The Fundamental Theorem for Line Integrals

Learning Objectives

By the end of this section, you will be able to:

1. Identify conservative vector fields and understand the condition $\frac{\partial P}{\partial y} = \frac{\partial Q}{\partial x}$ for path independence
2. Find potential functions for conservative fields by integrating component functions
3. Apply the Fundamental Theorem for Line Integrals to evaluate $\int_C \nabla f \cdot d\mathbf{r} = f(B) - f(A)$ using only endpoint values
4. Explain path independence and its connection to conservation laws in physics
5. Connect these ideas to optimization and understand why gradient descent finds minima regardless of the path taken

Why This Matters: The Fundamental Theorem for Line Integrals is the multivariable generalization of the Fundamental Theorem of Calculus. Just as $\int_a^b f'(x) \, dx = f(b) - f(a)$ tells us that the integral of a derivative depends only on boundary values, the line integral theorem tells us that the integral of a gradient depends only on endpoints—not on the path between them. This principle underlies conservation of energy in physics and path-independence in optimization.

The Big Picture

In the previous sections, we learned to compute line integrals $\int_C \mathbf{F} \cdot d\mathbf{r}$ by parametrizing the curve $C$ and integrating. This works but raises a natural question: does the answer depend on which path we take between two points?

For most vector fields, yes—different paths give different integrals. But for a special class called conservative vector fields, something remarkable happens: the integral depends only on where you start and where you end, not on how you get there. This is called path independence.

Historical Origins

The connection between path independence and potential functions emerged from the study of gravitational and electrical forces in the 18th and 19th centuries:

• Pierre-Simon Laplace (1749-1827) and Joseph-Louis Lagrange (1736-1813) developed the concept of gravitational potential, showing that gravitational force is the negative gradient of a potential energy function
• George Green (1793-1841) introduced the term "potential" and developed Green's theorem, which connects line integrals to area integrals
• Lord Kelvin (William Thomson, 1824-1907) applied potential theory to electrostatics and thermodynamics, showing how path independence leads to conservation laws

The physical insight was profound: when lifting a rock from one height to another, the work done depends only on the height difference—not on whether you lift it straight up, carry it diagonally, or take a winding path. Gravity is a conservative force.

Conservative Vector Fields

Definition and Examples

A vector field $\mathbf{F}$ is called conservative (or gradient field) if there exists a scalar function $f$ such that:

$\mathbf{F} = \nabla f = \left( \frac{\partial f}{\partial x}, \frac{\partial f}{\partial y} \right)$

The function $f$ is called the potential function (or scalar potential) of $\mathbf{F}$.

Testing for Conservative Fields

How can we tell if a field is conservative without guessing a potential function? For a field $\mathbf{F} = (P, Q)$ defined on a simply connected region:

$\mathbf{F} \text{ is conservative} \quad \Longleftrightarrow \quad \frac{\partial P}{\partial y} = \frac{\partial Q}{\partial x}$

Example: Is $\mathbf{F} = (2xy + 3, x^2 - 2y)$ conservative?

Here $P = 2xy + 3$ and $Q = x^2 - 2y$.

• $\frac{\partial P}{\partial y} = 2x$
• $\frac{\partial Q}{\partial x} = 2x$

Since $\frac{\partial P}{\partial y} = \frac{\partial Q}{\partial x}$, the field is conservative!

Potential Functions

Finding Potential Functions

Once we know a field is conservative, how do we find its potential function? The method is systematic:

Example: Find the potential for $\mathbf{F} = (2xy + 3, x^2 - 2y)$

1. Step 1: Integrate $P = 2xy + 3$ with respect to $x$:

   $f(x, y) = \int (2xy + 3) \, dx = x^2 y + 3x + g(y)$

   The "constant" of integration $g(y)$ may depend on $y$.
2. Step 2: Differentiate with respect to $y$ and set equal to $Q$:

