A wxMaxima Guide for Calculus Students

This guide follows A Maxima Guide for Calculus Students very closely.  Details are at the end of this document. I thank Professor Moses Glasner for allowing me to use the material that he had developed. I am responsible for errors or shortcomings where the material below deviates from Professor Glasner's guide. Please contact me (wmixon at berry dot edu) with corrections or suggestions for improvement.

By default Maxima comes with its own graphical user interface, Xmaxima. People familiar with Emacs may be more comfortable using Maxima under the Emacs maxima-mode, which provides a powerful environment for combining computations with the LaTeX composition of a document. For very brief usage not requiring extensive interaction, Maxima can be run from a console, such as Xterm under Linux, or DOS prompt under Windows.

Maxima can be run in two other more elaborate environments: TeXmacs and wxMaxima. The former gives a typeset appearance to Maxima's output and provides automatic generation of LaTeX source whereas the latter provides a menu driven environment for using Maxima. This tutorial employs wxMaxima, which can be used to  export both html and LaTeX files (though the exported files will likely require editing before  being ready for publication).

We assume that you have loaded Maxima onto your computer and can run it from wxMaxima. Be aware that the F1 key or the Help menu can be used to open the entire Maxima manual. Maxima also provides help from the command line in response to the commands describe(something) or example(something). Maxima may ask additional questions before responding to  these help requests and the response should be ended with a semicolon ; and then pressing the ENTER key.  If you are not comfortable with wxMaxima, see the Help option at wxmaxima.sourceforge.net/. In particular, work through "10 Minute (wx)Maxima tutorial" (expect to spend more than 10 minutes on this).

When Maxima is ready to process an instruction it displays a %i followed by an integer on the screen: for example, (%i1). The instruction given to Maxima is typed in at this point. It must end with a semicolon or a dollar sign followed by a press on the ENTER key.

Maxima is case sensitive. The cell below contains three Maxima commands, each ending with a semicolon.  The first command yields the expected result. In the other two an improper capitalization precludes computation. The second command contains an unknown command Sin. The third command contains the command to  return the sine of an unknown quantity (%Pi/2). When Maxima cannot interpret and command, it simply repeats  that command as output.

Note that %pi is Maxima's way of rendering the transcendental number with that name. Entering pi in wxMaxima results in the Greek letter of that name. Enter the command sin(pi/2); and confirm that Maxima does not evaluate it.

Ending a command with a dollar sign results in the command being carried out without printing the results. In the cell below, the values 1 and 2 are bound to the names a and b. Both commands end with the dollar sign so the  results are not printed. The third command, which spreads over two lines, ends with a semicolon, so its result is  printed. Maxima ignores spaces and line breaks.

One might interpret the two expressions above as equivalent to a = 1 and b = 2. Maxima does not. The equal sign is reserved for expressions that are temporarily true, but the statement of equality is not retained indefinitely. Consider the two expressions below.

The equality below is expressed with the equal sign. The solve command is applied to the expression, using the % symbol to refer to the most recently generated output. Maxima no longer retains any memory of the expression 2*x + 1 = 7.

We can bind the expression to a name, such as expr1. and then use expr1 in place of the expression itself. The command below binds this expression to a name, then binds the solution of the expression to another name, and finally evaluates the expression given the solution.

A very simple use of Maxima is as an infinite precision scientific calculator. To illustrate this use, consider the following, where bfloat invokes Maxima's bigfloat option, and fpprintprec:60 sets the floating point print precision.

When Maxima completes a calculation all labels remain in use until either Maxima is instructed to free them or the user quits Maxima. If the user is not aware of this when starting a new calculation unexpected results can occur. The instruction kill(all) destroys all information, including loaded packages. Freeing labels with this instruction may be wasteful if some of the objects currently attached or some of the loaded packages will be needed in subsequent calculations. The instructions values and functions ask Maxima to display all labels currently attached to expressions or functions. The the following dialogue indicates a procedure for freeing specific labels while keeping the rest intact.

