GATE Mechanical: Calculus and Vector Calculus (Engineering Mathematics), previous-year questions solved
Calculus is the biggest Engineering Mathematics section in GATE Mechanical. The five keyed papers from 2022 to 2026 set 16 calculus questions (20 marks), more than linear algebra (7) and differential equations (8) together. Here they all are: limits, maxima and minima, integrals, Fourier series and vector calculus, each solved with the exam shortcut and the trap behind the wrong options.
Questions from official GATE ME papers (IIT Kharagpur 2022, IISc Bengaluru 2024, IIT Roorkee 2025, IIT Guwahati 2026); solutions by Skill-Lync. · Updated 16 Nov 2026 · 16 questions
Limit as x → π of (x² + αx + 2π²)/(x − π + 2 sin x): which α makes it finite, and what is p?
- (A), and
- (B), and
- (C), and
- (D), and
Answer: (A) Official key: A
- At the denominator is . A finite limit needs the numerator to vanish there too, otherwise the ratio blows up.
- Numerator at : , so .
- Now it is a form, so use L'Hôpital's rule:
- and : option (A).
Trap: option (B) pairs with a plausible-looking , but then the numerator at is , not zero. One millionth away from the ratio is already close to in size: no finite limit. We evaluated all four options on both sides of ; only (A) settles on a finite value, and it is .
L'Hôpital-type limits are behind formulas engineers use without noticing: the sinc function in signal processing, the damped-vibration response as damping tends to the critical value, and the log-mean temperature difference of a heat exchanger when both end differences become equal (it tends to that common difference). Simulation codes special-case these points because evaluating the raw formula there gives .
Taylor expansion at a strict local minimum: what are the gradient term and the second-order term?
- (A) and
- (B) and
- (C) and
- (D) and
Answer: (B) Official key: B
- At an interior local minimum of a differentiable function the gradient is zero (first-order condition), so for every . That rules out (A).
- With the gradient term gone, the expansion reads .
- Strict local minimum means for every small . So the right side must be positive: : option (B). (C) would make flat around , and (D) would make a local maximum.
A subtle point the question gets right: the Hessian is evaluated at , not at . For , the origin is a strict minimum but the Hessian at the origin is the zero matrix. The textbook test (Hessian at positive definite) is sufficient, not necessary; the statement in (B) holds either way. We checked both cases numerically: zero gradient at , and a positive remainder for 500 random small steps each.
This is the logic inside every gradient-based optimiser: design optimisation in ANSYS, Altair OptiStruct or HyperStudy, least-squares fitting of material models to test curves, and the training of machine-learning models. The solver stops when the gradient is (near) zero and uses curvature, the Hessian or an approximation such as BFGS, to tell a minimum from a saddle point or maximum.
Minimum of f(x) = 2x³ − 9x² + 12x on the closed interval [0, 3]
- (A)0
- (B)4
- (C)5
- (D)9
Answer: (A) Official key: A
- Stationary points: , so and .
- : (local maximum), (local minimum).
- On a closed interval the minimum can also sit at an end point, so compare all four values: , , , .
- The smallest is : option (A).
Trap: stopping at the second-derivative test gives the local minimum , option (B). The other two options are also values from the list: the local maximum 5 and the end value 9 (the maximum on ). A grid of 300,001 points on finds the minimum 0 at .
Real design limits are closed intervals: a shaft diameter between 20 and 40 mm, a pressure between two ratings, an operating speed band. Optimisers in tools like HyperStudy or MATLAB's fmincon check the bounds as well as the stationary points for exactly this reason, and in practice the best design often sits on a bound (thinnest allowed wall, lowest allowed mass).
A polynomial has a triple root at s = −σ: which derivatives must vanish there?
- (A) at all the three values of satisfying
- (B), , and at
- (C), , and at
- (D), and at
Answer: (B) Official key: B
- A triple root at means for some polynomial with (exactly triple).
- Differentiate: still has a factor , and still has a factor . All three vanish at : option (B).
- The third derivative is , which equals at the root. So (D) is false for any exact triple root, and (C) skips .
Trap (A): has one real root and two complex roots . A triple root at says nothing about at those complex points. We tested 30 random real polynomials of degree 4 to 6 with a triple root: only the conditions in (B) held every time.
Repeated roots matter in vibration and control. A characteristic polynomial with a double root is a critically damped system (fastest return without overshoot), the target for door closers, gauge needles and many suspension and servo tunings. Repeated poles also change the time response from to and , the same -style extra factor seen in Euler-Cauchy equations.
Double integral of 1/(x + y)² over 3 ≤ x ≤ 4, 1 ≤ y ≤ 2 equals ln(a/24): what is a?
