Hostname: page-component-7b7567c8c-wxvwp Total loading time: 0 Render date: 2026-11-17T07:10:19.167Z Has data issue: false hasContentIssue false

The hunt for the Kármán ‘constant’ revisited

Published online by Cambridge University Press:  17 July 2023

Peter A. Monkewitz*
Affiliation:
École Polytechnique Fédérale de Lausanne (EPFL), CH-1015 Lausanne, Switzerland
Hassan M. Nagib
Affiliation:
Illinois Institute of Technology (IIT), Chicago, IL 60616, USA
*
†Email address for correspondence: peter.monkewitz@epfl.ch

Abstract

The log law of the wall, joining the inner, near-wall mean velocity profile (MVP) in wall-bounded turbulent flows to the outer region, has been a permanent fixture of turbulence research for over hundred years, but there is still no general agreement on the value of the prefactor, the inverse of the Kármán ‘constant’ $\kappa$, or on its universality. The choice diagnostic tool to locate logarithmic parts of the MVP is to look for regions where the indicator function $\varXi$ (equal to the wall-normal coordinate $y^+$ times the mean velocity derivative $\mathrm {d} U^+/\mathrm {d} y^+$) is constant. In pressure-driven flows, however, such as channel and pipe flows, $\varXi$ is significantly affected by a term proportional to the wall-normal coordinate, of order $O({Re}_{\tau }^{-1})$ in the inner expansion, but moving up across the overlap to the leading $O(1)$ in the outer expansion. Here we show that, due to this linear overlap term, ${Re}_{\tau }$ values well beyond $10^5$ are required to produce one decade of near constant $\varXi$ in channels and pipes. The problem is resolved by considering the common part of the inner asymptotic expansion carried to $O({Re}_{\tau }^{-1})$, and the leading order of the outer expansion. This common part contains a superposition of the log law and a linear term $S_0 \,y^+{Re}_{\tau }^{-1}$, and corresponds to the linear part of $\varXi$, which, in channel and pipe, is concealed up to $y^+ \approx 500\unicode{x2013}1000$ by terms of the inner expansion. A new and robust method is devised to simultaneously determine $\kappa$ and $S_0$ in pressure-driven flows at currently accessible ${Re}_{\tau }$ values, yielding $\kappa$ values which are consistent with the $\kappa$ values deduced from the Reynolds number dependence of centreline velocities. A comparison with the zero-pressure-gradient turbulent boundary layer, further clarifies the issues and improves our understanding.

Information

Type
JFM Papers
Creative Commons
Creative Common License - CCCreative Common License - BY
This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http://creativecommons.org/licenses/by/4.0), which permits unrestricted re-use, distribution and reproduction, provided the original article is properly cited.
Copyright
© The Author(s), 2023. Published by Cambridge University Press.
Figure 0

Figure 1.Illustration of the effect of the linear term $S_0\, y^+/{Re}_{\tau }$ on the estimate of $\kappa ^{-1}$ from the overlap $\varXi _{\mathrm {OL}}$ (1.2) with (red) $- - -$, $-\cdot -\cdot$, $-\cdot \cdot \,-$, $\kappa$ errors of $-1\,\%$, $-2\,\%$ and $-3\,\%$ relative to $\kappa _{ref}=0.417$. Linear deviations $S_0\, y^+/{Re}_{\tau }$ from the baseline log law for ${Re}_{\tau } = 10^4$, $10^5$, $10^6$ and $10^7$ (increasingly dark blue) with $S_0=1.15$ (—) and $S_0=2.5$ ($\cdot \cdot \cdot$); (red) $\bullet$, locations where the linear terms with $S_0 = 1.15$ induce a $-1\,\%$ error of $\kappa$.

Figure 1

Figure 2.Successive approximations of mean velocity derivative $\mathrm {d} U^+_{\mathrm {DNS}}/\mathrm {d} Y$ for eight channel DNS: $- \cdot -$ (grey), ${Re}_{\tau } = 934$ (del Alamo et al.2004); $- - -$ (grey), ${Re}_{\tau } = 1001$ (Lee & Moser 2015); — (grey), ${Re}_{\tau } = 1995$ (Lee & Moser 2015); — (dark orange), ${Re}_{\tau } = 3986$ (Yamamoto & Tsuji 2018); — (yellow), ${Re}_{\tau } = 4079$ (Bernardini, Pirozzoli & Orlandi 2014); — (red), ${Re}_{\tau } = 5186$ (Lee & Moser 2015); — (violet), ${Re}_{\tau } = 8000$ (Yamamoto & Tsuji 2018); — (light grey), ${Re}_{\tau } = 10049$ (Hoyas et al.2022). (a) Graph of $\mathrm {d} U^+_{\mathrm {DNS}}/\mathrm {d} Y - (0.417Y)^{-1}$; (b) profiles in (a) minus constant $(1.15 + 380\,{Re}_{\tau }^{-1})$ , $\bullet \bullet \bullet$ (blue), wake fit $[\kappa ^{-1} + 1.15 + 380\,{Re}_{\tau }^{-1}] (\mathrm {d} W/\mathrm {d} Y)$ (2.1), (2.2a–d); (c) profiles in (b) minus wake fit.

