1. Introduction
Turbulent boundary layers (TBLs) are one of the focuses of fluid dynamics studies, and many exciting studies have focused on zero pressure gradient (ZPG) TBLs. When an adverse pressure gradient (APG) is present, the properties of the boundary layer are significantly altered. The combined effect of APG and viscosity can cause flow separation, which has received sufficient attention in engineering fields such as aerospace (Simpson Reference Simpson1989). How to elucidate the properties of APG TBLs based on studies of ZPG TBLs is a key factor in studying APG TBLs.
The relationship between ZPG and APG can be seen from the governing equations. For two-dimensional incompressible TBLs, the basic governing equations include the mean continuity equation and the mean momentum equation. The pressure gradient term is included in the mean momentum equation. If the pressure gradient term is ignored, then the governing equation for ZPG TBLs is obtained. Therefore, a basic idea is to start from the governing equation and study the effect of APG relative to ZPG conditions. This result can be obtained by analysing the mean momentum balance (MMB) equation previously used to study wall turbulence (Wei et al. Reference Wei, Fife, Klewicki and Mcmurtry2005; Klewicki, Fife & Wei Reference Klewicki, Fife and Wei2009; Klewicki Reference Klewicki2013; Morrill-Winter, Philip & Klewicki Reference Morrill-Winter, Philip and Klewicki2017). For APG TBLs, Romero et al. (Reference Romero, Zimmerman, Philip, White and Klewicki2022) used it to study the properties of the logarithmic region and the distance-from-the-wall scaling. Han, Ma & Yan (Reference Han, Ma and Yan2024) attributed the widely observed outward shift of the Reynolds shear stress peak to the insufficient inertia term to balance the pressure gradient term by analysing the MMB equation. The MMB equation can be used to understand the influence of APG on different positions in the normal direction of TBLs. If we focus on the overall effect of APG at specific streamwise direction stations, then it involves integrals of the governing equations in the normal direction.
The wall-normal integral is defined as the integration of physical quantities along the normal direction, starting from the wall or boundary layer edge for a flat plate. The choice of integral limits and the specific direction can give different results. The wall-normal integral is a powerful tool for obtaining additional quantitative information. In recent years, it has been found that the wall-normal integral can be used to obtain the relationship between the TBL edge velocity and the friction velocity
$u_{\tau }=\sqrt {\tau _{w}/\rho }$
in ZPG and APG TBLs. Here,
$\tau _{w}$
is the wall shear stress, and
$\rho$
is the density. Wei & Klewicki (Reference Wei and Klewicki2016) derived the relationships among the TBL edge streamwise velocity
$U_{e}$
, normal velocity
$V_{e}$
, shape factor
$H=\delta ^{*}/\theta$
and friction velocity
$u_{\tau }$
under ZPG conditions at infinite Reynolds numbers. Here,
$\delta ^{*}$
is the displacement thickness, and
$\theta$
is the momentum thickness. Wei, Maciel & Klewicki (Reference Wei, Maciel and Klewicki2017) extended this relationship to the equilibrium APG TBLs. Wei & Klewicki (Reference Wei and Klewicki2023) extended this relationship to ZPG TBLs with streamwise variation. Moreover, Wei, Li & Wang (Reference Wei, Li and Wang2023b
) and Wei et al. (Reference Wei, Li, Knopp and Vinuesa2023a
) provided the analytical equations between
$V_{e}$
and
$U_{e}$
under ZPG and APG conditions, respectively, which included the streamwise derivative of
$\delta ^{*}$
. Wall-normal integrals are also widely used in the study of wall friction decomposition (Fukagata, Iwamoto & Kasagi Reference Fukagata, Iwamoto and Kasagi2002; Renard & Deck Reference Renard and Deck2016; Elnahhas & Johnson Reference Elnahhas and Johnson2022). These equations have been gradually extended to various fields, including APG TBLs (Fan et al. Reference Fan, Li, Atzori, Pozuelo, Schlatter and Vinuesa2020; Atzori et al. Reference Atzori, Mallor, Pozuelo, Fukagata, Vinuesa and Schlatter2023). The above studies indicate that the wall-normal integral can extend the study of mean velocity under ZPG conditions to APG TBLs.
As an important part of wall turbulence, study of the mean velocity is usually accompanied by the Reynolds stress. A widely accepted theory in ZPG TBLs is the emergence of a logarithmic law at high Reynolds numbers (Pope Reference Pope2000). At this point, there is a non-dimensional Reynolds shear stress
$-\langle uv\rangle ^{+}$
close to unity. Here,
$+$
refers to the use of friction-viscous scaling. The Reynolds shear stress, which appears in the mean momentum equation, serves as a bridge between the mean flow and the fluctuating flow. The variation in the Reynolds shear stress can be considered a momentum cascade process (Chen, Hussain & She Reference Chen, Hussain and She2019). An interesting question is whether it is possible to establish a more quantitative relationship between
$-\langle uv\rangle$
and the mean flow, which can also make it more convenient and accurate to predict total stress or wall friction.
The wall-normal integral from the wall to the interior of the TBLs allows the Reynolds shear stress to be expressed as a function of mean flow by the mean momentum equation. Volino & Schultz (Reference Volino and Schultz2018) proposed a method to determine the Reynolds shear stress
$-\langle uv\rangle$
using the mean velocity
$U$
. Womack, Meneveau & Schultz (Reference Womack, Meneveau and Schultz2019) extended the above study by using only one streamwise position in the analysis. Kumar & Mahesh (Reference Kumar and Mahesh2021) modelled the total stress under ZPG conditions using the mean normal velocity
$V$
. Kumar & Mahesh (Reference Kumar and Mahesh2022) extended the above study to APG TBLs. Based on the study of the analytical equation of the mean normal velocity
$V$
under ZPG conditions, Wei, Li & Wang (Reference Wei, Li and Wang2023b
) provided an analytical equation for the Reynolds shear stress
$-\langle uv\rangle$
.
Notably, most of the above studies used ZPG or equilibrium APG TBL data, which makes the method less applicable to more general pressure gradient distributions. We attempt to establish a relationship between
$-\langle uv\rangle$
and the mean velocity for non-equilibrium APG TBLs via the wall-normal integral. This helps to link the study of
$-\langle uv\rangle$
and the mean velocity. Due to the complexity of non-equilibrium APG TBLs, all derivations must be treated with caution, and potential limitations must be clarified. The organisation of the paper is as follows. Section 2 introduces the databases used in this paper. Section 3 derives the relationship between the TBL edge velocity and the wall-normal integral in non-equilibrium APG TBLs. Then, starting from the mean momentum equation, an approximate relationship between the Reynolds shear stress and mean velocity is proposed. Section 4 discusses the results presented in
$\S$
3, and emphasises the range of applicability of the pressure gradient parameter. Section 5 summarises the main conclusions of this paper.
