8 topics. Each one opens with its key idea, then the detail you need to use it.
1. Verifying solutions: general and particular
Key idea: Substitute into both sides and simplify; a solution must match for every $x$, and an initial condition then fixes $C$.
To verify a solution, compute the derivatives the equation uses. Then substitute the function into the left side and the right side separately. Take $y = Ce^{-x} + 2$ and $\frac{dy}{dx} = 2 - y$. The left side is $-Ce^{-x}$, and the right side is $2 - (Ce^{-x} + 2) = -Ce^{-x}$. They match for every value of $C$, so this formula is the general solution, a whole family of curves.
Matching at a few points is not enough. For $\frac{dy}{dx} = 3y$, the function $y = x^3$ has derivative $3x^2$, while $3y = 3x^3$. The two sides agree at $x = 0$ and $x = 1$ but nowhere else, so $x^3$ fails. A constant multiplier is harmless here, because $y = 5e^{-2x}$ solves $\frac{dy}{dx} = -2y$. An added constant is not, because $y = e^{-2x} + 5$ puts a $-10$ on the right side that the derivative never produces.
Second-order equations work the same way. For $y'' + 4y = 0$, try $y = \sin(2x)$. Then $y'' = -4\sin(2x)$, and the sum is zero. With $\sin(4x)$, $y'' = -16\sin(4x)$ and the sum is $-12\sin(4x)$, so it fails. Implicit relations are checked by implicit differentiation. From $x^2 + y^2 = 25$ you get $2x + 2y\frac{dy}{dx} = 0$, so the circle solves $\frac{dy}{dx} = -\frac{x}{y}$.
A particular solution is the one member of the family that passes through a given point. If $y(0) = 7$ in the first example, then $7 = C + 2$, so $C = 5$.
2. Slope fields and how to reason with them
Key idea: Test the axes, look for column or row patterns, and find where the slope is zero; a flat row across the field is an equilibrium solution.
A slope field is a grid of short segments. At each grid point, the segment has the slope the equation gives there. To sketch one, evaluate $\frac{dy}{dx}$ at each listed point and draw a short segment with that slope. Zero slopes are flat, and large slopes are steep.
To match a field to an equation, run three quick checks. First, test the axes. For $\frac{dy}{dx} = x$, every point on the $y$-axis has $x = 0$, so every segment there is flat. Second, look for patterns. If the slope depends only on $x$, the segments in each vertical column are identical. If it depends only on $y$, the segments in each horizontal row are identical. Third, find where the slope is zero. For $\frac{dy}{dx} = x - y$, the flat segments lie along the line $y = x$, with positive slopes below it and negative slopes above it.
A row of flat segments running across the whole field marks an equilibrium solution. For $\frac{dy}{dx} = y(2 - y)$, the lines $y = 0$ and $y = 2$ are solutions. Between them the slope is positive, so a curve starting at $y = 1$ rises toward 2. Above 2 the slope is negative, so curves fall back toward 2. This is how you answer a question about $\lim_{x\to\infty} y$ straight from a slope field.
To sketch a particular solution, start at the given point and follow the segments. Your curve should run along the segments it passes, not cut across them.
3. Separation of variables and the domain of a solution
Key idea: Separate, integrate with one constant, apply the initial condition, then solve for $y$; $e^{x + C}$ becomes $Ae^{x}$, not $e^{x} + C$.
Separation works when the equation can be written as $\frac{dy}{dx} = g(x)h(y)$. Move every $y$ to the side with $dy$ and every $x$ to the side with $dx$. Then integrate both sides, add one constant, apply the initial condition, and solve for $y$.
Take $\frac{dy}{dx} = 2y$ with $y(0) = 3$. Separating gives $\frac{1}{y}\,dy = 2\,dx$, and integrating gives $\ln|y| = 2x + C$. Exponentiate to get $|y| = e^{2x + C} = e^{C}e^{2x}$. Because $e^{C}$ is just a constant, write $y = Ae^{2x}$. Then $3 = Ae^{0}$, so $y = 3e^{2x}$. The constant multiplies $e^{2x}$ because $e^{2x + C}$ splits into a product. That is why $y = e^{2x} + 2$ is wrong, even though it also passes through $(0, 3)$.
A particular solution also needs a domain. It is an open interval that contains the initial $x$-value, on which the solution is defined and differentiable. For $\frac{dy}{dx} = xy^2$ with $y(0) = 1$, separating gives $y^{-2}\,dy = x\,dx$, so $-\frac{1}{y} = \frac{x^2}{2} + C$. The point $(0, 1)$ gives $C = -1$, and solving gives $y = \frac{2}{2 - x^2}$. The formula breaks at $x = \pm\sqrt{2}$. The solution curve cannot jump across a break, so its domain is $-\sqrt{2} < x < \sqrt{2}$, not every $x$ except $\pm\sqrt{2}$.
