betaFeedback

WorksPhysics and Astronomy

Reynolds number asymptotics of wall-turbulence fluctuations

Abstract

In furtherance of our earlier work (Chen & Sreenivasan, J. Fluid Mech. 908, 2021, p. R3; 933, 2022, p. A20 – together referred to as CS hereafter), we present a self-consistent Reynolds number asymptotics for wall-normal profiles of variances of streamwise and spanwise velocity fluctuations as well as root-mean-square pressure, across the entire flow region of channel and pipe flows and flat-plate boundary layers. It is first shown that, when normalized by peak values, the Reynolds number dependence and wall-normal variation of all three profiles can be decoupled, in excellent agreement with available data, sharing the common inner expansion of the type ϕ+(y+)=f0(y+)+f1(y+)/Reτ1/4ϕ^+(y^+)=f_0(y^+)+f_1(y^+)/Re^{1/4}_τ, where ϕ+ϕ^+ is one of the quantities just mentioned, and the functions f0f_0 and f1f_1 depend only on y+y^+. Here, the superscript ++ indicates normalization by wall variables. Secondly, by matching the above inner expansion and the inviscid outer flow similarity form, a bounded variation ϕ+(y∗)=αϕ−βϕy∗1/4ϕ^+(y^\ast)=α_ϕ-β_ϕy^{\ast{{1}/{4}}} is derived beyond the peak where, for each ϕ+ϕ^+, the constants αϕα_ϕ and βϕβ_ϕ are independent of ReτRe_τ and y∗y^\ast (=y+/Reτ=y^+/Re_τ, ReτRe_τ being the Reynolds number based on the friction velocity) – also in excellent agreement with simulations and experimental data. One of the predictions of the analysis is that, for asymptotically high Reynolds numbers, a finite plateau ϕ+≈αϕϕ^+\approxα_ϕ appears in the outer region. This result sheds light on the intriguing issue of the outer shoulder of the variance of the streamwise velocity fluctuation, which should be bounded by the asymptotic plateau of about 10.

Introduction

The Reynolds number dependencies of the variances of streamwise and spanwise velocity fluctuations as well as pressure fluctuations are thought to present exceptional challenges for the classical notion of wall scaling [19, 43]. A salient example is that, when scaled in wall units, the peak values of these quantities near the wall grow with increasing Reynolds number (though the peak locations are remarkably invariant; see, e.g., Sreenivasan [45]). In CS, the growth of these peaks was cast as a finite Reynolds number effect, and it was shown that a bounded growth model (discussed below) fits the data better. In this paper, we turn attention to wall-normal profiles of the variances of these fluctuations. This work is an alternative to the attached-eddy hypothesis by Townsend [46], which ascribes a logarithmic decay for fluctuations in the outer flow as

ϕ+(y∗)=Bϕ−Aϕln⁡y∗.(1)\phi^{+}(y^{*}) = B_{\phi} - A_{\phi}\ln y^{*}. \tag*{(1)}

Here, ϕ+\phi^{+} represents the variance of ⟨uw⟩+\langle uw\rangle^{+} or ⟨vw⟩+\langle vw\rangle^{+}; the superscript + indicates normalization by uτu_{\tau} and ν\nu, and u,v,wu,v,w for fluctuation velocities in the streamwise (xx), wall-normal (yy) and spanwise or azimuthal (zz) directions; y∗=y/δy^{*}=y/\delta where δ\delta is the flow thickness; the slope AϕA_{\phi} and intercept BϕB_{\phi} are constants independent of y∗y^{*} and the friction Reynolds number Reτ=uτδ/νRe_\tau= u_\tau\delta/\nu, where uτ≡(τw/ρ)1/2u_\tau\equiv(\tau_w/\rho)^{1/2} is the friction velocity, but may depend on ϕ\phi.

The rationale behind (1), as discussed by Marusic & Monty [20], is that the number density of the attached eddies that contribute to turbulent fluctuations varies inversely with y∗y^*, and an integration with respect to y∗y^* leads to the total fluctuation intensity given by (1). Some consequences of this idea have been explored in laboratory measurements (EXP) (Metzger & Klewicki [24]; Hultmark et al. [12]; Vincenti et al. [50]; Willert et al. [51]; Samie et al. [35]; Ono et al. [31]) as well as direct numerical simulations (DNS) (Wu & Moin 2009; Schlatter & Orlü [37]; Jimenez et al. [14]; Lee & Moser [17]; Pirozzoli et al. [33]; Hoyas et al. [10]; Yao et al. [56]). The resulting findings have been discussed in terms of mixed scaling (DeGraaff & Eaton [6]), multi-regime of the power-law spectrum (Vassilicos et al. [49]; Diaz-Daniel et al. [7]), inner-outer interactions (Marusic et al. [18]) and random addictive process (Yang & Lozano-Durán [54]). The notion of attached eddies has been extended to study high-order moments of single point velocity fluctuations (Meneveau & Marusic [22]) as well as to velocity structure functions (de Silva et al. [41]).

Pressure fluctuations have also received attention in the past sixty years (Bradshaw [1]; Klewicki et al. [16]; Panton et al. [32]), in part because of their importance for aircraft cabin noise. By extending Townsend’s attached-eddy hypothesis, Bradshaw [1] obtained a k−1k^{-1} spectrum by an inner-outer matching in wavenumber space and hence an ln⁡Reτ\ln Re_\tau growth of wall pressure fluctuation. The k−1k^{-1} spectrum so deduced is marginally detected in laboratory boundary layers at Reτ≈6000Re_\tau\approx6000 (Tsuji et al. [47]), but not in the DNS data so far. This unsatisfactory situation prompted Panton et al. [32] to develop alternative matching analysis in the spatial domain, also yielding the log-profile of (1). This is reminiscent of Hultmark [11] who derived the ln⁡y∗\ln y^* variation in pipes by matching ⟨uu⟩+\langle uu\rangle^+ between the inner and outer regions. It is worth noting that the ReτRe_\tau effects included in these models are not part of Townsend’s original attached-eddy hypothesis.

While the above works suggest a boundless growth of turbulence peaks as Reτ→∞Re_\tau\to\infty, CS argued that the observed variations are bounded at very high Reynolds numbers and follow a defect law of the type

ϕp+(Reτ)=ϕ∞+−cϕ,∞Reτ−1/4.(2)\phi_p^+(Re_\tau) = \phi_\infty^+ - c_{\phi,\infty}Re_\tau^{-1/4}. \tag*{(2)}

Here, ϕ∞+\phi_\infty^+ the asymptotically bounded value of ϕp+(Reτ)\phi_p^+(Re_\tau) and cϕ,∞c_{\phi,\infty} are the fixed coefficients. The underlying physics depends on the slight imbalance that exists between wall dissipation and maximum production in the turbulent energy budget at any finite Reynolds number, and on their tendency to eventually balance each other. Subsequently, Monkewitz [26] showed that an asymptotic expansion of ⟨uu⟩+\langle uu\rangle^+ profiles with the Reτ−1/4Re_\tau^{-1/4} gauge function from CS reproduced data better than ln⁡Reτ\ln Re_\tau (Smits et al. [42]). Recent measurements of Ono et al. [31] in pipes for ReτRe_\tau ranging from 990 to 20750 are also supportive of the bounded behavior. The results of CS have been checked against DNS data in the open channel (Yao et al. [55]) and compressible channel (Gerolymos & Vallet [8]), indicating the universality of the bounded behavior for different flow conditions. Indeed, Hoyas et al. [10] reported that the wall pressure fluctuation in their DNS channel data for ReτRe_\tau up to 10410^4 might also be bounded.

Yet, to differentiate between the two sets of results beyond doubt, one clearly requires much higher Reynolds numbers than currently covered (or likely to be covered for the foreseeable future) in laboratory experiments or DNS (Nagib et al. [29]). Measurements in the atmospheric boundary layer (Metzger & Klewicki [24]; Metzger et al. [23]) might be thought of as helpful but various uncertainties characteristic of field measurements prevent a decisive conclusion there also. At the current stage, new theoretical ideas are highly desired to provide additional insights. As noted by Klewicki [15], the bounded growth, once accepted, would necessitate a reassessment of a number of earlier empirical findings. In this spirit, we obtain an alternative to (1.1) by using the bounded behavior of (1.2), providing a more complete description of the asymptotic behavior of wall turbulence (including pressure).

