Calculus: Difference Quotients, Instantaneous Velocity, and Formal Derivatives
Tangent Lines and the Difference Quotient
The slope of a tangent line to a curve y=f(x) at a point x0 is defined as the limit of the difference quotient as x approaches x0: mtangent=limx→x0x−x0f(x)−f(x0)
Once the slope mtangent is determined, the equation of the tangent line is found using the point-slope form: y=f(x0)+mtangent(x−x0) or y−y0=mtangent(x−x0)
Derivation Example: Finding the tangent line equation for f(x)=6x2+3 at x=−1:
Curve characteristics: A parabola crossing the y-axis at (0,3).
Coordinates at x0=−1: f(−1)=6(−1)2+3=9 Target point: (−1,9).
Slope calculation via difference quotient: mtangent=limx→−1x−(−1)(6x2+3)−9mtangent=limx→−1x+16x2−6mtangent=limx→−1x+16(x+1)(x−1)mtangent=limx→−16(x−1)mtangent=6(−1−1)=−12
Tangent line equation derivation: y=9+(−12)(x−(−1))y=9−12(x+1)y=−12x−3
Geometric property: The tangent line crosses the y-axis at (0,−3).
Average Speed, Instantaneous Speed, and Instantaneous Velocity
Definitions:
Average Speed: Total distance traveled divided by total elapsed time: Average Speed=ΔtΔd
Instantaneous Speed: Limit of average speed as the time interval Δt approaches zero: Instantaneous Speed=limΔt→0(Average Speed)
Instantaneous Velocity: Vector quantity representing rate of position change with direction. Motion to the right or upward is positive; motion to the left or downward is negative.
Relationship between speed and velocity: Instantaneous Speed=∣Instantaneous Velocity∣
Geometric and Algebraic Formulation:
Initial position at time t1: x=f(t1).
Position after time increment h (where h=Δt): x=f(t1+h).
Change in distance: Δx=f(t1+h)−f(t1).
Change in time: Δt=(t1+h)−t1=h.
Limit formulation for instantaneous velocity at t1: v(t1)=limh→0hf(t1+h)−f(t1)
Instantaneous Velocity Examples and Kinematics
Kinematics Example A: Position function s(t)=78.4−4.9t2 (meters, seconds):
Initial position at t=0s: s(0)=78.4−4.9(0)2=78.4m
Note on Domain Baseline and Negative Time: The time t=0 represents an arbitrary reference baseline defined by the problem. Negative time (t<0) represents motion occurring prior to the reference point (such as an object thrown from ground level reaching its peak at t=0).
Ground impact time (s(t)=0): 78.4−4.9t2=0⟹4.9t2=78.4⟹t2=16⟹t=4s
Average speed over interval [0,4]: Average Speed=ΔtΔd=4s78.4m=19.6m/s
Average velocity over interval [0,4]: Average Velocity=−19.6m/s
Approximating Instantaneous Velocity at t=3s using interval [2.9,3.1]: s(3.1)=78.4−4.9(3.1)2=31.311ms(2.9)=78.4−4.9(2.9)2=37.191mApproximate Velocity=3.1−2.9s(3.1)−s(2.9)=0.231.311−37.191=0.2−5.88=−29.4m/s
Exact Instantaneous Velocity at t=3s using limit definition: v(3)=limh→0hs(3+h)−s(3)v(3)=limh→0h(78.4−4.9(3+h)2)−(78.4−4.9(3)2)v(3)=limh→0h−4.9(9+6h+h2)−(−44.1)v(3)=limh→0h−29.4h−4.9h2v(3)=limh→0−4.9(6+h)=−4.9×6=−29.4m/s
Approximation vs. Exact Values: Approximating instantaneous rate of change via finite intervals (e.g., Δt=0.2) approaches the exact value as interval size shrinks (e.g., [2.95,3.05]).
