701号时钟草 · Bilibili · 18:45
This clip introduces the concept of the Jacobian matrix through an intuitive geometric example. The video first draws an irregularly shaped region in the xy-plane and points out that directly computing its area—i.e., evaluating the double integral of 1 over that region—is difficult. Next, it draws a simple rectangular region D in the uv-plane and establishes a coordinate transformation T from the uv-plane to the xy-plane. By representing the complicated region as the image of the simple region D under transformation T, the video raises a central question: How can we convert the area integral over into an integral over D? It notes that the integrand after conversion is precisely the focus of the upcoming discussion. This video explains the idea of coordinate transformation in double integrals using a whiteboard. It begins by writing dxdy = ? dudv, raising the issue of converting the integration region from the xy-plane to the uv-plane. Then, using the region bounded by xy=q, xy=p, y=ax, and y=bx as an example, it sets u=xy and , transforming the original curved-sided region into the rectangle [p,q]×[a,b]. Next, the video asks how area elements relate before and after the transformation, and moves to the conceptual level: affine transformations map lines to lines, whereas curvilinear transformations do not necessarily do so; thus, curvilinear transformations generalize affine transformations. Finally, it writes a general curvilinear transformation as T: , (u,v)↦(x,y), where and , and gives the example . This segment introduces differentiability for transformations from two dimensions to two dimensions. The explanation first states that the domain and codomain of such transformations are both the two dimensional plane, and drawing a complete graph would require four dimensional space, so instead we represent the transformation using "two boxes plus an arrow". Then points on the plane are uniformly viewed as column vectors, and we take x vector sub zero in the domain and its image y vector sub zero.0 and its image y vector sub zero,0 then use small increments delta x vector and delta y vector to convert the problem from "region to region" to "how local vectors change". Next we review the one variable case where "the differential is the linear principal part of the function increment", writing delta y equals A times delta x plus little o of delta x, and analogously obtain delta y vector equals A times delta x vector plus little o of delta x vector. Finally we explain that A must be a two by two matrix, and little o of delta x vector is a vector valued infinitesimal remainder term satisfying that when divided by the magnitude of delta x vector it tends to the zero vector. This segment uses a whiteboard to explain differentials and derivatives of two-dimensional vector-valued functions. First, define 'differentiability' via , denote the linear principal part as , and thus obtain , emphasizing that the derivative A of a vector-valued function is a matrix. Then expand and componentwise into (du, dv)^T and (dx, dy)^T, use matrix multiplication to write the total differential forms of du and dv, fill in ∂u/∂x, ∂u/∂y, ∂v/∂x, ∂v/∂y in order, and obtain the Jacobian matrix J. Finally, explain that when J is a square matrix, its determinant can be taken — this is the Jacobian determinant — and begin introducing the special notation ∂(u,v)/∂(x,y). The video first reviews the definition of the two-dimensional Jacobian determinant, emphasizing how the ordering of variables affects the determinant's value. Then, through the vivid 'grass-growing on a plane' metaphor, it intuitively illustrates that a vector-valued function assigns a derivative matrix to each point. Next, the theory is generalized to higher-dimensional spaces, noting that a mapping from to has a derivative that is an matrix. Finally, it examines the special case where the output dimension , i.e., the Jacobian matrix of a multivariable scalar function appears as a row vector of partial derivatives. This segment first points out that when the codomain dimension of a multivariable function is 1, its Jacobian matrix degenerates into the gradient vector, and both can be viewed as the derivative of the function. Then, using a geometric diagram illustrating the coordinate transformation from the UV-plane to the XY-plane, it intuitively derives the area element transformation relation in the change-of-variables formula for multiple integrals: the local infinitesimal rectangular area dudv maps to a parallelogram area dxdy equal to the original area multiplied by the absolute value of the Jacobian determinant, i.e., dxdy = |J| dudv. This segment focuses on area elements under curve transformations. First, present dxdy = |∂(x,y)/∂(u,v)| dudv and stress that comparing areas requires the absolute value of the Jacobian determinant. Then integrate the area element to obtain dxdy = |∂(x,y)/∂(u,v)| dudv, thereby answering the earlier question "What is the integrand?" Next, generalize the conclusion to the general double integral substitution formula dxdy = |∂(x,y)/∂(u,v)| dudv, and draw an analogy with the single-variable substitution dx = (dx/du)du. Finally, note that this theorem extends to triple integrals and n-fold integrals. This segment primarily introduces three commonly used coordinate transformations and their Jacobian determinants: polar coordinate transformation (Jacobian determinant is r), cylindrical coordinate transformation (Jacobian determinant is r), and spherical coordinate transformation (Jacobian determinant is 〔r² sin φ〕). The video clearly presents the mathematical expressions and geometric meanings of these transformations through formulas written on a whiteboard and simple three-dimensional sketches.
Use the learning inspector for key ideas and moments, or open the reading tabs for the complete notes.
Generated from the video's visuals and explanation; not verbatim speech.
At the beginning, the video states its goal: to compute the area of an irregularly shaped region in the xy-plane. The most natural method is to evaluate the double integral of 1 over that region.
However, because the boundary of this region is complicated, setting up integration limits directly in the xy-plane would be difficult. Therefore, the instructor introduces a change of perspective: finding a simpler coordinate system.
On the left side of the screen, draw the uv-plane and sketch a simple rectangular region D inside it. Then, establish a mapping from the uv-plane to the xy-plane, i.e., the coordinate transformation and .
The instructor explains the meaning of this transformation: for any point in the uv-plane, substituting it into the transformation formulas yields a unique corresponding point in the xy-plane.
Denote the entire transformation as . Under this transformation, the rectangular region in the uv-plane is mapped onto an irregular region in the xy-plane, denoted .
Assuming region and transformation are well-behaved, then is also a region suitable for integration. The current goal becomes computing the area of .
By the definition of area, we need to compute . But to take advantage of the simple shape of , we wish to convert this integral into an integral over , namely .
The video pauses here with suspense: what exactly is the transformed integrand '?'? This is precisely the key question that leads to the Jacobian determinant.
At the beginning of the video, the problem is framed within the context of change of variables in double integrals: the equation displayed is dxdy = ?\,dudv, meaning that the integration region in the xy-plane is to be transformed via some mapping T into the uv-plane, but the scaling factor relating the area elements must still be determined.
To illustrate the effect of such a transformation, the video first draws two hyperbolas, xy = q and xy = p, on the right-hand xy-plane, with the condition ; it then draws two rays, y = ax and y = bx, with . These four curves enclose a curvilinear region, and the video states that direct integration over this region would be difficult to handle.
Next, the uv-plane is set up on the left, and new variables are defined as u = xy and . The motivation is straightforward: u corresponds exactly to the constant in the family of hyperbolas xy = constant, and v corresponds exactly to the slope in the family of rays y = (slope) . Thus, the original boundaries xy = p and xy = q become and , while the boundaries y = ax and y = bx become and .
From the diagram, it is clear that the complicated region bounded by hyperbolas and rays on the right becomes the rectangle [p,q] on the left. The video uses this to emphasize that the primary benefit of a coordinate transformation is simplifying both the limits of integration and the shape of the region into a form more amenable to computation.
However, once the region becomes simpler, a new issue arises: the area elements dxdy and dudv before and after the transformation are not obviously equal. At this point, the video shifts attention from "region shape" to "area relationship", laying the groundwork for introducing the Jacobian factor.
Next comes a conceptual clarification. The video first recalls affine transformations from linear algebra and explicitly states that affine transformations map straight lines in the plane to straight lines. In contrast, curved transformations do not preserve this linearity; for instance, the straight boundaries in the uv-plane correspond in the xy-plane to curves such as hyperbolas or rays.