Answer: 25 Official key: 25 to 25
- Inner integral in : .
- Outer integral in : .
- , so .
Check: the integral is 0.040822 (Simpson's rule in both directions), and . A sign slip in the inner integral () gives , which can't be written as with a positive integer : a useful self-check.
Double integrals over a region are how section properties are computed: area, centroid and second moment of area of a beam section. CAD tools (mass properties in SolidWorks or CATIA) and FEA solvers evaluate them numerically, element by element, with Gauss quadrature, which is the same integration done with a few well-chosen points.
Integral of y over the region bounded by x² = 4y, x = 2 and the x axis
The value of is
- (A)
- (B)
- (C)
- (D)
Answer: (A) Official key: A
- Sketch first. The parabola passes through and . The region is under the parabola, above the axis, from to .
- Vertical strips: runs from to , then from to :
- : option (A).
Trap: in the other order, the strip at height runs from the parabola to the line . Writing the limits as to integrates over the region above the parabola instead and gives . The correct order, , agrees with the first method.
is the first moment of area about the axis. Divided by the area ( here) it gives the height of the centroid, . The same integral locates the neutral axis of a beam section, the centre of pressure on a submerged gate, and it is the in the shear stress formula .
Given ∫e^(−x²)dx = √π, what is ∫e^(−a(x+b)²)dx over the whole real line?
If and are positive integers, the value of is ________.
- (A)
- (B)
- (C)
- (D)
Answer: (B) Official key: B
- Substitute , so . The limits stay to .
- : option (B).
The shift only slides the bell curve along the axis; over an infinite range the area is unchanged, so can't appear in the answer (traps C and D). Scaling by makes the curve narrower, so the area must fall as grows, which rules out (A). We integrated numerically for to and to : every value matched .
This integral is the normal distribution: with the area is the normalising constant. It is used in tolerance analysis (statistical stack-ups and Cpk), measurement uncertainty, and in heat transfer: the error function solution for a suddenly heated thick wall, used for quenching and welding estimates, is a Gaussian integral.
Fourier series of x³ on −1 ≤ x < 1: sine terms, cosine terms or both?
- (A)only sine terms
- (B)only cosine terms
- (C)both sine and cosine terms
- (D)only sine terms and a non-zero constant
Answer: (A) Official key: A
- is odd: . On the symmetric interval , its product with an even function is odd and integrates to zero, so every , including the constant , is zero.
- The product with is even, so the survive: , with .
- Only sine terms: option (A).
Trap (D): a non-zero constant would mean a non-zero average over one period, but an odd function's average over is zero. Numerically, to all came out below and to matched the formula above.
Fourier series are how vibration engineers read signals. An FFT of an accelerometer trace splits engine, gearbox or bearing vibration into harmonics, and symmetry shows up directly: a waveform with odd symmetry has only sine content, one with half-wave symmetry only odd harmonics. Order analysis in NVH testing and harmonic analysis in ANSYS both rest on this decomposition.
Square wave between 4 and 0, 1 s at each level: what is the constant Fourier term?
- (A)1
- (B)2
- (C)3
- (D)4
Answer: (B) Official key: B
- The constant term of a Fourier series is the average value of the function over one period.
- Period s: for 1 s and for 1 s. Average : option (B).
Trap (D): with the series written as , the coefficient is , and the constant term is . Typing itself gives 4. Averaging the wave numerically over a period gives 2.
The constant (DC) term is the mean of a periodic signal: the average torque of an engine over a cycle, the mean voltage of a PWM motor drive (duty cycle times supply voltage, here 50% of 4), the mean stress in a fatigue load cycle. Fatigue checks use exactly this split: mean stress (the DC term) plus alternating stress (the harmonics).
Component of a = 5i + 7j + 2k orthogonal to b = 3i − j + 6k, in their plane: what is its magnitude?
Answer: 8.32 Official key: 7.90 to 8.70
- Split into a part along and a part perpendicular to it. Both parts lie in the plane of and .
- , , .
- Parallel part: .
- Perpendicular part, by Pythagoras: , which rounds to 8.32.
Same answer another way: , since the cross product's magnitude is . Trap: the parallel part 2.95 is the projection, the opposite of what is asked.
Splitting a vector into components along and across a direction is everyday mechanics: normal and tangential force on an inclined contact (friction uses the normal part), axial and shear load on a bolt, and the normal and shear stress on a weld throat or a fibre direction in a composite. FEA post-processors do this projection when you ask for stress in a local coordinate system.
Flux of ∇φ, φ = (x² + y² + z²)/2, out of a unit sphere
- (A)
- (B)
- (C)
- (D)0
Answer: (A) Official key: A
- , the position vector.
