Solve separable and exponential models

Apply initial conditions, preserve equilibrium solutions and interpret domains

What to practice

  • Apply initial conditions, preserve equilibrium solutions and interpret domains

Predict before you read. The function y=11−xy=\frac{1}{1-x} solves dydx=y2\frac{dy}{dx}=y^{2} with y(0)=1y(0)=1. The formula gives y=−1y=-1 at x=2x=2. Is the point (2,−1)(2,-1) on this particular solution? Write yes or no. Example 18.3 settles it.

Chapter 17 read solutions from equations and slope fields. This chapter finds solution formulas by separation of variables, states the interval on which each particular solution lives, and applies the results to exponential growth and decay.

When separation applies

Some differential equations can be solved by separation of variables, and antidifferentiation then produces a general solution (course framework). The method needs the right side to factor as g(x) h(y)g(x)\,h(y), a function of xx times a function of yy:

  • dydx=x(y−2)\frac{dy}{dx}=x(y-2) is separable, with g(x)=xg(x)=x and h(y)=y−2h(y)=y-2.
  • dydx=ex−y\frac{dy}{dx}=e^{x-y} is separable, because ex−y=exe−ye^{x-y}=e^{x}e^{-y}.
  • dydx=x+y\frac{dy}{dx}=x+y is not separable: a sum does not factor into an xx part times a yy part.

Figure 18.1 lays out the procedure. Step 2 is the one most often skipped.

Flowchart with six steps in a column. Step 1 asks whether the right side can be written as g of x times h of y; a "no" arrow leads to a dashed box saying the equation is not separable, so use the slope field, a given solution or a tangent line. A "yes" arrow leads to step 2: find every y-value with h of y equal to 0, each of which gives a constant solution. A side note explains that dividing by h of y assumes it is not 0, so those constant solutions never appear later. Step 3: where h of y is not 0, divide and integrate both sides, the integral of 1 over h of y with respect to y equals the integral of g of x with respect to x, with one constant C. Step 4: substitute the initial point to find C. Step 5: solve for y and keep the branch through the initial point. Step 6: state the largest open interval containing the initial x-value on which the formula is defined, differentiable and satisfies the equation.

Figure 18.1. Six steps for solving dydx=g(x) h(y)\frac{dy}{dx}=g(x)\,h(y).

Example 18.1: a general solution and a particular solution

Solve dydx=x(y−2)\frac{dy}{dx}=x(y-2) with y(0)=5y(0)=5.

  1. Constant solutions. h(y)=y−2h(y)=y-2 is 0 at y=2y=2, so y=2y=2 is a solution.
  2. Separate and integrate. For y≠2y\neq 2: ∫1y−2 dy=∫x dx\int\frac{1}{y-2}\,dy=\int x\,dx, so ln⁡∣y−2∣=x22+C\ln|y-2|=\frac{x^{2}}{2}+C.
  3. Find the constant now. At (0,5)(0,5): ln⁡3=0+C\ln 3=0+C, so C=ln⁡3C=\ln 3.
  4. Solve for yy. ∣y−2∣=3ex2/2|y-2|=3e^{x^{2}/2}. At the initial point y−2=3>0y-2=3>0, so y=2+3ex2/2y=2+3e^{x^{2}/2}.
  5. Interval. The formula is defined and differentiable for every real xx.

Check: y′=3xex2/2y'=3xe^{x^{2}/2} and x(y−2)=3xex2/2x(y-2)=3xe^{x^{2}/2}. The general solution is y=2+Aex2/2y=2+Ae^{x^{2}/2} for any constant AA; A=0A=0 restores the constant solution that dividing by y−2y-2 set aside.

Why find CC before solving for yy? With the logarithm still in place, one substitution gives CC, and the sign of y−2y-2 at the initial point removes the absolute value. Solving first creates a ±\pm that must be sorted out later.

Particular solutions live on intervals

A general solution may describe infinitely many solutions. The course framework adds that only one particular solution passes through a given point. That holds for every equation in this chapter's examples, whose right sides are differentiable near each initial point, but it can fail where the right side is not differentiable: dhdt=−kh\frac{dh}{dt}=-k\sqrt{h} from Chapter 17 has two solutions through a point where h=0h=0, the constant h=0h=0 and a tank that empties at that moment. Solutions may also be subject to domain restrictions (course framework). A particular solution is a differentiable function on one open interval that contains the initial input.

To find that interval, list every input that the formula or the equation cannot accept: a zero denominator, the even root of a negative number, the logarithm of a number that is not positive, or a point where dydx\frac{dy}{dx} is undefined. Remove those inputs and keep the single open interval that contains the initial input. Any other part of the formula's domain belongs to other solutions, even where the formula still satisfies the equation.

