinging and grazing trajectories. The particle forcing term qq, which drives the amplitude factor c(x1)c(x_1) in eq. (2), contains three non-zero contributions associated with the streamwise-momentum (q2q_2), wall-normal-momentum (q4q_4), and energy (q6q_6) source terms. Figure 4 shows the corresponding forcing distributions prior to wall-normal integration, together with their associated adjoint eigenfunctions (EF), at several streamwise locations for the ρp=1.0\rho_p = 1.0 and ρp=0.5\rho_p = 0.5 at f∗=452f^* = 452 kHz (chosen given its high ASDASD downstream). Each adjoint EF is normalised by its maximum magnitude. Within each column, the bilinear-product distributions, (Gi,B0,i)(G_i, B_{0,i}), are normalised by the maximum among the three rows to identify the locally dominant forcing contribution. The bars in the top row show the maxima of the unaltered (Gi,B0,i)/Γ(G_i, B_{0,i})/\Gamma distributions at each streamwise location and therefore provide a global comparison of their relative magnitudes. When the particle is above the boundary layer, indicated by purple dashed-dot line, the wall-normal-momentum component is locally dominant (e.g., R=2095R = 2095). As the particle enters the boundary layer near R=2200R = 2200, the streamwise relative velocity, urel=(up−ux)u_{rel} = (u_p - u_x), increases from approximately zero. At the same time, the particle crosses the critical layer, indicated by pink dashed line, where the adjoint EF components B0,2B_{0,2}, B0,4B_{0,4}, and B0,6B_{0,6}, attain large values. Because urel<1u_{rel} < 1, the streamwise-momentum forcing, which scales linearly with urelu_{rel}, exceeds the energy contribution containing urel2u_{rel}^{2} and, thus, becomes the dominant local contribution. As the particle approaches the wall at R=2247R = 2247, urelu_{rel} becomes O(1)O(1) while the energy adjoint EF component B0,6B_{0,6} exhibits a second peak. Their combined effect causes the energy forcing to become locally dominant. The maxima shown in the top row of Figure 4 further demonstrate that the wall-normal-momentum contribution remains relatively small to the streamwise-momentum and energy contributions once the particle has penetrated the boundary layer.

Forcing distributions for $\rho_p=1.0$ and $\rho_p=0.5$, showing three forcing components across particle trajectories

Figure 4. Forcing distributions for ρp=1.0\rho_p = 1.0 (top) and 0.5 (bottom) at f∗=452f^* = 452 kHz.

A similar receptivity mechanism is observed for the grazing ρp=0.5\rho_p = 0.5 trajectory, shown in the lower half of Figure 4. The principal difference is the absence of a substantial energy contribution to qq, because the particle cannot penetrate deep enough into the boundary layer to attain a sufficiently large relative velocity. Nevertheless, the wall-normal-momentum forcing outside and the streamwise-momentum forcing generated at the critical layer remain active and provide an effective mechanism for disturbance excitation.

Having established that the relative importance of particle-forcing components is dictated essentially by the particle’s position relative to corresponding adjoint EF, we now examine how they combine to complete the receptivity process. The individual components of qq in equation (2), obtained by integrating the profiles of (Gi,B0,i)/Γ(G_i, B_{0,i})/\Gamma in Figure 4, are plotted in Figure 5, again at f∗=452f^* = 452 kHz for both trajectory scenarios. For the impinging ρp=1.0\rho_p = 1.0 case, the dominant contributions to qq arise successively from q4q_4 outside the boundary layer, q2q_2 as the particle crosses the critical layer, and q6q_6 near the wall. The corresponding distributions are roughly symmetric about the impingement location. Each of these forcing regions produces a pronounced increase in the amplitude factor c(x1)c(x_1), indicating that the receptivity to an impinging particle proceeds through three distinct stages.

Components of forcing for $\rho_p=1.0$ and $\rho_p=0.5$

Figure 5. Components of forcing at f∗=452f^* = 452 kHz.