Figure 2

Figure 3.Channel indicator functions $\varXi$ from DNS for the six highest ${Re}_{\tau }$ cases of figure 2 (${Re}_{\tau } = 1995$ and up). Same colour scheme as in figure 2. (a) Lines and symbols: —, total $\varXi$ versus $y^{+}$; $\bullet \bullet \bullet$, linear overlap fits $[\kappa ^{-1} + (S_0 + S_1\,{Re}_{\tau }^{-1})Y]$ ; - - -, complete outer fits of $\varXi$ for ${Re}_{\tau } = 4079$ and up (2.1), (2.2a–d); — (blue), $(1/0.417)$; ${\blacksquare}$ (green), $y^+ = 1200$ marking the approximate start of the overlap; ${\blacksquare}$ (red), $Y = 0.45$ marking the approximate end of the overlap. Note that for the lowest ${Re}_{\tau } = 1995$, the overlap ends before it starts. (b) Indicator functions in (a), corrected for finite ${Re}_{\tau }$ effects according to (2.1), (2.2a–d), i.e. $\varXi _\infty = \varXi - S_1Y[1- \mathrm {d} W/\mathrm {d} Y]/{Re}_{\tau }$, versus outer Y; $\bullet \bullet \bullet$ (blue), leading-order linear fit $(1/0.417) + 1.15Y$; grey band, variation of linear fit for $0.407 \leq \kappa \leq 0.427$.

Figure 3

Figure 4.Indicator function $\varXi _\infty (Y)$ and $U^+_\infty (Y)$ minus the linear part of the overlap for two channel/duct experiments. The subscript ‘$\infty$’ indicates that all data are corrected for finite Reynolds number effects with the ${Re}_{\tau }^{-1}$ corrections in (2.1)–(2.4). Hot wire data of Zanoun et al. (2003): $+$ (black), ${Re}_{\tau } = 1167$, 1543, $1851$; $\blacktriangle$ (grey), ${Re}_{\tau } = 2155$, $2573$, $2888$; ${\blacksquare}$ (blue), ${Re}_{\tau } = 3046,$ 3386, 3698, 3903; $\bullet$ (dark blue), ${Re}_{\tau } = 4040$, 4605, $4783$. The laser Doppler anemometry data of Schultz & Flack (2013): $\lozenge$ (red, increasing size), ${Re}_{\tau } = 1010$, 1956, 4048, 5895. (a) Graph of $\varXi _\infty (Y)$: — (green), linear fit $(1/0.417) + 1.1Y$ (note that the fitted $S_0 = 1.1$ is slightly reduced relative to the best DNS fit in (2.2a–d)). (b) Blowup of $\varXi _\infty$ versus $y^+$, with linear fits $[(1/0.417) + 1.1\,y^+/{Re}_{\tau }]$ for ${Re}_{\tau } = 1167$ and 5895. (c) Corrected $U^+_\infty (Y)$ minus linear fit $[(1/0.417)\ln {Re}_{\tau } + 1.1\,Y]$; - - - (green), resulting log law $[(1/0.417)\,\ln Y + 5.5]$.

Figure 4

Figure 5.Indicator functions $\varXi$ for selected pipe DNS (not corrected for finite ${Re}_{\tau }$): — (yellow), ${Re}_{\tau } = 999$ of El Khoury et al. (2013); $-\cdot \cdot -$, $-\cdot -$, — (violet), ${Re}_{\tau } =$ 1976, 3028 and 6019 of Pirozzoli et al. (2021); — (red), ${Re}_{\tau } = 5197$ of Yao et al. (2023); $\bullet \bullet \bullet$ (violet), linear fit $(1/0.385) + 1.95Y$ of ${Re}_{\tau } = 6019$ profile; $\bullet \bullet \bullet$ (pink), linear fit $(1/0.425) + 2.75Y$ of ${Re}_{\tau } = 5197$ profile.

Figure 5

Figure 6.Indicator function $\varXi (Y)$ and $U^+(Y)$ minus linear part of overlap (3.1) for the Superpipe data of McKeon (2003) and Bailey et al. (2013): $+$ (black), ${Re}_{\tau } < 5.10^3$; $\blacktriangle$ (grey), $5.10^3 < {Re}_{\tau } < 10^4$; ${\blacksquare}$ (blue), $10^4 < {Re}_{\tau } < 5.10^4$; $\bullet$ (red), $5.10^4 < {Re}_{\tau } < 2.10^5$; ${\blacklozenge}$ (purple), $2.10^5 < {Re}_{\tau } < 5.3\,10^5$. (a) Graph of $\varXi (Y)$: — (green), linear fit $(1/0.433) + 2.5Y$. (b) Blowup of $\varXi$ versus $y^+$, with linear fits $[(1/0.433) + 2.5\,y^+/{Re}_{\tau }]$ for ${Re}_{\tau } = 2000$, 25 000 and 250 000. (c) Graph of $U^+(Y) - [(1/0.433)\ln {Re}_{\tau } + 2.5Y]$; - - - (green), resulting log law $(1/0.433)\ln Y + 5.8$.

