Solve Linear Systems With Complex Coefficients

You set up the augmented matrix for a linear system as usual: coefficients on the left, constants on the right. The matrix for (1+i)x + (2-i)y = 3+2i and (2+i)x + (1+2i)y = 4+3i looks like any other 2x3 array. The difference hits at the first elimination step: you must multiply a row by a complex scalar. That scalar is (2+i)/(1+i), and computing it requires the conjugate method: multiply numerator and denominator by (1-i), get (3-i)/2 = 1.5 - 0.5i. Subtract that multiple of the first row from the second row. This is the moment that separates real linear systems from those that solve linear systems with complex coefficients. Every operation from this point forward uses complex arithmetic: i² = -1, and division always uses the conjugate of the denominator.

The steps of Gaussian elimination do not change. You still choose pivots, eliminate below, and back-substitute. What changes is the arithmetic inside each step. Addition and subtraction treat the real and imaginary parts separately but combine them in the same operation. Multiplication of a row by a complex scalar multiplies both parts, with the cross term producing a new imaginary contribution via i² = -1. Division of one complex number by another requires the conjugate: (a+bi)/(c+di) = [(a+bi)(c-di)]/(c²+d²). This doubles the number of multiplications per operation compared to a real-only system.

Pivot Selection and Complex Zero

A pivot is zero only when both its real part and its imaginary part are zero: 0+0i. If a pivot position contains, say, 0+2i, it is not zero. That pivot can be used directly. You do not need to swap rows unless the entry is exactly 0+0i. This is a common checkpoint where students misjudge: a non-zero imaginary part means the pivot is valid, even if the real part is zero.

The Augmented Matrix Stays the Same Shape

The augmented matrix for a 2x2 system with complex coefficients is still 2 rows by 3 columns. Each entry is a complex number, written as a+bi or in column vector form. The operations on the matrix are identical to the real case: row swapping, scaling, and addition of a multiple of one row to another. The only difference is that the scalar and the results are complex.

Work through the system introduced above: (1+i)x + (2-i)y = 3+2i and (2+i)x + (1+2i)y = 4+3i. Write the augmented matrix:

[1+i, 2-i | 3+2i]
[2+i, 1+2i | 4+3i]

Eliminate x from the second row. Multiply the first row by (2+i)/(1+i). Compute the scalar: (2+i)/(1+i). Multiply numerator and denominator by the conjugate (1-i): (2+i)(1-i) / (1+1) = (2 -2i + i -i²) / 2 = (2 -2i + i +1) / 2 = (3 - i)/2 = 1.5 - 0.5i. Subtract 1.5 - 0.5i times the first row from the second row. First compute the product for the y coefficient: (1.5 - 0.5i) * (2 - i) = 3 - 1.5i - i + 0.5i² = 3 - 2.5i - 0.5 = 2.5 - 2.5i. Subtract this from the second row's y coefficient: (1+2i) - (2.5 - 2.5i) = (1 - 2.5) + (2i + 2.5i) = -1.5 + 4.5i. Now the constant term: (1.5 - 0.5i) * (3+2i) = 4.5 + 3i - 1.5i - i² = 4.5 + 1.5i + 1 = 5.5 + 1.5i. Subtract from the second row's constant: (4+3i) - (5.5 + 1.5i) = (4 - 5.5) + (3i - 1.5i) = -1.5 + 1.5i.

The second row now reads: [0, -1.5 + 4.5i | -1.5 + 1.5i]. Solve for y: y = (-1.5 + 1.5i) / (-1.5 + 4.5i). Multiply numerator and denominator by the conjugate of the denominator: (-1.5 + 1.5i) * (-1.5 - 4.5i) / ((-1.5)² + (4.5)²). Compute numerator: (-1.5)(-1.5) + (-1.5)(-4.5i) + (1.5i)(-1.5) + (1.5i)(-4.5i) = 2.25 + 6.75i - 2.25i - 6.75i² = 2.25 + 4.5i + 6.75 = 9 + 4.5i. Denominator: 2.25 + 20.25 = 22.5. So y = (9 + 4.5i) / 22.5 = 0.4 + 0.2i.

Substitute y into the first equation: (1+i)x + (2-i)(0.4+0.2i) = 3+2i. Compute (2-i)(0.4+0.2i) = 0.8 + 0.4i - 0.4i - 0.2i² = 0.8 + 0.2 = 1.0. So (1+i)x + 1 = 3+2i, giving (1+i)x = 2+2i. Divide: x = (2+2i)/(1+i). Multiply numerator and denominator by (1-i): (2+2i)(1-i) / 2 = (2 -2i +2i -2i²)/2 = (2 +2)/2 = 2. So x = 2.

