Direction fields and Euler's method — visual - numerical intuition first: Question bank

4.6.2 · D5 · Maths › Ordinary Differential Equations › Direction fields and Euler's method — visual - numerical int

Spot the error

Har statement mein ek flaw hai. Use name karo aur fix karo.

"Euler step:

y_{n + 1} = y_{n} + f (x_{n}, y_{n})

Step size missing hai.

f

ek slope hai, rise nahi; tumhe ise run

h

se multiply karna hoga:

y_{n + 1} = y_{n} + h f (x_{n}, y_{n})

. Units check karo:

Δ y = (Δ y /Δ x) \cdot Δ x

"Accurate hone ke liye, slope ko us point par evaluate karo jahan tum ja rahe ho:

f (x_{n + 1}, y_{n + 1})

Ye ek alag method hai — implicit (backward) Euler. Explicit Euler ko us point par slope use karni chahiye jahan se wo nikal raha hai,

f (x_{n}, y_{n})

, kyunki

y_{n + 1}

abhi pata nahi hai.

"Direction field arrows wo direction dikhate hain jisme particle time ke through move karta hai."

Ye velocity vectors nahi hain. Har segment sirf us point par solution ki tangent direction mark karta hai; iska koi speed nahi hai aur koi time-orientation nahi hai.

"Kyunki

y^{'} = x + y

(0, 1)

slope

1

deta hai,

h = 0.5

ke step ke baad new

y

1 + 1 = 2

hai."

h

se scale karna bhool gaye. Rise

= h \cdot slope = 0.5 \cdot 1 = 0.5

, isliye

y_{1} = 1.5

hai,

2

nahi.

"Ek isocline wo curve hai jahan solution ki slope zero hoti hai."

Ye sirf special isocline

f = 0

hai. Ek isocline

f (x, y) = m

hoti hai kisi bhi fixed

m

ke liye;

m = 0

wali (horizontal dashes) sirf family ka ek member hai.

"Kyunki Euler convergent hai, finite

h

ke liye iska answer exact solution ke barabar hai."

Convergence ka matlab hai error

\to 0

jaise

h \to 0

, ye ek limiting statement hai. Kisi bhi finite

h

ke liye genuine

O (h)

error hoti hai; convergent

\neq =

exact.

y^{''} > 0

ka matlab hai curve upar slope karti hai, isliye Euler overshoots karta hai."

y^{''} > 0

ka matlab hai curve convex hai (upar ki taraf mudi hui), ye nahi ki wo rise karti hai. Ek convex curve ke tangents uske neeche hote hain, isliye Euler undershoots karta hai, overshoots nahi.

Why questions

ODE ek single curve ki jagah slopes ka field kyun deta hai?

Kyunki

f (x, y)

plane mein har point

(x, y)

ko ek slope assign karta hai; tum sirf ek curve pin down karte ho jab tum ek initial condition bhi fix kar lete ho ki kahan se start karna hai.

Hum Euler paane ke liye

\frac{h ^{2}}{2} y^{''} (ξ)

term ko kyun drop karte hain?

Hum usually

y^{''}

nahi jaante, aur chhote

h

ke liye wo term

h y^{'}

ki tulna mein tiny hoti hai. Use drop karna exact Taylor expansion ko ek computable one-step rule banata hai — ek known

O (h^{2})

per-step error ki cost par.

Global error ek order worse (

O (h)

) kyun hai local error (

O (h^{2})

) se?

Length

L

ke fixed interval ko cross karne ke liye tum

N = L / h

steps lete ho, har step

\sim h^{2}

contribute karta hai: total

\sim (L / h) \cdot h^{2} = L h = O (h)

. Zyada steps extra accuracy ko dilute kar dete hain.

Explicit Euler left endpoint ki slope kyun use karta hai?

Kyunki stepping ke waqt sirf

y_{n}

pata hota hai; slope

f (x_{n}, y_{n})

computable hai, jabki

f (x_{n + 1}, y_{n + 1})

ke liye unknown

y_{n + 1}

chahiye jo tum dhundh rahe ho.