2. Databases
2.1. Equilibrium and non-equilibrium
We verify the results via seven databases, including four non-equilibrium APG databases and three near-equilibrium APG databases. The exact definition of equilibrium and non-equilibrium flow is important, to the extent that it is still debated. Since the concept of equilibrium flow was established (Rotta Reference Rotta1953; Clauser Reference Clauser1954; Townsend Reference Townsend1956), it has been accompanied by studies of APG TBLs. The concept of equilibrium is often associated with self-similarity. Their relationship is considered to be that if a consistent set of scales can be used to bring all variables in the flow to a self-similar state, then the flow is in equilibrium (Devenport & Lowe Reference Devenport and Lowe2022). Self-similarity requires that the physical quantities be appropriately scaled so that the collapse profiles do not change with the streamwise direction (Gungor, Gungor & Maciel Reference Gungor, Gungor and Maciel2024). Based on different degrees of similarity analysis, different requirements can be derived (Mellor & Gibson Reference Mellor and Gibson1966; Castillo & George Reference Castillo and George2001; Maciel, Rossignol & Lemay Reference Maciel, Rossignol and Lemay2006; Kitsios et al. Reference Kitsios, Sekimoto, Atkinson, Sillero, Borrell, Gungor, Jiménez and Soria2017). Since only sink flow satisfies the strictest definition of equilibrium (Townsend Reference Townsend1956), flow is defined as near-equilibrium in the databases that we use (Bobke et al. Reference Bobke, Vinuesa, Örlü and Schlatter2017; Pozuelo et al. Reference Pozuelo, Li, Schlatter and Vinuesa2022). This requires that the free flow velocity is defined by the power-law
$U_{\infty }(x)=C ( x-x_0 ) ^{m}$
(Mellor & Gibson Reference Mellor and Gibson1966; Townsend Reference Townsend1976). Here,
$C$
is a constant,
$x_0$
is the virtual origin, and
$m$
is the power-law exponent. In APG TBLs,
$-1/3\lt m\lt 0$
. Under these conditions,
$\delta ^{*}$
and
$\theta$
grow linearly along the streamwise direction, while their ratio
$H$
remains constant (Maciel et al. Reference Maciel, Rossignol and Lemay2006).
Non-equilibrium flow is affected by pressure gradients that constantly change with the streamwise direction and cannot satisfy the similarity conditions. A characteristic of this type of flow is that the shape factor
$H$
varies with the streamwise direction. The shape factor is thought to reflect the cumulation of history effects in the boundary layer (Maciel et al. Reference Maciel, Wei, Gungor and Simens2018). Gungor et al. (Reference Gungor, Gungor and Maciel2024) suggested that an increase in the shape factor indicates that TBLs gain momentum. Due to the different upstream evolution of non-equilibrium APG TBLs, the accumulated history effects will affect the downstream flow. Even with the same Reynolds number and pressure gradient parameter
$\beta$
, the velocity and Reynolds stress profiles of non-equilibrium and near-equilibrium flows are not the same (Bobke et al. Reference Bobke, Vinuesa, Örlü and Schlatter2017; Devenport & Lowe Reference Devenport and Lowe2022). This suggests that extending the relevant theories of near-equilibrium flow to non-equilibrium flow requires consideration of the complex history effects, and applying the theory of non-equilibrium flow to equilibrium flow also requires careful examination.
2.2. Near-equilibrium and non-equilibrium databases
If the pressure gradient parameter
$\beta$
remains constant during the evolution of the streamwise direction, then there is no history effect. Bobke et al. (Reference Bobke, Vinuesa, Örlü and Schlatter2017) provided a detailed study of this type of flow. Their b1 (
$x_{0}=110$
,
$m=-0.14$
) and b2 (
$x_{0}=110$
,
$m=-0.18$
) databases are used as near-equilibrium APG TBLs without history effects. Pozuelo et al. (Reference Pozuelo, Li, Schlatter and Vinuesa2022) provided near-equilibrium data at higher Reynolds numbers, shown here as b1.4 (
$x_{0}=60$
,
$m=-0.16$
). Figure 1(a) shows the variation of the physical quantities with the streamwise direction, using the b2 database as an example of a near-equilibrium database. The shape factor remains constant, the boundary layer thickness increases linearly, and
$\beta$
remains approximately constant. The length and velocity scales of the above databases are non-dimensionalised using
$\delta _\infty ^*$
and
$U_\infty$
. Here,
$\delta _\infty ^*$
is the displacement thickness of the laminar inflow, and
$U_\infty$
is the free-stream velocity at the inlet.
Streamwise variation of the flow physical quantities: (a) data from b2; (b) data from DNS2018c.

A common method for direct numerical simulations (DNS) of non-equilibrium APG TBLs is to use blowing and suction boundary conditions at the top of the computational domain. Four sets of non-equilibrium databases with different Reynolds numbers from Coleman, Rumsey & Spalart (Reference Coleman, Rumsey and Spalart2018), Coleman (Reference Coleman2021) and Gungor et al. (Reference Gungor, Gungor and Maciel2024) are used, labelled DNS2018b, DNS2018c, DNS2021 and DNS2024. The Reynolds number of the first three databases gradually increases. Their APG is generated by the blowing transpiration profile
$V_{\textit{top}}(x)$
at a fixed distance
$Y$
offset from the no-slip wall. Figure 1(b) shows the variation of the physical quantities with the streamwise direction, using DNS2018c as an example of a non-equilibrium database. The shape factor, boundary layer thickness and pressure gradient increase nonlinearly. In these three databases, the length scales are non-dimensional with
$Y$
, and the velocity scales are non-dimensional with
$U_\infty$
. Here,
$U_\infty$
is the free-stream velocity of the ZPG TBL in the databases. The data in DNS2024 are not normalised.
Table 1 provides more detailed information on these databases, and compares different pressure gradient parameters
$\beta =({\delta ^{*} }/{\rho u_{\tau }^{2} }) ({\textrm{d} P_{e} }/{\textrm{d} x})$
,
$P^{+}= ({\nu }/{\rho u_{\tau }^{3} } )({\textrm{d} P_{w} }/{\textrm{d} x})$
and
$\beta _{zs} =({\delta }/{\rho U_{zs}^{2} }) ({\textrm{d} P_{e} }/{\textrm{d} x})$
. Here,
$P_{e}$
is pressure at the TBL edge,
$\nu$
is the kinematic viscosity,
$P_{w}$
is the pressure at the wall, and
$U_{zs} = U_{e} ({\delta ^{*} }/{\delta })$
is the Zagarola–Smits scaling (Zagarola & Smits Reference Zagarola and Smits1998). Note that only numerical simulation databases covering the low to moderate Reynolds number range are used. Although the Reynolds number effects on the results will be discussed in subsequent analyses, the results obtained under such limited Reynolds number conditions require further evaluation when the Reynolds number is further increased. Rkein & Laval (Reference Rkein and Laval2023) summarised the relationship between pressure gradient and Reynolds number in other databases in recent years. The results of this paper are applicable to small velocity defect situations, and the effect of further increasing the pressure gradient is discussed in the following sections. At the same time, it remains to be investigated how the current results can be extended to the studies with curvature. Overall, for the range of databases on which we are currently focusing, the Reynolds number is limited and the velocity defect is small. The results of the study will need to be re-examined as the Reynolds number or the velocity defect increases further.