   $\frac{\partial f}{\partial y} = x^2 + g'(y) = x^2 - 2y$

3. Step 3: Solve for $g'(y)$:

   $g'(y) = -2y \quad \Rightarrow \quad g(y) = -y^2 + C$

4. Result: $f(x, y) = x^2 y + 3x - y^2 + C$

Interactive: Potential Function Explorer

Explore different potential functions and see how the gradient field (shown as arrows) relates to the surface. Notice that gradient arrows always point "uphill" on the potential surface:

The Fundamental Theorem

Statement of the Theorem

Now we state the central result that connects conservative fields, potential functions, and line integrals:

This is the multivariable analog of the Fundamental Theorem of Calculus. Just as $\int_a^b f'(x) \, dx = f(b) - f(a)$, the line integral of a gradient depends only on the values at the endpoints.

Proof Sketch

The proof uses the chain rule. If $\mathbf{r}(t) = (x(t), y(t))$, then:

$\int_C \nabla f \cdot d\mathbf{r} = \int_a^b \nabla f(\mathbf{r}(t)) \cdot \mathbf{r}'(t) \, dt$

By the chain rule, the integrand is exactly:

$\nabla f \cdot \mathbf{r}'(t) = \frac{\partial f}{\partial x}\frac{dx}{dt} + \frac{\partial f}{\partial y}\frac{dy}{dt} = \frac{d}{dt}[f(\mathbf{r}(t))]$

So the integral becomes:

$\int_a^b \frac{d}{dt}[f(\mathbf{r}(t))] \, dt = f(\mathbf{r}(b)) - f(\mathbf{r}(a))$

Worked Example: The Easy Way vs The Hard Way

Words are persuasive, but a single example does more than a paragraph of theory. We will compute the same line integral two ways: once by parametrizing the curve and grinding through the integral, once by reading the potential at the endpoints. They must agree, and we will see exactly how much labor the Fundamental Theorem saves.

The setup. Let $\mathbf{F}(x,y) = (2xy + 3,\; x^2 - 2y)$. We already showed this field is conservative with potential $f(x,y) = x^2 y + 3x - y^2$. Compute $\int_C \mathbf{F} \cdot d\mathbf{r}$ along the curved path $\mathbf{r}(t) = (1+t,\; 3t^2)$ for $0 \leq t \leq 1$.

Notice the endpoints: $\mathbf{r}(0) = (1, 0) = A$ and $\mathbf{r}(1) = (2, 3) = B$.

Path Independence

An immediate consequence of the Fundamental Theorem is that line integrals of conservative fields are path-independent. If $\mathbf{F} = \nabla f$, then $\int_C \mathbf{F} \cdot d\mathbf{r}$ has the same value for any curve $C$ connecting $A$ to $B$.

This is remarkable: whether you take a straight line, a curved arc, or a winding path, you get the same answer! The work done against a conservative force depends only on position change, not on the path taken.

Interactive: Path Independence Demo

Select different paths between the same endpoints and compare their line integral values. For the conservative field, all paths give the same result. For the non-conservative field, different paths give different values:

Closed Curves and Circulation

Another key consequence: for a conservative field, the line integral around any closed curve is zero. If $C$ is a closed curve (starting and ending at the same point $A$), then:

$\oint_C \nabla f \cdot d\mathbf{r} = f(A) - f(A) = 0$

This gives us an equivalent characterization of conservative fields:

$\mathbf{F} \text{ is conservative} \quad \Longleftrightarrow \quad \oint_C \mathbf{F} \cdot d\mathbf{r} = 0 \text{ for all closed curves } C$

Real-World Applications

Physics: Work and Energy

In physics, when a force $\mathbf{F}$ moves an object along a path $C$, the work done is $W = \int_C \mathbf{F} \cdot d\mathbf{r}$.

For a conservative force (like gravity or an ideal spring):

• The force is the negative gradient of potential energy: $\mathbf{F} = -\nabla U$
• Work done: $W = -\int_C \nabla U \cdot d\mathbf{r} = -(U(B) - U(A)) = U(A) - U(B)$
• Work equals the decrease in potential energy—independent of path!