Specifically, we first decouple the ReτRe_\tau dependence from the wall-normal variation for the root-mean-square (rms) profiles by using peak values for normalization. Then we develop a matching procedure between the inner viscous and outer inviscid regions, yielding in the outer flow a defect law of the type

ϕ+(y∗)=αϕ−βϕy∗1/4.(3)\phi^{+}(y^{*}) = \alpha_{\phi} - \beta_{\phi} y^{*1/4}. \tag*{(3)}

Here, ϕ+\phi^{+} represents not only ⟨uu⟩+\langle uu\rangle^{+} and ⟨ww⟩+\langle ww\rangle^{+} but also the rms of pressure fluctuation p′+p^{\prime+} (superscript prime denotes the rms).

To verify (1.3), DNS data sets are collected for those with clear ReτRe_\tau trend for ⟨uu⟩+\langle uu\rangle^{+}, ⟨ww⟩+\langle ww\rangle^{+} and p′+p^{\prime+}, all publicly available. In particular, we use the DNS on channels by Lee & Moser [17] for ReτRe_\tau from 550 to 5200, on pipes by Pirozzoli et al. [33] for ReτRe_\tau from 500 to 6000, and on TBLs by Schlatter et al. [38, 36] for ReτRe_\tau from 490 to 1270. Higher ReτRe_\tau data in the literature ([40, 10]) are also included for comparison. For experiments, we select ⟨uu⟩+\langle uu\rangle^{+} data from the Princeton pipe by Hultmark et al. [12] for ReτRe_\tau from 5411 to 98187, from the Princeton TBLs by Vallikivi et al. [48] for ReτRe_\tau from 4635 to 25062, and from the Melbourne TBLs by Samie et al. [35] for ReτRe_\tau from 6000 to 20000. Channel experiments are not collected here because their (limited) ReτRe_\tau variation has been covered by the DNS of Lee & Moser [17]. Note that the data uncertainty, especially concerning the probe resolution in experiments and grid resolution in the DNS, are not addressed in this paper (see CS for a brief discussion). We do wish to state, however, that there is much need for better-resolved data.

The paper is organized as follows. Section 2 presents data collapse for the inner flow region, which leads to the uniform expansion scheme presented there. Section 3 begins with the verification of inviscid similarity in the outer region, followed by the derivation of the defect decay, using comprehensive data comparisons. Section 4 is devoted to a discussion of the geometry effect. A perspective and summary of the results are given in section 5.

ReτRe_\tau-scaling for near wall region

When scaled by wall variables, an asymptotic expansion for ⟨uu⟩+\langle uu\rangle^{+}, ⟨ww⟩+\langle ww\rangle^{+}, and p′+p^{\prime+}, represented by ϕ+\phi^{+}, can be written as

ϕ+(y+,Reτ)=f0(y+)+f1(y+)g(Reτ)+f2(y+)g2(Reτ)+h.o.t.,(4)\phi^{+}(y^{+},Re_\tau) = f_{0}(y^{+}) + f_{1}(y^{+})g(Re_\tau) + f_{2}(y^{+})g^{2}(Re_\tau) + h.o.t., \tag*{(4)}

where gg is the gauge function of ReτRe_\tau; f0f_{0}, f1f_{1} and f2f_{2} (as well as ff introduced below in (2.2)) are general functions depending merely on the wall-normal distance y+y^{+}, and h.o.t.h.o.t. indicates high order terms. For the streamwise mean velocity ϕ+=U+\phi^{+}=U^{+}, a first order truncation of (2.1) is fairly accurate near the wall, as already remarked, for turbulent fluctuations, the wall scaling has received challenges, reflected in the notable ReτRe_\tau-dependence for the near wall ⟨uu⟩+\langle uu\rangle^{+} and ⟨ww⟩+\langle ww\rangle^{+} as well as pressure fluctuation p′+p^{\prime+}. Note that ⟨uv⟩+\langle uv\rangle^{+} and ⟨vv⟩+\langle vv\rangle^{+} are thought as ‘active’ motions by Townsend [46], with bounded maximum values, with the latest evidence being given, for example, in [42, 55]. In the rest of this section, we first show data collapse of ⟨uu⟩+\langle uu\rangle^{+}, ⟨ww⟩+\langle ww\rangle^{+} and p′+p^{\prime+} after normalization by their corresponding peak values, and then summarize a common expansion for these quantities, which is actually a second-order truncation of (4).

Data collapse for the inner flow region

Figure 1 shows the profiles of ⟨uu⟩+\langle uu\rangle^{+}, top panels for the channel, middle for the pipe and bottom for the TBL. While the left column displays marked ReτRe_{\tau} variations, the right column illustrates excellent data collapse after normalization by peak values. That is,

Six-panel plot of wall-normal streamwise velocity fluctuations for DNS channel, pipe, and TBL flows, with unnormalized and peak-normalized profiles

Figure 1. Wall-normal dependence of streamwise velocity fluctuation scaled in viscous units (abscissa in logarithmic scale) for a series of ReτRe_\tau’s in channels (top panels), pipes (middle panels) and TBL flows (bottom panels). Left column: ⟨uu⟩+\langle uu\rangle^+ versus y+y^+. Right column: ⟨uu⟩+\langle uu\rangle^+ normalized by its (inner) peak value ⟨uu⟩p+\langle uu\rangle_p^+, showing very good collapse. Colored lines indicate DNS data at different ReτRe_\tau’s marked in the figure legends, for channels by Lee & Moser [17], for pipes by Pirozzoli et al. [33], and for TBLs by Schlatter et al. [38, 36].

⟨uu⟩+(y+,Reτ)=⟨uu⟩p+(Reτ)f(y+).(5)\langle uu\rangle^{+}(y^{+},Re_{\tau})=\langle uu\rangle_{p}^{+}(Re_{\tau})f(y^{+}). \tag*{(5)}

Note that according to CS, the peak location is an invariant at yp+≈15y_{p}^{+}\approx15. This is generally accepted as correct (at least since Sreenivasan [45]); see Smits et al. [42]. On the other hand, Pirozzoli et al. [33] commented that the invariant peak location in CS is violated by their pipe data, which shows that yp+y_{p}^{+} slightly increases from 14.28 at Reτ≈500Re_{\tau}\approx500 to 15.14 at Reτ=6000Re_{\tau}=6000. Nevertheless, using a finer near-wall resolutions than in Pirozzoli et al. [33], a later study by Yao et al. [56] found no such variation of yp+y_{p}^{+} with ReτRe_{\tau} (yp+=15.07,15.03,15.50y_{p}^{+}=15.07,15.03,15.50 for Reτ=180,2000,5000Re_{\tau}=180,2000,5000, respectively). This range of variation is typically seen by others as well ([28, 14, 5]), but is regarded as small here, owing possibly to secondary reasons such as the grid and probe resolution.

Similar to (5), data collapse for the spanwise velocity fluctuation is achieved via

⟨ww⟩+(y+,Reτ)=⟨ww⟩p+(Reτ)h(y+),(6)\langle ww\rangle^{+}(y^{+},Re_{\tau})=\langle ww\rangle_{p}^{+}(Re_{\tau})h(y^{+}), \tag*{(6)}

where ⟨ww⟩p+\langle ww\rangle_{p}^{+} is the peak value, and hh is a y+y^{+}-dependent function. As shown in Figure 2, different ReτRe_{\tau} curves are in close agreement with each other, with the self-preserving range from the wall to the peak (at y+≈45y^{+}\approx45) or beyond. We note a marginal ReτRe_{\tau} dependence on this peak location; it is not clear whether they arise from numerical uncertainty or physical modulation by outer flow structures.

Six plots of spanwise velocity fluctuations, arranged in three rows and two columns

Figure 2. The same plots as in figure 1 but for the spanwise velocity fluctuations. That is, top panels are for channels, middle for pipes and bottom for TBLs; left column for ⟨ww⟩+\langle ww\rangle^{+} while right for ⟨ww⟩+/⟨ww⟩p+\langle ww\rangle^{+}/\langle ww\rangle_{p}^{+} versus y+y^{+}. Lines represent the same DNS data as in figure 1.

