Experimental Error 3

Experimental Error

[Courtesy 3M Company, St. Paul, MN.]

Some laboratory errors are more obvious than others, but there is error associated with
every measurement. There is no way to measure the “true value” of anything. The best
we can do in a chemical analysis is to carefully apply a technique that experience tells
us is reliable. Repetition of one method of measurement several times tells us the preci-
sion (reproducibility) of the measurement. If the results of measuring the same quantity
by different methods agree with each other, then we become confident that the results
are accurate, which means they are near the “true” value.

S uppose that you determine the density of a mineral by measuring its mass
(4.635  0.002 g) and volume (1.13  0.05 mL). Density is mass per unit volume:
4.635 g / 1.13 mL  4.101 8 g/mL. The uncertainties in measured mass and volume are

45
46 0.002 g and 0.05 mL, but what is the uncertainty in the computed density? And how
many significant figures should be used for the density? This chapter discusses the
3 Experimental Error propagation of uncertainty in lab calculations.

3-1 Significant Figures

Significant figures: minimum number The number of significant figures is the minimum number of digits needed to write a
of digits required to express a value in given value in scientific notation without loss of accuracy. The number 142.7 has four
scientific notation without loss of significant figures, because it can be written 1.427  102. If you write 1.427 0  102,
accuracy you imply that you know the value of the digit after 7, which is not the case for the
number 142.7. The number 1.427 0  102 has five significant figures.
The number 6.302  106 has four significant figures, because all four digits are
necessary. You could write the same number as 0.000 006 302, which also has just four
significant figures. The zeros to the left of the 6 are merely holding decimal places. The
number 92 500 is ambiguous. It could mean any of the following:
9.25  104 3 significant figures
9.250  104 4 significant figures
9.250 0  104 5 significant figures
You should write one of the three numbers above, instead of 92 500, to indicate how
many figures are actually known.
Significant zeros below are bold: Zeros are significant when they occur (1) in the middle of a number or (2) at the
106 0.010 6 0.106 0.106 0 end of a number on the right-hand side of a decimal point.
The last significant digit (farthest to the right) in a measured quantity always has
some associated uncertainty. The minimum uncertainty is 1 in the last digit. The scale
of a Spectronic 20 spectrophotometer is drawn in Figure 3-1. The needle in the figure
appears to be at an absorbance value of 0.234. We say that this number has three signif-
icant figures because the numbers 2 and 3 are completely certain and the number 4 is an
estimate. The value might be read 0.233 or 0.235 by other people. The percent trans-
mittance is near 58.3. Because the transmittance scale is smaller than the absorbance
scale at this point, there is more uncertainty in the last digit of transmittance. A reason-
able estimate of uncertainty might be 58.3  0.2. There are three significant figures in
the number 58.3.
Interpolation: Estimate all readings to When reading the scale of any apparatus, try to estimate to the nearest tenth of
the nearest tenth of the distance a division. On a 50-mL buret, which is graduated to 0.1 mL, read the level to the
between scale divisions. nearest 0.01 mL. For a ruler calibrated in millimeters, estimate distances to the near-
est 0.1 mm.
There is uncertainty in any measured quantity, even if the measuring instrument
has a digital readout that does not fluctuate. When a digital pH meter indicates a pH of
3.51, there is uncertainty in the digit 1 (and maybe even in the digit 5). By contrast,
some numbers are exact—with an infinite number of unwritten significant digits. To
calculate the average height of four people, you would divide the sum of heights (which
is a measured quantity with some uncertainty) by the integer 4. There are exactly 4 peo-
ple, not 4.000  0.002 people!

Percent transmittance

0 10 20 30 40 50 60 70 80 90 100

Figure 3-1 Scale of a Bausch and

Lomb Spectronic 20 spectrophotometer.
Percent transmittance is a linear scale • 2 1.0 0.5 0.4 0.3 0.2 0.1 0.05 0
and absorbance is a logarithmic scale. Absorbance
3-2 Significant Figures in Arithmetic 47
We now consider how many digits to retain in the answer after you have performed 3-2 Significant Figures in Arithmetic
arithmetic operations with your data. Rounding should only be done on the final answer
(not intermediate results), to avoid accumulating round-off errors.

Addition and Subtraction

If the numbers to be added or subtracted have equal numbers of digits, the answer goes
to the same decimal place as in any of the individual numbers:
1.362  104
 3.111  104
4.473  104
The number of significant figures in the answer may exceed or be less than that in the
original data.
5.345 7.26  1014
 6.728 6.69  1014
12.073 0.57  1014
If the numbers being added do not have the same number of significant figures, we Inspect the legend of the periodic table
are limited by the least certain one. For example, the molecular mass of KrF2 is known inside the cover of this book. Be sure
only to the second decimal place, because we only know the atomic mass of Kr to two you can interpret uncertainties in atomic
masses. For F and Kr, the atomic
decimal places:
masses are
18.998 403 2 (F) F: 18.998 403 2  0.000 000 5
 18.998 403 2 (F) Kr: 83.80  0.01

 83.80 (Kr)
121.796 806 4
123
Not significant

The number 121.796 806 4 should be rounded to 121.80 as the final answer.
When rounding off, look at all the digits beyond the last place desired. In the pre- Rules for rounding off numbers
ceding example, the digits 6 806 4 lie beyond the last significant decimal place.
Because this number is more than halfway to the next higher digit, we round the 9 up to
10 (i.e., we round up to 121.80 instead of down to 121.79). If the insignificant figures
were less than halfway, we would round down. For example, 121.794 8 is correctly
rounded to 121.79.
In the special case where the number is exactly halfway, round to the nearest even
digit. Thus, 43.550 00 is rounded to 43.6, if we can only have three significant figures.
If we are retaining only three figures, 1.425  109 becomes 1.42  109. The number
1.425 01  109 would become 1.43  109, because 501 is more than halfway to the
next digit. The rationale for rounding to an even digit is to avoid systematically increas-
ing or decreasing results through successive round-off errors. Half the round-offs will
be up and half down.
In adding or subtracting numbers expressed in scientific notation, all numbers Addition and subtraction: Express all
should first be expressed with the same exponent: numbers with the same exponent and
align all numbers with respect to the
1.632  105 1.632  105 decimal point. Round off the answer
 4.107  103 1  0.041 07  105 according to the number of decimal
places in the number with the fewest
 0.984  106  9.84  105 decimal places.
11.51  105
The sum 11.513 07  105 is rounded to 11.51  105 because the number 9.84  105
limits us to two decimal places when all numbers are expressed as multiples of 105.
48 Multiplication and Division
3 Experimental Error In multiplication and division, we are normally limited to the number of digits con-
tained in the number with the fewest significant figures:
3.26  105 4.317 9  1012 34.60
 1.78  3.6  1019  2.462 87
5.80  105 1.6  106 14.05
The power of 10 has no influence on the number of figures that should be retained.