Numerical simulation databases used in this paper. The first four are non-equilibrium TBL databases, and the last three are near-equilibrium TBL databases.

Both
$H$
and
$G=(U_{e}/u_{\tau })(1-1/H)$
are considered to characterise the cumulative effect of APG (Maciel et al. Reference Maciel, Wei, Gungor and Simens2018; Gungor et al. Reference Gungor, Gungor and Maciel2024). The difference is that
$H$
is considered to be more sensitive to the Reynolds number, while
$G$
no longer maintains a finite value as the flow tends towards separation (Maciel et al. Reference Maciel, Wei, Gungor and Simens2018). By plotting the defect shape factor
$G$
as a function of
$\beta$
, the degree of deviation from equilibrium can be obtained (Mellor & Gibson Reference Mellor and Gibson1966). As shown in figure 2, in the databases that we used, all four non-equilibrium databases deviate from three near-equilibrium databases, with DNS2018c having the largest degree of deviation. Meanwhile, a lag in the mean flow response to APG can be observed (Knopp et al. Reference Knopp, Reuther, Novara, Schanz, Schülein, Schröder and Kähler2021; Gungor et al. Reference Gungor, Gungor and Maciel2024).
The defect shape factor
$G$
as a function of
$\beta$
.

Mean velocity and Reynolds stress profiles normalised with friction-viscous scaling as functions of
$y^{+}$
: (a,c,e) from the near-equilibrium database b2; (b,d,f) from the non-equilibrium database DNS2018c.

To gain a fuller understanding of the databases, information on the mean velocity profiles and Reynolds stresses affected by the APG is required. Figure 3(a,c,e) are from the near-equilibrium database b2, where
$\beta$
decreases along the streamwise direction in the region of interest. The differences in the profiles of the physical quantities can be attributed to Reynolds number effects (Bobke et al. Reference Bobke, Vinuesa, Örlü and Schlatter2017). In contrast, figure 3(b,d,f) are from the non-equilibrium database DNS2018c, where
$\beta$
increases along the streamwise direction in the region of interest. The profiles of the physical quantities are simultaneously influenced by the pressure gradient, the Reynolds number and the upstream history effects. The effect of the APG can be visually observed in the results from the non-equilibrium database. For the mean velocity profiles, the most prominent feature is the appearance of a larger wake region. For the Reynolds shear stress
$-\langle uv\rangle$
, the APG induces an increasingly pronounced peak. For almost the same wall normal position, the Reynolds normal stress
$\langle uu\rangle$
profiles show that the magnitude of the outer peaks gradually approaches the inner peaks as the APG increases. In the outer region of the TBL, none of the profiles show any similarity. Relevant studies suggest that
$u_{\tau }$
is no longer an appropriate scaling for the large velocity defect (Gungor et al. Reference Gungor, Maciel, Simens and Soria2016; Maciel et al. Reference Maciel, Wei, Gungor and Simens2018; Gungor, Maciel & Gungor Reference Gungor, Maciel and Gungor2022; Romero et al. Reference Romero, Zimmerman, Philip, White and Klewicki2022). One possible explanation is that as the APG increases,
$u_{\tau }$
decreases, leading to a sharp increase in the normalised physical quantity. In the next section, we construct a dimensionless quantity
$u_{\tau }^{2}/ ( {U_{e}V_{e}} )$
, and use its decomposition form to illustrate the extent to which
$u_{\tau }$
decreases relative to the TBL edge velocity.
The definition of the TBL edge is important in related studies, and the non-equilibrium APG conditions make this issue more complex (Griffin, Fu & Moin Reference Griffin, Fu and Moin2021). Among the three near-equilibrium databases, Bobke et al. (Reference Bobke, Vinuesa, Örlü and Schlatter2017) and Pozuelo et al. (Reference Pozuelo, Li, Schlatter and Vinuesa2022) used the method of Vinuesa et al. (Reference Vinuesa, Bobke, Örlü and Schlatter2016) to determine TBL thickness based on diagnostic maps (Alfredsson, Segalini & Örlü Reference Alfredsson, Segalini and Örlü2011). The three non-equilibrium databases provided by Coleman et al. (Reference Coleman, Rumsey and Spalart2018) and Coleman (Reference Coleman2021) used the wall-normal integral of the spanwise vorticity to define the generalised velocity, then defined the thickness of the TBL from the wall to the position where the generalised velocity reaches 99.5 % of the generalised free flow velocity. The non-equilibrium database provided by Gungor et al. (Reference Gungor, Gungor and Maciel2024) computed the boundary layer thickness using the method of Griffin et al. (Reference Griffin, Fu and Moin2021). To demonstrate that different definitions of the TBL edge do not affect the results presented in the next section, in Appendix A, the TBL edge definition in DNS2018c is changed to that proposed by Wei & Knopp (Reference Wei and Knopp2023), and the analysis is performed.
3. Results
In this section, we first extend the relationship between the TBL edge velocity and the wall-normal integral to the non-equilibrium APG conditions. Then, using this relationship, an approximate equation for the Reynolds shear stress in non-equilibrium flow is given. Finally, the above results are applied to near-equilibrium APG TBLs and compared with results of Kumar & Mahesh (Reference Kumar and Mahesh2022). All the physical quantities used are in a non-dimensional manner according to the length and velocity scales in the corresponding databases. DNS2024 is not normalised.
3.1. Development of non-equilibrium APG TBLs
The APG causes changes in the velocity and Reynolds stress normal profiles of TBLs compared to ZPG conditions, and these profiles typically also vary with the streamwise direction. There is an inherent relationship between the changes in APG TBLs in these two directions. To understand this phenomenon, establishing a quantitative relationship between the physical quantities in these two directions is important. The displacement thickness
$\delta ^{*}$
and the momentum thickness
$\theta$
have always been important physical quantities in TBLs. They are obtained by wall-normal integral, and the variation of the velocity profile in the normal direction is reflected in these two physical quantities. In APG TBLs, the velocity defect and the boundary layer thickness increase, while the TBL edge velocity decreases, resulting in an increase in the displacement thickness. The degree of increase is influenced by the APG. Therefore, the displacement thickness and momentum thickness as streamwise direction functions reflect the development of APG TBLs to some extent. Their ratio, the shape factor
$H$
, is considered to reflect the degree of TBL distortion caused by the cumulative effect of APG (Maciel et al. Reference Maciel, Wei, Gungor and Simens2018).