Therefore, the video characterizes curved transformations as a generalization of affine transformations: affine transformations are a special class of planar mappings, whereas curved transformations allow more general, bent correspondences. This step is significant because, when later discussing area scaling, one cannot rely solely on the linear-algebraic intuition for affine maps.
Then the video gives the general definition: a curved transformation on the plane is written as T:\begin{cases}\\y = This formula needs review. Please report it using the page feedback control., which is essentially T:\mathbb{R}^, mapping a point (u,v) in the uv-plane to a point (x,y) in the xy-plane. The video adds that the domain and codomain may each be taken as subregions of the plane.
Finally, the video presents a concrete example: = (,\ ), using it to reiterate that a coordinate transformation takes points in the plane as input and outputs points in the plane. This segment ends here, without yet presenting the formal formula linking area elements to the Jacobian determinant.
The central question of this segment is: given a transformation from the two-dimensional plane to the two-dimensional plane, we cannot grasp its overall behavior directly via ordinary graphs, as we do for single-variable functions. The board first displays and writes down . The explanation then points out that both the domain and the codomain are two-dimensional planes; to draw the full graph directly would require observing in four-dimensional space, so we need an alternative approach.
To visualize abstract transformations, the explanation draws two rectangular boxes connected by a central arrow , with both boxes labeled underneath. Mathematically, the left box represents the domain, the right box the codomain, and the arrow represents the transformation itself. Thus, the problem of "function graphs" is recast as the problem of "correspondence between two planes".
Next, the explanation stipulates that points on the plane are all represented as vectors—and specifically as column vectors. This convention may seem merely a notational choice, but it actually affects subsequent algebraic structure: since points are column vectors, objects that linearly map one vector to another must be written on the left—that is, matrices multiply vectors from the left.
Select a point in the left box and its image in the right box. This step first establishes a point-to-point correspondence. Then the explanation raises a more crucial question: knowing only how a single point maps to a single point is insufficient for discussing region-to-region correspondence; therefore, we must further examine how the image point changes when a small displacement occurs near a given point.
We thus introduce the increment vector. If moves a small distance to , then its image changes from to . The two arrows in the diagram, left and right, represent and respectively. The core idea here is to reduce a region-level change problem to a local vector increment problem.
To characterize this local increment relationship, the explanation first reviews the differential concept for single-variable functions and writes down . The verbal definition is: "the differential is the linear principal part of the function increment", meaning that the dependent variable's increment is approximated by a linear expression in the independent variable's increment. The structure of this formula is "linear principal part + higher-order remainder".
Next, this same structure is extended to the two-dimensional vector case, yielding . This step is not a rigorous proof but an analogy-based generalization: if a single-variable increment can be written as a linear principal part plus a higher-order term, then a two-dimensional vector increment can likewise be attempted in a similar form. The board also notes that both and are "2-dimensional vectors".
Then the explanation addresses the first question arising from this analogous form: what is ? The answer given is that we need an object that maps two-dimensional vectors linearly to two-dimensional vectors—and such an object is a "second-order matrix". Therefore, is no longer an ordinary scalar constant, but rather a linear mapping matrix acting on a column vector. This also explains why the matrix is written on the left.
Finally, the explanation handles the second question: how to interpret . It emphasizes that since the entire left-hand side of the equation is a vector and the linear principal part is also a vector, the remainder must likewise be a vector-valued term—not a scalar infinitesimal. Its "infinitesimal" meaning is: when this remainder vector is divided by , the result tends to the zero vector. By this point, the segment has prepared all three elements needed for the formal definition of differentiability: vector increment, linear principal part matrix, and vector-valued higher-order remainder.
The video first gives the definition of differentiability for a vector-valued function: if , where A is independent of , then the function is said to be differentiable at that point. This structure mirrors single-variable differentiation, but upgrades the increments, independent variables, and linear coefficient to vectors and matrices.
Next, the video highlights as the linear principal part using a red circle, and defines it as the differential at that point. It also denotes the increment of the independent variable as , yielding the key relation , which can formally be written as .
From this, the video explicitly states that the "derivative" of a vector-valued function is not a scalar, but rather this matrix A. This step is the crucial conceptual shift; all subsequent content aims to write A explicitly.
To find the entries of A, the explanation expands and componentwise into column vectors: =(du,dv)^T, . Thus the matrix equation becomes a problem of comparing coefficients in two scalar equations.
After expanding via matrix multiplication, du and dv each appear as linear combinations of dx and dy. The video points out that this is precisely the familiar form of the total differential for multivariable functions, so the four unknown coefficients must correspond to the four first-order partial derivatives.
Then, , , , and are filled in order, and arranged into an matrix. This matrix is the derivative matrix at that point, named the Jacobian matrix J.
Finally, the video further explains that when the Jacobian matrix is square, its determinant can be taken—this is the Jacobian determinant. The blackboard writes this determinant in blue and introduces the special notation , emphasizing its importance in later applications.
When constructing the Jacobian determinant , the dependent variables are placed in the numerator and the independent variables in the denominator. This order must be strictly observed, because determinants are antisymmetric: swapping any two rows or two columns changes the sign of the determinant. Thus, the ordering of variables directly determines the accuracy of the result.
Next, we introduce the concepts of differential and derivative for vector-valued functions. Although we discuss this at a single point, each point in the domain corresponds to a local linear transformation (matrix). To gain geometric intuition, imagine that at every point in the two-dimensional plane, a small line segment or matrix—representing the derivative at that point—'grows', just like grass growing on a field; this vividly depicts the distribution of a matrix field.
Generalize the perspective to higher dimensions and consider the mapping . This mapping transforms dimensional vectors into dimensional vectors. To express the output increment in terms of the input increment , we need a linear operator. By the definition of differentiation , this linear operator must be a matrix, which is the derivative of the vector-valued function (the Jacobian matrix).
Finally, we examine a special case: when the output dimension is , the mapping degenerates into a multivariate scalar function, such as . In this case, its Jacobian matrix is no longer square, but rather a -row vector. The entries of this row are the partial derivatives of the function with respect to each independent variable, in order: . This illustrates the matrix-based foundation of the gradient concept for multivariate functions.
In multivariable calculus, when the dimension of the target space of a mapping is one the Jacobian matrix of the function contains only one row of partial derivatives. In this case, the Jacobian matrix formally coincides exactly with the gradient vector of the function as shown on the video blackboard This means that for a scalar-valued multivariable function, the gradient is essentially its derivative, while the Jacobian matrix is the generalization of the derivative for vector-valued functions.
Next, we examine the area scaling issue in change of variables for multiple integrals. Imagine an infinitesimal rectangular region in the plane, whose side lengths are given by the tiny vectors and . Under the coordinate transformation (whose local linear approximation is described by the Jacobian matrix), this rectangle maps to the plane, approximately forming a parallelogram spanned by the vectors and . According to the geometric meaning of determinants in linear algebra, the ratio of areas before and after the transformation—the area scaling factor—equals the absolute value of the determinant of the Jacobian matrix. Thus, we obtain the core area element formula for change of variables in multiple integrals: . The video concludes by noting that since area must be nonnegative, the absolute value of the determinant must be taken in practice, accounting for its sign.
At the beginning of this segment, local diagrams are drawn side by side on the whiteboard: one in the uv-plane and the other in the xy-plane; at the bottom, dxdy = |J| dudv = |∂(x,y)/∂(u,v)| dudv is written. The instructor adds, "Regarding sign, we must include an absolute value," explaining that when comparing areas, we cannot use J directly but must take the absolute value of the Jacobian determinant.
Next, this relationship is named "the expression for the area element under a curvilinear transformation." Mathematically, this means: a small area dxdy in the xy-plane corresponds, after the change of variables, to a small area dudv in the uv-plane, differing by a local scaling factor |∂(x,y)/∂(u,v)|.
