GATE Electrical: Calculus and Differential Equations (Engineering Mathematics), previous-year questions solved
Every calculus and differential-equation question from the last five GATE Electrical papers: where a function rises or falls, convexity and minima, differentiability, a symmetric improper integral, and linear versus nonlinear ODEs with their solutions. Each of the 11 solutions matches the official key, was checked numerically (Simpson integration, RK4, counterexamples), names the trap and shows where the idea is used in electrical engineering.
Questions from official GATE EE papers (IIT Kharagpur 2022, IIT Kanpur 2023, IISc Bengaluru 2024, IIT Roorkee 2025, IIT Guwahati 2026); solutions by Skill-Lync. · Updated 16 Nov 2026 · 11 questions
f(x) = ∫ from 0 to x of eᵗ(t − 1)(t − 2) dt: in which interval does f decrease?
- (A)
- (B)
- (C)
- (D)
Answer: (A) Official key: A
- Don't integrate. By the fundamental theorem of calculus, .
- always, so the sign of is the sign of : negative only between the roots, for . Option (A).
- On and both factors are negative, so and rises; on both are positive and rises again.
Check by numerical integration: and , so does fall between and , and it falls on no other option's interval.
Energy is the integral of power: . Whenever the integrand changes sign, the running integral turns. In an inductor or capacitor fed with AC, is negative for part of each cycle, which is exactly when the stored energy is handed back to the source; the turning points of sit where or crosses zero, just as turns at and here.
Curve |x₁|ᵖ + |x₂|ᵖ = 1: which statements about the enclosed area are true?
- (A)When , the area enclosed by the curve is .
- (B)When tends to , the area enclosed by the curve tends to 4.
- (C)When tends to 0, the area enclosed by the curve is 1.
- (D)When , the area enclosed by the curve is 2.
Answer: (A), (B) and (D) Official key: A, B, D
- Sketch the curve for the easy values of . : , the unit circle, area . (A) is true.
- : , a square standing on a corner with vertices and . Its diagonals are 2 long, so its area is . (D) is true.
- : the larger of dominates, so the curve becomes , the square with corners . Area . (B) is true.
- : the curve is pulled in towards the two axes (a four-pointed star), and the area shrinks towards 0, not 1. (C) is false. Answer (A), (B) and (D).
The exact area is . It gives and 2 for and , 3.997 for , 0.667 for and only 0.00002 for . Trap: (C) looks plausible because , but at the equation reads and has no points at all.
These are the unit balls of the -norms. The 2-norm (circle) is RMS, the 1-norm (diamond) is the sum of absolute errors, and the -norm (square) is the peak error. Power engineers use all three: an RMS (2-norm) current sets heating, a peak (-norm) voltage sets insulation and semiconductor ratings, and a sum of absolute deviations (1-norm) is often the cost function in robust state estimation, because it is less thrown by one bad meter reading.
f″(t) > 0 for all t: which statements are always true?
- (A) has at least one local minimum.
- (B) cannot have two distinct local minima.
- (C) has at least one local maximum.
- (D)The minimum value of cannot be negative.
Answer: (B) Official key: B
- means is strictly increasing, so can be zero at most once. A local minimum of a smooth function needs , so there is at most one: (B) is always true.
- (A) fails for : , but is never zero, so there is no minimum at all (the function only approaches 0 as ).
- (C) fails for every such function: a local maximum needs there.
- (D) fails for : convex, with minimum value . Answer (B).
We ran the same tests on , , , and (all with ): none had two local minima or any local maximum, had no minimum and went negative. Trap: (A) is what you'd say from a picture of a parabola; convex does not mean it must turn back up.
Optimal power flow and economic dispatch rely on exactly this property. With quadratic generator cost curves and , the total cost is convex, so a solver that finds a point where the incremental costs are equal has found the one global minimum and need not search further. Losses and valve-point effects make the real curves non-convex, which is why practical OPF can get stuck in local minima.
f(t) = (max(0, t))²: differentiable? Is the derivative continuous? Differentiable twice?
- (A) is not differentiable.
- (B) is differentiable and its derivative is continuous.
- (C) is differentiable but its derivative is not continuous.
- (D) and its derivative are differentiable.
Answer: (B) Official key: B
- Write it in pieces: for and for . Only needs checking.
- Left derivative at 0: 0. Right derivative: . They agree, so is differentiable everywhere, with .
- is continuous (it is a ramp). (B) is true; (A) and (C) are false.
- But has a corner at 0: its left slope is 0 and its right slope is 2. So is not differentiable at 0 and (D) is false. Answer (B).
Trap: itself has a corner at 0, and it is tempting to say (A). Squaring it flattens the corner once; it would take a cube to make the second derivative continuous too.
This is the shape of a soft-start or ramp command: position follows when the acceleration steps from 0 to a constant. Velocity (the first derivative) is continuous, but acceleration jumps, so jerk is infinite at the switch-on instant. Motor drive and elevator controllers use S-curve profiles, which add one more continuous derivative, to avoid that jerk and the torque spike it causes.
Minimise (x + y − 1)² + (x + y)²: how many points minimise it?
- (A)The number of elements in the set is finite and more than one.
- (B)The number of elements in the set is infinite.
- (C)The set is empty.
- (D)The number of elements in the set is exactly one.
Answer: (B) Official key: B
- depends on and only through : .
- Minimise in : , so and the minimum value is .
- Every point on the line gives that minimum: , , and so on. Infinitely many: option (B).