This leads to conservation of mechanical energy: the total $E = K + U$ (kinetic + potential) remains constant for conservative forces.

Machine Learning: Optimization Landscapes

In machine learning, we minimize a loss function $L(\theta)$ over parameters $\theta$. The gradient $\nabla L$ points toward steeper increase, so we move opposite to it:

$\theta_{\text{new}} = \theta_{\text{old}} - \alpha \nabla L$

The Fundamental Theorem tells us that the "work" done against the gradient field—the change in loss—depends only on where we start and end. The path through parameter space doesn't matter for the total change in $L$:

$\int_C \nabla L \cdot d\theta = L(\theta_{\text{end}}) - L(\theta_{\text{start}})$

Interactive: Conservative Field Explorer

Explore different vector fields and compute line integrals between points you choose. Compare the numerical integral with the theoretical value $f(B) - f(A)$ from the potential function:

Python Implementation

Here's a complete implementation for testing conservative fields, finding potential differences, and computing line integrals:

1import numpy as np
2import matplotlib.pyplot as plt
3from scipy.integrate import quad_vec
4
5def is_conservative_2d(P, Q, x, y, h=1e-7):
6    """
7    Test if a 2D vector field F = (P, Q) is conservative.
8
9    For F to be conservative: ∂P/∂y = ∂Q/∂x
10
11    This is equivalent to curl(F) = 0 in 2D.
12
13    Args:
14        P, Q: Component functions of the vector field
15        x, y: Point to test
16        h: Step size for numerical differentiation
17
18    Returns:
19        is_conservative: Boolean
20        curl_value: The curl value (should be ~0 if conservative)
21    """
22    # Compute ∂P/∂y numerically
23    dP_dy = (P(x, y + h) - P(x, y - h)) / (2 * h)
24
25    # Compute ∂Q/∂x numerically
26    dQ_dx = (Q(x + h, y) - Q(x - h, y)) / (2 * h)
27
28    # Curl in 2D: ∂Q/∂x - ∂P/∂y
29    curl = dQ_dx - dP_dy
30
31    is_conservative = abs(curl) < 1e-5
32    return is_conservative, curl
33
34def find_potential_numerical(F, start, end, num_points=100):
35    """
36    Find potential difference using line integral.
37
38    For conservative F = ∇f:
39    f(end) - f(start) = ∫_C F · dr
40
41    This works because line integrals are path-independent
42    for conservative fields.
43
44    Args:
45        F: Vector field function F(x, y) -> (Fx, Fy)
46        start: Starting point (x0, y0)
47        end: Ending point (x1, y1)
48        num_points: Number of points for integration
49
50    Returns:
51        potential_difference: f(end) - f(start)
52    """
53    # Parametrize straight line: r(t) = start + t*(end - start)
54    t_values = np.linspace(0, 1, num_points)
55    dt = 1 / (num_points - 1)
56
57    integral = 0
58    for i in range(len(t_values) - 1):
59        t = (t_values[i] + t_values[i+1]) / 2
60
61        # Position on curve
62        x = start[0] + t * (end[0] - start[0])
63        y = start[1] + t * (end[1] - start[1])
64
65        # Velocity (tangent) vector dr/dt
66        dx = end[0] - start[0]
67        dy = end[1] - start[1]
68
69        # Vector field at this point
70        Fx, Fy = F(x, y)
71
72        # F · dr = F · (dr/dt) dt
73        integral += (Fx * dx + Fy * dy) * dt
74
75    return integral
76
77def line_integral_curved(F, parametric_curve, t_start=0, t_end=1, n=200):
78    """
79    Compute line integral along a parametric curve.
