You're staring at a graph. Because of that, the function climbs steadily from left to right. So naturally, you draw rectangles under the curve using the left endpoint of each subinterval. The tops of those rectangles sit below the curve. Every single one That's the part that actually makes a difference. Worth knowing..
Sound familiar? That's the left Riemann sum doing its thing — and it's giving you an underestimate.
But here's the catch: that's not always true. In practice, flip the function upside down — make it decrease — and suddenly those same left-endpoint rectangles tower above the curve. Now you've got an overestimate.
So when exactly is the left Riemann sum an underestimate? Now, ** But there's more to it than a one-line rule. In practice, the short answer: **when the function is increasing on the interval. Let's unpack it.
What Is a Left Riemann Sum
A Riemann sum approximates the area under a curve by chopping the interval into subintervals and building rectangles on each one. The left Riemann sum uses the function value at the left endpoint of each subinterval to determine the rectangle's height.
So for an interval [a, b] divided into n equal subintervals of width Δx = (b - a)/n, the left Riemann sum is:
Lₙ = Σ f(xᵢ) Δx, where xᵢ = a + iΔx for i = 0, 1, ..., n-1
That's the formula. But formulas don't build intuition. Pictures do Practical, not theoretical..
Imagine f(x) = x² on [0, 2] with four subintervals. Empty space under the curve. The left endpoints are 0, 0.Which means 25, 1, 2. 25. The rectangles sit at heights 0, 0.Still, the rest of the rectangle? Each rectangle's top-right corner touches the curve. Because of that, 5. 5, 1, 1.That missing area is why the sum falls short.
The Visual Test
Here's a quick mental check: draw the function. This leads to draw the left-endpoint rectangles. Ask yourself — does the curve sit above the rectangle tops, or below?
- Curve above rectangles → left sum underestimates
- Curve below rectangles → left sum overestimates
- Curve matches rectangles (flat line) → left sum is exact
That's the whole game.
Why It Matters / Why People Care
You might wonder — why does this distinction even matter? Isn't the integral the "real" answer anyway?
Sure. But in practice, you're not always integrating analytically. Sometimes you're:
- Writing numerical integration code and need to know which direction your error goes
- Estimating a definite integral by hand on an exam with no calculator
- Building a Riemann sum approximation in a spreadsheet or simulation
- Trying to bound an integral without computing it exactly
Knowing whether your approximation is high or low lets you bracket the true value. Also, if left sum underestimates and right sum overestimates (for increasing functions), the actual integral lives between them. That's powerful.
It also shows up in AP Calculus, college calc courses, and technical interviews. That said, the question "Is this an overestimate or underestimate? " is a classic — and it's testing whether you understand the geometry, not just the algebra.
How It Works: The Monotonicity Rule
The direction of the error depends entirely on whether the function is increasing or decreasing. Let's break it down.
Increasing Functions → Left Sum Underestimates
If f is increasing on [a, b], then for any subinterval [xᵢ, xᵢ₊₁]:
f(xᵢ) ≤ f(x) for all x in [xᵢ, xᵢ₊₁]
The left endpoint gives the minimum value on that subinterval. So the rectangle fits entirely under the curve. Still, the rectangle height is the smallest the function gets. Area of rectangle < area under curve on that subinterval Surprisingly effective..
Sum over all subintervals → total left sum < true integral.
Example: f(x) = √x on [0, 4]. Increasing. Left sum with n=4: endpoints 0, 1, 2, 3. Heights 0, 1, √2, √3. Each rectangle sits under the curve. Total ≈ 4.15. True integral = 16/3 ≈ 5.33. Underestimate confirmed The details matter here. Surprisingly effective..
Decreasing Functions → Left Sum Overestimates
Flip it. If f is decreasing on [a, b], then on each subinterval:
f(xᵢ) ≥ f(x) for all x in [xᵢ, xᵢ₊₁]
The left endpoint gives the maximum value. The rectangle towers over the curve. Area of rectangle > area under curve Most people skip this — try not to..
Example: f(x) = 1/x on [1, 4]. Decreasing. Left sum with n=3: endpoints 1, 2, 3. Heights 1, 1/2, 1/3. Rectangles stick out above the curve. Left sum ≈ 1.83. True integral = ln(4) ≈ 1.39. Overestimate.
Constant Functions → Left Sum Is Exact
If f(x) = c (constant), then f(xᵢ) = c everywhere. Consider this: rectangles match the curve perfectly. Left sum = right sum = midpoint sum = true integral = c(b - a).
This is the trivial case — but it's worth remembering. Monotonicity is the key. Constant functions are both increasing and decreasing (non-strictly), and the approximation is exact It's one of those things that adds up. Still holds up..