The discussion then moves to applications. Since the area element is now known, we can directly integrate it over a region to compute area. The screen begins writing dxdy, representing integration of the area element over the region in the xy-plane.
The instructor emphasizes "substitution requires changing the limits of integration," which here specifically means reverting the domain of integration from back to D. Thus, dxdy is replaced by |∂(x,y)/∂(u,v)| dudv, yielding dxdy = |∂(x,y)/∂(u,v)| dudv. This step elevates the earlier local area relationship into a global area formula for the entire region.
The screen then scrolls back to a previous problem page, where dxdy = ? dudv had been written. Now the answer can be filled in directly: the integrand represented by the question mark is the absolute value of the Jacobian determinant, |J|. This shows that the previously unresolved question—"What is the integrand?"—is precisely the core conclusion derived in this segment.
The instructor further summarizes the method's significance: if we need to compute the area of a complicated region that arises as the image of some simpler region D under a curvilinear transformation, we need not integrate directly over the complicated region; instead, we integrate over the simple region D, using |∂(x,y)/∂(u,v)| as the integrand.
Next, the conclusion is generalized from "computing area" to "evaluating general double integrals." The whiteboard displays dxdy, indicating the standard form of a double integral. Substitution involves not only transforming the differential element but also the function itself and the region of integration.
The specific procedure proceeds in three simultaneous steps: first, replace the region with D; second, rewrite x and y inside as and ; third, replace dxdy with |∂(x,y)/∂(u,v)| dudv. This yields the complete formula: dxdy = |∂(x,y)/∂(u,v)| dudv.
To aid understanding of this formula's structure, the instructor recalls the single-variable substitution formula dx = (dx/du) du. A shared feature is that a coefficient must be inserted between the differentials of the two variables.
In the single-variable case, this coefficient is the derivative dx/du; in the two-variable case, it is the absolute value of the Jacobian determinant |∂(x,y)/∂(u,v)|. Thus, change of variables for double integrals is viewed as the natural higher-dimensional generalization of single-variable substitution.
Finally, the instructor names this result the 'Change of Variables Theorem for Multiple Integrals' and adds that it can be further generalized to triple integrals and n-fold integrals, though proofs in higher dimensions become more complex.
Next, we introduce several coordinate transformations frequently used in change of variables for multiple integrals, along with their corresponding Jacobian determinants.
First is the polar coordinate transformation in the plane. We convert Cartesian coordinates into polar coordinates , with transformation formulas and . Computing its Jacobian determinant yields .
Extending polar coordinates to three-dimensional space, we keep the -axis coordinate unchanged, i.e., , yielding the cylindrical coordinate transformation. Since the coordinate undergoes no mixed transformation, its partial derivatives in the Jacobian matrix leave the determinant’s value unchanged; thus, the cylindrical coordinate Jacobian determinant remains .
Another important three-dimensional coordinate transformation is spherical coordinates. We introduce radial distance , polar angle , and azimuthal angle (i.e., the angle between the line connecting the point to the origin and the positive direction of the axis). Its transformation formulas are , , and . The corresponding Jacobian determinant is .
To summarize, today we explored applications of the Jacobian matrix and Jacobian determinant in differentiation and integration. The Jacobian matrix is essentially the derivative of a vector-valued function; precisely because it represents the scaling factor of the local linear transformation, it appears in the change-of-variables formula for multiple integrals.
The coordinate transformation maps a figure from one plane to another. In this example, it maps the simple rectangular region D in the uv-plane onto the complicated region in the xy-plane. Such mappings form the foundation of the change-of-variables method for multiple integrals.
The most direct way to compute the area of a planar region is to perform a double integral of the constant 1 over that region. When the shape of the integration region is complicated, direct integration becomes difficult, prompting us to seek a method that converts integration over a complicated region into integration over a simpler region.
When we try to convert into , the key issue is determining the new integrand . This function reflects how the coordinate transformation scales the local area element—namely, the absolute value of the Jacobian determinant, which we will introduce later.
At the start, the video uses dxdy = ?\,dudv to highlight the central task: not only must an appropriate transformation T be found, but the scaling factor relating the area elements before and after the transformation must also be determined.
The video draws two hyperbolas and two rays in the xy-plane on the right, forming a curvilinear region, and explains that integrating directly over this region is difficult.
After defining u as xy and v as , the original boundaries correspond respectively to , , , and ; thus, the region of integration becomes a rectangle in the uv-plane.
After transforming the region into a rectangle, the video raises a key question: what is the relationship between the areas of figures before and after the transformation? This is precisely preparing for the subsequent introduction of the Jacobian factor.
The video explicitly contrasts two types of transformations: affine transformations map straight lines in the plane to straight lines, whereas curvilinear transformations can map straight lines to curves such as hyperbolas, making curvilinear transformations more general.
The video uses the board writing 'curvilinear transformation ←— affine transformation' and verbal explanation to indicate that curvilinear transformations are a more general concept obtained by relaxing the conditions of affine transformations.
The video writes a general curvilinear transformation as , , and further interprets it as T: , i.e., a mapping from the plane to the plane.
At the end of the video, F of x comma y equals x plus sine y comma y plus cosine x is given to illustrate that both the input and output of a coordinate transformation are points in the plane
The segment first points out that if the transformation is , both its domain and codomain are two-dimensional planes; visualizing the full graph would require entering four-dimensional space, making the "graph-reading" method used for single-variable functions unsuitable.
The explanation redraws the abstract transformation as two rectangular boxes placed side by side, connected by an arrow , with labels beneath both boxes. The left box represents the domain, the right box the codomain, and the arrow indicates the transformation relation.
To facilitate subsequent discussion of linear mappings, the explanation stipulates that points in the plane are all represented as vectors—and specifically as column vectors. This convention determines that matrices must be written on the left of vectors.
Pick in the domain and its image in the codomain. This step first establishes a pointwise correspondence, but the explanation immediately points out that to study region-to-region correspondence, we must also examine how local increments change.
If changes to , then the image point changes to . Both and are two-dimensional vectors, converting 'how regions change' into 'how small vectors near a point change'.
The explanation reviews differentiation for single-variable functions, defining it as 'the linear principal part of the function's increment', i.e., representing the dependent-variable increment using a linear expression in the independent-variable increment; a typical form is .
Inspired by single-variable differentiation, the explanation writes the vector form . Its structure remains 'linear principal part + higher-order remainder term', except that the increment has changed from a scalar to a two-dimensional vector.
In , is not an ordinary scalar constant, but rather an object that maps two-dimensional vectors linearly to two-dimensional vectors; thus the explanation explicitly calls it a 'second-order matrix'. Since vectors are column vectors, matrix multiplication occurs on the left.
The explanation emphasizes that is also a vector, not a scalar. Its infinitesimal meaning is: dividing this remainder vector by yields the zero vector in the limit. This prepares for the formal definition of differentiability later.
If the increment of the dependent variable of a vector-valued function can be written as , where A is a matrix independent of , then the function is said to be differentiable at that point. The video uses this definition to generalize single-variable differentiation to the vector-valued case.
In the differentiability expansion, only the linear term is called the differential, denoted ; correspondingly, the independent variable differential is denoted . Thus the differential is not the full increment, but rather the linear approximation part of the increment.
From , one may formally write . The video accordingly calls A the derivative, and especially emphasizes that this derivative is a matrix, not a scalar.
To determine the entries of matrix A, write and as column vectors (du, dv)^T and (dx, dy)^T. This allows converting the abstract relation into two scalar linear combinations via matrix multiplication.
Explore conditions, steps and evidence. Supplementary explanations are labeled separately from content shown in the video.