Note the use of f(x) := as a third way to relate an expression to a name. In this case a functional relationship is  established.

In the next cell two values and two functions are defined. One value and one function are killed, leaving the value b and the function g(x).

2 Finding Limits with Maxima

Maxima can be very helpful in checking limits of functions. Before we consider the limits, however, we examine the functional notation more closely. The function f(x) = sin(x)/x is specified and then evaluated for four values of x. Note the use of a list of commands and the corresponding list of output values.

Inspection of f(x) indicates that f(0) is not defined. We may wish, however, to determine the value toward which f(x) tends as x approaches zero. The cell below confirms that f(0) is not defined, and it reports that the limit of f(x) as x approaches 0 is "infinity". This is an unsatisfactory reply in that this value is Maxima's way of indicating a complex infinity. Maxima indicates a positive infinity with inf and a negative infinity with minf. wxMaxima converts these to the standard symbols as we see below.

The difficulty that results in the imprecise "infinity" response above is removed when we tell Maxima whether x approaches 0 from above ("plus") or below ("minus").

2.1 Plotting Despite Asymptotes

Graphing this function presents a challenge, because f(x) grows without limit as x approaches 1. The use of draw's scenario creation and graphing feature below shows how to circumvent  this difficulty. The first graph places no limits on the f(x) range, so the values being graphed over the relevant range cannot be discerned. Forcing draw to consider only f(x) values between -1 and 1 allows for the drawing of the graph that we require. The x and y axes are optional.

Maxima can deal with functions that involve logical expressions. The transpose( matrix( ) ) command is used to generate a table and is not pertinent to the analysis.

Maxima can also be used as a graphing calculator, and it can calculate more than one function. Observe that the two graphs are not plotted over the same range. Note the use of string to make the functional expression the key entry.

2.2 The Limit of the Difference Quotient

Being able to determine the limit of a difference quotient can be quite helpful. The expression that characterizes the change in f(x) per unit change in x is defined below (h is the size of the change in x). The limit of this expression is determined. It is a quite long expression, but one that can be simplified, as shown.  The ratsimp command results in simplification of a rational expression.

The limit of the difference quotient is the derivative of the function. This equality is confirmed for the example below. The next section discusses the notation for the diff (derivative) command.

The graph below shows the function f(x) and its derivative for x from -10 to 10. In general, care should be taken when graphing f(x) and its derivatives, because often the y units are different for the two.

3 Differentiation Rules

The example below shows that Maxima can provide an expression for the derivative of a particular function. It can do more: Maxima can show the rules for differentiation, independent of the specific expression. We ensure that Maxima understands the dependencies in f(x) and g(x) but do not specify the functional forms.

Maxima knows the sum, product, and quotient rules (and many others).1 The quotient rule is  stored in a somewhat different form than the one that is optimal for humans. The command depends(f,x) is required, to tell Maxima that f is a function of x.  It can then deal with its derivative in a symbolic way. If we had not specified these  dependencies, then instruction diff(f,x) would have evoked the response 0 from Maxima because it would have thought that f and x are simply two independent variables.

The cell below contains the same commands as above, but the depends command has  been "commented out." Not knowing about these dependencies, Maxima returns 0's.

3.1 Extracting and Manipulating Results

A variety of instructions control the final form of the answer that Maxima returns. If we wish the answer to be over a common denominator, for example, then the instruction is factor.

To access just parts of an answer provided by Maxima, request that Maxima give you labels with which you can access the individual parts. The instruction for this is pickapart.  For example, if you wish to find where the derivative in the example above is  equal to 0, then you need a label with which to access the numerator. This is achieved with  the following. Note that the additional variable 2 tells Maxima to what depth  the  expression should broken into parts. Use the smallest value that gets the desired label.

