Hostname: page-component-7b7567c8c-2n9c9 Total loading time: 0 Render date: 2026-11-16T20:56:07.377Z Has data issue: false hasContentIssue false

Local and global instability of the rotating-disk boundary layer of a rotor-stator cavity

Published online by Cambridge University Press:  22 October 2025

Yaguang Xie
Affiliation:
Advanced Gas Turbine Laboratory, Institute of Engineering Thermophysics, Chinese Academy of Sciences, Beijing 100190, PR ChinaUniversity of Chinese Academy of Sciences, Beijing 100049, PR ChinaNational Key Laboratory of Science and Technology on Advanced Light-duty Gas-turbine, Beijing 100190, PR China
Zihao Zhu
Affiliation:
Department of Mechanical Engineering, City University of Hong Kong, Hong Kong 999077, PR China
Lei Xie
Affiliation:
Advanced Gas Turbine Laboratory, Institute of Engineering Thermophysics, Chinese Academy of Sciences, Beijing 100190, PR ChinaUniversity of Chinese Academy of Sciences, Beijing 100049, PR ChinaNational Key Laboratory of Science and Technology on Advanced Light-duty Gas-turbine, Beijing 100190, PR China
Ruonan Wang
Affiliation:
Advanced Gas Turbine Laboratory, Institute of Engineering Thermophysics, Chinese Academy of Sciences, Beijing 100190, PR ChinaUniversity of Chinese Academy of Sciences, Beijing 100049, PR ChinaNational Key Laboratory of Science and Technology on Advanced Light-duty Gas-turbine, Beijing 100190, PR China
Siyi Li
Affiliation:
Advanced Gas Turbine Laboratory, Institute of Engineering Thermophysics, Chinese Academy of Sciences, Beijing 100190, PR ChinaUniversity of Chinese Academy of Sciences, Beijing 100049, PR ChinaNational Key Laboratory of Science and Technology on Advanced Light-duty Gas-turbine, Beijing 100190, PR China
Guangzu Zhang
Affiliation:
Advanced Gas Turbine Laboratory, Institute of Engineering Thermophysics, Chinese Academy of Sciences, Beijing 100190, PR ChinaUniversity of Chinese Academy of Sciences, Beijing 100049, PR ChinaNational Key Laboratory of Science and Technology on Advanced Light-duty Gas-turbine, Beijing 100190, PR China
Qiang Du*
Affiliation:
Advanced Gas Turbine Laboratory, Institute of Engineering Thermophysics, Chinese Academy of Sciences, Beijing 100190, PR ChinaUniversity of Chinese Academy of Sciences, Beijing 100049, PR ChinaNational Key Laboratory of Science and Technology on Advanced Light-duty Gas-turbine, Beijing 100190, PR ChinaQingdao Institute of Aeronautical Technology, Qingdao 266400, PR China
Junqiang Zhu
Affiliation:
Advanced Gas Turbine Laboratory, Institute of Engineering Thermophysics, Chinese Academy of Sciences, Beijing 100190, PR ChinaUniversity of Chinese Academy of Sciences, Beijing 100049, PR ChinaNational Key Laboratory of Science and Technology on Advanced Light-duty Gas-turbine, Beijing 100190, PR China
*
Corresponding author: Qiang Du, duqiang@iet.cn