80
81    ∫_C F · dr = ∫_a^b F(r(t)) · r'(t) dt
82
83    Args:
84        F: Vector field F(x, y) -> (Fx, Fy)
85        parametric_curve: Function r(t) -> (x(t), y(t))
86        t_start, t_end: Parameter bounds
87        n: Number of integration points
88
89    Returns:
90        integral_value: The line integral
91    """
92    dt = (t_end - t_start) / n
93    integral = 0
94
95    for i in range(n):
96        t = t_start + (i + 0.5) * dt
97
98        # Position
99        x, y = parametric_curve(t)
100
101        # Velocity (numerical derivative)
102        h = dt / 10
103        x1, y1 = parametric_curve(t - h)
104        x2, y2 = parametric_curve(t + h)
105        dx_dt = (x2 - x1) / (2 * h)
106        dy_dt = (y2 - y1) / (2 * h)
107
108        # Field value
109        Fx, Fy = F(x, y)
110
111        # Integrate F · r'(t) dt
112        integral += (Fx * dx_dt + Fy * dy_dt) * dt
113
114    return integral
115
116# Example: Conservative field F = (2x, 2y) = ∇(x² + y²)
117def F_conservative(x, y):
118    return 2*x, 2*y
119
120def potential_f(x, y):
121    return x**2 + y**2
122
123# Example: Non-conservative field F = (-y, x)
124def F_non_conservative(x, y):
125    return -y, x
126
127# Test conservative field
128print("=== Conservative Field F = (2x, 2y) ===")
129is_cons, curl = is_conservative_2d(
130    lambda x, y: 2*x,
131    lambda x, y: 2*y,
132    1.0, 1.0
133)
134print(f"Is conservative: {is_cons}, curl = {curl:.6f}")
135
136# Compute potential difference via line integral
137start, end = (0, 0), (2, 3)
138integral_value = find_potential_numerical(F_conservative, start, end)
139theoretical = potential_f(*end) - potential_f(*start)
140print(f"Line integral from {start} to {end}: {integral_value:.4f}")
141print(f"Theoretical f(B) - f(A): {theoretical:.4f}")
142
143# Test non-conservative field
144print("\n=== Non-Conservative Field F = (-y, x) ===")
145is_cons, curl = is_conservative_2d(
146    lambda x, y: -y,
147    lambda x, y: x,
148    1.0, 1.0
149)
150print(f"Is conservative: {is_cons}, curl = {curl:.6f}")
151
152# Path 1: Straight line
153path1 = lambda t: (2*t, 3*t)
154integral_straight = line_integral_curved(F_non_conservative, path1)
155
156# Path 2: Curved path (arc)
157path2 = lambda t: (2*t, 3*t + 2*np.sin(np.pi*t))
158integral_curved = line_integral_curved(F_non_conservative, path2)
159
160print(f"Straight path integral: {integral_straight:.4f}")
161print(f"Curved path integral: {integral_curved:.4f}")
162print(f"Difference: {abs(integral_straight - integral_curved):.4f}")
163print("→ Different values prove the field is NOT conservative!")

PyTorch: Conservative Fields via Autograd

Plain Python made the mechanics concrete. PyTorch makes them automatic. Every conservative field comes from a scalar potential $f$, and PyTorch's autograd engine is, at its core, a machine for computing gradients of scalars. So we can manufacture a conservative field on demand: write down a potential, and let autograd produce $\mathbf{F} = \nabla f$ for us.

The same machinery that trains neural networks — reverse-mode automatic differentiation — is doing a calculus exercise here. The connection is not metaphorical: gradient descent in optimization is integrating a conservative field (the gradient of the loss), and the Fundamental Theorem is the reason the total descent depends only on the start and end loss values.