Kinematics Example B: Position function s(t)=112−16t2:
General velocity derivation v(t): v(t)=limh→0hs(t+h)−s(t)v(t)=limh→0h(112−16(t+h)2)−(112−16t2)v(t)=limh→0h112−16(t2+2th+h2)−112+16t2v(t)=limh→0h−32th−16h2v(t)=limh→0(−32t−16h)=−32t
Velocity evaluation at specific times:
At t=2s: v(2)=−32(2)=−64m/s
At t=4s: v(4)=−32(4)=−128m/s
Peak height identification: Parabolic vertex occurs at t=0s, giving maximum height s(0)=112m.
Ground impact velocity:
Time of impact: s(t)=0⟹112−16t2=0⟹t2=7⟹t=7s.
Velocity at impact: v(7)=−327m/s.
Limit setup at impact point: v(7)=limh→0hs(7+h)−s(7)=limh→0hs(7+h)−0
General Definitions of Rates of Change and Derivatives
Average Rate of Change: Generalization of average speed for any function f(x) over an interval [x1,x2]: Average Rate of Change=x2−x1f(x2)−f(x1)
Instantaneous Rate of Change: Generalization of instantaneous velocity at x1: Instantaneous Rate of Change=limx→x1x−x1f(x)−f(x1)=limh→0hf(x1+h)−f(x1)
Derivative Definition: Given a function f, the derivative f′(x) is defined as: f′(x)=limh→0hf(x+h)−f(x) where h=Δx
Notational Equivalences: f′(x)=dxd[f(x)]=dxdy
Differentiability Criteria:
Derivative at a specific point x=c: f′(c)=limh→0hf(c+h)−f(c)
If f′(c) exists, f is differentiable at c.
If f′(c) exists for all points in an open interval, f is differentiable on that interval.
Derivative Calculations and Tangent Line Equations
General derivative f′(x) calculation: f′(x)=limh→0h[2(x+h)2−(x+h)+3]−[2x2−x+3]f′(x)=limh→0h[2(x2+2xh+h2)−x−h+3]−[2x2−x+3]f′(x)=limh→0h2x2+4xh+2h2−x−h+3−2x2+x−3f′(x)=limh→0h4xh+2h2−hf′(x)=limh→0(4x+2h−1)=4x−1
Specific derivative at x=4: f′(4)=4(4)−1=15
Advanced Derivative Derivations via Conjugate Rationalization
Example: Evaluating dxdy at x=9 for y=4x+1:
Function values: f(9+h)=49+h+1f(9)=49+1=4(3)+1=13
Evaluating limit as h→0: 49+0+1216=4(3)+1216=12+1216=2416=32
Questions and Discussion
Tangent Line Final Equation Format:
Prompt: Should line equations be left in point-slope form or simplified further?
Response: All specific numeric values (x0, f(x0), and mtangent) must be explicitly substituted into point-slope form y=f(x0)+mtangent(x−x0). Do not leave placeholders like x0 uninstantiated. When arithmetic values are clean, simplify down to slope-intercept form (y=mx+b).
Equivalence of Derivative Limit Definitions:
Prompt: Can derivatives be computed using limx→x1x−x1f(x)−f(x1) instead of the h→0 formula?
Response: Both formulations are mathematically equivalent. The definition limx→x1x−x1f(x)−f(x1) works efficiently when computing a derivative value at a specific point x1. The definition $ rối \lim_{h \to 0} \frac{f(x + h) - f(x)}{h}issuperiorwhenconstructinggeneralfunctionalderivativeexpressionsf'(x)$$.
Limit Definitions vs. Shortcut Rules:
Prompt: Is writing out full limit definitions required when finding derivatives?
Response: While shortcut operational rules (e.g., Power Rule, Product Rule, Quotient Rule) exist to evaluate derivatives quickly, mastering the full limit derivation is essential to understand the underlying mechanics of instantaneous rates of change. Formal assessment items will explicitly require full limit evaluations before transitioning to computational rules.