Logarithms and Antilogarithms

The base 10 logarithm of n is the number a, whose value is such that n  10a.

Logarithm of n: n  10ameans that log n  a (3-1)

For example, 2 is the logarithm of 100 because 100  102. The logarithm of 0.001 is
3 because 0.001  103. To find the logarithm of a number with your calculator,
enter the number and press the log function.
1 1
103    0.001 In Equation 3-1, the number n is said to be the antilogarithm of a. That is, the
103 1 000
antilogarithm of 2 is 100 because 102  100, and the antilogarithm of 3 is 0.001
because 103  0.001. Your calculator has either a 10x key or an antilog key. To find the
antilogarithm of a number, enter it in your calculator and press 10x (or antilog).
A logarithm is composed of a characteristic and a mantissa. The characteristic is
the integer part and the mantissa is the decimal part:
log 339  2.530log (3.39  105 )  4.470
{{ {{
Characteristic Mantissa Characteristic Mantissa
2  0.530  4  0.470

Number of digits in mantissa of log x  The number 339 can be written 3.39  102. The number of digits in the mantissa
number of significant figures in x: of log 339 should equal the number of significant figures in 339. The logarithm of 339
log (5.403  108)  7.267 4 is properly expressed as 2.530. The characteristic, 2, corresponds to the exponent in
123 123
3.39  102.
4 digits 4 digits
To see that the third decimal place is the last significant place, consider the follow-
ing results:
102.531  340 (339.6)
102.530  339 (338.8)
102.529  338 (338.1)
The numbers in parentheses are the results prior to rounding to three figures. Changing
the exponent in the third decimal place changes the answer in the third place of 339.
Number of digits in antilog x ( 10x)  In converting a logarithm to its antilogarithm, the number of significant figures in
number of significant figures in the antilogarithm should equal the number of digits in the mantissa. Thus,
mantissa of x:
{  10 {  3.8
{  10
antilog (3.42) 3.42 4
{  1.39
106.142 {  10
6
2 digits 2 digits 2 digits
3 digits 3 digits
Here are several examples showing the proper use of significant figures:
log 0.001 237  2.907 6 antilog 4.37  2.3  104
log 1 237  3.092 4 104.37  2.3  104
log 3.2  0.51 102.600  2.51  103

3-3 Significant Figures and Graphs

When drawing a graph on a computer, consider whether the graph is meant to display
qualitative behavior of the data (Figure 3-2) or precise values. If someone will use the
graph (such as Figure 3-3) to read points, it should at least have tic marks on both sides
of the horizontal and vertical scales. Better still is a fine grid superimposed on the graph.
3-4 Types of Error 49

0.8 + 0.04

Correction (mL)
0.4 + 0.02

y 0.0 0.00
29.43
mL
–0.4 – 0.02

–0.8 – 0.04
0 5 10 15 20 25 10 20 30 40 50
x Volume delivered (mL)

Figure 3-2 Example of a graph Figure 3-3 Calibration curve for a 50-mL buret. The volume delivered can be read to
intended to show the qualitative the nearest 0.1 mL. If your buret reading is 29.43 mL, you can find the correction factor
behavior of the function y  ex/6 cos x. accurately enough by locating 29.4 mL on the graph. The correction factor on the ordinate
You are not expected to be able to read (y-axis) for 29.4 mL on the abscissa (x-axis) is 0.03 mL (to the nearest 0.01 mL).
coordinates accurately on this graph.

Rulings on a graph should be compatible with the number of significant figures of

the coordinates. The graph in Figure 3-4a has reasonable rulings for the points (0.53,
0.63) and (1.08, 1.47). Rulings correspond to every 0.1 unit, so you can estimate the
position of the 0.01 unit. The graph in Figure 3-4b is the same size but is not ruled fine
enough to estimate the position of the 0.01 unit. Plan the axes of a graph so that data are
spread over as much of the sheet of paper as possible.

1.50 1.50

1.00 1.00

0.50
0.50
Figure 3-4 Graphs demonstrating
choice of rulings in relation to significant
figures in the data. Graph (b) does not
have divisions fine enough to allow you
0.50 1.00 1.50 to plot data that are accurate to the
0.50 1.00 1.50
(a) ( b) hundredths place.

3-4 Types of Error

Every measurement has some uncertainty, which is called experimental error. Conclu-
sions can be expressed with a high or a low degree of confidence, but never with com-
plete certainty. Experimental error is classified as either systematic or random.

Systematic Error
Systematic error, also called determinate error, arises from a flaw in equipment or Systematic error is a consistent error that
the design of an experiment. If you conduct the experiment again in exactly the same can be detected and corrected. Box 3-1
manner, the error is reproducible. In principle, systematic error can be discovered and describes Standard Reference Materials
designed to reduce systematic errors.
corrected, although this may not be easy.
50 3 Experimental Error

Box 3-1 Standard Reference Materials

Inaccurate laboratory measurements can mean wrong med- Before introduction of this reference material, five labo-
ical diagnosis and treatment, lost production time, wasted Retracked to gain
ratories 1 line identical samples reported a range of
analyzing
energy and materials, manufacturing rejects, and product lia- results with relative errors of 40 to 110% of the expected
bility problems. To minimize errors in laboratory measure- value. After distribution of the reference material, the error
ments, the U.S. National Institute of Standards and Technol- was reduced to 20 to 40%.
ogy distributes more than 1 000 Standard Reference Materials,
such as metals, chemicals, rubber, plastics, engineering
materials, radioactive substances, and environmental and
clinical standards that can be used to test the accuracy of 120
Before
analytical procedures used in different laboratories.1
100 After

