Product of 1F1 Functions: Formulas, Identities, and Applications Explained
Confluent hypergeometric functions, commonly written as 1F1(a; b; z) and often typed as 1f1, are also called Kummer functions. A single 1F1 function already encodes a rich family of behaviors, from exponential growth to oscillatory decay, depending on its parameters. When two of these functions are multiplied, the resulting expression, the product of 1F1 functions, becomes even more expressive and more delicate to handle. Understanding that product means understanding how two hypergeometric series, two parameter sets, and possibly two different arguments interact.
The product of 1F1 functions appears in radial integrals for Coulomb problems, transition amplitudes in quantum optics, moment generating functions in probability, and symbolic reductions in computer algebra. It is rarely enough to simply multiply two numerical values. Analysts usually need a series representation, an integral identity, a recurrence relation, or an asymptotic expansion that reveals the structure of the product. The goal of this article is to map those routes and to show where the product of 1F1 functions simplifies, where it remains genuinely multivariate, and how to compute it reliably.
Understanding the Product of Two 1F1 Functions
The confluent hypergeometric function is defined by the power series 1F1(a; b; z) = sum from n = 0 to infinity of (a)_n / (b)_n times z^n / n!, where (a)_n is the Pochhammer symbol. It is a solution of Kummer's differential equation and is closely related to Laguerre polynomials, Bessel functions, and the incomplete gamma function. Because it has one upper parameter and one lower parameter, its product with another 1F1 function involves two upper parameters and two lower parameters. That doubling is the source of both the product's flexibility and its complexity.
If the two functions share the same argument z, their product can be written as a convolution of two coefficient sequences. If the arguments differ, say z and w, the product becomes a double series in z and w. In that case, it belongs naturally to the family of Kampé de Fériet functions or bivariate hypergeometric series. When the arguments coincide, some of that bivariate structure collapses, but the product still does not generally reduce to a single 1F1 function. Special parameter choices can produce dramatic simplifications, including exponential, polynomial, or Bessel-type forms.
Series Expansion and Coefficients
For the product 1F1(a; b; z) times 1F1(c; d; z), the coefficient of z^n is a finite sum over k from 0 to n: (a)_k (c)_{n-k} divided by (b)_k (d)_{n-k} k! (n-k)!. This expression is the convolution of the two original coefficient sequences. It shows explicitly how the parameters a, b, c, and d are mixed at each order. For small or moderate n, this sum is often the most practical way to evaluate the product, especially when the individual 1F1 functions are not suffering from cancellation.
The convolution coefficient can often be rewritten using a terminating 3F2 function at unit argument. That transformation is valuable because it connects the product of 1F1 functions to well-studied generalized hypergeometric identities, including Saalschütz-type and Watson-type theorems. When one of the upper parameters is a negative integer, the corresponding 1F1 function becomes a polynomial, and the product coefficient often reduces to a finite combinatorial sum. These reductions are the basis of many exact formulas in quantum mechanics and combinatorial enumeration.
Integral Representations and Multiplication
One of the most useful representations of 1F1(a; b; z) is the integral 1F1(a; b; z) = Gamma(b) / (Gamma(a) Gamma(b - a)) times the integral from 0 to 1 of e^{z t} t^{a - 1} (1 - t)^{b - a - 1} dt, valid for Re b > Re a > 0. Multiplying two such functions turns the product into a double integral over the unit square. The exponential factors combine into e^{z(t + u)} when the arguments are the same, which can then be manipulated by changing variables or by expanding the exponential.
These integral representations are especially effective for deriving product formulas with Laguerre polynomials and Bessel functions. They also provide stable numerical algorithms when the series expansion converges slowly or suffers from severe cancellation. In some cases, the double integral can be reduced to a single integral by integrating out one Beta-distributed variable. The result is a compact formula that reveals the product as a weighted average of simpler exponentials, a viewpoint that is often more intuitive than the raw series.
Applications and Computational Strategies
The product of 1F1 functions is not a purely abstract curiosity. It appears whenever two independent hypergeometric processes are combined, whether they describe electron wavefunctions, photon statistics, or stochastic transition rates. In physics, products of Kummer functions arise in Coulomb scattering, hydrogen-like atoms, and strong-field laser problems. In statistics, they appear in the normalization constants of noncentral distributions and in Bayesian posterior calculations where confluent hypergeometric priors are used.
Computationally, the product can be treated in three broad ways: direct multiplication of two accurately evaluated 1F1 values, expansion in a multivariate or convolution series, or reduction to a differential equation and recurrence. The right choice depends on the parameter regime, the magnitude of the arguments, and the required precision. A method that works beautifully for small real arguments may fail completely for large complex arguments, so robust implementations usually combine several strategies with error estimates.
Physics and Quantum Mechanics
In nonrelativistic quantum mechanics, the regular Coulomb wavefunction is proportional to e^{-i rho} rho^{L + 1} 1F1(L + 1 - i eta; 2L + 2; 2i rho). Products of such functions appear in matrix elements of multipole operators, scattering phase shifts, and photoionization cross sections. The product carries information about the overlap of two continuum states, and its analytic structure determines selection rules and threshold behavior. Similar structures appear in the radial wavefunctions of the hydrogen atom, where Laguerre polynomials can be written as 1F1 functions with negative integer upper parameters.
Quantum optics provides another natural setting. The Jaynes-Cummings model and its generalizations involve overlaps of displaced number states, which are expressed through products of confluent hypergeometric functions. When the field is intense, these products can be approximated by asymptotic formulas; when the photon number is small, the polynomial reduction is exact. This dual behavior is a recurring theme in the product of 1F1 functions: the same expression supports both exact finite sums and smooth asymptotic limits.
Numerical Computation and Symbolic Simplification
Numerically, the most direct approach is to compute each 1F1 function with a reliable library and multiply the results. This works well when both values are moderate in magnitude and the parameters are not near singularities. For large arguments, however, 1F1 functions can grow like e^z or decay like z^{-a}, and direct multiplication may overflow or lose precision. In those regimes, it is better to use scaled versions, logarithmic gamma functions, and asymptotic expansions that combine the dominant exponentials before multiplying.
Symbolic simplification is equally important. Kummer's transformation, 1F1(a; b; z) = e^z 1F1(b - a; b; -z), can move exponential growth from the series to a prefactor, often making the product easier to analyze. When a or c is a negative integer, the product becomes a polynomial, and recurrence relations in the parameters can generate exact coefficients efficiently. Computer algebra systems such as Mathematica, Maple, and SymPy can evaluate many such products, while specialized packages for holonomic functions can derive recurrence equations and creative telescoping identities automatically.
Whether you are evaluating a radial integral, building a symbolic pipeline, or approximating a transition probability, the product of 1F1 functions rewards a structured approach. Start by checking for polynomial or exponential reductions, then choose between convolution series, integral representations, and asymptotic methods based on the parameter and argument ranges. With the right combination of identities and numerics, these products become not just manageable but genuinely illuminating, opening a path to exact results and stable approximations in the next problem you encounter.