{
  "id": "tangent-line",
  "title": "Tangent Line",
  "lang": "en",
  "source_lang": "en",
  "summary": "A tangent line describes the local direction of a curve at a point and, at regular points, gives its first-order straight-line approximation.",
  "category": "mathematics",
  "subcategory": "geometry",
  "type": "concept",
  "aliases": [
    "Tangent"
  ],
  "generated_at": "2026-11-03",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/tangent-line/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/tangent-line/",
      "json": "https://aiwiki.page/zh/tangent-line.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "A tangent line is a straight line that represents the local direction of a curve at a specified point, called the point of tangency. In geometry, it is commonly understood as the limiting position of lines through two nearby points on the curve. For the graph of a function that has a finite derivative, the tangent line has slope equal to that derivative. Tangency is a local property: a tangent may cross the curve at the point of tangency or meet it again elsewhere."
    },
    {
      "heading": "Definition through secant lines",
      "text": "A secant line passes through two distinct points of a curve. Fix a point \\(P\\) on the curve and let another point \\(Q\\) approach \\(P\\) along it. If the direction of the secant line \\(PQ\\) approaches a unique limiting direction, the line through \\(P\\) with that direction is a tangent line.\n\nThis construction uses a limit of line directions rather than a requirement about how many times the line intersects the curve. For a general curve, a line can meet the curve at exactly one point without being tangent, and a tangent can have several intersections with the curve.\n\nFor the graph \\(y=f(x)\\), take\n\n\\[P=(a,f(a)),\\qquad Q=(a+h,f(a+h)).\\]\n\nWhen \\(h\\ne 0\\), the secant slope is\n\n\\[\\frac{f(a+h)-f(a)}{h}.\\]\n\nIf this expression has a finite limit as \\(h\\to 0\\), then\n\n\\[f'(a)=\\lim_{h\\to 0}\\frac{f(a+h)-f(a)}{h},\\]\n\nand the tangent line is\n\n\\[\\boxed{y-f(a)=f'(a)(x-a).}\\]\n\nAt an endpoint of a curve, a tangent may instead be defined using a one-sided limit."
    },
    {
      "heading": "Tangents to function graphs",
      "text": "Basic examples\n\nFor the parabola \\(y=x^2\\), the derivative is \\(2x\\). At the point \\((a,a^2)\\), the tangent is\n\n\\[y-a^2=2a(x-a),\\]\n\nor\n\n\\[y=2ax-a^2.\\]\n\nIts slope is positive for \\(a>0\\), negative for \\(a<0\\), and zero at the vertex.\n\nFor \\(y=x^3\\), the tangent at the origin is \\(y=0\\). The curve lies below this line for negative \\(x\\) and above it for positive \\(x\\), so the tangent crosses the curve at its point of tangency. This shows why “touches but does not cross” is not a general definition of a tangent.\n\nHorizontal and vertical tangents\n\nA tangent is horizontal when\n\n\\[f'(a)=0.\\]\n\nA horizontal tangent does not necessarily indicate a local maximum or minimum; \\(y=x^3\\) at the origin is an example.\n\nA vertical tangent has equation\n\n\\[x=a.\\]\n\nIt cannot be represented by a finite slope. For example, the curve \\(y=x^{1/3}\\) has a vertical tangent at the origin. Away from zero,\n\n\\[f'(x)=\\frac{1}{3x^{2/3}},\\]\n\nwhich grows without bound as \\(x\\to 0\\). The curve can also be parametrized as \\((x,y)=(t^3,t)\\), making its vertical direction at \\(t=0\\) explicit.\n\nThus, the existence of a geometric tangent does not always imply the existence of a finite derivative when the curve is expressed as \\(y=f(x)\\).\n\nCorners, cusps, and failure of tangency\n\nA continuous graph need not have a tangent at every point. For \\(y=|x|\\), the left-hand secant slopes at the origin tend to \\(-1\\), while the right-hand slopes tend to \\(1\\). The two branches have different limiting directions, so there is no single tangent line there.\n\nCusps require more care. The curve\n\n\\[y^2=x^3\\]\n\nhas two branches meeting at the origin, both with the same limiting tangent line \\(y=0\\). It therefore has a tangent in the limiting-line sense, although the cusp is not a regular smooth point.\n\nOther curves can oscillate so rapidly near a point that no limiting secant direction exists. Consequently, continuity alone does not ensure tangency."