Error (% of target value)

For example, in treating patients with epilepsy, physi-
cians depend on laboratory tests to measure concentrations of 80
anticonvulsant drugs in blood serum. Drug levels that are too
low lead to seizures; high levels are toxic. Because tests of 60

identical serum specimens at different laboratories were giv- 40

ing an unacceptably wide range of results, the National Insti-
tute of Standards and Technology developed a Standard Ref- 20
erence Material containing known levels of antiepilepsy drugs
0
in serum. The reference material now enables different labo- Phenytoin Phenobarbital Primidone Ethosuximide
ratories to detect and correct errors in their assay procedures. (Mysoline) (Zarontin)

Ways to detect systematic error: For example, using a pH meter that has been standardized incorrectly produces
1. Analyze samples of known composi- a systematic error. Suppose you think that the pH of the buffer used to standardize
tion, such as a Standard Reference the meter is 7.00, but it is really 7.08. Then all your pH readings will be 0.08 pH
Material. Your method should repro- unit too low. When you read a pH of 5.60, the actual pH of the sample is 5.68. This
duce the known answer. (See Box systematic error could be discovered by using a second buffer of known pH to test
15-1 for an example.)
the meter.
2. Analyze “blank” samples containing
none of the analyte being sought. If
Another systematic error arises when you use an uncalibrated buret. The manufac-
you observe a nonzero result, your turer’s tolerance for a Class A 50-mL buret is 0.05 mL. When you think you have
method responds to more than you delivered 29.43 mL, the real volume could be anywhere from 29.38 to 29.48 mL and
intend. Section 29-3 discusses differ- still be within tolerance. One way to correct for an error of this type is to construct a
ent kinds of blanks. calibration curve, such as that in Figure 3-3, by the procedure on page 44. To do this,
3. Use different analytical methods to deliver distilled water from the buret into a flask and weigh it. Determine the volume
measure the same quantity. If the
of water from its mass by using Table 2-7. Figure 3-3 tells us to apply a correction fac-
results do not agree, there is error in
one (or more) of the methods. tor of 0.03 mL to the measured value of 29.43 mL. The actual volume delivered is
4. Round robin experiment: Assign dif- 29.43  0.03  29.40 mL.
ferent people in several laboratories A key feature of systematic error is that it is reproducible. For the buret just dis-
to analyze identical samples by the cussed, the error is always 0.03 mL when the buret reading is 29.43 mL. Systematic
same or different methods. error may always be positive in some regions and always negative in others. With care
Disagreement beyond the estimated and cleverness, you can detect and correct a systematic error.
random error is systematic error.

Random Error
Random error, also called indeterminate error, arises from the effects of uncon-
trolled (and maybe uncontrollable) variables in the measurement. Random error has an
equal chance of being positive or negative. It is always present and cannot be corrected.
Random error cannot be eliminated, but There is random error associated with reading a scale. Different people reading the
it might be reduced by a better scale in Figure 3-1 report a range of values representing their subjective interpolation
experiment. between the markings. One person reading the same instrument several times might
report several different readings. Another indeterminate error results from random elec-
trical noise in an instrument. Positive and negative fluctuations occur with approxi-
51
mately equal frequency and cannot be completely eliminated. 3-5 Propagation of Uncertainty

Precision and Accuracy

Precision describes the reproducibility of a result. If you measure a quantity several Precision: reproducibility
times and the values agree closely with one another, your measurement is precise. If the Accuracy: nearness to the “truth”
values vary widely, your measurement is not very precise. Accuracy describes how
close a measured value is to the “true” value. If a known standard is available (such as
a Standard Reference Material, described in Box 3-1), accuracy is how close your value
is to the known value.
A measurement might be reproducible, but wrong. For example, if you made a
mistake while preparing a solution for a titration, the solution would not have the
desired concentration. You might then do a series of reproducible titrations but report an
incorrect result because the concentration of the titrating solution was not what you
intended. In this case, the precision is good but the accuracy is poor. Conversely, it is
possible to make poorly reproducible measurements clustered around the correct value.
In this case, the precision is poor but the accuracy is good. An ideal procedure is both
precise and accurate.
Accuracy is defined as nearness to the “true” value. The word true is in quotes
because somebody must measure the “true” value, and there is error associated with
every measurement. The “true” value is best obtained by an experienced person using a
well-tested procedure. It is desirable to test the result by using different procedures,
because, even though each method might be precise, systematic error could lead to poor
agreement between methods. Good agreement among several methods affords us confi-
dence, but never proof, that results are accurate.

Absolute and Relative Uncertainty

Absolute uncertainty expresses the margin of uncertainty associated with a measure- An uncertainty of 0.02 means that,
ment. If the estimated uncertainty in reading a calibrated buret is 0.02 mL, we say when the reading is 13.33, the true
that 0.02 mL is the absolute uncertainty associated with the reading. value could be anywhere in the range
Relative uncertainty compares the size of the absolute uncertainty with the size of 13.31 to 13.35.
its associated measurement. The relative uncertainty of a buret reading of 12.35  0.02
mL is a dimensionless quotient:

Relative absolute uncertainty

Relative uncertainty  (3-2)
uncertainty: magnitude of measurement

0.02 mL
  0.002 If you use a 50-mL buret, design your
12.35 mL
titration to require 20–40 mL of reagent
The percent relative uncertainty is simply to produce a small relative uncertainty
of 0.1–0.05%.
Percent In a gravimetric analysis, plan to
relative Percent relative uncertainty  100  relative uncertainty (3-3) have enough precipitate for a low
uncertainty: relative uncertainty. If weighing precision
 100  0.002  0.2% is 0.3 mg, a 100-mg precipitate has a
relative weighing error of 0.3% and a
If the absolute uncertainty in reading a buret is constant at 0.02 mL, the percent rela- 300-mg precipitate has an uncertainty
tive uncertainty is 0.2% for a volume of 10 mL and 0.1% for a volume of 20 mL. of 0.1%.

