Linear fractional transformations: the formula, how to map three points, and practice
Learn the formula, the building blocks and the cross-ratio method, see how lines and circles map, and work through four practice questions with answers.
On this page
A linear fractional transformation, also called a Möbius transformation, is a map T(z) = (az + b)/(cz + d) with ad − bc ≠ 0. This guide explains that formula and the simple maps it's built from, then shows how to use the cross-ratio to find the one map that sends three points to three points. It finishes with worked examples and four practice questions with full working.
What is a linear fractional transformation?
A linear fractional transformation is any map of the form
T(z) = (az + b)/(cz + d), with a, b, c, d complex and ad − bc ≠ 0.
You'll also see it called a Möbius transformation, a bilinear transformation or a fractional linear transformation. All four names mean the same map.
The condition ad − bc ≠ 0 stops T from collapsing to a constant. The derivative shows why:
T′(z) = (ad − bc)/(cz + d)²
If ad − bc = 0, the derivative is zero everywhere and T is constant. If ad − bc ≠ 0, then T′ is never zero, so T is conformal: it preserves angles wherever it's defined.
It's easiest to work on the extended complex plane, which is the complex numbers plus one extra point, ∞. There, T is a one-to-one map of the extended plane onto itself, and the awkward points have definite images.
| Point z | Image T(z) | When |
|---|---|---|
| −d/c | ∞ | c ≠ 0 |
| ∞ | a/c | c ≠ 0 |
| ∞ | ∞ | c = 0 |
| −b/a | 0 | a ≠ 0 |
Source: Orloff's notes for MIT 18.04, section 11.7. Last checked: November 17, 2026.
The inverse is also a linear fractional transformation: T⁻¹(w) = (dw − b)/(−cw + a). You can find it by solving w = (az + b)/(cz + d) for z, or with the matrix shortcut below.
Building blocks: translation, dilation and inversion
Every linear fractional transformation is a chain of three simple maps.
| Map | Formula | Matrix | Effect |
|---|---|---|---|
| Translation | z + b | [1 b; 0 1] | Shifts by b |
| Dilation | az | [a 0; 0 1] | Scales and rotates |
| Inversion | 1/z | [0 1; 1 0] | Swaps 0 and ∞ |
Source: MIT 18.04 notes, topic 10. Last checked: November 17, 2026.
Inversion does the interesting work. Since 1/z = z̄/|z|², it's a reflection in the unit circle followed by a reflection in the real axis.
If c = 0, then T(z) = (a/d)z + b/d, which is a dilation followed by a translation. If c ≠ 0, divide out to get
T(z) = a/c + (bc − ad)/(c(cz + d))
Read it from the inside out: multiply by c and add d, invert, multiply by (bc − ad)/c, then add a/c. For example, (2z + 1)/(z + 1) = 2 − 1/(z + 1), which breaks down as:
- z ↦ z + 1 (translate by 1).
- z + 1 ↦ 1/(z + 1) (invert).
- 1/(z + 1) ↦ −1/(z + 1) (rotate by π).
- −1/(z + 1) ↦ 2 − 1/(z + 1) (translate by 2).
Composing Möbius transformations with 2×2 matrices
Match T(z) = (az + b)/(cz + d) with the matrix [a b; c d]. Composing maps then becomes matrix multiplication: if A is the matrix of T and B is the matrix of S, then T∘S has matrix AB.
Two facts make this useful. Any nonzero multiple of a matrix gives the same map, so [2 2; 0 2] and [1 1; 0 1] are both z + 1. And the inverse matrix gives the inverse map, and since you can drop the 1/(ad − bc) factor, T⁻¹ matches [d −b; −c a]. This also explains the condition on ad − bc. It's the determinant, so it can't be zero.
Order matters. Take S(z) = 1/z and T(z) = z + 1:
- T∘S has matrix [1 1; 0 1][0 1; 1 0] = [1 1; 1 0], so T∘S(z) = (z + 1)/z.
- S∘T has matrix [0 1; 1 0][1 1; 0 1] = [0 1; 1 1], so S∘T(z) = 1/(z + 1).
Check both by substituting directly. They agree, and they're clearly different maps.
How to find the map that sends three points to three points
Use the cross-ratio. For distinct points z₁, z₂, z₃, define
(z, z₁, z₂, z₃) = (z − z₁)(z₂ − z₃) / ((z − z₃)(z₂ − z₁))
As a function of z, this is a linear fractional transformation that sends z₁ to 0, z₂ to 1 and z₃ to ∞. To send z₁, z₂, z₃ to w₁, w₂, w₃, set the two cross-ratios equal and solve for w:
(w, w₁, w₂, w₃) = (z, z₁, z₂, z₃)
Both sides send matched points to 0, 1 and ∞. A Möbius map that fixes three points is the identity, so the answer is unique. If one of the points is ∞, cancel the two factors that contain it. With z₃ = ∞, for example, the cross-ratio becomes (z − z₁)/(z₂ − z₁). MIT's lecture 7 for 18.112 covers the cross-ratio in more depth.