    },
    {
      "heading": "Tangency as first-order approximation",
      "text": "For a differentiable function, the tangent line provides a first-order local approximation:\n\n\\[f(a+h)=f(a)+f'(a)h+o(|h|).\\]\n\nThe error term means\n\n\\[\\lim_{h\\to 0} \\frac{f(a+h)-f(a)-f'(a)h}{|h|}=0.\\]\n\nThus the approximation error becomes negligible relative to the size of the horizontal displacement.\n\nThe function\n\n\\[L(x)=f(a)+f'(a)(x-a)\\]\n\nis often called the linearization of \\(f\\) at \\(a\\). Strictly speaking, it is an affine function: it is a linear expression plus a constant.\n\nThis interpretation connects tangency with Taylor expansion. If \\(f\\) is twice continuously differentiable near \\(a\\), then\n\n\\[f(a+h) =f(a)+f'(a)h+\\frac12f''(a)h^2+o(h^2).\\]\n\nThe tangent captures the constant and first-order terms, while higher-order terms describe the curve’s departure from the line. A big-O estimate \\(O(h^2)\\) for the tangent-line error is also available when the second derivative is bounded nearby."
    },
    {
      "heading": "Parametric and implicit descriptions",
      "text": "Parametric curves\n\nLet a curve in Euclidean space be given by a differentiable parametrization\n\n\\[\\mathbf r(t)=\\bigl(x_1(t),\\ldots,x_n(t)\\bigr).\\]\n\nIf\n\n\\[\\mathbf r'(t_0)\\ne\\mathbf 0,\\]\n\nits tangent line at \\(\\mathbf r(t_0)\\) is\n\n\\[\\boxed{\\mathbf x=\\mathbf r(t_0)+s\\,\\mathbf r'(t_0), \\qquad s\\in\\mathbb R.}\\]\n\nThe vector \\(\\mathbf r'(t_0)\\) supplies the tangent direction. For a moving particle, it is the instantaneous velocity vector.\n\nIn the plane, if \\(x'(t_0)\\ne 0\\), the tangent slope is\n\n\\[\\frac{dy}{dx}=\\frac{y'(t_0)}{x'(t_0)}.\\]\n\nIf \\(x'(t_0)=0\\) but \\(y'(t_0)\\ne 0\\), the tangent is vertical.\n\nA zero derivative vector makes this formula inconclusive, but does not by itself rule out a tangent. For example, \\(\\mathbf r(t)=(t^3,0)\\) has zero derivative at \\(t=0\\), yet its image is the horizontal axis.\n\nAt a self-intersection, distinct parameter values may describe the same point with different tangent directions. Tangency is then associated with each local branch rather than necessarily with the point alone.\n\nImplicit curves\n\nSuppose a plane curve is defined by\n\n\\[F(x,y)=0,\\]\n\nwhere \\(F\\) is continuously differentiable. At a point \\((a,b)\\) satisfying \\(F(a,b)=0\\) and\n\n\\[\\nabla F(a,b)\\ne\\mathbf 0,\\]\n\nthe tangent line is\n\n\\[\\boxed{F_x(a,b)(x-a)+F_y(a,b)(y-b)=0.}\\]\n\nHere \\(F_x\\) and \\(F_y\\) are partial derivatives. The gradient is perpendicular to the tangent line.\n\nWhen \\(F_y(a,b)\\ne 0\\), implicit differentiation gives\n\n\\[\\frac{dy}{dx}=-\\frac{F_x(a,b)}{F_y(a,b)}.\\]\n\nIf both partial derivatives vanish, the first-order equation supplies no tangent direction. Higher-order terms or a parametrization must then be examined. Such a point may have a unique tangent, several branch tangents, or no real curve passing through it except the point itself."