As shown in table 2, wall-normal integrals were used to obtain the relationship between boundary layer edge velocity and friction velocity, as well as the relationship between Reynolds shear stress and mean velocity. The subscript ‘out’ indicates that the equation applies to the outer region of TBLs. This paper aims to complete this table based on a review of these existing results. It is easy to see that
$U_{e}V_{e}$
can be decomposed into two parts under equilibrium APG conditions (Wei et al. Reference Wei, Maciel and Klewicki2017). One is a ZPG part, and the other contains a pressure gradient term. The ZPG part
$Hu_{\tau }^{2}$
is consistent with the result obtained by Wei & Klewicki (Reference Wei and Klewicki2016) under the infinite Reynolds number condition. For finite Reynolds numbers, an additional term
$U_{e}^{2} ( {\textrm{d} \delta ^{*}}/{\textrm{d} x}-H\,{\textrm{d} \theta }/{\textrm{d} x} )$
should be added to the ZPG part because of the variation in the streamwise direction (Wei & Klewicki Reference Wei and Klewicki2023). Note the presence of the TBL edge-normal velocity
$V_{e}$
in these equations. The relationship between
$V_{e}$
and the integral of the TBL has been obtained for both ZPG and non-equilibrium APG TBLs (Wei et al. Reference Wei, Li, Knopp and Vinuesa2023a
,Reference Wei, Li and Wang
b
). The above equations are obtained by integrating the two-dimensional mean governing equation. The difference is whether pressure gradients or variations in the streamwise direction are neglected. Here, we summarise the above studies and extend the results of Wei et al. (Reference Wei, Maciel and Klewicki2017) to non-equilibrium APG TBLs.
Previous study of the TBL edge velocity and Reynolds shear stress.

For a two-dimensional TBL, the mean continuity equation can be expressed as
The streamwise mean momentum equation can be expressed as
Here, it is assumed that the normal pressure gradient is zero and that the Reynolds normal stress related term is neglected. This is true when the pressure gradient is insignificant, and the full mean momentum equation and discussion can be found in Devenport & Lowe (Reference Devenport and Lowe2022) and Wei, Li & Wang (Reference Wei, Li and Wang2024). We then make the equations non-dimensional without making any similarity assumptions, and the processing steps here are the same as those in Wei, Li & Wang (Reference Wei, Li and Wang2023b ). The normalised variables are
Here,
$L$
is a length scale. This causes some terms in the governing equations to become
Substituting (3.4) into the mean continuity equation yields
The first term is neglected for ZPG conditions, and the second term is neglected for self-similar flows. This allows the subsequent analysis to be extended to more conditions. Integrating (3.5) from the wall to the TBL edge, the results for some terms are
Then the wall-normal integral of (3.5) can be written as
This result corresponds to the second row of table 2, already obtained by Wei, Li & Wang (Reference Wei, Li and Wang2023b ) and Wei et al. (Reference Wei, Li, Knopp and Vinuesa2023a ). The reason for not further simplifying (3.7) is that it allows a more intuitive comparison with the results for the equilibrium TBLs.
Substituting (3.4) and (3.5) into the mean momentum equation yields
After the wall-normal integral, some terms are
Then the wall-normal integral of (3.8) can be written as
This result is a combination of the Kármán momentum integral and (3.7). By substituting one of the two into (3.10), the other can be derived. The streamwise derivative of the momentum thickness can be written in the form
Here,
$H$
is the shape factor. Substituting (3.7) and (3.11) into (3.10) yields
Equation (3.12) can fill the blank space in the first row of table 2 for non-equilibrium APG TBLs.
Under ZPG conditions, (3.12) returns to the equation in the first row and first column of table 2 (Wei & Klewicki Reference Wei and Klewicki2023). Under equilibrium conditions, (3.12) returns to the equation in the first row and second column of table 2 (Wei et al. Reference Wei, Maciel and Klewicki2017). Based on the above analysis, the relationship between the TBL edge velocity and wall-normal integral for non-equilibrium APG TBLs can be expressed as a combination of the ZPG term, the pressure gradient term and the streamwise variation term.
Equation (3.12) can be converted to
\begin{equation} \frac {u_{\tau }^{2}}{U_{e}V_{e}}=\underbrace {\frac {1}{H}}_{\text{term 1}}\underbrace {-\frac {\delta ^{*}+\theta + \frac{\delta }{H}}{\rho U_{e}V_{e}} \frac {\textrm{d} P_{e}}{\textrm{d} x}}_{\text{term 2}}\underbrace {+\frac {U_{e}}{V_{e}}\left ( \frac {\textrm{d} \theta }{\textrm{d} x} -\frac {1}{H} \frac {\textrm{d} \delta ^{*}}{\textrm{d} x} \right )}_{\text{term 3}}\!. \end{equation}
Maciel et al. (Reference Maciel, Wei, Gungor and Simens2018) proposed that
${U_{e}V_{e}}/u_{\tau }^{2}$
represents the ratio of the order of magnitude of the inertial force to the order of magnitude of the apparent turbulent force in the outer region when
$H\lt 1.6$
. In this paper, we consider
$u_{\tau }^{2}/ ( {U_{e}V_{e}} )$
to reflect the degree of scale separation. In (3.13), the term
${U_{e}V_{e}}/u_{\tau }^{2}$
is decomposed into three terms that quantify the effect of non-equilibrium conditions on scale separation relative to ZPG and near-equilibrium conditions. The classical description of scale separation is that the ratio of the outer length scale
$\delta$
to the inner length scale
$\nu /u_{\tau }$
increases with increasing Reynolds number (Klewicki Reference Klewicki2010). In the presence of the APG, the difference between the inner and outer region scales is not only influenced by the Reynolds number. It is known that
$u_{\tau }$
is a classical inner velocity scale, while
$U_{e}$
and
$V_{e}$
are used to measure velocities in the outer region and perform order-of-magnitude analysis (Wei & Maciel Reference Wei and Maciel2018). In the databases that we used,
$u_{\tau }^{2}$
decreases along the streamwise direction, and
$U_{e}V_{e}$
increases along the streamwise direction. Both are affected by the APG. Another outer scale option is
$U_{e}^{2}$
, but as shown in figure 1, it decreases along the streamwise direction. We believe that
$U_{e}V_{e}$
, which increases along the streamwise direction, may help to better reflect the different scales between the inner and outer regions. As there are many scales between the outer and inner regions, we do not exclude other better choices or forms of decomposition. A similar example is the different decompositions of the mean friction drag (Renard & Deck Reference Renard and Deck2016; Elnahhas & Johnson Reference Elnahhas and Johnson2022). The key is the physical interpretation of each decomposition.