3-5 Propagation of Uncertainty

We can usually estimate or measure the random error associated with a measurement,
such as the length of an object or the temperature of a solution. The uncertainty might
 be based on how well we can read an instrument or on our experience with a particular
52 method. If possible, uncertainty is expressed as the standard deviation or as a confi-
3 Experimental Error dence interval; these parameters are based on a series of replicate measurements. The
following discussion applies only to random error. We assume that systematic error has
Standard deviation and confidence been detected and corrected.
interval are discussed in the next In most experiments, it is necessary to perform arithmetic operations on several
chapter. numbers, each of which has an associated random error. The most likely uncertainty in
the result is not simply the sum of the individual errors, because some of these are
likely to be positive and some negative. We expect some cancellation of errors.

Addition and Subtraction

Suppose you wish to perform the following arithmetic, in which the experimental
uncertainties, designated e1, e2, and e3, are given in parentheses.
1.76 (0.03) d e1
 1.89 (0.02) d e2
(3-4)
 0.59 (0.02) d e3
3.06 (e4 )
The arithmetic answer is 3.06. But what is the uncertainty associated with this result?
For addition and subtraction, the uncertainty in the answer is obtained from the
absolute uncertainties of the individual terms as follows:
For addition and subtraction, use Uncertainty in addition
absolute uncertainty. and subtraction: e4  2e21  e22  e23 (3-5)

For the sum in Equation 3-4, we can write

e4  2 (0.03) 2  (0.02) 2  (0.02) 2  0.041
The absolute uncertainty e4 is 0.04, and we can write the answer as 3.06  0.04.
Although there is only one significant figure in the uncertainty, we wrote it initially as
0.041, with the first insignificant figure subscripted. We retain one or more insignificant
figures to avoid introducing round-off errors into later calculations through the number
0.041. The insignificant figure was subscripted to remind us where the last significant
figure should be at the conclusion of the calculations.
To find the percent relative uncertainty in the sum of Equation 3-4, we write
0.041
Percent relative uncertainty   100  1.3 %
3.06
The uncertainty, 0.041, is 1.3% of the result, 3.06. The subscript 3 in 1.3% is not signif-
icant. It is sensible to drop the insignificant figures now and express the final result as
For addition and subtraction, use 3.06 (0.04) (absolute uncertainty)
absolute uncertainty. Relative 3.06 (1%) (relative uncertainty)
uncertainty can be found at the end of
the calculation.
Example Uncertainty in a Buret Reading
The volume delivered by a buret is the difference between the final reading and the
initial reading. If the uncertainty in each reading is 0.02 mL, what is the uncertainty
in the volume delivered?
SOLUTION Suppose that the initial reading is 0.05 (0.02) mL and the final reading
is 17.88 (0.02) mL. The volume delivered is the difference:
17.88 (0.02)
 0.05 (0.02)
17.83 (e) e  20.022  0.022  0.028  0.03
Regardless of the initial and final readings, if the uncertainty in each one is 0.02 mL, 53
the uncertainty in volume delivered is 0.03 mL.
3-5 Propagation of Uncertainty
Multiplication and Division
For multiplication and division, first convert all uncertainties to percent relative uncer-
tainties. Then calculate the error of the product or quotient as follows:
Uncertainty in multiplication
%e4  2 (%e1 ) 2  (%e2 ) 2  (%e3 ) 2 (3-6) For multiplication and division, use
and division: percent relative uncertainty.
For example, consider the following operations:
1.76 (0.03)  1.89 (0.02)
 5.64  e4
0.59 (0.02)
First convert absolute uncertainties to percent relative uncertainties. Advice Retain one or more extra
insignificant figures until you have
1.76 (1.7 %)  1.89 (1.1 %) finished your entire calculation. Then
 5.64  e4
0.59 (3.4 %) round to the correct number of digits.
When storing intermediate results in a
Then find the percent relative uncertainty of the answer by using Equation 3-6. calculator, keep all digits without
rounding.
%e4  2 (1.7 ) 2  (1.1 ) 2  (3.4 ) 2  4.0 %
The answer is 5.64 (4.0%).
To convert relative uncertainty to absolute uncertainty, find 4.0% of the answer.
4.0%  5.64  0.040  5.64  0.23
The answer is 5.64 (0.23). Finally, drop the insignificant digits.
5.6 (0.2) (absolute uncertainty) For multiplication and division, use
5.6 (4%) (relative uncertainty) percent relative uncertainty. Absolute
uncertainty can be found at the end of
The denominator of the original problem, 0.59, limits the answer to two digits. the calculation.

Mixed Operations
Now consider a computation containing subtraction and division:
1.76 (0.03)  0.59 (0.02)
 0.6190  ?
1.89 (0.02)
First work out the difference in the numerator, using absolute uncertainties. Thus,
1.76 (0.03)  0.59 (0.02)  1.17 (0.036)
because 2 (0.03) 2  (0.02) 2  0.036.
Then convert to percent relative uncertainties. Thus,
1.17 (0.036 ) 1.17 (3.1 %)
  0.6190 (3.3 %)
1.89 (0.02) 1.89 (1.1 %)
because 2 (3.1 % ) 2  (1.1 % ) 2  3.3%.
The percent relative uncertainty is 3.3%, so the absolute uncertainty is 0.033 
0.6190  0.020. The final answer can be written as
0.619 (0.020) (absolute uncertainty)
0.619 (3.3%) (relative uncertainty)
Because the uncertainty begins in the 0.01 decimal place, it is reasonable to round the The result of a calculation ought to be
result to the 0.01 decimal place: written in a manner consistent with its
uncertainty.
0.62 (0.02) (absolute uncertainty)
0.62 (3%) (relative uncertainty)
54 The Real Rule for Significant Figures
3 Experimental Error The first uncertain figure of the answer is the last significant figure. For example, in the
quotient
The real rule: The first uncertain figure 0.002 364 (0.000 003)
is the last significant figure.  0.094 6 (0.000 2)
0.025 00 (0.000 05)
the uncertainty (0.000 2) occurs in the fourth decimal place. Therefore, the answer is
properly expressed with three significant figures, even though the original data have
four figures. The first uncertain figure of the answer is the last significant figure. The
quotient
0.002 664 (0.000 003)
 0.106 6 (0.000 2)
0.025 00 (0.000 05)
is expressed with four significant figures because the uncertainty occurs in the fourth
place. The quotient
0.821 (0.002)
 1.022 (0.004)
0.803 (0.002)
is expressed with four figures even though the dividend and divisor each have three
figures.