For the grazing ρp=0.5\rho_p = 0.5 trajectory, shown on the right-hand side of Figure 5, the dominant contributions to qq arise from q4q_4 outside the boundary layer and q2q_2 inside the boundary layer. As noted previously, the grazing particle does not penetrate sufficiently close to the wall for q6q_6 to make significant contributions, and therefore, the receptivity of grazing particles can be interpreted as a two-stage process. Only the qualitative behaviour of qq and c(x1)c(x_1) is compared between the two trajectories because their magnitudes depend on the upstream integration limit, x1,ux_{1,u}, which differs between the two cases. The reconstructed amplitude A(x1)A(x_1), however, is independent of the choice of x1,ux_{1,u}, and may therefore be compared directly

[7]. A key feature of the distributed receptivity process in both cases is the downstream plateau of c(x1)c(x_1). This behaviour arises because the factor K−1K^{-1} (grey dash-dot line in Figure 5) weighs the forcing contribution most strongly near the neutral point and decays away from it in both streamwise directions. Once the particle has moved sufficiently far downstream of this receptivity region, its forcing is overwhelmed by exponential growth.

Discussion

Having developed a theoretical framework capable of predicting both the receptivity and downstream amplification of wavepackets generated by impinging and grazing particles, we now assess the extent to which each trajectory class may contribute to laminar–turbulent

Impinging (Rp∗=2 μmR_p^* = 2\,\mu\mathrm{m})Grazing (Rp∗=0.275 μmR_p^* = 0.275\,\mu\mathrm{m})Flight
xtr∗x_{tr}^* (m)0.711.151.29

caption: Comparison of transition onset locations predicted by impinging and grazing particles 26.95 seconds into flight of HIFLIER vehicle [2].

:::

transition in hypersonic free flight. Equation 2.15 of Chuvakhov & Fedorov [6] is used to estimate the root-mean-square (RMS) wall pressure amplitude of second mode waves excited by particles of a prescribed radius. To extend the formulation to grazing particles, a discrete set of admissible grazing trajectories is searched at each saddle point, (ωs,βs)(\omega_s, \beta_s), and the trajectory producing the maximum growth is selected. Transition onset is then estimated by applying a threshold criterion to the resulting wall-pressure RMS. Predictions are presented in table 2 for particles with radii Rp∗=2 μmR_p^* = 2\,\mu\mathrm{m} (all trajectories impinge the cone wall) and Rp∗=0.275 μmR_p^* = 0.275\,\mu\mathrm{m} (no trajectories impinge the cone wall). The transition-onset location inferred from the flight measurements of Borg et al. [2] is also included for comparison. The transition-onset location predicted for grazing particles, defined as the position at which the wall-pressure RMS reaches 10% of the mean boundary-layer-edge pressure, agrees more closely with that inferred from the flight measurements at this stage of HIFLIER’s ascent. Moreover, the freestream concentration of submicron particles can exceed that of particles with Rp=2 μmR_p = 2\,\mu\mathrm{m} by two orders of magnitude [9]. Although each submicron grazing particle generates a lower-amplitude wavepacket, their substantially greater abundance suggests that they were more likely than larger impinging particles to sustain the transition front during this portion of the HIFLIER hypersonic flight.

Conclusion

This study sought to clarify the role that submicron atmospheric particles following grazing trajectories may play in initiating LTT during hypersonic free flight. A parametric investigation was conducted of Mack’s second mode wavepackets excited by solid, spherical particles within the boundary layer of a 7∘7^\circ straight cone flying Mach 6 at an altitude of 21.5 km, corresponding to the HIFLIER vehicle at 26.95 s into its ascent. Grazing particles have previously been expected to produce substantially weaker disturbances than larger particles that traverse the boundary layer and impinge on the wall [6]. The present NLDE simulations demonstrate, however, that the downstream wavepacket amplitudes generated by the grazing trajectories considered are at most 45% lower than those produced by impinging particles, even though the grazing particles penetrate only partially through the boundary layer before the receptivity process is essentially complete.