Draw a Cartesian coordinate system with the horizontal axis labeled 'x' and the vertical axis labeled 'y'.
xy
The standard Cartesian coordinate plane.
real numbers
Draw another Cartesian coordinate system with the horizontal axis labeled 'u' and the vertical axis labeled 'v'.
uv
An alternative Cartesian coordinate plane for transformations.
real numbers
Label the rectangular region in the uv-plane as 'D'.
D
A simple rectangular region in the uv-plane.
a subset of two-dimensional real space
Write the transformation equations x equals x as a function of u and v, y equals y as a function of u and v, and label the mapping arrow as 'T'.
T
A coordinate transformation that maps points from the uv-plane to the xy-plane.
a function from two-dimensional real space to two-dimensional real space
Label the irregular quadrilateral region in the xy-plane as ''.
The image of region D under the transformation T, located in the xy-plane.
a subset of the two-dimensional real space
The top of the screen displays dxdy = ? dudv
dxdy = ? dudv
The form of the double integral before and after the coordinate transformation, with the integrand on the right temporarily marked as a question mark
Double integral expression
A horizontal axis is drawn in the upper left corner labeled u
u
The horizontal coordinate of the uv-plane, taken here as xy
Real-valued coordinate variable
The vertical axis of the left coordinate system is labeled v, followed by
v
The vertical coordinate of the uv-plane, taken here as
Real-valued coordinate variable
A horizontal axis is drawn in the upper right corner labeled x
x
The horizontal coordinate of the xy-plane
Real-valued coordinate variable
The vertical axis of the right coordinate system is labeled y.
y
The vertical coordinate of the xy-plane.
A real-number coordinate variable.
Write on screen and mark p and q on the horizontal axis of the uv-plane.
Here, q is greater than p, which is greater than 0.
p,q
Parameters of the hyperbolas xy = p and xy = q, and also the endpoint values of u = xy.
Positive real-number parameters.
Write on screen and mark a and b on the vertical axis of the uv-plane.
Here, b is greater than a, which is greater than 0.
a,b
Slopes of the rays y = ax and y = bx, and also the endpoint values of .
Positive real-number parameters.
The instructor explains that a connection between the two planes can be established by considering a coordinate transformation.
Write x as a function of u and v, and y as a function of u and v.
A mapping that transforms coordinates (u, v) in one plane into coordinates (x, y) in another plane via the functions and .
The functions x and y must be well-defined at all points in the domain.
The instructor points out that to find the area of region , one can compute the double integral over that region with integrand 1.
Write the double integral over , with integrand dx dy.
The area of a two-dimensional region can be computed by integrating the constant function equal to 1 over that region.
The region R must be measurable.
How magical the method of coordinate transformation is
The equation dxdy = ? dudv has already been written above
At the beginning of the video, coordinate transformation is introduced as a technique for handling double integrals, and an equation is shown: the integration region in the xy-plane is transformed via T to the uv-plane, and the conversion factor between the area elements dxdy and dudv must be determined.
The topic under discussion is double integrals
There exists a transformation T from the plane to the plane
The goal is to convert a complicated region into a simpler, more manageable one
The screen displays T: { , } and T: -> , (u,v) -> (x,y)
A general planar curve transformation is, in essence, simply a mapping from the plane to the plane
The video defines a general planar curve transformation as a mapping that sends a point (u,v) in the uv-plane to a point (x,y) in the xy-plane, where x and y are both expressed as functions of u and v.
The input is a point (u,v) in the plane
The output is a point (x,y) in the plane
Both the domain and the codomain can be taken as subregions of the plane
Affine transformations map lines to lines, but curve transformations do not necessarily do so.
Write 'curve transformation ←— affine transformation' at the bottom of the screen.
The video explains by contrast: affine transformations preserve lines as lines, whereas curve transformations allow mapping lines to curves; thus, curve transformations generalize affine transformations.
The objects being compared are planar transformations
A special property of affine transformations is that they map lines to lines
Curve transformations do not require preserving linearity
So now the question is: what is the relationship between the areas of a figure before and after this coordinate transformation?
The right-hand side of the equation above remains ? dudv
After simplifying the shape of the region, the video clarifies the central question: for the same figure, what relationship must hold between the area elements dxdy and dudv before and after a coordinate transformation?
The transformation T has already been established
It is already known that the region of integration is simpler in the uv-plane
The transformation T: from the two-dimensional real space to the two-dimensional real space, mapping the vector (u, v) to the vector (x, y)
“Our domain and codomain are both the two-dimensional plane”
This segment first sets the object of study as a transformation T from the two-dimensional plane to the two-dimensional plane, with input variables denoted (u, v) and output variables denoted (x, y). The explanation emphasizes that both the domain and codomain of such an object are the two-dimensional plane, so it cannot be fully visualized using ordinary two-dimensional graphs as in the case of single-variable functions.
The domain is
The codomain is
Next, we represent points in the plane using vectors, and stipulate that each vector is a column vector.
Points in the left and right boxes are labeled and , respectively; increments are labeled and .
To discuss correspondences between regions, the explanation treats all points in the plane uniformly as vectors and explicitly stipulates that these vectors are column vectors. This convention directly determines that matrices must appear on the left side of vectors in matrix multiplication.
Points in the plane are treated as vectors
All vectors are handled as column vectors
We assume there is a vector x0 in the domain. Then its image under the transformation T is the vector y0.
The left box shows , the right box shows , and an arrow labeled T connects them.
The explanation first selects a point in the domain, then selects its image in the codomain, establishing a 'point-to-point' correspondence. It then raises the question: given a point-to-point correspondence, how can we further discuss a 'region-to-region' correspondence?
belongs to the domain
is the image of under T
If x0 moves a small distance to , we say it has an increment Δx; then y0 changes to .
The left arrow points from to ; the right arrow points from to .
Writing and on the board
When the independent variable point undergoes a small displacement, its increment is denoted , so the new point is ; correspondingly, the image point changes from to . The explanation explicitly states that both and are two-dimensional vectors.
is a two-dimensional increment vector
is a two-dimensional increment vector
If the function under discussion is a single-variable function, there is exactly one concept designed to handle this situation: the differential. The differential is the linear principal part of the function's increment—it uses a linear expression in the independent variable's increment to represent the dependent variable's increment.
The board displays 'Differentiation' and
The explanation first reviews the concept of differentiation for single-variable functions, defining it as 'the linear principal part of the function's increment'—that is, approximating the dependent variable's increment using a linear expression in the independent variable's increment. Then this idea is extended to the two-dimensional vector case.
When the object is a single-variable function, this form is introduced first
The core idea is to describe increments using their linear principal parts
The board displays
Then we can imitate the form written above
This segment only presents the analogous form, without formally stating the complete definition here
Inspired by the single-variable differentiation form, the explanation expresses the increment vector of the dependent variable as a linear expression in the increment vector of the independent variable plus a higher-order term: . This step extends the idea of 'linear principal part' from scalar increments to vector increments.
and are two-dimensional vectors
A is an undetermined linear coefficient object
An affine transformation maps straight lines in the plane to straight lines
An affine transformation maps straight lines in the plane to straight lines.
The discussion concerns affine transformations in the plane
This holds for all straight lines in the plane
We can regard curve transformations as a generalization of affine transformations
On-screen text: "Curve transformation ←— Affine transformation"
Curve transformation can be understood as a generalization of affine transformation.
Curve transformation is understood, as defined in the video, as a mapping from the plane to the plane.
Affine transformation is understood, as described in the video, as a planar transformation preserving linearity.
This holds within the conceptual framework adopted in the video.
For example, the transformation we just discussed maps a line to a hyperbola, and thus is not an affine transformation.