Example 1: mapping three points
Find T with T(0) = i, T(1) = 1 and T(∞) = −i.
- Left side: z₃ = ∞, so (z, 0, 1, ∞) = (z − 0)/(1 − 0) = z.
- Right side: (w, i, 1, −i) = (w − i)(1 + i)/((w + i)(1 − i)). Since (1 + i)/(1 − i) = i, this is i(w − i)/(w + i).
- Set them equal: i(w − i)/(w + i) = z, so iw + 1 = zw + iz.
- Collect the w terms: w(i − z) = iz − 1, so w = (1 − iz)/(z − i).
- Check: T(0) = 1/(−i) = i, T(1) = (1 − i)/(1 − i) = 1, and T(∞) = a/c = −i. Also ad − bc = (−i)(−i) − 1 = −2, which isn't zero.
Textbooks order the cross-ratio differently, and some send z₁ to 1 instead of 0. Because the map is unique, you get the same answer with any convention, as long as you use the same one on both sides. You can also substitute the three points into (az + b)/(cz + d) and solve the linear equations. It's slower, but it's a good check.
Why lines and circles map to lines and circles
Think of a line as a circle through ∞. Every circle and line in the plane satisfies
A(x² + y²) + Bx + Cy + D = 0
for real A, B, C and D, where A = 0 gives a line. Translations and dilations clearly keep circles as circles and lines as lines, so only inversion needs checking.
Under w = 1/z, write w = u + iv. Then x = u/(u² + v²), y = −v/(u² + v²) and x² + y² = 1/(u² + v²). Substitute and multiply through by u² + v²:
D(u² + v²) + Bu − Cv + A = 0
That's the same form, so the image is a circle or a line. Every linear fractional transformation is a chain of these building blocks, so the result holds for all of them. The image is a line exactly when the original curve passes through −d/c, the point sent to ∞. For 1/z, that's the origin, which is the case D = 0.
Example 2: the image of a circle
Find the image of the circle |z − 1| = 1 under w = 1/z.
Expanding gives x² + y² − 2x = 0, so A = 1, B = −2, C = 0 and D = 0. The image is 1 − 2u = 0, which is the line Re w = 1/2. To see where the inside of the circle goes, test one point: z = 1 gives w = 1, and Re w = 1 > 1/2. So the disk maps to the half-plane Re w > 1/2.
Worked examples
These three examples are fully worked, so read them step by step. The practice questions further down are for you to try on your own first.
Worked example: finding fixed points
Find the fixed points of T(z) = (2z + 1)/(z + 2).
- Check: ad − bc = 2 × 2 − 1 × 1 = 3 ≠ 0.
- Set T(z) = z: (2z + 1)/(z + 2) = z, so 2z + 1 = z² + 2z.
- Simplify: z² = 1, so z = 1 or z = −1.
- Check: T(1) = 3/3 = 1 and T(−1) = −1/1 = −1.
Since c ≠ 0, ∞ isn't fixed: T(∞) = a/c = 2.
Worked example: composing a map with itself
Let T(z) = (z + 1)/(z − 1). Use matrices to find T∘T.
- The matrix of T is [1 1; 1 −1].
- Square it: [1 1; 1 −1][1 1; 1 −1] = [1 + 1, 1 − 1; 1 − 1, 1 + 1] = [2 0; 0 2].
- [2 0; 0 2] is a multiple of the identity matrix, so T∘T(z) = 2z/2 = z.
- So T is its own inverse. Check with T(2) = 3/1 = 3 and T(3) = 4/2 = 2.
Worked example: the image of a half-plane
Find the image of the right half-plane Re z > 0 under T(z) = (z − 1)/(z + 1).
- The boundary is the imaginary axis. T sends −1 to ∞, and −1 isn't on the imaginary axis, so the image of the axis is a circle.
- Find three image points: T(0) = −1, T(∞) = a/c = 1 and T(i) = (i − 1)/(i + 1) = i.
- The circle through −1, 1 and i is the unit circle, |w| = 1.
- Test a point in the right half-plane: T(1) = 0/2 = 0, which is inside the unit circle.
- So the right half-plane maps to the unit disk, |w| < 1.
Mistakes that cost points on Möbius transformation questions
Most lost points on these questions come from a handful of slips:
- Skipping the ad − bc check. If it's zero, the formula gives a constant and none of the theory applies.
- Composing in the wrong order. T∘S means do S first, and its matrix is AB, with T's matrix on the left.
- Using one cross-ratio convention on the z side and another on the w side.
- Leaving ∞ in a formula. Cancel the two factors that contain it.