    },
    {
      "heading": "Circles and conic sections",
      "text": "For a circle with center \\((h,k)\\) and radius \\(r>0\\),\n\n\\[(x-h)^2+(y-k)^2=r^2,\\]\n\nthe tangent at a point \\((a,b)\\) on the circle is\n\n\\[(a-h)(x-a)+(b-k)(y-b)=0.\\]\n\nEquivalently,\n\n\\[(a-h)(x-h)+(b-k)(y-k)=r^2.\\]\n\nThis expresses the classical fact that the tangent is perpendicular to the radius drawn to the point of tangency.\n\nFor the ellipse\n\n\\[\\frac{x^2}{A^2}+\\frac{y^2}{B^2}=1,\\]\n\nthe tangent at \\((a,b)\\) is\n\n\\[\\frac{ax}{A^2}+\\frac{by}{B^2}=1.\\]\n\nA line tangent to a nondegenerate conic section has a repeated intersection at the point of contact, understood algebraically with multiplicity. In algebraic geometry, this viewpoint extends tangency beyond ordinary real plane curves, including to complex and projective settings."
    },
    {
      "heading": "Normal lines and tangent spaces",
      "text": "For a regular plane curve, the normal line at a point is the line through that point perpendicular to the tangent. If the tangent has finite nonzero slope \\(m\\), the normal has slope \\(-1/m\\). A horizontal tangent has a vertical normal, and a vertical tangent has a horizontal normal.\n\nThe distinction between a tangent line and a tangent space is important. The geometric tangent line passes through the point on the curve. Its associated tangent space consists of tangent vectors based at that point and is a one-dimensional vector space. For an embedded curve, translating this vector space to the point produces the geometric tangent line.\n\nThis construction generalizes in differential geometry: a smooth surface has a tangent plane, and a smooth manifold has a tangent space whose dimension equals that of the manifold."
    },
    {
      "heading": "Applications and historical development",
      "text": "Tangent constructions were studied in ancient Greek geometry, particularly for circles and conic sections. Euclid’s Elements includes propositions relating circle tangents to radii, while Archimedes studied a tangent to the spiral bearing his name. In the seventeenth century, methods associated with Pierre de Fermat and the development of calculus connected tangent problems with instantaneous rates of change.\n\nA central numerical application is Newton’s method. Starting from \\(x_n\\), it uses the zero of the tangent-line approximation to obtain\n\n\\[x_{n+1}=x_n-\\frac{f(x_n)}{f'(x_n)}.\\]\n\nThe method requires a nonzero derivative at each step and can fail to converge when the initial estimate or the function’s local behavior is unsuitable.\n\nTangents also express necessary conditions in optimization. At an interior local extremum of a differentiable real function, the tangent is horizontal. The converse is false: a horizontal tangent need not mark an extremum.\n\nFor a differentiable convex function on an interval, every tangent line lies below the graph:\n\n\\[f(x)\\ge f(a)+f'(a)(x-a).\\]\n\nHere the tangent is not only a local approximation but also a global supporting line on the interval—a property that does not hold for arbitrary differentiable functions."
    }
  ],
  "links": [
    "geometry",
    "graph-of-a-function",
    "derivative",
    "limit",
    "parabola",
    "continuous-function",
    "affine-map",
    "taylor-series",
    "big-o-notation",
    "euclidean-space",
    "partial-derivative",
    "gradient",
    "circle",
    "ellipse",
    "conic-section",
    "algebraic-geometry",
    "tangent-space",
    "vector-space",
    "differential-geometry",
    "manifold",
    "euclids-elements",
    "archimedes",
    "calculus",
    "newtons-method",
    "mathematical-optimization",
    "convex-function"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}