Check: on a grid from to 2 in steps of 0.02, the smallest value of is exactly 0.5, and it is reached all along that line. The Hessian is , singular, so the second-derivative test cannot give a single point. Trap: setting both partial derivatives to zero gives the same equation twice, ; that is one line, not a unique solution (D).
Least-squares problems hit this when two unknowns only ever appear together, for example two parallel lines or two identical generators whose individual flows the meters cannot tell apart. The normal-equation matrix is singular and the estimator has infinitely many equally good answers. In power-system state estimation this is called an unobservable network: the fix is more measurements (or a pseudo-measurement), not a better solver.
(1/π) ∫ from 0 to ∞ of x²⁰²⁶ / ((1 + x²⁰²⁶)(1 + x²)) dx
Answer: 0.25 Official key: 0.25 to 0.25
- Call the integral (without the ) , and substitute , . The limits swap, and .
- So as well.
- Add the two forms: .
- , and 0.25.
The 2026 is decoration: any power gives the same answer. Numerically (with the integrand becomes a steep step at ) we get 0.250000. A rough 'the integrand is 0 below and above it' estimate also lands at 0.2500, which is why the answer is exactly a quarter.
The same symmetry appears in circuit theory as frequency inversion: replacing by turns a low-pass response into a high-pass one. Filter designers use it to design one normalised low-pass prototype (Butterworth, Chebyshev) and get high-pass and band-pass filters by substitution, and integrals of the response over all frequencies, such as noise bandwidth, are often found with the same trick.
dy/dt = e^(−αt)/(2 + αt), α = 0.01, y(0) = 0: find y(1)
Answer: 0.50 Official key: 0.48 to 0.52
- The right side depends on only, so .
- With the integrand barely changes: at and at .
- Trapezoidal rule (one step): , which is already the answer to 2 decimals: 0.50.
RK4 with 2000 steps gives . Even one Euler step () lands inside the key range of 0.48 to 0.52, which is why the question says 'numerically obtained' and accepts a band.
Every circuit and power-system simulator (SPICE, PSCAD, PLECS, Simulink) solves differential equations numerically like this, step by step. The step size and method matter when the solution changes fast, for example the switching edges of a converter, which is why those tools use variable-step solvers and why a fixed-step Euler simulation of a stiff circuit can blow up.
Which of these four first-order differential equations are nonlinear?
- (A)
- (B)
- (C)
- (D)
Answer: (B) and (D) Official key: B; D
- An ODE is linear when and its derivatives appear only to the first power, are not multiplied together, and are not inside another function. The coefficients may depend on .
- (A): . Coefficient multiplies : linear (time-varying).
- (B): is a product of the unknown and its derivative: nonlinear.
- (C): and are coefficients that depend on only: linear.
- (D): puts the derivative inside an exponential: nonlinear. Answer (B) and (D).
We tested each equation by superposition: subtract the forcing term, then check that the left side for equals the sum of the left sides, and that scaling by scales it by . (A) and (C) pass at every test point; (B) and (D) fail. Trap: , and look nonlinear, but they are functions of the independent variable , not of .
Linear time-varying equations like (A) and (C) still obey superposition, so the methods of circuit analysis still apply; nonlinear ones do not. In power electronics the switch makes the converter time-varying, and the averaged model is linearised around an operating point before a compensator is designed. An inductor whose core saturates is nonlinear in the sense of (B) and (D): its inductance depends on its own current.
ẋ₁ = 2x₂, ẋ₂ = r(t) (unit step), x₁(0) = 1, x₂(0) = 0: find x₁(1)
Answer: 2 Official key: 2 to 2
- Solve from the bottom up. with gives .
- Then , so .
- At : 2.
RK4 gives and . Trap: forgetting the initial condition gives 1.
This is a double integrator in state-space form, the model of a mass driven by a force or a motor shaft driven by torque (position and speed as the states). A step in torque makes speed ramp and position grow as . Motion controllers for servo drives are designed on exactly this model before friction and load are added.
t² y″ + 7t y′ + 8t y = 10 sin t: linear, nonlinear, time-invariant or partial?
- (A)It is a linear differential equation
- (B)It is a nonlinear differential equation
- (C)It is a time-invariant differential equation
- (D)It is a second-order partial differential equation
Answer: (A) Official key: A
- , and each appear to the first power and are not multiplied together; the coefficients , and depend only on . So it is linear: option (A).
- It is not time-invariant (C): the coefficients change with , so delaying the input does not just delay the output.
- It is not a PDE (D): there is only one independent variable, .
A superposition test on the left side (minus the forcing) holds at , 1.7 and 3.2. Trap: the and terms look like products, but is the independent variable; only products of with or its derivatives make an equation nonlinear.
Linear time-varying systems are common in electrical engineering: a capacitor whose capacitance is varied (a varactor or a capacitive sensor), a machine whose inductances change with rotor angle, a switched converter. They are still linear, so superposition works for each input, but frequency-domain transfer functions do not apply directly; Park's transform is used on machines precisely to turn the time-varying inductance equations into time-invariant ones.
y″ + y′ + y = 0, y(0) = 1, y′(0) = 1: which solution is correct?
- (A)
- (B)
- (C)
- (D)
Answer: (A) Official key: A
- Characteristic equation : . So .
- gives (all four options already have it).
- , so . Option (A).
RK4 from , gives , and only option (A) matches it (to ) at , 1, 2 and 4. Trap: forgetting the that the exponential contributes to gives , which is in no option; option (B) is the solution for .
This is the series RLC circuit, , with : damping ratio , so it rings at while decaying as . A second-order step response with overshoots by about 16%, a common design target for voltage and current control loops.