Maxima has attached labels (%t5) and (%t6) to the numerator and denominator of the factored form of the answer. (These numbers vary, depending on whether Maxima has produced any such labels before this pickapart command.) To see what happens if the level 2 is replaces by 1, 3, and so forth.

To find the zeros of the numerator, do as follows:

The resulting output, named soln, is a list with two elements. (Actually everything in Maxima is a list because it is written in the computer language LISP that is based on list processing.) You can access the elements that are in the form of equations x = something with the instruction first and then value of this solution can be retrieved with the instruction first(), or with the appropriate subscript, indicated by square brackets.

Each item in soln is an expression. We can use the command rhs (right-hand side) to extract the value. It can be bound to a name.

3.2 Derivatives of Trigonometric Functions

Maxima can be quite helpful in differentiating trigonometric functions. However, a couple of commands specific to trig functions are required in order to instruct Maxima to apply trig identities in simplifications. For example consider the following dialogue:

The instruction trigsimp instructs Maxima to make the obvious simplification using the Pythagorean identity.  The other Maxima instruction is trigreduce which allows using the multiple angle formulas to reduce powers, e. g:

The latter expression might not appear to be simpler than the one we started with. It is invaluable, however, for integration, the inverse process of differentiation.

3.3 The Chain Rule

Before Maxima can apply the chain rule, it must know that the relevant dependencies exist. In the example below, f(x):= defines an explicit relationship between x and f(x). The existence of relationship between x and u is asserted, but the relationship is not defined.

The dialogue above uses the functional notation to define f and uses the instruction depends to inform Maxima that x is a function of u. It did not, however, provide a specific formula for this dependency. We can, however, specify the dependency of x on u as follows:

Alternatively, we can use functional notation for both functions and get Maxima to differentiate their composition. Note that in this case g(u) and not x must be entered as the expression to be differentiated.

Note that diff(f(x), u) results in 0, because Maxima no longer remembers the dependency of x on u.

The instruction kill was needed to remove the relationships that had been set. If a variable has multiple dependencies and only one of them is to be removed, then an instruction like remove([u,x],dependency) can be used.

3.4 Implicit Differentiation; Higher Derivatives

Maxima can easily compute derivatives of implicit functions. Consider the following dialog that instructs Maxima to find dy/dx given the equation x² + y² = 25.

Note the new symbol appearing in the solve instruction above. Normally the first argument of solve is an equation or list of equations and the second is a variable and a list of variables.  Here dy/dx is the second argument. Also, note the single quote in front of diff(y,x). A single quote in front of a symbol tells Maxima not to evaluate the symbol but to treat it as an unknown quantity. For example,

Likewise, then, the instruction solve(deriv_of_eqn, 'diff(y,x)) tells Maxima not try to evaluate the derivative of y with respect to x directly (which it really cannot do anyway) but to regard diff(y,x) as an unknown quantity and solve for it from the differentiated expression, named deriv_of_eqn.

Higher Derivatives. The Maxima instruction to find higher order derivatives is the same as that for finding the first derivative except for a third argument indicating the order. The first command below is equivalent to diff(x^n, x, 1). (Confirm this.) The second in the list of commands calls for the fourth derivative. The third command calls for the n-th derivative. Until n is specified, this expression cannot be evaluated.

Exercise: Confirm that for the expression x⁸ the 8-th derivative equals 8!. Does this  generalize to any value of n? Explain.

3.5 Related Rates

Maxima can help in solving related rates problems. For example consider the problem  of finding the rate of change of the area of a circle given that the rate of rate of change of  the circle's radius is dr/dt = 60 when t = 2 and r = 120:

Maxima must be told which variables are time-dependent with the instruction depends([a list of variables], t). The instruction subst instructs Maxima to substitute a list of equations appearing as the first argument of subst, enclosed in square brackets, into the expression appearing as its second argument.