4. Exponential growth and decay, and modeling from words
Key idea: $\frac{dy}{dt} = ky$ means $y = y_0e^{kt}$: the starting value goes in front, the rate constant goes in the exponent.
The sentence "the rate of change of $y$ is proportional to $y$" translates to $\frac{dy}{dt} = ky$. Its solutions are exactly the functions $y = y_0e^{kt}$, where $y_0$ is the value at $t = 0$. If $k > 0$ the quantity grows, and if $k < 0$ it decays. The starting value sits in front and the rate constant sits in the exponent, and swapping them is a classic wrong answer.
You usually find $k$ from a second data point. Suppose a culture of 200 bacteria triples in 5 hours. Then $600 = 200e^{5k}$, so $e^{5k} = 3$ and $k = \frac{\ln 3}{5} \approx 0.22$ per hour. Decay works the same way. A half-life of $h$ years means $\frac{1}{2} = e^{kh}$, so $k = -\frac{\ln 2}{h}$.
Do not confuse a constant rate with a proportional one. $\frac{dy}{dt} = 2$ gives a straight line, $y = 2t + y_0$. But $\frac{dy}{dt} = 2y$ gives an exponential, because the rate grows as $y$ grows.
Modeling questions test translation more than solving. "Proportional to the square root of $V$" means $\frac{dV}{dt} = k\sqrt{V}$. "Inversely proportional to $y$" means $\frac{dy}{dt} = \frac{k}{y}$. "Proportional to the difference between 100 and $y$" means $\frac{dy}{dt} = k(100 - y)$. "Jointly proportional to $P$ and $500 - P$" means $\frac{dP}{dt} = kP(500 - P)$. Once you have the equation, the rest is separation or a solution form you already know.
5. Newton's Law of Cooling
Key idea: $T = A + (T_0 - A)e^{-kt}$: the gap to the surrounding temperature decays exponentially, so the rate is largest when the gap is largest.
Newton's Law of Cooling says an object's temperature changes at a rate proportional to the gap between its temperature and the surrounding temperature. With surrounding temperature $A$, write $\frac{dT}{dt} = -k(T - A)$, where $k > 0$. If $T > A$, the right side is negative, so the object cools. If $T < A$, the right side is positive, so it warms. Either way it moves toward $A$.
Substitute $u = T - A$ and the equation becomes $\frac{du}{dt} = -ku$, so $u = u_0e^{-kt}$. Back in terms of $T$, the solution is $T = A + (T_0 - A)e^{-kt}$. The gap shrinks exponentially, so $T$ approaches $A$ but never reaches it.
Here is a full example. Coffee at 90°C sits in a 20°C room and reads 60°C after 10 minutes. The gap starts at 70 and falls to 40, so $40 = 70e^{-10k}$. That gives $k = \frac{\ln(7/4)}{10} \approx 0.056$ per minute. After 20 minutes the gap is $70\left(\frac{4}{7}\right)^2 \approx 22.9$, so the coffee is about 42.9°C.
The rate is fastest at the start, when the gap is largest. At $t = 0$ the coffee cools at about $0.056 \times 70 \approx 3.9$ degrees per minute. At 60°C it cools at about $0.056 \times 40 \approx 2.2$ degrees per minute. That shrinking rate is why a cooling curve is concave up. Many questions ask for the rate at a given temperature, and you get it straight from the equation without solving.
6. Logistic growth
Key idea: Carrying capacity is the nonzero zero of $\frac{dP}{dt}$, $P$ approaches it in the long run, and growth is fastest at half of it.
Logistic growth models a population that grows fast at first and then levels off at a carrying capacity $L$. The equation is $\frac{dP}{dt} = kP\left(1 - \frac{P}{L}\right)$ with $k > 0$. When $P$ is small, the factor in parentheses is close to 1, so growth is nearly exponential. As $P$ nears $L$, that factor nears 0, so growth stalls.
You can read most answers straight from the equation. The carrying capacity is the nonzero value of $P$ where $\frac{dP}{dt} = 0$. If $0 < P(0) < L$, the population increases toward $L$, so $\lim_{t\to\infty} P(t) = L$. If $P(0) > L$, the rate is negative, and $P$ decreases toward $L$. The population grows fastest when $P = \frac{L}{2}$. That is because $kP\left(1 - \frac{P}{L}\right)$ is a downward parabola in $P$, with its vertex halfway between its zeros at 0 and $L$. If $0 < P(0) < \frac{L}{2}$, the graph of $P$ against $t$ has its inflection point where $P = \frac{L}{2}$. A population that starts between $\frac{L}{2}$ and $L$ never passes $\frac{L}{2}$, so its graph has no inflection point.