In (3.13), term 1 is present under ZPG conditions, term 2 is caused by APG, and term 3 is caused by
${\partial U^{-}}/{\partial x^{*}}$
. Figure 4 illustrates how the three terms in (3.13) affect this scale separation. The verified pressure gradient parameter
$\beta$
ranges from 0.03 to 7.98, and
${Re}_\theta$
ranges from 868 to 5689. Term 2 (the pressure gradient term) plays a dominant role in the contribution of scale separation. In the range
$0\lt \beta \lt 1$
, it causes
$u_{\tau }^{2}/ ( {U_{e}V_{e}} )$
to decrease rapidly. The pressure gradient term is absent in ZPG TBLs. This suggests that there are differences between the inner and outer regions of APG TBLs compared with those of ZPG TBLs. Previous studies suggested that the inner region of APG TBLs tends to respond promptly to changes in the streamwise direction, while the outer region retains some of the upstream states (Knopp et al. Reference Knopp, Reuther, Novara, Schanz, Schülein, Schröder and Kähler2021; Devenport & Lowe Reference Devenport and Lowe2022; Parthasarathy & Saxton-Fox Reference Parthasarathy and Saxton-Fox2023; Gungor et al. Reference Gungor, Gungor and Maciel2024). This difference requires separate consideration of appropriate scaling for the inner and outer regions (Wei & Knopp Reference Wei and Knopp2023; Han et al. Reference Han, Ma and Yan2024).
Term 3 is called the streamwise variation term, which is absent at equilibrium flow. Although its value is relatively small compared with values of the other terms, its role in the non-equilibrium TBLs gradually increases as the APG or
${Re}_\theta$
increases. Its absolute value approaches
$u_{\tau }^{2}/ ( {U_{e}V_{e}} )$
under larger APG conditions. By comparison, we can see the effect of term 3 on the accuracy of (3.13). As
$\beta$
or
${Re}_\theta$
increases, the purple symbols (term 1 + term 2, suitable for equilibrium flow) have a lower accuracy than the black symbols. This intuitively demonstrates that not only the APG, but also the variation in the streamwise direction it induces, can have an effect on physical quantities. The streamwise variation term was first derived by Wei & Klewicki (Reference Wei and Klewicki2023). They used this term as a complement to the transition region in ZPG conditions. We show here that it is equally important for non-equilibrium APG TBLs.
The above decomposition helps to clarify how different effects are superimposed in non-equilibrium flow. The present study is not in a position to provide a further answer to the widespread history effect in non-equilibrium flows. The difficulty in studying the history effect lies in the need to reflect the cumulative effects of variations in the streamwise direction, while distinguishing them from local effects. In our study, the governing equations are integrated along the wall-normal for a fixed position. Because the governing equation contains the history effect to some degree (Devenport & Lowe Reference Devenport and Lowe2022), (3.13) can work, but cannot distinguish the history effect from local effects. Gungor et al. (Reference Gungor, Gungor and Maciel2024) proposed to distinguish between local pressure gradient effects, local non-equilibrium effects and history effects. We believe that this facilitates further analysis of the history effect. The ultimate solution to this problem requires further study.
Plots of
$u_{\tau }^{2}/( {U_{e}V_{e}} )$
decomposed into three terms. Symbols of the same shape belong to the same database. Sum means terms 1 + 2 + 3. (a) Plots as a function of
$\beta$
; (b) plots as a function of
${Re}_\theta$
.

3.2. Reynolds shear stress
Now, we turn to the study of the Reynolds shear stress
$-\langle uv\rangle$
. The relationship between the Reynolds shear stress and mean flow in table 2 shows a similar form of decomposition. The equation in the third row and second column of table 2 (Kumar & Mahesh Reference Kumar and Mahesh2022) can be simplified to the equation in the third row and first column under the ZPG condition (Wei, Li & Wang Reference Wei, Li and Wang2023b
). We consider providing an approximate equation for the Reynolds shear stress in non-equilibrium APG TBLs.
In the outer region of the TBLs, the viscous term in the mean momentum equation (3.2) can be neglected. The mean momentum equation is then integrated inwards from the TBL edge:
The subscript ‘out’ is the same as in table 2, indicating that the equation applies to the outer region of the TBL. This equation is then decomposed to obtain a part that is the same as shown in the third row and first column of table 2:
\begin{align} -\langle uv\rangle _{\textit{out}}&=\left ( u_{\tau }^{2}-\frac {u_{\tau }^{2}}{U_{e}V_{e}}\textit{UV} \right )+\left ( 1- \frac {u_{\tau }^{2}}{U_{e}V_{e}}\right ) \left ( U_{e}V_{e}-UV \right ) +2\int _{\delta }^{y } V\frac {\partial U}{\partial y} \,\textrm{d}y\nonumber\\&\quad +\frac {y-\delta }{\rho } \frac {\textrm{d} P_{e}}{\textrm{d} x}. \end{align}
By combining the third and fourth terms, we have
\begin{align} {-}\langle uv\rangle _{\textit{out}}&=\left ( u_{\tau }^{2}-\frac {u_{\tau }^{2}}{U_{e}V_{e}}\textit{UV} \right )+\int _{\delta }^{y } \left (\left ( \frac {u_{\tau }^{2}}{U_{e}V_{e}}+1\right )V\frac {\partial U}{\partial y}+\left ( \frac {u_{\tau }^{2}}{U_{e}V_{e}}-1\right )U\frac {\partial V}{\partial y}\right)\textrm{d}y\nonumber\\&\quad +\frac {y-\delta }{\rho } \frac {\textrm{d} P_{e}}{\textrm{d} x}. \end{align}
After processing, we convert the integral of the inertial term to the above form. The coefficients of both terms in the integral include
$u_{\tau }^{2}/ ( {U_{e}V_{e}} )$
, which is the start of the approximation. By substituting (3.13) into the integrated term of (3.16), it can be decomposed into eight terms:
\begin{align} \left (\! \frac {u_{\tau }^{2}}{U_{e}V_{e}}+1\!\right )V\frac {\partial U}{\partial y}&=\underbrace {\frac {1}{H}V\frac {\partial U}{\partial y}}_{\text{term 1}}\underbrace {{}-\frac {\delta ^{*}+\theta + \frac{\delta }{H}}{\rho U_{e}V_{e}} \frac {\textrm{d} P_{e}}{\textrm{d} x}V\frac {\partial U}{\partial y}}_{\text{term 2}}\underbrace {{}+\frac {U_{e}}{V_{e}}\left ( \frac {\textrm{d} \theta }{\textrm{d} x} -\frac {1}{H} \frac {\textrm{d} \delta ^{*}}{\textrm{d} x} \right )V\frac {\partial U}{\partial y}}_{\text{term 3}}\nonumber\\&\quad \underbrace {{}+V\frac {\partial U}{\partial y}}_{\text{term 4}},\end{align}
\begin{align} \left (\! \frac {u_{\tau }^{2}}{U_{e}V_{e}}-1\!\right )U\frac {\partial V}{\partial y}&=\underbrace {\frac {1}{H}U\frac {\partial V}{\partial y}}_{\text{term 5}}\underbrace {{}-\frac {\delta ^{*}+\theta + \frac{\delta }{H}}{\rho U_{e}V_{e}} \frac {\textrm{d} P_{e}}{\textrm{d} x}U\frac {\partial V}{\partial y}}_{\text{term 6}}\underbrace {{}+\frac {U_{e}}{V_{e}}\left ( \frac {\textrm{d} \theta }{\textrm{d} x} -\frac {1}{H} \frac {\textrm{d} \delta ^{*}}{\textrm{d} x} \right )U\frac {\partial V}{\partial y}}_{\text{term 7}}\nonumber\\&\quad \underbrace {{}-U\frac {\partial V}{\partial y}}_{\text{term 8}}.\end{align}
Normal distribution of the inertia term combination. Sum means terms 1 + 2 + 3 + 4 + 5 + 6 + 7 + 8. Plots for (a) DNS2018b, (b) DNS2018c, (c) DNS2021, (d) DNS2024.