1import torch
2
3# ── 1. Define the SCALAR potential f(x, y) = x²y + 3x − y² ──────────────
4def potential(p):
5    """
6    p has shape (..., 2) with p[..., 0] = x, p[..., 1] = y.
7    Returns a scalar (or tensor of scalars) f(x, y).
8    """
9    x, y = p[..., 0], p[..., 1]
10    return x**2 * y + 3*x - y**2
11
12# Endpoints A = (1, 0) and B = (2, 3).
13A = torch.tensor([1.0, 0.0])
14B = torch.tensor([2.0, 3.0])
15
16# ── 2. FTLI — the "easy way" in three lines ─────────────────────────────
17ftli_value = potential(B) - potential(A)
18print(f"f(B) - f(A) = {ftli_value.item():.4f}")  # 6.0000
19
20# ── 3. Build the CONSERVATIVE field F = ∇f using autograd ───────────────
21def conservative_F(p):
22    """
23    Return the gradient of f at point p. Because F = ∇f, this IS the
24    conservative vector field — no hand-derivation of partials needed.
25    """
26    p = p.clone().detach().requires_grad_(True)
27    val = potential(p)
28    grad = torch.autograd.grad(val, p)[0]
29    return grad
30
31# Sanity check at (1, 2): ∂f/∂x = 2xy+3 = 7,  ∂f/∂y = x²−2y = −3.
32print("F(1, 2) =", conservative_F(torch.tensor([1.0, 2.0])))
33# tensor([ 7., -3.])
34
35# ── 4. Direct line integral along a CURVED path r(t) = (1+t, 3t²) ───────
36n = 500
37t = torch.linspace(0.0, 1.0, n)
38x_curve = 1.0 + t
39y_curve = 3.0 * t**2
40
41# Tangent r'(t) = (1, 6t).
42drdt_curve = torch.stack([torch.ones_like(t), 6.0 * t], dim=-1)
43
44# Sample F along the path one point at a time.
45points_curve = torch.stack([x_curve, y_curve], dim=-1)
46F_curve = torch.stack([conservative_F(p) for p in points_curve])
47
48# Integrand: F(r(t)) · r'(t)
49integrand_curve = (F_curve * drdt_curve).sum(dim=-1)
50
51# Trapezoidal rule on [0, 1]
52integral_curve = torch.trapz(integrand_curve, t)
53print(f"Curved path ∫ ∇f·dr = {integral_curve.item():.4f}")  # ≈6.0000
54
55# ── 5. Same endpoints, different path: STRAIGHT line r(t) = (1+t, 3t) ───
56x_line = 1.0 + t
57y_line = 3.0 * t
58points_line = torch.stack([x_line, y_line], dim=-1)
59F_line = torch.stack([conservative_F(p) for p in points_line])
60drdt_line = torch.stack([torch.ones_like(t), 3.0 * torch.ones_like(t)], dim=-1)
61integral_line = torch.trapz((F_line * drdt_line).sum(dim=-1), t)
62print(f"Straight path ∫ ∇f·dr = {integral_line.item():.4f}")  # ≈6.0000
63
64# ── 6. All three numbers must agree — that IS path independence ─────────
65print("\nPath independence ✓" if torch.allclose(
66    integral_curve, ftli_value, atol=1e-2
67) else "MISMATCH ✗")

Common Pitfalls

Test Your Understanding

Summary

The Fundamental Theorem for Line Integrals is one of the most important results in vector calculus. Here are the key takeaways:

1. Conservative Fields: A field $\mathbf{F}$ is conservative if $\mathbf{F} = \nabla f$ for some potential function $f$. Test: $\frac{\partial P}{\partial y} = \frac{\partial Q}{\partial x}$.
2. The Fundamental Theorem: $\int_C \nabla f \cdot d\mathbf{r} = f(B) - f(A)$. The line integral depends only on endpoints, not on the path.
3. Path Independence: For conservative fields, all paths between two points give the same integral value.
4. Closed Curves: The circulation around any closed curve is zero for conservative fields.
5. Physical Meaning: Conservative forces conserve mechanical energy; work done depends only on position change.

Looking Ahead: In the next section, we'll study Green's Theorem, which connects line integrals around closed curves to double integrals over the enclosed region. This provides another way to test whether fields are conservative and leads to powerful computational techniques.