Abstract

This study investigates the stability characteristics of rotating-disk boundary layers in rotor–stator cavities under the frameworks of local linear, global linear and global nonlinear analyses. The local linear stability analysis uses the Chebyshev polynomial method, the global linear stability analysis relies on the linearised incompressible Navier–Stokes (N–S) equations and the global nonlinear analysis involves directly solving the complete incompressible N–S equations. In the local linear framework, the velocity profile derived from the laminar self-similar solution on the rotating-disk side of an infinite rotor–stator cavity is mapped to the Bödewadt–Ekman–von Kármán theoretical model to establish a unified analytical framework. For the global stability study, we extend the methodological framework proposed by Appelquist et al. (J. Fluid Mech.,vol 765, 2015, pp. 612–631) for the von Kármán boundary layer, implementing pulsed disturbances and constructing a radial sponge layer to effectively capture the spatiotemporal evolution of perturbation dynamics while mitigating boundary reflection effects. The analysis reveals that the rotating-disk boundary layer exhibits two distinct instability regimes: convective instability emerges at ${\textit{Re}}=r^*/\sqrt {\nu ^*/\varOmega ^*}=204$ (where $r^*$ is the radius, $\nu ^*$ is the kinematic viscosity and $\varOmega ^*$ is the rotation rate of the system) with azimuthal wavenumber $\beta =27$, while absolute instability emerges at ${\textit{Re}}=409.6$ with azimuthal wavenumber $\beta =85$. Under pulsed disturbance excitation, an initial convective instability behaviour dominates in regions exceeding the absolute instability threshold. As perturbations propagate into the sponge layer’s influence domain, upstream mode excitation triggers the emergence of a global unstable mode, characterised by a minimum critical Reynolds number ${\textit{Re}}_{\textit{end}}=484.4$. Further analysis confirms that this global mode is an inherent property of the rotating-disk boundary layer and is independent of the characteristics of the sponge layer. Frequency-domain analysis establishes that the global mode frequency is governed by local stability characteristics at ${\textit{Re}}_{\textit{end}}$, while its growth rate evolution aligns with absolute instability trends. By further incorporating nonlinear effects, it was observed that the global properties of the global nonlinear mode remain governed by ${\textit{Re}}_{\textit{end}}$. The global temporal frequency corresponds to ${\textit{Re}}_{\textit{end}}=471.8$. When ${\textit{Re}}$ approaches 517.2, the spiral waves spontaneously generate ring-like vortices, which subsequently trigger localised turbulence. This investigation provides novel insights into the fundamental mechanisms governing stability transitions in the rotating-disk boundary layer of the rotor–stator cavity.

Information

Type
JFM Papers
Copyright
© The Author(s), 2025. Published by Cambridge University Press
Figure 0

Figure 1.A sketch of the computational domain for rotor–stator cavity flow. Here $\varOmega _d^*$ is the angular velocity and $2h^*$ is the distance between stator and rotor.

Figure 1

Figure 2.Laminar self-similar velocity profiles of an infinite rotor–stator cavity for different Reynolds numbers ${\textit{Re}}_h$.

Figure 2

Figure 3.The distribution of radial velocity $\widetilde U$, azimuthal velocity $\widetilde V$ and axial velocity $\widetilde W$ as a function of the axial coordinate $z_\delta$ for different Reynolds numbers ${\textit{Re}}_h$.

Figure 3

Figure 4.The distribution of $\omega$ in the complex $\alpha$ plane under the conditions ${\textit{Re}}=530$ and $\beta =67$ for the von Kármán boundary layer. ($a$) Temporal frequency $\omega _r$, with the red $*$ and blue $*$ denoting the convergence trajectories initiated at $\alpha =0.28+\mathrm{i}0.15$ and $\alpha =0.12-\mathrm{i}0.20$, respectively. These trajectories converge to the critical point $\partial \omega / \partial \alpha =0$ following the steepest descent method. The black arrows indicate the direction of convergence, and the critical condition $\partial \omega / \partial \alpha =0$ is achieved at $\alpha =0.213-\mathrm{i}0.123$ and $\omega =0.031098+\mathrm{i}0.000347$. ($b$) Temporal growth rate $\omega _i$, where the three solid black contours correspond to $\omega _i=0.01$, $0.004$ and $0.000289$, respectively. The flow conditions discussed here are derived from figure 7 in Lingwood (1995).

Figure 4

Table 1.The absolute instability mode of the von Kármán boundary layer at ${\textit{Re}}=530$ and $\beta =67$, showing the corresponding $\alpha$ and $\omega$ obtained with a spectral method based on different Chebyshev polynomial orders. For comparison, the results from Lingwood (1995) were obtained using the Shooting method.

Figure 5

Table 2.Critical values for the onset of absolute instability in the von Kármán boundary layer.

Figure 6

Figure 5.The eigenvalue spectra $(\alpha _r, -\alpha _i)$ obtained under spatial mode analysis for $\beta =45$ and $\omega =0$ at Reynolds numbers ${\textit{Re}}=200, 250, 300, 350, 400, 450, 500, 550$.

Figure 7

Figure 6.Neutral ($\omega _i= 0, \alpha _i= 0$) stability diagrams for stationary modes.