Maxima's default is to produce exact calculations, not numeric approximations. Thus, da/dt = 14400π. The float(%) command instructs Maxima to provide the floating-point approximation to the exact value. You might want to see how many digits of %pi (Maxima's way of denoting the number as opposed to the Greek letter) Maxima can find by giving the instructions fpprec= 1000$ fpprintprec:1000$ bfloat(%pi);. The commands fpprec and fpprintprec set the floating-point precision and the floating-point print precision. These need not have the same value.

3.6 Linear Approximations and Differentials

We can use the diff operation to express the differential of an equation as a linear function of its arguments. The next cell defines f(x) and a differential equation based on that function. The command diff does not specify a variable with which to take a derivative, so Maxima returns dy = (dy/dx) dx, where dx is denoted del(x).

The expression is picked apart for use below.

To create an equation that produces a line tangent to f(x) at the value x = π/3, we evaluate the term dy/dx at that value. Then we create a linear function with that slope that passes through the point (π/3, f(π/3)).

Maxima has a very powerful built-in tool for finding linear approximations to functions called the Taylor expansion of order 1.

The ellipsis (...) after the output denotes a residual, and the /T/ before the output indicates that the terms reported constitute a truncated representation of a polynomial of higher degree. We use taytorat to convert the Taylor expansion to a rational expression. The second and third commands below are used to express the result in a manner that makes it comparable to the result above.

The graph below shows the linear approximation to the function, along with the function itself. The graph makes obvious one important aspect of the linear approximation: It is best near the value for which the Taylor expansion is computed.

Taylor polynomials come in all degrees. You will work with them in a systematic way later. The last argument in the Maxima instruction Taylor specifies the degree.

3.7 Locating Maxima and Minima

Maxima can be a great help in checking the process of locating critical points of functions, as the following dialogue shows.

Examination of output %o2, the derivative function, shows that df/dx is not defined when x = 0. The graphs below show the behavior of f and df/dx. The derivative grows without bound as x approaches 0 (use the limit command to confirm this), but f is continuous.

Have we found a maximum or a minimum? The condition that df/dx = 0 is a necessary condition for a value of f(x) to be a (local) maximum. It is also a necessary condition for the value of f(x) to be a (local) minimum. To determine which of the two has been discovered requires that we examine the second derivative of f with respect to x. If this second derivative is negative, then the critical value of x yields a (local) maximum value of f(x); if it is positive, a (local) minimum has been found. In our case, the result is a local (and global, the graph suggests) maximum value of f(x).

Inspection of the expression above shows that the second derivative is negative. We can determine its exact value by using subst or ev. Note the alternative ways of converting the exact answer to a floating point value.

3.8 Optimization: Finding Maximum or Minimum Values of Functions Subject to a Constraint 

We use Maxima to solve the following. A rectangular window is surmounted by a semicircle. The rectangle is of clear glass while the semicircle is of tinted glass that transmits only half as much light per unit area as clear glass does. The total perimeter is fixed. Find the proportions of the window that will admit the most light. Neglect the thickness of the frame.

Let r be the radius of the semicircle. Then 2r is the base of the rectangle. Denote the height of the rectangle by h. Let P be the total perimeter of the window. Then P is a fixed constant. We have the following equations connecting r and h and expressing the amount of light L in terms of r and h, assuming that L is one for each unit of area in the rectangular part and L is 1/2 for each unit of area in the semicircular part of the window. We denote the fixed value of P as P0.

Extract the right hand side of the solution and assign it the name h_is.  Substitute this  into the L expression, yielding L as a function of r (recall that P has a constant value P0).

Taking the derivative of L_fcn_r and setting the derivative equal to zero yields an expression for r: r is proportional to P0.

Assign the right hand side of the solution the name r_is and substitute r_is into the expression h_is, yielding the expression h_is_now, which shows that h, like r, is proportional to P.

That that the critical values of both h and r are proportional to P0 implies that they are proportional to each other. The third command shows the ratio of height to the radius.