Watch for the second common form, $\frac{dP}{dt} = kP(L - P)$. The carrying capacity is still the nonzero zero, $L$. But near $P = 0$ the rate is about $kL \cdot P$, so the near-exponential growth rate is $kL$, not $k$. For $\frac{dP}{dt} = 0.002P(800 - P)$, the carrying capacity is 800, growth is fastest at $P = 400$, and early growth runs at a relative rate of about $0.002 \times 800 = 1.6$.
For the first form, the explicit solution is $P(t) = \frac{L}{1 + Ae^{-kt}}$, where $A = \frac{L - P(0)}{P(0)}$. You need it only when a question asks for $P$ at a specific time.
7. Euler's method
Key idea: New $y$ = old $y$ + slope at the old point times $\Delta x$; concave up means Euler underestimates, concave down means it overestimates.
Euler's method approximates a solution by walking along short tangent lines. Start at the known point $(x_0, y_0)$ and find the slope there from the equation. Step right by $\Delta x$, and change $y$ by the slope times the step. In symbols, $y_{\text{new}} = y_{\text{old}} + m\,\Delta x$, where $m$ is $\frac{dy}{dx}$ evaluated at the old point. Then repeat from the new point.
Take $\frac{dy}{dx} = x + y$ with $y(0) = 1$, and estimate $y(1)$ with two steps of size 0.5. At $(0, 1)$ the slope is 1, so $y \approx 1 + 1(0.5) = 1.5$ at $x = 0.5$. At $(0.5, 1.5)$ the slope is 2, so $y \approx 1.5 + 2(0.5) = 2.5$ at $x = 1$. Keep a small table with columns for $x$, $y$, slope and change in $y$. It stops the most common slip, which is reusing the starting $y$ when computing the second slope.
The true value is $y(1) = 2e - 2 \approx 3.44$, so this estimate came out low. That is predictable. Here $\frac{d^2y}{dx^2} = 1 + x + y$, which is positive near these points, so the solution is concave up. Tangent lines sit below a concave-up curve, so each step lands under it. A concave-down solution gives an overestimate instead. Smaller steps shrink the error, so ten steps of 0.1 would land closer.
One step of Euler's method is the same calculation as a tangent line approximation. Exam questions often ask for both in the same problem.
8. Second-derivative analysis
Key idea: Differentiate implicitly, substitute the original $\frac{dy}{dx}$, and use the sign of $\frac{d^2y}{dx^2}$ to judge concavity, extrema and over- or underestimates.
You can find concavity without solving the equation. Differentiate $\frac{dy}{dx}$ with respect to $x$, treating $y$ as a function of $x$. Every time you differentiate a $y$, the chain rule attaches a factor of $\frac{dy}{dx}$. Then replace $\frac{dy}{dx}$ with the original right side, so the answer is in terms of $x$ and $y$ only.
Take $\frac{dy}{dx} = xy$. The product rule gives $\frac{d^2y}{dx^2} = y + x\frac{dy}{dx}$. Substituting gives $y + x(xy) = y(1 + x^2)$. Since $1 + x^2 > 0$, solution curves are concave up wherever $y > 0$ and concave down wherever $y < 0$.
The exam then asks you to use that sign. A typical stem says $(1, 2)$ is on a solution of $\frac{dy}{dx} = xy$ and asks about the tangent line there. The slope is $1 \cdot 2 = 2$, so the tangent line is $y = 2 + 2(x - 1)$, which estimates $y(1.1) \approx 2.2$. At $(1, 2)$, $\frac{d^2y}{dx^2} = 2(1 + 1) = 4 > 0$. The curve is concave up, so near $x = 1$ the tangent line lies below it, and 2.2 is an underestimate.
The same work classifies critical points of a solution. If $\frac{dy}{dx} = 0$ at a point and $\frac{d^2y}{dx^2} > 0$ there, the solution has a relative minimum. If $\frac{d^2y}{dx^2} < 0$, it has a relative maximum. For a logistic equation, $\frac{d^2P}{dt^2} = 0$ at $P = 0$, $P = \frac{L}{2}$ and $P = L$. Only $P = \frac{L}{2}$ is an inflection point, because $P = 0$ and $P = L$ are equilibrium solutions whose graphs are flat lines.