The coordinate transformation example presented earlier in the video is not an affine transformation.
This example maps a line in the uv-plane to a hyperbola or a ray-like curve in the xy-plane.
Affine transformations require that lines map to lines.
This holds for the specific example given in the video.
But here, both the domain and codomain are two-dimensional planes; to draw its graph clearly, you would need four-dimensional space, which is very difficult.
For a transformation whose domain and codomain are both two-dimensional planes, directly drawing its full graph requires observation in four-dimensional space, and is therefore extremely difficult.
The domain of the transformation is
The codomain of the transformation is
Regarding the two-to-two dimensional transformation discussed in this segment
Now that we understand point-to-point correspondence, how do we discuss region-to-region correspondence? To answer this, we must first determine how y changes when our input variable x0 undergoes a small increment.
To move beyond point-to-point correspondence and discuss region-to-region correspondence, we must first examine how the output changes when the input undergoes a small increment.
We already have the point-to-point correspondence of transformation T
Consider a small displacement at some point in the domain.
The incremental behavior of transformation T near a local point.
Is there something that maps two-dimensional vectors to two-dimensional vectors and is preferably linear? Yes: second-order matrices.
On the board, label A as "2-order matrix".
In the form , if A is to map two-dimensional vectors linearly to two-dimensional vectors, then A must be a second-order matrix.
and are two-dimensional column vectors.
A acts as a linear mapping.
For the case of two-dimensional vector increments discussed in this segment.
This vector divided by the magnitude of should approach the zero vector.
The board displays and an explanation relating it to ||.
The limit expression is not fully written; only verbal and schematic board explanations are given.
In Δy⃗ = A Δx⃗ + o(Δx⃗), the remainder term o(Δx⃗) is a vector, and it tends to the zero vector when divided by |Δx⃗|.
o(Δx⃗) denotes the vector-valued remainder term.
|Δx⃗| denotes the magnitude of the increment vector of the independent variable.
As Δx⃗ tends to zero.
The speaker explicitly states, "The derivative of a vector-valued function is a matrix."
A appears in the position of and is labeled "derivative".
For a vector-valued function differentiable at a point, its derivative is not a single number but rather the matrix A satisfying .
The function is differentiable at that point.
The notation used in the video.
Holds for the vector-valued function at the point discussed in the video.
The speaker says, "We call this matrix the derivative of the vector-valued function at this point, or the Jacobian matrix."
Blue board writing shows J and "Jacobian matrix".
In the above two-dimensional case, the matrix formed by the partial derivatives is both the derivative matrix at that point and is also called the Jacobian matrix.
The function is differentiable.
Denote the components as u and v, and the independent variables as x and y.
Applies to the two-dimensional vector-valued function discussed in this segment.
The speaker says, "If the Jacobian matrix is a square matrix, then we often discuss its determinant."
The blue board displays a determinant composed of four partial derivatives.
When the Jacobian matrix is square, we may take its determinant; this determinant is called the Jacobian determinant.
The Jacobian matrix is square.
Holds for the Jacobian matrix in the square-matrix case.
We just said that the Jacobian matrix serves as the derivative of a vector-valued function; in that sense, the gradient is the derivative of a multivariable function.
The Jacobian matrix is the derivative of a vector-valued function, whereas the gradient is the derivative of a multivariable function (a scalar-valued function).
The function is a multivariable function.
For a multivariable function,
Therefore, dxdy equals the absolute value of the Jacobian determinant... This is the expression for the area element under a curvilinear transformation.
Write on the board: dxdy = |J| dudv = |∂(x,y)/∂(u,v)| dudv
Under a curvilinear transformation, the area element in the xy-plane satisfies dxdy = |∂(x,y)/∂(u,v)| dudv.
There exists a curvilinear transformation from (u,v) to (x,y)
We discuss the scaling relationship between area elements
Holds for local area elements
First, consider two hyperbolas in the xy-plane: xy=q and xy=p; then two rays: y=ax and y=bx. The range of u is from p to q, and the range of slopes is from a to b. Thus, the integration region becomes a rectangle.
On the right, draw the curved region in the xy-plane bounded by xy=q, xy=p, y=ax, and y=bx; on the left, draw the rectangle [p,q]×[a,b] in the uv-plane.
First, present two hyperbolas in the xy-plane as one pair of boundaries of the region.
Verbally stated and simultaneously written on the board in the video.
Next, present two rays through the origin as the other pair of boundaries of the region.
Verbally stated and simultaneously written on the board in the video.
Define the horizontal coordinate in the uv-plane as xy, i.e., the parameter in the family of hyperbolas.
The video explicitly says, “Think of it as xy, that is, the coefficient in the hyperbola.”
Define the vertical coordinate in the uv-plane as , i.e., the slope of the family of rays.
The video explicitly says, “Think of it as the slope of the ray, y over x.”
Since the original region lies between xy=p and xy=q, the range of u in the new coordinates is from p to q.
u = xy corresponds directly to the boundary equations.
Since the original region lies between y = ax and y = bx, the range of v in the new coordinates is from a to b.
corresponds directly to the definition of the slope of a ray.
Thus, the region of integration becomes a rectangle in the uv-plane.
The Cartesian product of intervals for two independent variables forms a rectangular region.
By setting u = xy and , the original complicated region bounded by hyperbolas and rays is transformed into the rectangle [p, q] × [a, b] in the uv-plane.
The screen displays T: { , } and T: , (u, v) ↦ (x, y).
It is essentially a mapping from one plane to another... I map the point (u, v) in the uv-plane to the point (x, y) in the xy-plane.
First write a general curve transformation in component function form.
The definition form given on the video blackboard.
Then interpret this transformation as a mapping from the two-dimensional plane to the two-dimensional plane.
Video verbal explanation: "a mapping from a plane to a plane."
Ultimately, this is realized as a point-to-point correspondence: a point in the uv-plane is mapped to a point in the xy-plane.
Video verbal explanation and illustration of point-mapping notation.
The essence of a general planar curve transformation is a mapping from (u,v) to (x,y), whose components can be written as and .
First explain that the differential of a single-variable function is the linear principal part of its increment; then say, 'We can imitate the form written above.'
On the board, and Δy⃗ = A Δx⃗ + o(Δx⃗) appear sequentially.
This is an analogy-based generalization, not a rigorous proof; this segment does not prove that this form necessarily exists.
First write the standard form stating that 'the differential of a single-variable function is the linear principal part of its increment.'
A retrospective explanation directly given in the video.
Replace the scalar increment with a vector increment to obtain the analogous form for the two-dimensional case.
The explanation explicitly states, 'Follow the form written above.'
It then asks what A is and identifies A as a second-order matrix, since a linear transformation from a two-dimensional vector to a two-dimensional vector is required.
Oral reasoning and board annotations in the video
Next, is explained as a vector-valued infinitesimal term, and it is stated that dividing it by || yields the zero vector in the limit.
Oral explanation in the video of the meaning of infinitesimal quantities
By analogy, the idea of the 'linear principal part' for single-variable functions is generalized to the two-dimensional vector increment form , where A is interpreted as a second-order matrix and the remainder term is interpreted as a vector-valued infinitesimal.
Starting from d y⃗ = A d x⃗, expand it into (du, dv)^, then write the two-row expressions for du and dv, and finally fill in the four partial derivatives.
The speaker explains step by step: "Let us expand it... We use matrix multiplication... This expression readily reminds us of the total differential of a multivariable function... Here is ∂u/∂x..."
Begin from the already-defined vector differential relationship.
Use the differential definition given earlier.
Write the differentials of the two vectors componentwise as column vectors.
Substitute component notation without changing the meaning of the equation.
Expand according to matrix multiplication rules to obtain two scalar linear combinations.
