Inverse Hyperbolics as Logarithms

Solve for the input and a logarithm appears.

What arsinh x means

arsinh x is the number whose sinh is x. Call that number y. Then y = arsinh x says exactly the same as x = sinh y, and sinh has a definition in exponentials: x = (eʸ − e⁻ʸ)/2.

That turns the question into solving an equation for y. The aim is a formula for y in terms of x, and since y sits in an exponent, a logarithm will be needed at the end to bring it down.

Clear the negative power

Multiply both sides by 2: 2x = eʸ − e⁻ʸ. The term e⁻ʸ is 1/eʸ, so multiplying every term by eʸ clears it, because eʸ × e⁻ʸ = e⁰ = 1. The left side becomes 2x eʸ and the right side becomes e²ʸ − 1.

Collect everything on one side: e²ʸ − 2x eʸ − 1 = 0.

A quadratic in eʸ

Write u for eʸ. Then e²ʸ = (eʸ)² = u², and the equation reads u² − 2x u − 1 = 0, a quadratic in u with leading coefficient 1, middle coefficient −2x and constant −1.

The quadratic formula gives u = (2x ± √(4x² + 4))/2. Take 4 out of the root as 2: √(4x² + 4) = 2√(x² + 1). Dividing by 2 leaves eʸ = x ± √(x² + 1).

xy

The two roots as x varies. The gold curve, x + √(x² + 1), stays above the x-axis for every x; the plain curve, x − √(x² + 1), stays below it. At x = 0.75 the roots are 2 and −0.5, and at x = 0 they are 1 and −1: each pair multiplies to −1.

Reject the negative root

The two roots cannot both be values of eʸ. Since x² + 1 is bigger than x², its square root is bigger than √(x²) = |x|, and |x| is never less than x. So √(x² + 1) > x for every x, and the root x − √(x² + 1) is always negative.

An exponential is never negative, so eʸ cannot equal that root. Another way to see it: the two roots of u² − 2x u − 1 = 0 multiply to the constant term, −1, so one is positive and the other negative whatever x is. The positive one is eʸ = x + √(x² + 1).

Take logarithms

Taking the natural logarithm of both sides brings y down from the exponent: y = ln(x + √(x² + 1)). So arsinh x = ln(x + √(x² + 1)), for every real x. The bracket is always positive, so the logarithm always exists.

Try x = 0.75. Then x² + 1 = 1.5625, whose square root is 1.25, so the bracket is 0.75 + 1.25 = 2 and arsinh 0.75 = ln 2. Check it the other way: sinh(ln 2) = (2 − ½)/2 = 3/4, since e to the power ln 2 is 2 and e to the power −ln 2 is ½.

arcosh: one sign changes

For y = arcosh x, start from x = cosh y = (eʸ + e⁻ʸ)/2. The same steps give 2x eʸ = e²ʸ + 1, so e²ʸ − 2x eʸ + 1 = 0. Only the constant has changed sign, and the quadratic formula gives eʸ = x ± √(x² − 1). The root needs x ≥ 1, which is where arcosh is defined.

This time neither root is negative. The two roots multiply to the constant, +1, so they are reciprocals, and their logarithms are a number and its negative. At x = 2 the roots are 2 + √3 = 3.7321 and 2 − √3 = 0.2679, with logarithms 1.3170 and −1.3170. These are the two inputs where cosh is 2, one on each side of the y-axis.

arcosh takes the non-negative input, which comes from the larger root. So arcosh x = ln(x + √(x² − 1)) for x ≥ 1, and arcosh 2 = ln(2 + √3) = 1.3170.

artanh: no quadratic at all

For y = artanh x, start from x = tanh y = (eʸ − e⁻ʸ)/(eʸ + e⁻ʸ). Multiply the top and the bottom by eʸ: x = (e²ʸ − 1)/(e²ʸ + 1).

Multiply out: x e²ʸ + x = e²ʸ − 1. Gather the e²ʸ terms on one side: e²ʸ(1 − x) = 1 + x, so e²ʸ = (1 + x)/(1 − x). This is a linear equation in e²ʸ, with one solution and nothing to reject.

Take logarithms and halve: y = ½ ln((1 + x)/(1 − x)). The fraction is positive exactly when −1 < x < 1, the inputs artanh accepts. At x = 0.6 the fraction is 1.6/0.4 = 4, so artanh 0.6 = ½ ln 4 = ln 2.