Finally, we can determine the maximum light, given the window's perimeter.

An alternative approach to solving problem like this one is to differentiate the constraint equation. This is significantly simpler than solving the constraint equation for one of the variables in cases where the constraint equation is complicated or perhaps not even solvable for either of the variables.

In the above problem we differentiate the equation for the perimeter with respect to r. This yields a linear equation in the unknown derivative dh/dr which can easily be solved. Then we differentiate the formula for L with respect to r and this would also involve the  unknown derivative dh/dr which now can be eliminated. And then one could determine the relationship between r and h that yields a critical point. The following dialogue indicates how Maxima can be instructed to carry out this computation.

3.9 Optimization with More than One Variable

Suppose that we wish to construct a box that has a given surface area in a configuration that yields the maximum volume. The Lagrangian approach to solving a problem like this is to create an augmented expression that incorporates the constraint. Expression A in the cell below adds an undetermined multiplier mu (μ) times the constraint to the expression for volume. The constraint is expressed so that the value by which mu (μ) is being multiplied is zero.

Take all possible derivatives of the augmented expression, including a derivative with respect to μ, which is the constraint. The last line is added only to make the output easier to read.

Solve the system of equations for the derivatives with respect to L, W, and H. The value of μ will be implied, as we see below.

Maxima returns two solutions. The first, L = W = H = 0, yields a minimum. In this case, μ must equal zero in order for S0 - S = 0 to hold (S0 > 0 and S = 0). The interpretation is the adding to S does not further decrease V; a minimum has been attained.

The second solution provides the conditions for a maximum V. Here, L = W = H = 4*μ. The example below illustrates this result. Suppose that S = 60 meters². Then 2*(3*L²) = 60, so L = √(10) meters.

The resulting volume is 10^(3/2) meters³, approximately 31.6228 meters³.

The value of μ is sqrt(10)/4, approximately 0.7906. Therefore a change in S from 60 to 61 meters² should increase V by approximately 0.7906 meters³. The results below show that the actual increase is approximately 0.7939 meters³. Confirm that the discrepancy between L0/4 = μ and the actual value decreases as h decreases by using smaller values of h0.

We can confirm that μ has the same units as L, W, and H: μ is the ratio of the change in volume (meters³) to change in surface (meters²). Thus, the unit for μ is meters, the same as for L, W, and H.

The Lagrangian technique can be applied to multiple constraints. Suppose that the height of the box is required to equal one-half the length. We define a new augmented function that adds this constraint to augmented function A. We then generate the set of derivatives and solve for L, W, H, and kappa (κ), the new multiplier.

Again, Maxima produces two solutions. The first shows the conditions for minimizing the volume, and the second for maximizing it. Now the values of L, W, and H are all different from each other.

The condition that κ = 24μ³ warrants attention. The value of κ is proportional to the box's volume (recall that μ is in the same units as L, W, and H).

The value of μ falls from approximately 0.7906 to approximately 0.7454, so less volume is added per unit change in the surface area. We reevaluate L2, W2, H2, and κ2 given the positive value of μ -- soln_mu2[2] -- and then compute the new volume.

The value of κ estimates that a 0.01 unit increase in the ratio (from 0.5 to 0.51) will cause volume to rise by approximately 0.0994 meters³. Replace 0.5 with 0.51 above and confirm that the actual increase is approximately 0.0973.

For S = 60, the maximum area is now approximately 29.8142 meters³, down from approximately 32.4166 meters³ when the relationships among L, W, and H are not constrained.

4 Integration

The syntax of indefinite integration resembles that of differentiation, with integrate replacing diff, except that multiple integration requires multiple calls to integrate. The general expressions below show that Maxima does not report a constant of integration. It is up to the analyst to keep track.

With the same function f (x, y) = x⁴/y, we find the indefinite and definite integrals of the expression with respect to x. The definite integral is over the range x = 0 to x = 5. Then the corresponding integrals of the resulting expression with respect to y, with the definite integral taken over the range y = 1 to y = 2.