Example 18.2: choose the branch

Solve dydx=3x22y\frac{dy}{dx}=\frac{3x^{2}}{2y} with y(1)=−2y(1)=-2.

Separate: ∫2y dy=∫3x2 dx\int 2y\,dy=\int 3x^{2}\,dx, so y2=x3+Cy^{2}=x^{3}+C. At (1,−2)(1,-2): 4=1+C4=1+C, so C=3C=3 and y2=x3+3y^{2}=x^{3}+3. Two branches satisfy this, y=x3+3y=\sqrt{x^{3}+3} and y=−x3+3y=-\sqrt{x^{3}+3}. The initial value −2-2 is negative, so y=−x3+3y=-\sqrt{x^{3}+3}. Check: y′=−3x22x3+3y'=-\frac{3x^{2}}{2\sqrt{x^{3}+3}}, which equals 3x22y\frac{3x^{2}}{2y} because y=−x3+3y=-\sqrt{x^{3}+3}.

Interval: the equation divides by yy, so yy cannot be 0, and the root requires x3+3>0x^{3}+3>0. The solution lives on x>−33x>-\sqrt[3]{3}, roughly x>−1.442x>-1.442. At x=−33x=-\sqrt[3]{3} the formula gives y=0y=0, where dydx\frac{dy}{dx} is undefined.

Example 18.3: a solution that stops

Solve dydx=y2\frac{dy}{dx}=y^{2} with y(0)=1y(0)=1.

The constant solution is y=0y=0. For y≠0y\neq 0: ∫y−2 dy=∫dx\int y^{-2}\,dy=\int dx, so −1y=x+C-\frac{1}{y}=x+C. At (0,1)(0,1), C=−1C=-1, so y=11−xy=\frac{1}{1-x}.

Graph for x from negative 3 to 4. A solid curve passes through the marked point (0, 1), rising slowly from near 0 on the left and then increasing without bound as x approaches 1 from the left. A dotted vertical line marks x equals 1. To the right of that line a lighter dashed curve rises from far below toward the x-axis, staying below it.

Figure 18.2. The formula y=11−xy=\frac{1}{1-x} has two branches; the particular solution is only the branch through (0,1)(0,1).

The formula is undefined at x=1x=1, and yy increases without bound as x→1−x\to 1^{-}. The particular solution is the solid branch on x<1x<1. The dashed branch also satisfies y′=y2y'=y^{2}, but it is a different solution, the one through (2,−1)(2,-1). So the answer to the opening question is no: the formula's value at x=2x=2 does not belong to this solution.

Common error. Reporting the domain as "all x≠1x\neq 1." A solution must be differentiable on one interval through the initial point, so the answer is x<1x<1.

Example 18.4: when the antiderivative has no formula

Solve dydx=1+x3\frac{dy}{dx}=\sqrt{1+x^{3}} with y(2)=5y(2)=5, and find y(3)y(3).

No elementary antiderivative exists, so write the solution as an accumulation function:

y(x)=5+∫2x1+t3 dty(x)=5+\int_{2}^{x}\sqrt{1+t^{3}}\,dt

By the Fundamental Theorem of Calculus, y′(x)=1+x3y'(x)=\sqrt{1+x^{3}}, and y(2)=5+0=5y(2)=5+0=5. With a calculator, y(3)=5+∫231+t3 dt≈5+4.100=9.100y(3)=5+\int_{2}^{3}\sqrt{1+t^{3}}\,dt\approx 5+4.100=9.100. The representation changed from formula to integral, and the solution is still exact.

Exponential models

"The rate of change of a quantity is proportional to the size of the quantity" translates to dydt=ky\frac{dy}{dt}=ky. Separating gives ln⁡∣y∣=kt+C\ln|y|=kt+C, and with y(0)=y0y(0)=y_{0} the solution is y=y0ekty=y_{0}e^{kt}. When k>0k>0 the quantity grows; when k<0k<0 it decays. The same model covers motion along a line, such as a velocity that satisfies dvdt=−0.5v\frac{dv}{dt}=-0.5v.

Over equal time steps an exponential quantity is multiplied by the same factor ek Δte^{k\,\Delta t}. A table with a constant ratio of successive values fits this model; a constant difference fits a linear one. The doubling time is ln⁡2k\frac{\ln 2}{k} for growth, and the half-life is ln⁡2∣k∣\frac{\ln 2}{|k|} for decay.

Common error. Reading "grows 5% per year" as k=0.05k=0.05. That phrase gives a yearly factor of 1.05, so ek=1.05e^{k}=1.05 and k=ln⁡1.05≈0.04879k=\ln 1.05\approx 0.04879. The equation dydt=0.05y\frac{dy}{dt}=0.05y describes a different model, one whose yearly factor is e0.05≈1.05127e^{0.05}\approx 1.05127.