Definition of matrix multiplication.
Compare with the total differential form of a multivariable function to identify the four coefficients as the corresponding partial derivatives.
Comparison between the total differential formula and the coefficients.
Place the four partial derivatives back into their matrix positions to obtain the explicit form of the derivative matrix.
Follows directly from the coefficient correspondence in the previous step.
The derivative matrix of a two-dimensional vector-valued function at that point is the Jacobian matrix composed of the four first-order partial derivatives.
The blue chalkboard first writes the determinant , then begins writing the notation ∂(u,v)/∂(x,y).
The speaker says, "Because this determinant is widely used, it has a special notation. We also write it in derivative form."
At the end of the clip, the special notation is not yet fully written; only the beginning ∂(u,v) is confirmed, and the speaker points to a notation similar to that for derivatives.
First, confirm the Jacobian matrix itself.
Already obtained in the previous derivation.
Because it is a square matrix, its determinant can be taken.
Definition of the determinant of a square matrix.
The video further introduces a dedicated notation to represent this determinant.
The speaker states that it has a special notation and begins writing it on the board.
The Jacobian determinant can be written as the determinant shown above, and is commonly denoted by specialized notations such as ∂(u,v)/∂(x,y); this clip only shows the initial writing of the notation.
Draw a small rectangle in the UV-plane and a parallelogram in the XY-plane.
We implemented the coordinate transformation using a matrix; what is the scaling factor for area before and after this transformation? It is the determinant of this matrix.
Select an infinitesimal rectangle in the UV-plane with side vectors du and dv.
Define a local small region
Through the coordinate transformation T (described by the Jacobian matrix J), du and dv are mapped to vectors dx and dy in the XY plane
The Jacobian matrix acts as a linear transformation mapping basis vectors to new basis vectors
The area of the parallelogram spanned by dx and dy in the XY plane corresponds to the image of the original rectangle's area
Local linear approximation
The area of the parallelogram equals the original rectangle's area multiplied by the absolute value of the determinant of the transformation matrix
Geometric meaning of a matrix determinant in linear algebra (area scaling factor)
Derive the area element relationship for change of variables in multiple integrals: dxdy = |J| dudv
Now that we know the area element, we can directly compute the area by integrating it over the region.
The blackboard step-by-step writes dxdy = |∂(x,y)/∂(u,v)| dudv.
Start with the previously obtained area element relation dxdy = |∂(x,y)/∂(u,v)| dudv.
The area element formula from the earlier board work.
Take the double integral of the area element over the region in the xy-plane.
The video states, "Directly integrating it yields the area."
Replace dxdy with |∂(x,y)/∂(u,v)| dudv and change the region of integration from to D.
The video says, "When changing variables, change the limits… here, that means replacing the region."
Obtain the area expression before and after coordinate transformation: dxdy = |∂(x,y)/∂(u,v)| dudv.
"We formally write down such a formula… this is a general double integral."
The board writes dxdy = |∂(x,y)/∂(u,v)| dudv.
Start from a general double integral, not just integrating the constant 1.
The video says, "This is a general double integral."
Substitute the expressions for x and y in terms of the new variables into the integrand.
The video states, "We need to express it using expressions of u and v."
Substitute the area differential using the area element formula derived earlier.
Use the established relationship between dxdy and dudv.
Transform the integration region from in the xy-plane back to D in the uv-plane.
The video emphasizes, "When changing variables, the limits must also change—the region must be changed."
Obtain the change-of-variables formula for double integrals: dxdy = |∂(x,y)/∂(u,v)| dudv.
Draw an irregular quadrilateral in the xy-plane and a rectangle in the uv-plane, and connect them via the transformation T.
The instructor poses a question: find the area of a complicated region by transforming it from a simpler rectangular region.
Use the coordinate transformation T to relate the irregular region in the xy-plane to the simple rectangular region D in the uv-plane, thereby computing the area of .
The region D is a rectangle in the uv-plane.
The transformation T maps (u, v) to (x, y), where x is a function of u and v, and y is a function of u and v.
The region is the image of D in the xy-plane.
Express the integral for the area of as an integral over D.
Set up a double integral for the area of the target region in the xy-plane.
Define the area using a double integral with integrand 1.
Propose transforming the integral to the simpler uv-plane region D, but the new integrand remains unclear.
We hope to integrate over a region with a simpler geometric shape (a rectangle).
At the end of the video, a question is raised: what should this unknown integrand '?' be? This leads to the concept of the Jacobian determinant.
No content
"I want to compute the area of the region bounded by these four curves... and under this change of variables, the integration region becomes very simple."
A curvilinear region is drawn on the right in the xy-plane, and a rectangular region is drawn on the left in the uv-plane.
Consider in the xy-plane the region bounded by xy = q, xy = p, y = ax, and y = bx, and examine its shape under the coordinate transformation.
x times y equals q
x times y equals p
y equals a times x
y equals b times x
q is greater than p, and p is greater than zero
b is greater than a, and a is greater than zero
Find suitable u and v so that the integration region becomes simple in the uv-plane.
Let u be the hyperbolic parameter.
Then xy = p and xy = q correspond directly to and .
Let v be the slope of the ray.
Then y = ax and y = bx correspond directly to and .
Read off the transformed range of values.
Obtained by term-by-term correspondence between the original region’s boundaries and the definitions of the new variables.
The transformed integration region is a rectangular region in the uv-plane.
When two variables each independently take an interval, their Cartesian product is a rectangle.
The transformed integration region is the rectangle 〔p,q〕×〔a,b〕.
The video illustrates through side-by-side diagrams: the curved region on the right corresponds to the rectangular region on the left.
The screen displays
Let us consider a concrete example: this is a curve transformation in the plane.
Here is a concrete example of a mapping from the plane to itself.
F of x comma y equals the ordered pair x plus sine of y comma y plus cosine of x
This explains that both the input and output of a curve transformation are points in the plane.
Write down a binary vector-valued function directly.
The example written on the video's blackboard.
Interpret this expression as mapping a point in the plane to another point in the plane.
The video then explains: "For a coordinate transformation, both its input and output are points in the plane."
=(,\ ) is an example of a curve transformation in the plane.
The video presents it directly as a concrete example and verbally classifies it under curve transformations.
The board displays
No dedicated computation or expansion of this example occurs within this segment.
This formula appears on the board, but no problem-solving example is built around it in the segment; thus, it serves only as an illustrative board note, not a complete worked example.
Some symbols in the handwritten notation have limited legibility, but the expression is overall readable as (x sin y, )
The board presents a two-dimensional vector-valued function example, , to accompany the theme of 'two-dimensional transformations.'
The value of the function F at the point (x, y) is (x times the sine of y, y plus the cosine of x).
No explicit goal to solve is given in this segment.
Only record the example function as it appears on the board.
Directly from the board writing.
No computation result for this example is given in the segment.
There is no verifiable final answer; only the example formula is confirmed to have been written on the board.
"Now, what about that question we raised at the beginning—what is the integrand? Well, now we know: the integrand is the absolute value of the Jacobian determinant."
The screen returns to the earlier page displaying dxdy = ? dudv, and then |J| is filled in for the question mark.
In the area transformation problem raised earlier, to write dxdy as a double integral over the uv-plane, we must determine the integrand represented by the question mark.
We already have the area element relationship dxdy = |∂(x,y)/∂(u,v)| dudv.
The original problem is stated as dxdy = ? dudv.
Determine the explicit expression for ?.
Replace dx dy using the area element formula.
Directly use the previously obtained dx dy = |∂(x,y)/∂(u,v)| du dv.
Thus, the integrand over the uv-plane is the absolute value of the Jacobian determinant.
The video explicitly states, "This integrand is the absolute value of the Jacobian determinant."