The output can also be generated in a single nested command. Setting the limits as floating point numbers invokes a set of statements specifying the substitutions that Maxima made while carrying out the operation. Adding the command ratprint: false as the first line would suppress this output, which is often not useful.

The limit of a Riemann sum, if it exists, defines the definite integral of a function on [a, b]. The actual evaluation of the limit of the Riemann sums associated to a function on an interval is tedious. Maxima can help. Maxima's instruction for summing a series is sum. The sum command takes four arguments: the expression, the index, the beginning value of the index, and the ending value of the index.

As before, the print command is not required, but it makes the result easier to read.

Maxima's default is not to carry out the evaluation of the sum of a series. The instruction to change this is simpsum : true.

The result of the corresponding definite integral confirms that the integral does equal the limit of the Riemann sum.

4.1 The Fundamental Theorem of Calculus

That Maxima knows the Fundamental Theorem is illustrated below. The extensions, labeled "Chain rule illustrations," show the flexibility of the lambda function. Note the use of the single-quote operator to suppress evaluation.

Without the command assume(x > 0) Maxima will ask the user to declare the sign of x.

The name of the dummy variable t in the definition of S cannot be chosen to be x even though that is perfectly acceptable in mathematical communications.

4.2 The Commands desolve and ode2

The Maxima instruction integrate serves either of two purposes depending on the number of arguments that are passed to it. It is used for telling Maxima to find either the indefinite or definite integral. If two arguments are passed then Maxima finds the indefinite integral of the function in the first argument with respect to the variable in the second. If four arguments are passed, then Maxima finds the definite integral over the interval starting at the third and ending at the fourth.

Another instruction that can be used to achieve the same purpose is ode2. It can be used to instruct Maxima to solve the equation dy/dx = f (x), or dy = f(x) dx for y. This is a very powerful instruction but in this case using a sledge hammer to drive a tack does no harm. The following indicates how this is done. The ode2 command follows the familiar integrate command.

Whenever Maxima is handed an expression it automatically tries to evaluate it. Therefore the quote before the diff tells Maxima not to waste time trying to evaluate dy/dx. Maxima is being told what this derivative is and the next instruction will ask it to solve for y. Note that when you use ode2 Maxima includes the arbitrary constant %c in its response.

Here ode stands for ordinary differential equation. From the Maxima Manual: "The function ode2 solves an ordinary differential equation (ODE) of first or second order." Another Maxima command, desolve, can be used to solve systems of one or more ordinary differential equations.

4.3 Intersections and Areas Between Curves

Maxima can be very helpful in finding the points of intersection of curves. The cell below demonstrates how.

Maxima does not integrate absolute value. Thus to get Maxima to do a problem where neither graph remains above the other on the entire interval, we first find the points of intersection and then instruct Maxima to sum the integrals of the upper function minus the lower function over the relevant interval(s).

The draw2d command below has much more detail than is required to solve this problem, but such detail can be useful. The Maxima Manual (Chapter 48) contains more information regarding options for the draw command.

Use solve to determine the range for which f2 > f1.

The area to be determined is the integral of f2 - f1 over this interval.

4.4 Volumes by Slicing: Using Trigsimp

Maxima cannot help with finding the formula for the area of the cross-section of a typical slice A(x) but it does help in checking the ensuing integration.

For example, finding the volume generated by revolving the area below y = sec(x), above y = 1 and between x = 1 and x = -1 requires evaluating the integral of the expression V = sec²(x) - 1 over the indicated range.

The command trigsimp (trigonometric simplification) is used for trigonometric expressions as ratsimp is used for rational expressions.

As the output indicates, Maxima could have found this integral without the instruction trigsimp, but one should be aware that it is available. This instruction causes Maxima to use the Pythagorean identities to simplify an expression. An instruction like ratsimp would have no effect here. The units in Maxima are radians, so tan(1) is the tangent when x = 1 radian, or π/180 degrees.