Coming now to pressure fluctuations, the left column of Figure 3 shows ReτRe_{\tau} dependence of p′+p^{\prime+}. The best collapse is obtained by plotting pp′+−p′+p_{p}^{\prime+}-p^{\prime+}, as shown in the right column. On this basis, we may write

Wall-normal dependence plots for pressure fluctuations in channels, pipes and TBL flows

Figure 3. Wall-normal dependence for the rms of pressure fluctuation p′+=⟨pp⟩+1/2p^{\prime+} = \langle pp\rangle^{+1/2} in channels (top panels), pipes (middle panels) and TBL flows (bottom panels). Left column for p′+p^{\prime+} while right for pp′+−p′+p_{p}^{\prime+} - p^{\prime+} versus y+y^{+}. Lines are the same DNS data as in figure 1.

p′+(y+,Reτ)=pp′+(Reτ)−j(y+),(7)p^{\prime+}(y^{+},Re_{\tau})=p_{p}^{\prime+}(Re_{\tau})-j(y^{+}), \tag*{(7)}

where jj is (in general) a y+y^{+}-dependent function. The collapse extends from wall to the peak (at y+≈30y^{+}\approx30), with pp′+−pw′+p_{p}^{\prime+}-p_{w}^{\prime+} a constant around 0.4. This constancy inspires us to postulate (7). From (7) one has p′+/pp′+=1−j(y+)/pp′+(Reτ)p^{\prime+}/p_{p}^{\prime+}=1-j(y^{+})/p_{p}^{\prime+}(Re_{\tau}), in which an increasing pp′+p_{p}^{\prime+} with ReτRe_{\tau} would eventually spoil the data collapse if one plotted p′+/pp′+p^{\prime+}/p_{p}^{\prime+}. That is the reason why (5) or (6) is not applied to p′+p^{\prime+}. Note that Panton et al. [32] attempted another data collapse by using ⟨pp⟩+−⟨pp⟩p+\langle pp\rangle^{+}-\langle pp\rangle_{p}^{+}, but it is not as satisfactory as (7) in Figure 3, as discussed later in section 3.

Summary for the near wall scaling

The above comparisons demonstrate that the ReτRe_{\tau} and y+y^{+} dependencies could be decoupled after a proper normalization by peak values. Recalling (1.2) for the ReτRe_{\tau}-scaling of the peak values and substituting it for ⟨uu⟩p+\langle uu\rangle_{p}^{+}, ⟨ww⟩p+\langle ww\rangle_{p}^{+} and pp′+p_{p}^{\prime+} into (5), (6) and (7), respectively, one has a uniform expansion

ϕ(y+)=ϕ0(y+)+ϕ1(y+)/Reτ1/4,(8)\phi(y^{+})=\phi_{0}(y^{+})+\phi_{1}(y^{+})/Re_{\tau}^{1/4}, \tag*{(8)}

where ϕ0(y+)=ϕ∞+f(y+)\phi_{0}(y^{+})=\phi_{\infty}^{+}f(y^{+}) and ϕ1(y+)=−cϕ,∞f(y+)\phi_{1}(y^{+})=-c_{\phi,\infty}f(y^{+}) for uu; ϕ0(y+)=ϕ∞+h(y+)\phi_{0}(y^{+})=\phi_{\infty}^{+}h(y^{+}) and ϕ1(y+)=−cϕ,∞h(y+)\phi_{1}(y^{+})=-c_{\phi,\infty}h(y^{+}) for ww; ϕ0(y+)=ϕ∞−j(y+)\phi_{0}(y^{+})=\phi_{\infty}-j(y^{+}) and ϕ1(y+)=−cϕ,∞\phi_{1}(y^{+})=-c_{\phi,\infty} for pressure fluctuations.

Note that (2.5) is a specific case of

ϕ+(y+,Reτ)=f0(y+)+f1(y+)g(Reτ),(9)\phi^{+}(y^{+},Re_{\tau})=f_{0}(y^{+})+f_{1}(y^{+})g(Re_{\tau}), \tag*{(9)}

which is a second-order truncation of (2.1) that was initiated first by Spalart & Abe [44] and Monkewitz [26]. If f0=0f_{0}=0, the above (2.6) reduces to

ϕ+(y+,Reτ)=f1(y+)g(Reτ),(10)\phi^{+}(y^{+},Re_{\tau})=f_{1}(y^{+})g(Re_{\tau}), \tag*{(10)}

which is the scaling proposed by Smits et al. [42] for ⟨uu⟩+\langle uu\rangle^{+} with g(Reτ)=ϵx−w+g(Re_{\tau})=\epsilon_{x-w}^{+} (streamwise wall dissipation). However, Smits et al. [42] found that their proposal did not work as well for ⟨ww⟩+\langle ww\rangle^{+}, which they speculated was due to different superposition and modulation enforced by outer flow structures. Here, we show that replacing wall dissipation by peak value, i.e. g=ϕp+g=\phi_p^+, (10) applies for both ⟨uu⟩+\langle uu\rangle^+ and ⟨ww⟩+\langle ww\rangle^+. Even so, (10) is not proper for pressure due to a constancy pt′+−pw′+p_t^{\prime+}-p_w^{\prime+} as explained earlier. Thus, a nonzero f0f_0 is needed in (9) when taking pressure into consideration, which is missed in Smits et al. [42].

Finally, we recall that from (9), Monkewitz [26] developed a composite model for ⟨uu⟩+\langle uu\rangle^+, which shows that g=Reτ−1/4g=Re_\tau^{-1/4} yields a better data description than the alternative g=ln⁡Reτg=\ln Re_\tau by Smits et al. [42]. The gauge function g=Reτ−1/4g=Re_\tau^{-1/4} in (9) restores wall scaling for asymptotically high ReτRe_\tau, and we will use it below to derive an outer decay profile, which has not not achieved before.

Re-scaling for outer region and the defect law for fluctuations

Similar to (4) for the inner region, an asymptotic expansion for the outer flow reads

ϕ+(y∗,Reτ)=F0(y∗)+F1(y∗)G(Reτ)+F2(y∗)G2(Reτ)+h.o.t.,(11)\phi^+(y^*,Re_\tau)=F_0(y^*)+F_1(y^*)G(Re_\tau)+F_2(y^*)G^2(Re_\tau)+\text{h.o.t.}, \tag*{(11)}

where y∗=y/δ=y+/Reτy^*=y/\delta=y^+/Re_\tau is the outer unit; F0F_0, F1F_1 and F2F_2 are general functions depending on y∗y^*, and G(Reτ)G(Re_\tau) is the gauge function. In analogy to law of the wall, the inviscid outer flow similarity corresponds to the first order truncation in (11), i.e.

ϕ+(y∗,Reτ)=F0(y∗).(12)\phi^+(y^*,Re_\tau)=F_0(y^*). \tag*{(12)}

When we take ϕ+=Ue+−U+\phi^+=U_e^+-U^+, the resulting equation for the mean velocity is known as the velocity defect law. Here, the subscript ee indicates the value at y=δy=\delta or y∗=1y^*=1, i.e., the centerline for channel and pipe flows, and the boundary layer edge for TBL. This law has been tested extensively in the literature—see Nagib et al. [30] and She et al. [39] for recent efforts. We shall now develop the equation for the fluctuating quantities.

An assessment of outer similarity for the three profiles is shown in Figure 4, with the top panels for the channel, middle for the pipe and bottom for the TBL. It is remarkable that the pressure fluctuation displays an excellent data collapse from y∗=0.1y^*=0.1 to y∗=1y^*=1. In the same flow region, the spanwise velocity variance data also collapse well, with a deviation of the order of 0.1 (scaled on uτ2u_\tau^2). For streamwise velocity, the outer similarity holds better towards the outer edge though discernible ReτRe_\tau dependence exists for small ReτRe_\tau profiles. For Reτ>1000Re_\tau>1000, the streamwise variance profiles also collapse together closely (Figure 5-Figure 6), thus supporting the inviscid similarity of (12) at high ReτRe_\tau.

Wall-normal dependence of turbulence fluctuations in outer length unit

Figure 4. Wall-normal dependence of turbulence fluctuations in outer length unit y∗=y/δy^{*}=y/\delta (the abscissa in linear scale). Top panels for channels, middle for pipes and bottom for TBLs. Left column for ⟨uu⟩+\langle uu\rangle^{+}; middle for ⟨ww⟩+\langle ww\rangle^{+} and right for p′+p^{\prime+}. Lines are the same DNS data as in figure 1.