Checks

At zero the arsinh formula gives ln(0 + √1) = ln 1 = 0, and sinh 0 is 0, so arsinh 0 = 0 as it should be. The arcosh formula at 1 gives ln(1 + 0) = 0, and cosh 0 = 1. The artanh formula at 0 gives ½ ln 1 = 0, and tanh 0 = 0.

Three values give ln 2: arsinh(3/4), arcosh(5/4) and artanh(3/5). For arcosh, 5/4 + √(25/16 − 1) = 5/4 + 3/4 = 2. Going back, e to the power ln 2 is 2 and e to the power −ln 2 is ½, so sinh(ln 2) = (2 − ½)/2 = 3/4, cosh(ln 2) = (2 + ½)/2 = 5/4 and tanh(ln 2) = 1.5/2.5 = 3/5.

For large x, √(x² + 1) is very close to x, so arsinh x is close to ln 2x. At x = 10, arsinh 10 = ln(10 + √101) = ln 20.0499 = 2.9982, and ln 20 = 2.9957.

x² + y² = 1area = t/2 = 0.25x² − y² = 1area = t/2 = 0.25(cosh t, sinh t) = (1.13, 0.52)t = 0.5

the hyperbolic sector and the circular sector both have area t/2, so t is an angle for the hyperbola: (cosh t, sinh t) is to x² − y² = 1 what (cos t, sin t) is to x² + y² = 1

Sweep t until both sectors have area ½

The point (cosh t, sinh t) on the hyperbola x² − y² = 1, beside the point at angle t on the unit circle; each shaded sector has area t/2. At t = 0.5 the point is (1.1276, 0.5211), and 1.1276 + 0.5211 = 1.6487 = e^0.5. Drag t to 1 and the point is (1.5431, 1.1752), whose coordinates add to 2.7183, which is e. Since cosh t + sinh t = eᵗ for every t, t = ln(x + y) recovers t from the point.

The usual mistakes

Keeping both roots. eʸ = x − √(x² + 1) is negative, and no exponential is, so for arsinh only the plus sign gives an answer.

Mixing up the signs under the root. arsinh has x² + 1, because its quadratic has constant −1. arcosh has x² − 1, because its quadratic has constant +1.

Taking the logarithm of x alone. eʸ equals the whole of x + √(x² + 1), so the whole of it goes inside the logarithm: ln(x + √(x² + 1)), not ln x + √(x² + 1).

Losing the half, or turning the fraction upside down, in artanh. The equation gives e²ʸ, not eʸ, so the logarithm is halved, and the fraction is (1 + x)/(1 − x), which is bigger than 1 when x is positive.

A cable and two boosts

In the first application below, a chairlift cable hangs as a cosh curve, and arcosh written as a logarithm finds where it reaches a tower. In the second, the speed of a particle is c tanh w, and artanh written as half a logarithm turns each speed into the quantity w that adds.

Worked example: A Chairlift Cable Over a Valley: Where Along It the Cable Reaches the Top of a Tower

Question A chairlift's cable hangs as y = 45cosh(x45), where y is the height in meters above the valley floor and x is the horizontal distance in meters from the lowest point. A tower holds the cable at a height of 53 m. (a) How far horizontally is that tower from the lowest point of the cable? (b) How long is the cable between the lowest point and the tower? Give the first answer to three significant figures.

  1. 1.Put the height into the model and make the cosh the subject: 45cosh(x45) = 53, so cosh(x45) = 5345.

    42465054−27027x, meters from the lowest pointheight above the valley floor, m53 m45 cosh(x/45) = 53, so cosh(x/45) = 53/45
    42465054−27027x, meters from the lowest pointheight above the valley floor, m53 m45 cosh(x/45) = 53, so cosh(x/45) = 53/45
    The tower holds the cable at 53 m, so cosh(x45) = 5345.
  2. 2.Turn that round with the inverse hyperbolic cosine, and write it as a logarithm: x45 = arcosh(5345), and arcosh u = ln(u + √u2 − 1) for u ≥ 1.

    42465054−27027x, meters from the lowest pointheight above the valley floor, m53 m45 cosh(x/45) = 53, so cosh(x/45) = 53/45x/45 = arcosh(53/45), a logarithm
    42465054−27027x, meters from the lowest pointheight above the valley floor, m53 m45 cosh(x/45) = 53, so cosh(x/45) = 53/45x/45 = arcosh(53/45), a logarithm
    Turn that round with arcosh u = ln(u + √u2 − 1).
  3. 3.Work the root out exactly. (5345)2 − 1 = 2809 − 20252025 = 7842025, whose square root is 2845. So u + √u2 − 1 = 53 + 2845 = 8145 = 1.8.