Example Significant Figures in Laboratory Work

The rationale for finding the uncertainty You prepared a 0.250 M NH3 solution by diluting 8.45 (0.04) mL of 28.0 (0.5)
in the molecular mass of NH3 is wt % NH3 [density  0.899 (0.003) g/mL] up to 500.0 (0.2) mL. Find the
explained at the end of this chapter: uncertainty in 0.250 M. The molecular mass of NH3, 17.030 6 g/mol, has negligible
N:  14.006 74  0.000 07 uncertainty relative to other uncertainties in this problem.
3H:3(1.007 94  0.000 07)
SOLUTION To find the uncertainty in molarity, we need to find the uncertainty in moles
N: 14.006 74  0.000 07 delivered to the 500-mL flask. The concentrated reagent contains 0.899 (0.003) g
3H:3.023 82  0.000 21 of solution per milliliter. The weight percent tells us that the reagent contains
NH3:  17.030 56  20.000 072  0.000 212 0.280 (0.005) g of NH3 per gram of solution. In the following calculations, we
 17.030 56  0.000 22
retain extra insignificant digits and round off only at the end.
 17.030 6  0.000 2
Grams of NH3 per g solution g NH3
mL in concentrated  0.899 (0.003) mL  0.280 (0.005)
g solution
reagent
g solution g NH3
 0.899 (0.334%)  0.280 (1.79%)
mL g solution
g NH3
 0.251 7 (1.82%)
mL
Convert absolute uncertainty to %
relative uncertainty for multiplication. because 2 (0.334% ) 2  (1.79% ) 2  1.82%.
Next, we find the moles of ammonia contained in 8.45 (0.04) mL of
concentrated reagent. The relative uncertainty in volume is 0.04/8.45  0.473%.
g NH3
0.251 7 (1.82%)  8.45 (0.473%) mL
mL
mol NH3 
g NH3
17.030 6 (0%)
mol
 0.124 9 (1.88%) mol
because 2 (1.82% ) 2  (0.473% ) 2  (0% ) 2  1.88%.
This much ammonia was diluted to 0.500 0 (0.000 2) L. The relative
uncertainty in the final volume is 0.000 2/0.500 0  0.04%. The molarity is
 mol NH3 0.124 9 (1.88%) mol
 55
L 0.500 0 (0.04%) L
3-5 Propagation of Uncertainty
 0.249 8 (1.88%) M

because 2 (1.88% ) 2  (0.04% ) 2  1.88%. The absolute uncertainty is 1.88% of

0.249 8 M  0.004 7 M. The uncertainty in molarity is in the third decimal place, so
our final, rounded answer is

[NH3]  0.250 (0.005) M

Exponents and Logarithms

For the function y  xa, the percent relative uncertainty in y(%ey) is equal to a times the To calculate a power or root on your
percent relative uncertainty in x(%ex): calculator, use the y x button. For
example, to find a cube root (y 1/3), raise
Uncertainty for powers y to the 0.333 333 333 . . . power with
and roots: y  xa1 %ey  a(%ex ) (3-7) the y x button.

For example, if y  2 x  x1/2, a 2% uncertainty in x will yield a 112 2 (2%)  1% uncer-

tainty in y. If y  x2, a 3% uncertainty in x leads to a (2)(3%)  6% uncertainty in y
(Box 3-2).
If y is the base 10 logarithm of x, then the absolute uncertainty in y (ey) is propor-
tional to the relative uncertainty in x (ex /x):

Uncertainty for 1 ex ex
y  log x1 ey   0.434 29 (3-8)
logarithm: ln 10 x x
Use relative uncertainty (ex /x),
You should not work with percent relative uncertainty [100  (ex /x)] in calculations not percent relative uncertainty
with logs and antilogs, because one side of Equation 3-8 has relative uncertainty and [100  (ex /x )], in calculations involving
the other has absolute uncertainty. log x, ln x, 10x, and ex.

Box 3-2 Propagation of Uncertainty in the Product xx

Table 3-1 says that the uncertainty in the function y  xa is other. In the product x  z, the measured value of x could be
%ey  a(%ex). If y  x2, then %ey  2(%ex). A 3% uncer- high sometimes and the measured value of z could be low
tainty in x leads to a (2)(3%)  6% uncertainty in y. sometimes. In the majority of cases, uncertainty in the prod-
But what if we just apply the multiplication formula 3-6 uct x  z is not as great as the uncertainty in x2.
to the product xx? Example. The distance traveled by a falling object in time
t is 12 gt 2, where g is the acceleration of gravity. If t has an uncer-
x(e1)  x (e2)  x2 (e3) tainty of 1%, the uncertainty in t 2 is 2(%et)  2(1%)  2%.
%e3  2 (%e1 ) 2  (%e2 ) 2 The uncertainty in distance computed from 12 gt 2 will also be
 2 (3%) 2  (3%) 2  4.2 % 2%. If you (incorrectly) used Equation 3-6, you would com-
pute an uncertainty in distance of 2 1% 2  1% 2  1.4%.
Which uncertainty is correct, 6% from Table 3-1 or 4.2%
from Equation 3-6?
Table 3-1 (6%) is correct. In the formula y  x2, the
error in a measured value of x is always positive or always
negative. If the true value of x is 1.00 and the measured value
is 1.01, the computed value of x2 is (1.01)2  1.02. That is, if
the measured x is high by 1%, the computed value of x2 is
high by 2% because we are multiplying the high value by the
high value.
Relative Relative Relative Relative
Equation 3-6 presumes that the uncertainty in each fac- uncertainty uncertainty uncertainty uncertainty
tor of the product x  z is random and independent of the in x in z in x · z in x 2
56 Table 3-1 Summary of rules for propagation of uncertainty
3 Experimental Error Function Uncertainty Function Uncertainty

y  x1  x2 ey  2 e2x1  e2x2 y  xa %ey  a %ex

1 ex ex
y  x1  x2 ey  2 e2x1  e2x2 y  log x ey   0.434 29
ln 10 x x
ex
y  x1 x2 %ey  2 %e2x1  %e2x2 y  ln x ey 
x
x1 ey
y %ey  2 %e2x1  %e2x2 y  10x  (ln 10) ex  2.302 6 ex
x2 y
ey
y  ex  ex
y
NOTE: x represents a variable and a represents a constant that has no uncertainty. ex /x is the relative error in x
and %ex is 100  ex /x.