These eight terms are numbered from 1 to 8. The distributions of their combinations in the normal direction of the TBLs is shown in figure 5. The results indicate that combination one (terms 1 + 2 + 3 + 4 + 6 + 7) approximately maintains a constant value over a large region of the outer boundary layer. Note that since the streamwise derivative cannot be calculated from the DNS2024 database, terms 1 + 2 + 3 + 4 + 6 + 7 is replaced by
$ ( ({u_{\tau }^{2}}/{U_{e}V_{e}})+1 )V ({\partial U}/{\partial y})+ ( ({u_{\tau }^{2}}/{U_{e}V_{e}})-( {1}/{H}) )U ({\partial V}/{\partial y})$
in figure 5(d) according to (3.17). By integrating this combination inwards from the TBL edge, the integral of the inertia term of (3.16) can be decomposed into a linear part
$a(y-\delta )$
and a nonlinear part. Rewriting it gives the approximate form
Parameter
$a$
can be chosen as the value of combination one (terms 1 + 2 + 3 + 4 + 6 + 7) at the TBL edge. The second term of (3.18) can be rewritten as
Equation (3.22) is compared with the Reynolds shear stress
$-\langle uv\rangle$
data: (a) DNS2018b, (b) DNS2018c, (c) DNS2021, (d) DNS2024.

Note that (3.19) still contains the integral term
$\int _{\delta }^{y } V ({\partial U}/{\partial y}) \,\textrm{d}y$
, which is consistent with the original (3.15). By combining these two equations and (3.18), the integral term
$\int _{\delta }^{y } V({\partial U}/{\partial y}) \,\textrm{d}y$
can be solved as
\begin{equation} \int _{\delta }^{y } V\frac {\partial U}{\partial y} \,\textrm{d}y\approx \frac { \frac {u_{\tau }^{2}}{U_{e} V_{e}}H-1}{H+1}\left (U_{e} V_{e}-U V\right )+\frac {aH}{H+1}(y-\delta ). \end{equation}
By substituting (3.20) into (3.15), we obtain an approximate equation for the Reynolds shear stress
$-\langle uv\rangle _{\textit{out}}$
in the outer region of non-equilibrium APG TBLs:
\begin{align} {-}\langle uv\rangle _{\textit{out}}&\approx \left ( u_{\tau }^{2}-\frac {u_{\tau }^{2}}{U_{e}V_{e}}\textit{UV} \right )+\left ( 1+ \frac {u_{\tau }^{2}}{U_{e}V_{e}}\right )\frac {H-1}{H+1} \left ( U_{e}V_{e}-\textit{UV} \kern1pt \right )\nonumber\\&\quad +\left ( \frac {2aH}{H+1} +\frac {1}{\rho } \frac {\textrm{d} P_{e}}{\textrm{d} x} \right )\left ( y-\delta \right ). \end{align}
Equation (3.21) also contains three terms. The first term is the same as the ZPG condition part (Wei et al. Reference Wei, Li and Wang2023b ), the second term includes the mean velocity that varies in the streamwise direction, and the third term includes the pressure gradient term. This equation can fill the blank space in the third row of table 2. Equation (3.21) can be simplified into a more concise form:
\begin{equation} {-}\langle uv\rangle _{\textit{out}}\approx \frac {\left ( \frac {2u_{\tau }^{2}}{U_{e}V_{e}}+1\right )H-1}{H+1} \left ( U_{e}V_{e}-\textit{UV} \right )+\left ( \frac {2aH}{H+1} +\frac {1}{\rho } \frac {\textrm{d} P_{e}}{\textrm{d} x} \right )\left ( y-\delta \right ). \end{equation}
Comparison of the Reynolds shear stresses in different databases.

Figure 6 shows the comparison between (3.22) and the Reynolds shear stress
$-\langle uv\rangle$
, where the maximum value of
$\beta$
is approximately unity. Equation (3.22) is closer to
$-\langle uv\rangle$
in most outer regions of the TBLs, whereas a further approximation is required in the inner region. Notably, (3.22) is zero at the TBL edge, but this is not the true reality. This is due to the APG TBL edge
$ {\partial U}/{\partial y}\neq 0$
(Pozuelo et al. Reference Pozuelo, Li, Schlatter and Vinuesa2022; Wei et al. Reference Wei, Li, Knopp and Vinuesa2023a
), which is simultaneously
$-\langle uv\rangle \neq 0$
(Wei & Knopp Reference Wei and Knopp2023). These small differences make it difficult to provide a more accurate equation up to the APG TBL edge. Figure 7 shows the effects of different Reynolds numbers on the
$-\langle uv\rangle$
profile. Although the outer-scaled profiles from the same database are similar within the pressure gradient range mentioned above, an increase in the Reynolds number results in a decrease in the value of the Reynolds shear stress. Changes in the Reynolds number within the databases do not affect the accuracy of the results.
Wei et al. (Reference Wei, Li, Knopp and Vinuesa2023a
) presented the relationship between the mean normal velocity
$V$
and the mean streamwise velocity
$U$
in APG TBLs. By substitution, (3.22) can be transformed to include only the mean streamwise velocity
$U$
, but the form is complex. Whether a concise form for
$-\langle uv\rangle$
and
$U$
could be derived, as Wei, Li & Wang (Reference Wei, Li and Wang2023b
) did for ZPG TBLs, remains to be investigated. We have demonstrated the possibility of decomposing the Reynolds shear stress
$-\langle uv\rangle$
in the outer region of the APG TBL. This helps to combine the study of mean velocity with the Reynolds shear stress. We cannot rule out the possibility of a better approximate equation. Further discussion of the results follows in the next section.
3.3. Application to the near-equilibrium flow
To test the applicability of the results to a wide range of APG TBLs, we used near-equilibrium databases with a nearly constant
$\beta$
. For this type of flow, the streamwise variation term can be approximately neglected. The changes in the results from the first two subsections need to be considered.
Decomposition of
$u_\tau^2/(U_eV_e)$
into two terms: b1,
$\beta =0.96$
; b2,
$\beta =2.07$
; b1.4,
$\beta =1.41$
.