Figure 8

Figure 7.($a$) Variation of the two spatial modes with the largest radial growth rates, $-\alpha _i$, as a function of ${\textit{Re}}$ under the condition $\omega = 0$. ($b$) The two modes with the largest growth rates exhibit a crossing. In this schematic, at the crossing point ${\textit{Re}}$, the eigenvector $\hat {\boldsymbol{q}}$ corresponding to the larger $-\alpha _i$ is labelled as $a$, while the eigenvector corresponding to the smaller $-\alpha _i$ is labelled as $A$. By incrementally varying the Reynolds number with a step size $\Delta {\textit{Re}} = 1$, the corresponding eigenvectors are defined as the letters shown in the figure.

Figure 9

Table 3.The dot product between various eigenvectors for selected representative wavenumbers, as illustrated in Figure 7($b$), is computed as ${ab} = \hat {\boldsymbol{q}}_a^{H} \hat {\boldsymbol{q}}_b$. The values highlighted in bold denote the four largest entries among the eight results, which are also the closest to 1, signifying a high degree of similarity between the corresponding eigenvectors.

Figure 10

Figure 8.Radial wavenumber $\alpha$ for disturbances of the normal mode form for $\beta = 45$ and $\omega = 0$.

Figure 11

Figure 9.Under the local linear stability analysis of the spatial mode, the evolution of the radial growth rate $-\alpha _i$ of type I and type III modes with respect to the Reynolds number ${\textit{Re}}$, given $\omega _{r_c} = -0.038617$.

Figure 12

Figure 10.The disturbance velocity distribution of the type I mode corresponding to ${\textit{Re}}=409.6,\beta =85$ and $\omega _r = -0.038617$.

Figure 13

Figure 11.Absolute instability of the rotating-disk boundary layer. ($a$) Neutral ($\omega _i= 0, \partial \omega /\partial \alpha = 0$) stability diagrams for absolute instability, ($b$) Variation of the complex frequency $\omega = \omega _r + \mathrm{i}\omega _i$ of absolute instability with the Reynolds number ${\textit{Re}}$ for $\beta = 85$.

Figure 14

Figure 12.Meridional plane $(r, z)$ schematic of the computational model. ($a$) The computational domain model includes a rotating disk and a stationary disk. At ${\textit{Re}}_h=2000$, the rotating disk is located at $z=0$ and the stationary disk at $z=89.4$. To avoid the influence of the stationary-disk side, only the pink region is selected as the computational domain for the simulation. ($b$) Computational domain for global linear simulation L5. The boundary condition $v$ is Dirichlet, and ON indicates a mixture of both Dirichlet and homogeneous Neumann boundary conditions. The axial computational domain extends over $0 \leqslant z \leqslant 20$. ($c$) Computational domain for DNS NL1, the axial computational domain extends over $0 \leqslant z \leqslant 49.3$. ($d$) Computational domain for DNS NL2, where SYM represents a symmetry condition. The grey shaded region indicates the smooth growth area of the sponge function, while the black shaded region indicates the area where the sponge function has reached its maximum value.

Figure 15

Table 4.Summary of the parameters for the linear simulations. Here L1–L8 represent the linear simulations and NL1 and NL2 represent the DNS in this study. The radial outer boundary of the domain is indicated by $r_d$, and a sponge layer is applied at the outer boundary starting from $r_s$, ${\textit{Re}}_d$ and ${\textit{Re}}_s$ are the Reynolds numbers corresponding to $r_d$ and $r_s$, respectively. Here ${\textit{Re}}_{\textit{end}}$ represents the end of the linear region, while $r_{\textit{ex}}$ denotes the radial location of the initial impulsive disturbance, ${\textit{Re}}_{\textit{ex}}$ is the Reynolds numbers corresponding to $r_{\textit{ex}}$; $N_r$ and $N_{nel}$ represent the number of spectral elements in the radial direction and the total number of spectral elements in the simulation, respectively.

Figure 16

Figure 13.Space–temporal diagram from L1 ($r_d=780, {\textit{Re}}_{d}=560.3$) and L5 ($r_d=820, {\textit{Re}}_{d}=589.1$); ($a$) $\mathrm{log}_{10}(V_{\textit{rms}})$ from L1, ($b$) $\mathrm{dln}(V_{\textit{rms}}(t))/\mathrm{d}t$ from L1, ($c$) $\mathrm{log}_{10}(V_{\textit{rms}})$ from L5, ($d$) $\mathrm{dln}(V_{\textit{rms}}(t))/\mathrm{d}t$ from L5.