The natural logarithm (ln) of x is the number y, whose value is such that x  ey,
where e ( 2.718 28 . . . ) is called the base of the natural logarithm. The absolute
uncertainty in y is equal to the relative uncertainty in x.
Uncertainty for ex
y  ln x1 ey  (3-9)
natural logarithm: x
Now consider y  antilog x, which is the same as saying y  10x. In this case, the
relative uncertainty in y is proportional to the absolute uncertainty in x.
Uncertainty ey
y  10x1   (ln 10) ex  2.302 6 ex (3-10)
for 10x: y
If y  ex, the relative uncertainty in y equals the absolute uncertainty in x.
Uncertainty ey
y  ex1   ex (3-11)
for e x: y
Appendix C gives a general rule for Table 3-1 summarizes rules for propagation of uncertainty. You need not memorize the
propagation of uncertainty for any rules for exponents, logs, and antilogs, but you should be able to use them.
function.

Example Uncertainty in H Concentration

Consider the function pH  log[H], where [H] is the molarity of H. For
pH  5.21  0.03, find [H] and its uncertainty.
SOLUTION First solve the equation pH  log[H] for [H]: Whenever a  b, then
10a  10b. If pH  log[H], then log[H]  pH and 10log[H]  10pH. But
10log[H]  [H]. We therefore need to find the uncertainty in the equation
[H]  10pH  10(5.21  0.03)
In Table 3-1, the relevant function is y  10x, in which y  [H] and x  (5.21  0.03).
For y  10x, the table tells us that ey /y  2.302 6 ex.
ey
 2.302 6 ex  (2.302 6)(0.03)  0.069 1 (3-12)
y
The relative uncertainty in y ( ey /y) is 0.069 1. Inserting the value y  105.21 
6.17  106 into Equation 3-12 gives the answer:
ey ey
  0.069 11 ey  4.26  107
y 6.17  106
The concentration of H is 6.17 (0.426)  106  6.2 (0.4)  106 M. An 57
uncertainty of 0.03 in pH gives an uncertainty of 7% in H. Notice that extra digits were
retained in the intermediate results and were not rounded off until the final answer. 3-5 Propagation of Uncertainty

Uncertainty in Molecular Mass

What is the uncertainty in the molecular mass of O2? On the inside cover of this book,
we find that the atomic mass of oxygen is 15.999 4  0.000 3 g/mol. Without thinking,
I always used Equation 3-5 for the uncertainty of a sum:
15.999 4 0.000 3
 15.999 4 0.000 3
31.998 8e  20.000 32  0.000 32  0.000 42
14444244443
The rule for sums in Equation 3-5 The wrong way!

Equation 3-5 is appropriate when the errors in each term are random. One might be
positive and one might be negative. In most cases, the uncertainty in the sum is less
than 0.000 3  0.000 3  0.000 6.
One day a student named Ian told me that he did not know the true mass of a mole of
oxygen, but he was quite sure that every mole of oxygen has the same mass. The uncer-
tainty of 0.000 3 means that the true mass is in the range 15.999 1 to 15.999 7. If the
true mass were 15.999 7, then the mass of O2 is 2  15.999 7  31.999 4 g/mol. If the
true mass is 15.999 1, then the mass of O2 is 2  15.999 1  31.998 2 g/mol. The mass
of O2 is somewhere in the range 31.998 8  0.000 6. The uncertainty is not
 2 0.000 32  0.000 32  0.000 42. It is 2  (0.000 3)  0.000 6. The uncer-
tainty of the mass of n atoms is n  (uncertainty of one atom).
Let’s apply this reasoning to find the molecular mass of C2H4:
2C:2(12.010 7  0.000 8)  24.021 4  0.001 6 d 2  0.000 8 Uncertainty in mass of n identical atoms
 n  (uncertainty in atomic mass).
4H:4(1.007 94  0.000 07)  4.031 76  0.000 28 d 4  0.000 07
28.053 16  ? (3-13)
To find the uncertainty in the sum of the masses of 2C  4H, we do use Equation 3-5
because the uncertainties in the masses of C and H are independent of each other. One
might be positive and one might be negative. So the molecular mass of C2H4 is
28.053 16  20.001 62  0.000 282 To find the uncertainty in the sum in
Equation 3-13, we do use Equation 3-5.
28.053 16  0.001 6
28.053  0.002 g/mol

Multiple Deliveries from a Pipet

A 25-mL Class A volumetric pipet delivers 25.00  0.03 mL. The actual volume can
be in the range of 24.97 to 25.03 mL. The precision of delivering this volume is much
better—typically 0.005 mL. If you deliver 100 mL with 4 aliquots of 25 mL,
the uncertainty in the total volume is approximately 4  0.03  0.12 mL, not
 2 0.032  0.032  0.032  0.032  0.06 mL. The reason is the same as the rea-
son why the uncertainty in the mass of 4 H atoms is 4  0.000 07 g/mol. Whatever the
error of one pipet is, it is always in the same direction.
You can greatly improve the accuracy of using a pipet by calibrating it as described By calibrating the pipet, we reduce
in Section 2-9. For example, a particular pipet delivers a mean volume of 24.991 mL with the uncertainty from 0.12 mL to
a standard deviation of 0.006 mL in replicate deliveries. If you used this pipet to deliver 0.012 mL.
4 aliquots, the actual volume delivered would be 4  24.991  99.964 mL. The uncer-
tainty is not 4  0.006  0.024 mL because the delivery error is sometimes positive and
sometimes negative. The uncertainty is  2 0.0062  0.0062  0.0062  0.0062 
0.012 mL.
58 3 Experimental Error