    42465054−27027x, meters from the lowest pointheight above the valley floor, m53 m45 cosh(x/45) = 53, so cosh(x/45) = 53/45x/45 = arcosh(53/45), a logarithm53 × 53 − 45 × 45 = 784, root 28(53 + 28)/45 = 1.8
    42465054−27027x, meters from the lowest pointheight above the valley floor, m53 m45 cosh(x/45) = 53, so cosh(x/45) = 53/45x/45 = arcosh(53/45), a logarithm53 × 53 − 45 × 45 = 784, root 28(53 + 28)/45 = 1.8
    532 − 452 = 784, whose root is 28, so u + √u2 − 1 = 8145 = 1.8.
  4. 4.(a) Therefore x45 = ln 1.8 = 0.587787, and x = 45 × 0.587787 = 26.450 m, which is 26.5 m to three significant figures.

    42465054−27027x, meters from the lowest pointheight above the valley floor, m53 m26.5 m45 cosh(x/45) = 53, so cosh(x/45) = 53/45x/45 = arcosh(53/45), a logarithm53 × 53 − 45 × 45 = 784, root 28(53 + 28)/45 = 1.8(a) x = 45 ln 1.8 = 45 × 0.587787 = 26.5 m
    42465054−27027x, meters from the lowest pointheight above the valley floor, m53 m26.5 m45 cosh(x/45) = 53, so cosh(x/45) = 53/45x/45 = arcosh(53/45), a logarithm53 × 53 − 45 × 45 = 784, root 28(53 + 28)/45 = 1.8(a) x = 45 ln 1.8 = 45 × 0.587787 = 26.5 m
    (a) x = 45ln 1.8 = 26.450 m, which is 26.5 m to three significant figures.
  5. 5.(b) The length along the cable from the lowest point is 45sinh(x45), and with ex/45 = 1.8 that is 45 × 1.8 − 11.82 = 45 × 0.62222 = 28 m. Check with the identity: multiplying cosh2 u − sinh2 u = 1 by 452 gives 532 − (45sinhx45)2 = 452, so the length squared is 2809 − 2025 = 784 and the length is exactly 28 m.

    42465054−27027x, meters from the lowest pointheight above the valley floor, m53 m26.5 m28 m of cable45 cosh(x/45) = 53, so cosh(x/45) = 53/45x/45 = arcosh(53/45), a logarithm53 × 53 − 45 × 45 = 784, root 28(53 + 28)/45 = 1.8(a) x = 45 ln 1.8 = 45 × 0.587787 = 26.5 m(b) 45 sinh(x/45) = 45 × 0.62222 = 28 m
    42465054−27027x, meters from the lowest pointheight above the valley floor, m53 m26.5 m28 m of cable45 cosh(x/45) = 53, so cosh(x/45) = 53/45x/45 = arcosh(53/45), a logarithm53 × 53 − 45 × 45 = 784, root 28(53 + 28)/45 = 1.8(a) x = 45 ln 1.8 = 45 × 0.587787 = 26.5 m(b) 45 sinh(x/45) = 45 × 0.62222 = 28 m
    (b) The cable from the lowest point to the tower is 45sinh(x45) = 28 m exactly.

Answer: (a) 26.5 m; (b) exactly 28 m of cable

Common mistakes

  • Writing arcosh as 1cosh. The inverse function and the reciprocal are different things: 1cosh(5345) is about 0.55 and answers nothing here, while arcosh(5345) = 0.588 is the number whose cosh is 5345.
  • Keeping both signs from the square root, as one would for a quadratic. The logarithm form with the minus sign gives ln2545, a negative number, which names the matching point on the other side of the lowest point. The tower asked about is to one side, so the positive value is the one to report.

More hyperbolic functions problems, worked step by step →

Worked example: Two Boosts in a Particle Accelerator: The Quantity That Adds When Speeds Do Not

Question A particle in an accelerator is boosted to 0.6c, and is then given a second boost of 0.8c as measured in a frame moving with it. Speeds do not simply add here; the quantity that does is the rapidity w, defined by vc = tanh w. (a) What is the rapidity of each boost, and what do the two add to? (b) What is the particle's final speed, as a multiple of c? Give the rapidities to four significant figures.

  1. 1.Turn the definition round. From vc = tanh w the rapidity is w = artanh(vc), and the logarithm form of that is artanh k = 12ln1 + k1 − k for −1 < k < 1.

