Why the Starting Value Matters
Newton’s method repeatedly replaces a current estimate by the horizontal intercept of the tangent line to the graph. The method can be very effective, but the first tangent line depends on the initial guess. A poor starting value can send the iteration far from the root of interest, even when that root is easy to locate on a graph.
Choosing an initial value is therefore not just a matter of picking a convenient number. First identify which root you want. Then use the graph, known root intervals, and the slope near a possible starting value to judge whether the first tangent step is likely to be useful.
The formula requires that the derivative at the current iterate is nonzero. It does not, by itself, guarantee that the iterates approach a root. This tutorial focuses on how to make an informed initial choice, not on a general convergence guarantee.
Use a Graph to Identify the Root of Interest
A graph gives a useful first picture of where the zeros of \(f\) lie. A zero, or root, is a value \(r\) for which \(f(r)=0\); graphically, the curve meets or touches the horizontal axis at \(x=r\). If there is more than one root, decide which one the calculation is intended to approximate before choosing \(x_0\).
A graph is a guide, not a precise calculation. It can suggest a root interval, show whether the graph is steep or nearly horizontal, and reveal nearby roots or turns. The numerical values of the function at selected endpoints can then confirm whether an interval contains a root.
This result makes a sign-changing interval a useful anchor for choosing a starting value. If the graph or other information indicates that the intended root lies between \(a\) and \(b\), choose an initial value inside that interval and check the first update. A sign change alone does not force Newton’s iterates to stay inside the interval, so monitor where each step goes.
The Tangent-Line View of the First Step
The Newton update is obtained by setting the tangent line at \(x_0\) equal to zero. That tangent line has equation \(y=f(x_0)+f'(x_0)(x-x_0)\). Setting \(y=0\) and solving for \(x\) gives the next estimate.
This explains why the slope matters. The size of the first step is \(\left|f(x_0)/f'(x_0)\right|\). If the slope is small while \(f(x_0)\) is not close to zero, the quotient can be large. The tangent line may meet the axis far from the starting point, possibly outside the root interval or in a region where the graph behaves quite differently.
Conversely, a steep slope can make the first step small, but that alone is not proof that the starting value is good: the tangent may still point away from the intended root. Check the sign and size of the update, and compare its location with the graph and known interval.
Worked Example: A Starting Value Inside a Root Interval
Worked Example: A Starting Value Inside a Root Interval
Use Newton’s method to approximate the positive root of \(f(x)=x^2-2\), starting from the midpoint of the interval \([1,2]\). First verify that the interval identifies a root.
Evaluate the endpoints:
The signs differ, and \(f\) is continuous, so there is at least one root in \((1,2)\). The midpoint is \(x_0=(1+2)/2=3/2\). Since \(f'(x)=2x\), the first update is
Check the new value by substitution:
The second update is therefore
Substituting this iterate back into the function gives
The residual is much smaller than at \(x_1\), and both iterates lie in the interval \((1,2)\). In this example, the midpoint is a reasonable start: it is inside the interval, and the first tangent step moves toward the positive root.
When a Small Slope Makes a Risky Start
The next example shows why being near a root interval is not enough. The starting point can still have a small derivative, creating a large first step. In that situation, inspect the update before assuming the iteration will remain near the target root.
Worked Example: A Small Slope Produces a Large Step
For \(f(x)=x^2-1\), choose the positive root, which lies in \((0,2)\), and start at \(x_0=1/10\). The endpoint values confirm a sign change:
At \(x_0=1/10\), the function value is \(f(x_0)=1/100-1=-99/100\), while the derivative \(f'(x)=2x\) gives \(f'(x_0)=1/5\). Thus
The first step lands at \(101/20=5.05\), well outside the interval \((0,2)\). Substitution verifies this large jump has not brought us close to a root:
For completeness, the next iterate is
Its function value is still large:
The initial derivative \(1/5\) was small compared with the size of \(f(x_0)\), so the tangent-line intercept was far away. This example does not establish that every such iteration must fail; it shows why a candidate start should be assessed by its slope and predicted first step, not only by its distance from a root interval.
Root Intervals Help Select Which Root to Find
If a function has several roots, a starting value can influence which one the iteration approaches. A graph and sign-changing intervals help make the target explicit. They also provide a reference against which to judge whether the first iterate has moved in a useful direction.
Worked Example: Targeting One of Two Roots
Consider \(f(x)=x^2+x-2\). It factors as \(f(x)=(x-1)(x+2)\), so its roots are \(1\) and \(-2\). To target \(1\), use the interval \((0,2)\). Its endpoint values are \(f(0)=-2\) and \(f(2)=4\), confirming a sign change. Choose \(x_0=3/2\).
The derivative is \(f'(x)=2x+1\). At the starting value,
Therefore,
Check the function and derivative at this iterate:
The next update is
Substitution gives
Both iterates remain in \((0,2)\) and move toward the selected root \(1\), rather than the other root \(-2\). The interval makes the target clear; it does not alone guarantee that every Newton iterate stays there.
A Practical Selection Process
Before carrying out several iterations, use the following checks to decide whether a starting value is sensible. These are diagnostic steps, not a theorem that Newton’s method will converge.
Use the problem context or graph to decide which zero is wanted, especially if several roots are present.
Look for endpoint values with opposite signs. For a continuous function, this confirms at least one root in the interval, though not necessarily a unique one.
A point where the derivative is zero cannot be used for a Newton update. A point with a small derivative can produce a large step if the function value is not comparably small.
Compute \(x_1=x_0-f(x_0)/f'(x_0)\). Check whether it stays in a useful region and moves toward the intended root.
If the step leaves the root interval or lands in a region with very different graph behavior, reconsider the start rather than treating the next iterate as automatically reliable.
Common Mistake: Treating a Sign Change as a Guarantee
A sign change on \([a,b]\) guarantees that a continuous function has a root between the endpoints. It does not guarantee that Newton’s method, started anywhere in that interval, will stay inside it. For example, the calculation for \(x^2-1\) above started inside \((0,2)\) but produced \(x_1=5.05\).
The mistaken reasoning would be: “There is a root in \((0,2)\), and \(x_0\) is in \((0,2)\), so all Newton iterates will remain in \((0,2)\).” The example disproves that claim. The correction is to separate two facts: the sign change establishes root existence, while the Newton update determines the next point. Calculate that update and check it directly.
Check Your Understanding
Use the graph, interval, and tangent-step ideas to answer these questions.
- For a continuous function, what does a sign change between the endpoints of an interval guarantee, and what does it not guarantee?
- Why can a small value of \(\lvert f'(x_0)\rvert\) lead to a large first Newton step?
- If a function has two roots, what information can help you choose a starting value aimed at one root rather than the other?
- For \(f(x)=x^2-2\), verify that \(f(1)<0\) and \(f(2)>0\). What does this establish about roots in \((1,2)\)?
- Why should you check whether the first iterate remains in a useful region instead of assuming that it does?