Figure 17

Figure 14.The space–temporal diagram is derived from L2 ($r_{\textit{ex}}=560,{\textit{Re}}_{\textit{ex}}=402.3$), L3 ($r_{\textit{ex}}=600,{\textit{Re}}_{\textit{ex}}=431.0$) and L4 ($r_{\textit{ex}}=640,{\textit{Re}}_{\textit{ex}}=460.0$), each with $r_d=800$, ${\textit{Re}}_d=574.4$; ($a$) $\mathrm{log}_{10}(V_{\textit{rms}})$ from L2, ($b$) $\mathrm{dln}(V_{\textit{rms}}(t))/\mathrm{d}t$ from L2, ($c$) $\mathrm{log}_{10}(V_{\textit{rms}})$ from L3, ($d$) $\mathrm{dln}(V_{\textit{rms}}(t))/\mathrm{d}t$ from L3, ($e$) $\mathrm{log}_{10}(V_{\textit{rms}})$ from L4, ($f$) $\mathrm{dln}(V_{\textit{rms}}(t))/\mathrm{d}t$ from L4.

Figure 18

Figure 15.($a$) Space–temporal diagram of $\mathrm{log}_{10}(V_{\textit{rms}})$ from L6 ($r_d=820,{\textit{Re}}_d=589.1,r_s=740. {\textit{Re}}_s=531.6$). ($b$) Space–temporal diagram of $\mathrm{log}_{10}(V_{\textit{rms}})$ from L7 ($r_d=820,{\textit{Re}}_d=589.1,r_s=760.{\textit{Re}}_s=546.0$).

Figure 19

Figure 16.Radial growth rate $\mathrm{dln} V_{\textit{rms}}(r_\delta )/\mathrm{ d } r_\delta$ from L5 ($r_d=820,{\textit{Re}}_d=589.1,r_s=720.{\textit{Re}}_s=517.2$), L6 ($r_d=820,{\textit{Re}}_d=589.1,r_s=740.{\textit{Re}}_s=531.6$), L7 ($r_d=820,{\textit{Re}}_d=589.1,r_s=760.{\textit{Re}}_s=546.0$) and L8 ($r_d=840,{\textit{Re}}_d=603.4,r_s=740.{\textit{Re}}_s=531.6$).

Figure 20

Figure 17.($a$) Temporal growth rate $\mathrm{dln} V_{\textit{rms}}(t)/\textrm {d} t$ from L1 to L5 and L8 at ${\textit{Re}}=450$. ($b$) Radial growth rate $\textrm {dln} V_{\textit{rms}}(r_\delta )/\textrm {d} r_\delta$ from L1, L3, L5 and L8 at $t/(2\unicode {x03C0}) =3$, NL1 at $t/(2\unicode {x03C0}) = 9$ (black line) and local theory data ($-\alpha _i$) for the absolute instability pinch points for $\beta =85$ are also included (black line, $+-$) together with data from the spatial modes contributing to the pinch points at end for each simulation (grey lines, $-\!\,-$).

Figure 21

Figure 18.Variation of the global frequency with ${\textit{Re}}$ for $\beta =85$ together with linear simulation data at ${\textit{Re}}=450$, the black circle represents NL1. ($a$) The final growth rates of the global modes, $\omega _{gi}$. ($b$) The real temporal frequency $\omega _{gr}$.

Figure 22

Figure 19.Space–temporal diagram from NL1 ($r_d=900,{\textit{Re}}_d=646.5$) and NL2 ($r_d=900,{\textit{Re}}_d=646.5$); ($a$) $\mathrm{log}_{10}(V_{\textit{rms}})$ from NL1, ($b$) $\mathrm{log}_{10}(V_{\textit{rms}})$ from NL2.

Figure 23

Table 5.For NL1, when $t/(2\unicode {x03C0}) = 6, 7, 8$ and $9$, the linear regime endpoints ${\textit{Re}}_{\textit{end}}$ are determined based on the intersection points between their radial growth rate ${{\rm dln}} V_{\textit{rms}}(r_\delta )/\textrm {d} r_\delta$ and the local theory data ($-\alpha _i$) for the absolute instability pinch points for $\beta = 85$.