Terms to Understand
absolute uncertainty characteristic logarithm precision significant figure
accuracy determinate error mantissa random error systematic error
antilogarithm indeterminate error natural logarithm relative uncertainty

Summary
The number of significant digits in a number is the minimum (determinate) error affects the accuracy (nearness to the
required to write the number in scientific notation. The first “true” value). Systematic error can be discovered and elimi-
uncertain digit is the last significant figure. In addition and nated by a clever person, but some random error is always

requires absolute uncertainties 1e3  2 e21  e22 2 , whereas

subtraction, the last significant figure is determined by the present. Propagation of uncertainty in addition and subtraction
number with the fewest decimal places (when all exponents

1 %e3  2 %e21  %e22 2 . Other rules for propagation of error

are equal). In multiplication and division, the number of fig- multiplication and division utilize relative uncertainties
ures is usually limited by the factor with the smallest number
of digits. The number of figures in the mantissa of the loga- are found in Table 3-1. Always retain more digits than neces-
rithm of a quantity should equal the number of significant sary during a calculation and round off to the appropriate num-
figures in the quantity. Random (indeterminate) error affects ber of digits at the end.
the precision (reproducibility) of a result, whereas systematic

Exercises
3-A. Write each answer with a reasonable number of figures. 3-C. We have a 37.0 (0.5) wt % HCl solution with a density
Find the absolute and percent relative uncertainty for each of 1.18 (0.01) g/mL. To deliver 0.050 0 mol of HCl requires
answer. 4.18 mL of solution. If the uncertainty that can be tolerated in
(a) [12.41 (0.09)  4.16 (0.01)]  7.068 2 (0.000 4)  ? 0.050 0 mol is 2%, how big can the absolute uncertainty in
(b) [3.26 (0.10)  8.47 (0.05)]  0.18 (0.06)  ? 4.18 mL be? (Caution: In this problem, you have to work
(c) 6.843 (0.008)  104  [2.09 (0.04)  1.63 (0.01)]  ? backward. You would normally compute the uncertainty in
(d) 2 3.24  0.08  ? mol HCl from the uncertainty in volume:
(e) (3.24  0.08)4  ?
g solution g HCl
(f) log(3.24  0.08)  ? mL solution  
(g) 103.24  0.08  ? mL solution g solution
mol HCl 
3-B. (a) You have a bottle labeled “53.4 (0.4) wt % NaOH— g HCl
density  1.52 (0.01) g/mL.” How many milliliters of mol HCl
53.4 wt % NaOH will you need to prepare 2.000 L of
0.169 M NaOH? But here we know the uncertainty in mol HCl (2%) and we need
(b) If the uncertainty in delivering NaOH is 0.01 mL, to find what uncertainty in mL solution leads to that 2% uncer-
calculate the absolute uncertainty in the molarity (0.169 M). tainty. The arithmetic has the form a  b  c  d, for which
Assume there is negligible uncertainty in the formula mass %e2a  %e2b  %e2c  %e2d. If we know %ea, %ec, and %ed, we
of NaOH and in the final volume (2.000 L). can find %eb by subtraction: %e2b  %e2a  %e2c  %e2d.)

Problems
Significant Figures 3-2. Round each number as indicated:
3-1. How many significant figures are there in the following (a) 1.236 7 to 4 significant figures
numbers? (b) 1.238 4 to 4 significant figures
(a) 1.903 0 (c) 0.135 2 to 3 significant figures
(b) 0.039 10 (d) 2.051 to 2 significant figures
(c) 1.40  104 (e) 2.005 0 to 3 significant figures
 Problems 59
Reading Digit 6 on lower scale aligned
(1.46 mm) with marking on upper scale
0 1 2 3 4 5
(a )
0 1 2 3 4 5 6 7 8 9 0

0 1 2 3 4 5
(b )
0 1 2 3 4 5 6 7 8 9 0
0 1 2 3 4 5
(c )
0 1 2 3 4 5 6 7 8 9 0

Figure for Problem 3-4.

3-3. Round each number to three significant figures: (c) 107.868  (2.113  102)  (5.623  103)  5 519.568
(a) 0.216 74 (d) (26.14/37.62)  4.38  3.043 413
(b) 0.216 5 (e) (26.14/(37.62  108))  (4.38  102)  3.043 413  1010
(c) 0.216 500 3 (f) (26.14/3.38)  4.2  11.933 7
3-4. Vernier scale. The figure above shows a scale found on (g) log(3.98  104)  4.599 9
instruments such as a micrometer caliper used for accurately (h) 106.31  4.897 79  107
measuring dimensions of objects. The lower scale slides Types of Error
along the upper scale and is used to interpolate between the 3-8. Why do we use quotation marks around the word true in
markings on the upper scale. In (a) the reading (at the left- the statement that accuracy refers to how close a measured
hand 0 of the lower scale) is between 1.4 and 1.5 on the upper value is to the “true” value?
scale. To find the exact reading, observe which mark on the
lower scale is aligned with a mark on the upper scale. 3-9. Explain the difference between systematic and random
Because the 6 on the lower scale is aligned with the upper errors.
scale, the correct reading is 1.46. Write the correct readings in 3-10. Suppose that in a gravimetric analysis, you forget to dry
(b) and (c) and indicate how many significant figures are in the filter crucibles before collecting precipitate. After filtering
each reading. the product, you dry the product and crucible thoroughly
before weighing them. Is the error in the mass of product that
3-5. Write each answer with the correct number of digits.
you report a systematic or a random error? Is the mass of prod-
(a) 1.021  2.69  3.711
uct always high or always low?
(b) 12.3  1.63  10.67
(c) 4.34  9.2  39.928 3-11. State whether the errors in (a)–(d) are random or sys-
(d) 0.060 2  (2.113  104)  2.849 03  106 tematic:
(e) log(4.218  1012)  ? (a) A 25-mL transfer pipet consistently delivers 25.031 
(f) antilog(3.22)  ? 0.009 mL when drained from the mark.
(g) 102.384  ? (b) A 10-mL buret consistently delivers 1.98  0.01 mL when
drained from exactly 0 to exactly 2 mL and consistently deliv-
3-6. Using the correct number of significant figures, find the
ers 2.03 mL  0.02 mL when drained from 2 to 4 mL.
formula mass of (a) BaCl2 and (b) C6H4O4. Use the periodic
(c) A 10-mL buret delivered 1.983 9 g of water when drained
table inside the cover of this book to find uncertainties in
from exactly 0.00 to 2.00 mL. The next time I delivered water
atomic masses.
from the 0.00- to the 2.00-mL mark, the delivered mass was
3-7. Write each answer with the correct number of significant 1.990 0 g.
figures. (d) Four consecutive 20.0- L injections of a solution into a
(a) 1.0  2.1  3.4  5.8  12.300 0 chromatograph were made, and the area of a particular peak
(b) 106.9  31.4  75.500 0 was 4 383, 4 410, 4 401, and 4 390 units.
60 3 Experimental Error

