Non-Newtonian Fluid Seminar 01 · Universität Heidelberg

This post is the second part of the script of my seminar talk From Maxwell’s Fluid to the Oldroyd-B Model, translated from Typst to Markdown. The series:

  1. From Maxwell's Fluid to Oldroyd-B (I): Maxwell's Fluid
  2. Objective Derivatives (this post)
  3. From Maxwell's Fluid to Oldroyd-B (III): Oldroyd-B Model and Weissenberg Effect

In the last post we saw that the tensorial Maxwell model σ+λσ˙=2ηD fails: the material derivative σ˙ is not frame invariant, so different observers see different dynamics. We need a time derivative that every observer agrees on.

1. A Boat Ride to Objective Derivatives

For the introduction to objective derivatives, let’s take a boat ride. At time t=0 , Alice and Bob stand in a laboratory on a lake, at the point X . They want to observe some quantity A^ . To them, a tensor such as A^ is a magical black box: it eats two vectors and spits out a number. Their only tools are the arrows of an orthonormal Cartesian basis {ei} , painted on the surface of the water. To measure A^ , they feed it the pair ei,ej and record

(1.1)A^{ei};ij=⟨ei,A^ej⟩.

This is the observation of A^ in {ei} .

Then Bob goes for a boat ride. He does not row; he simply drifts with the water. After a time t he arrives at x=φ(X,t) (Figure 1). The arrows painted on the water have drifted with him, but the flowing lake has stretched and sheared them into

(1.2)gi=Fei,F=∂φ∂X,

which are no longer orthonormal. To read off components with such arrows, Bob also needs the dual basis gi=F−Tei , defined by ⟨gi,gj⟩=δij .

Figure 1. Left: the basis eᵢ painted on the water at X. Right: after drifting to x, Bob finds the arrows carried and deformed by the flow, gᵢ = F eᵢ (blue), together with their dual basis gⁱ = F⁻ᵀ eᵢ (red), which is normal to the deformed material lines.
Figure 1. Left: the basis eᵢ painted on the water at X. Right: after drifting to x, Bob finds the arrows carried and deformed by the flow, gᵢ = F eᵢ (blue), together with their dual basis gⁱ = F⁻ᵀ eᵢ (red), which is normal to the deformed material lines.

2. Push-Forward and Pull-Back

To Bob, the quantity now looks like a different tensor A , although it is essentially the same thing. Bob therefore insists that the observation be preserved: measured with his deformed arrows, A must give the same numbers that A^ gave Alice,

(2.1)A^{ei};ij=A{gi};ij=⟨gi,Agj⟩.

Writing gi=F−Tei and moving F into the ket,

(2.2)⟨gi,Agj⟩=⟨F−Tei,AF−Tej⟩=⟨ei,F−1AF−Tej⟩.

Comparing with Alice’s observation, Bob finds A=FA^FT .

Push-Forward and Pull-Back

The map

(2.3)Φ(⋅):=F(⋅)FT

is called the push-forward, and

(2.4)Φ−1(⋅):=F−1(⋅)F−T

the pull-back.

Since F is exactly how line elements are carried by the flow, this form tells us that A changes the way line elements do, just like a polymer chain stretched by the flow.

3. Upper-Convected Derivative

Now Bob wants to know how fast A changes. His idea is simple: go back to the one fixed frame he shares with Alice. He pulls A back to A^ in {ei} , takes the time derivative there, and pushes the result forward. With the velocity gradient L=∇v , the material derivative A˙ , and F˙=LF , ddtF−1=−F−1L , this gives the upper-convected derivative:

Upper-Convected Derivative (UCD)
(3.1)A∇=Φ(ddtΦ−1(A))=F(ddt(F−1AF−T))FT=F(−F−1LAF−T+F−1A˙F−T−F−1ALTF−T)FT=A˙−LA−ALT.

In short: pull back, differentiate, push forward. The UCD measures change relative to the flow.

4. Lower-Convected Derivative

With different choices of the push-forward and pull-back, we can define other objective derivatives. For example, the lower-convected derivative is defined by

Lower-Convected Derivative (LCD)
(4.1)AΔ=A˙+LTA+AL.

The same story produces it. Suppose Bob had measured with the carried arrows gi instead of the dual ones, and insisted on ⟨gi,Agj⟩=A^{ei};ij . Since ⟨gi,Agj⟩=⟨ei,FTAFej⟩ , he would find A=F−TA^F−1 . Pulling back, differentiating with ddtFT=FTLT , and pushing forward gives

(4.2)F−T(ddt(FTAF))F−1=A˙+LTA+AL=AΔ.
UCD vs. LCD

The dual arrows gi are normal to the deformed material lines, so the LCD carries A like normals of material surfaces, while the UCD carries it like line elements. A polymer chain is a line element; this is the microscopic reason for choosing the UCD.

With the UCD at hand, we are ready to repair the Maxwell model in the next post.