Variance plots of streamwise velocity fluctuation for DNS channel, pipe, and turbulent boundary layer data

Figure 5. Variance of streamwise velocity fluctuation of DNS data at high ReτRe_{\tau}. Channel data in top panels are Reτ=2000,5200Re_{\tau}=2000, 5200 from Lee & Moser [17], Reτ=104Re_{\tau}=10^{4} from Hoyas et al. [10]. Pipe data in middle panels from Pirozzoli et al. [33]. Bottom panels for TBL, Reτ=1270Re_{\tau}=1270 from Schlatter & Örlü [37], Reτ=1570,1990Re_{\tau}=1570,1990 from Sillero et al. [40]. Left column is for abscissa in linear and the right in logarithmic outer units. Dotted (green) line in top panels indicate 1.61−1.25ln⁡(y∗)1.61-1.25\ln(y^{*}) by Hultmark et al. [12]; 2.2−1.26ln⁡(y∗)2.2-1.26\ln(y^{*}) in middle by Marusic et al. [21] and 1.95−1.26ln⁡(y∗)1.95-1.26\ln(y^{*}) in the bottom by Samie et al. [35]. Dashed (red) lines indicate the present relation (3), i.e. 10−9.3(y∗)1/410-9.3(y^{*})^{1/4}, the same for all the flows here.

Variance of streamwise velocity fluctuation in high $Re_\tau$ experiments

Figure 6. Variance of streamwise velocity fluctuation in high ReτRe_\tau experiments. Data in top panels from Princeton pipe [12], middle from Princeton TBL [48], and bottom from Melbourne TBL [35]. Dotted (green) line in top panels indicate 1.61−1.25ln⁡(y∗)1.61 - 1.25\ln(y^*) by Hultmark et al. [12]; 2.2−1.26ln⁡(y∗)2.2 - 1.26\ln(y^*) in middle by Marusic et al. [21] and 1.95−1.26ln⁡(y∗)1.95 - 1.26\ln(y^*) in the bottom by Samie et al. [35]. Dashed (red) lines indicate (3), i.e. 10−9.3(y∗)1/410 - 9.3(y^*)^{1/4} in all the figures.

Matching for the defect law for fluctuations

Based on the above inner viscous expansion and outer inviscid similarity, we now develop a matching procedure to derive the analytical form in the intermediate zone. To start with, one may invoke the derivation of Millikan [25] for the log-law, which has been extended (for example) for ⟨uu⟩+\langle uu\rangle^+ by Hultmark [11] and for ⟨pp⟩+\langle pp\rangle^+ by Panton et al. [32]. Nevertheless, different orders of matching would lead to different scaling proposals, and here we present a short account of matching to obtain the defect law for fluctuations. Later we will address the question of how the current defect law is consistent with Millikan’s matching analysis.

Specifically, as (9) applies near the wall (with g=Reτ−1/4g=Re_\tau^{-1/4}) and (12) holds in the outer, we match (12) directly with (9) in the outer region. That is,

F0(y∗)=ϕ+(y+)=f0(y+)+f1(y+)/Reτ1/4=f0(y+)+h1(y+)y∗1/4(13)\begin{aligned} F_0(y^*)&=\phi^+(y^+)=f_0(y^+)+f_1(y^+)/Re_\tau^{1/4}\\ &=f_0(y^+)+h_1(y^+)y^{*1/4} \tag*{(13)} \end{aligned}

with h1(y+)=f1(y+)/y+1/4h_1(y^{+})=f_1(y^{+})/y^{+1/4}. Note that towards the wall, f0=ϕ0f_0=\phi_0 and h1=ϕ1/y+1/4h_1=\phi_1/y^{+1/4} so that (13) approaches (8); f0f_0 and h1h_1 are to be determined towards the outer region. At this stage, assuming the scale-separation between y∗y^{*} and y+y^{+} for asymptotically high ReτRe_{\tau}, one has ∂F0(y∗)/∂y+=0\partial F_0(y^{*})/\partial y^{+}=0 in (13), which leads to

f0(y+)=c0;(14)f_0(y^{+})=c_0; \tag*{(14)}
h1(y+)=c1,(15)h_1(y^{+})=c_1, \tag*{(15)}

where c0c_0 and c1c_1 are constants independent of y∗y^{*}, y+y^{+} and ReτRe_{\tau}, but may depend on the precise quantity ϕ\phi. Denoting c0=αϕc_0=\alpha_{\phi} and c1=−βϕc_1=-\beta_{\phi}, we obtain (3) from (13) and (3.4).

It is readily verified that (3) matches (12) in the outer region and (9) in the inner, hence offering a common description for the overlap region. It should be mentioned that a logarithmic decay of (1) can also be obtained if the gauge function Reτ−1/4Re_{\tau}^{-1/4} in (13) is replaced by ln⁡Reτ\ln Re_{\tau}. In other words, mere matching analysis cannot preclude (3) or (1), unless the gauge function is justified rigorously.

Before turning to the data in evidence for (3), we discuss how the defect law conforms with the classical matching analysis. In fact, by following the matching of Millikan [25], as discussed for example in Pope [34], one has

∂ϕ+∂ln⁡y∗=A0+A1(Reτ)+h.o.t.,(16)\frac{\partial\phi^{+}}{\partial\ln y^{*}} = A_{0} + A_{1}(Re_{\tau}) + h.o.t., \tag*{(16)}

where A0A_{0} and A1A_{1} are coefficients independent of y+y^{+} or y∗y^{*}. Note that A1A_{1} involving the ReτRe_{\tau}-dependence is a higher-order modification compared to the leading order coefficient A0A_{0}, which has been introduced for the streamwise mean velocity by Wosnik et al. [52] and Monkewitz & Nagib [27]. Here, particularly for A0<0A_{0}<0, (16) yields the logarithmic decay, e.g. for ⟨uu⟩+\langle uu\rangle^{+} as in Hultmark [11], and for ⟨pp⟩+\langle pp\rangle^{+} as in Panton et al. [32].

On the other hand, if A0=0A_{0}=0 so that A1(Reτ)=o(1)A_{1}(Re_{\tau})=o(1) dominates (16), a wall-normal integration of (16) leads to

ϕ+(y∗)=A1(Reτ)ln⁡(y∗)+B0.(17)\phi^{+}(y^{*}) = A_{1}(Re_{\tau})\ln(y^{*}) + B_{0}. \tag*{(17)}

Further, with A1(Reτ)=O(1/ln⁡Reτ)A_{1}(Re_{\tau})=O(1/\ln Re_{\tau}), one has

ϕ+(ys+)=A1(Reτ)ln⁡ys+−A1(Reτ)ln⁡(Reτ)+B0=O(1),(18)\phi^{+}(y_{s}^{+}) = A_{1}(Re_{\tau})\ln y_{s}^{+} - A_{1}(Re_{\tau})\ln(Re_{\tau}) + B_{0} = O(1), \tag*{(18)}

which means that ϕ+\phi^{+} is bounded at any specific ys+y_{s}^{+} location. Conversely, a bounded ϕ+\phi^{+} would exclude a constant logarithmic slope but require the slope to decrease with ReτRe_{\tau} according to O(1/ln⁡Reτ)O(1/\ln Re_{\tau}).

One can easily check that (3) falls into this category of (18) as follows. Calculate the local logarithmic slope of (3) yielding

Alocal(y∗)≡∂ϕ+∂ln⁡y∗=(βϕ/4)(y∗)1/4=(βϕ/4)(y+)1/4Reτ−1/4,(19)A_{local}(y^{*}) \equiv\frac{\partial\phi^{+}}{\partial\ln y^{*}} = (\beta_{\phi}/4)(y^{*})^{1/4} = (\beta_{\phi}/4)(y^{+})^{1/4}Re_{\tau}^{-1/4}, \tag*{(19)}

which agrees with (18) because Alocal∝Reτ−1/4=O(1/ln⁡Reτ)A_{local}\propto Re_{\tau}^{-1/4}=O(1/\ln Re_{\tau}) at any given position ys+y_{s}^{+}. Moreover, as y∗y^{*} increases, AlocalA_{local} in (19) becomes larger, indicating a steeper logarithmic slope towards the outer flow. This picture is indeed supported by data in the comparison below, in contrast to (1) with a constant slope valid in a narrower flow domain.

Comparison with data

Figures 5-8 show the data comparisons with (1) and (3) for ⟨uu⟩+\langle uu\rangle^{+}, ⟨ww⟩+\langle ww\rangle^{+} and p′+p^{\prime+}, respectively. Table 1 collects all the parameters for the three profiles, arising from these fits to the data, which will now be discussed in greater detail.