Cheryl Cynthia Carmen Chastity Figure for Problem 3-12.

3-12. Cheryl, Cynthia, Carmen, and Chastity shot the targets 3-19. (a) Show that the formula mass of NaCl is 58.442 5
above at Girl Scout camp. Match each target with the proper (0.000 9) g/mol.
description. (b) To prepare a solution of NaCl, you weigh out
(a) accurate and precise (c) precise but not accurate 2.634 (0.002) g and dissolve it in a volumetric flask whose vol-
(b) accurate but not precise (d) neither precise nor accurate ume is 100.00 (0.08) mL. Express the molarity of the solution,
3-13. Rewrite the number 3.123 56 (0.167 89%) in the forms along with its uncertainty, with an appropriate number of digits.
(a) number ( absolute uncertainty) and (b) number ( per- 3-20. What is the true mass of water weighed at 24°C in the air
cent relative uncertainty) with an appropriate number of digits. if the apparent mass is 1.034 6  0.000 2 g? The density of air
is 0.001 2  0.000 1 g/mL and the density of balance weights
Propagation of Uncertainty
is 8.0  0.5 g/mL. The uncertainty in the density of water in
3-14. Find the absolute and percent relative uncertainty and Table 2-7 is negligible in comparison to the uncertainty in the
express each answer with a reasonable number of significant density of air.
figures.
3-21. Twelve dietary iron tables were analyzed by the gravi-
(a) 6.2 (0.2)  4.1 (0.1)  ?
metric procedure in Section 1-4 and the final mass of Fe2O3
(b) 9.43 (0.05)  0.016 (0.001)  ?
(FM 159.688) was 0.2774  0.0018 g. Find the average mass
(c) [6.2 (0.2)  4.1 (0.1)]  9.43 (0.05)  ?
of Fe per tablet. (Relative uncertainties in atomic masses are
(d) 9.43 (0.05)  {[6.2 (0.2)  103] 
small compared with relative uncertainty in the mass of Fe2O3.
[4.1 (0.1)  103]}  ?
Neglect uncertainties in atomic masses in this problem.)
3-15. Find the absolute and percent relative uncertainty and
3-22. We can measure the concentration of HCl solution (a pro-
express each answer with a reasonable number of significant
cedure called standardizing the solution) by reaction with pure
figures.
sodium carbonate: 2H  Na2CO3 B 2Na  H2O  CO2.
(a) 9.23 (0.03)  4.21 (0.02)  3.26 (0.06)  ?
A volume of 27.35  0.04 mL of HCl solution was required
(b) 91.3 (1.0)  40.3 (0.2) / 21.1 (0.2)  ?
for complete reaction with 0.967 4  0.000 9 g of Na2CO3
(c) [4.97 (0.05)  1.86 (0.01)] / 21.1 (0.2)  ?
(FM 105.988  0.001). Find the molarity of the HCl and its
(d) 2.016 4 (0.000 8)  1.233 (0.002)  4.61 (0.01)  ?
absolute uncertainty.
(e) 2.016 4 (0.000 8)  103  1.233 (0.002)  102 
3-23. Avogadro’s number can be computed from the following
4.61 (0.01)  101  ?
measurable quantities for a particular element:2 (1) the atomic
(f) [3.14 (0.05)]1/3  ?
mass of the element (obtained from the mass and abundance of
(g) log[3.14 (0.05)]  ?
each isotope), (2) the density of a crystal of the element, (3)
3-16. Verify the following calculations:
the size of the unit cell (the smallest repeating structural unit)
(a) 2 3.141 5 (0.001 1)  1.772 43 (0.000 31) in the crystal, and (4) the number of atoms in the unit cell.
(b) log[3.141 5 (0.001 1)]  0.497 14 (0.000 15) Extremely accurate measurements for silicon give an atomic
(c) antilog[3.141 5 (0.001 1)]  1.3852 (0.0035)  103 mass of mSi  28.085 384 2 (35) g/mol, where 35 is the
(d) ln[3.141 5 (0.001 1)]  1.144 70 (0.000 35) standard deviation in the last two digits. The density is

2 0.104 (0.006)
(e) loga b  0.800 (0.015)
2.329 031 9 (18) g/cm3, the size of the cubic unit cell is c0 
0.051 1 (0.000 9) 5.431 020 36 (33)  108 cm, and there are 8 atoms per unit
3-17. When using Class 1 standard weights (Table 2-1) to cal- cell. Avogadro’s number is computed from the equation
ibrate your balance, you want to place exactly 1 g on the bal- mSi
ance. Is it more accurate to use a single 1-g weight or the com- NA 
(ρc03 )>8
bination 500 mg  200 mg  200 mg  100 mg? Compare
the uncertainties in mass for the two choices. From the measured properties and their uncertainties (standard
3-18. Express the molecular mass (uncertainty) of deviations), compute Avogadro’s number and its uncertainty. To
C9H9O6N3 with the correct number of significant figures. find the uncertainty of c03, use the function y  xa in Table 3-1.