- Assuming a circle maps to a circle. First check whether it passes through −d/c.
- Shading the wrong region. Test one point from inside the original region.
- Not checking the final map. Plug in all three points before you move on.
If your course uses its own notation, upload your lecture notes to Medly. It builds practice tests from your own lectures, in the style your exams use.
Practice questions
Now it's your turn. Try each question yourself before you look at the answer. The full answer and working for each one are hidden in the Answer dropdown directly below it.
Question 1: (2 points)
Which of these is a linear fractional transformation?
A. T(z) = (2z + 4)/(z + 2)
B. T(z) = (z − i)/(z + i)
Answer
- For A, a = 2, b = 4, c = 1 and d = 2, so ad − bc = 4 − 4 = 0. It isn't one: T(z) = 2 for every z ≠ −2.
- For B, a = 1, b = −i, c = 1 and d = i, so ad − bc = i − (−i) = 2i ≠ 0. It is one.
Map B sends the upper half-plane to the unit disk. For example, T(i) = 0.
Question 2: (3 points)
Let T(z) = (3z + 1)/(z + 2). Find T(∞), the point that T sends to ∞, and T⁻¹(w).
Answer
- Check: ad − bc = 3 × 2 − 1 × 1 = 5 ≠ 0.
- T(∞) = a/c = 3 ÷ 1 = 3.
- The denominator is zero at z = −2, so T(−2) = ∞.
- Solve w(z + 2) = 3z + 1: wz − 3z = 1 − 2w, so z = (1 − 2w)/(w − 3).
- T⁻¹(w) = (2w − 1)/(3 − w). This matches the matrix [d −b; −c a] = [2 −1; −1 3].
Question 3: (4 points)
Find the linear fractional transformation that sends −1, 0 and 1 to −1, i and 1 respectively.
Answer
- Left side: (z, −1, 0, 1) = (z + 1)(0 − 1)/((z − 1)(0 + 1)) = (z + 1)/(1 − z).
- Right side: (w, −1, i, 1) = (w + 1)(i − 1)/((w − 1)(i + 1)). Since (i − 1)/(i + 1) = i, this is i(w + 1)/(w − 1).
- Cross-multiply: i(w + 1)(1 − z) = (z + 1)(w − 1).
- Expand: iw − iwz + i − iz = zw − z + w − 1.
- Collect the w terms: w(i − iz − z − 1) = iz − z − 1 − i.
- So w = ((i − 1)z − (1 + i))/(−(1 + i)z + (i − 1)).
- Divide the top and bottom by i − 1, using (1 + i)/(i − 1) = −i: w = (z + i)/(iz + 1).
- Check: T(−1) = (i − 1)/(1 − i) = −1, T(0) = i and T(1) = (1 + i)/(1 + i) = 1.
Question 4: (4 points)
Find the image of the unit disk |z| < 1 under T(z) = (z + 1)/(z − 1).
Answer
- The unit circle passes through z = 1, where the denominator is zero, so its image is a line.
- Find three image points: T(−1) = 0, T(i) = (1 + i)/(i − 1) = −i and T(−i) = (1 − i)/(−1 − i) = i.
- The line through 0, −i and i is the imaginary axis, Re w = 0.
- Test a point inside the disk: T(0) = 1/(−1) = −1, which has negative real part.
- So the unit disk maps to the left half-plane, Re w < 0.
The usual slip is guessing the right half-plane. Testing z = 0 settles it.
For more questions like these, upload your complex analysis lectures and notes to Medly. It builds custom practice tests from your own course material, and you can highlight any part of your slides and ask about it. It's free to start, with no card needed.
Frequently asked questions
Is a Möbius transformation the same as a linear fractional transformation?
Yes. Möbius transformation, linear fractional transformation, fractional linear transformation and bilinear transformation all mean T(z) = (az + b)/(cz + d) with ad − bc ≠ 0.
Why must ad − bc be nonzero?
If ad − bc = 0, the derivative (ad − bc)/(cz + d)² is zero everywhere, so the map is constant. In matrix terms, ad − bc is the determinant, and the map is only invertible when the determinant is nonzero.
How many points determine a linear fractional transformation?
Three. Any three distinct points and any three distinct targets determine exactly one linear fractional transformation, which you can find with the cross-ratio.
Is a linear fractional transformation a linear map?
No, not in the linear algebra sense, unless c = 0 and b = 0. The name comes from the numerator and denominator both being linear in z.
How many fixed points does a linear fractional transformation have?
At most two, unless it's the identity. Setting T(z) = z gives cz² + (d − a)z − b = 0. When c = 0, ∞ is also a fixed point.
What is the inverse of a linear fractional transformation?
T⁻¹(w) = (dw − b)/(−cw + a). It comes from the inverse matrix [d −b; −c a], with the determinant factor dropped.