Example 18.5: half-life from a description

A dose of 80 mg of a medication leaves the body at a rate proportional to the amount present and halves every 5 hours (constructed practice data). Find the amount after 12 hours and the rate at 10 hours.

From 40=80e5k40=80e^{5k}, k=−ln⁡25≈−0.138629k=-\frac{\ln 2}{5}\approx -0.138629. So N(t)=80ektN(t)=80e^{kt}, which also equals 80(12)t/580(\frac{1}{2})^{t/5}. After two half-lives N(10)=20N(10)=20 mg, and N(12)≈15.157N(12)\approx 15.157 mg. At t=10t=10, dNdt=kN=−0.138629(20)≈−2.773\frac{dN}{dt}=kN=-0.138629(20)\approx -2.773, so the amount is decreasing at about 2.773 mg per hour.

A linear model that loses 8 mg per hour matches the first 5 hours but predicts 0 mg at 10 hours and −16-16 mg at 12 hours, which is impossible. The proportional model removes the same fraction, not the same amount, in each 5-hour period.

Example 18.6: test a table for an exponential model

A culture is counted every 2 hours (constructed practice data).

tt (hours)0246
yy (thousands of cells)500600720864

The differences are 100, 120 and 144, so a linear model does not fit. The ratios are 600500=720600=864720=1.2\frac{600}{500}=\frac{720}{600}=\frac{864}{720}=1.2, the same over each 2-hour step, so y=500ekty=500e^{kt} fits with e2k=1.2e^{2k}=1.2, which gives k=ln⁡1.22≈0.091161k=\frac{\ln 1.2}{2}\approx 0.091161 per hour. The model predicts y(8)=500(1.2)4=1036.8y(8)=500(1.2)^{4}=1036.8 thousand cells. It satisfies dydt=ky\frac{dy}{dt}=ky, so at t=4t=4 the culture grows at about 0.091161(720)≈65.6360.091161(720)\approx 65.636 thousand cells per hour. The table, a list of numbers, became a differential equation and a formula.

Example 18.7: motion along a line

A boat coasts with its engine off. Its velocity satisfies dvdt=−0.5v\frac{dv}{dt}=-0.5v, with vv in meters per second and tt in seconds, and v(0)=12v(0)=12 (constructed practice data). The model gives v(t)=12e−0.5tv(t)=12e^{-0.5t}: the velocity halves every ln⁡20.5≈1.386\frac{\ln 2}{0.5}\approx 1.386 seconds and never reaches 0. In the first 4 seconds the boat travels ∫0412e−0.5t dt=24(1−e−2)≈20.752\int_{0}^{4}12e^{-0.5t}\,dt=24(1-e^{-2})\approx 20.752 meters. After TT seconds it has traveled 24(1−e−0.5T)24(1-e^{-0.5T}) meters, which is always less than 24.

Example 18.8: a shifted exponential

Chapter 17, Example 17.6, read the cooling model dTdt=−0.08(T−20)\frac{dT}{dt}=-0.08(T-20) from its slope field. Solve it with T(0)=80T(0)=80.

The constant solution is T=20T=20. For T≠20T\neq 20: ∫1T−20 dT=∫−0.08 dt\int\frac{1}{T-20}\,dT=\int -0.08\,dt, so ln⁡∣T−20∣=−0.08t+C\ln|T-20|=-0.08t+C. At t=0t=0, T−20=60>0T-20=60>0, so T=20+60e−0.08tT=20+60e^{-0.08t}. Then T(10)≈46.960T(10)\approx 46.960 degrees, and T=30T=30 when 60e−0.08t=1060e^{-0.08t}=10, at t=ln⁡60.08≈22.397t=\frac{\ln 6}{0.08}\approx 22.397 minutes.

The formula confirms what Chapter 17 read from the field: the temperature decreases, the graph is concave up, and the curve approaches 20 without crossing it. With any starting temperature T0T_{0}, the same steps give T=20+(T0−20)e−0.08tT=20+(T_{0}-20)e^{-0.08t}. Figure 18.3 shows three cases.

Graph for t from 0 to 40 minutes and T from 0 to 80 degrees. A solid curve starts at (0, 80) and falls toward 20, passing a marked point at (22.40, 30). A dash-dot horizontal line starts at (0, 20) and stays at 20. A dashed curve starts at (0, 5) and rises toward 20. Both curves approach the line T equals 20 without crossing it.

Figure 18.3. Particular solutions of dTdt=−0.08(T−20)\frac{dT}{dt}=-0.08(T-20) for starting temperatures 80, 20 and 5 degrees Celsius, shown for 0≤t≤400\leq t\leq 40 minutes.