The note "Principal Value" indicates how Maxima approaches this problem. "Principal Value" refers to the Cauchy principal value, which is a finite integral of a function about a point c, within the range of integration. It is the limit of the function shown below as epsilon approaches zero from above.

4.5 Volumes by Cylindrical Shells; Using Trigreduce and Trigexpand

Let find the volume of the solid generated by revolving the region between y = cos x, y = 0  and the lines x = 0, x = π/2 about the x-axis. Using the shell method, the integral to be calculated is as below:

This is a difficult integral that requires integration by parts. The disk method yields an integral that can be evaluated:

If you are working problems like this, you will need to recall double/half angle formulas. Maxima can help.

This dialogue shows that trigreduce tells Maxima to use the half-angle formulas to simplify an expression and trigexpand tells Maxima to use the double-angle formulas.

5 Some Numerical Methods in Maxima

In many cases we cannot find an analytical solution to a problem (one might not exist). Maxima offers a range of numerical methods for finding approximate values in such cases.

5.1 Solving Equations

Finding solutions for equations or systems of equations is an important application of numerical analysis. Consider the equation below, for which
solve does not return a useful analytical solution.

When using numerical methods to find roots, one should graph the expression(s) to be solved. The graph can show whether multiple roots are to be expected over the relevant range. It can also guide the selection of guesses that serve as initial values for iterative search methods.

The graph below indicates that a roots occurs around x = 0 and for around x = 40. The function is not defined for x = 0. Why?

The find_root command is used first. This command requires the expression to be evaluated, the expression's argument, and the interval endpoints. If the sign of f(x) is the same at both endpoints, an error message results.

Shorten the interval to [.0001, 10]. (Why not [0, 10]?) The first root is x = 1.2750, approximately. The resulting f(x, a0, b0) is nearly zero. The discrepancy reflects  error in the search process.

The second root -- x = 1.2749, approximately -- is found the same way. Again, the resulting f(x, a0, b0) is approximately 0; 0.0 is a floating-point approximation of 0.

An alternative approach is to use the Newton search process. Doing so requires loading the module newton1. This is the module for applying the Newton process to a single equation. The newton command differs slightly from find_root. It requires the expression, the expression's argument, an initial guess of the value, and a tolerance level. The process works better if the guess is near to the root, and this is where the graph provides guidance.

Newton's method can be extended to multiple equations. As long as the number of equations (and unknowns) is less than 4, graphing can still be used to guide the search process. Consider the two equations below. To avoid three dimensional graphs, we use the implicit command within draw2d. The graph shows x, y pairs for which f(x, y) = 0 and for which g(x, y) = 0. Because f(0, y) is not defined, the implicit function is discontinuous at x = 0.

The graph shows that f(x, y) = g(x, y) = 0 at two values of x: just over -1.0 , and just over +1.0.

To determine the roots for multiple expressions, we load the mnewton module and enter the mnewton command. This command's arguments consist of three lists: a list of expressions, a list of variables, and a list of guesses. The tolerance level must be set explicitly if one wishes to use a value other than the default.

Because the graph identifies two x, y pairs that are roots for both equations, we execute mnewton twice.

We extract the critical values above and check the function values. The solutions are expressed as embedded lists, so extracting these values requires double indices (even though the top-level lists contains only one item).

5.2 Recursive Functions

It may be a good exercise to define your own Maxima function which implements Newton's method. Doing so shows a way to deal with recursive functions in Maxima.

This definition of my_newt is recursive. That is, the definition of the function calls the function itself until the desired precision is attained. From the viewpoint of program clarity this is the method of choice. When calling my_newt, a lambda definition of function must be used because the name of a function f is being passed to my_newt.