Figure 6

Figure 7.The ZPG TBL Indicator function $\varXi (Y)$ (top) and $U^+(Y)$ minus log law $[(1/0.384)\ln y^+ + 4.17]$ (bottom): $\bullet$ (yellow, dark yellow, red, dark red), data of Samie et al. (2018) for ${Re}_{\tau } = 6$, 10, 14.5 and $20 \times 10^3$; $\bullet$ (light blue, blue, dark blue), data of Österlund (1999) for ${Re}_{\tau } = 5.5$, 6.6 and $7.9 \times 10^3$; $\bullet \bullet \bullet$ (increasingly dark green), data of Nagib et al. (2007) for ${Re}_{\tau } = 12.6$, 16 and $22.5 \times 10^3$ (for this last set, the log law constant has been increased from 4.17 to 4.32). Fits: — (light blue), $\varXi = 1/0.384$; — (light green), linear part $\varXi = (1/0.384) + 7.7(Y-0.11)$ for $0.11 \leqq Y \lessapprox 0.45$; — (lavender), full fit of $\varXi$ (4.1), (4.2); - - - (light green), fit $7.7[Y - 0.11 - 0.11\ln (Y/0.11)]$ corresponding to the linear part of $\varXi$. - - - (lavender), full fit of $U^+$ – log law (numerical integration of (4.1), (4.2)).

Figure 7

Figure 8.Dependence of the overlap parameters $\kappa$ (a) and the slope of the linear term $S_0$ (b) on the pressure gradient parameter $\beta \equiv -(\hat {\mathcal {L}}/\widehat {\tau _w}) (\mathrm {d} \hat {p}/\mathrm {d} \hat {x})$, equal to 1 and 2 for channel and pipe. Blue vertical bars, range of values deduced from the DNS of figures 3 and 5; red ${\blacksquare}$, baseline fits of the experimental data of figures 7 (ZPG TBL), 4 (channel) and 6 (pipe) with uncertainty estimates elaborated in the supplementary material. - - - (grey), tentative linear fits.

Figure 8

Figure 9.Contributions of order $O({Re}_{\tau }^{-1})$ to the channel $\mathrm {d} U^+/\mathrm {d} Y$, taken from figure 4(b) of Monkewitz (2021): $\bullet \bullet \bullet$ (blue), new fit (2.1), (2.2a–d); $\cdots$ (black), previous fit in Monkewitz (2021).

Figure 9

Figure 10.The DNS mean velocity for the channel DNS of figure 2, minus the outer fit (2.3), (2.4). Same colour scheme as in figure 2.

Figure 10

Figure 11.Grid spacing $\Delta y^+$ of different channel DNS (right vertical axis) versus $y^+$ compared with two indicator functions of Yamamoto & Tsuji (2018) (left axis). Left axis: — (orange), $\varXi ({Re}_{\tau } = 3986)$; $\cdot \cdot \cdot$ (light brown), linear fit $(1/0.424) + 1.2Y$; - - - (light brown), fit $(1/0.395) + 0.65Y$; $\bullet$ (light brown), switch between the two linear fits at $y^+ \approxeq 1300$. — (red), $\varXi ({Re}_{\tau } = 8000)$; $\cdot \cdot \cdot$ (dark red), linear fit $(1/0.424) + 1.3Y$; - - - (dark red), fit $(1/0.405) + 0.85Y$; $\bullet$ (dark red), switch between the two linear fits at $y^+ \approxeq 1900$. Right axis: grid spacing for ${Re}_{\tau } = 3986$ (thick orange dashes) and 8000 (thick red dashes) of Yamamoto & Tsuji (2018) with location of the change of linear slope of $\varXi$ indicated by bullets. Comparison grid spacings shown for ${Re}_{\tau }$ = 934 (del Alamo et al.2004, short green dashes), 1001 (Lee & Moser 2015, long green dashes), 1995 (Lee & Moser 2015, short blue dashes), 2004 (Hoyas & Jiménez 2006, long blue dashes), 4079 (Bernardini et al.2014, long violet dashes), 4179 (Lozano-Durán & Jiménez 2014), 5186 (Lee & Moser 2015, pink solid line), 10046 (Hoyas et al.2022, pink dashes).

Figure 11

Figure 12.Detail of $\varXi (Y)$ [(red) —] for the pipe DNS of Yao et al. (2023, figure 5) at ${Re}_{\tau } = 5197$. Lines: (black) $\cdots$, uncertainty estimates given by Yao et al.; (pink) $\bullet \bullet \bullet$, overall fit $(1/0.425) + 2.75Y$ of figure 5; (green) - - -, best fit $(1/0.408) + 2.34Y$ for $Y\in [0.1, 0.29]$; (blue) $-\cdot -$, best fit $(1/0.443) + 3.00Y$ for $Y\in [0.29, 0.45]$.

Supplementary material: PDF

Monkewitz and Nagib supplementary material

Monkewitz and Nagib supplementary material

Download Monkewitz and Nagib supplementary material(PDF)
PDF 2 MB