Diagnose incorrect solutions

Example 18.9: the constant in the wrong place

A student solves dydx=2xy\frac{dy}{dx}=2xy with y(0)=3y(0)=3:

"∫1y dy=∫2x dx\int\frac{1}{y}\,dy=\int 2x\,dx, so ln⁡y=x2\ln y=x^{2} and y=ex2+Cy=e^{x^{2}}+C. Then y(0)=3y(0)=3 gives C=2C=2, so y=ex2+2y=e^{x^{2}}+2."

Check by substitution: y′=2xex2y'=2xe^{x^{2}}, but 2xy=2xex2+4x2xy=2xe^{x^{2}}+4x. The two differ by 4x4x, so the answer fails. The constant belongs with the antiderivatives, before exponentiating: ln⁡∣y∣=x2+C\ln|y|=x^{2}+C, so ∣y∣=eCex2|y|=e^{C}e^{x^{2}}, and y=Aex2y=Ae^{x^{2}}. Then y(0)=3y(0)=3 gives y=3ex2y=3e^{x^{2}}, and y′=6xex2=2x(3ex2)y'=6xe^{x^{2}}=2x(3e^{x^{2}}). The division by yy also set aside the constant solution y=0y=0, which is A=0A=0.

Common error. Adding CC after exponentiating. An additive constant on the logarithm becomes a multiplying constant on yy.

Try it

Try it 18.1. Solve dydx=eycos⁡x\frac{dy}{dx}=e^{y}\cos x with y(0)=0y(0)=0, and state the interval of the solution.

Try it 18.2. A population grows at a rate proportional to its size and doubles every 9 years. How long does it take to triple?

Try it 18.3. Solve dPdt=0.5(P−40)\frac{dP}{dt}=0.5(P-40) with P(0)=50P(0)=50 (Try it 17.3 described its slope field), and find when P=100P=100.

Try it 18.4. For dydx=(y−1)cos⁡x\frac{dy}{dx}=(y-1)\cos x, find the particular solutions with y(0)=3y(0)=3 and with y(0)=1y(0)=1.

Check your work

18.1. ∫e−y dy=∫cos⁡x dx\int e^{-y}\,dy=\int\cos x\,dx gives −e−y=sin⁡x+C-e^{-y}=\sin x+C, and (0,0)(0,0) gives C=−1C=-1. So e−y=1−sin⁡xe^{-y}=1-\sin x and y=−ln⁡(1−sin⁡x)y=-\ln(1-\sin x). The logarithm needs sin⁡x<1\sin x<1, so the solution lives on −3π2<x<π2-\frac{3\pi}{2}<x<\frac{\pi}{2}.

18.2. k=ln⁡29k=\frac{\ln 2}{9}, and ekt=3e^{kt}=3 gives t=9ln⁡3ln⁡2≈14.265t=\frac{9\ln 3}{\ln 2}\approx 14.265 years.

18.3. P=40+10e0.5tP=40+10e^{0.5t}. Then 10e0.5t=6010e^{0.5t}=60 gives t=2ln⁡6≈3.584t=2\ln 6\approx 3.584.

18.4. For y(0)=3y(0)=3: ln⁡∣y−1∣=sin⁡x+C\ln|y-1|=\sin x+C with C=ln⁡2C=\ln 2, so y=1+2esin⁡xy=1+2e^{\sin x} for all xx. For y(0)=1y(0)=1 the solution is the constant y=1y=1, which the division by y−1y-1 would have lost.

Practice

Chapter 18 practice is items APCAL-C18-M01 to APCAL-C18-M09 and the written tasks APCAL-C18-F01 and APCAL-C18-F02. They are printed after this chapter's companion panel as Questions 233–241 and Written tasks 25 and 26. Questions 239, 240 and 241 and Written task 26 require a graphing calculator in radian mode; the others and Written task 25 do not. Answers, full reasoning, the error behind each wrong choice, a Repair and retry line and the written tasks' scoring rubrics are in Answers, explanations and written responses at the back of the book, under Chapter 18: answers and explanations.

Before you move on

Close the book and redraw Figure 18.1 from memory, including the reason step 2 comes before the division. Then solve Example 18.1 again with y(0)=0y(0)=0 and state its interval. In two or three days, without the chapter, rework Example 18.5 for a 120 mg dose with a 6-hour half-life, and explain in one sentence why a linear model fails.

Use this chapter companion

Review the lesson, attempt the original practice, then check the explanation. Your printed study guide remains complete without the website.

Independence

Skill by Skill Prep is independent. SAT® and AP® are trademarks registered by the College Board, which is not affiliated with, and does not endorse, this website or book. PSAT/NMSQT® is a registered trademark of the College Board and the National Merit Scholarship Corporation, which are not affiliated with, and do not endorse, this site. Bluebook™ is a trademark of the College Board. ACT® is a trademark of ACT Education Corp., which is not affiliated with and does not endorse this website or book. Desmos is a trademark of Desmos Studio PBC. Practice results describe these questions, not official scores or admission predictions.

Last updated .