5.3 Numerical Integration

Maxima provides a set of options for conducting numerical integration. To illustrate two of these, we use the function below. It can be integrated, so the numerical approach need not be taken, but we use it as a basis for comparison.

A relatively simple method to implement is that of Romberg. Maxima's romberg command has the same syntax as the integrate command. For the  function above, romberg cannot evaluate the integral over the entire range because it cannot deal with f(0). We use f(1e-100) instead. The romberg command can handle values as small as f(1e-320) or so. For this function the romberg command yields a solution with errors after the fourth decimal place. Confirm that moving the beginning value closer to zero by using 1e-300 does not appreciably change the result.

A more sophisticated method (set of methods, actually) is available in QUADPACK. We consider one example, using the command quad_qag. This command contains at least five arguments. (The Maxima Manual describes other, optional arguments.) The first four are the same as those of integrate and romberg. The fifth argument is a key that determines details of the search method and depends on the user's knowledge of the function's characteristics.

In this example quad_qag provides a closer approximation to the correct value than does romberg. The other three pieces of output are these: the estimated error of the method (quite small here), the number of iterations taken, and an error code (0 means no errors).

The Romberg technique can be applied to multidimensional integration, as in this example. The assume(x > 0) command is entered to avoid dealing with a dialog. Remove this line and determine the values once more.

5.4 Numerical Solution of ODEs

Some ordinary differential equations (ODEs), or systems of ODEs, cannot be solved analytically. Numerical solutions of these systems can provide insights that are otherwise not attainable. A classic example is the Lorenz equations that were used to represent the convective motion of a fluid cell that is being warmed from below and cooled from above. This same set of equations can be applied to the analysis of dynamos and laser light. Numerous books analyze this system's behavior, most famously James Gleck's Chaos.

The equations that comprise this system are named dxdt (for dx/dt), dydt, and dzdt. The variable t is time. The parameters a, b, and c are all positive: a is called the Prandtl number, b is called the Rayleigh number, and c is a factor of proportionality.

The values of x, y, and z are not coordinates in physical space. Rather the value of x is proportional to the intensity of convective motion, and y is proportional to the temperature difference between ascending and descending currents. Finally, z is proportional to the distortion of the vertical temperature profile from linearity. Of course, x, y, and z have quite different interpretations in other applications of this set of equations.

The cell below shows how Maxima uses the Runge-Kutte (rk) method to provide numerical solutions for the values of x, y, and z at points in time. The results of 1000 rk solutions are placed into a list named data. The rk command has these arguments: the name(s) of the expression(s), the name(s) or the variable(s), the initial value(s) or the variable(s), and the domain over which the solution is to occur. The domain itself has four elements: the name of the independent variable (time, here), that variable's initial value, its final value, and the increment. In the example below time ranges from t = 0 to t = 100, with increments of 0.01, so Maxima produces a list of 10,000 sets of values.

The first, second, 9999th, and 10000th lists of values appear as output. Each of these four items shows the time, the x value, the y value, and the z value.

Below, we create a list of the 10,000 values of variables x, y, and z and show their interactions. The graph shows that the system does not tend toward any single point, rather it oscillates around two points. This points are called strange attractors. The y axis is labeled; the positions of the x and z labels can be deduced. The graph is produced with the points option.

Suggested exercise: change one of the initial values in the rk command and trace the effects of that change. Make the change small. For example, let the new initial value of x be 15.1 or 14.9 rather than 15. 

It can be instructive to remove the wx from the draw command and execute the command again. This produces a gnuplot window that lets you rotate the graph. This window must be closed before wxMaxima can proceed to the next command.

A Maxima Guide for Calculus Students
c January 2, 2007 by Moses Glasner
e-mail: glasner AT math DOT psu DOT edu
Address:Department of Mathematics
Pennsylvania State University
University Park, PA 16802


Created with wxMaxima.
Edited with Kompozer.
Converted to PDF with Winnovative HTML to PDF Converter.