Figure 24

Table 6.For NL2, when $t/(2\unicode {x03C0}) = 8, 10, 12$ and $14$, the linear regime endpoints ${\textit{Re}}_{\textit{end}}$ are determined based on the intersection points between their radial growth rate ${{\rm dln}} V_{\textit{rms}}(r_\delta )/\textrm {d} r_\delta$ and the local theory data ($-\alpha _i$) for the absolute instability pinch points for $\beta = 85$.

Figure 25

Figure 20.Amplitude $\mathrm{log_{10}}(P)$ for the temporal frequencies of the signals. The frequencies are calculated over the last three cycles of the simulation time in the reference frame rotating with $\varOmega _{\!f}$; ($a$) NL1, where the black dashed line represents the critical Reynolds number for the energy-filled full spectrum; ($b$) NL2.

Figure 26

Figure 21.For NL1, FFT analysis along the azimuthal direction $\theta$ is performed at each time step for $t/(2\unicode {x03C0}) = 6 - 9$, yielding modes corresponding to azimuthal wavenumbers $\beta = 85, 170, 255, 340$ and $425$. The figure presents the $\mathrm{log_{10}}(V_{\textit{rms}})$ of these modes, averaged over time, denoted as $\mathrm{log_{10}}(\overline {V}_{\!\textit{rms}})$. ($a$) The distribution of modal energy $\mathrm{log{10}}(\overline {V}_{\!\textit{rms}})$ as a function of ${\textit{Re}}$. ($b$) The distribution of modal energy $\mathrm{log{10}}(\overline {V}_{\!\textit{rms}})$ as a function of the dominant instability mode energy $\mathrm{log{10}}(\overline {V}_{\!\textit{rms}}(\beta = 85))$.

Figure 27

Figure 22.Direct numerical simulation results for NL1. ($a$) The dimensionless azimuthally averaged velocity $\widetilde {V}$ as a function of the axial position for various radial locations. The results are presented in the same dimensionless format as figure 3, with the red solid line indicating the laminar self-similarity profile, consistent with the results in figure 3. The corresponding radial positions, from left to right, are: $\mathrm{A}: {\textit{Re}}=400$, $\mathrm{B}: {\textit{Re}}=420$, $\mathrm{C}: {\textit{Re}}=440$, $\mathrm{D}: {\textit{Re}}=460$, $\mathrm{E}: {\textit{Re}}=480$, $\mathrm{F}: {\textit{Re}}=500$, $\mathrm{G}: {\textit{Re}}=520$, $\mathrm{H}: {\textit{Re}}=540$, $\mathrm{I}: {\textit{Re}}=560$, $\mathrm{J}: {\textit{Re}}=580$. ($b$) The distribution of the dimensionless instantaneous perturbation velocity $(V-\overline {V})/(\varOmega _d r)$ in the (${\textit{Re}},z_\delta$) plane at $\theta =0$, $\overline {V}$ is the azimuthal average of the azimuthal velocity $V$. ($c$) The Q-criterion vortex structure visualisation: $Q=20$, colour represents dimensionless axial velocity $W/(\varOmega _d r)$, the black isosurface corresponds to ${\textit{Re}} = 517.2$, the results correspond to two $2\unicode {x03C0}/85$ sectors.

Figure 28

Table 7.Comparison of the present results with the data for $\textit{Ro}=-1$ and $\textit{Ro}=-0.71$, where the absolute instability corresponds to $\textit{Ro}=-0.71$ and the base flow is the DNS average flow of the rotor–stator cavity. Here ${\textit{Re}}_c$ represents the critical Reynolds number for convective instability type I, while ${\textit{Re}}_{c}$ denotes the critical Reynolds number for absolute instability; ${\textit{Re}}_{r}$ is the critical Reynolds number converted to the rotational Reynolds number, defined as ${\textit{Re}}_{r}=\varOmega _dr^{2}/\nu$, based on the local radius $r$; $\beta$ symbolises the azimuthal wavenumber, and $\omega _r$ and $-\alpha _i$ represent the temporal frequency and radial growth rate, respectively, corresponding to the critical Reynolds number.