When there is no streamwise variation term, (3.7) becomes
This equation can fill the blank space in the second row of table 2. The corresponding mean momentum equation is integrated as
The implicit condition of these two integrals is
Substituting (3.25) into (3.23) gives the same form as the simplified (3.7). Therefore, the analytical equation for
$V_{e}$
under APG conditions can be uniformly expressed in the simplified form of (3.7) (Wei et al. Reference Wei, Li, Knopp and Vinuesa2023a
). Substituting the Kármán momentum integral and (3.25) into (3.24) also yields (3.23). Equation (3.24) is now converted into a form similar to (3.12):
This is consistent with Wei et al. (Reference Wei, Maciel and Klewicki2017). It also can be converted to
\begin{equation} \frac {u_{\tau }^{2}}{U_{e}V_{e}}=\underbrace {\frac {1}{H}}_{\text{term 1}}\underbrace {{}-\frac {\delta ^{*}+\theta +\frac{\delta }{H}}{\rho U_{e}V_{e}} \frac {\textrm{d} P_{e}}{\textrm{d} x}}_{\text{term 2}}. \end{equation}
Figure 8 shows the results. Three databases with different Reynolds numbers and APGs demonstrate the effectiveness of (3.27). Compared with the non-equilibrium flow in figure 4, the pressure gradient term here does not change much. It can only be confirmed that the APG accelerates scale separation between the inner and outer regions of TBLs compared with ZPG conditions. The scale referred to here is the same as in the previous subsection, with
$u_{\tau }^{2}$
and
$U_{e}V_{e}$
used for the inner and outer regions, respectively, and
$u_{\tau }^{2}/ ( {U_{e}V_{e}} )$
representing the degree of scale separation. However, due to complex properties such as history effects in non-equilibrium flow, it is difficult to strictly compare the contribution of the streamwise variation term relative to the equilibrium flow under the same conditions (Gungor et al. Reference Gungor, Gungor and Maciel2024).
For the Reynolds shear stress profile, an approximate equation can be obtained via the same method, where the decomposition of (3.17) becomes six terms. According to (3.27), there is no streamwise variation term. This means that the value of parameter
$a$
in (3.18) becomes the value of terms 1 + 2 + 4 + 6 at the boundary layer edge, while the form of (3.22) remains unchanged. As shown in figure 9, the approximate equation is also applicable to the near-equilibrium APG TBLs. Compared with figure 6, the approximate equation is applicable to a smaller range of outer regions. This emphasises that although the APG causes similar changes in equilibrium and non-equilibrium flows, the extension of any conclusion should be carefully considered, and the applicable APG and Reynolds number range should be clearly defined. The equation given by Kumar & Mahesh (Reference Kumar and Mahesh2022) is used for comparison:
As shown in figure 9, there is a difference between the two approximate equations near the wall. The main reason for the difference between the approximate equations and the Reynolds shear stress
$-\langle uv\rangle$
in the outer region is that
$-\langle uv\rangle \neq 0$
at the TBL edge.
Equation (3.28) can be rewritten in a form similar to (3.22):
This is related to the theme of this paper, which is that the integral analysis shows that the Reynolds shear stress
$-\langle uv\rangle$
can be decomposed into a mean velocity related term and a pressure gradient term in a given
$\beta$
range. As shown in (3.22) and (3.29), the coefficients of these two terms vary according to different approximations.
4. Discussion
As the classical and most widely used pressure gradient parameter,
$\beta$
can be considered as the ratio of the time scale of the TBL outer region to the potential flow (Maciel et al. Reference Maciel, Rossignol and Lemay2006; Romero et al. Reference Romero, Zimmerman, Philip, White and Klewicki2022). It can also be derived from the ratio of the additional pressure on the right-hand side of the Kármán momentum integral equation to the wall shear stress (Devenport & Lowe Reference Devenport and Lowe2022). Wei et al. (Reference Wei, Li and Wang2024) discussed the influence of the APG on the Kármán momentum integral, and noted that the error in the Kármán momentum integral comes from the boundary layer approximation under APG conditions. This is relevant to our study because the Kármán momentum integral is used in the derivation of (3.13).
The non-dimensional physical quantity
$u_{\tau }^{2}/ ( {U_{e}V_{e}} )$
in (3.13) can also be obtained via (3.7):
\begin{equation} \frac {u_{\tau }^{2} }{U_{e}V_{e}} =\frac {u_{\tau }^{2}}{U_{e}^{2}\left(\frac{\textrm{d} \delta ^{*}}{\textrm{d} x} \right)+\left ( \delta ^{*}-\delta \right ) U_{e} \left(\frac{\textrm{d} U_{e}}{\textrm{d} x} \right)} . \end{equation}
This avoids the use of the Kármán momentum integral. Comparing this equation with (3.13), as shown in figure 10, both are close to the
$u_{\tau }^{2}/ ( {U_{e}V_{e}} )$
data under the given pressure gradient. Although Wei et al. (Reference Wei, Li and Wang2024) reported that the accuracy of the Kármán momentum integral decreases with increasing APG in non-equilibrium APG TBLs,
$u_{\tau }^{2}/ ( {U_{e}V_{e}} )$
also decreases with increasing APG. The derivation of (3.13) includes the integral of the mean continuity equation. If the
$U_{e}V_{e}$
calculation is more accurate under APG conditions, then this reduction in ratio allows (3.13), which implicitly uses the Kármán momentum integral, to be applied to a wider range of APGs.
The ratio of the response distances to non-equilibrium conditions to the boundary layer thickness as a function of
$\beta$
.

We now return to (3.22). Note that (3.22) has
$\beta \approx 1$
as the upper limit for application. Devenport & Lowe (Reference Devenport and Lowe2022) suggested that for non-equilibrium APG TBLs, attention needs to be paid to the response distance of the boundary layer to non-equilibrium changes in pressure gradient or wall shear stress. They proposed a length scale
$L_{\tau }= \theta U_{e}^{2}/u_{\tau }^{2}$
for the boundary layer defect to be re-made by the wall shear stress, and a length scale
$L_{P}= \theta U_{e}^{2}/ ( ({\delta ^{*}}/{\rho }) ({\textrm{d} P_{e}}/{\textrm{d} x} ))$
for the boundary layer defect to be re-made by the pressure gradient. The discussion here is about mean flow, and Devenport & Lowe (Reference Devenport and Lowe2022) proposed a different length scale for the turbulent structure. Figure 11 shows the ratio of these two scales to the boundary layer thickness as a function of pressure gradient and Reynolds number. As the APG increases,
$L_{\tau }/\delta$
gradually increases, while
$L_{P}/\delta$
gradually decreases. This indicates that an increase in APG accelerates the response of the boundary layer to the pressure gradient while slowing the response to wall shear. The boundary layer state changes from being dominated by wall shear under ZPG conditions to being dominated by the pressure gradient. The effect of the Reynolds number is reflected in the fact that both response distances increase with increasing Reynolds number. This indicates that the enhancement of the inertial effect leads to a delay in the response of the boundary layer to changes in wall shear stress and pressure gradient. Han et al. (Reference Han, Ma and Yan2024) considered that
$\delta /L_{P}$
can play a role in the measurement of history effects, and extended it to the Zagarola–Smits scaling for non-equilibrium flow. In the recent study by Gungor et al. (Reference Gungor, Gungor and Maciel2024), they also expressed interest in the response distance of the boundary layer to non-equilibrium conditions. When
$0\lt \beta \lt 1$
,
$L_{P}/\delta$
changes rapidly with
$\beta$
. As shown in figure 4, the contribution of the pressure gradient term to the scale separation increases rapidly as
$0\lt \beta \lt 1$
. These analyses highlight the underlying physical laws behind the use of
$\beta \approx 1$
as the upper limit of (3.22), where for non-equilibrium flow, the response distance of the boundary layer to changes in wall shear and pressure gradient is the same. It is approximately 60–80 times the boundary layer thickness within the Reynolds number range of the data used. Although the current results are based on the databases of small velocity defect and limited Reynolds number ranges, the properties of the non-equilibrium flow are already evident as described above. We try to take some of these properties into account in the derivation of (3.22). This is reflected in the importance of the streamwise variation term in the previous section.