Quantity ϕ\phiαϕ(CH)\alpha_{\phi}^{(CH)}αϕ(Pipe)\alpha_{\phi}^{(Pipe)}αϕ(TBL)\alpha_{\phi}^{(TBL)}βϕ(CH)\beta_{\phi}^{(CH)}βϕ(Pipe)\beta_{\phi}^{(Pipe)}βϕ(TBL)\beta_{\phi}^{(TBL)}
⟨uu⟩+\langle uu\rangle^{+}1010109.39.39.3
⟨ww⟩+\langle ww\rangle^{+}3.93.93.63.453.452.6
p′+p^{\prime+}4.844.64.654.13.73.15

Table 1. Parameters in (1.3), for different fluctuations. Superscripts ‘CH’, ‘Pipe’ and ‘TBL’ represent channel, pipe and boundary layer flows, respectively. Note that both αϕ\alpha_{\phi} and βϕ\beta_{\phi} vary only modestly among different flows, implying that essentially the same mechanisms applies for all flows. Moreover, βϕ\beta_{\phi} is quite close to αϕ\alpha_{\phi}, as ϕ\phi at the boundary layer edge y∗=1y^{*}=1 is fairly small.

Similar to figures 1-3, top panels of Figure 5-Figure 8 are for channel, middle for pipe and bottom for boundary layer flows; the left column is for abscissa that are in linear units while right are for logarithmic units. A difference from Figure 1-Figure 3 is that the data in Figure 5-Figure 8 are denoted by symbols with lines (black), so that (1.1) and (1.3) are better marked to guide the eye.

Figure 8. R.M.S. of pressure fluctuation in channel (top), pipe (middle) and TBL (bottom) flows scaled in the outer unit. Dashed (red) lines indicate the bounded decay (3) with parameters in table 1. Dotted (green) lines indicate the logarithmic decay in (20), i.e. 0.27−2.56ln⁡(y∗)\sqrt{0.27 - 2.56\ln(y^*)} for channel, 0.6−2.45ln⁡(y∗)\sqrt{0.6 - 2.45\ln(y^*)} for pipe and 2.5−2.45ln⁡(y∗)\sqrt{2.5 - 2.45\ln(y^*)} for TBL. Symbols with lines are the same DNS data as in Figure 7.

Particularly for ⟨uu⟩+\langle uu\rangle^{+}, to avoid distractions by data scatter at small ReτRe_{\tau}, we collect in Figure 5 only high ReτRe_{\tau} profiles from DNS, namely, ReτRe_{\tau} from 2000 to 10410^{4} for channel; 2000 to 6000 for pipe; and 1270 to 1990 for TBL. Compared to Figure 4a,Figure 4d,Figure 4g, it is clear in Figure 5a,Figure 5c,Figure 5e that all high ReτRe_{\tau} profiles collapse on each other in the flow range for 0.1≲y∗≲10.1 \lesssim y^{*} \lesssim1, thus bearing out the inviscid similarity. This is confirmed again by experimental data in Figure 6 corresponding to higher ReτRe_{\tau}.

Note that the logarithmic behavior advanced in the literature [12, 21, 35] is indicated by green dotted line. Although it characterizes data in the region from 0.1≲y∗≲0.30.1 \lesssim y^{*} \lesssim0.3, the value of the intercept BϕB_{\phi} needs to be adjusted for the three flows from 1.61 to 2.2, while AϕA_{\phi} holds constant around 1.26. In contrast, the red dashed line represents (1.3) with the same αϕ=10\alpha_{\phi}=10 and βϕ=9.3\beta_{\phi}=9.3, which reproduces data well for channel, pipe and TBL flows, covering not only the logarithmic range but also the so-called wake region, almost all the way to the centerline of channel and pipe flows.

One may imagine that the data in Figure 5 have not reached the asymptotic state and that (1.1) might agree with data better for higher ReτRe_{\tau}. But experimental data from Princeton pipes [12] with ReτRe_{\tau} covering one more decade, e.g. ReτRe_{\tau} from about 5000 to 10510^{5}, do not show any improvement of the fit to (1.1); see Figure 6a-Figure 6b. A similar observation is also true for the TBL, as shown in Figure 6c-Figure 6f.

Moreover, for ⟨ww⟩+\langle ww\rangle^{+} and p′+p^{\prime+}, (1.3) extends almost all the way to the centerline of channel and pipe flows. Here, data in Figure 7 and Figure 8 are the same DNS groups as in Figure 1 and contain those low ReτRe_{\tau} profiles. The agreement with (1.3) is excellent at the smallest Reτ≈500Re_{\tau}\approx500 for channel and pipe, in contrast to the log variation that agrees with data only for Reτ≳2000Re_{\tau}\gtrsim2000. Therefore, in both y∗y^{*} and ReτRe_{\tau} ranges, (1.3) covers a wider range than (1.1).

Variance of spanwise velocity fluctuation in channel, pipe and TBL flows

Figure 7. Variance of spanwise velocity fluctuation in channel (top), pipe (middle) and TBL (bottom) flows scaled in outer unit. Dashed (red) lines indicate the bounded decay (3) with parameters in Table 1. Dotted (green) lines indicate the logarithmic decay (1), i.e. 1.08 − 0.387 ln(y∗y^*) for channel and pipe, and 1.23 − 0.387 ln(y∗y^*) for TBL. Symbols with lines are the same DNS data as in Figure 1, i.e. channel from Lee & Moser [17], pipe from Pirozzoli et al. [33], TBL from Schlatter et al. [38, 36].

Three further points will —rednow be discussed. First, the difference between (1.3) and (1.1) is more vital for asymptotically high ReτRe_{\tau}. For (1.1), an infinitely large ϕ+∝ln⁡Reτ\phi^{+}\propto\ln Re_{\tau} would arise as y∗→0y^{*}\to0 and Reτ→∞Re_{\tau}\to\infty. In contrast, (1.3) assigns a plateau of ϕ+≈αϕ\phi^{+}\approx\alpha_{\phi} in the same limit. Such an asymptotic plateau implies that turbulent eddies in the bulk would be in a quasi-equilibrium state in the sense that their contribution to ϕ+\phi^{+} is invariant when y∗y^{*} changes. Mimicking Townsend’s terminology, the decreasing influence of attached eddies away from the wall would be compensated by the increasing influence of detached eddies, so that their total contribution to ϕ+\phi^{+} remains invariant. Note also that according to (1.3), the outer peak of ⟨uu⟩+\langle uu\rangle^{+}, if it exists, should be bounded by ⟨uu⟩+≈10\langle uu\rangle^{+}\approx10. Clarification of such differences of perspectives in the asymptotic state require future work.

Second, while (1.3) adheres closely with TBL data of Vallikivi et al. [48] up to y∗=1y^{*}=1 (Figure 6c), it is slightly and uniformly higher than TBL data of Samie et al. [35] for y∗>0.6y^{*}>0.6 (Figure 6e). This is due to the fact that former set of data are obtained for flow over a flat plate mounted in the same pipe in which data of Hultmark et al. [12] are measured. Therefore, the data of Vallikivi et al. [48] in its outer wake region resemble the center behavior of pipe data by Hultmark et al. [12], both in agreement with (1.3) up to y∗=1y^{*}=1. In contrast, TBL data of Samie et al. [35] are measured in the Melbourne wind tunnel with the normal free stream boundary condition, so that a vanishing ⟨uw⟩+≈0\langle uw\rangle^{+}\approx0 towards boundary layer edge is observed (Figure 6e&Figure 6f), which is

lower than (1.3). Such a difference reflects the wake influence on ⟨uu⟩+\langle uu\rangle^+ in TBL. In fact, the wake influence is much sharper for ⟨ww⟩+\langle ww\rangle^+ and p′+p^{\prime+} in TBL (Figure 7e & Figure 8e). This issue will be addressed in the section 4.2.