What About Non-Monotonic Functions?
Here's where it gets messy. Even so, if f increases on some subintervals and decreases on others, the left sum might overestimate on some parts and underestimate on others. The net error could go either way — or cancel out almost perfectly Still holds up..
Example: f(x) = sin(x) on [0, 2π]. Increases on [0, π/2], decreases on [π/2, 3π/2], increases on [3π/2, 2π]. Left sum with n=4? Endpoints 0, π/2, π, 3π/2. Heights 0, 1, 0, -1. The positive and negative errors don't cleanly cancel. You can't predict the direction without computing.
Bottom line: The clean "left sum underestimates for increasing functions" rule only applies when the function is monotonic (entirely increasing or entirely decreasing) on the whole interval.
Common Mistakes / What Most People Get Wrong
Mistake 1: Assuming "Left = Underestimate" Always
This is the big one. In real terms, students memorize "left Riemann sum underestimates" without the condition. Then they apply it to a decreasing function and get the wrong error direction.
The rule isn't about left vs. Worth adding: right. It's about where the function sits relative to the rectangle.
Mistake 2: Confusing Concavity with Monotonicity
Concavity affects midpoint and trapezoidal sums. For left and right sums, only monotonicity matters. A function can be increasing and concave down (like √x) —
left sums still underestimate because the function's increasing nature dominates, not its concavity. In real terms, g. Conversely, a decreasing concave-up function (e., e<sup>−*x_2_</sup> on [0,1]) will have left sums overestimate, regardless of its curvature The details matter here. Which is the point..
Final Conclusion
The left Riemann sum’s accuracy hinges on the function’s monotonicity, not concavity or oscillation. For strictly increasing functions, left sums underestimate the integral; for strictly decreasing, they overestimate. Constant functions yield exact results. Non-monotonic functions defy this pattern, as their varying slopes create mixed errors. This principle is foundational for understanding Riemann sums’ behavior but must be applied cautiously—always verify the function’s behavior across the entire interval before drawing conclusions.
Building on that observation, the behavior of the left Riemann sum can be formalized in terms of the variation of the function over the subintervals. If a function possesses a bounded total variation on ([a,b]), then the error incurred by a left sum with mesh size (|P|) is bounded by the product of the mesh size and the total variation:
[ \bigl|L(f,P)-\int_a^b f(x),dx\bigr|\le V_a^b(f),|P|, ]
where (V_a^b(f)) denotes the total variation of (f) on ([a,b]). In practice, this inequality makes explicit what intuition already suggests: the more the function wiggles, the larger the possible discrepancy between the left‑endpoint approximation and the true integral, regardless of whether any single subinterval is increasing or decreasing. Worth calling out: for a piecewise monotone function the error can be decomposed into contributions from each monotonic piece, allowing a precise prediction of the sign of the error on each piece while still acknowledging that the net error may be positive, negative, or zero depending on the relative widths of those pieces It's one of those things that adds up..
A practical corollary of this viewpoint is that, when designing numerical integration schemes, one can deliberately choose a partition that aligns with the intervals of monotonicity of (f). If (f) is increasing on ([a,c]) and decreasing on ([c,b]), a partition that places a node at (c) will isolate the two behaviours, and the left sum on ([a,c]) will underestimate while the left sum on ([c,b]) will overestimate. The overall error is then the algebraic sum of these two contributions. By refining the partition on the subinterval where the error dominates—typically the one with larger variation—one can systematically reduce the total error to any prescribed tolerance The details matter here..
The same principle extends beyond left sums to right, midpoint, and trapezoidal rules. For a monotone function, the ordering of the four classical Riemann sums is fixed: the left sum provides a lower bound, the right sum an upper bound, and the trapezoidal and midpoint sums lie between them. When monotonicity is lost, the ordering can break down, but the variation‑based error bound still guarantees that the deviation of any of these sums from the integral is proportional to the mesh size and the total variation. This means adaptive quadrature algorithms often monitor the local variation of the integrand and insert nodes where the variation is high, thereby restoring the predictable error structure that monotonicity supplies It's one of those things that adds up..
Worth pausing on this one.
Boiling it down, the left Riemann sum’s tendency to underestimate or overestimate is not an intrinsic property of the “left” endpoint alone; it is a manifestation of the function’s monotonic behaviour across the entire integration domain. But by recognizing that monotonicity governs the direction of the error, that total variation quantifies the magnitude of that error, and that strategic partitioning can isolate and control each source of deviation, we obtain a strong framework for both analyzing and applying Riemann sums in practice. This understanding not only clarifies why simple rules of thumb work for elementary functions but also equips us to handle more nuanced integrands with confidence, ensuring that numerical approximations remain both accurate and interpretable.