Devenport & Lowe (Reference Devenport and Lowe2022) described
$\beta =1$
as a moderate pressure gradient. For non-equilibrium APG TBLs with a larger
$\beta$
, there may also be an approximate decomposition of the inertia term similar to (3.18). However, the key is that the neglected terms in the momentum (3.2) also play a role in this condition (Wei et al. Reference Wei, Li and Wang2024). The Reynolds shear stress profile approximation based on the integral of the momentum equation can no longer neglect these terms. This requires further study in the future. For near-equilibrium flow, the restriction on
$\beta$
can be relaxed appropriately because of the self-similarity of the flow in the streamwise direction.
5. Conclusion
The wall-normal integral is defined as the integration of physical quantities along the normal direction, starting from the wall or boundary layer edge for a flat plate. The main result of this study is the use of the wall-normal integral to supplement the missing entries in table 2. Specifically, the analytical equation for
$V_{e}$
proposed by Wei et al. (Reference Wei, Li, Knopp and Vinuesa2023a
) is re-derived. By combining it with the Kármán momentum integral, the relationship between
$U_{e}V_{e}$
and
$u_{\tau }^{2}$
obtained by Wei et al. (Reference Wei, Maciel and Klewicki2017) under near-equilibrium conditions is extended to non-equilibrium conditions. The above derivation decomposes
$u_{\tau }^{2}/ ( {U_{e}V_{e}} )$
, which describes the scale separation in this paper, into three terms. It includes a ZPG term, a pressure gradient term, and a streamwise variation term. The pressure gradient term significantly reduces
$u_{\tau }^{2}/ ( {U_{e}V_{e}} )$
and promotes scale separation. Previously, Wei & Klewicki (Reference Wei and Klewicki2023) proposed adding the streamwise variation term to the wall-normal integral analysis of ZPG TBLs under finite Reynolds number conditions. Our study reveals that in non-equilibrium flows, the contribution of the streamwise variation term to scale separation increases as the APG increases. Then, substituting the decomposition of
$u_{\tau }^{2}/ ( {U_{e}V_{e}} )$
into the integral of the inertia term of the mean momentum equation, it is found that the combination of some of its terms approximates a constant over a wide range in the outer region of the TBLs. An analytical (3.22) for the Reynolds shear stress in the outer region with respect to the mean velocity is obtained by this approximation. This result was not obtained in previous studies of non-equilibrium APG TBLs. It has
$\beta \approx 1$
as the upper limit for application. The analysis of the response distance indicates that the non-equilibrium effect is already evident when
$\beta =1$
.
The above results focus on the unique nature of the non-equilibrium flow. By distinguishing the effects of the streamwise variation term, it is shown that it plays a non-negligible role in obtaining quantitative results. Gungor et al. (Reference Gungor, Gungor and Maciel2024) suggested distinguishing between local pressure gradient effects, local non-equilibrium effects and history effects in non-equilibrium APG TBLs. The governing equation contains the history effect to some degree. Therefore, the wall-normal integral plays a role in this study. However, it cannot yet distinguish between the history effect and local effects. It is expected that this study, which focuses on limited Reynolds numbers and small velocity defect, will be verified or improved under more complex conditions in the future.
Acknowledgements
We would like to acknowledge the authors of the data used in this paper, and to express our sincere gratitude to the anonymous reviewers for their efforts in improving the quality of the paper.
Funding
This study is supported by a grant from the National Natural Science Foundation of China (grant no. 92252201).
Declaration of interests
The authors report no conflict of interest.
Appendix A. Definition of the TBL edge
The definition of the TBL edge becomes more important and complex for APG TBLs (Vinuesa et al. Reference Vinuesa, Bobke, Örlü and Schlatter2016; Griffin et al. Reference Griffin, Fu and Moin2021). In their recent study, Wei & Knopp (Reference Wei and Knopp2023) proposed to define the TBL edge as the point where the Reynolds shear stress decreases to 1 % or 5 % of its maximum value, depending on the spatial resolution. Wei et al. (Reference Wei, Li, Knopp and Vinuesa2023a
) compared the aforementioned definition with the one proposed by Vinuesa et al. (Reference Vinuesa, Bobke, Örlü and Schlatter2016). They found that the different definitions affect the TBL thickness, but show less sensitivity in terms of TBL edge velocity and displacement thickness. Their proposed analytical expression of
$V_{e}$
, shown in the second row and third column of table 2 of this paper, is not affected by the definition of the TBL edge. This expression is also used in this study.
Results of this study after changing the definition of the TBL edge; data from DNS2018c. (a) Comparison of boundary layer edges. (b) Decomposition of
$u_{\tau }^{2}/( {U_{e}V_{e}} )$
into three terms. (c) Normal distribution of the inertia term combination. (d) Equation (3.22) compared with the Reynolds shear stress
$-\langle uv\rangle$
data.

We change the original TBL edge definition in DNS2018c to the position where the Reynolds shear stress decreases to 5 % of its maximum value. As shown in figure 12(a), the definition proposed by Wei & Knopp (Reference Wei and Knopp2023) results in a more inward boundary layer edge compared to the original definition in DNS2018c. The difference between the two definitions tends to increase as
$\beta$
increases. Although the two definitions result in different boundary layer edges, the results of this study are not sensitive to this. Comparing figure 12(b) with figure 4, it can be seen that changing the definition of the TBL edge does not affect the accuracy of (3.13). Note that the values on both sides of (3.13) have changed. Comparing figure 12(c) with figure 5, it can be seen that the specific combinations of terms maintain a constant value over a wide range. Comparing figure 12(d) with figure 6, it can be seen that (3.22) can still predict the Reynolds shear stress in the outer region. At this point, the Reynolds shear stress at the edge of the boundary layer is still not zero, resulting in a discrepancy with (3.22). Therefore, for the small velocity defect range focused on in this study, the results are not sensitive to different definitions. The differences introduced by the definitions of TBL edge increase as the APG increases (Wei et al. Reference Wei, Li, Knopp and Vinuesa2023a
), at which point the study needs to be more cautious.






