The third and final point is that for p′+p^{\prime+}, the green dotted line in figure 8 represents

p′+(y∗)=Bϕ−Aϕln⁡y∗,(20)p^{\prime+}(y^*) = \sqrt{B_\phi- A_\phi\ln y^*}, \tag*{(20)}

which is the square root of (1.1) for pressure variance

⟨pp⟩+(y∗)=Bϕ−Aϕln⁡y∗.(21)\langle pp\rangle^+(y^*) = B_\phi- A_\phi\ln y^*. \tag*{(21)}

This equation is obtained by Panton et al. [32] via an inner-outer matching (i.e. a viscous inner layer overlapping with an inviscid outer layer). It is almost indistinguishable from (1.3) in Figure 8. Nevertheless, as shown in Figure 9a-Figure 9c, ⟨pp⟩w+−⟨pp⟩p+\langle pp\rangle_w^+ - \langle pp\rangle_p^+ versus y+y^+ produces no data collapse for y+>5y^+ > 5. Particularly for the trough located at about y+=30y^+ = 30, the data are markedly lower for increasing ReτRe_\tau, thus creating a challenge for the inner-outer matching analysis. To reconcile this challenge, Panton et al. [32] introduced two logarithmic slopes, i.e. ACP=2.56A^{CP} = 2.56 for the common part of presumed log profile in the overlap layer, and another Aw=2.24A^w = 2.24 for the ReτRe_\tau variation of the wall pressure. Following this fix, one can estimate

Four-panel plots showing the wall-normal dependence and Reynolds-number dependence of pressure fluctuations

Figure 9. Wall-normal dependence of the variance of pressure fluctuations is shown in the plots of ⟨pp⟩w+−⟨pp⟩+\langle pp\rangle_w^+ - \langle pp\rangle^+ versus y+y^+, where ⟨pp⟩w+\langle pp\rangle_w^+ is the wall-value of ⟨pp⟩+\langle pp\rangle^+. Lines are the same DNS data as in Figure 3, (a) for channels, (b) for pipes and (c) for TBLs. Note that lines depart markedly from each other with increasing ReτRe_\tau in the region y+>5y^+ > 5. (d) Difference between the peak and wall values of pressure variance, i.e. ⟨pp⟩p+−⟨pp⟩w+\langle pp\rangle_p^+ - \langle pp\rangle_w^+, for channel, pipe and TBL flows for a series of ReτRe_\tau values. Dotted line (green) indicates the logarithmic growth by (22), i.e. 0.32ln⁡Reτ+0.330.32\ln Re_\tau+ 0.33. Dashed line (red) indicates the bounded variation of (23) according to CS. Symbols are DNS data, squares for channel ([17]), circles for pipe ([33]) and diamonds for TBL ([38]). Note that DNS channel data at Reτ=150,300,400Re_\tau= 150, 300, 400 of [13], and Reτ=180,550,944,2000Re_\tau= 180, 550, 944, 2000 of [9] are also included here.

⟨pp⟩p+−⟨pp⟩w+∝(ACP−Aw)ln⁡Reτ≈0.32ln⁡Reτ,(22)\langle pp\rangle_p^+ - \langle pp\rangle_w^+ \propto(A^{CP} - A^w)\ln Re_\tau\approx0.32\ln Re_\tau, \tag*{(22)}

which would break the wall scaling completely.

As a comparison, Figure 3b,Figure 3d,Figure 3f show that data of pw′+−pp′+p_w^{\prime+} - p_p^{\prime+} collapsed —redwell up to the trough, better than ⟨pp⟩w+−⟨pp⟩p+\langle pp\rangle_w^+ - \langle pp\rangle_p^+ in Figure 9a-Figure 9c. Moreover, via the bounding relation pw′+≈4.4−10.5/Reτ1/4p_w^{\prime+} \approx4.4 - 10.5/Re_\tau^{1/4} and pp′+≈4.84−10.5/Reτ1/4p_p^{\prime+} \approx4.84 - 10.5/Re_\tau^{1/4} given in CS, we have

⟨pp⟩p+−⟨pp⟩w+=pp′+2−pw′+2≈4.07−9.24/Reτ1/4,(23)\langle pp\rangle_p^+ - \langle pp\rangle_w^+ = p_p^{\prime+2} - p_w^{\prime+2} \approx4.07 - 9.24/Re_\tau^{1/4}, \tag*{(23)}

which depicts data satisfactorily in a wider ReτRe_\tau range than (3.11) in Figure 9d.

Discussion on flow geometry influence

Near wall universality

We focus here on the geometry effects. First, a universal data collapse is summarized by unifying (2.2), (2.3) and (2.4) together, i.e.

ϕ+(y+,Reτ)−ϕw+(Reτ)ϕp+(Reτ)−ϕw+(Reτ)=fϕ(y+),(24)\frac{\phi^+(y^+, Re_\tau)-\phi_w^+(Re_\tau)}{\phi_p^+(Re_\tau)-\phi_w^+(Re_\tau)}=f_\phi(y^+), \tag*{(24)}

where fϕ(y+)f_\phi(y^+) depends on y+y^+ and ϕ\phi but is independent of ReτRe_\tau. For ⟨uu⟩+\langle uu\rangle^+, ϕw+=0\phi_w^+=0 so that (4.1) reduces to (2.2); the same is true for ⟨ww⟩+\langle ww\rangle^+. For p′+p^{\prime+}, as ϕp+−ϕw+\phi_p^+ - \phi_w^+ is a constant independent of ReτRe_\tau shown in Figure 3, (4.1) reproduces (2.4) with fϕ=1−j(y+)/j(0)f_\phi=1-j(y^+)/j(0).

Moreover, as y+y^+ moves from the wall to the peak location, it is interesting to check whether fϕ(y+)f_\phi(y^+) is universal for channel, pipe and TBL flows. This is indeed verified in Figure 10a, Figure 10b & Figure 10c, for ⟨uu⟩+\langle uu\rangle^+, ⟨ww⟩+\langle ww\rangle^+ and p′+p^{\prime+}, respectively. Profiles from these three wall flows collapsed together from the wall to the peak location, which means that ReτRe_\tau-dependence and geometry influence are canceled by the ratio of relative variations composed of ϕw+\phi_w^+ and ϕp+\phi_p^+. This is conceivable if the near wall region is viewed as a self-organized entity, so that superposition and modulation effects enforced by the outer flow structures are characterized to the first order by wall and peak values.

Wall-normal dependence for fluctuations after normalization by the corresponding peaks

Figure 10. Wall-normal dependence for fluctuations after normalization by the corresponding peaks. (a) ⟨uu⟩+/⟨uu⟩p+\langle uu\rangle^{+}/\langle uu\rangle_{p}^{+}; (b) ⟨ww⟩+/⟨ww⟩p+\langle ww\rangle^{+}/\langle ww\rangle_{p}^{+}; (c) pp′+−p′+p_{p}^{\prime+}-p^{\prime+}. Lines are DNS data of channel at Reτ=1000Re_{\tau}=1000 [17], of pipe at Reτ=1140Re_{\tau}=1140 [33], and of TBL at Reτ≈1270Re_{\tau}\approx1270 [36], all of which collapse closely on each other.

Wake modification in TBL

Note that in Figure 7 and Figure 8, ⟨ww⟩+\langle ww\rangle^+ and p′+p^{\prime+} depart from (1.3) for y∗≳0.5y^* \gtrsim0.5 in TBL, but the agreement persists all the way to the centerline of channel and pipe flows. Recall the findings by Chen et al. [2] that the total shear stress τ+\tau^+ and the Reynolds shear stress ⟨−uv⟩+\langle-uv\rangle^+ in the wake of the TBL differ notably from that in channel and pipe flows. The reason for this difference, according to Chen et al. [2], is the nonzero mean momentum transport in the wall-normal direction of TBL, i.e. V/Ve∝(y∗)3/2V/V_e \propto(y^*)^{3/2} (the subscript ee indicates the value at the boundary layer edge, as noted earlier). The latter leads to ⟨−uv⟩+≈τ+≈1−(y∗)3/2\langle-uv\rangle^+ \approx\tau^+ \approx1-(y^*)^{3/2} in TBL, differing from ⟨−uv⟩+=1−y∗\langle-uv\rangle^+=1-y^* in channel and pipe flows for which V=0V=0.