? equals the absolute value of the Jacobian determinant of x and y with respect to u and v, that is, the absolute value of the Jacobian determinant.
Consistent with the subsequently written |∂(x,y)/∂(u,v)| du dv.
Hand-draw the x-axis and y-axis, then sketch an irregular quadrilateral in the first quadrant.
x-axis
y-axis
irregular quadrilateral region
The coordinate axes have been drawn.
The complicated region has been sketched.
Identify the actual space whose area needs to be computed, and highlight its geometric complexity.
Hand-draw the u-axis and v-axis, then sketch a simple rectangle in the first quadrant.
u-axis
v-axis
rectangular region
The coordinate axes have been drawn.
A simple rectangular region has been sketched.
Introduce an alternative coordinate system with a geometrically simple region to facilitate integration.
Draw an arrow from the uv-plane to the xy-plane, write the transformation equations, label the mapping as T, and label the two regions as D and .
mapping arrow
transformation equations
label D and
The relationship between two planes is explicitly illustrated and mathematically defined.
It demonstrates how a point or region in the uv-plane maps to a point or region in the xy-plane, preparing for change of variables.
First, the u-axis and x-axis appear at the top of the screen; then the y-axis of the right coordinate system is added, forming two planes side by side.
u-axis at the top left
x-axis at the top right
y-axis on the right
Integral equality at the top
First present the u-axis and x-axis
Then add the y-axis to form the xy-plane reference.
The equation dxdy = ? dudv remains displayed at the top throughout.
The video prepares for comparing the region in the xy-plane with the region in the uv-plane using a side-by-side layout.
On the right, the curves xy=q, xy=p, y=ax, and y=bx are drawn sequentially, and the region enclosed by these four curves is shaded.
The hyperbola xy = q
The hyperbola xy = p
The ray y = ax
The ray y = bx
The shaded region
First draw the two hyperbolas
Then draw the two rays
Finally shade the region enclosed by them
The parameter conditions and remain unchanged
This figure shows that the original integration region is a complicated curved-region formed by the intersection of two families of curves.
On the left, draw the uv-coordinate system, write u = xy and , label p and q on the horizontal axis, label a and b on the vertical axis, and finally draw and shade a rectangle
u-axis
v-axis
tick marks p, q
tick marks a, b
shaded rectangular region
First set up the uv-coordinate system
Then write u = xy,
Next label p, q, a, b
Finally draw the rectangular region
The rectangle's boundaries correspond to , , ,
This figure transforms the complicated region on the right into the regular rectangle on the left, illustrating intuitively that the integration limits become simpler after the change of variables.
Draw an arrow between two planes and label it T
Left rectangle
Right curved-region
Arrow T
Add a mapping arrow from uv to xy between the two diagrams
The left and right regions still represent the pre-transform and post-transform shapes respectively
Arrow T lifts the earlier correspondence into an explicit coordinate transformation relationship.
Write 'curved transformation ← affine transformation' at the bottom of the screen, then write the component form of T, T: , (u,v)→(x,y), and finally write
Text 'curved transformation'
Text 'affine transformation'
mapping T
example F
first write the conceptual comparison
then write the general definition of curve transformation
finally add concrete examples
the original figure and integral equation above remain as background references
this segment abstracts the earlier geometric examples into the language of general mappings and concludes with a concrete function example.
The instructor draws two rectangular boxes connected by an arrow in the middle, with T written above the arrow and written below both boxes.
Let us first draw two boxes; these boxes represent the domain and codomain of this transformation.
The left rectangular box.
The right rectangular box.
The arrow T in the middle.
two labels of
first draw the left box
then draw the right box
add the middle arrow T
label below both boxes
the two boxes represent the domain and codomain respectively
the arrow direction always goes from left to right
this diagram rewrites the abstract two-dimensional to two-dimensional transformation as the correspondence 'the left plane is mapped by T to the right plane', facilitating subsequent discussion of point and region mappings.
a point appears inside the left box labeled , and a point appears inside the right box labeled
there is a vector x0 in the domain … its image under the transformation T is the vector y0
initial vector x zero
initial vector y zero
left and right boxes
first mark in the left box
then mark in the right box
always lies on the domain side
always lies on the codomain side
this step concretizes the transformation as a point-to-point correspondence.
in the left box, draw an arrow from to ; in the right box, draw an arrow from to
if x0 moves a small amount to , then y0 changes to
arrow
arrow
the initial vector x zero plus the vector increment Δx
the initial vector y zero plus the vector increment Δy
first draw the small displacement of the independent variable in the left box
then draw the corresponding displacement of the image point in the right box
the left and right increments are related by the same transformation T
all increments are treated as vectors
the diagram transforms the question of how a region changes into the question of how small vectors near a point are affected by the transformation
the board displays "differential " and ""
label A as "2by dimensional vector" next to delta x vector and delta y vector2dimensional vector"
formula
formula
Annotation: "2-dimensional vector"
annotate "2by matrix"
First write the univariate form.
Then write the vector form.
Add annotations clarifying the identities of A and the increment vector.
Both expressions capture the structure "linear principal part plus higher-order terms".
Visual annotation grounds abstract symbols in concrete objects: the increment is a two-dimensional vector, and the linear coefficient is a second-order matrix.
Affine transformations map lines in the plane to lines, but curve transformations do not necessarily do so. For instance, the example we just gave maps a line to a hyperbola, which is therefore not an affine transformation.
assuming that all plane coordinate transformations, like affine transformations, map lines to lines.
the video emphasizes that affine transformations preserve linearity, whereas curve transformations can map lines to curves such as hyperbolas, so the two are distinct, and curve transformation is the more general concept.
first state 'here A is a constant independent of Δx', then ask 'what is this A?' and answer 'it is a second-order matrix'
on the board, '2-order matrix' is written beneath A
when seeing , one tends to apply the intuition from single-variable calculus and interpret A as an ordinary constant
In the two-dimensional vector case of this segment, the explanation explicitly identifies A as a 'second-order matrix' because it must linearly map two-dimensional vectors to two-dimensional vectors.
This is an expression for a vector, so this infinitesimal must also be a vector.
In single-variable functions, is often regarded as a scalar higher-order infinitesimal, making it easy to mistakenly interpret in the two-dimensional case as scalar.
The explanation emphasizes that this infinitesimal is a vector-valued remainder, and must be understood as 'dividing by || yields the zero vector in the limit'.
The speaker explicitly states, "That is, the derivative of a vector-valued function is a matrix."
⃗ / d x⃗ is explicitly written in matrix form
When hearing the term 'derivative', one may instinctively apply experience from single-variable functions and assume the result is still a scalar.
This segment explicitly states that the derivative of a vector-valued function at a point is the matrix A, not a single number.
First write , then rename the term alone as .
The speaker distinguishes between 'the increment of the dependent variable' and 'this part... the differential'.
It is easy to mistakenly treat the entire 〔Δy⃗〕 as the differential.
The video decomposes 〔Δy⃗〕 into the linear principal part 〔AΔx⃗〕 and the higher-order remainder term 〔o(Δx⃗)〕, where only the linear principal part is called the differential 〔d y⃗〕.
Emphasize that swapping the positions of two variables in the Jacobian determinant changes the determinant.
Believing that the order of variables in the numerator of the Jacobian determinant can be arbitrarily reversed without affecting the result.
Determinants are antisymmetric: swapping two rows (or two columns) changes the sign. Therefore, the ordering of dependent and independent variables must correspond strictly; otherwise, the computed determinant values will be incorrect (negatives of each other).
Of course, if we ignore sign...
the video cuts off here without fully explaining the geometric meaning of the sign
treating the area scaling factor as equal to the Jacobian determinant itself, ignoring the possibility that it may be negative.