The solution is x = 2, y = 0.4 + 0.2i. You can verify by substituting back: first equation: 2(1+i) + (2-i)(0.4+0.2i) = 2+2i + 1 = 3+2i. Correct.

When you reach the point of solving for y above, the calculation involves a division of two complex numbers. This is where a gaussian elimination calculator becomes necessary if you need to check your work or handle larger systems. The calculator performs the conjugate multiplication and division automatically, but you must understand the steps to catch errors.

Quadrant Error in Complex Division

The most frequent error in manual calculation is the quadrant slip. When you compute the argument of a complex number for division in polar form, arctan(b/a) only gives the correct angle when a > 0. If a is negative, you must add π (or 180°) to the result. In the example above, the denominator -1.5 + 4.5i has a negative real part. If you use arctan(4.5 / -1.5) = arctan(-3) ≈ -71.6°, the true argument is -71.6° + 180° = 108.4°. Failing to add the 180° flips the sign of the division result entirely.

Conjugate Sign Slip

A second common failure is writing the conjugate with the wrong sign. For a denominator a+bi, the conjugate is a-bi. If you write a+bi instead, the denominator does not become real, and the division gives a wrong complex number. In the y calculation above, the denominator was -1.5 + 4.5i. Its conjugate is -1.5 - 4.5i. A slip writing it as 1.5 - 4.5i or -1.5 + 4.5i would produce a different numerator and denominator, leading to an incorrect y.

The explanation for why row reduction works with complex coefficients is the same as for real coefficients: the operations are linear and reversible. Adding a multiple of one row to another does not change the solution set. Scaling a row by a non-zero complex number is reversible by scaling by its reciprocal. The only requirement is that the scalar be non-zero, meaning its modulus is greater than zero. A complex number like 3+0i is non-zero, as is 0+5i.

Arithmetic Complexity Increases

Each multiplication of two complex numbers requires four real multiplications and two additions. Division adds a conjugate multiplication and a real division. For a 3x3 system, the number of operations roughly quadruples compared to a real system. This makes manual calculation tedious and error-prone beyond 2x2 systems. The recommendation is to use exact fractions or rationalized forms at each step, not decimal approximations, until the final answer.

Failure Case: Pivot Swapping with Complex Zero

A specific failure case occurs when the first pivot is 0+0i. In that situation, you must swap rows. But if the pivot is, say, 0+0.001i, it is not zero. The elimination proceeds, but the pivot is very small. This leads to numerical instability: multiplying by a large scalar to eliminate below amplifies rounding errors. In manual calculation, you can avoid this by swapping rows anyway, even if the pivot is technically non-zero. The choice is a judgement call based on the magnitude of the pivot's modulus.

Step	Operation	Result
1	Write augmented matrix	[1+i, 2-i | 3+2i; 2+i, 1+2i | 4+3i]
2	Compute scalar (2+i)/(1+i)	1.5 - 0.5i
3	Multiply row 1 by scalar, subtract from row 2	New row 2: [0, -1.5+4.5i | -1.5+1.5i]
4	Solve for y using conjugate division	y = 0.4 + 0.2i
5	Back-substitute into row 1	x = 2

The arithmetic rules that change from real to complex systems are few but critical:

• Addition and subtraction: Combine real with real, imaginary with imaginary. (a+bi) + (c+di) = (a+c) + (b+d)i.
• Multiplication: Use FOIL, then replace i² with -1. (a+bi)(c+di) = ac + adi + bci + bdi² = (ac - bd) + (ad + bc)i.
• Division: Multiply numerator and denominator by the conjugate of the denominator. (a+bi)/(c+di) = [(a+bi)(c-di)]/(c²+d²).
• Zero check: A complex number is zero only when both real and imaginary parts are zero.

These rules are the only new content. The elimination algorithm itself is identical to the real case.

This approach to solving linear systems with complex coefficients suits precalculus and college algebra students who need to perform complex arithmetic in a structured algorithm. Engineering undergraduates in circuits or signals courses will use these techniques for AC steady-state analysis. Physics students working with oscillating functions will encounter similar systems. Exam revisers for A-level Further Maths or IB will need fast, accurate manipulation of complex numbers in elimination.

Skip this subject if you are a pure mathematician researching complex analysis; the algebraic row reduction here is far below that level. High school students who have not yet seen the quadratic formula lack the prerequisite algebra. Professionals doing high-precision numerical computation should use MATLAB or Python's cmath, not manual calculation.

The single thing that most often goes wrong is the quadrant error in the argument. When you compute arctan(b/a) and the real part a is negative, you must add 180° (or π radians). Forgetting that step flips the sign of the entire elimination result.