Once we accept the difference in the ⟨−uv⟩+\langle-uv\rangle^{+} behavior, we may expect a similar wake modification on ⟨ww⟩+\langle ww\rangle^{+} and p′+p^{\prime+} by the nonzero VV in TBL. That is,

ϕ+−ϕe+∝ϕuv+−ϕuv,e+,(25)\phi^{+}-\phi_{e}^{+}\propto\phi_{uv}^{+}-\phi_{uv,e}^{+}, \tag*{(25)}

where ϕ+\phi^{+} represents ⟨ww⟩+\langle ww\rangle^{+} or p′+p^{\prime+}, and ϕuv+\phi_{uv}^{+} represents ⟨−uv⟩+\langle-uv\rangle^{+}. If so, substituting ϕuv+≈1−(y∗)3/2\phi_{uv}^{+}\approx1-(y^{*})^{3/2} into (25) yields

ϕ+(y∗)−ϕe+=cϕ(ϕuv+−ϕuv,e+)≈cϕ[1−(y∗)3/2],(26)\phi^{+}(y^{*})-\phi_{e}^{+}=c_{\phi}(\phi_{uv}^{+}-\phi_{uv,e}^{+})\approx c_{\phi}[1-(y^{*})^{3/2}], \tag*{(26)}

where the proportionality coefficient cϕc_{\phi} is independent of y∗y^{*} but may depend on ϕ\phi. Verification of (26) for TBL is provided in Figure 11. The agreement with data is quite satisfactory for y∗>0.2y^{*}>0.2, and the fitting parameters are ϕe+=0\phi_{e}^{+}=0 and cϕ=2.02c_{\phi}=2.02 for ⟨ww⟩+\langle ww\rangle^{+}, while ϕe+=0.8\phi_{e}^{+}=0.8 and cϕ=1.8c_{\phi}=1.8 for p′+p^{\prime+}. This model for the wake flow could also be applied to describe ⟨uu⟩+\langle uu\rangle^{+} towards the free stream of TBL, but the deviation is fairly small as shown in Figure 5e and Figure 6e, and will not be considered further here.

Image not extracted

Figure 11. Same plot as Figure 7(e) for ⟨ww⟩+\langle ww\rangle^+, and as Figure 8(e) for p′+p^{\prime+}, with newly added solid (cyan) lines (26) for wake modification in TBL. That is, ⟨ww⟩+=2.02[1−(y∗)3/2]\langle ww\rangle^+ = 2.02[1-(y^*)^{3/2}] in the left panel and p′+=0.8+1.8[1−(y∗)3/2]p^{\prime+}=0.8+1.8[1-(y^*)^{3/2}] in the right.

Perspective and conclusions

New methods of analysis and generations of new experiments and simulations have revealed deeper layers of interesting questions on wall flow dynamics. Previously unthinkable questions, such as the universality of the Kármán constant in the mean flow description and the scaling of fluctuations in these flows, as well as the implications of the behavior of fluctuations for the mean velocity itself, can now be asked, and reasonable answers for them can be attempted. In contrast to the mean velocity, concerted effort to understand the scaling of fluctuations is relatively new. This paper, when taken together with our earlier work [3, 4], provides a self-consistent description of fluctuations in streamwise and spanwise velocity, as well as pressure fluctuations. One of the main qualitative conclusions of this work is that wall-normalized fluctuations are bounded even when Reτ→∞Re_\tau\to\infty, thus restoring the validity of the standard law of the wall. The alternative scenario of attached eddy hypothesis and its consequences lead to a different conclusion.

Aiming for an asymptotic description of fluctuations in canonical wall flows, we have obtained several new results, summarized as follows. First, excellent data collapse is achieved for the near-wall rms profiles of streamwise and spanwise velocity fluctuations (⟨uu⟩+\langle uu\rangle^+ and ⟨ww⟩+\langle ww\rangle^+) as well as pressure fluctuations (p′+p^{\prime+}). Their spatial variations and the Reynolds number dependence are decoupled via the normalization through peak values. With the defect law for the peaks given in CS, a universal near wall expansion (9) is obtained with the specific gauge function g=Reτ−1/4g=Re_\tau^{-1/4}, consistent with that developed by Monkewitz [26].

Moreover, a defect decay (3) is derived by matching (9) with the outer inviscid similarity (12). Compared to the log-profile by Townsend’s attached eddy hypothesis, it is shown that (3) reproduces the data better, not only over a wider ReτRe_\tau domain but also in a larger flow region. As indicated by (3), there would appear an asymptotic plateau as y∗→0y^* \to0 and Reτ→∞Re_\tau\to\infty, which implies a quasi-equilibrium state with contributions to fluctuations coming from all associated eddies that are invariant as wall-normal position changes. If so, the intriguing outer peak of streamwise fluctuation, if one exists, would be bounded by ⟨uu⟩+≈10\langle uu\rangle^+ \approx10.

Finally, a near wall universality (24) is obtained independent of both Reynolds number and flow geometries. In addition, a wake flow modification in TBL is introduced for ⟨ww⟩+\langle ww\rangle^+ and p′+p^{\prime+}, which shows close agreement with data towards the boundary layer edge. There is no gainsaying that more and better data are required to put all these results on a firmer foundation. It is exciting to await cleaner data at higher Reynolds numbers with improved resolution.

Acknowledgement. We thank all the authors cited in the figures for making their data available. X. Chen acknowledges the support by the National Natural Science Foundation of China, No. 12072012, 11721202 and 91952302.