    00.69311.792rapidityv/c = tanh w, so w = artanh(v/c)
    00.69311.792rapidityv/c = tanh w, so w = artanh(v/c)
    The rapidity of a speed is w = artanh(vc) = 12ln1 + k1 − k.
  2. 2.The first boost has k = 0.6, so w1 = 12ln1.60.4 = 12ln 4 = ln 2 = 0.6931.

    00.69311.792rapidity0.6931 = ln 2v/c = tanh w, so w = artanh(v/c)artanh 0.6 = half ln 4 = ln 2 = 0.6931
    00.69311.792rapidity0.6931 = ln 2v/c = tanh w, so w = artanh(v/c)artanh 0.6 = half ln 4 = ln 2 = 0.6931
    The first boost has rapidity 12ln 4 = ln 2 = 0.6931.
  3. 3.The second boost has k = 0.8, so w2 = 12ln1.80.2 = 12ln 9 = ln 3 = 1.099.

    00.69311.792rapidity0.6931 = ln 21.099 = ln 3v/c = tanh w, so w = artanh(v/c)artanh 0.6 = half ln 4 = ln 2 = 0.6931artanh 0.8 = half ln 9 = ln 3 = 1.099
    00.69311.792rapidity0.6931 = ln 21.099 = ln 3v/c = tanh w, so w = artanh(v/c)artanh 0.6 = half ln 4 = ln 2 = 0.6931artanh 0.8 = half ln 9 = ln 3 = 1.099
    The second boost has rapidity 12ln 9 = ln 3 = 1.099.
  4. 4.(a) Rapidities add, so the total is w = ln 2 + ln 3 = ln 6 = 1.792.

    00.69311.792rapidity0.6931 = ln 21.099 = ln 31.792 = ln 6v/c = tanh w, so w = artanh(v/c)artanh 0.6 = half ln 4 = ln 2 = 0.6931artanh 0.8 = half ln 9 = ln 3 = 1.099(a) 0.6931 + 1.099 = 1.792 = ln 6
    00.69311.792rapidity0.6931 = ln 21.099 = ln 31.792 = ln 6v/c = tanh w, so w = artanh(v/c)artanh 0.6 = half ln 4 = ln 2 = 0.6931artanh 0.8 = half ln 9 = ln 3 = 1.099(a) 0.6931 + 1.099 = 1.792 = ln 6
    (a) Laid end to end the two rapidities make ln 2 + ln 3 = ln 6 = 1.792.
  5. 5.(b) Turn the total back into a speed. With ew = 6, tanh w = 6 − 166 + 16 = 3537 = 0.946, so the particle ends at 3537c, about 0.946c. Check against the usual addition law: 0.6 + 0.81 + 0.6 × 0.8 = 1.41.48 = 3537, the same figure.

    00.69311.792rapidity0.6931 = ln 21.099 = ln 31.792 = ln 60.946cv/c = tanh w, so w = artanh(v/c)artanh 0.6 = half ln 4 = ln 2 = 0.6931artanh 0.8 = half ln 9 = ln 3 = 1.099(a) 0.6931 + 1.099 = 1.792 = ln 6(b) tanh(ln 6) = 35/37, a speed of 0.946c
    00.69311.792rapidity0.6931 = ln 21.099 = ln 31.792 = ln 60.946cv/c = tanh w, so w = artanh(v/c)artanh 0.6 = half ln 4 = ln 2 = 0.6931artanh 0.8 = half ln 9 = ln 3 = 1.099(a) 0.6931 + 1.099 = 1.792 = ln 6(b) tanh(ln 6) = 35/37, a speed of 0.946c
    (b) tanh(ln 6) = 3537 = 0.946, so the particle ends at 0.946c.

Answer: (a) ln 2 = 0.6931 and ln 3 = 1.099, which add to ln 6 = 1.792; (b) 3537c, which is 0.946c

Common mistakes

  • Adding the speeds: 0.6c + 0.8c = 1.4c. No speed in this model can pass c, because tanh w lies between −1 and 1 for every w. It is the rapidities that add, and they add to 1.792, whose tanh is 0.946.
  • Adding the rapidities and then forgetting the last step, reporting 1.792 as the speed. A rapidity is not a speed and has no units of speed; the speed is ctanh of it, which is 0.946c.

More hyperbolic functions problems, worked step by step →

Practice Inverse Hyperbolics as Logarithms in the app