Area is a scalar and non-negative, so the area scaling factor should be the absolute value |J| of the Jacobian determinant.
If orientation matters, we must also include an absolute value.
Using J directly as a coefficient in area elements or multiple integrals under change of variables.
The video emphasizes that when orientation matters, comparing areas requires using the absolute value |J| of the Jacobian determinant.
Then we replace all the xy with u and v... yet the xy remains xy, but we express it using expressions in u and v. And dxdy becomes this expression.
Assuming that changing variables in multiple integrals only requires replacing dxdy with |J|dudv.
The video explains that we must also rewrite x and y in as and , and revert the integration region from back to D.
The instructor explains that the transformation T maps each point in region D to a point in the xy-plane, thereby forming the region .
Coordinate transformations are tools that connect complicated integration regions with simpler ones; this is the fundamental idea behind the change-of-variables formula.
above, first give dxdy = ? dudv
Then uses a curved-region example in the xy-plane and a rectangular-region example in the uv-plane.
The video first introduces the topic via a general integral substitution problem, then demonstrates concretely how coordinate transformations simplify regions using an example bounded by hyperbolas and rays.
To discuss this problem, we must first examine the nature of coordinate transformations.
Then writes T: {, } and T: .
The concrete example first shows how substitution simplifies the region; only afterward does the video abstract the general definition of a curvilinear transformation to proceed with discussing the relationship between area elements.
We previously studied affine transformations in linear algebra… We can regard curvilinear transformations as a generalization of affine transformations.
The screen displays "curvilinear transformation ←— affine transformation".
The video explicitly states that curvilinear transformations are a generalization of affine transformations; thus, affine transformations are a special case of curvilinear transformations.
So the question now is: what is the relationship between the areas of figures before and after this coordinate transformation?
The right-hand side of the equation above still retains ? dudv.
To answer the relationship between dxdy and dudv, we must first clarify the meaning of the curvilinear transformation T as a point mapping; the video proceeds precisely in this order.
First writes the general form T: {, }, then writes .
A concrete example: F is a direct instantiation of the general curve transformation, used to illustrate that both input and output are points in the plane.
First define T: -> , then stipulate 'each vector is a column vector'.
Only after viewing the object as a transformation from two dimensions to two dimensions does one further introduce the convention of writing points in the plane uniformly as column vectors.
As we just stated, all vectors here are column vectors, so matrix multiplication acts on the left—this is entirely reasonable.
The column-vector convention determines that, in the linear principal part, matrix A must appear to the left of the increment vector, thereby supporting the identification of A as a second-order matrix.
First discuss 'point-to-point correspondence', then ask 'how to discuss region-to-region correspondence', and subsequently introduce x0 moving a small step to .
Region-level correspondence must first be studied through the behavior of small increments near a point; thus, point-to-point correspondence is the prerequisite for introducing the increment vector.
On the board, first write , then write .
We can follow the form written above.
The vector form in two dimensions is a direct generalization of the idea for single-variable functions that 'the differential is the linear principal part of the increment.'
After writing the vector form, we ask 'What exactly is this A?' and then answer 'It is a second-order matrix.'
The linear principal part expression in vector form contains further identification of the coefficient A's identity.
After writing , we continue by explaining 'How do we understand this infinitesimal quantity?'
This formula not only includes the linear principal part but also calls for an explanation of the higher-order remainder term .
The speaker asks, 'Then who is the integrand? This is the question we will discuss today.'
"How magical this coordinate transformation method is!" "The integration region then becomes very simple."
T: {, } and T: -> , (u,v)->(x,y)
"An affine transformation maps lines in the plane to lines... but a curve transformation does not necessarily do so."
"What is the relationship between the areas of the figure before and after this coordinate transformation?"
The curvilinear region on the right corresponds to the rectangle [p,q]×[a,b] on the left.
To draw its graph clearly, you would need to work in four-dimensional space, which is very difficult.
Now that we understand point-to-point correspondences, how do we discuss region-to-region correspondences?
The first question: What is this A? … Yes, it is a second-order matrix.
This vector divided by the magnitude of Δx should approach the zero vector.
All vectors here are column vectors, so matrix multiplication acts on the left—this is very natural.
The speaker defines, "If a vector-valued function … then we say this vector-valued function is differentiable at this point."
The vector Δy equals the matrix A multiplied by the vector Δx, plus an infinitesimal term of higher order than Δx.
Covered · This segment as a whole introduces double integrals using coordinate transformations to simplify area computation, ultimately leading to the central question about the new integrand.
Covered · Introducing coordinate transformations and the goal of change of variables in double integrals.
Covered · Use an example region bounded by a hyperbola and rays to illustrate how substitution transforms the region into a rectangle.
Covered · Raise the question of how area elements relate before and after the transformation.
Covered · Compare affine transformations with curvilinear transformations, emphasizing that the latter generalize the former.
Covered · Give the mapping definition of a general curvilinear transformation.
Covered · Provide a concrete example of a planar curvilinear transformation: .
Covered · Explain that visualizing a two-dimensional to two-dimensional transformation directly is difficult, and introduce an alternative approach.
Covered · Draw two boxes representing the domain and codomain, and specify that points are represented as column vectors.
Covered · Label and its image , and raise the question of how to advance from pointwise correspondence to regionwise correspondence.
Covered · Use arrows to depict and , illustrating the local increment relationship.
Covered · Review that for single-variable functions, the differential is the linear principal part of the increment, and write the vector analog.
Covered · ask what A means and confirm it is a two by two matrix.
Covered · Explain that is a vector-valued infinitesimal term, and that dividing it by || yields a vector approaching the zero vector.
Covered · Explain the definition of differentiability for vector-valued functions, and use a red circle to highlight the linear principal part, which is the differential.
Covered · Write , define A as the derivative, and emphasize that the derivative is a matrix.
Covered · After turning the page, expand the vector differential componentwise into (du, dv)^T and (dx, dy)^T.
Covered · Use matrix multiplication to obtain the two-row expressions for du and dv, and relate them to the total differentials of multivariable functions.
Covered · Fill in the four partial derivatives in order, and arrange them into the Jacobian matrix J.
Covered · Take the determinant of a square matrix, and introduce the Jacobian determinant along with its special notation.
Covered · Explaining the definition of the two-dimensional Jacobian determinant and the importance of variable order.
Covered · Introducing the concepts of differentiation and derivative for vector-valued functions, with a geometric intuitive explanation using the 'grass-growing' metaphor.
Covered · Generalizing to higher-dimensional spaces and defining the derivative matrix for mappings from to .
Covered · Discussing the special case where , i.e., the Jacobian matrix of a multivariable function (a row vector of partial derivatives).
Covered · Explain how the Jacobian matrix degenerates into the gradient when , and their shared interpretation as derivatives.
Covered · Derive the area scaling factor formula dxdy = |J| dudv for change of variables in multiple integrals using a geometric illustration of the mapping from the UV-plane to the XY-plane.
Covered · Explain the area element formula and emphasize taking the absolute value.
Covered · Derive the region area transformation formula by integrating the area element.
Covered · Revisit the earlier question and clarify that the integrand is |J|.
Covered · Summarize that the area of a complicated region can be expressed as a multiple integral over a simpler region.
Covered · Generalize to the general double integral substitution formula.
Covered · Draw an analogy with the single-variable substitution formula.
Covered · Explain that the theorem extends to triple integrals and n-fold integrals.
Covered · This segment fully covers the explanations of polar, cylindrical, and spherical coordinate transformations and their Jacobian determinants.
Reviewed subject paths
Continue with related videos in the same subject.
Bilibili13:2.7000000000000455
Bilibili11:02
Bilibili3:12.810000000000002