References

  1. [1]Bradshaw, P. 1967 ‘Inactive’ motion and pressure fluctuations in turbulent boundary layers. J. Fluid Mech. 30, 241–258.DOI
  2. [2]Chen, X., Hussain, F. & She, Z.S. 2019 Non-universal scaling transition of momentum cascade in wall turbulence. J. Fluid Mech. 871, R2.DOI
  3. [3]Chen, X. & Sreenivasan, K.R. 2021 Reynolds number scaling of the peak turbulence intensity in wall flows. J. Fluid Mech. 908, R3.DOI
  4. [4]Chen, X. & Sreenivasan, K.R. 2022 Law of bounded dissipation and its consequences in turbulent wall flows. J. Fluid Mech. 933, A20.
  5. [5]Chin, C., Philip, J., Klewicki, J., Ooi, A. & Marusic, I. 2014 Reynolds-number-dependent turbulent inertia and onset of log region in pipe flows. J. Fluid Mech. 757, 747–769.DOI
  6. [6]DeGraaff, D. B. & Eaton, J. K. 2000 Reynolds-number scaling of the flat plate turbulent boundary layer. J. Fluid Mech. 422, 319–346.DOI
  7. [7]Diaz-Daniel, C., Laizet, S. & Vassilicos, J. C. 2017 Wall shear stress fluctuations: Mixed scaling and their effects on velocity fluctuations in a turbulent boundary layer. Phys. Fluids 29, 055102.DOI
  8. [8]Gerolymos, G.A. & Vallet, I. 2023 Scaling of pressure fluctuations in compressible turbulent plane channel flow. J. Fluid Mech. 958, A19.DOI
  9. [9]Hoyas, S. & Jimenez, J. 2006 Scaling of the velocity fluctuations in turbulent channels up to Reτ=2003. Phys. Fluids 18, 011702.DOI
  10. [10]Hoyas, S., Oberlack, M., Alcantara-Avila, F., Kraheberger, S.V. & Laux, J. 2022 Wall turbulence at high friction Reynolds numbers. Phys. Rev. Fluids 7, 046202.DOI
  11. [11]Hultmark, M. 2012 A theory for the streamwise turbulent fluctuations in high Reynolds number pipe flow. J. Fluid Mech. 707, 575–584.DOI
  12. [12]Hultmark, M., Vallikivi, M., Bailey, S.C.C. & Smits, A. J. 2012 Turbulent pipe flow at extreme Reynolds numbers. Phys. Rev. Lett. 108, 094501.pubmed.ncbi.nlm.nih.gov/22463643
  13. [13]Iwamoto, K., Suzuki, Y. & Kasagi, N. 2002 Database of fully developed channel flow. Tech. Rep. ILR-0201, see http://www.ihtlab.utokyo. ac.jp. Japan.
  14. [14]Jimenez, J., Hoyas, S., Simens, M.P. & Mizuno, Y. 2010 Turbulent boundary layers and channels at moderate Reynolds numbers. J. Fluid Mech. 657, 335–360.DOI
  15. [15]Klewicki, J.C. 2022 Bounded dissipation predicts finite asymptotic state of near-wall turbulence. J. Fluid Mech. 940, F1.DOI
  16. [16]Klewicki, J.C., Priyadarshana, P.J.A. & Metzger, M.M. 2008 Statistical structure of the fluctuating wall pressure and its in-plane gradients at high Reynolds number. J. Fluid Mech. 609, 195–220.DOI
  17. [17]Lee, M. & Moser, R.D. 2015 Direct numerical simulation of turbulent channel flow up to Reτ = 5200. J. Fluid Mech. 774, 395–415.DOI
  18. [18]Marusic, I., Baars, W.J. & Hutchins, N. 2017 Scaling of the streamwise turbulence intensity in the context of inner–outer interactions in wall turbulence. Phys. Rev. Fluids 2, 100502.DOI
  19. [19]Marusic, I., McKeon, B. J., Monkewitz, P. A., Nagib, H. M., Smits, A. J. & Sreenivasan, K. R. 2010 Wall-bounded turbulent flows at high Reynolds numbers: Recent advances and key issues. Phy. Fluids 22, 065103.DOI
  20. [20]Marusic, I. & Monty, J.P. 2019 Attached eddy model of wall turbulence. Annu. Rev. Fluid Mech. 51, 49–74.DOI
  21. [21]Marusic, I., Monty, J.P., Hultmark, M. & Smits, A.J. 2013 On the logarithmic region in wall turbulence. J. Fluid Mech. 716, R3.DOI
  22. [22]Meneveau, C. & Marusic, I. 2013 Generalized logarithmic law for high-order moments in turbulent boundary layers. J. Fluid Mech. 719, R1.DOI
  23. [23]Metzger, M, McKeon, B.J & Holmes, H 2007 The near-neutral atmospheric surface layer: turbulence and non-stationarity. Phil. Trans. Roy. Soc. Lond. A 365 (1852), 859–876.pubmed.ncbi.nlm.nih.gov/17244589
  24. [24]Metzger, M. M. & Klewicki, J. 2001 A comparative study of near-wall turbulence in high and low Reynolds number boundary layers. Phys. Fluids 13, 692–701.DOI
  25. [25]Millikan, C.M. 1938 A critical discussion of turbulent flows in channels and circular tubes. In Proc. 5th Intl Congress of Applied Mechanics. John Wiley and Sons. p. 386–392.
  26. [26]Monkewitz, P. 2022 Asymptotics of streamwise Reynolds stress in wall turbulence. J. Fluid Mech. 931, A18.DOI
  27. [27]Monkewitz, P. & Nagib, H.M. 2023 The hunt for the Kármán “constant” revisited. J. Fluid Mech. (to appear) .DOI
  28. [28]Moser, R.D., Kim, J. & Mansour, N.N. 1999 Direct numerical simulation of turbulent channel flow up to Reτ = 590. Physics of Fluids 11, 943–945.DOI
  29. [29]Nagib, H., Monkewitz, P. & Sreenivasan, K.R. 2022 Reynolds number required to accurately discriminate between proposed trends of peak normal stress in wall turbulence. APS bulletin: 75th Annual Meeting of the Division of Fluid Dynamics, Indianapolis, Indiana .arxiv.org/abs/2312.01184
  30. [30]Nagib, H. M., Chauhan, K. A. & Monkewitz, P. A. 2007 Approach to an asymptotic state for zero pressure gradient turbulent boundary layers. Phil. Trans. R. Soc. A 365, 755–770.pubmed.ncbi.nlm.nih.gov/17244582
  31. [31]Ono, M., Furuichi, N., Wada, Y., Kurihara, N. & Tsujii, Y. 2022 Reynolds number dependence of inner peak turbulence intensity in pipe flow. Phys. Fluids 34, 045103.DOI
  32. [32]Panton, R. L., Lee, M. & Moser, R. D. 2017 Correlation of pressure fluctuations in turbulent wall layers. Phys. Rev. Fluids 2, 094604.DOI
  33. [33]Pirozzoli, S., Romero, J., Fatica, M., Verzicco, R. & Orlandi, P. 2021 One-point statistics for turbulent pipe flow up to Reτ ≈ 6000. J. Fluid Mech. 926, A28.DOI
  34. [34]Pope, S. B. 2000 Turbulent flows. Cambridge University Press.
  35. [35]Samie, M., Marusic, I., Hutchins, N., Fu, M. K., Fan, Y., Hultmark, M. & Smits, A. J. 2018 Fully resolved measurements of turbulent boundary layer flows up to Reτ = 20000. J. Fluid Mech. 851, 391–415.
  36. [36]Schlatter, P., Li, Q., Brethouwer, G., Johansson, A. V. & Henningson, D. S. 2010 Simulations of spatially evolving turbulent boundary layers up to Reθ = 4300. Int. J. Heat Fluid Flow 31 (3), 251–261.
  37. [37]Schlatter, P. & Örlü, R. 2010 Assessment of direct numerical simulation data of turbulent boundary layers. J. Fluid Mech. 659, 116–126.DOI
  38. [38]Schlatter, P., Örlü, R., Li, Q., Brethouwer, G., Fransson, J. H. M., Johansson, A. V., Alfredsson, P. H. & Henningson, D. S. 2009 Turbulent boundary layers up to Reθ=2500 studied through simulation and experiment. Phys. Fluids 11, 051702.DOI
  39. [39]She, Z. S., Chen, X. & Hussain, F. 2017 Quantifying wall turbulence via a symmetry approach: A Lie group theory. J. Fluid Mech. 827, 322–356.DOI
  40. [40]Sillero, J. A., Jimenez, J. & Moser, R. 2013 One-point statistics for turbulent wall-bounded flows at Reynolds numbers up to δ+ = 2000. Phys. Fluids 25, 105102.DOI
  41. [41]de Silva, C. M., Marusic, I., Woodcock, J. D. & Meneveau, C. 2015 Scaling of second- and higher-order structure functions in turbulent boundary layers. J. Fluid Mech. 769, 654–686.DOI
  42. [42]Smits, A.J., Hultmark, M., Lee, M., Pirozzoli, S. & Wu, X.H. 2021 Reynolds stress scaling in the near-wall region of wall-bounded turbulence. J. Fluid Mech. 926, A31.DOI
  43. [43]Smits, A. J., McKeon, B. J. & Marusic, I. 2011 High-Reynolds number wall turbulence. Annu. Rev. Fluid Mech. 43, 353–375.
  44. [44]Spalart, P. R. & Abe, H. 2021 Empirical scaling laws for wall-bounded turbulence deduced from direct numerical simulations. Phys. Rev. Fluids 6, 044604.DOI
  45. [45]Sreenivasan, K.R. 1989 The turbulent boundary layer. In Frontiers in Experimental Fluid Mechanics, edited by M. Gad-el-Hak, pp. 159–209.
  46. [46]Townsend, A. A. 1956 The structure of turbulent shear flow. Cambridge University Press.
  47. [47]Tsujii, Y., Fransson, J. H. M., Alfredsson, P. H. & Johansson, A. V. 2007 Pressure statistics and their scaling in high-reynolds-number turbulent boundary layers. J. Fluid Mech. 585, 1–40.
  48. [48]Vallikivi, M., Ganapathisubramani, B. & Smits, A. J. 2015 Spectral scaling in boundary layers and pipes at very high Reynolds numbers. J. Fluid Mech. 771, 303–326.
  49. [49]Vassilicos, J. C., Laval, J.-P., Foucaut, J.-M. & Stanislas, M. 2015 The streamwise turbulence intensity in the intermediate layer of turbulent pipe flow. J. Fluid Mech. 774, 324–341.DOI
  50. [50]Vincenti, P., Klewicki, J., Morrill-Winter, C., White, C. M. & Wosnik, M. 2013 Streamwise velocity statistics in turbulent boundary layers that spatially develop to high Reynolds number. Exp. Fluids 54, 1629.DOI
  51. [51]Willert, C., Soria, J., Stanislas, M., Klinner, J., Amili, O., Eisfelder, M., Cuvier, C., Bellani, G., Fiorini, T. & Talamelli, A. 2017 Near-wall statistics of a turbulent pipe flow at shear Reynolds numbers up to 40 000. J. Fluid Mech. 826, R5.DOI
  52. [52]Wosnik, M., Castillo, L. & George, W.K. 2000 A theory for turbulent pipe and channel flows. J. Fluid Mech. 421, 115–145.DOI
  53. [54]Yang, X. I. A. & Lozano-Durán, A. 2017 A multifractal model for the momentum transfer process in wall-bounded flows. J. Fluid Mech. 824, R2.DOI
  54. [55]Yao, J., Chen, X. & Hussain, F. 2022 Direct numerical simulation of turbulent open channel flows at moderately high Reynolds numbers. J. Fluid Mech. 953, A19.DOI
  55. [56]Yao, J., Rezaeiravesh, S., Schlatter, P. & Hussain, F. 2023 Direct numerical simulations of turbulent pipe flow up to Reτ ≈ 5200. J. Fluid Mech. 956, A18.DOI

Paper details

vlnc:W_a7hewaf9
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

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