- Direct route: on the unit sphere (radius 1), so everywhere. The integral is the surface area, .
- Divergence theorem route: , times the volume gives : option (A).
Trap (C): is the sphere's volume, what you get if you forget the divergence is 3, not 1. A direct numerical surface integral over the sphere gives 12.566460 ().
The divergence theorem is the foundation of the finite volume method used by ANSYS Fluent, OpenFOAM and STAR-CCM+: the change of mass, momentum or energy inside each cell is computed from the fluxes through its faces. It is also Gauss's law for heat: the total heat leaving a closed surface equals the heat generated inside it.
Closed surface integral of z dx dy over a sphere of radius R
- (A)0
- (B)
- (C)
- (D)
Answer: (C) Official key: C
- In surface-integral notation, means the flux of the field : is the projection of the surface element.
- Divergence theorem: , so the integral equals the enclosed volume, : option (C).
Trap (A): the integrand is odd, which tempts you to answer 0. But is odd too (negative on the lower half, where is negative), so is positive everywhere and nothing cancels. A numerical surface integral of for gives 4.18899 against .
Flux integrals like this give a body's volume, buoyancy and pressure force from its surface alone: integrating pressure times the normal over a hull's wetted surface gives Archimedes' buoyancy. CAD systems compute volume, mass and centroid of a solid model the same way, by integrating over its boundary surfaces.
Flux of F = 3x i + 5y j + 6z k out of a unit cube centred at (1, 2, 3)
- (A)14
- (B)27
- (C)28
- (D)31
Answer: (A) Official key: A
- Divergence theorem: , a constant.
- Flux : option (A).
The centroid is a distraction: with a constant divergence, only the volume matters. Face by face it is just as quick: the two faces are 1 apart, so their net flux is ; likewise and for and . Trap (D): plugging the centroid into gives , and adding the components gives 31. A face-by-face numerical integral gives 14 with the cube at and 14 at the origin.
This is a finite volume cell in miniature. A CFD solver checks that, for every cell, the net flux through its faces balances what is created or stored inside; a constant divergence means a uniform source, like uniform heat generation in a battery cell or a resistive conductor. Mass imbalance across cells is one of the residuals you watch converge in Fluent or OpenFOAM.
What is the divergence of the curl of a vector field?
- (A)the magnitude of this vector field
- (B)the argument of this vector field
- (C)the magnitude of the curl of this vector field
- (D)zero
Answer: (D) Official key: D
- Write in components and take its divergence:
- The six mixed second derivatives cancel in pairs ( against , and so on) for any field with continuous second derivatives. The result is identically zero: option (D).
Companion identity worth memorising: , the curl of a gradient is zero. We checked the first one numerically on the field at three points: the curl itself is far from zero, its divergence is zero to rounding error.
Because , a magnetic field written as the curl of a vector potential automatically has no divergence, which is how electromagnetic FEA (motor and transformer design in tools like Ansys Maxwell) guarantees physical fields. In fluids, the vorticity (the curl of velocity) is divergence-free for the same reason, a fact used in vortex methods and in checking CFD results.
Directional derivative of f = x² + xy² at (1, 0) along (i + √3 j)/2
Answer: 1.0 Official key: 0.9 to 1.1
- Gradient: . At : .
- The direction already has length , so no normalising is needed.
- , so the answer is 1.0.
Trap: is the largest rate of change, along the gradient itself, not along the given direction (60° away from it, and ). A finite-difference step along gives 1.000000.
Directional derivatives tell you how fast temperature, stress or pressure changes along a particular line: the temperature gradient through a weld along the plate, the stress gradient at a notch used in fatigue (notch sensitivity), or a sensitivity in design optimisation, where the solver needs the change in mass or stress along a proposed design change.
Is the gradient ∇φ perpendicular or parallel to a surface of constant φ?
- (A)always perpendicular to the surface of constant
- (B)always parallel to the surface of constant
- (C)the minimum rate of change of scalar
- (D)always zero
Answer: (A) Official key: A
- Move a small step within a surface of constant . Then doesn't change: .
- So is perpendicular to every direction lying in the surface: it is normal to the surface of constant , option (A).
- (C) is backwards: is the maximum rate of change, reached along . (D) holds only at stationary points.
We checked it on the ellipsoid : at 200 random points, the gradient's dot product with both surface tangent vectors was zero to of their lengths.
Heat flows along , perpendicular to isotherms, which is why heat-flux arrows in a thermal FEA plot cross the temperature contours at right angles. CAD uses the same fact to get the normal of an implicit surface, and contact algorithms in crash and forming simulation use surface normals from gradients to push penetrating nodes back out.