Isoclines haath se sketching itni fast kyun banate hain?

Ek isocline ke along har dash ka same tilt hota hai, isliye tum ek saath parallel dashes ki puri family draw karte ho bajaye point by point

f

compute karne ke — field almost free mein appear hoti hai.

RK4 jaise methods Euler se behtar kyun hain?

Wo har step ke andar kai points par slope sample karte hain aur unhe blend karte hain, zyada Taylor terms cancel karte hain, isliye unka error

O (h^{4})

hota hai — same

h

ke liye dramatically chhota.

Euler kisi point ke through solution follow kyun nahi kar sakta jahan

f

blow up karta hai?

Infinite slope ka koi finite tangent step nahi hota; local linear approximation breakdown ho jaati hai, aur aise singularities ke paas uniqueness bhi fail ho sakti hai — "follow your nose" wali picture meaningful nahi rehti.

Edge cases

Agar

h = 0

hai, toh ek Euler step kya deta hai?

y_{n + 1} = y_{n} + 0 = y_{n}

— tum kabhi move nahi karte. Zero step degenerate hai: koi progress nahi, koi error bhi nahi, aur kahin bhi jaane ke liye infinitely many steps.

Jahan

f (x, y) = 0

ho wahan direction field kaisi dikhti hai?

Ek horizontal dash (slope

0

). Wahan se guzarne wale solutions momentarily flat hote hain; agar

f = 0

poori curve par hai toh wo ek equilibrium ho sakta hai jis taraf solution level off karta hai.

y^{'} =

constant ki exact straight-line solution par Euler ka kya hota hai?

Har step par ye exact hai, kyunki

y^{''} = 0

truncation term ko khatam kar deta hai. Euler aur true solution

h

ki parwah kiye bina coincide karte hain.

Ek separable ODE ke liye jise tum exactly solve kar sakte ho, Euler se kyun bother karo?

Numerical method ko known answer ke against verify karne ke liye aur intuition banane ke liye; lekin agar closed form exist karta hai, exact solution prefer ki jaati hai — Euler tab shine karta hai jab koi formula available nahi hota.

Ek stiff equation par "small"

h

ke saath bhi kya ho sakta hai?

Explicit Euler unstable ho sakta hai aur oscillate ya blow up kar sakta hai jab tak

h

extremely tiny na ho; step size ko stability limit respect karni hoti hai, sirf accuracy nahi — ye Stability and Stiff Equations ka domain hai.

Ek point par jahan do solution curves touch karni padti hain lekin cross nahi, kya field ambiguous hai?

Agar uniqueness hold karta hai, wahan sirf ek slope exist karta hai isliye koi ambiguity nahi; ambiguity signal karta hai ki

f

\partial f / \partial y

misbehave kar raha hai, jahan single-arrow picture trustworthy nahi hoti.

Agar true solution interval par har jagah concave hai (

y^{''} < 0

), toh Euler kis taraf error karta hai?

Ye overshoots: concave curve ke tangents uske upar hote hain, isliye har step thoda zyada upar land karta hai, jo convex undershoot ka mirror image hai.

Active recall

Recall Self-quiz ke liye one-line traps

Segment length ka matlab? → Kuch nahi; sirf tilt matter karta hai.
Uniqueness ke under curves crossing? → Impossible.
Galat Euler formula mein missing factor? → Step $h$ .
Straight-line solution par Euler? → Exact.
Convex curve → Euler? → Undershoots.
Concave curve → Euler? → Overshoots.
Global error order? → $O (h)$ .

Connections

Direction fields and Euler's method — visual - numerical intuition first — wo parent jise ye bank drill karta hai.
Taylor's Theorem — kyun dropped term $O (h^{2})$ hai.
Runge-Kutta Methods (RK4) — accurate successor.
Existence and Uniqueness (Picard–Lindelöf) — kyun curves cross nahi karte.
Stability and Stiff Equations — kyun "zyada steps" phir bhi explode ho sakta hai.
Separable First-Order ODEs — test karne ke liye exact answers.

Question bank — Direction fields and Euler's method — visual - numerical intuition first

True or false — justify