To explain this behaviour, a spatial formulation of the particle-receptivity problem was developed using the method of multiple scales. When combined with the method of steepest descent, the formulation accurately reproduces the downstream wavepacket amplitudes obtained from the NLDE simulations. The principal difference between the impinging three-stage and grazing two-stage receptivity mechanisms is the absence of a substantial energy-forcing contribution near the wall. Nevertheless, both trajectories produce strong streamwise-momentum forcing as the particle crosses the critical layer, where the boundary layer is most receptive. This critical-layer interaction accounts for the comparable disturbance amplitudes generated by the two trajectory classes. Finally, the statistical model of Chuvakhov & Fedorov [6] was used to estimate the RMS wall-pressure amplitude of particlegenerated second-mode disturbances. The transition-onset location predicted for grazing particles agrees more closely with that inferred from the HIFLIER flight measurements. Moreover, submicron particles may occur at freestream concentrations up to two orders of magnitude greater than larger particles with Rp=2 μmR_p = 2\ \mu\mathrm{m} [9]. Although an individual submicron grazing particle generates a lower-amplitude wavepacket, their substantially greater abundance, together with the strong receptivity associated with critical-layer crossing, suggests that grazing particles may have played a leading role in sustaining the transition front during this stage of the HIFLIER ascent.

Acknowledgements. The authors would like to thank Dr. Kevin Luna for helpful discussions.

Funding. Sean Dungan was supported by the DoD through the NDSEG Fellowship Program.

Declaration of interests. The authors report no conflict of interest.

References

  1. [1]Al Hasnine, S.M.A., Russo, V., Saikia, B. & Brehm, C. 2023 A model for particle-induced transition for hypersonic boundary layer flows. In AIAA AVIATION 2023 Forum.
  2. [2]Borg, M., Adamczak, D. & Tufts, M 2025 Hifl ier hypersonic flight experiment: Ascent results. In AIAA SciTech Forum 2025.DOI
  3. [3]Browne, O. M. F., Al Hasnine, S. M. A. & Brehm, C. 2021 Numerical method for particulate-induced high-speed boundary-layer transition simulations. AIAA Journal 59 (4), 1196–1213.DOI
  4. [4]Bushnell, D. 1990 Notes on initial disturbance fields for the transition problem. In Instability and Transition (ed. M. Y. Hussaini & R. G. Voigt), pp. 217–232. New York, NY: Springer US.DOI
  5. [5]Chou, A. & Marineau, E. C. 2024 A summary of the 2023 joint high-speed boundary layer transition workshop. Tech. Rep., NASA.
  6. [6]Chuvakhov, P. V. & Fedorov, A. V. 2025 Statistical model of laminar-turbulent transition onset caused by atmospheric particulates. JFM 1007.DOI
  7. [7]Edwards, L. D. & Tumin, A. 2019 Model of distributed receptivity to kinetic fluctuations in high-speed boundary layers. AIAA Journal 57, 4750–4763.DOI
  8. [8]Fedorov, A. V. 2013 Receptivity of a supersonic boundary layer to solid particulates. JFM 737, 105–131.DOI
  9. [9]Habeck, J.B., Flaten, J. & Candler, G.V. 2020 High-altitude balloon measurements of atmospheric particulates. In AIAA Scitech 2020 Forum.DOI
  10. [10]Habeck, J.B., Kroells, M.D., Knutson, A.L. & Candler, G.V. 2026 Numerical study on the role of aerosol particles in hypersonic boundary-layer transition. AIAA Journal 64, 1235–1252.
  11. [11]Mack, Leslie M 1977 Transition and laminar instability. Contractor Report NASA-CR-153203. NASA, California Institute of Technology, Pasadena, CA, USA.
  12. [12]McQuaid, J.A. & Brehm, C. 2024 Heat flux predictions for hypersonic flows with an overset near body solver on an adaptive block-structured cartesian off-body grid. Computers & Fluids 269, 106121.DOI
  13. [13]Nayfeh, Ali H. 1980 Stability of three-dimensional boundary layers. AIAA Journal 18 (4), 406–416.
  14. [14]Schneider, Steven P. 2008 Development of hypersonic quiet tunnels. JSR 45 (4), 641–664.DOI
  15. [15]Singh, N., Kroells, M., Li, C., Ching, E., Ihme, M., Hogan, C.J. & Schwartzent r u ber, T.E. 2022 General drag coefficient for flow over spherical particles. AIAA Journal 60, 587–597.DOI
  16. [16]Tumin, Anatoli 2007 Three-dimensional spatial normal modes in compressible boundary layers. JFM 586.

Paper details

vlnc:W_g8p2dkbt
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 2

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents